Clarion University Clarion, PA Proceedings of the 2012 Pennsylvania Economic Association Conference Proceedings of the 2012 Pennsylvania Economic Association Conference PROCEEDINGS OF THE PENNSYLVANIA ECONOMIC ASSOCIATION 2012 CONFERENCE May 31-June 2, 2012 Clarion University Clarion, Pennsylvania William K. Bellinger, Editor Dickinson College Visit the Pennsylvania Economic Association Home Page at http://aux.edinboro.edu/pea/index.html Proceedings of the 2012 Pennsylvania Economic Association Conference Pennsylvania Economic Association: 2011-2012 Executive Board • • • • • • • • • • President: Orhan Kara, West Chester University President-Designate: Tracy Miller, Grove City College Vice President, Program: William Bellinger, Dickinson College Vice President, Publicity: Sandra Trejos, Clarion University Vice President, Membership: Natalie Reaves, Rowan University Secretary: Stephanie Brewer, Indiana University of Pennsylvania Treasurer: Steven Andelin, Penn State - Schuylkill Co-Editors, Pennsylvania Economic Review: Thomas Tolin & Orhan Kara, West Chester University Webmaster: Michael Hannan, Edinboro University of Pennsylvania Immediate Past President: James Jozefowicz, Indiana University of Pennsylvania Board of Directors • • • • • • • • • • Ron Baker, Millersville University Charles Bennett, Gannon University Deborah Gougeon, University of Scranton Kosin Isariyawongse, Edinboro University John McCollough, Penn State - Lehigh Valley Brian O'Roark, Robert Morris University Mark Schweitzer, Federal Reserve Bank of Cleveland Yaya Sissoko, Indiana University of Pennsylvania Luke Tilley, Federal Reserve Bank of Philadelphia Roger White, Franklin and Marshall College Ex-Officio Directors • • • • • • • Thomas O. Armstrong, Pennsylvania Department of Community & Economic Development Gerald Baumgardner, Pennsylvania College of Technology David Culp, Slippery Rock University Donald Dale, Muhlenberg College Robert D'Intino, Rowan University James Dunn, Edinboro University Andrew Economopoulos, Ursinus College • • • • • • • • Joseph Eisenhauer, Wright State University Mark Eschenfelder, Robert Morris University Andrew Hill, Federal Reserve Bank of Philadelphia Elizabeth Hill, Penn State-Mont Alto Mehdi Hojjat, Neuman College Tahereh Hojjat, DeSales University Ioannis N. Kallianiotis, University of Scranton Donna Kish-Goodling, Muhlenberg College Proceedings of the 2012 Pennsylvania Economic Association Conference • • • • • • • • • • • Richard Lang, Federal Reserve Bank of Philadelphia (retired) Daniel Y. Lee, Shippensburg University Robert Liebler, King's College Johnnie B. Linn III, Concord University Patrick Litzinger, Robert Morris College Stanley G. Long, University of Pittsburgh/Johnstown Jacquelynne McLellan, Frostburg State University Lawrence Moore, Potomac State College of West Virginia University Gayle Morris, Edinboro University Heather O'Neill, Ursinus College Brian O’Roark, Robert Morris University • • • • • • • • • • • • Abdul Pathan, Pennsylvania College of Technology William F. Railing, Gettysburg College Margarita M. Rose, King's College William Sanders, Clarion University John A. Sinisi, Penn State University-Schuylkill Brian Sloboda, US Postal Service Kenneth Smith, Millersville University of Pennsylvania Lynn Smith, Clarion University Osman Suliman, Millersville University Paul Woodburne, Clarion University Bijou Yang-Lester, Drexel University David Yerger, Indiana University of Pennsylvania Proceedings of the 2012 Pennsylvania Economic Association Conference Editor’s Introduction and Acknowledgements The papers published in this volume were presented at the 2012 Annual Conference of the Pennsylvania Economic Association held at Clarion University of Pennsylvania from May 31 to June 2, 2012. The program lists all scheduled presenters, session chairs and discussants. Only the papers and comments submitted according to the manuscript guidelines are included in the Proceedings. Each research paper listed in the Table of Contents and Author Index can be accessed directly by clicking on the listing. The 2012 Conference was apparently a great success. Participants gathered from across Pennsylvania, several other states, and multiple nations in order to share their research. Faculty, other professionals, and both graduate and undergraduate students participated in the conference. The conference also featured three general presentations; “Stratification Economics” by Dr. Sue Stockly, a presentation on the regional economy by Mark Schweitzer, Research Director at the Federal Reserve Bank of Cleveland, and a session on teaching economics with Excel by Dr. Humberto Barreto. I particularly wish to thank Dr. Sandra Trejos of Clarion University for her time and energy in coordinating all local arrangements. We also owe special thanks to the Federal Reserve Bank of Cleveland and Dr. James Pesek, Dean of the College of Business at Clarion University of Pennsylvania for their support of this conference. Additional thanks go to the PEA board for their advice and effort in making the conference a success. Lastly, thanks to all of the participants who contributed interesting ideas and a friendly atmosphere to the conference. Proceedings of the 2012 Pennsylvania Economic Association Conference Table of Contents Conference Agenda page 1 RISK AVERSION AND BUSINESS CYCLES: AN EMPIRICAL ANALYSIS Cristian Pardo page 16 RESPONSIVENESS OF THE U.S. TRADE FLOWS TO CHANGES IN CHINESE CURRENCY Orhan Kara page 37 RECENT PENNSYLVANIA JOB TRENDS: EFFECTS OF SHALE? (2012) Jay Bryson, Tim Quinlan and Joe Seydl page 48 MULTIGENERATIONAL DISCOUNTING: MERGING INTERGENERATIONAL EQUITY AND INDIVIDUAL TIME PREFERENCE William Bellinger page 55 AID EFFECTIVENESS IN SUB-SAHARAN AFRICA Yaya Sissoko and Niloufer Sohrabji page 66 ECONOMIC ANALYSIS OF ALTERNATIVE WAYS TO REFORM MEDICAL MALPRACTICE Tracy C. Miller page 85 DOES A FIRM’S DIVIDEND INITIATION AFFECT ITS RISK? Henry F. Check, Jr., John S. Walker and Karen L. Randall page 90 THE BUREAU OF MOTOR FUEL TAXES COMPLIANCE STRATEGY: PENNSYLVANIA DEPARTMENT OF REVENUE Thomas O. Armstrong, Daniel Meuser, James Dehnert, and Nic Banting page 101 A PANEL STUDY ANALYSIS OF ECONOMIC GROWTH IN SOUTH EAST ASIA Tai McNaughton page 117 THE DODD-FRANK “WALL STREET REFORM” ACT OF 2010: IS OUR FINANCIAL SYSTEM MORE STABLE NOW? Adora D. Holstein page 125 GOVERNMENT POLICY AND RESULTANT EFFECT ON NICHE INDUSTRIES: THE CASE OF USPS “EVERY DOOR DIRECT MAIL” Brenda Ponsford and William R. Hawkins page 145 FOREIGN AID EFFECTS ON GROWTH IN LATIN AMERICA Tai McNaughton page 148 OFFSETS IN THE DEFENSE TRADE: COUNTERING COMPARATIVE ADVANTAGE INERNATIONAL BUSINESS Brenda Ponsford and William R. Hawkins page 156 ACADEMIC PERFORMANCE IN GRADUATE MANAGERIAL ECONOMICS Rod D. Raehsler page 164 DECOMPOSING RECENT MONEY SUPPLY CHANGES WITH IMPLICATIONS FOR CURRENT FED POLICY Richard Robinson and Marwan El Nasser page 177 Proceedings of the 2012 Pennsylvania Economic Association Conference WORLD OIL PRICES: ECONOMIC IMPACT AND ECONOMETRIC FORECAST Carrie R. Williams page 187 ADVANTAGEOUS SELECTION IN HEALTH INSURANCE Paul Sangrey page 196 THE EFFECT OF RELIGION AND EMPOWERMENT OF WOMEN ON FERTILITY Denae A. Heath and Cameron D. McConnell page 202 THE EFFECT OF HUMAN DEVELOPMENT LEVEL ON THE RELATIONSHIP BETWEEN ENERGY CONSUMPTION AND ECONOMIC GROWTH: A CHAOS THEORY APPROACH Ezatollah Abbasian and Maysam Nasrindoost page 208 LIVING IN KEYNES’S LONG RUN: THE EFFECTS OF THE OVERUSE OF ECONOMIC STIMULUS David Nugent page 216 DISCUSSANT COMMENT ON LIVING IN KEYNES'S LONG RUN: THE EFFECTS OF THE OVERUSE OF ECONOMIC STIMULUS Michael J. Hannan page 222 KANTIAN MARKETS, BOYCOTTS, AND EFFICIENCY Richard Robinson page 223 THE ECONOMIC IMPACT OF ALVERNIA UNIVERSITY Tufan Tiglioglu, Lisa Cooper, Bari Dzomba, Rachel Gifford, and Joseph Hess page 232 A STUDY OF FACTORS AFFECTING THE ECONOMIC FEASIBILITY OF THE IMPLEMENTATION OF TORREFACTION TECHNOLOGY BY THE PENNSYLVANIA WOOD PELLET INDUSTRY Robert F. Brooker and Harry R. Diz page 243 THE COLLEGE EDUCATED AND THE PUBLIC/PRIVATE SALARY DIFFERENTIAL: CAN OCCUPATION EXPLAIN DIFFERENCES? Mary Ellen Benedict, Michael Bajic, and David McClough page 251 Author Index Proceedings of the 2012 Pennsylvania Economic Association Conference page 267 Pennsylvania Economic Association 2012 CONFERENCE AGENDA THURSDAY, May 31, 2012 • • • 4:00 pm - 9:00 pm Registration (Moore Hall) 5:00 pm - 8:00 pm Board of Directors Dinner/Meeting (Eagle Commons 107-108) 6:00 pm - 10:00 pm Reception (Moore Hall) FRIDAY, June 1, 2012 • • • • • • • • 8:00 am – noon, 2 pm – 4 pm Registration (Still Hall Lobby) 7:30 am - 10:30 am Continental Breakfast (Still Hall Lobby) 9:00 am - 10:15 am Concurrent Sessions (Still Hall) 10:30 am - 11:45 pm Concurrent Sessions (Still Hall) 12 noon - 1:45 pm Luncheon with Speaker- “Stratification Economics" Dr. Sue Stockly (Gemmell Complex-Multipurpose Room) 2:15 pm - 3:30 pm Concurrent Sessions (Still Hall) 3:45 pm - 4:45 pm Fed Lecture: Mark Schweitzer, Research Director at The Federal Reserve Bank of Cleveland (Still Hall 112) 5 pm - 8 pm Fed Sponsored Reception (Still Hall Lobby) SATURDAY, June 2, 2012 • • • • • 7:30 am - 10:30 am Registration & Continental Breakfast (Still Hall Lobby) 9:00 am - 10:15 am Concurrent Sessions (Still Hall) 10:30 am - 11:30 am Plenary Session- Teaching Economics with Excel by Dr. Humberto Barreto (Still Hall 112) 11:45-12:45 General Membership Meeting (Still Hall 112) 12:45 am Closing Proceedings of the 2012 Pennsylvania Economic Association Conference 1 FRIDAY, June 1, 2012 Conference Registration 8:00 a.m. – noon, 2-4 p.m., Still Hall Lobby 7:30 a.m. – 10:30 a.m. Continental Breakfast(Still Hall Lobby) Sessions F1: Friday, June 1, 2012, 9:00 a.m. – 10:15 a.m. Session F1A: Development Economics Location: Still Hall 111 Chair: Orhan Kara, West Chester University The effect of human development level on the relationship between energy consumption and economic growth: a chaos theory approach Ezatollah Abbasian Bu-Ali Sina University Is Foreign Debt a Threat or a Promise to Least Developed Countries? Evelyn Wamboye Pennsylvania State University Aid Effectiveness in Sub-Saharan Africa Yaya Sissoko Indiana University of Pennsylvania Niloufer Sohrabji Simmons College Discussants: Morteza Sameti, University of Isfahan, Iran Kiril Tochkov, Texas Christian University Orhan Kara, West Chester University Session F1B: Student Session I Location: Still Hall 102 Chair: Tracy Miller Grove City College A Dissimilarity of the Federal Reserve and the European Central Bank Mandates and an Investigation of the ZeroBound Rule during the Recent Global Financial Crisis of 2008 Emily Abbondanza Edinboro University Industrial Composition of U.S. Counties as a Determent of Income: A Case Study of Pennsylvania Deepra Yusuf Franklin & Marshall College Advantageous Selection in Health Insurance Paul Sangrey Grove City College Discussants: Tracy Miller, Grove City College Thomas Armstrong, PA Department of Revenue Proceedings of the 2012 Pennsylvania Economic Association Conference 2 Session F1C: Women and Labor Location: Still Hall 202 Chair: Mary Ellen Benedict, Bowling Green University Models for Forecasting Local Employment Trends in Pennsylvania Sandra McPherson Millersville University Determinants of Youth Unemployment in the UAE Samer Kherfi American University of Sharjah Informal social networks and household wellbeing in rural Ethiopia Abera Birhanu Demeke University Canada West Tax Incentives, Education and Migration: The Case of Puerto Rico Carlos Liard Central Connecticut State University Discussants: Jay Bryson, Wells Fargo Kevin Quinn, Bowling Green State University Mary Ellen Benedict, Bowling Green University Session F1D: Microeconomics in an International Context Location: Still Hall 203 Chair: Ronald Baker, Millersville University of Pennsylvania Elasticity Estimation for the Turkish Manufacturing Industry Zeynep Deniz Dervisen Kadir Has University Cournot and Bertrand Competition when Advertising Rotates Demand: The Case of Honda and Scion Kosin Isariyawongse Edinboro University of Pennsylvania Ecological Efficiency and Economic Development in Latin America SandraTrejos Clarion University of PA Discussants: Ronald Baker, Millersville University of Pennsylvania Thomas Andrews, West Chester University Soloman Kone, City University of New York Session F1E: Macroeconomics Location: Still Hall 205 Chair: Michael Hannon, Edinboro University of Pennsylvania Risk Aversion and Business Cycles: An Empirical Analysis Cristian Pardo Saint Joseph's University Using Options to Measure Monetary Policy Credibility Timothy Kearney Misericordia University The Volcker Rule, Wall Street Reform, and the Subprime Financial Crisis Adora Holstein and Nicole Baird Robert Morris University Changing Patterns of Credit Usage from the Survey of Consumer Expenditures: An Exploratory Study of Type and Level of Credit Before and After 9/11 Heather Kirkwood-Mazik Clarion University of Pennsylvania Proceedings of the 2012 Pennsylvania Economic Association Conference 3 Discussants: Richard Robinson Shuang Feng Michael Hannan Marwan El Nasser SUNY-Fredonia Edinboro University of PA Edinboro University of PA SUNY-Fredonia Session F1F: The Economics of Education Location: Still Hall 206 Chair: Lynn Smith, Clarion University What Distinguishes Winners from Losers? An Analysis of School Choice in Massachusetts Soma Ghosh Albright College Financial Aid, College Tuition and Student Retention Frederick Tannery Slippery Rock University The Economic Impact of Alvernia University Tufan Tiglioglu Alvernia University Discussants: Lynn Smith, Clarion University of Pennsylvania Liang Ding, Alvernia University William Bellinger, Dickinson College Sessions F2: Friday, June 1, 2012 10:30 a.m. – 11:45 a.m. Session F2A: Health Economics Location: Still Hall 111 Chair: David McClough, Ohio Northern University The Effect of Safe Water, Sanitation, and Human Capital on Child Nutrition and Health: A Quantile Regressions Approach Divya Balasubramaniam St. Joseph's University Economic Analysis of Alternative ways to reform medical malpractice Tracy Miller Grove City College The Economic Costs of Racial Disparities in Breast Cancer Kemi Oyewole Spelman College Discussants: Samer Kherfi, American University of Sharjah Charles Telly, SUNY Fredonia David McClough, Ohio Northern University Proceedings of the 2012 Pennsylvania Economic Association Conference 4 Session F2B: Student Session II, Empirical Analysis Location: Still Hall 102 Chair: Rod Raehsler Clarion University A Panel Study Analysis of Economic Growth in Southeast Asia McNaughton, Tai Clarion University The Determinants of Flight Delays Nadejda Sergheeva Villanova University World oil Prices: Economic Impact and Econometric Forecast Williams, Carrie Clarion University Discussants: Wenting Yu Rod Raehsler Texas Christian University Clarion University Session F2C: Sustainability in the Business Curriculum Location: Still Hall 202 Chair: David Culp, Slippery Rock University A Matrix Model for Integrating Sustainability into the Business Curriculum: A Case Study David Culp Slippery Rock University Green Accounting: Measuring and Reporting the Triple Bottom Line Anna Lusher Slippery Rock University How a Sustainable Business Accelerator Can Enhance Delivery of Sustainability in the Business Curriculum John Golden Slippery Rock University Discussants: Robert Leibler, Kings College Robert F. Brooker, Gannon University Steven Andelin, Penn State University-Schuylkill Session F2D: International Macro Topics Location: Still Hall 203 Chair: Lei Zhang, Edinboro University of PA Convergence and Persistence of Prices within the European Union Olena Ogrokhina University Of Houston Interest Rates and the Demand for Credit in Ghana Eric Fosu Oteng-Abayie Kwame Nkrumah University Is Foreign Debt a Threat or a Promise to Least Developed Countries? Evelyn Wamboye Pennsylvania State University Discussants: Lei Zhang, Edinboro University of PA Cristian Pardo, Saint Joseph's University Xuebing Yang, Penn State University-Altoona Proceedings of the 2012 Pennsylvania Economic Association Conference 5 Session F2E: Public Economics Location: Still Hall 205 Chair: Michael Hannan , Edinboro University of PA The Bureau of Motor Fuels Tax Compliance Strategy: Pennsylvania Department of Revenue Thomas Armstrong PA Department of Revenue Can Responsibility Centered Budgeting Provide Healing from Budget Cuts? Robert Balough Clarion University of PA Tax Incentives and Capital Spending Revisited Arthur Schiller Casimir Western New England University Discussants: Heather Kirkwood-Mazik, Clarion University of Pennsylvania Michael Hannan , Edinboro University of PA John Walker, Kutztown University Session F2F: Miscellaneous Topics Location: Still Hall 206 Chair: Roger White, Franklin and Marshall Does Economic Freedom Mean Freedom to Grow? Ranajoy Ray-Chaudhuri The Ohio State University Sources and Management of Stress Among Non-Academic Staff in Public Universities in Ghana: The Case of Kwame Nkrumah University of Science and Technology Henry Kofi Mensah Kwame Nkrumah University of Science and Technology The Reciprocal Relation of Marginal Cost of Public Funds: Inequality Aversion and Economic Growth in Iran Morteza Sameti University of Isfahan, Iran Estimation Window and the Power of Event Study Liang Ding and Tufan Tiglioglu Alvernia University Discussants: Roger White, Franklin & Marshall David Nugent, Not Affiliated Ezatollah Abbasian Bu-Ali Sina University Abera Birhanu Demeke , University Canada West Proceedings of the 2012 Pennsylvania Economic Association Conference 6 LUNCHEON AND SPEAKER 12:00 Noon – 1:45 P.M. “Stratification Economics" Dr. Sue Stockly Gemmell Complex-Multipurpose Room) Dr. Sue K. Stockly is an Associate Professor of Economics at Eastern New Mexico University. A native New Mexican, she completed a B.A. and an M.B.A. at the College of Santa Fe. She received a Ph.D. in Economics from the University of Texas at Austin in 1999. After working as an associate economist with the RAND Corporation for five years, she returned to New Mexico where she teaches Economics and Business Research Methods. Her current research interests include stratification economics, regional economic development, teaching of economics, and minority student performance in higher education. Proceedings of the 2012 Pennsylvania Economic Association Conference 7 Sessions F3: Friday, June 1, 2012 2:15 p.m. – 3:30 p.m. Student Poster Session: Location: Still Hall Lobby Nobel Laureate James Meade Kathleen Coen Clarion University Antoine-Augustin Cournot Matthew Bauer Clarion University Jeremy Bentham Samantha Myler Clarion University Session F3A: Resource Economics Location: Still Hall 111 Chair: Sandra McPherson, Millersville University Recent Pennsylvania Job Trends: Effects of Shale? Jay Bryson, Tim Quinlan and Joe Seydl, Wells Fargo Anthracite Production in Northeast Pennsylvania Steven Andelin Pennsylvania State University Determinants of Distributed Photovoltaic Installations in the United States Thomas Andrews West Chester University A study of factors affecting the economic feasibility of the implementation of torre faction technology by the Pennsylvania wood pellet industry Robert F. Brooker Gannon University Discussants: Sandra McPherson, Millersville University Arthur Schiller Casimir, Western New England University Tufan Tiglioglu, Alvernia University Steven Andelin, Penn State University-Schuylkill Session F3B: Student Session III Location: Still Hall 102 Chair: Sandra Trejos, Clarion University Small Business and the United States Economy: The Key to Job Creation and Growth Miranda Mease Clarion University Sheltering Our Poor: Natural Disasters and the Economies of Bangladesh and Haiti Mary Jo Milford Clarion University Militarization and Economic Development in Brazil Braden Picardi Clarion University Discussants: Sandra Trejos Clarion University Anna Lusher, Slippery Rock University Proceedings of the 2012 Pennsylvania Economic Association Conference 8 Session F3C: Economic Education Location: Still Hall 202 Chair: Abdul Pathan, PA College of Technology Academic Performance in Graduate Managerial Economics Rod Raehsler Clarion University An Outline of Some Thoughts on Teaching Money, Banking, and Monetary Economics in Ex-Soviet Countries in Transition Marwan El Nasser SUNY-Fredonia Testing Procedures Robert Liebler King's College Approaches to Lecture Preparation Presentation and Assessing Student Learning for Economics Abdul Pathan PA College of Technology Discussants: David Culp, Slippery Rock University Robert Balough, Clarion University of PA Soma Ghosh, Albright College John Golden, Slippery Rock University Session F3D: Microeconomics Location: Still Hall 203 Chair: Kosin Isariyawongse, Edinboro University of Pennsylvania Comparing the effectiveness of leadership gifts and matching funds in the provision of public goods under alternative laboratory methodologies Ronald Baker Millersville University of Pennsylvania Multigenerational Discounting: Merging Intergenerational Equity and Individual Time Preference William Bellinger Dickinson College Kantian Markets, Boycotts, and Efficiency Richard Robinson SUNY Fredonia Discussants: Kosin Isariyawongse, Edinboro University of Pennsylvania Zeynep Deniz Dervisen, Kadir Has University Brenda Ponsford, Clarion University Proceedings of the 2012 Pennsylvania Economic Association Conference 9 Session F3E: Miscellaneous Topics Location: Still Hall 205 Chair: Tracy Miller, Grove City College The College Educated and the Public/Private Salary Differential: Can Occupation Explain Differences? Mary Ellen Benedict and Michael Bajic, Bowling Green State University" David McClough Ohio Northern University Innovation by London's Water Companies: Internalizing public health externalities Nicola Tynan Dickinson College Adam Smith on Education Kevin Quinn Bowling Green State University Smith’s Invisible Hand is Exonerated by Hayek’s Concept in the Fatal Conceit Charles Telly SUNY Fredonia Discussants: Lynn Smith, Clarion University Tracy Miller, Grove City College Charles Telly, SUNY Fredonia Kevin Quinn, Bowling Green State University Session F3F: International Economics Location: Still Hall 206 Chair: Kiril Tochkov Texas Christian University Responsiveness of the U.S. Trade Flows to Changes in Chinese Currency Orhan Kara West Chester University Impact of the Global Financial Crisis the ECOWAS Countries Yaya Sissoko Indiana University of Pennsylvania Soloman Kone City University of New York The Transforming Dragon: Labor Productivity Change in China's Three Economic Sectors Wenting Yu Texas Christian University Discussants: Xuebing Yang, Penn State University-Altoona Adora Holstein, Robert Morris University Ranajoy Ray-Chaudhuri, The Ohio State University 3:45 p.m. – 4:45 p.m. Still Hall 112 Presentation on the Regional Economy, Mark Schweitzer, Research Director at the Federal Reserve Bank of Cleveland 5 –8 P.M. Still Hall Lobby Reception hosted by the Federal Reserve Bank of Cleveland Proceedings of the 2012 Pennsylvania Economic Association Conference 10 SATURDAY, June 2, 2012 7:30 – 10:30 A.M. Conference Registration & Continental Breakfast: Still Hall Lobby ____________________________________________ Sessions S1: Saturday, June 2, 2012 9:00 a.m. – 10:15 a.m. Session S1A: International Trade Location: Still Hall 111 Chair: Shuang Feng, Edinboro University of PA Regional Supplier Associations as the Producer of Transitional Public Goods in Latin America and the Caribbean Sunita Mondal University of Pittsburgh Losing Comparative Advantage? Calibrating the DFS Model for US Trade, 1970-2008 Roger White Franklin & Marshall College An Eaton-Kortum Openness Index Xuebing Yang Penn State University at Altoona Discussants: Ronald Baker, Millersville University of Pennsylvania Orhan Kara, West Chester University Shuang Feng, Edinboro University of PA Session S1B: Monetary Economics Location: Still Hall 102 Chair: Robert Balough, Clarion University of PA Decomposing Recent Money Supply Changes with Implications for Current Fed Policy Richard Robinson and Marwan El Nasser, SUNY at Fredonia The Bank Lending Channel: a FAVAR Analysis Lei Zhang Edinboro University of PA Living in Keynes's Long Run: The Effects of the Overuse of Economic Stimulus David Nugent Not Affiliated Discussants: Timothy Kearney, Misericordia University Robert Balough, Clarion University of PA Michael Hannan, Edinboro University of PA Proceedings of the 2012 Pennsylvania Economic Association Conference 11 Session S1C: Student Session IV Location: Still Hall 202 Chair: Abdul Pathan, Pennsylvania College of Technology Impacts of the Recession on Women: Analysis of a Local Economy Deepra Yusuf Franklin and Marshall College The Effect of Immigration on State Wages Jennifer Johnson Indiana University of Pennsylvania Foreign Aid and Its Effect on Human Functioning In Africa Nathan Wolbert Clarion University Why Do Dragons Breathe Fire? The Economic Growth and Development of Eastern Asia Michael Bartley Clarion University Discussants: Abdul Pathan, Pennsylvania College of Technology Lynn Smith, Clarion University Session S1D: Miscellaneous Topics Location: Still Hall 203 Chair: William Bellinger, Dickinson College Government Policy and Resultant Effect on Niche Industries Brenda Ponsford Clarion University William R. Hawkins U.S. House of Representatives Can a Change in Dividend Policy Affect Risk? John Walker Kutztown University Offsets in the Defense Trade: Brenda Ponsford Clarion University William R. Hawkins U.S. House of Representatives Discussants: William Bellinger, Dickinson College Thomas Andrews, West Chester University Nicola Tynan, Dickinson College Session S1E: Student Session V Location: Still Hall 205 Chair: Yaya Sissoko, Indiana University of PA Promised Land? The Impact of Promise Scholarship Programs on Housing Prices Michael LeGower University of Pittsburgh Analysis of the Time Series Properties of U.S. Unemployment Kristen Workley Clarion University Recovery From Current Recession: Is It Different From Previous Recessions? Yang Yang Clarion University Proceedings of the 2012 Pennsylvania Economic Association Conference 12 Discussants: Heather Kirkwood-Mazik, Clarion University Yaya Sissoko, Indiana University of PA Session S1F: Student Session VI (Development) Location: Still Hall 206 Chair: Sandra Trejos Clarion University Religion and Empowerment of Women in Brazil Deane Heath Clarion University The Effects of Political Stability and Disease in Sub-Saharan Africa Jared Bruce Clarion University Leadership, Democracy and Economic Development in Africa Chris Myers Clarion University Foreign Aid Effects on Growth in Latin America McNaughton, Tai Clarion University Discussants: Sandra Trejos, Clarion University Rod Raehsler, Clarion University 10:30 am - 11:30 am Plenary Session, Still Hall 112 Dr. Humberto Barreto, Teaching Economics with Excel 11:45-12:45 General Membership Meeting Still Hall 112 This Annual Business Meeting of the General Membership of the Pennsylvania Economic Association is open to the entire membership of the PEA, including all registrants at the conference. Door prizes will be awarded. SATURDAY, June 2, 2012 CLOSING Program Author & Participant Index First Name Last Name E-mail Sessions Ezatollah Abbasian abbasian@basu.ac.ir F1A, F2F Emily Abbondanza abbondanzaemily@yahoo.com F1B Steven Andelin sla7@psu.edu F2C, F3A Thomas Andrews tandrews@wcupa.edu F1D, F3A, S1D Thomas Armstrong thoarmstro@state.pa.us F1B, F2E Ronald Baker ronald.baker@millersville.edu F1D, F3D, S1A Divya Balasubramaniam dbalasub@sju.edu F2A Robert Balough balough@clarion.edu F2E, F3C, S1B Michael Bartley m.c.bartley@eagle.clarion.edu S1C Matthew Bauer m.j.Bauer1@eagle.clarion.edu Poster Sess. Proceedings of the 2012 Pennsylvania Economic Association Conference 13 William Bellinger bellinge@dickinson.edu F3D, S1D Mary Ellen Benedict mbenedi@bgsu.edu F1C, F3E Robert Brooker brooker@gannon.edu F2C, F3A Jared Bruce j.m.bruce@eagle.clarion.edu S1F Jay Bryson jay.bryson@wellsfargo.com F1C, F3A Arthur Schiller Casimir scasimir@wne.edu F2E, F3A Kathleen Coen K.A.Coen@eagle.clarion.edu Poster Sess. David Culp david.culp@sru.edu F2C, F3C Abera Birhanu Demeke abera.birhanu@yahoo.com F1C, F2F Zeynep Deniz Dervisen zeynepdeniz.dervisen@khas.edu.tr F1D, F3D Liang Ding liang.ding@alvernia.edu F1F, F2F Marwan El Nasser Marwan.ElNasser@fredonia.edu F1E, F3C, S1B Shuang Feng sfeng@edinboro.edu F1E, S1A Soma Ghosh sghosh@alb.edu F1F, F3C John Golden john.golden@sru.edu F2C, F3C Deborah Gougeon gougeond1@scranton.edu Michael Hannon hannan@edinboro.edu F1E, F2E, S1B Deane Heath D.A.Heath@eagle.clarion.edu S1F Adora Holstein holstein@rmu.edu F1E, F3F Kosin Isariyawongse kisariyawongse@edinboro.edu F1D, F3D Jennifer Johnson jennifermjohnson0811@hotmail.com S1C Orhan Kara dearorhankara@gmail.com F1A, F3F, S1A Timothy Kearney tkearney@misericordia.edu F1E, S1B Samer Kherfi skherfi@aus.edu F1C, F2A Heather Kirkwood-Mazik hmazik@clarion.edu F1E, F2E, S1E Soloman Kone skone@lagcc.cuny.edu F1D, F3F Michael LeGower mjl88@pitt.edu S1E Carlos Liard liardcaf@ccsu.edu F1C, F2D Robert Liebler robertliebler@kings.edu F2C, F3C Anna Lusher anna.lusher@sru.edu F2C, F3B David McClough d-mcclough@onu.edu F2A, F3E Tai McNaughton t.k.mcnaughton@eagle.clarion.edu F2B, F3B Sandra McPherson smcpherson@millersville.edu F1C, F3A Miranda Mease M.E.Mease@eagle.clarion.edu F3B Henry Kofi Mensah henbil25@yahoo.com F2F Mary Jo Milford m.j.milford@eagle.clarion.edu F3B Tracy Miller tcmiller@gcc.edu F1B, F2A, F3E Sunita Mondal sum46@pitt.edu S1A Chris Myers Samantha Myler S.L.Myler@eagle.clarion.edu Poster Sess. David Nugent davidanugent@hotmail.com F2F, S1B S1F Proceedings of the 2012 Pennsylvania Economic Association Conference 14 Olena Ogrokhina oogrokhi@mail.uh.edu F2D Eric Fosu Oteng-Abayie efoteng-abayie.socs@knust.edu.gh F2D Kemi Oyewole koyewole@gmail.com F2A Cristian Pardo cpardo@sju.edu F1E, F2D Abdul Pathan apathan@pct.edu F3C, S1C Braden Picardi B.C.Picardi@eagle.clarion.edu F3B Brenda Ponsford bponsford@clarion.edu F3D, S1D Kevin Quinn kquinn@bgsu.edu F1C, F2A, F3E Rod Raehsler rraehsler@clarion.edu F2B, F3C, S1F Ranajoy Ray-Chaudhuri ray-chaudhuri.4@buckeyemail.osu.edu F2F, F3F Richard Robinson Richard.Robinson@fredonia.edu F1E, F3D, S1B Morteza Sameti msameti@gmail.com F1A, F2F Paul Sangrey Sangreypm1@gcc.edu F1B Arthur Casimir Schiller scasimir@wne.edu F1E, F3A Nadejda Sergheeva nsergh01@villanova.edu F2B Joseph Seydl joseph.seydl@wellsfargo.com F3A Yaya Sissoko ysissoko@iup.edu F1A, F3F, S1E Lynn Smith lsmith@clarion.edu F3E, S1C Frederick Tannery frederick.tannery@sru.edu S2G Charles Telly tabak@fredonia.edu F2A, F3E Tufan Tiglioglu tufan.tiglioglu@alvernia.edu F2F, F3A,F1F Sandra Trejos strejos@clarion.edu F1D, F3B, S1F Kiril Tochkov k.tochkov@tcu.edu F1A, F3F Nicola Tynan tynann@dickinson.edu F3E, S1D John Walker walker@kutztown.edu F2E, S1D Evelyn Wamboye efw10@psu.edu F1A, F2D Roger White roger.white@fandm.edu F2F, S1A Carrie Williams C.R.Williams1@eagle.clarion.edu F2B Nathan Wolbert n.a.wolbert1@eagle.clarion.edu S1C Kristen Workley K.M.Workley@eagle.clarion.edu S1E Yang Yang morningyang@yahoo.com S1E Xuebing Yang xyang@psu.edu F2D, F3F, S1A Wenting Yu wentingyu@tcu.edu F2B, F3F Deepra Yusuf deepra.yusuf@fandm.edu F1B, S1C Lei Zhang lzhang@edinboro.edu F2D, S1B Proceedings of the 2012 Pennsylvania Economic Association Conference 15 RISK AVERSION AND BUSINESS CYCLES: AN EMPIRICAL ANALYSIS Cristian Pardo Department of Economics Saint Joseph's University Philadelphia, PA 19131 ABSTRACT Private entrepreneurs usually lack access to complete riskpooling for their idiosyncratic risk. Consequently, risk-averse entrepreneurs internalize volatility and demand a private equity premium, which is capable of amplifying business cycles due to its sensitivity to shocks and subsequent impact on investment and output. Therefore, economies with larger private entrepreneurial sectors should present higher volatility. I test this prediction by (1) conducting a reducedform analysis that shows that volatility is negatively associated with the importance of the corporate vs. privatelyheld sectors; and (2) estimating the model's structural parameters, where positive risk-aversion coefficients should be found where private entrepreneurs are predominant. INTRODUCTION Differences in business cycle fluctuations in emerging markets relative to developed economies have been extensively studied by economists. Prasad, Agenor and McDermott (1999), among many others, provide empirical evidence supporting much higher average output volatility in emerging economies than in industrialized economies.1 In the theoretical arena, numerous models rely on financial imperfections as a primary stylized fact to motivate this discussion, where low levels of development in financial markets observed in emerging markets are often cited. For instance, Calvo and Reinhart (2000) and Chang and Velasco (2000) focus on the role of “dollarized liabilities." Namely, due to currency mismatching, real exchange rate depreciations may negatively impact firms' and banks' balance sheets by asymmetrically increasing the value of outstanding debt relative to revenues. Another stylized fact commonly analyzed is the presence of a more important privately-held sector relative to corporations in emerging markets. Market capitalization as a fraction of GDP, for instance, is about 100 percent or more in highincome countries, while 35 percent or less in low and middleincome countries (The World Bank, 2010). The fact that privately-owned firms tend to rely more on debt rather than equity to finance their investments may play an additional role in creating frictions in emerging economies. Bernanke, Gertler and Gilchrist (1999), for example, study the impact of information asymmetries in the borrower-lender relationship. The authors show that the agency problems that arise from the positive probability of costly default imply that lenders optimally charge entrepreneurs an external finance premium. This premium is endogenous to firms' balance sheet in that a higher reliance on external funds raises the aforementioned agency costs. As business cycles affect entrepreneurs' net worth, the resulting external finance premium is countercyclical and may become a mechanism that magnifies the impact of real shocks over time. Pardo (2010) examines additional stylized facts about emerging markets, including that private entrepreneurial activity in particular is also often very volatile. An illustrative example is the Chilean case. While it has boasted one of the most robust financial systems in its region, its economy still reacted strongly to the effects of the Asian crisis of the 1990's. A reason often cited for this response was that the entrepreneurial sector moved quickly from an early-to-mid 1990's boom euphoria to a deep depression in the following years. Evidence like this may make it worthwhile to further examine facts and assumptions about the private entrepreneurial sector. In general, the simplifying assumption of risk neutrality on agents makes sense in some cases. Gale and Hellwig (1985), for instance, point out that “risk neutrality is not an unreasonable assumption to make in the case of investors since it can be justified as a consequence of risk-pooling." That is, by investing large amounts of funds and thus taking advantage of economies of scale, lenders tend to successfully maintain highly diversified portfolios that allow them to significantly reduce the exposure to risk. However, the authors also emphasize the fact that the risk-neutrality assumption “makes less sense in the case of entrepreneurs." As shown by Moskowitz and Vissing-Jørgensen (2002), the high concentration of ownership of privately-held companies and their importance in their owners' portfolios, leave private entrepreneurs highly vulnerable to project-specific, uninsurable risks. That is, the lack of access to complete riskpooling for their idiosyncratic risks leaves private entrepreneurs with no other option but to internalize the cost of volatility. Therefore, assuming risk neutrality (that is, assuming that risk can be ignored) seems to be a stronger assumption in this case. Pardo (2010) builds on Bernanke et al. (1999) to show that introducing risk aversion among private entrepreneurs modifies the optimal relationship with lenders mainly by Proceedings of the 2012 Pennsylvania Economic Association Conference 16 incorporating a private equity premium, or the positive risk premium that risk-averse entrepreneurs demand due to the stochastic nature of the uninsurable part of their investment returns. This premium may lead to further magnifying the aggregate effects of real shocks. The mechanism works as follows: a real shock that decreases entrepreneurial profits and net worth, reduces entrepreneurs' minimum guaranteed level of consumption (the insurable part of the entrepreneur's returns). Consequently, their effective risk aversion and so the private equity premium rise. In response to the increased internal and external costs, entrepreneurs increase the rental rate of capital to final goods firms, producing a contraction in the supply of capital and thus additional impact of shocks on investment and production. Finally, Pardo (2012) extends the previous model into a small open economy framework. In this context, following Chang and Velasco (2000), shocks not only affect entrepreneurs' net worth directly, but also indirectly through the increase in the value of debt following the corresponding real exchange rate adjustment. As wealth falls, the aforementioned private equity premium rises, therefore generating the known amplifying impacts of shocks over time. Consequently, a direct implication of this model is that economies where the private entrepreneurial sector is a relatively important actor in the financial market (for instance, where family-owned businesses are predominant), and therefore the economy as a whole is more vulnerable to uninsurable risk, all else equal, should present higher output volatility than economies where the corporate sector, whose ownership structure is highly diversified at all levels, is more important. Considerable debate has taken place among economists about the role of entrepreneurial activity in affecting economic growth. Arguments about the impact of entrepreneurship on important factors affecting long-run growth often discuss innovation, productivity and knowledge spillovers (van Stel, Carree and Thurik, 2005). Much less attention has been given to the relationship between entrepreneurship and output growth volatility. Apart from Pardo (2012), another exception is Rampini (2004), who provides a theoretical framework in which entrepreneurial activity is procyclical and produces amplification and propagation of shocks. In the empirical arena, however, to the best of my knowledge there have been no studies examining the relationship between the ownership structure of the real sector and output volatility. Entrepreneurial activity is a plausible index of the economic importance of entrepreneurship. The Global Entrepreneurship Monitor (GEM) builds the Total Entrepreneurial Activity rate (TEA), which measures the “relative amount of nascent entrepreneurs and business owners of young firms for a range of countries" (van Stel et al., 2005). Figure 1 illustrates the simple direct (unconditional) correlation between total entrepreneurial activity, as measured by the TEA index, and output volatility, as measured by the standard deviation of the per capita real GDP growth, for 46 countries. At least as preliminary evidence, there seems to be a positive correlation between those two variables.2 The objective of this paper is to empirically test whether the volatility-inducing frictions that risk aversion introduces are more likely to be present the greater the relative size of the private entrepreneurial sector is. I test this prediction through two alternative approaches. First, I examine the statistical significance of the correlation between GDP growth volatility and the ownership structure of the productive sector through a simple reduced-form analysis. In particular, using international data, I find that output volatility is negatively associated with the relative importance of the corporate sector in the financial market, all else equal. The investmentto-capital ratio (as a measure of financial leverage), indices of financial structure and development plus other measures of cross-country sources of uncertainty are used as control variables. The reduced-form approach's main limitation in this case is that a resulting statistical relationship between output volatility and ownership structure may not provide empirical significance of the magnitude of the impact imposed by private entrepreneurs in promoting sharper business cycle volatility. Consequently, I also conduct a structural analysis through which instead of using a measure of ownership of the real sector as a proxy for the relevance of entrepreneurial risk aversion, I estimate the model's risk-aversion coefficient (γ) that is consistent with an economy's observed fluctuation of output. That is, I estimate the structural parameters of the dynamic model that I introduce in Section 2 using observed data for the model's main variables; namely output, investment and consumption. Intuitively, if an economy is mostly composed of corporations (such as the U.S.) and so the importance of frictions imposed by risk-averse entrepreneurs is relatively small, we should see this economy behave more closely to one with risk neutral agents (that is, the estimated γ should not be significantly different from zero). Conversely, an economy where private entrepreneurs are predominant and so risk aversion is likely to impose stronger impacts, a positive coefficient of risk aversion should be found. The outline of this paper is as follows. The next section describes the dynamic model and its theoretical implications; the following two sections introduce a reduced-form empirical analysis using international data and then conducts a structural empirical analysis of the risk-aversion assumption. The last section 4 provides some concluding remarks. THE SUPPLY OF CAPITAL, AGGREGATE EFFECTS AND DYNAMICS Proceedings of the 2012 Pennsylvania Economic Association Conference 17 The theoretical framework is based on Pardo (2012), who builds upon Cespedes, Chang and Velasco (2004) and Bernanke et al. (1999). In order to focus this paper's attention on the model's empirical implication, which is that GDP growth volatility, all else equal, should be associated with the ownership structure of the productive sector, this section briefly describes the model in which frictions between riskaverse entrepreneurs and lenders may lead to the aforementioned aggregate effects, while leaving further discussion and a more detailed description of the model in 5. The first step is to introduce the context in which an international lender and a domestic borrower optimally interact and, as a consequence of asymmetric information and entrepreneurial risk aversion, a risk premium arises. The contract between entrepreneurs and lenders, which eventually yields the economy's capital supply, involves the following features: (i) entrepreneurs produce capital goods (such as plant and equipment) that domestic firms rent in order to produce final goods; (ii) entrepreneurs (or capitalists) finance their capital investments using both their internal net worth and borrowed funds denominated in foreign currency; (iii) competitive risk-neutral international lenders provide financing to entrepreneurs for which they receive a contracted interest rate; and (iv) the contractual relationship between borrowers and lenders is subject to informational frictions in that the effective profitability of capital production is ex-ante unknown and ex-post observed only by entrepreneurs, unless lenders pay an auditing cost (Townsend, 1979). The optimal contract implies the following equilibrium results: (i) lenders charge a premium that is higher than the opportunity cost of lending due to the agency costs that arise from costly entrepreneurial bankruptcy, denoted the external finance premium; (ii) risk-averse entrepreneurs seek selfinsurance against low state of nature; (iii) hedging is incomplete as entrepreneurs optimally and willingly accept some uncertainty in order to maintain some information as private, namely the upper side of the distribution of capital returns; (iv) risk-averse entrepreneurs internalize the utility associated with facing uncertain returns and for which they require a premium on top of the pecuniary borrowing costs, denoted the private equity premium; (v) the capital rental rate that entrepreneurs charge captures both the external finance premium and the private equity premium; (vi) all else equal, a risk-averse entrepreneur requires a higher expected capital return than a risk-neutral entrepreneur, or equivalently, given a capital rental rate, an entrepreneur is willing to supply less capital to final goods firms; and (vii) any drop in entrepreneurial net worth, say as a consequence of an adverse shock, affects both premiums: the external finance premium due to increased agency costs as entrepreneurs rely more on external rather than internal funds, and the private equity premium, as the entrepreneurs' minimum guaranteed consumption level falls and so their effective level of risk aversion increases. The next step is to introduce the dynamics and aggregate effects of the derived supply of capital in the context of an open economy dynamic general equilibrium framework. The main elements that this model comprises include: (i) the domestic economy produces a final good using labor and capital supplied by workers and entrepreneurs, respectively; (ii) domestic firms maximize profits by optimally choosing capital, labor and output; (iii) the firm optimality conditions yield the economy's demand for capital, which together with the supply of capital, determine the economy's equilibrium capital investment and rental rate; (iv) workers maximize utility over consumption and leisure subject to a budget constraint; (v) the monetary authority uses its policy instruments to maintain the domestic price level constant while letting the exchange rate freely fluctuate; and (vi) the model closes by imposing domestic and external market clearing conditions. The predicted effects of an unexpected shock on the economy are as follows. Say that an unexpected decrease in the foreign demand for domestic goods takes place. The drop in exports produces a decrease in output demand and, in particular, in planned capital investment demand. The corresponding drop in domestic expenditure produces an instantaneous real depreciation of the local currency, which reduces the entrepreneur's net worth by increasing the domestic currency value of debt repayment, which is denominated in foreign currency. As a consequence, both the external finance premium and the private equity premium increase, further impacting planned investment decisions and the real exchange rate. That is, the shock together with entrepreneurial risk aversion should produce a magnified short-term effect on capital investment and output. The predicted implication suggests that business cycles should be more volatile in economies where non-publicly traded private firms are relatively more important than in economies where publicly-traded corporations are relatively more important, all else equal. This result arises from the fact that risk aversion and thus the private equity premium are more relevant to privately-held companies than to, say, corporations, which are usually owned by multiple large investors. Recall that given the observed high concentration of ownership of privately-held companies and their importance in the owners' portfolios, private entrepreneurs become highly vulnerable to project-specific risks and thus are forced to internalize the cost of volatility more (Moskowitz and Vissing-Jørgensen, 2002). EMPIRICAL EVIDENCE: REDUCED-FORM APPROACH Proceedings of the 2012 Pennsylvania Economic Association Conference 18 This model, like others that examine the impact of financial frictions on business cycles, predict that more leveraged economies tend to present stronger business cycle fluctuations, as higher reliance on debt rather than internal wealth translates into a higher external finance premium. In addition, the model in this paper particularly stresses the role played by risk-averse entrepreneurs in that the private equity premium may also be responsible for creating even sharper business cycle fluctuations. This section attempts to empirically test this model's main predictions. In particular, using international data, I look at the reduced-form cross-sectional association between business cycle volatility and the ownership structure of the economy's capital-producing sector, controlling for the financial development, leverage and other sources of uncertainty. Specifically, I test whether business cycle volatility, all else equal, is more pronounced where the privately-owned private sector represents a larger fraction of their economic activity of a country, relative to where the corporate sector predominates. The Data I use the standard deviation of per capita GDP growth to measure business cycle volatility.3 I obtain data for this variable from the International Financial Statistics (IFS) of the International Monetary Fund. To capture leverage, I employ the investment to capital ratio estimated by Albuquerque and Wang (2008) (QK/PN).4 In order to account for the relative importance of the privately-owned sector, I use cross-country data on a financial structure definition constructed by Levine (2002). This measure, called Structuresize (STSIZE), is an index of the relative size of stock markets vs. bank credit to the private sector. The main principle behind this measure is that an economy will have a larger index the greater the importance of the corporate sector relative to the privately-owned sector in its financial structure. Therefore, this model predicts a negative coefficient on this variable. This measure, however, harbors some limitations. For instance, even though this index measures the relative importance of the two markets, it can only serve as a proxy for the relative importance of the flow of new equity and credit to businesses. No clear bias can be ex-ante identified as consequence of this measurement error. In addition, due to its role of facilitating the optimal allocation of funds and risk pooling, the different levels of development of financial markets should in theory imply differences in how risk is managed and diversified, thus also affecting GDP growth volatility. To this goal, I have included an index of financial development as an explanatory variable both by itself and multiplied by the financial structure variables in order to explore an interaction effect. I use as the index of financial development the ratio of the stock market liquidity to costs in the banking sector as a measure of efficiency in the financial market, as in Levine (2002). Empirically, however, there is no clear evidence of a significant link between financial development and economic growth volatility. Some studies have found that financial development does dampen GDP growth volatility (Wahid and Jalil, 2010; Denizer, Iyigun and Owen, 2002), while others have obtained no robust relationship (Beck, Lundberg and Majnoni, 2006). Consequently, there is no unambiguous exante way that differences in financial development may affect GDP growth volatility in the context of this model. As in previous studies, I control for other cross-country sources of uncertainty. To account for aggregate monetary uncertainty, I use the average inflation rate for the period 1985-2006 (INFAVE) and inflation volatility for the same period (INFSTD). In terms of uncertainty associated with international trade, I use the volatility of the real exchange rate (RERSTD) and the degree of openness of the economy, as measured by the share of the sum of exports and imports to GDP (TRADE). All these series come from either the International Financial Statistics (IFS) of the IMF or the World Development Indicators (WDI) of the World Bank. Finally, in order to account for the uncertainty caused by government institutions, as in Albuquerque and Wang (2008), I employ a measure of government spending volatility (GOVSTD), an index of corruption (CORRUPT) and an index of government stability (GOVSTAB).5 The first series is obtained from the IFS while the last two series come from the International Country Risk Guide (ICRG) of the Political Risk Service Group. Results Table 1 shows the results from estimating the relationship between output growth volatility, firm leverage (QK/PN) and the measure of relative importance of the corporate sector in the financial market (STSIZE), while controlling for different sets of exogenous sources of volatility. In regression (1), I examine the impact of leverage and the relative size of the corporate sector on volatility, initially controlling only for monetary shock variables: the average inflation level (INFAVE) and its volatility (INFSTD). Only the estimated coefficients for the inflation variables have the predicted sign and statistical significance. Regression (2) include the volatility induced by openness to international markets, as captured by the volatility of the real exchange rate (RERSTD) and the trade to GDP ratio (TRADE). Regressions (3) additionally considers the volatility induced by the government. In both regressions (2) and (3), the coefficients for the relative importance of the corporate sector have the expected sign and statistical significance. In regressions (4) and (5), I split the sample in order to track Proceedings of the 2012 Pennsylvania Economic Association Conference 19 groups with high and low entrepreneurial importance base on their “Total Entrepreneurial Activity" index (TEA) separately. Results are not unambiguous. While the volatilityreducing impact of the relative importance of the corporate sector is stronger for countries with lower entrepreneurial activity (STSIZE1) than those with higher entrepreneurial activity (STSIZE2), the difference of these values, however, is not statistically significant. The lack of significance maybe partially explained by the low number of observations. Finally, regression (6) considers the impact of differences in financial development and GDP growth volatility across countries, either directly or through its interaction with the financial structure variable. As commonly found in the literature, results do not show evidence of a significant role played by capital markets in affecting growth volatility. The above results suggest that the volatility of GDP growth is indeed negatively associated with the relative importance of the corporate sector in the financial market, holding overall leverage, financial development and other sources of volatility constant. That is, these results are consistent with this model's main prediction that a relatively larger entrepreneurial sector should increase volatility, all else equal.6 In addition, results are also consistent with the more standard prediction that more leverage should significantly increase volatility. The main limitations of the reduced-form approach is that the existence of a statistical relationship between volatility and ownership structure does not indicate how strong the role of entrepreneurial risk aversion is in promoting sharper business cycles fluctuations. This shortcoming can be addressed through a structural empirical analysis. EMPIRICAL EVIDENCE: STRUCTURAL APPROACH Given the reduced-form analysis' limitation in providing clear empirical evidence about the magnitude of the role of entrepreneurial risk aversion in affecting business cycles, this section attempts to estimate the coefficient of risk aversion ( γ ) that would be consistent with observed fluctuations in output and other variables experienced by individual economies. A positive coefficient would signal that entrepreneurial risk aversion is indeed a necessary feature of the model in order to replicate the observed path of real variables. A coefficient not significantly different from zero, on the other hand, would imply that risk aversion is irrelevant in explaining a country's business cycle. Consequently, the objective of this section is two-fold: (i) to estimate the main structural parameters of the model, in particular the risk-aversion coefficients ( γ ) that are consistent with 12 economies' observed paths of real variables such as output, consumption and investment: Argentina, Brazil, Canada, Chile, France, Germany, Japan, South Korea, Mexico, Thailand, the U.K. and the U.S.7; (ii) to provide cross-country comparisons for consideration. As results in Section 3.1 suggest, the role of risk-averse entrepreneurs in causing sharper business cycle fluctuations is likely to be more significant in economies where the privately-owned private sector represents a relatively larger fraction of the real sector. In contrast, for countries like the U.S., where the corporate sector is more important, we should observe this economy behave more closely to one with riskneutral agents. Therefore, the null hypotheses are (i) that the estimated risk-aversion coefficient is significantly greater than zero, and (ii) that it is higher for countries in which privately-held economies predominate like, say, Mexico, Chile, Thailand or Korea, when compared to corporateintensive economies like the U.S. In short, the way this estimation procedure works is the following. The model's non-standard parameters are initially calibrated using observed data on each economy. Calibration details are presented in 7. Subsequently, the mathematical package DYNARE, which is commonly used for simulation and estimation of rational expectation models, solves the dynamic model presented in 5 and obtains the policy function of its choice variables, given the realization of the stochastic shocks and an initial choice of parameters values. Then, using Bayesian methods, the package uses real data in order to estimate the value of select parameters that best match the observed path followed by some of the model's main variables. The following section introduces the dataset and the parameters to be estimated. The Data Besides the risk-aversion coefficient (γ), the parameters I estimate in this section are the share of domestic goods in total expenditure (θ) and the three parameters that are initially obtained through calibration: the default cost rate (μ), the entrepreneurial saving rate (δ) and the variance of capital returns (σω2). Given that the model has three shock parameters (the multifactor productivity parameter, A; the international riskfree rate, ρ; and foreign demand for domestic goods, X), only three data vectors can be used in order to avoid stochastic singularity (Juillard, 2004). As a result, I use time series on real output (Y), consumption (C) and physical capital investment (K) to conduct the estimations.8 Data for France, Germany, Japan and the U.K. come from Eurostat. Data for Brazil, Korea and Mexico were obtained from the IMF's International Financial Statistics (IFS). In addition, I obtain Proceedings of the 2012 Pennsylvania Economic Association Conference 20 Argentinean data from Argentina's Instituto Nacional de Estadística y Censos (INDEC), Chilean data from the Central Bank of Chile, Canadian data from OECD.Stat Extracts, Thailand's data from Office of The National Economic and Social Development Board, Thailand, and U.S. data comes the U.S. Bureau of Economic Analysis (BEA). The frequency of data is quarterly and, in order to facilitate cross-country comparison, the data range from 1995.1 to 2011.4 for all 12 economies. In addition, I use the HodrickPrescott filter in order to extract the trend of all series and then transform the series in terms of the percentage deviations from their trends. Priors The structural estimation procedure requires setting prior assumptions for the parameters' distribution, mean and standard deviation. Most of the mean values were obtained or chosen in the calibration procedure presented in 7 (see Table 4 for a summary of results). For the risk-aversion coefficient (γ), for instance, I use the normal distribution around the initially assumed mean value of 2 for all countries. The standard deviation is set high enough in order to allow for higher flexibility in the estimation. For the expenditure share of domestic goods in home consumption parameter (θ), I also assume a normal distribution with a prior mean of 0.6 and standard deviation of 0.3 for all countries.9 Mean and standard deviation values for the remaining variables to be estimated were set according to calibration and specific to each country. The bankruptcy cost parameter (μ) and the variance of capital returns (σω2) were assumed to follow an inverse gamma distribution in order to ensure positive estimates. The standard deviations for these variables were set at values equal to roughly one half of their prior means. Finally, the entrepreneurial saving rate (δ) follows a normal distribution with rather low standard deviations (between 0.05 and 0.1), in order to avoid the possibility of saving rates greater than one. With respect to the shock parameters, I use uniform distribution for their standard deviations (σ) and persistence parameters (φ). Their lower and upper bounds were set loosely equal to 0 and 1, respectively. The first four columns of Table 2 summarize the parameters' prior distribution, mean and standard deviation. Results The last three columns of Table 2 present the parameter estimations computed using the Bayesian method. The first column contains the estimated mean, while the last two show the 5th and 95th percentiles of the posterior distributions. The bottom of each table contains the value of the marginal likelihood, computed with the Laplace approximation. Turning to the risk-aversion coefficient, as observed in Table 2, while the posterior values for γ are positive for Argentina, Brazil, Canada, Chile, Germany, Korea, Mexico and Thailand, they are not statistically different from zero for the cases of France, Japan, the U.S. and the U.K. One may have not expected Canada and Germany to belong to the group with positive γ parameters; however one may also notice that their values are statistically lower than the values for γ of any country in that group. In general, the fact that the posterior γ is statistically higher for countries like Argentina, Brazil, Chile, Korea, Mexico and Thailand than for Canada, France, Germany, the U.K. and the U.S. suggests that this model requires a higher risk-aversion coefficient in order to explain the responses to shocks followed by some economies than for others. Similarly, one may observe that not all the estimated mean values for γ can be statistically distinguished among countries. For instance, though France presents the lowest mean value for γ, its estimate is not statistically different from those found for Japan, the U.S. or the U.K. Equivalently, though Korea presents the highest mean values for γ, its estimate is not statistically different from those found for Argentina or Brazil. In addition, the estimate of γ for Japan cannot be statistically distinguished from any of the countries in the group of economies with positive γ, except for Korea. The result above supports the idea that entrepreneurial risk neutrality may be an assumption that more closely reassembles economies that are more corporate-intensive than predominantly privately-held. Put differently, entrepreneurial risk aversion is realistic and statistically significant in explaining output fluctuations for economies that are more heavily composed of privately-held firms, but not so much for economies where the corporate sector is relatively more important. That is, these results suggest that risk aversion on the part of entrepreneurs not only makes sense in theory for some economies, but also presents empirical relevance. The values of the other coefficients are mostly close to their corresponding prior values, which came from the calibration procedures described in 7. One exception is the coefficient on the share of domestic goods in consumption (θ), which tends to be greater than the uniformly assumed value of 0.6 for developed countries and lower for less developed countries. This result is consistent with the empirical evidence that bias towards domestic goods weakens in emerging markets relative to developed countries. Wang and Chen (2004) discuss and provide evidence that issues like lack of trust of goods manufactured in developing countries and the Proceedings of the 2012 Pennsylvania Economic Association Conference 21 documented consumption home bias among OECD countries are behind this result. With respect to the shock variables, the estimated standard deviations and persistence parameters tend to differ with respect to the shock and across countries. In general, however, shocks are on average stronger in emerging economies than in more advanced economies (as captured by their standard deviations σX, σρ and σA), however less persistent (as given by their autocorrelation coefficients φX, φρ and φA. The difference in volatility and persistence of shocks towards stronger but less persistent values for less developed economies is also consistent with previous literature. For instance, Mendoza (1995) finds that terms of trade shocks are on average almost three times more volatile however less persistent in emerging markets relative to developed countries. The parameter μ is one that seemingly departs from previously estimates values. For instance, Bernanke et al. (1999) and Carlstrom and Fuerst (1997) find that for the United States, μ = 0.12 and μ = 0.20, respectively, while I found from both calibration and estimation than its value is considerably lower. One of the factors that may explain this difference is that in both Bernanke et al. (1999) and Carlstrom and Fuerst (1997), μ represents costs of bankruptcy, while in this model μ captures the costs of default. Given that bankruptcy implies default while default does not necessarily imply bankruptcy, it follows that the costs of default should be lower than those of bankruptcy. CONCLUDING REMARKS Evidence has shown that economies can still respond strongly to shocks even in the absence of noticeable flaws in their financial and/or productive sectors. A possible answer to this apparent puzzle may lie in the role of private entrepreneurs. This sector has evidenced signals of rapid transitions from euphoria during booms to deep depression in even mild recessions in some economies. Consequently, their behavior towards the exposure to risk may be an additional source of amplification of shocks in an economy. Risk aversion on the part of private entrepreneurs implies that entrepreneurs internalize the cost of the uncertainty of their investment returns, for which they require an extra premium: the private equity premium. A microeconomic analysis of the optimal relationship between lenders and entrepreneurs under asymmetric information shows that the rate at which entrepreneurs rent capital to final goods firms (R) incorporates this premium. As a consequence, a positive premium translates into a more constrained capital supply. may respond to shocks as changes in net worth would impact their effective level of risk aversion in the opposite direction. The corresponding effect on the private equity premium further impacts the economy's capital supply and thus production. A corollary to this result is the fact that higher output volatility should be more prevalent in countries where private entrepreneurs are likely to be relatively more important actors of an economy. That is, economies with a relatively higher share of privately-held companies, all else equal, should exhibit sharper business cycles than economies with a private sector composed more importantly of publicly traded companies. Put differently, two otherwise similar economies subject to similar real shocks may respond differently to these shocks due to differences in the ownership structure of the real sector. This paper empirically tests the above implication and finds evidence to suggest the hypothesis that risk aversion may matter. I first approach this issue by conducting a simple reduced-form regression analysis to examine the crosscountry empirical connection between GDP growth volatility and the general ownership structure of the real sector. Results suggest that a statistically significant relationship exists when controlling for other relevant variables, like financial development, leverage and other cross-country sources of uncertainty. To provide further evidence on the magnitude of the role of entrepreneurial risk aversion in affecting business cycles fluctuations, I also execute an empirical analysis of the structural parameters of the model. In particular, I estimate the risk aversion coefficient that would generate business cycles which best match the observed real fluctuations. Empirical results for 12 countries signal that the risk-aversion assumption is macroeconomically more relevant for small open economies like Argentina, Brazil, Chile, Mexico, Korea and Thailand (which present relatively stronger privatelyheld sectors) than for Canada, France, Germany, the U.K. and the U.S. (where the corporate sector is relatively more important). In terms of policy implications, improvements in information technology and transparency in the privately-owned sector that could alleviate agency costs may reduce the importance of asymmetric information in generating volatility. In addition, entrepreneurial risk aversion will play a lesser role as more established businesses become public. Therefore, policies encouraging the participation of investors in the ownership of firms, for instance through private equity firms, may help reduce the volatility associated with entrepreneurial risk aversion. In addition, this premium may further amplify shocks in general equilibrium. Specifically, entrepreneurial activity Proceedings of the 2012 Pennsylvania Economic Association Conference 22 Figure 1: Scatter plot and linear fit of the volatility of GDP growth on total entrepreneurial activity. Proceedings of the 2012 Pennsylvania Economic Association Conference 23 Table 1: OLS Estimates of Standard Deviation of Per Capita Real GDP Growth Variable QK/PN STSIZE (1) -6.1030 12.065 -0.0514 0.0501 (2) 3.1022** 1.2218 -0.0093** 0.0038 (3) 4.700** 1.5297 -0.0108** 0.0038 STSIZE1 (4) 5.6532 2.2419 INFAVE 0.2770** 0.0402 1.2297** 0.1012 0.7893** 0.0461 0.0836 0.1019 0.0003 0.0009 0.0002* 0.0001 0.7535** 0.0550 0.2753** 0.1336 0.0004 0.0009 0.0002* 0.0001 0.0026 0.0055 -0.3378 0.3256 0.0075 0.0050 0.6324** 0.1135 0.5384 0.3169 0.0017 0.0014 0.0002* 0.0001 -0.0192 0.0099 -0.5655 0.3433 0.0200* 0.0083 -0.0093 0.0093 1.3930** 0.8700 -0.1047 0.8218 0.001 0.0018 3.6E-6 0.0004 0.0145 0.0103 0.3571 1.7343 0.0025 0.0102 0.6140 1.3061 34 0.9939 -0.3378** 0.1250 28 0.9822 -0.5652** 0.1656 28 0.9835 -0.5677** 0.2325 14 0.9955 -0.7808** 0.3421 14 0.7044 RERSTD TRADE GOVSTAB GOVSTD CORRUPT FDEV STSIZE*FDEV Intercept # of Obs. Adjusted R2 (6) 3.9861** 1.9727 -0.0144** 0.0066 -0.0131** 0.0048 STSIZE2 INFSTD (5) 5.9117* 3.0158 0.7349** 0.0857 0.3201* 0.1899 0.0004 0.0009 0.0002* 0.0001 0.0005 0.0070 -0.2655 0.3712 0.0073 0.0053 0.0033 0.0051 0.0021 0.0050 -0.4800** 0.2197 28 0.9820 Notes: Variables are the capital to net worth ratio (QK/PN), the measure of the relative importance of the corporate sector (STSIZE), the volatility of inflation (INFSTD), average inflation (INFAVE), the volatility of the real exchange rate (RERSTD), trade openness (TRADE), an index of government stability (GOVSTAB), the volatility of real government spending (GOVSTD), the degree of corruption (CORRUPT), and an index of efficiency of the domestic financial market (FDEV). Numbers below the coefficient estimates correspond to the standard errors. Proceedings of the 2012 Pennsylvania Economic Association Conference 24 Table 2: Bayesian Estimation ARGENTINA Prior Parameter Distrib. Mean Std. dev Mean γ normal 2 2 2.7859 σ2ω inv gamma 0.009 0.0044 0.0077 μ inv gamma 0.028 0.014 0.0274 δ gamma 0.947 0.05 0.8431 θ normal 0.6 0.05 0.6684 φX uniform 0.5 0.2887 0.6522 φA uniform 0.5 0.2887 0.1455 φρ uniform 0.5 0.2887 0.2809 σX uniform 0.5 0.2887 0.2919 σA uniform 0.5 0.2887 0.0586 σρ uniform 0.5 0.2887 0.0151 Marginal Likelihood: 345.39 BRAZIL Prior Parameter Distrib. Mean Std. dev γ normal 2 3 σ2ω inv gamma 0.055 0.0276 μ inv gamma 0.775 0.3875 δ gamma 0.52 0.05 θ normal 0.6 0.3 φX uniform 0.5 0.2887 φA uniform 0.5 0.2887 φρ uniform 0.5 0.2887 σX uniform 0.5 0.2887 σA uniform 0.5 0.2887 σρ uniform 0.5 0.2887 Marginal Likelihood: 375.43 CANADA Prior Parameter Distrib. Mean Std. dev γ normal 2 3 σ2ω inv gamma 0.055 0.0276 μ inv gamma 0.775 0.3875 δ gamma 0.52 0.05 θ normal 0.6 0.3 φX uniform 0.5 0.2887 φA uniform 0.5 0.2887 φρ uniform 0.5 0.2887 σX uniform 0.5 0.2887 σA uniform 0.5 0.2887 σρ uniform 0.5 0.2887 Marginal Likelihood: 662.56 Posterior Confidence interval 1.9004 3.9440 0.0049 0.0102 0.0187 0.0340 0.7976 0.8924 0.6423 0.7063 0.6098 0.6873 0.0232 0.2810 0.0790 0.4858 0.2578 0.3435 0.0516 0.0663 0.0133 0.0167 Mean 2.5528 0.0812 0.6583 0.4276 0.5472 0.2375 0.2785 0.3492 0.1393 0.0349 0.0062 Posterior Confidence interval 1.8430 3.4458 0.0695 0.0927 0.3889 0.8846 0.3739 0.4876 0.4773 0.6285 0.1687 0.3136 0.2025 0.3528 0.1552 0.5505 0.1193 0.1621 0.0295 0.0393 0.0033 0.0096 Mean 0.8780 0.1240 0.5409 0.4505 0.8584 0.6853 0.9633 0.7214 0.1337 0.0052 0.0030 Posterior Confidence interval 0.8364 0.9153 0.1208 0.1277 0.5215 0.5649 0.4461 0.4548 0.8450 0.8674 0.6701 0.7017 0.9483 0.9847 0.6716 0.7625 0.1244 0.1460 0.0047 0.0057 0.0025 0.0036 Proceedings of the 2012 Pennsylvania Economic Association Conference 25 CHILE Prior Parameter Distrib. Mean Std. dev γ normal 2 1 σ2ω inv gamma 0.007 0.0035 μ inv gamma 0.018 0.0092 δ gamma 0.959 0.05 θ normal 0.6 0.3 φX uniform 0.5 0.2887 φA uniform 0.5 0.2887 φρ uniform 0.5 0.2887 σX uniform 0.5 0.2887 σA uniform 0.5 0.2887 σρ uniform 0.5 0.2887 Marginal Likelihood: 399.34 FRANCE Prior Parameter Distrib. Mean Std. dev γ normal 2 4 σ2ω inv gamma 0.01 0.0048 μ inv gamma 0.029 0.0146 δ gamma 0.939 0.05 θ normal 0.6 0.3 φX uniform 0.5 0.2887 φA uniform 0.5 0.2887 φρ uniform 0.5 0.2887 σX uniform 0.5 0.2887 σA uniform 0.5 0.2887 σρ uniform 0.5 0.2887 Marginal Likelihood: 721.64 GERMANY Prior Parameter Distrib. Mean Std. dev γ normal 2 3 σ2ω inv gamma 0.042 0.0208 μ inv gamma 0.059 0.0296 δ gamma 0.915 0.05 θ normal 0.6 0.3 φX uniform 0.5 0.2887 φA uniform 0.5 0.2887 φρ uniform 0.5 0.2887 σX uniform 0.5 0.2887 σA uniform 0.5 0.2887 σρ uniform 0.5 0.2887 Marginal Likelihood: 599.62 Mean 1.9325 0.0057 0.0635 0.9353 0.6257 0.6912 0.2463 0.0553 0.2035 0.0175 0.0310 Posterior Confidence interval 0.9202 3.0959 0.0037 0.0074 0.0256 0.0887 0.8902 0.9850 0.4984 0.7474 0.5973 0.7934 0.0753 0.4331 0.0003 0.1217 0.1505 0.2615 0.0149 0.0197 0.0263 0.0358 Mean -0.2102 0.0168 0.0469 0.9483 0.9667 0.9563 0.7402 0.8195 0.2437 0.0047 0.0044 Posterior Confidence interval -0.5691 0.0535 0.0160 0.0178 0.0440 0.0484 0.9414 0.9540 0.9605 0.9736 0.9357 0.9692 0.6981 0.7868 0.7811 0.8603 0.2050 0.2710 0.0041 0.0054 0.0037 0.0052 Mean 0.8205 0.0405 0.0424 0.9001 0.6868 0.6410 0.7979 0.5944 0.0685 0.0084 0.0052 Posterior Confidence interval 0.7550 0.9037 0.0352 0.0460 0.0313 0.0515 0.8838 0.9173 0.6679 0.7015 0.6239 0.6604 0.7066 0.8763 0.4271 0.7053 0.0588 0.0777 0.0075 0.0093 0.0045 0.0059 Proceedings of the 2012 Pennsylvania Economic Association Conference 26 JAPAN Prior Parameter Distrib. Mean Std. dev γ normal 2 4 σ2ω inv gamma 0.008 0.0038 μ inv gamma 0.02 0.01 δ gamma 0.961 0.05 θ normal 0.6 0.3 φX uniform 0.5 0.2887 φA uniform 0.5 0.2887 φρ uniform 0.5 0.2887 σX uniform 0.5 0.2887 σA uniform 0.5 0.2887 σρ uniform 0.5 0.2887 Marginal Likelihood: 643.23 KOREA Prior Parameter Distrib. Mean Std. dev γ normal 2 2 σ2ω inv gamma 0.008 0.0042 μ inv gamma 0.039 0.0193 δ gamma 0.924 0.05 θ normal 0.6 0.3 φX uniform 0.5 0.2887 φA uniform 0.5 0.2887 φρ uniform 0.5 0.2887 σX uniform 0.5 0.2887 σA uniform 0.5 0.2887 σρ uniform 0.5 0.2887 Marginal Likelihood: 275.91 MEXICO Prior Parameter Distrib. Mean Std. dev γ normal 2 4 σ2ω inv gamma 0.006 0.0031 μ inv gamma 0.015 0.0075 δ gamma 0.967 0.05 θ normal 0.6 0.3 φX uniform 0.5 0.2887 φA uniform 0.5 0.2887 φρ uniform 0.5 0.2887 σX uniform 0.5 0.2887 σA uniform 0.5 0.2887 σρ uniform 0.5 0.2887 Marginal Likelihood: 393.38 Mean 0.9690 0.0068 0.0208 0.9196 0.7167 0.7156 0.7008 0.6579 0.0544 0.0097 0.0054 Posterior Confidence interval -0.4808 2.2994 0.0041 0.0095 0.0103 0.0348 0.8530 0.9615 0.6239 0.8095 0.6208 0.8187 0.5985 0.8184 0.4766 0.7976 0.0390 0.0700 0.0085 0.0111 0.0047 0.0065 Mean 3.9768 0.0073 0.0306 0.8696 0.3667 0.1559 0.0552 0.7975 0.2897 0.0534 0.0238 Posterior Confidence interval 3.2552 4.8196 0.0062 0.0081 0.0174 0.0473 0.8249 0.9171 0.2895 0.4796 0.0636 0.2640 0.0004 0.0956 0.7070 0.8854 0.2475 0.3218 0.0468 0.0628 0.0203 0.0285 Mean 2.6745 0.0069 0.0179 0.8884 0.8574 0.8310 0.0838 0.3178 0.2408 0.0308 0.0328 Posterior Confidence interval 2.2630 3.0079 0.0050 0.0092 0.0136 0.0229 0.8168 0.9314 0.8253 0.8881 0.7959 0.8682 0.0102 0.1652 0.1565 0.4718 0.1891 0.3051 0.0274 0.0349 0.0279 0.0362 Proceedings of the 2012 Pennsylvania Economic Association Conference 27 THAILAND Prior Parameter Distrib. Mean Std. dev γ normal 2 1 σ2ω inv gamma 0.014 0.0071 μ inv gamma 0.053 0.0264 δ gamma 0.902 0.05 θ normal 0.6 0.3 φX uniform 0.5 0.2887 φA uniform 0.5 0.2887 φρ uniform 0.5 0.2887 σX uniform 0.5 0.2887 σA uniform 0.5 0.2887 σρ uniform 0.5 0.2887 Marginal Likelihood: 338.43 U.K. Prior Parameter Distrib. Mean Std. dev γ normal 2 4 σ2ω inv gamma 0.007 0.0033 μ inv gamma 0.013 0.0067 δ gamma 0.973 0.05 θ normal 0.6 0.3 φX uniform 0.5 0.2887 φA uniform 0.5 0.2887 φρ uniform 0.5 0.2887 σX uniform 0.5 0.2887 σA uniform 0.5 0.2887 σρ uniform 0.5 0.2887 Marginal Likelihood: 659.46 U.S. Prior Parameter Distrib. Mean Std. dev γ normal 2 3 μ inv gamma 0.02 0.0101 σ2ω inv gamma 0.005 0.0024 δ gamma 0.948 0.05 φA uniform 0.5 0.2887 σA uniform 0.5 0.2887 Marginal Likelihood: 190.14 Mean 1.5726 0.0112 0.0372 0.8822 0.4703 0.5507 0.4518 0.0563 0.2136 0.0377 0.0209 Posterior Confidence interval 0.9820 2.3990 0.0079 0.0145 0.0234 0.0480 0.8122 0.9558 0.4085 0.5452 0.4983 0.6070 0.2849 0.6767 0.0035 0.0921 0.1862 0.2460 0.0346 0.0417 0.0194 0.0236 Mean -0.0299 0.0048 0.021 0.9579 0.7311 0.7318 0.8345 0.5564 0.0813 0.0052 0.0051 Posterior Confidence interval -0.8155 0.9308 0.0044 0.0052 0.0197 0.0222 0.9411 0.9753 0.6736 0.8040 0.6567 0.8063 0.7130 0.9361 0.5124 0.6078 0.0627 0.1041 0.0045 0.0060 0.0042 0.0058 Mean -0.1481 0.0159 0.0039 0.9711 0.0028 0.0075 Posterior Confidence interval -1.0452 0.4725 0.0125 0.0200 0.0035 0.0046 0.9511 1.0020 0.0001 0.0050 0.0039 0.0124 Note: Dependent variables: Y, K and C. The intervals correspond to the 5th and 95th percentiles of the posterior distributions. Proceedings of the 2012 Pennsylvania Economic Association Conference 28 Table 3: Calibration Parameter Values and Sources Country Argentina Argentina Parameter Default probability Risk premium Value 3% 2.85% Argentina Brazil Brazil Brazil Canada Canada Canada Capital to net worth ratio Default probability Risk premium Capital to net worth ratio Default probability Risk premium Capital to net worth ratio 1.646 4% 36.62% 2.7 7% 1.521% 1.463 Chile Default probability Chile Chile France France France Germany Germany Germany Japan Japan Japan Korea Korea Korea Mexico Mexico Mexico Thailand Thailand Thailand U.K. U.K. Risk premium Capital to net worth ratio Default probability Risk premium Capital to net worth ratio Default probability Risk premium Capital to net worth ratio Default probability Risk premium Capital to net worth ratio Default probability Risk premium Capital to net worth ratio Default probability Risk premium Capital to net worth ratio Default probability Risk premium Capital to net worth ratio Default probability Risk premium 2.12% 1.612 6% 3.145% 1.818 9% 6.727% 1.314 4% 2.13% 1.481 2.72% 3.5% 2.1 4.74% 1.747% 1.424 8.7% 5.369% 1.988 3% 1.499% U.K. U.S. U.S. U.S. Capital to net worth ratio Default probability Risk premium Capital to net worth ratio 1.202 2% 2% 2 9% Sources Lizarazo (2011) Ministerio de Economia (Argentina); Glen and Mondragon-Velez (2011) Cespedes, Gonzalez and Molina (2010) Souza-Sobrinho (2010) World Development Indicators (WDI) Terra (2009) Aretz and Pope (2007) World Development Indicators (WDI) Goyal and Wang (2011) Calvo, Drexler, Flores and Pacheco (2009) Central Bank of Chile Fernandez (2005) Aretz and Pope (2007) World Development Indicators (WDI) Mefteh and Oliver (2007) Aretz and Pope (2007) World Development Indicators (WDI) Niu (2009) Aretz and Pope (2007) World Development Indicators (WDI) Bartram, Brown and Fehle (2009) Gertler, Gilchrist and Natalucci (2007) Gertler et al. (2007) Gertler et al. (2007) Rocha and Garcia (2004) World Development Indicators (WDI) Cespedes et al. (2010) Sandstrom (2006) World Development Indicators (WDI) Chorruk and Worthington (2010) Aretz and Pope (2007) UK Debt Management Oο¬ο¬ce; Bank of England Niu (2009) Bernanke et al. (1999); Fisher (1998) Bernanke et al. (1999) Bernanke et al. (1999) Proceedings of the 2012 Pennsylvania Economic Association Conference 29 Table 4: Calibration. Parameter α γ β Υ θ Common Parameters Description share of capital to output coefficient of risk aversion discount factor labor supply elasticity parameter domestic goods to total expenditure ratio Value 0.35 2 0.99 4/3 0.6 Country Parameters Parameter µ δ σω Description default cost rate entrepreneur saving rate volatility Argentina 0.0285 0.9471 0.0936 µ δ σω default cost rate entrepreneur saving rate volatility France 0.0293 0.9390 0.0980 µ δ σω default cost rate entrepreneur saving rate volatility Mexico 0.0148 0.9676 0.0781 Parameter κ H(Ε΅) R − (1 + ρ) QK/Y QC/Y SX/Y x κ H(Ε΅) R − (1 + ρ) QK/Y QC/Y SX/Y x κ H(Ε΅) R − (1 + ρ) QK/Y QC/Y SX/Y x Country Brazil Canada 0.7750 0.0127 0.5205 0.9698 0.2350 0.0876 Country Germany Japan 0.0592 0.0200 0.9147 0.9610 0.2039 0.0875 Country Thailand U.K. 0.0528 0.0134 0.9020 0.9726 0.1193 0.0814 Table 5: Implied steady state values Description Value Argentina Brazil Canada capital to net worth ratio 1.647 2.694 1.437 default rate 2.999% 4.017% 6.923% overall risk premium 2.849% 36.603% 1.537% capital to output ratio 33.70% 25.50% 34.13% consumption to output ratio 65.00% 65.00% 65.00% exports to output ratio 40.78% 45.70% 40.52% insurance rate 476.38% 43.62% 88.32% France Germany Japan capital to net worth ratio 1.860 1.306 1.463 default rate 6.053% 8.975% 4.025% overall risk premium 3.086% 6.756% 2.178% capital to output ratio 33.62% 32.48% 33.93% consumption to output ratio 65.00% 65.00% 65.00% exports to output ratio 40.83% 41.51% 40.64% insurance rate 79.25% 81.93% 81.30% Mexico Thailand U.K. capital to net worth ratio 1.445 2.032 1.187 default rate 4.769% 8.800% 2.969% overall risk premium 1.717% 5.230% 1.554% capital to output ratio 34.07% 32.94% 34.12% consumption to output ratio 65.00% 65.00% 65.00% exports to output ratio 40.56% 41.24% 40.53% insurance rate 83.94% 75.55% 84.47% Chile 0.0185 0.9594 0.0842 Korea 0.0385 0.9244 0.0916 U.S. 0.0193 0.9600 0.0705 Chile 1.617 7.93% 2.14% 33.94% 65.0% 40.64% 83.88% Korea 2.1 2.72% 3.5% 33.49% 65.0% 40.91% 70.67% U.S. 2 2.15% 1.98% 34.11% 65.0% 40.54% 75.98% Proceedings of the 2012 Pennsylvania Economic Association Conference 30 ENDNOTES 1 See also Backus and Kehoe (1992). The limitation of this index is that it is measured in absolute terms and not relative to the importance of the corporate sector. That is, a country with a higher entrepreneurial activity index does not necessarily have lower participation corporate sector participation in the productive sector. In section 3.1, I use a different variable that overcomes this limitation. 3 Data are for the period 1985-2006 and are for the following sources Argentina, Australia, Austria, Belgium, Brazil, Canada, Chile, Colombia, Denmark, Finland, France, Germany, Greece, India, Ireland, Israel, Italy, Japan, Malaysia, Mexico, Netherlands, New Zealand, Norway, Pakistan, Peru, Philippines, Portugal, South Africa, Spain, Sweden, Switzerland, Thailand, the United Kingdom and the United States. 4 These authors use real GDP data from the World Development Indicators of the World Bank and follow King and Levine (1994) to estimate this variable. 5 This index measures both the government’s ability to carry out its declared program(s), and its ability to stay in office. 6 A potential reverse causality situation in which GDP growth volatility would favor private entrepreneurial activity over corporate activity is not a concern. Given corporations' greater ability to manage and diversify risk over private entrepreneurs, a resulting higher GDP growth volatility would make private entrepreneurial activity relatively less important rather than more predominant. As common in the entrepreneurial activity literature, the existence of idiosyncratic uninsurable risk may affect this sector and in turn cause amplified output responses to shocks, but the resulting increased aggregate volatility does not translate into higher idiosyncratic risk (Rampini, 2004; Angeletos and Calvet, 2006). 7 I include these countries as representative economies with different degrees of openness, income level, geographic location and degree of participation of publicly-traded businesses in physical capital formation. 8 Given its size and lower degree of openness, I estimate a closedeconomy version of the model for the case of the U.S. Since there are no export or international interest rate shocks in a closedeconomy framework, only technological shocks can be considered. Consequently, only one variable series can be used to estimate the model; in this case, I used capital investment ( K ) because it is the variable that is most responsive to shocks in this model. 9 See Cespedes et al. (2004). 2 APPENDIX 1 The Model This section briefly describes the optimal contract between an international lender and a domestic borrower in order to examine the risk premium that arises from asymmetric information and entrepreneurial risk aversion. As in Bernanke et al. (1999), entrepreneurs produce capital goods (such as plant and equipment) that domestic firms employ to manufacture final goods. The effective amount of finished capital, however, after entrepreneur j’s purchase of Kjt+1 units of raw capital, is given by ωjKjt+, where ωj is a stochastic variable with a known distribution, Et(ωj) = 1 and variance σω2. The entrepreneur finances the investment using both his internal net worth (N) and funds borrowed from international lenders denominated in foreign currency (B), St Bt j+1 = Pt K K t j+1 − Pt N t j (1) where net worth is measured in terms of the domestic good priced P, PK is the price of capital, and S is the nominal exchange rate. All but borrowing (B) are denominated in local currency. Competitive international lenders charge a contracted gross interest rate Zjt+1. The contract is subject to informational frictions in that the ex-post realization of ωj is private information and idiosyncratic to each entrepreneur. Assuming costly state verification as in Townsend (1979), lenders can only observe the return to capital by paying an auditing cost. This framework ensures that in non-default states, the lender optimally does not monitor and thus the entrepreneur maintains the true realization of ωj as private information and repays the lender the contracted interest rate. Therefore, profits for the entrepreneur in high states from producing capital are ωjRt+1Kjt+1 – St+1Zjt+1Bjt+1, where Rt+1 is the nominal rental rate denominated in domestic currency that final goods firms pay to entrepreneurs for renting capital. In low states of nature where default takes place, however, the lender does monitor and states become observable by both parties. In such a case, the risk-neutral lender optimally purchases the project from the risk-averse entrepreneur at some fixed price equal to, say, a fraction xjt+1 ≤ 1 of his initial net worth PtNjt. That is, the optimal contract insures the risk-averse entrepreneur with positive consumption in all states. Therefore, entrepreneur j maximizes the expected utility: ωˆ j xtj Pt N t j U ( ) dH (ω ) + ∫0 S t +1 (2) ∞ ω j Rt +1 K t j+1 − S t +1 Z t j+1 Bt j+1 ) dH (ω ) ∫ωˆ jU ( S t +1 where St+1 also denotes the price of entrepreneurial consumption, assuming for simplicity that they only consume imports. Let ωΜ j be the cut-off value of ωj such that the entrepreneur is indifferent between defaulting or not. That is, ωΜ j solves St+1Zjt+1Bjt+1 = ωΜ j Rt+1Kjt+1 – xjtPtNjt and the entrepreneur's expected utility can be re-expressed as Proceedings of the 2012 Pennsylvania Economic Association Conference 31 ωˆ j ∫ 0 U( which imply the following optimal relationship: xtj Pt N t j ) dH (ω ) + S t +1 xtj Pt N t j + (ω j − ωˆ j ) Rt +1 K t j+1 ) dH (ω ) ∫ωˆ jU ( S t +1 (3) ∞ Note that the entrepreneur obtains x tPtN t regardless of ω . In addition, if ωj > ωΜ j the entrepreneur receives an extra random return ωjRt+1Kjt+1 and repays a fixed amount ωΜ j Rt+1Kjt+1 to the lender. Note also that xtPtNjt is the price at which it would be optimal for the entrepreneur not to face risk and thus sell his business to investors. Since the entrepreneur would agree to take that payment only when the state of nature is “low enough," then xjt can be interpreted as an insurance rate against idiosyncratic risk. j j j Risk-neutral international lenders are willing to take part in this contract only when their participation constraint holds: ωˆ j ∫ 0 [ ∞ ω ˆ j Rt +1 K t j+1 (1 − µ ) ωRt +1 K t j+1 ] dH (ω ) + ∫ˆ j [ ] dH (ω ) ω S t +1 S t +1 (4) xtj Pt N t j j − ≥ (1 + ρ t +1 ) Bt +1 S t +1 That is, the lender participates only when the expected returns from lending net of default costs (as given by auditing costs and captured by the parameter μ) are not lower than the opportunity cost of funds, where ρ is the international risk-free interest rate. In equilibrium, due to default costs, lenders charge a external finance premium. The optimal contract is given by the choice of Kjt+1, ωΜ j and xjt that maximizes the entrepreneur's expected utility subject to the lender participation constraint. The optimality conditions are: ∞ ∫ω (ω − ωˆ ) dH (ω ) R t +1 = ˆ ′ E (U ( ) |ω < ω ) (1 + ρ t +1 ) Pt N t S t +1 xt Pt N t − [1 + ][ ] (5) E (U ′( ) |ω > ωˆ ) K t +1 St K t +1 Rt +1 − Cov{U ′( ), ω} E (U ′( ) |ω > ωˆ ) ∞ xt Pt N t + [1 − H (ωˆ )]∫ˆ ωRt +1 dH (ω ) = ω K t +1 ∞ S (1 + ρ t +1 ) Pt N t [1 − H (ωˆ )]∫ˆ ωˆ Rt +1 dH (ω ) + t +1 ω St K t +1 [1 − H (ωˆ ) − µωˆ h(ωˆ )] − ⋅ Cov{U ′(⋅), ωRt +1} + Ο E (U ′(⋅) |ω > ωˆ ) (8) Equation (8) captures the entrepreneur's optimal investment decision that in equilibrium equates the entrepreneur's expected marginal revenues (left hand side) and expected marginal costs (right hand side). Revenues are the sum of the insurance guaranteed in all states of nature, and the expected net return in non-default states. Marginal costs, on the other hand, consist of the combination of four terms. The first two are, respectively, the unit cost of capital repayment in non-default states and the entrepreneur's opportunity cost of funds. The third term, the covariance between capital return and the entrepreneur's marginal utility of consumption, captures the drop in utility that risk-averse entrepreneurs experience from facing uncertain returns. This additional cost that for which entrepreneurs demand a premium corresponds to the private equity premium (the fourth term φ is negligible). Equation (8) also captures the economy's supply of capital as a function of its price (R) and the other predetermined variables. In general equilibrium, the entrepreneurs' capital supply and final goods firms' capital demand yield the equilibrium R and the economy's optimal capital investment (K). In terms of the general equilibrium framework, the domestic economy produces a good through a standard constant returns to scale Cobb-Douglas production function, employing capital (K) supplied by entrepreneurs and employing labor (L) supplied by workers, ˆ [1 + 1 E (U ′( ) | ω < ωˆ ) ]= E (U ′( ) | ω > ωˆ ) [1 − H (ωˆ )] − µωˆ h(ωˆ ) j ωˆ j (1 − µ ) ωR K xtj Pt N t j t +1 t +1 =∫ [ ] dH (ω ) + 0 St +1 St +1 ∞ ω ˆ j Rt +1 K t j+1 j ∫ωˆ j[ St +1 ] dH (ω ) − (1 + ρt +1 ) Bt +1 (6) (7) Yt = At K tα L1t−α (9) where At is a aggregate multifactor productivity shock parameter. Domestic firms maximize profits by optimally choosing capital, labor and output. The standard optimality conditions determine the economy's demand for capital and labor: Rt K t = αPtYt Wt Lt = (1 − α ) PtYt (10) (11) where Wt is the nominal wage rate, which is denominated Proceedings of the 2012 Pennsylvania Economic Association Conference 32 in local currency. Firm's profits are zero in equilibrium. The representative worker consumption and leisure, maximizes utility N t = δ {( Rt /Pt ) K t − (1 + ρ t )( S t /Pt ) Bt over ωˆ − µ ∫ ω ( Rt /Pt ) K t dH (ω )} (17) 0 St CtE = 1 log Ct − Lυt υ subject to the budget constraint Wt Lt = Pt CtH + Pt F CtF where Ct = [θθ(1 – θ)(1 – θ)]-1(CtH)θ(CtF)(1 – θ) is a CES aggregate of domestic goods (CH) and foreign goods (CF), υ > 0 determines the elasticity of labor supply, θ is the weight of domestic goods in total consumption and PtF is the domestic price of imports. Assuming that the law of one price holds and the price of the foreign good is the numeraire (PF = 1), then PF = S. Consumption cost minimization and the definition of aggregate consumption Ct, produce the economy's labor supply and demand for consumption goods 1−δ Pt N t δ Finally, the model closes by imposing domestic and external market clearing conditions. For the money market, the monetary authority uses its policy instruments to maintain a constant domestic price level while letting the exchange rate freely fluctuate. The goods market equilibrium condition imply that: PtYt = θQt ( K t +1 + Ct ) + St X t (19) That is, proceeds from selling output can be consumed, invested in capital or exported, where Xt is exports, which are exogenous to the model. The stochastic behavior of the model is driven by three structural shocks At = φ A At −1 + ε tA Wt = Lυt −1 Qt Ct Qt Ct = Wt Lt (18) ρ ρ t +1 = φρ ρ t + ε t (12) X t = φ X X t −1 + ε (13) (20) (21) X t (22) where εtA, εtρ and εtX are independent and identically distributed innovations with εtA ~ N(μi, σi2), i = {A, ρ, X}. where Qt = Ptθ S t1−θ (14) Assuming that capital production uses domestic and foreign goods in proportions θ and 1 – θ, respectively, then Qt also represents the domestic price of capital production (i.e., PK = Q). Then, equation (1) can be re-expressed as: Pt N t + St Bt +1 = Qt K t +1 Therefore, the rational expectations stochastic dynamic general equilibrium is defined by equations (5), (6), (7), (9), (10), (11), (14), (13), (12), (15), (17) and (19), along with the stochastic processes in equations (20), (21) and (22), that solve for Yt, Lt, Kt+1, Qt, St, Rt, ωΜ j , Wt, xt, Nt, Bt+1 and Ct,. (15) From equations (3) and (4), the entrepreneurial sector aggregate equity (Vt+1) accrued from renting capital to final goods firms is: ωˆ Vt +1 = Rt +1 K t +1 − (1 + ρ t +1 ) S t +1 Bt +1 − µ ∫ ωRt +1 K t +1 dH (ω ) 0 (16) Following an overlapping generations framework as in Pardo (2010), entrepreneurs consume fraction 1 - δ of Vt, and save and invest the rest in a risky project that yields aggregate equity Vt+1 in the following period. As a result, entrepreneurial net worth and consumption (PE) at period t evolve respectively as: APPENDIX 2 Calibration This section presents the values of the parameters used to solve the dynamic model. I select most parameters in a standard manner. For instance, following Bernanke et al. (1999), the share of capital in production (α) is set at 0.35, the discount factor β is 0.99 and the labor supply elasticity is equal to 3 (which implies a value for the parameter υ equal to 4/3). In addition, the idiosyncratic return ω is assumed to be log-normally distributed with a mean of one and standard deviation of σω. As in Cespedes et al. (2004), the expenditure on domestic goods in total consumption, θ, is set at 0.6. Finally, as in most dynamic macroeconomic models, I initially set the constant relative risk-aversion Proceedings of the 2012 Pennsylvania Economic Association Conference 33 coefficient (γ) equal to 2. I select the remaining parameters so that they imply steadystate values that match previously estimated measures for the 12 economies under study. In particular, I choose the values of the volatility of returns σω, the default cost rate μ and the entrepreneurial saving rate δ so that the steadystate values of the default rate H(ω), the overall risk premium R – (1 + ρ) and the ratio of capital to net worth QK/PN match previous estimates. Table 3 presents the values for each of those parameters found in the literature and their sources. To match such values, Table 4 summarizes the values for σω, μ and δ that I obtained from the calibration procedure. Table 5 displays the resulting calibrated steady state values of the model's main variables, which are used for the structural estimation of the model. REFERENCES Albuquerue, R., and Wang, N. (2008). Agency Conflicts, Investment, and Asset Pricing. Journal of Finance, 63 , 1-40. Angeletos, G.-M., and Calvet, L.-E. (2006). Idiosyncratic production risk, growth and the business cycle. Journal of Monetary Economics, 53 , 1095-1115. Aretz, K., and Pope, P. F. (2007). Common Factors in Default Risk Across Countries and Industries. SSRN eLibrary. Backus, D. K., and Kehoe, P. J. (1992). International Evidence of the Historical Properties of Business Cycles. American Economic Review, 82 , 864-88. Bartram, S. M., Brown, G. W., and Fehle, F. R. (2009). International Evidence on Financial Derivatives Usage. Financial Management, 38 , 185-206. Beck, T., Lundberg, M., and Majnoni, G. (2006). Financial intermediary development and growth volatility: Do intermediaries dampen or magnify shocks? Journal of International Money and Finance, 25 , 1146-1167. Bernanke, B. S., Gertler, M., and Gilchrist, S. (1999). The financial accelerator in a quantitative business cycle framework. In J. B. Taylor, and M.Woodford (Eds.), Handbook of Macroeconomics chapter 21. (pp. 1341-1393). Elsevier volume 1, Part C. (1st ed.). Calvo, D., Drexler, A., Flores, C., and Pacheco, D. (2009). The Effect of the Number of Lending Banks on the Liquidity Constraints of Firms: Evidence From a Quasi-Experiment. Working Papers Central Bank of Chile 528 Central Bank of Chile. Calvo, G., and Reinhart, C. (2000). When Capital Inows Come to a Sudden Stop: Consequences and Policy Options. MPRA Paper 6982 University Library of Munich, Germany. Carlstrom, C. T., and Fuerst, T. S. (1997). Agency Costs, NetWorth, and Business Fluctuations: A Computable General Equilibrium Analysis. American Economic Review, 87, 893-910. Cespedes, J., Gonzalez, M., and Molina, C. A. (2010). Ownership and capital structure in Latin America. Journal of Business Research, 63 , 248-254. Cespedes, L. F., Chang, R., and Velasco, A. (2004). Balance Sheets and Exchange Rate Policy. American Economic Review, 94, 1183-1193. Chang, R., and Velasco, A. (2000). Exchange-Rate Policy for Developing Countries. American Economic Review, 90 , 71-75. Chorruk, J., and Worthington, A. (2010). Firm-specific determinants and outcomes of initial public offerings in Thailand, 20012007. Discussion Papers in Finance finance:201002 Griffith University, Department of Accounting, Finance and Economics. Denizer, C. A., Iyigun, M. F., and Owen, A. (2002). Finance and Macroeconomic Volatility. The B.E. Journal of Macroeconomics. Proceedings of the 2012 Pennsylvania Economic Association Conference 34 Fernandez, V. (2005). Determinants of Firm Leverage in Chile: Evidence from Panel Data. Estudios de Administracion 1 Central Bank of Chile. Fisher, J. D. (1998). Credit market imperfections and the heterogeneous response of firms to monetary shocks. Working Paper Series, Macroeconomic Issues 96-23 Federal Reserve Bank of Chicago. Gale, D., and Hellwig, M. (1985). Incentive-Compatible Debt Contracts: The One-Period Problem. Review of Economic Studies, 52 , 647-63. Gertler, M., Gilchrist, S., and Natalucci, F. M. (2007). External Constraints on Monetary Policy and the Financial Accelerator. Journal of Money, Credit and Banking, 39 , 295-330. Glen, J., and Mondragon-Velez, C. (2011). Business cycle effects on commercial bank loan portfolio performance in developing economies. Review of Development Finance, 1 , 150 - 165. Goyal, V. K., and Wang, W. (2011). Debt Maturity and Asymmetric Information: Evidence from Default Risk Changes. Journal of Financial and Quantitative Analysis (JFQA), Forthcoming, . Juillard, M. (2004). Dynare Manual, version 4.1.0 . Technical Report. King, R. G., and Levine, R. (1994). Capital fundamentalism, economic development, and economic growth. CarnegieRochester Conference Series on Public Policy, 40 , 259-292. Levine, R. (2002). Bank-Based or Market-Based Financial Systems: Which Is Better? Journal of Financial Intermediation, 11 , 398-428. Lizarazo, S. (2011). Default risk and risk averse international investors. MPRA Paper 20794 University Library of Munich, Germany. Mefteh, S., and Oliver, B. R. (2007). Capital structure choice: the influence of confidence in France. In European Financial Management Association 2007 Annual Meeting. Mendoza, E. G. (1995). The Terms of Trade, the Real Exchange Rate, and Economic Fluctuations. International Economic Review, 36 , 101-37. Moskowitz, T. J., and Vissing-Jørgensen, A. (2002). The Returns to Entrepreneurial Investment: A Private Equity Premium Puzzle? American Economic Review, 92 , 745-778. Niu, X. (2009). Do Institutional Differences Affect Leverage Choice? International Business Research, 1. Pardo, C. (2010). Risk Aversion, Net Worth Effects and Real Fluctuations. Mimeo, Department of Economics, St. Joseph's University. Pardo, C. (2012). Risk Aversion, Business Volatility and Exchange Rate Regimes in Small Open Economies. Eastern Economic Journal, 38 , 167-188. Prasad, E., Agenor, P.-R., and McDermott, C. J. (1999). Macroeconomic Fluctuations in Developing Countries - Some Stylized Facts. IMF Working Papers 99/35 International Monetary Fund. Rampini, A. A. (2004). Entrepreneurial activity, risk, and the business cycle. Journal of Monetary Economics, 51 , 555-573. Rocha, K., and Garcia, F. A. (2004). The Term Structure of Sovereign Spreads in Emerging Markets: A Calibration Approach for Structural Models. Sandstrom, A. (2006). Political Risk and Firm Default Probability: Exploring export credits to high-risk countries. Technical Proceedings of the 2012 Pennsylvania Economic Association Conference 35 Report Department of Finance and Statistics, Swedish School of Economics and Business Administration. Souza-Sobrinho, N. (2010). Macroeconomics of bank interest spreads: evidence from Brazil. Annals of Finance, 6 , 1-32. van Stel, A., Carree, M., and Thurik, R. (2005). The Effect of Entrepreneurial Activity on National Economic Growth. Small Business Economics, 24, 311-321. Terra, P. R. S. (2009). Are leverage and debt maturity complements or substitutes? Evidence from Latin America. RAM. Revista de Administracao Mackenzie, 10 , 4 - 24. The World Bank (2010). World development indicators 2010. Townsend, R. M. (1979). Optimal contracts and competitive markets with costly state verification. Journal of Economic Theory, 21 , 265-293. Wahid, A. N. M., and Jalil, A. (2010). Financial Development and GDP Volatility in China. Economic Notes, 39 , 27-41. Wang, C. L., and Chen, Z. X. (2004). Consumer ethnocentrism and willingness to buy domestic products in a developing country setting: testing moderating effects. Journal of Consumer Marketing, 21 , 391-400. Proceedings of the 2012 Pennsylvania Economic Association Conference 36 RESPONSIVENESS OF THE U.S. TRADE FLOWS TO CHANGES IN CHINESE CURRENCY Orhan Kara Department of Economics and Finance West Chester University West Chester, PA 19383 ABSTRACT This study examines the trade flows between the U.S. and China to find the effect of changes in the renminbi on the U.S. trade deficit. The conclusions reveal that the U.S. imports from China are more sensitive to changes in the U.S. income than the U.S. exports to China to the changes in the Chinese income level. Furthermore, a one percent appreciation of the renminbi leads to an about 0.82 percent decrease in the value of the U.S. imports from China and an about two percent increase in the value of the U.S. exports to China. INTRODUCTION While encouraging a weaker dollar, Krugman (2011) stated the following: “But sensible policy makers have long known that sometimes a weaker currency means a stronger economy, and have acted on that knowledge. Switzerland, for example, has intervened massively to keep the franc from getting too strong against the euro. Israel intervened ever more forcefully to weaker the shekel.” Since the U.S. has high trade deficit, especially with China, the argument for a weaker dollar by Krugman (2011) was against China. China pegged her currency to the U.S. dollar and then to the British pound during the 1950s and 1960s until the breakdown of Bretton Woods Agreement (Eichengreen, 2006). Then it pegged to a basket of currencies. China started a dual exchange rate policy consisting of an official rate and market rate (Jin, 2009). In trade, China used the market rate (1 dollar = 2.8 yuan) to promote exports and the official rate was used for non-trade transactions, mainly remittance and tourism (Eichengreen, 2006; Jin, 2009). However, China abandoned the dual exchange rate system and fixed the renminbi at the prevailing market rate of 8.7 to the dollar and allowed to float in a narrow margin ( Eichengreen, 2006). On July 21, 2005 the Chinese government implemented a new currency reform, pegging the renminbi a basket of eleven currencies that were important in terms of their shares in Chinese trade (Sun, 2010). With a tightly controlled exchange rate system and export promoting strategies, China has become the leading exporter of the world (Chen, Rau, and Chiu; 2011). Due to a growing high volume of trade flows between the U.S. and China, the U.S. trade deficit has increased over the years. Several public officials and commentators speculated that China has been engaging in currency manipulation and that various proposals appeared to take action against China (Staiger and Sykes, 2010). Specifically China has been accused of keeping her currency undervalued which led to the increase of China’s export. Scholars examining the degree of undervaluation of Chinese currency estimated that the renminbi (RMB), the Chinese currency, is undervalued as high as fifty percent (Tyers and Zhang; 2011). Under the political pressure, the U.S. government has entered exchange negotiations with China (Cao, Cao, Prasad, and Shen; 2011). Recently, several proposals in Washington has been put forward, ranging from insisting that the Treasury Department designate China as a ‘currency manipulator’ and take this matter to the International monetary Fund, having the U.S. trade Representative file a complaint to the World trade organization, and imposing antidumping and countervailing duties on China’s exports to the U.S. (Staiger and Sykes, 2010). However, there are other studies arguing that the Chinese currency is not undervalued and revaluing the RMB would not reduce the U.S. trade deficit with China (McKinnon and Schabl; 2009, 2008). Therefore, this study investigates the trade flows between China and the U.S., focusing on the effect of exchange rates. Specifically, this study examines the short run and long run effect of changes in dollar renminbi exchange rate on trade flows by estimating export and import demand equations. To that end, the next section reviews the literature. Then, the model and methodology are explained. The section after that presents the results. The final section concludes the study. LITERATURE REVIEW China’s export promoting growth strategy led to a remarkable economic development over thirty years Proceedings of the 2012 Pennsylvania Economic Association Conference 37 with a 9.8% average growth rate (Xu, 2010). This reduced the number of people living below the poverty level and increased prosperity of the middle class. Although this great success story also brought many problems as the trade deficits of several countries with China increased, particularly the U.S. As a result, rich literature has been developed . The main criticism is that China has manipulated her currency with keeping low at the expense of her trading partners. With respect to currency manipulation, many studies tried to estimate how undervalued the Chinese currency, renminbi. According to Cheung, Chinn, and Fujii (2010), renminbi was undervalued as much as fifty percent with the pre-2005 international Comparison program benchmark data. Cheung et al. (2010) also presented other estimates undertaken by different researches. For example, Cline and Williamson (2010) found that Chinese currency was thirty three percent undervalued. Similarly, Wang and Hu (2010) estimated a 17.5 percent undervaluation while Subramanian (2010) reported a value of thirty one percent undervalued renminbi. Goldstein and Lardy (2009) argued that renminbi was undervalued twenty one percent and in another study Goldstein (2004) found that the renminbi was undervalued by 15-30 percent. An estimate of twenty three percent undervalued renminbi was found by Coudert and Couharde (2004). While Tenengauzer (2010) found ten percent undervalued renminbi, Stupnytska, Stolper, and Meechan (2009) estimated only 2.56% undervaluation. On the other hand, Bergsten (2010) argued that the renminbi was undervalued by forty percent against dollar, and about twenty five percent on a trade weighted basis. However, there are also studies that the renminbi is actually overvalued or no evidence of undervalued renminbi. For instance, Goh and Kim (2006), employing a data set from 1972 to 2002, found no evidence that the renminbi was significantly undervalued. Wang (2004) claimed that the renminbi was actually overvalued. Similarly, Funke and Rahn ( 2005) claimed that the renminbi was not substantially undervalued. McKinnon and Schnabl (2008) also argued that the Chinese currency was not undervalued. Hu and Chen (2010) found that the Chinese currency was overvalued by 13.4 percent. Hoover, Cheung et al. (2010) estimated an overvaluation of 16.8 percent based on the real U.S. exchange rate and thirty six percent on M trade weighted exchange rate basis. Although no agreed upon conclusion has been reached as to whether the Chinese currency is undervalued and how much it was undervalued, some researchers (for example Krugman, 2011) suggest the Chinese government appreciate her currency. In addition, the U.S. government has been pressuring for appreciation. After finding that the renminbi was undervalued, Goldstein (2006) suggested a ten to twenty five percent appreciation. According to Stager and Sykes (2010) Paul Krugman and C. Fred Bergsten claimed that the renminbi must appreciate twenty five to forty percent to restore global imbalances. Furthermore, the U.S. government has been pressuring and negotiating with the Chinese government to increase the value of renminbi and the negotiations are likely to continue (Cao, Cao, Prasad, and Shen, 2011). However, studies are also divided if an appreciation of the renminbi is going to help to balance global imbalances in general and the U.S. trade deficit in particular. Bahmani-Oskooee and Wang (2006) investigated the short run and long run effect of depreciation of the renminbi on the trade balance of China with thirteen trading partners and concluded that depreciation had a favorable impact on the trade balance with the U.S. in the long run implying an appreciation would deteriorate the trade balance, there was no short run effect. Similarly, Thorbecke (2006) found that an appreciation of renminbi would reduce the trade deficit of the U.S. with China and that an appreciation of the Chinese currency would also lead to a generalized appreciation of Asian currencies that reduce the U.S. trade deficit further. Moreover, Chen, Rau, and Chiu (2011) examined the determinants of China’s exports to the U.S. and Japan and empirically tested the response to the real appreciation of the renminbi with quarterly data on seventy one industries for 1997 to 2007. They concluded that appreciation of the Chinese currency had negative effect on the exports to the U.S. On the other hand, several studies provided compelling arguments, both theoretically and empirically, that an appreciation of renminbi would not lead to an improvement in the U.S. trade balance. For example, the Chinese currency appreciated about twenty percent between 2005 and 2008, leading to an increase in the imports of China; however, the U.S. imports from China also increased at a much faster rate (Cheung et al.,2010). Another argument suggests that even a forty percent appreciation in renminbi would not reduce the U.S. trade deficit since the U.S. will replace the imports from China with goods from other countries as was the case when the Japanese Proceedings of the 2012 Pennsylvania Economic Association Conference 38 currency appreciated, the U.S. replaced imports from Japan with imports from other countries (Woo, 2008). Witte (2009) noted that there was a significant deflationary trend in Chinese exports during the appreciation of the renminbi after 2005, due to the high market share of Chinese exporters, and that appreciation of the renminbi ‘resulting from a more flexible exchange rate regime, would have a little impact on the Chinese export prices to the U.S. Likewise, Zhang, Fung and Kummar (2006) offered several explanations as to why the appreciation of renminbi would not solve the U.S. trade deficit. Among the explanations were China’s exports to the U.S. were necessity goods, that the U.S. trade deficit was due to the fiscal deficit and low saving rate, and the U.S. have been increasingly importing more manufacturing goods as the country moved to a service oriented economy. Furthermore, Chinese government officials vehemently denied the allegations that the renminbi was undervalued by the government (Stranger and Sykes, 2010: Moosa, 2011). Chinese government policies in terms of monetary and exchange rate were also supported by academics. For instance, McKinnon and Schnabl (2008) and McKinnon, Lee, and Wang (2010) argued in favor of Chinese currency policy and its monetary policies to control inflation and sustaining favorable economic growth. Given the controversies surrounding the Chinese currency and calls for appreciation of renminbi, only did few studies estimate the effect of the renminbi appreciation on the U.S. trade deficit (Thorbecke,2006). Thorbecke (2006) estimated that a ten percent appreciation of the renminbi in 2005, the gap between exports would have been reduced to 1.4 percent of the U.S. GDP from 1.6 percent. As Thorbecke (2006) pointed out other studies reported similar results. For example, ten percent valuation of the renminbi would cause a fifteen billion dollar decline in the Chinese trade surplus (Park ,2004). Kamada and Takagawa (2005) asserted that a ten percent increase in the value of renminbi would decrease the Chinese trade surplus by 0.5 percent of Chinese GDP. Similarly, Marquez and Schindler (2007) concluded that ten percent appreciation of the renminbi decreased the Chinese exports by one percentage point while the effect on the imports was negligible. Finally, the price elasticities of exports and imports were estimated around one (Thorbecke ,2006). In summary, several studies provided evidence of misalignment of the renminbi. Estimates of the amount of undervaluation were as high as fifty percent; whereas, some claimed overvaluation of the currency up to thirty six percent. The Chinese government refused any currency manipulation, supporter by academics while the U.S. government and several scholars urged an appreciation of the renminbi. Even the renminbi is appreciated by some amount in response to calls, the evidence as to how much appreciation would improve the U. S. trade deficit is scarce and no agreed upon conclusion has been reached. Therefore, this study aims at providing further evidence by empirically estimating the U.S. export and import demand functions to determine the effect of changes in the renminbi on the U.S. trade deficit with China. THE MODEL AND METHODOLOGY To determine the effect of exchange rate on exports and imports, we develop two equations: import and export demand functions that are applied to the data for the U.S and China trade flows. The import and export demand functions adopted in this study are conventional in the sense that they were also employed by previous studies. In formulating the import demand function we follow BahmaniOskooee and Kara (2008) and use the following log linear formulation:  PM ο£Ά ln M td = α + β ln Yt + φ ln ο£· + Ο ln Et ο£ PD ο£Έ (1) Where Md is demand for imports, Y is national income, PM is price of imports, PD is price of domestic goods, and E is exchange rate. We would expect an estimate of β to be positive indicating that at high level of income we import more. However, if the increase in income is due to an increase in production of import substitute goods, a country may import less yielding a negative estimate of β. An increase in import price relative to domestic price levels is expected to depress the import volume resulting in a negative estimate for αΆ². Finally, since the exchange rate is defined as number of units of foreign currency per unit of domestic currency, a decrease in E or a depreciation of the domestic currency is expected to reduce the import volume, thus estimate of α΅ is expected to be negative. Equation (1) indicates the long-run relationship among the variables of import demand function. However, since we are also interested in short run response of imports to changes in the exchange rate, we need to introduce the short-run dynamics into equation (1). In order to do so, we follow Pesaran et al. (1996, 1999, & 2009) and express the equation in Proceedings of the 2012 Pennsylvania Economic Association Conference 39 an Autoregressive Distributed Lag (ARDL) format as follows. n n i =0 i =0 β ln M td = α + ∑ β i β ln Yt −i + ∑ γ i β ln( n + ∑θ i β ln M td−i + δ1 ln Yt −1 + δ 2 ln( i =1 n PM )t −i + ∑ λi β ln Et −i PD i =0 PM ) t −1 + δ 3 ln Et −1 PD (2) + δ 4 ln M t −1 + ut Again, the variables are defined as before. The error term, ut, is assumed iid(0,σu2 ). The export demand function is structurally very similar to the import demand function and is assumed to be the function of income level of a country’s trading partner, export prices of the country, export prices of the country’s trading partners, and exchange rate. As formulated by Bahmani-Oskooee and Kara (2008), the export demand equation in log linear form as follows: ln X td = a + b ln YWt + c ln( PX ) t + d ln Et (3) PXW Where Xd is demand for exports, YW is income level of trading partner (China), PX is export prices of the U.S., PXW is export prices of other countries, and E is exchange rate. We assume that the Chinese demand for the U.S. exports (Xd) has a positive relation with the income (YW). However, this relation could be negative if the increase in the income is due to an increase in import substitute goods in China. Exports are expected to have a negative relation with the relative price of a country's export price over world export price (PX/PXW). Finally, given the definition of exchange rate, E, if currency depreciation is to stimulate exports, we would expect an estimate of d to be negative. However, equation (3) only provides the long-run estimates. Since we are interested in short run response, we incorporate short-run dynamics into equation (3) and express it in ARDL. m m PX β ln X td = α + ∑ β i β ln YWt −i + ∑ γ i β ln( ) t −i PXW i =0 i =0 m m i =0 i =1 + ∑ λi β ln Et −i + ∑ φi β ln X td−i + θ1 ln YWt −1 PX + θ 2 ln( ) t −1 + θ 3 ln Et −1 + θ 4 ln X t −1 + vt (4) PXW In equation (4), the variables are defined the same as before. The error term is assumed to be iid(0,σv2 ). Due to a relative ease in estimation, we use ARDL approach developed by Pesaran et al. (1996, 2009). First, we construct two null hypotheses (stating the non-existence of cointegration) to determine the cointegration among variables in each equation and test them according to the test statistics obtained from the models. H1: δ1≠0, δ2≠0, δ3≠0, δ4≠0 (5) H0: δ1=δ2=δ3=δ4=0 H0: θ1=θ2=θ3=θ4=0 H1: θ1≠0, θ2≠0, θ3≠0, θ4≠0 (6) After, we make a decision, we proceed to the second stage of the ARDL method and estimate equations (2) and (4). In this stage, we estimate four models based on the following criteria: the r bar square criterion, Akaika information criterion (AIC), Schwarz Bayesian criterion (SBC), and HannanQuinn criterion (HQC). Then we select one model based on examining diagnostic statistics. The data for this study cover the period from 2000 to 2011. All data were collected on a quarterly basis, ranging from the first quarter of 2000 to the last quarter of 2011 (2000Q1-2011Q4). Part of data used in this study was extracted from the “IMF Financial Statistics and IMF Direction of Trade Statistics, online”. The International Monetary Fund, (IMF), collects data on countries for economic variables on an annual, quarterly, and monthly basis. RESULTS The error-correction models were estimated as outlined by equations (2) & (4) using quarterly data over 2000Q1-2011Q4 period for trade flows between the United States and China. W first tested for the joint hypotheses in equations (5) and (6). Since we have quarterly data, we imposed four lags. As the testing can be executed based on; a) no intercept, no trend variable, b) intercept and no trend variable, and c) intercept and trend variables, we performed three F tests. In most cases the F-statistic was significant. For the case of intercept and no trend variable, the F values were much higher than upper bound critical value at 95% significance level, which indicated the existence of cointegration and our estimates are based on intercept and no trend equations (Table 1). Since the variables in both models were cointegrated, we moved to the second stage of estimation and estimated equation (2) by imposing eight maximum lags on each first differenced variable. Using the ARDL Approach, ARDL(6,6,0,6) was selected based on Hannan-Quinn Criterion, since it gave the best results. Table 2 reports the long run estimates along with ARDL estimates. Long run estimates confirm our expected signs. All of the variables are statistically significant, except for price variable. The biggest effect on imports comes from income, followed by the exchange rate. Although it is not statistically significant, the effect of price is relatively Proceedings of the 2012 Pennsylvania Economic Association Conference 40 small. Since we have the log of variables, the coefficients can also be interpreted as elasticities. Therefore, the U.S. imports are income elastic, but they are close to unit elastic with respect to prices and exchange rate. The estimated coefficient for the exchange rate suggests that a one percent depreciation in the Chinese currency leads to about 0.8 percent increase in the imports. In other words, ten percent appreciation of the renminbi reduces the U.S. imports from China by eight percent. If the Chinese government answers the calls by the critics and decrease the value of its currency against the dollar, its exports to the U.S. would go down by about twenty percent, all else will stay constant. ARDL estimates indicate that some of the lagged variables are statistically significant. As indicated with changing signs in the lags, the variables adjust over time to long run values. Diagnostic tests show the robustness of the estimates. Under the variables column in the table, the number in parenthesis indicates the lags and L means log of the variable, which is the same in the following tables in the study. Table 3 shows the error correction estimates. Again, under the variables column in the table, the symbol Δ in front of the variables refers to the differencing. For example, ΔLE means one period difference. The number after the variable refers to the further differencing in the variables. For instance, ΔLE1 indicates one period difference of ΔLE, and so on, which is the same in the remaining tables. Ecm(-1) is the error correction coefficient, which means how fast the economy returns to the equilibrium (long run values) once it is shocked. After a shock (change in exchange rates), an adjustment process takes place during which the economy returns to its long-run equilibrium values. The expected sign of the error correction coefficient is negative, indicating that when the short- run values overshoot the long-run equilibrium values, the adjustment is downward, as expected or vice versa (Greene, 2008). Tables 4 and 5 present the results from the export equation. ARDL(6,8,0,5) was selected based on RBar Criterion, since it provided the best results. The long run estimates along with ARDL estimates are given in Table 4. Similar to the import case, long run estimates confirm our expected signs. Again, all of the variables are statistically significant, except for price variable. In terms of the magnitudes of the estimated coefficients, the biggest effect on exports comes from income, followed by the exchange rate, although the values are very similar. The estimated coefficient for the exchange rate suggests that a one percent depreciation in the Chinese currency leads to about 2 percent decrease in the U.S. exports to China, or a ten percent appreciation of the renminbi increases the U.S. exports to China by about twenty percent, which is about the same effect that an increase in Chinese GDP would have. Compared to the imports from China to the U.S., responsiveness of the U.S. exports to China is twice as much to the changes in the exchange rate between the renminbi and dollar, and one third to the changes in the income. ARDL estimates illustrate that changing signs in the lags shows that variables adjust over time to long run values. Error correction estimates for the exports are given in table 5. The error correction coefficient in the export case is again statistically significant and negative, indicating that when the short- run values overshoot the long-run equilibrium values, the adjustment is downward. We have examined the diagnostic statistics (given in tables 2-5) for the models in this study as well as visual examination of the plot of residuals and fitted values. In order to check the structural stability of the models, we performed two additional tests: cumulative sum (CUSUM) and cumulative sum of squares (CUSUM of squares) of residuals. These tests were obtained by examining the plots of cumulative sum and cumulative sum of squares against time. In all four four cases of the model, all of them pass the CUSUM and CSUM of squares tests that the graphs do not cross the straight lines at 95 percent level. This is what we expected if we consider the diagnostic tests that were robust. CONCLUSIONS As the U.S. trade deficit has increased over the years with China, several public officials and commentators speculated that China has been engaging in currency manipulation and that various proposals appeared to take action against China. Although there seems to be some consensus that Chinese currency is undervalued, previous research often produced mixed result as to how much the currency is undervalued and some research even suggested overvaluation of the renminbi. However, the evidence as to how much appreciation would improve the U. S. trade deficit is scarce and no agreed upon conclusion has been reached. This study investigated the trade flows between the U.S. and China and estimated the effect of changes in the renminbi on the U.S. trade deficit with China. The main findings of this study can be summarized as follows. First, the U.S. imports from China are more sensitive to changes in the U.S. income than the U.S. exports to China to the Chinese income level. In Proceedings of the 2012 Pennsylvania Economic Association Conference 41 particular, a one percent increase in the U.S. income increases imports from China by about 6.9 percent while a one percent increase in the Chinese GDP raises the U.S. exports to China by about 2.1 percent. Second, the U.S. imports from China are less responsive to the changes in the renminbi-dollar exchange rates. It is found that a one percent appreciation of the renminbi leads to about a 0.82 percent decrease in the value of the U.S. imports from China. Third, the U.S. exports to China are more responsive to the changes in the renminbi. Specifically, a one percent appreciation of the Chinese currency causes a two percent increase in the value of the U.S. exports to China. Finally, the renminbi should appreciate about fifty six percent in order to eliminate the U.S. trade deficit with China based on 2011 trade flows. Moreover, a ten percent appreciation (as some scholars call for) would reduce the U.S. trade deficit with China by 28.5 percent and forty percent appreciation of the renminbi (as suggested by researchers and politicians) decreases it by approximately seventy percent. TABLE 1. F-TEST RESULTS FOR COINTEGRATION F-values* 8.3768 8.2151 Import Equation Export Equation *. Critical values for F-Test at 95%(intercept, No Trend): Lower Bound=2.850 & Upper Bound= 4.049 Variables LY LPMPD LE INPT TABLE 2. LONG-RUN COEFFICIENT ESTIMATES FOR IMPORT Coefficient Standard Error T-Ratio [P-value] 6.8971 0.2917 23.6416 [.000] -0.6077 0.2129 -2.8543 [.011] -0.8177 0.0753 -10.8547 [.000] -5.1525 1.4503 -3.5527 [.002] ARDL Estimates LM(-1) LM(-2) LM(-3) LM(-4) LM(-5) LM(-6) LY LY(-1) LY(-2) LY(-3) LY(-4) LY(-5) LY(-6) LPMPD LE LE(-1) LE(-2) LE(-3) LE(-4) LE(-5) LE(-6) INPT Coefficient Standard Error T-Ratio [P-value] 0.1451 -0.1306 -0.3336 0.5120 -0.4642 -0.3110 2.4640 4.4160 1.9379 1.4961 0.3293 -2.6426 2.9133 -0.9617 3.9383 -1.6512 -2.0906 1.6906 -0.7509 -4.8780 2.4479 -8.1534 0.1877 0.1260 0.1194 0.1454 0.1484 0.1786 1.4614 2.1505 2.0992 1.8894 1.9226 1.9744 1.1850 0.3599 1.2687 2.1690 2.1711 2.2402 2.0957 2.2625 1.5830 3.5668 .77272 -1.0369 -2.7928 3.5218 -3.1279 -1.7412 1.6861 2.0534 .92315 .79182 .17129 -1.3384 2.4586 -2.6720 3.1043 -.76130 -.96291 .75467 -.35828 -2.1560 1.5464 -2.2859 R-Bar-Squared F-Stat. Equation Log-likelihood DW-statistic [.450] [.313] [.012] [.002] [.006] [.099] [.109] [.055] [.368] [.439] [.866] [.197] [.024] [.016] [.006] [.456] [.348] [.460] [.724] [.045] [.139] [.035] 0.9948 354.1942 102.2746 2.0135 Proceedings of the 2012 Pennsylvania Economic Association Conference 42 TABLE 3. ERROR CORRECTION ESTIMATES FOR IMPORTS Coefficient Variables ΔLM1 ΔLM2 ΔLM3 ΔLM4 ΔLM5 ΔLY ΔLY1 ΔLY2 ΔLY3 ΔLY4 ΔLY5 ΔLPMPD ΔLE ΔLE1 ΔLE2 ΔLE3 ΔLE4 ΔLE5 ecm(-1) 0.7275 0.5968 0.2633 0.7752 0.3110 2.4640 -4.0340 -2.0961 -0.6000 -0.2707 -2.9133 -0.9617 3.9383 3.5810 1.4904 3.1810 2.4301 -2.4479 -1.5824 R-Bar-Squared F-Stat. Equation Log-likelihood DW-statistic Variables LYW LPXPXW LE INPT LX(-1) LX(-2) LX(-3) LX(-4) LX(-5) LX(-6) LYW LYW(-1) LYW(-2) LYW(-3) LYW(-4) LYW(-5) LYW(-6) LYW(-7) Standard Error 0.3148 0.3376 0.2631 0.1615 0.1786 1.4614 2.3600 2.0117 1.6652 1.5515 1.1850 0.3599 1.2687 1.5445 1.4820 1.4047 1.4268 1.5830 0.4052 T-Ratio [Pvalue] 2.3107 1.7680 1.0004 4.8001 1.7412 1.6861 -1.7093 -1.0420 -.36035 -.17448 -2.4586 -2.6720 3.1043 2.3186 1.0056 2.2646 1.7033 -1.5464 -3.9056 0.9540 43.7122 102.2746 2.0135 TABLE 4. LONG-RUN COEFFICIENT ESTIMATES FOR EXPORTS Coefficient Standard Error T-Ratio [Pvalue] 31.2098 2.0576 0.0659 -1.5567 -0.2960 0.1902 8.7183 1.9903 0.2283 12.9838 9.4101 0.7248 ARDL Estimates Coefficient Standard Error T-Ratio [Pvalue] 1.3391 0.2360 0.1763 -3.0484 -0.5450 0.1788 .17617 0.0347 0.1971 -3.5186 -0.6719 0.1910 -.42478 -0.0787 0.1853 -2.7762 -0.3877 0.1397 2.9823 2.9482 0.9886 1.1727 1.2734 1.0859 -.94614 -1.0676 1.1283 -1.3301 -1.6244 1.2213 .37886 0.5275 1.3923 -.19913 -0.2200 1.1050 2.1380 2.4221 1.1329 2.0530 2.5130 1.2241 Proceedings of the 2012 Pennsylvania Economic Association Conference 43 LYW(-8) -1.8083 LPXPXW -0.7142 LE -4.7293 LE(-1) 4.0603 LE(-2) 0.8726 LE(-3) 3.7070 LE(-4) -7.3483 LE(-5) 8.2392 INPT 22.7020 R-Bar-Squared F-Stat. Equation Log-likelihood DW-statistic 1.1368 0.4865 2.2919 4.0018 4.1307 4.0832 3.9137 2.5720 4.4326 -1.5907 -1.4679 -2.0635 1.0146 .21125 .90789 -1.8776 3.2035 5.1216 0.98889 158.0987 77.6229 2.0439 TABLE 5. ERROR CORRECTION ESTIMATES FOR EXPORTS Coefficient Standard Error T-Ratio value] ΔLX1 1.6485 0.37784 4.3630 ΔLX2 1.1036 0.33926 3.2528 ΔLX3 1.1383 0.26104 4.3606 ΔLX4 0.4664 0.18104 2.5762 ΔLX5 0.38771 0.13966 2.7762 ΔLYW 2.9482 0.98857 2.9823 ΔLYW1 -0.74232 1.2661 -.58630 ΔLYW2 -1.8099 1.386 -1.3058 ΔLYW3 -3.4343 1.6874 -2.0353 ΔLYW4 -2.9068 2.1062 -1.3801 ΔLYW5 -3.1269 1.8793 -1.6639 ΔLYW6 -0.70474 1.494 -.47172 ΔLYW7 1.8083 1.1368 1.5907 ΔLPXPXW -0.71418 0.48652 -1.4679 ΔLE -4.7293 2.2919 -2.0635 ΔLE1 -5.4705 2.8801 -1.8995 ΔLE2 -4.5979 2.5826 -1.7804 ΔLE3 -0.8909 2.6717 -.33346 ΔLE4 -8.2392 2.572 -3.2035 ecm(-1) -2.4125 0.44563 -5.4138 R-Bar-Squared 0.78943 F-Stat. 8.4104 Equation Log-likelihood 77.6229 DW-statistic 2.0439 Variables Proceedings of the 2012 Pennsylvania Economic Association Conference [P- 44 REFERENCES Bahmani-Oskooee, M. and Wang, Y. (2006). The J Curve: China Versus Her Trading Partners. Bulletin of Economic Research, Vol. 58(4), 323-343. Bahmani-Oskooee, M. and Kara, O. (2008). Relative Responsiveness of Trade Flows to A Change in Prices and Exchange Rate in Developing Countries. Journal of Economic Development, Vol. 33(1), 147-163. Cao, E. Y., Cao, Y., Prasad, R., & Shen, Z. (2011). U.S.- China Exchange Rate Negotiation: Stakeholders’ Pparticipation and Strategy Deployment. Business and Politics, Vol. 13(3), 1-23. Chen, K., Rau, H., & Chiu, R. (2011). Determinants of China’s Exports to the United States and Japan. The Chinese Economy, Vol. 44 (4), 19-41. Cheung, Y., Chinn, M.D., and Fujii, E. (2010a). Measuring Renminbi Misalignment: Where do we Stand? Korea & World Economy, Vol. 11(2), 263-296. Cheung, Y., Chinn, M.D., and Fujii, E. (2010b). Measuring Misalignment: Latest Estimates for the Chinese Renminbi. In Simon Evenett ed., The US-Sino Currency Dispute: New Insights from Economics, Politics and Law, A Vox EU.org Publication, Chapter 10, 79-90. Cline, W.R. and Williamson, J. (2010). Notes on Equilibrium Exchange Rates. Policy Brief PB 10-2, Peterson Institute for International Economics, Washington, DC. Coudert, V. and Couharde, C. (2005). Real Equilibrium Exchange Rate in China. Working Paper No: 2005-01.” CEPII, Paris. Coudert, V. and Couharde, C. (2007). Real Equilibrium Exchange Rate in China: Is the Renminbi Undervalued? Journal of Asian Economics, Vol. 18(4), 568-594. Eichengreen,B.(2006). Is a Change in the Renminbi Exchange Rate in China’s Interests? Asian Economic Papers 4:1, 40-75. Funke, M. and Rahn, J. (2005). Just How Undervalued is the Chinese Renminbi? Hamburg University Paper, Blackwell Publishing. 465-489. Greene, W. H., (2008). Econometric Analysis, 6th Ed. New Jersey: Pearson/Prentice Hall. Goh, M. H. and Kim, Y. (2006). Is the Chinese Renminbi Undervalued? Contemporary Economic Policy, Vol. 24(1), 116-126. Goldstein, M. (2004). Adjusting China’s Exchange Rate Policies. Working Paper: 04-1. Peterson Institute for International Economics, Washington, DC. Goldstein, M. (2006). Renminbi Controversies. Cato Journal, Vol. 26(2), 251-266. Goldstein, M. and Lardy, N. (2009). The Future of Chinas Exchange Rate Policy. Policy Analysis in International Economics, No. 87, Peterson Institute for International Economics, Washington DC. Hu, C. and Chen, Z. (2010). Renminbi Already Over appreciated: Evidence from FEERs (1994-2008). Chinese Economist, Vol. 26, 64-78. Jin, G. (2009). Examining the Exchange Rate Regime for China. International Research Journal of Finance and Economics, 25, 64-77. Proceedings of the 2012 Pennsylvania Economic Association Conference 45 Kamada, K. and Takagawa, I. (2005). Policy Coordination in East Asia and Across Pacific. Bank of Japan Working Paper Series No. 05-E-4. Tokyo. Krugman, P. (Oct. 2, 2011). Holding China to Account. The New York Times. Marquez, J. and Schindler, J. (2007). Exchange-rate Effects on China’s Trade. Review of International Economics, Vol. 15(5), 837-853. McKinnon, R. & Schnabl, G. (2008). China’s Exchange Rate Impasse and the Weak U.S. Dollar. CESinfo Working Paper Series, No. 2386. McKinnon, R. & Schnabl, G. (2009). The Case for Stabilizing China’s Exchange Rate: Setting the Stage for Fiscal Expansion. China & World Economy, Vol. 17(1), 475-485. McKinnon, R. & Lee, B., and, Wang, Y. D. (2010). The Global Credit Crisis and China’s exchange rate: Setting. The Singapore Economics Review, Vol. 55(2), 253-272. Moosa, I. (2011). On the U.S. –Chinese Trade Dispute. Journal of Post Keynesian Economics, Vol. 34(1) 85-111. Park, C. (2005). Coping with Global Imbalance and Asian Currencies (Asian Development Bank of Manila, Philippines http://www.adb.org/sites/default/files/pub/2005/PB037.pdf Accessed on May 21, 2012 Pesaran, B., & Pesaran, M.H., (2009). Time Series Econometrics. Oxford: Oxford University Press, 2009. Pesaran, M.H., Shin, Y & Smith, R. J., (1996). Testing For the Existence of a Long-run Relationship, In DAE Working paper, V. 9622, Department of Applied Economics, University of Cambridge, 1996. Pesaran, M.H., & Shin, Y., (1999). “An Autodistributed Lag Modeling Approach to Cointegration Analysis”, in (ed) S Strom, Econometrics and Economic Theory in the 20th Century: The Ragnar Frisch Centennial Symposium , Cambridge University Press, Cambridge, 1999. Schmitt, F. G., Ma, L., and Angounou, T. (2011). Multifractal Analysis of the Dollar-Yuan and Euro-Yuan Exchange Rates Before and After the Peg. Quantitative Finance, Vol. 17(4), 505-513. Staiger, R . W. & Sykes, A. O. (2010). ‘Currency Manipulation’ and World Trade. World Trade Review, Vol. 9(4), 583-627. Stupnytska, A., Stolper, T., and Meechan, M. (2009). GSDEER on Track: Our Improved FX Fair Value Model. Global Economics Weekly, No. 09/38, Goldman Sachs Global Economics. Subramanian, A. 2010. New PPP- Based Estimates of Renminbi Undervaluation and Policy Implications. Policy Brief PB 10-18, Peterson Institute for International Economics, Washington, DC. Sun, J. (2010). Retrospect of the Chinese Exchange Rate Regime after Reform: Stylized Facts during the Period from 2005 to 2010. China & World Economy, 18(6), 19-35. Tenergauzer ID. (2010). RMB: The People’s Currency. EM FX and Debt spotlight, Bank of America-Merrill Lynch. Thorbecke, W. (2006). How Would and Appreciation of the Renminbi Affect the U.S. Trade Deficit with China. Topics in Macroeconomics, Vol. 6 (3), 1-15. Tyers, R. & Zhang, Y. (2011). Appreciating the Renminbi. The World Economy, Vol. 34(2), 265-297. Wang, T. and Hu, H.(2010). How Undervalued Is the RMB? Asian Economic Perspectives, UBS Investment Research. Proceedings of the 2012 Pennsylvania Economic Association Conference 46 Woo, W. T. (2008). Understanding the Source of Friction in U.S.-China Trade Relations: The Exchange Rate Debate Diverts Attention from Optimum Adjustment. Asian Economic Papers: 7.3,61-95. Xu, X. (2010). China’s Export-led Growth Strategy. China & World Economy, Vol. 18(4), 18-33. Witte, M.D. (2009). Pricing to Market: Chinese Export Pricing to the USA after the Peg. China & World Economy, Vol. 17(2), 65-78. Zhang, J; Fung, H., and Kummer, D. (2006). Can Renminbi Appreciation Deduce the U.S. Trade Deficit? China & World Economy, Vol. 14(1), 44-56. Proceedings of the 2012 Pennsylvania Economic Association Conference 47 RECENT PENNSYLVANIA JOB TRENDS: EFFECTS OF SHALE? (2012) Jay Bryson, Tim Quinlan and Joe Seydl* Wells Fargo Securities, LLC ABSTRACT This paper examines employment trends in Pennsylvania, with an emphasis on investigating the effects that natural gas drilling has had on the commonwealth’s labor market in recent years. We consider county-level employment data and find that natural gas drilling is having a noticeable impact on employment in the counties in which most of the drilling is occurring. We also find statistical evidence to suggest that the employment effects of natural gas drilling in the state are beginning to spill over into counties in which drilling is not occurring. By constructing a regression based on our estimates of the spillover employment effects related to natural gas drilling, we then forecast employment growth for the overall commonwealth under three different scenarios: an optimistic scenario, a pessimistic scenario and a midpoint scenario. Under our midpoint scenario, which is predicated on what may be the most realistic set of assumptions, statewide employment would increase by about 570,000 jobs by the end of 2020, roughly equivalent to the job creation experience in Pennsylvania during the 1990s. Of the 570,000 new jobs created by 2020, about 250,000 would be related to the spillover effects of natural gas drilling. Whether the extra boost to employment that is tied to natural gas exploration and production in Pennsylvania is “worth it” depends on the reader’s assessments of the economic, environmental and social issues that are involved. INTRODUCTION Following the deepest downturn in the post-World War II era, economic expansion has taken hold in the United States again. Real Gross Domestic Product (GDP) has grown for 11 consecutive quarters, and recent data suggest that the expansion remains intact in the current quarter. Nationwide employment is also growing again, although it lags behind output. Whereas real GDP has finally surpassed its previous peak, nonfarm payrolls in the United States have, at the time of this writing, retraced less than half of the 8.8 million jobs that were shed between early 2008 and early 2010. Although we do not yet have Gross State Product data for Pennsylvania for 2011, available indicators suggest that the commonwealth’s economy grew during 2011 and that the expansion has continued into the current year. For example, the Pennsylvania Coincident Index, which is published by the Federal Reserve Bank of Philadelphia, is currently expanding at a pace faster than that at any point during the 2003-2007 expansion (Figure 1). Recent gains in nonfarm employment in Pennsylvania are also consistent with continued economic expansion. Since reaching a bottom in February 2010, nearly 160,000 jobs have been added in the commonwealth. This 2.8 percent increase in Pennsylvania payrolls is in line with the job growth experienced at the national level (Figure 2). Interestingly, job growth in the commonwealth relative to that at the national level has been much stronger in the current cycle than what had been the case in the 2003-2007 cycle, during which job growth in Pennsylvania lagged considerably behind job growth for the country as a whole. PENNSYLVANIA EMPLOYMENT: SECTORAL AND REGIONAL CHARACTERISTICS So what sectors are creating jobs in the Keystone State? Figure 3 shows that the gains have been generally broad based.1 The sector comprised of education and healthcare employs 20 percent of the commonwealth’s workers and has accounted for one-third of all the jobs that have been created in the state since employment bottomed in February 2010. Trade, transportation and utilities, professional and business services, manufacturing, leisure and hospitality, and construction have also generated jobs over the past two years. Notably, the natural resources and mining sector, which employs less than 1 percent of the commonwealth’s workers, has accounted for 8 percent of the overall job growth in Pennsylvania. In other words, the natural resources and mining sector, which includes the shale gas industry, is currently an outperformer in terms of employment growth. Since drilling in the Marcellus formation began in earnest during the past decade, there has been much speculation, both in the popular media and in academia, about the employment effects from the shale industry in Pennsylvania. Not surprisingly, some studies have found significant employment effects from the shale industry, while other studies have been less effusive.2 Before turning to our formal econometric analysis, we undertake a quick visual examination of the employment experience in the commonwealth over the past few years. To look at the local impact of the shale gas industry, we divide Pennsylvania into 14 counties that account for 90 percent of the shale gas wells drilled to date in the Keystone State and the commonwealth’s other 53 counties. We call the former group primary shale (PS) counties and the latter group we refer to as other counties (Figure 5). Between mid-2003 and early 2008, the period that marked the last cyclical employment upturn in Pennsylvania, there was little material Proceedings of the 2012 Pennsylvania Economic Association Conference 48 difference between the employment outcomes in these two groups of counties, especially once employment gains really started to gather pace in 2004 (Figure 4).3 However, in the current cycle, there has been a clear divergence in the two time series. Nonfarm employment in the PS counties rose nearly 5 percent between its trough in August 2009 and June 2011, while employment in the other 53 counties rose only 1.6 percent. Furthermore, employment in the PS counties has surpassed its previous cyclical peak, unlike the situation in the other Pennsylvania counties or in the nation at large. These 14 PS counties are generally remote and account for only 10 percent of the commonwealth’s employment base and 11 percent of its population of 12.7 million people. Therefore, it may seem reasonable to assume that employment in the PS counties has not had a significant effect on job growth in the other counties. A quick glance at Figure 4 also seems to lend some visual credence to this supposition. However, we wish to formally test the hypothesis that employment in the PS counties has not had a statistically significant effect on employment in the other counties. We now turn to our econometric analysis of employment growth in Pennsylvania. PENNSYLVANIA EMPLOYMENT: AN ECONOMETRIC APPROACH The 53 counties that are not PS counties account for 90 percent of the commonwealth’s jobs. Therefore, it seems reasonable that employment in these other counties would be related to more than just employment in the PS counties. Specifically, employment in these other 53 counties should be positively correlated with the national labor market as Figure 2 suggests. We performed cointegration tests on U.S. nonfarm payrolls, employment in the PS counties and employment in the other counties for our monthly data sample, which spans January 1990 to June 2011. A Johansen test indicated the existence of one cointegrating vector among these three variables for the entire sample. However, the residuals of an error correction model (ECM) showed significant volatility in late 1999, which could indicate the presence of one or more series breaks in the time-series data. Therefore, we limited our sample to January 2000 through June 2011, and subdivided this sample into two periods (January 2000 until December 2005, and January 2006 until June 2011) because annual time-series data indicate that shale gas production began to ramp up in Pennsylvania around 2005-2006.4 For the January 2000 to December 2005 sub-period, a Johansen test was not able to reject the null hypothesis of no cointegration among the three variables at the 95 percent confidence level (Table 1). However, the variables appear to be cointegrated in the January 2006 to June 2011 time period (Table 2). Because there is evidence to suggest that U.S. nonfarm payrolls, employment in the PS counties and employment in the other Pennsylvania counties have been cointegrated recently (i.e., during the January 2006 to June 2011 subperiod), we estimated the following ECM for this sub-period: β log(other ) t = − 0.0001 0.271β log( ps ) t 0.467β log(us ) t 0.315errort −1 + + − + εt ( −1.063) ( 4.936) (5.033) ( −2.822) R2 = 0.81 DW = 2.08 where (ps)t = employment in primary shale counties in Pennsylvania (other)t = employment in other counties in Pennsylvania (us)t = U.S. nonfarm employment (error)t-1 = error from the first-stage regression Our model results suggest that, everything else equal, a one percentage point change in employment growth in the PS counties between January 2006 and June 2011, a period during which shale gas exploration and production was ramping up significantly, was associated with a 0.271 percentage point change in employment growth in the other 53 Pennsylvania counties. In other words, we can reject the null hypothesis that employment in the PS counties has not had a statistically significant effect on employment in the commonwealth’s other 53 counties. In addition, a one percentage point change in U.S. nonfarm payrolls growth was associated with a 0.467 percentage point change in employment growth in the other 53 counties. We then use these model results to make projections about employment in the Keystone State until the end of 2020. However, we first need to make some assumptions regarding growth in the explanatory variables over the next few years. Recognizing that recent growth rates of employment in the PS counties may not be indicative of future trends, we considered three scenarios for our forecast: an optimistic scenario, a pessimistic scenario and a midpoint scenario. In our optimistic scenario, we assume that employment in the PS counties will grow over the next eight years at the rapid 2.9 percent per annum rate experienced between August 2009, when employment in the PS counties troughed after the last downturn, and June 2011. In addition, we assume that growth in U.S. nonfarm payrolls reverts to the 1.6 percent annual pace witnessed during the 2003-2008 expansion. Using these assumptions and the above regression results, we were able to produce a forecast of employment in the commonwealth’s other 53 counties. We then added this forecast to our trend estimate of employment in the PS counties to generate a forecast for total payrolls in Pennsylvania. This forecast is shown as the optimistic scenario in Figure 6.. Proceedings of the 2012 Pennsylvania Economic Association Conference 49 Under this optimistic scenario, we project that nonfarm payrolls in the Keystone State would grow by 825,000 by the end of 2020 (Figure 8). This 15 percent increase from the current level represents a 1.5 percent per annum rise, nearly twice as fast as the 0.8 percent average annual growth rate that was achieved in Pennsylvania between 2003 and 2008. Employment in the PS counties would shoot up by more than 165,000 by the end of 2020, a remarkable 32 percent increase for counties that historically have been among the commonwealth’s less prosperous areas.5 The commonwealth’s other 53 counties would also experience strong job growth. Employment in these counties would increase by more than 650,000 positions by the end of 2020 due to two components. First, the general U.S. macroeconomic expansion, which we proxy by the 1.6 percent per annum increase in U.S. nonfarm payrolls, would lead to an increase in jobs in the other counties in excess of 250,000 positions. The second component reflects very strong growth in the PS counties, which induces an extraordinary 400,000 jobs in the other counties over the next eight years. As the example of Williamsport shows, strong growth in PS counties could generate demand for leisure and hospitality jobs in other counties that are located near the PS counties. In addition, employment in other supporting industries likely would be induced in other counties by continued strong growth in the PS counties. We refer interested readers to the appendix, where we discuss the recent employment experience in Williamsport in more detail. However, the trajectory of recent history may not be sustainable for our entire forecast period, because natural gas prices have receded to their lowest levels in a decade (Figure 7). Consequently, some producers have announced plans to freeze production. For example, Chesapeake Energy recently announced plans to reduce production and to halve the number of dry gas drilling rigs this year.6 The unusually warm winter this year has clearly played a role in reducing the price for natural gas, as individuals have needed less gas to heat their households and businesses. That said, the downward trend in natural gas prices has been in place for much longer than what unseasonably warm winter weather this year would explain. The risk is that if gas prices remain depressed due to factors beyond just warmer weather, our optimistic scenario could prove to be, well, way too optimistic. Therefore, we consider a pessimistic scenario. Under this scenario, we assume that job growth in the PS counties will revert to its 2003-2008 trend, which coincided with the last cyclical employment upturn in Pennsylvania but which also largely preceded the significant increase in shale gas production in the commonwealth. During this period, employment growth in the PS counties equaled only 0.8 percent per annum. We also assume that U.S. nonfarm payrolls will continue to grow at the relatively sluggish 1.2 percent annualized pace that has prevailed since national employment bottomed in February 2010. Using these assumed employment growth rates along with our ECM estimates we produce the pessimistic scenario that is shown in Figure 9. Whereas a remarkable 825,000 jobs were created in Pennsylvania under our optimistic scenario, the pessimistic scenario leads to an increase of only 325,000 statewide positions by the end of 2020. The employment gain in the PS counties shrinks to less than 45,000 positions, almost onefourth of the jobs created under the optimistic scenario. The level of employment in the other counties increases by only 285,000 jobs. Not only does the sluggish nature of U.S. economic growth lead to slower job creation in the Keystone State, but the induced rise in employment in the other counties from economic growth in the PS counties also falls to a bit more than 100,000 jobs from the 400,000 positions that were created under the more optimistic scenario. As with most things in life, the “truth” probably lies somewhere between these two scenarios. Expecting that the recent breakneck pace of job creation in the PS counties will continue for the next eight years does not seem very credible, especially in light of the sharp decline in natural gas prices. On the other hand, it seems likely that U.S. employment growth will eventually strengthen from its slow pace of the past two years. Perhaps our pessimistic scenario is too gloomy. Therefore, we considered a midpoint scenario in which employment in the PS counties grows at a pace that is the average of its 2003-2008 and its 2009-2011 growth rates (i.e., we assumed that employment in these 14 counties will grow 1.9 percent per annum). We also assumed that U.S. nonfarm payrolls would grow at an annual average rate of 1.3 percent, the midpoint between our pessimistic and optimistic scenarios, over the next eight years. Under this scenario, which may be the most realistic of our three forecasts, total employment in Pennsylvania would rise by roughly 570,000 jobs by the end of 2020 (Figure 6). Approximately 100,000 of these positions would be created in the PS counties, and about 250,000 jobs (out of a total of 470,000) would be induced in the other 53 counties by strong growth in the PS counties. On a statewide basis, employment would rise by 1.1 percent per annum, which would be just shy of the 1.2 percent annual average growth rate that was achieved during the long expansion of the 1990s. CONCLUSION Economic growth in Pennsylvania appears to be positive at present, which is consistent with the expansion that has taken hold in the overall U.S. economy. Furthermore, growth in nonfarm payrolls in the Keystone State has kept pace over the past two years with employment growth in the overall national economy. There is plenty of anecdotal evidence to suggest that the shale gas industry is helping to boost Proceedings of the 2012 Pennsylvania Economic Association Conference 50 Under our optimistic scenario, in which employment in PS counties continues to grow at its robust pace of the past two years and U.S. nonfarm payroll growth reverts to its 20032008 trend, we estimate that overall employment in Pennsylvania will soar by about 825,000 jobs between now and the end of 2020, a remarkable 15 percent increase from its current level. Under more pessimistic assumptions, however, employment in Pennsylvania would increase by only 325,000 jobs over the next eight years, a disappointing 0.6 percent per annum increase. Our midpoint scenario leads to an increase in statewide employment of roughly 570,000 jobs over the next eight years, which would be roughly equivalent to the total job creation enjoyed during the 1990s. We are sometimes asked if shale gas exploration, drilling and production in Pennsylvania is “worth it.” In other words, do the benefits of shale gas outweigh the costs? Our midpoint scenario projects that there would be 250,000 more statewide jobs at the end of 2020 than there would be under our pessimistic scenario, which assumes slow economic growth in the PS counties as well as in the overall national economy. Of this 250,000 increase, about 60,000 more jobs would be created directly in the PS counties and roughly 140,000 more would be induced in the other counties by the stronger growth in the PS counties. In other words, the stronger growth rate we assumed for the PS counties under the midpoint scenario relative to the pessimistic scenario would lead to 200,000 more jobs across the commonwealth, which represents a 3 percent boost to employment, at the end of 2020. The 3 percent extra boost to employment that we referenced above, even if accrued over an eight-year period, is clearly a benefit. That said, shale gas exploration, drilling and production could potentially entail significant environmental costs. In addition, there may be some social negative externalities (e.g., there have been anecdotes of families becoming priced out of the rental market in shale regions). As economists, we believe that we have the expertise to address economic issues, and we hope that the analysis in this paper will help to enlighten the debate about the desirability of shale gas exploration, drilling and production in Pennsylvania. Because we do not have expertise in environmental and other social sciences, however, we need to leave an analysis of these issues to others. Thus, taking economic, environmental and social issues into consideration, we leave it up to our readers to decide whether the benefits of shale gas outweigh the costs. FIGURES AND TABLES Figure 1 Pennsylvania Coincident Index Three-Month Percent Change 4.0% 4.0% Pennsylvania: Mar @ 1.2% 3.0% 3.0% 2.0% 2.0% 1.0% 1.0% 0.0% 0.0% -1.0% -1.0% -2.0% -2.0% -3.0% -3.0% 80 82 84 88 90 86 92 94 96 98 02 04 00 06 08 10 12 Figure 2 Nonfarm Payrolls: PA vs. US Index, 100=Feb. 2010, From The Establishment Survey 108 108 United States: Apr @ 102.9 Pennsylvania: Mar @ 102.8 106 106 104 104 102 102 100 100 98 98 03 04 05 06 07 08 09 10 11 12 Figure 3 PA Employment Growth By Sector Percent, Since Feb. 2010 Trough, As of Dec. 2011 35% Percent of Job Growth Accounted For By Sector employment in Pennsylvania, and “hard” data confirm that job growth in the commonwealth’s 14 primary shale (PS) counties has indeed been strong over the past two years. Moreover, our statistical analysis shows that, everything else equal, growth in employment in the commonwealth’s 53 other counties has been associated directly with job growth in the PS counties. Significant drilling activity and gas production in the PS counties has probably increased demand for services such as leisure and hospitality, and engineering and surveying in adjacent counties. 45-Degree Line Ed/Health Svcs. 30% 25% Trade/Trans/Ut 20% 15% Leis/Hosp Mfg. 10% Prof/Bus Svcs. Construction Nat. Resources & Mining Other Svcs. 5% Info. Svcs. 0% 0% 5% 10% 15% 20% 25% 30% 35% Size of Sector - Percent of Total Payrolls Proceedings of the 2012 Pennsylvania Economic Association Conference 51 Figure 4 Table 2: Cointegration Tests (January 2006 to June 2011) Employment Growth in PA Index, 100=July 2003, SA, Quarterly Census of Employ. & Wages 106 106 Cointegrating Vectors Primary Shale Counties: Jun @ 105.0 Other Counties: Jun @ 100.2 104 Eigenvalue Trace Statistic Prob None 0.240 31.976 0.028 ≤1 0.187 13.890 0.086 ≤2 0.003 0.225 0.635 104 102 102 100 100 98 2003 98 2005 2007 2009 2011 Figure 5 Williamsport ERIE WARREN MCKEAN POTTER CRAWFORD FOREST ELK VENANGO CLINTON CLARION LYCOMING JEFFERSON BEAVER MIFFLIN INDIANA CAMBRIA SNYDER SOMERSET CARBON NORTH_ UMBERLAND H RT NO SCHUYLKILL DAUPHIN LEBANON CUMBERLAND FAYETTE MONROE BEDFORD BERKS M O N TG LANCASTER FULTON FRANKLIN CHESTER ADAMS N TO MP HA LEHIGH PERRY HUNTINGDON WASHINGTON GREENE PIKE LUZERNE JUNIATA BLAIR ALLEGHENY WESTMORELAND LACKAWANNA Montour UNION CENTRE ARMSTRONG WYOMING COLUMBIA CLEARFIELD BUTLER SUSQUEHANNA WAYNE SULLIVAN CAMERON MERCER LAWRENCE Pittsburgh BRADFORD TIOGA YORK BUCKS O M ER Y DELA- PH WARE IA PH EL AD IL Primary Shale Counties Other Counties Table 1: Cointegration Tests (January 2000 to December 2005) Cointegrating Vectors Eigenvalue Trace Statistic Prob None 0.232 28.157 0.076 ≤1 0.097 9.155 0.351 ≤2 0.025 1.799 0.180 Proceedings of the 2012 Pennsylvania Economic Association Conference 52 Figure 6 Figure 9 Pennsylvania Employment Projections Job Growth in Pessimistic Scenario Millions of Jobs, Seasonally Adjusted 6.3 6.2 Optimistic Scenario: 2020 @ 6.24 Million Midpoint Scenario: 2020 @ 5.98 Million 6.2 6.1 Pessimistic Scenario: 2020 @ 5.74 Million Total Nonfarm: 2011 @ 5.41 Million 6.1 6.0 6.0 5.9 5.9 5.8 5.8 5.7 5.7 5.6 5.6 5.5 5.5 5.4 5.4 5.3 5.3 5.2 2003 2008 2010 2013 2015 2018 Induced Statewide Change from PS Counties: 2020 @ 106.6K Primary Shale Counties: 2020 @ 42.3K 800 600 400 400 200 200 0 2012 0 2014 2016 2018 2020 Figure 10 Williamsport Employment Growth By Sector Natural Gas Henry Hub Spot, Dollars per MMBTU Percent, Since Feb. 2010 Trough, As of Dec. 2011 $16 $14 $14 $12 $12 $10 $10 $8 $8 $6 $6 $4 40% Percent of Job Growth Accounted For By Sector $16 $4 $2 $2 2007 2008 2009 2010 2011 Leisure & Hospitality 30% 20% Manufacturing 10% Edu. & Health Services Government Trans. & Utilities 0% $0 2006 Nat. Resources, Mining & Construction 0% Natural Gas: May @ $2.28 $0 2005 800 600 2020 Figure 7 1,000 Trend Growth in Other Counties: 2020 @ 177.9K 5.2 2005 Thousands of Jobs Created (Cumulative), Seasonally Adjusted 1,000 6.3 10% 20% 45-Degree Line 30% 40% Size of Sector - Percent of Total Payrolls 2012 Figure 8 ENDNOTES Job Growth in Optimistic Scenario 1,000 Thousands of Jobs Created (Cumulative), Seasonally Adjusted 1,000 Trend Growth in Other Counties: 2020 @ 258.5K Induced Statewide Change from PS Counties: 2020 @ 398.6K 800 Primary Shale Counties: 2020 @ 167.8K 800 600 600 400 400 200 200 * We would like to thank Tom Murphy from Penn University and Matthew Conlan and David Tameron Wells Fargo Securities for helpful discussions about gas production in Pennsylvania. We also thank Azhar from Wells Fargo Securities for econometric support. State from shale Iqbal 1 0 2012 0 2014 2016 2018 2020 Sectors that lie above the 45-degree line in Figure 3 have contributed more to employment growth than their weight in the overall workforce. For expositional simplicity, we have omitted sectors in which employment has contracted from Figure 3. Over the past two years, employment in the financial services industry, which accounts for 5 percent of Pennsylvania’s nonfarm payrolls, has contracted nearly 2 percent, while public sector employment (13 percent of payrolls) has plunged nearly 18 percent. 2 Considine, Watson and Blumsack (2010) estimate that the industry would create 111,000 jobs in 2011. However, Herzenberg (2011), and Timothy Kelsey et al (2011) find the employment-generating effects to be significantly smaller. Proceedings of the 2012 Pennsylvania Economic Association Conference 53 3 All county-level employment data considered in this report henceforward is obtained from The U.S. Department of Labor’s Quarterly Census of Employment and Wages. 4 According to data obtained from the Pennsylvania Department of Environmental Protection. 5 The number of jobs in these 14 counties would swell to nearly 700,000 at the end of 2020 from about 530,000 in June 2011. 6 See “CHK: Positive – More Shifting Away from Dry Gas”, Wells Fargo Securities, January 23, 2012. 7 As was the case with Figure 3, we have omitted sectors in which employment has contracted since February 2010 in Figure 10. In the case of Williamsport, the only sector in which jobs have declined since February 2010 is the “other services” sector, where employment is down only 0.2 percent. APPENDIX Williamsport is the county seat of Lycoming County, which is one of our PS counties. Since the statewide trough in employment in February 2010, payrolls in the Williamsport Metropolitan Statistical Area (MSA) have risen 2.9 percent, which is above the 2.4 percent gain witnessed at the broader state level. The increase in jobs in the natural resources, mining and construction sector has been especially impressive, surging by more than 25 percent since February 2010. Despite accounting for only 5 percent of the employment base in Williamsport, the natural resources, mining and construction sector has been responsible for nearly 40 percent of the total job gains witnessed in the metropolitan area since February 2010 (Figure 10).7 The leisure and hospitality sector is also having an outsized effect on overall employment growth in Williamsport. This sector, which makes up only 8 percent of the MSA’s employment base, has accounted for 35 percent of the job growth in Williamsport since February 2010. Shale gas drilling in Lycoming County and in some of the nearby PS counties is likely contributing to additional demand in the city’s hotels and restaurants, translating into 13 percent job growth in the leisure and hospitality sector over the past two years. REFERENCES Considine, Timothy, Watson, Robert and Blumsack, Seth, “The Economic Impacts of the Pennsylvania Marcellus Shale Natural Gas Play: An Update,” (May 24, 2010) Herzenberg, Stephen “Drilling Deeper into Job Claims: The Actual Contribution of Marcellus Shale to Pennsylvania Job Growth,” (June 2011) Kelsey, Timothy, Shields, Martin, Ladlee, James, and Ward, Melissa, “Economic Impacts of Marcellus Shale in Pennsylvania: Employment and Income in 2009,” (August 2011) Proceedings of the 2012 Pennsylvania Economic Association Conference 54 MULTIGENERATIONAL DISCOUNTING: MERGING INTERGENERATIONAL EQUITY AND INDIVIDUAL TIME PREFERENCE William Bellinger Department of Economics Dickinson College Carlisle, PA 17013 ABSTRACT The goal of this paper is to present a model that combines as distinct elements the intergenerational and intertemporal dimensions of present value discounting. The main components of the model are a basic social welfare function defined across overlapping annual cohorts, and an individual decision making model based on common assumptions of individual time preference. The most unique aspect of the model is the resultant adjustment to a new long run equilibrium caused by the implementation of a long-lived public policy decision. INTRODUCTION The adjustment process mentioned above may benefit from more explanation. When a long lived policy with public net benefits begins in the same year (year 0) for all living cohorts, the length of time the average cohort experiences net benefits gradually rises throughout the first lifetime of net benefits as cohorts living in year 0 are replaced with future cohorts who will experience the policy’s net benefits over their entire lifetimes. This gradual lessening of immediacy produces discounting over one lifetime even under utilitarian social welfare and restrictive steady state assumptions. While neither the basic structure nor individual components of this model are new (Bellinger 1991; Kula 1981, 1988), the analysis of the adjustment process created by a new policy over the first subsequent lifetime has some unique implications for the relationship between private and public good investments and the related distinction between private and social rates of discount. In addition to this adjustment process, the implications of the model for intergenerational equity and social welfare will also be explored. Considerable additional development of the model is needed and future goals for this project will be discussed in the conclusion. COMPONENTS OF THE MULTIGENERATIONAL VALUE MODEL Before presenting the full model, its two main components should be briefly discussed. With regard to the time preference of individual decision makers, discussion generally begins with the common discounted utility model of Samuelson (1937), and the closely related present value discounting formula, which will be referred to as an exponential discounting model in the following discussion. While the discounted utility model has been subjected to several technical and behavioral criticisms (Frederick, et. al, pp. 355-365), the most familiar controversy involving the pattern of individual time preference comes from the behavioral economics literature, where many experiments have found that individual time preference follows a pattern that has come to be known as hyperbolic discounting (Frederick, et. al.). While several formulas expressive of this discounting pattern have been proposed, (Frederick, et. al, pp. 360n), the main characteristic of hyperbolic discounting is that the typical individual’s marginal rate of discount has been found to decline with a lengthening of the time frame of the analysis. This concept currently lacks an accepted theoretical foundation and is unlikely to apply to decisions related to financial or real capital investment. However, the model of time preference developed in this paper will have something to say about this concept. The impact of hyperbolic individual time preference will be explored later in this paper. The social welfare and equity dimensions of long term policy issues involve considering the well-being of future as well as currently living cohorts. In some senses this is not fully feasible, as is widely known. Generations not yet born cannot be active participants in present contractual or political decisions (Georgescu-Roegen, 1975, Beckerman, p. 199-200). Therefore the inclusion of future generations in a long run public goods decision necessarily involves a more abstract decision rule based on broader ethical considerations. In this paper we combine a utilitarian social welfare function with a net benefit decision rule for both current and future cohorts. THE MULTIGENERATIONAL VALUE MODEL The presentation of this model begins by considering its basic assumptions and decision framework. The multigenerational value (MV) model is based on overlapping annual cohorts who analyze issues in different time frames. Each annual cohort includes N individuals. All individuals will have equal tastes with the usual properties, equal incomes unless otherwise noted, and equal lifetimes of L years. This analysis further assumes that policy alternatives will be in force T+1 (t=0,...,T) Proceedings of the 2012 Pennsylvania Economic Association Conference 55 years and that G cohorts (g=(-L+1),...,G) will live within a project's lifetime. All individuals will have equal tastes with the usual properties, equal incomes unless otherwise noted, and equal lifetimes of L years. There is a natural distinction to be drawn between those involved in an initial decision in year 0 and those not yet born who will have at least an implicit decision to reverse or continue the original policy. Future cohorts (g=1,...,G) are annual cohorts born in years 1 to T. The initial decision-makers include all cohorts alive at the time (t=0) of the initial policy decision, with birth years from -L+1 to 0. For convenience this group will be referred to as generation 0, The initial decision group's population equals (L-t)N for any year prior to t=L, starting with LN persons in year zero and declining to zero in year t=L and beyond as older cohorts die off. Each cohort alive during the duration of a policy, whether part of the original or future set of cohorts, is assumed to make a decision on that policy. Current decision making cohorts in generation 0 may choose to implement or not implement the policy. Future generations decide whether to alter or continue the policy. Because each generation lives for many years its private decision will be based on the present value of net benefits over its lifetime discounted to its decision-making period. Because different generations make their decisions at different times, however, they will discount their net benefits to different "present" values. 1 All those involved in the initial decision will discount their net benefits to the date of implementation (t=0). I will assume that future cohorts (generations 1-T) make their decisions during their birth year, when they first receive net benefits. 2 Each policy produces net benefits Bt in year t. Each living generation receives Bgt net benefits in year t, where Σg Bgt = Bt. Generational share coefficients (αgt), which must sum to one, may be defined as follows. G G G Σ αgt = Σ Bgt/Bt = Σ Ngt.(Bgt/Ngt)/Bt = 1 g=1 g=1 g=1 (1) where Ngt = the population of cohort g in year t and Bgt/Ngt = the per capita share of benefits for cohort g. This share coefficient will vary with the type of investment and, under broader assumptions, with the generation's population. For example, a purely private investment by a member of any cohort in generation 0 implies zero benefits and/or a zero share for all future generations. Also, equal per-capita benefits imply that a cohort’s benefit share will be proportional to its population. Equally dividing Bt implies equal shares of 1/L for each living cohort, and therefore for all living generations other than 0. Generation 0's share (while living) equals (L-t)/L, the number of cohorts living in any year t between 0 and L. Assuming for simplicity that its cohorts are equally weighted, generation 0's decision equals the sum of the choices of L cohorts from year 0 to their death, judged in current time. 1 MV0 = U(Bt/L)/(1+r)0 + Σ U(Bt/L)/(1+r)t +...+ t=0 L-1 L-1 Σ U(Bt/L)/(1+r)t = Σ U[Bt.(L-t)/L]/(1+r)t t=0 t=0 (2) where the first addend is the net benefit stream for the oldest cohort (–L+1), the last addend equals the benefit stream for the cohort born in year 0, MV0 = the multigenerational value for all cohorts in generation 0, U(Bt/L) = each cohort's utility from its share (Bt/L) of net benefits, 3 r is a fixed private discount rate, L-1 is the final year of life for the youngest cohort in generation 0, and (Lt)/L is generation 0's benefit share (α0t) in each year prior to L. Similarly, a future cohort (g>0) values its proportional share of the benefit stream as g+L-1 MVg = Σ U(Bt/L)/(1+r)t-g t=g (3) Society maximizes social welfare over a set of feasible policies. The social welfare function (W) will be defined in weighted linear form. The weight equals the social value of each generation's per capita utility, and is specified as λg = (1/1+τ)g, where τ is society's discount rate for the per capita utility of future generations. 4 This function is assumed to be constructed by an impartial observer with perfect information. 5 This function takes the form G G W = Σ λg U(αgt.Bt) = Σ U(αgt.Bt)/(1+τ)g g=0 g=0 (4) where all symbols are defined above. For a utilitarian social welfare function in a steady state world with fixed tastes, incomes, and population, τ = 0, λg =1, and the social welfare function reduces to G Σ U(αgt.Bt) g=0 (5) the common equally weighted utilitarian social welfare function. The multigenerational value for year t (MVt) is then derived by adding the weighted benefits of all living Proceedings of the 2012 Pennsylvania Economic Association Conference 56 cohorts. t MVt= Σ U(αgt.Bt)/(1+r)t-g(1+τ)g g=max{t-L,0} (6) The total multigenerational value (MV) of a project may then be obtained by summing over years 0 to T. 6 T T G MV = Σ MVt = Σ Σ U(αgt.Bt)/(1+r)t-g(1+τ)g t=0 t=0 g=0 (7) Summing over T years makes the asymmetric minimum value for g in (7) redundant. If benefits are equally divided, as in the public goods case, (7) becomes L-1 MV= Σ U[Bt.(L-t)/L]/(1+r)t + t=0 T G Σ Σ U(Bt/L)/(1+r)tg (1+τ)g, t=1g=1 (8) where the first addend refers to the decision of the cohorts in generation zero and the second to cohorts 1-G. In interpreting the second addend note that a generation born in year t ≥ 0 will experience benefits from year t to t+L. Equation (8) represents the general form of the multigenerational value (MV) model for policy analysis. For the public goods case, utilitarianism implies the following multigenerational value model. Combining the social welfare function from equation (5) with the exponential individual time preference assumption produces the following model of social time preference: MVU Now let us evaluate the time preference pattern of this utilitarian model. Let us first consider years beyond one lifetime (t > L). Beyond year L all generations are single cohorts with equal share coefficients of 1/L. The social value for year t equals t MVt = Σ U(Bt/L)/(1+r)t-g= U(Bt/L)/(1+r)L +...+ g=t-L U(Bt/L)/(1+r)0. (10) Similarly for year t+1, t+1 MVt+1 = Σ U(Bt+1/L)/(1+r)t+1-g = g=t+1-L U(Bt+1/L)/(1+r)L +..+ U(Bt+1/L)/(1+r)0. (11) If Bt+1 equals Bt, MVt+1 minus MVt equals zero. This proves that if Bt is fixed, utilitarian MVt is constant after one lifetime. L-1 = Σ U[Bt.(L-t)/L]/(1+r)t + t=0 T G Σ Σ U(Bt/L)/(1+r)t-g, t=1 g=1 utilitarian optimality may be identified via an "overtaking" rule (von Weizsäcker, 1965), a "catching up" rule (Gale, 1967), or by the assumption of a rate of discount based on the probability of world extinction (Dasgupta and Heal, 1979, pp. 273-81), although the last approach implies that expected population declines over time. For a review, see Dasgupta and Heal (1979, Ch. 9). The interpretation of these maximums that is least problematic is to assume an arbitrary finite maximum for G, and a corresponding value for T of G+L, the final year of net benefits for the final cohort being considered. Under this set of values the social value of each year greater than L will be constant until the policy analysis ends in year G+L. (9) where MVU is utilitarian multigenerational value, the first addend refers to generation 0, and the second addend sums all future cohorts. Before analyzing and displaying the pattern of social time preference exhibited by this model, the meanings of the maximum values of G and T should be discussed briefly. If the policy under consideration is infinite, and annual net benefits are positive, the utilitarian form will produce an infinite net present value. However, alternative criteria developed under utilitarianism may also be applied to MV discounting in this context. In an infinite time horizon, For years within one lifetime of the original policy decision (t < L), utilitarian discounting takes place at a slower rate than present value given equal interest rates. The social value for any year t < L is shown below: t MVt<L = Σ U(αgt.Bt)/(1+r)t-g = U[Bt.(L-t)/L]/(1+r)t g=0 t (12) + Σ U(Bt/L)/(1+r)t-g, g=1 where the right hand portion of (12) divides the decisions of generation 0 from those of other living generations. Since future generations discount to their birth year (or implicit decision year), future generations place a higher Proceedings of the 2012 Pennsylvania Economic Association Conference 57 present value on their net benefits for a given year than does generation zero, implying that MVt > PVt for all t > 0. Combining these results, utilitarian principles and restrictive steady state assumptions imply that the social value of each unit of net benefits will converge after L-1 years to a fixed value < 1. DISCOUNTING PATTERNS FOR PUBLIC GOODS IN THE MV MODEL Social rates of discount become both most relevant and most controversial in the context of long-lived public or quasi-public policy issues. Long run environmental decisions provide the context in which this issue has been most actively discussed during recent decades (Lind, et.al; Portney and Weyant). In this model the decision context involves multiple dimensions related to the overlapping cohorts, public net benefits, and steady state assumptions for income, tastes, and population. Into this constantly replicating and stable age and income distribution we introduce a policy that begins in year 0 for all living cohorts (for an 80 year lifespan, g = -79 through 0), and continues indefinitely with constant annual social net benefits. For every living generation the policy’s net present value will extend over its remaining lifespan. Each generation unborn in year zero will receive net benefits from the policy over their entire lives. The practical effect of this distinction between current and future cohorts is that over the first human lifetime of the policy’s existence the average length of the net benefit stream rises until it reaches the full lifetime L for all living cohorts after year L. As noted above, for the public goods case this lengthening of the average net benefit stream produces a quasi-hyperbolic valuation pattern over one lifetime followed by a constant mean present value produced by the overlapping cohorts associated with future generations. g+L U(B0g) = 1/L Σ U(B0t)/(1+r)t-g, t=g (13) where g = the birth year of future cohort g and U(B0g) is the present value of policy A0 to cohort g. Alternatively, cohort g may adopt its best alternative (Aîg) with PV g+L-1 U(Bîg) =1/L Σ [U(Bît)/(1+r)t-g] - 1/L U(Rît) t=g (14) where Bîg = the discounted net benefits of Aî to generation g, and 1/L Rît equals the cost to generation g of transferring resources from policy A0 to Aî. Generation g will reject policy A0 if g+L-1 Σ [U(Bît) - U(B0t)]/(1+r)t-g > U(Rît). t=g (15) The existence of reversal costs will have significant implications for the efficiency of PV discounting, as we will explore later. The next step in the exploration of the model is to explore the discounting patterns it produces under various assumptions. The basic discounting pattern produced by this model in the public goods context adds a superstructure associated with the cohorts in generation 0 to the long run pattern produced by the social welfare function. In the case of a steady state economy with equal income and population size over time, this model produces a quasi-hyperbolic discounting pattern over one lifetime (years 0 to L-1) followed by a constant present value beyond year L. This long run pattern can vary with assumptions about the social welfare function or the long run behavior of population, income, or cumulative risk, as will be discussed later in the paper. The decision process associated with this type of problem involves the effects of such policies on future as well as present generations, but also the degree to which there is a potential decision rule for future generations regarding such a policy. As can be seen in equations (8) and (9), those cohorts living at the initiation of the policy (generation 0) will choose to implement a policy if the net present value over their remaining lifetimes is positive. The net welfare gains for future generations depend on the net benefits of the original policy, those of the next best alternative, and any transactions costs associated with a policy modification. The Effect of the Length of Life Retaining the initial policy A0 produces the following present value for any future cohort: The aggregate MV discounting pattern produced in the bottom row is of interest for two related reasons. First, The first step in this section involves displaying the full discounting pattern for a very short lifespan. This step helps to clarify the pattern of net present value in the MV model, and to identify the sources of the distinction in present values between a common exponential PV model and the MV model. Table I displays the pattern of present value with a 5 year lifespan for each cohort, an infinitely lived policy whose net benefits begin in year 0, and an exponential individual discounting pattern with a 5% discount rate. Proceedings of the 2012 Pennsylvania Economic Association Conference 58 despite the assumption of an exponential discounting pattern for each individual cohort, the aggregate present value shows a time pattern more consistent with a hyperbolic form. A comparison of the exponential discounting pattern for cohort 0 and the MV values in Table I reveals this distinction. One hypothesis produced by this model is that individuals with a strong sense of altruism (i.e.: for other cohorts) may display a more hyperbolic individual discounting pattern than those who have more selfish tastes. Secondly, under the assumed static economy, all discounting ends after one lifetime in the MV model, a pattern that is partially consistent with the long run discounting pattern behavior of a utilitarian social welfare function in a steady state. There are several wellknown justifications for a degree of discounting across generations, including increasing mean incomes, increasing risk or uncertainty, and a cumulative long run risk of annihilation, all of which can be included in this model through the generational discount factor λ (see equation 9). notably, the distinctiveness of exponential and hyperbolic individual discounting streams is significantly reduced in this framework (See Table III below). The clear result displayed in Table III is that the overlapping cohort model substantially dampens the difference in present value created by exponential and hyperbolic discount factors. While the future value in year 100 for the exponential discount factor is only.0456 (about 1/20th) that of the hyperbolic formula, the corresponding ratio in the overlapping cohorts model is .63. However, under either formula hyperbolic individual time preference lessens the degradation of present value over time. The distinction between private and social discounting in the public goods case offers a potentially interesting hypothesis regarding the causes of the hyperbolic pattern. If an individual is perfectly altruistic with no distinction between her welfare and that of others, private time preference may be shown to equal social time preference. In this case the MV model implies a positive relation between a person’s altruism and the hyperbolic pattern of her time preference. We next consider the sensitivity of the model to the assumptions of lifespan. Under the restrictive assumptions outlined above, differing assumptions regarding the average lifespan have two effects on the present value pattern. First, longer lifespans mean that the substitution of future for current cohorts takes place more gradually, which implies (rather obviously) that long run social equilibrium will reappear more slowly. Second, and far more importantly, the long run equilibrium present value, one it is reached, will be inversely related to the length of life. This finding implies that an increasing average lifespan over the long run creates another source for long run social discounting. MULTIGENERATIONAL VALUE AND INTERGENERATIONAL EQUITY ISSUES Two of several unanswered questions in this draft relate to the relationship between this model and the basic social welfare and individual time preference models from which it is derived. These are (1) As L approaches infinity, does MV approach PV, and (2) As L approaches 0, does MV approach one, a constant social present value. Both are likely results which would help to place this model firmly in its broader context. Exponential Versus Hyperbolic Individual Discounting As noted above, one area of controversy related to discounting is the accuracy of the traditional assumption of exponential individual time preference. Considerable behavioral research has found that individuals tend to value the distant future more highly than an exponential present value model. While several formulas are capable of producing this pattern, the relatively basic formula used below is PVh = 1/(1+rt), where r is the discount rate and t is time. The MV model offers some surprising implications for the role of hyperbolic individual time preference. Most The equity implications of this model are somewhat narrow due to its reliance on the familiar utilitarian social welfare construct. However, the inclusion of individual as well as intergenerational discount factors raises some potentially interesting observations related to equity. By including intra-generational and inter-generational dimensions this model suggests a useful equity criterion, which will be labeled "moral consistency". Moral consistency requires that the same equity assumption be applied to all distributional dimensions unless the difference can be explicitly justified. Recall that for any year t > L, t (16) MVt = Σ αgt.U(Bt)/(1+τ)g(1+r)t-g. g=t-L Note from (16) that if τ = r, MV is equal to the present value model PV. Unless one can justify discounting future generations at rate r, PV will be inequitable. For any other value of τ including 0, MV is morally consistent over time while PV is not. However, it is also well known that a degree of generational discounting may be consistent with utilitarian social welfare maximization in some contexts. If per capita income grows over the lifetime of the project, and if marginal utility of income declines, a degree of positive generational discounting may be morally consistent. Moreover, the MV model violates moral consistency in one dimension. Specifically, utilitarian MV does not weigh net benefits equally across contemporaneous generations in a given year. To take an extreme example, Proceedings of the 2012 Pennsylvania Economic Association Conference 59 assume that the marginal utility of net benefits is constant, assume τ=0, and give all of Bt to the oldest living generation (g=t-L) by setting its share coefficient (αt-L) equal to 1. This maximizes the exponent of (1+r) and therefore minimizes MVt. MVt = U(Bt)/(1+0)t-L(1+r)L. (17) Conversely, if the share (αt) of the newborn generation (g=t) equals 1, MVt is maximized. While diminishing marginal utility may avert such a corner solution, contemporaneous generations are not valued equally under utilitarian MV. This result is peripherally related to the finding of Dasgupta, et. al., p. 64) that an individual’s decision regarding a long term policy is not consistent over time. However, (16) and (17) imply that if τ = r, as in the long run present value case, the distribution among living generations does not affect MVt. This result verifies that PV and utilitarian welfare are consistent in a crosssectional context both within and across living generations, despite their inter-temporal inconsistency. It also implies that MV will not be morally consistent for decisions involving the distribution of benefits across generations in a single time period. Because this overlapping cohorts approach retains a degree of discounting, critics content that it also retains some inequity (Price 1987, p. 500; Thompson 1988, pp. 171-72). While I have established that this complaint has validity for cross-sectional comparisons, I defend MV's equity over time. First, analyzing every decision in its own time frame eliminates one source of bias against future generations. Moreover, given a constant population and average income, the MV model is consistent with the equal weighing of each individual, and therefore with utilitarian equity. Furthermore, the MV model allows more explicit consideration of the efficiency costs of zero discounting. Early defenders of discounting often argued that zero discounting sacrifices efficiency by ignoring foregone interest (Baumol, 1968; Mikesell, 1977, pp. 18-21). Within the traditional formulation, this debate has been confused because single generation analyses (Marglin, 1963; Baumol, 1968) consider only r, while most multiple generation models (Dasgupta, 1974; Brown, 1987) consider only τ. Since both groups claim to be analyzing the discount rate, confusion has been unavoidable. The debate can be repeated here more explicitly. MV produces zero discounting if both the interest rate (r) and the generational discounting parameter (τ) equal zero. Setting only τ equal to zero provides equal generational weights, and produces the utilitarian MV discounting pattern. However, setting the intra-generational discount rate (r) to zero has no effect on the weights of different generations. It merely forces the analyst to ignore an important cost. Those who insist on equal outcomes, as opposed to fair process, are welcome to disagree. In conclusion, while the MV model is more equitable than PV in the inter-temporal dimension, discounting is not inherently unfair to future generations, and multigenerational discounting is not morally consistent in all contexts. Therefore, Kula's claim (1988) of moral superiority for multigenerational discounting is subdued considerably in this modification of his model. MULTIGENERATIONAL VALUE AND ALLOCATIVE EFFICIENCY In this section I explore the conditions under which present value may be consistent with the maximization of social welfare, thereby defining more carefully those cases for which PV is suboptimal and MV discounting may be appropriate. First, if a given policy is at least equally preferred to any alternative by all generations, then any choice of non-negative generational weights (λg > 0) will lead to a socially optimum choice. This policy criterion will be labeled multigenerational Pareto optimality (MPO). 7 Multigenerational Pareto optimality is extremely stringent, since an MPO policy has no preferred alternative for any generation. Yet some potentially common cases exist which may meet the MPO criterion. One such case is purely private investment. For example, if all net benefits go to an individual in generation zero, its share coefficient equals 1 and the social welfare function reduces to L-1 W = λ0 U(B0) = λ0 Σ U(Bt)/(1+r)t. t=0 (18) A private investment that provides positive net benefits to one individual in generation 0, all else equal, is a net gain to generation 0, and thus a net welfare gain to society if generation zero's social weight (λ0) is positive. A comparison of equation (18) with the public goods case developed above indicates that the MV model implies different social discount rates for inter-generationally private and public goods. 8 Specifically, for private goods the social (MV) and private rates of discount are identical, while only PV generational weights (τ=r) produce this result for public goods. Also, a policy decision by generation 0 may be MPO despite intergenerational externalities if the initial decision may be costlessly reversed. The existence of reversal costs (Rît) is a crucial factor in determining the efficiency of PV. 9 This argument involves revisiting the decision process for future generations in equations (13) through (15) above. Equations (13) through (15) imply that if policy A0 is perfectly reversible (R=0), future generations Proceedings of the 2012 Pennsylvania Economic Association Conference 60 will shift from policy A0 to policy Aî whenever the net benefits of the change are greater than 0. A selfish (PV) decision by generation zero will not restrict the decisions of future generations, and therefore will be MPO. Another case which may approximate costless policy modification involves liquid private bequests, which the recipient may invest as she sees fit. Therefore, the use of present value by generation zero may be MPO given costless reversibility. On the other hand, these equations imply that with positive reversal costs future generations may accept a policy which would be suboptimal for them in the absence of the original policy decision, thus violating the MPO criterion. More generally, when an initial policy is MPO, either through the provision of optimal benefits to each generation or through zero reversal costs, then many common equity assumptions are consistent with the maximization of social welfare, including the equity assumption implicit in PV. When a policy is not MPO, generational equity becomes crucial, and MV may be preferable. CONCLUSION The multigenerational value (MV) approach offered in this paper reconsiders the theoretical basis and applicability of social discounting in the context of overlapping cohorts. For public goods in a steady state world, the utilitarianbased MV model proposing that positive discounting should be limited to one human lifetime. Also, a few principles are suggested for applying MV. First, MV has different implications for analyzing public and private goods. Analyzing any type of good requires the proper application of MV. For private goods, PV is the appropriate form of MV. Secondly, multigenerational analysis assumes that different generations make decisions in different time frames. Therefore, each element of any decision sequence should be analyzed in its own time frame, and only over its effective length. Some new hypotheses arise from this analysis of the model. These include the possibility that individual time preference will be hyperbolic for highly altruistic individuals and the finding of a negative relation between lifespan and long run social present value. Finally, the explicit distinction between discount rates for private and long-lived public goods is also relatively unique to this approach. equity. Rather, it is a generalization of both aspects of inter-temporal welfare. Multigenerational discounting, and particularly the MV approach, is both a blessing and a curse in that it forces (or allows) analysts to face complex conceptual and empirical issues which traditional analysis sweeps under the rug. The applicability of MV to grey areas such as quasi-public goods and relatively brief time frames depends on our ability to address difficult empirical and conceptual issues. At this point it would be premature to prejudge our ability to do so. At a minimum, however, MV has many potential policy applications involving longlived public goods such as nuclear waste disposal, infrastructure, and wilderness development. Intergenerational equity is an important factor in weighing the net benefits of such policies, and the recommendations of the multigenerational value model may be quite different from those of present value. This paper represents a first step in more fully considering the implications of the much older paper that originally produced this model. The explicit presentation of valuation patterns under different lifespan and individual time preference assumptions is a useful addition. However, several important aspects of the model cannot be easily explored in discrete time, suggesting that a dynamic equilibrium analysis of these concepts is warranted. I suspect, but cannot yet prove, that this model with overlapping cohorts contains long run equilibrium properties that may solve problems arising from other types of time-based social welfare functions. For example, the policy choice related to long run equilibrium is likely to prove consistent with a median cohort voting model, lessening the problem of inconsistent individual decisions over time. This paper already demonstrates the much less significant role for alternative individual time preference patterns in the public goods context. The constant value under long run steady state equilibrium beyond one lifetime, while not equal to 1, can provide an explicit definition of the numeraire value for social welfare that varies with lifespan as well as discounting assumptions. . Reintroducing and analyzing the generational discount factor λ will permit the integration of the role of expected population, changes in per capita income, uncertainty, and other determinants of the social welfare discount factor into the model, producing a more or less complete model of individual and social time preference. Such a model could prove beneficial in bringing an increased degree of consistency to the analysis of time preference in a multitude of contexts. More generally, multigenerational value is neither a rejection of present value nor a political compromise between discounting and utilitarian intergenerational Proceedings of the 2012 Pennsylvania Economic Association Conference 61 ENDNOTES i. This approach differs subtly from Kula's. He (1981) assumes different time frames because of different birthdates. I assume different time frames because of different decision periods. While insignificant in the public goods case, this distinction provides considerable guidance toward the analysis of sequential decisions and other applications later in the paper. ii. Alternative decision periods are possible. For example, each generation could make its policy decision at its age of majority. Generation 0 would then include all adult cohorts, and each "future" cohort would make its decision and begin receiving net benefits upon reaching adulthood. In the utilitarian case considered later, this assumption produces social discounting over one adult lifetime, rather than an entire lifespan. This distinction doesn't affect the model's general characteristics. iii. The assumption of equal tastes removes the need to identify the utility function by generation. iv. This weighted linear social welfare function has the advantage of being directly comparable with the present value discounting formula under discrete time. For the purposes of this paper, this advantage outweighs any lack of generality. v. A version of the impartial observer was introduced by Adam Smith (1976, pp. 162, 352). A more idealized version of the impartial observer is described by Rawls (1971, pp. 186-87). While this observer is generally associated with utilitarianism, it also can be applied to the Rawlsian case (Dasgupta and Heal, 1979, pp. 271-72). vi. Alternatively one could determine the value of net benefits for each generation and then sum across the generations. vii. More formally, any policy alternative Ai ε A , where A is the set of all feasible policy alternatives, will be MPO if its net benefits Big > Bjg for all Aj ε A and all generations g. viii. This result is superficially similar to that of Marglin, but differs in some important respects. First, Marglin bases his analysis only on the preferences of currently living generations (1963, pp. 97) and, secondly, that his model is based largely on cross sectional and intergenerational altruism by generation 0. For effective criticisms of Marglin's approach, see Tullock (1964) and Baumol (1968). ix. Reversal costs in this context refer to the dollar costs of resource transfer and to any decision making costs associated with the political process for public goods. Irreversibility may be defined in at least the following three ways. Policy Ai will be irreversible if (1) Rig = 4 for all i and g, if (2) Rîg > Bîg - B0g for all g, or if (3) Rig is "large". The first definition is very strong, the second defines all multi-generationally optimal policies as irreversible, and the third is somewhat vague. The weak definition (2) is most useful in this context. Proceedings of the 2012 Pennsylvania Economic Association Conference 62 APPENDIX: PRESENT VALUE PATTERNS Table 1: MV for a 5 Year Lifespan Cohort birth year Yr. 0 -4 1 -3 1 0.95238 -2 1 0.95238 0.90703 -1 1 0.95238 0.90703 0.863838 0 1 0.95238 0.90703 0.863838 0.822702 1 0.95238 0.907029 0.863838 0.822702 1 0.952381 0.907029 0.863838 0.822702 1 0.952381 0.907029 0.863838 0.822702 1 0.952381 0.907029 0.863838 1 0.952381 0.907029 1 0.952381 Year 1 1 Year 2 2 Year 3 3 Year 4 4 5 Year 5 6 Year 6 7 Year 7 1 Σ PVit Mean PVit =MVt 5 4.80952 4.67347 4.587086 4.545951 4.545951 4.545951 4.545951 1 0.96191 0.93469 0.917417 0.90919 0.90919 0.90919 0.90919 Yr 80 Yr 100 Lifespan yr 0 Table 2: MV Present Values for Different Lifespans yr 1 yr 25 Yr 50 Yr 65 65 years 1 0.967766 0.413939 0.316369 0.31017 0.31017 0.31017 80 years 1 0.964881 0.391695 0.2734 0.259878 0.257456 0.257456 100 years 1 0.962381 0.372417 0.236161 0.216292 0.21 0.198479 Table III: PV and MV with Exponential and Hyperbolic Discounting Patterns Discounting Pattern yr 0 yr 1 yr 25 Yr 50 Yr 65 Yr 80 Yr 100 PV 5% PV 5% Hyperbolic 1 0.952381 0.295303 0.087204 0.041946 0.020177 0.0076 0.952381 0.444444 0.285714 0.235294 0.2 0.16667 MV, L = 80 MV, L = 80 Hyperbolic 1 0.964881 0.391695 0.2734 0.259878 0.257456 0.257456 1 0.964881 0.511802 0.424846 0.410676 0.407409 0.407409 Proceedings of the 2012 Pennsylvania Economic Association Conference 63 REFERENCES Baumol, William J. 1968 "On the Social Rate of Discount" American Economic Review 58 (September): 788-802. Beckerman, Wilfred, Economics as Applied Ethics, (London: Palgrave-MacMillan, 2011). Bellinger, 1991 “Multigenerational value: modifying the modified discounting method,” Project Appraisal, V.6 No 2, (June), pp. 101-108. Brown, S.P.A. 1987 "The fairness of discounting: A majority rule approach" Public Choice 55 (October): 215-26. Dasgupta, P.S. 1974 "On Optimum Population Size" in A. Mitra, ed. Economic Theory and Planning (London: Oxford U. Press). _______________, Karl-Göran Maler, and Scott Barrett, “Intergenerational Equity, Social Discount Rates, and Global Warming,” in Portney and Weyant, eds., Discounting and Intergenerational Equity, (Washington D.C.: Resources for the Future, 1999), pp. 51-78. ______________, and G. M. Heal 1979 Economic Theory and Exhaustible Resources (Cambridge, UK: Cambridge U. Press). Frederick, Shane, George Loewenstein, and Ted O’Donoghue, “Time Discounting and Time Preference: A Critical Review”, Journal of Economic Literature, V. XL (June 2002), pp. 351-401. Gale, D. 1967 "Optimal Development in a Multi-Sector Economy" Review of Economic Studies 34: 1-18. Georgescu-Roegen, Nicholas, “Energy and Economic Myths,” Southern Economic Journal, V. 41, no. 3 (January 1975), pp. 347–381. Kula, Erhun 1981 "Future generations and discounting rules in public sector investment appraisal" Environment and Planning A 13: 899-910. ___________ 1988 "Future generations: the modified discounting method" Project Appraisal 3 (June): 85-88. Lind, et. al., Discounting for Time and Risk in Energy Policy, (Washington, D.C.: Resources for the Future, 1982. Lind, Robert C. 1989 "Reassessing the Government's Discount Rate Policy In Light of New Theory and Data in a World Economy with a High Degree of Capital Mobility" (Ithaca, N.Y.: mimeo). Marglin, Steven A. "The Social Rate of Discount and the Optimal Rate of Investment" Quarterly Journal of Economics v. 77 (Feb. 1963), pp. 95-112. Mikesell, Raymond 1977 The Rate of Discount for Evaluating Public Projects (Washington, D.C.: American Enterprise Institute). Portney, Paul R., and John Weyant, eds., Discounting and Intergenerational Equity, (Washington D.C.: Resources for the Future, 1999). Price, Colin, 1987 "The Developing Framework for the Economic Evaluation of Forestry in the United Kingdom--A Comment" Journal of Agricultural Economics 24: 497-500. Rawls, John 1971 A Theory of Justice (Cambridge, MA: Harvard University Press). Smith, Adam 1976 The Theory of Moral Sentiments (Indianapolis: Liberty Classics). Proceedings of the 2012 Pennsylvania Economic Association Conference 64 Thomson, K.J. 1988 "Future generations: the modified discounting method--a reply" Project Appraisal 3 (September): 171-72. Tullock, Gordon 1964 "The Social Rate of Discount and the Optimal Rate of Investment: Comment," Quarterly Journal of Economics LXXVII (May), pp. 331-46. von Weizsäcker, Carl C. 1965 "Existence of Optimal Programs of Accumulation for an Infinite Time Horizon" Review of Economic Studies 32, pp. 85-104. Proceedings of the 2012 Pennsylvania Economic Association Conference 65 AID EFFECTIVENESS IN SUB-SAHARAN AFRICA Yaya Sissoko Department of Economics Indiana University of Pennsylvania Indiana, PA 15705 Niloufer Sohrabji Simmons College 300 The Fenway Boston, MA 02115 ABSTRACT There is considerable debate on the potential benefits and costs of aid. On the one hand, aid provides much needed resources to improve development efforts and alleviate poverty. On the other hand, aid can create dependencies. This paper analyzes the effectiveness of aid for countries in subSaharan Africa. Using a sample of eleven countries including Botswana, Cameroon, Cote d'Ivoire, Ethiopia, Ghana, Kenya, Nigeria, Senegal, South Africa, Tanzania and Togo we estimate a modified version of Barro’s (1991) growth model. Following the literature we use GMM estimation for panel data of our sample countries from 1975 to 2006. We find that aid is a positive and statistically significant determinant of growth for our sample. As expected however, it cannot be the sole component of a growth strategy as its effectiveness is complicated by other factors such as FDI flows and political regimes. Using our empirical results and case studies we analyze how aid has helped in some cases and not in others. obstacles to aid effectiveness, we study the impact of aid on growth in Botswana, Cameroon, Cote d'Ivoire, Ethiopia, Ghana, Kenya, Nigeria, Senegal, South Africa, Tanzania and Togo. Our sample includes large and small countries with varying levels of trade and foreign investment flows and different levels of political stability. This paper uses generalized method of moments (GMM) for panel data from 1975 to 2006 and combines these results with case study analysis to shed light on the importance of aid and other factors in determining growth. We find that aid has a positive (and statistically significant) impact on growth. However, large aid flows alone are not likely to be beneficial for countries that have low levels of FDI and/or have more autocratic regimes. The paper is organized as follows: the next section provides background on our sample countries. Section 3 describes the framework for estimating aid effectiveness which is followed by an analysis of the results. Section 5 is a comparative case study analysis of the impact of aid in our sample countries and the last section concludes. INTRODUCTION BACKGROUND The paper analyzes the effectiveness of aid in countries in sub-Saharan Africa. There is considerable debate on the beneficial and harmful effects of aid. On the one hand, Sachs et al. (2004) highlight the need for aid to reduce poverty. Aid can also support large scale investment in infrastructure for long-term growth and development benefits. However, aid funds can be wasted which leads others to argue that it does not promote growth (Easterly 2003; Collier, 2007; and Williamson, 2008). Moreover, in some cases not only is aid not helpful, but can create dependencies which actually hurt the country. The potential benefit to countries from aid is related to effective governance and strength of institutions (Radelet et al. 2005). Countries in sub-Saharan Africa (SSA) are among the poorest nations. As expected, they are also major recipients of aid. Of the approximately $ 129 billion in official development assistance (ODA) in 2010 from Development Assistance Committee (DAC) of OECD, $ 26.5 billion was given to SSA countries (OECD website). Tables 1 and 2 provide current macroeconomic information and aid data for our sample countries. Despite the debate on its effectiveness, aid flows continue. With challenges of high debt and low growth and development, countries in sub-Saharan Africa are major recipients of aid. Unfortunately, the region has witnessed failures in effective aid utilization. To understand the Overall, the most distressed economies in our sample are Côte d'Ivoire and Ethiopia. Côte d'Ivoire has a relatively low GDP per capita, high unemployment1 and the biggest debt burden (table 1). Moreover, it is the only country in our sample to experience negative growth in 2011 (table 1). Ethiopia is also a concern with a relatively low GDP per capita, high inflation and large debt burden (table 1). With a 40% unemployment rate and a 48.5% debt to GDP ratio Kenya also faces considerable challenges. The same is true Proceedings of the 2012 Pennsylvania Economic Association Conference 66 for Senegal with 48% unemployment rate and a 33.2% debt to GDP ratio. Botswana, Ghana and South Africa are the most successful in our sample. Ghana’s GDP per capita is one of the higher ones in our sample and with a growth of 13.5% (table 1) it is one of the fastest growing countries in the world. The most prosperous countries in our sample (based on GDP per capita) are Botswana and South Africa (table 1). Expectedly, Botswana has the least aid inflows (table 2). However, despite its high GDP per capita and status as a major emerging market,2 South Africa continues to have substantial aid inflows. Other major recipients are Kenya and Nigeria. These four countries are not part of the Heavily Indebted Poor Countries (HIPC) Initiative unlike the rest of the countries in the sample. Cameroon, Ethiopia, Ghana, Senegal, Tanzania and Togo have reached the Completion Point of the HIPC Initiative and Côte d'Ivoire is at the Decision Point.3 Despite their participation in the HIPC Initiative, Ethiopia and Tanzania are significant recipients of ODA from OECD DAC (table 2). For both countries, the IDA and the U.S. are major donors (table 2). Aid from OECD (and elsewhere) have been and remain high to countries in SSA. As noted above, some countries have experienced macroeconomic successes while others have continued to struggle. Both the successes and failures could be, but are not necessarily related to, aid. To analyze the link between aid and growth we discuss the framework to measure aid effectiveness. FRAMEWORK FOR ESTIMATING AID EFFECTIVENESS Barro (1991) provides a model for real GDP growth which includes factors such as the real GDP per capita for initial period, investment, education, civil wars and aid. Our model is a modified version of Barro’s (1991) where we also include money supply (based on Alvi, Mukherjee and Shukralla, 2008 and Armah and Nelson, 2008). We also, explore the importance of globalization by including both foreign direct investment and a measure of openness (the sum of exports and imports). The basic model to be estimated is as follows: grthit = βo + β1 GDPPC + β2 DI + β3 Openit + β4 FDIit + β5 MSit + β6 PIit + β7 Aidit + εit (1) where grth = rate of growth of real Gross Domestic Product (GDP) per capita, GDPPC = GDP per capita of initial year of each period, DI = domestic investment as a percentage of GDP, Open = exports and imports as a percentage of GDP, FDI = foreign direct investment as a percentage of GDP, MS = money supply as a percentage of GDP, PI = polity index (which measures extent of democracy), Aid = foreign aid as a percentage of Gross National Income (GNI). All variables are expected to have a positive impact on growth except aid which is ambiguous given the debate over aid effectiveness. In addition to the above, similar to Alvi, Mukherjee and Shukralla (2008) and Armah and Nelson (2008) we also examine the importance of interaction between the polity variable and aid and compare our results to theirs. The equation to be estimated is: grthit = βo + β1 GDPPC + β2 DI + β3 Openit + β4 FDIit + β5 MSit + β6 PIit + β7 Aidit + β8 Aidit ∗ PIit + εit (2) Given our focus on the role of globalization, specifically trade, we also examine the interaction between aid and trade as it impacts growth. This allows us to analyze if aid and trade are complementing each other in the pursuit of growth. The equation to be estimated is given as: grthit = βo + β1 GDPPC + β2 DI + β3 Openit + β4 FDIit + β5 MSit + β6 PIit + β7 Aidit + β8 Aidit ∗ PIit + β9 Aidit ∗ Openit + εit (3) Similar to the literature, this paper uses Generalized Method of Moments (GMM) estimation. As noted by Alvi, Mukherjee and Shukralla (2008), this technique “removes country specific effects by taking first differences and makes use of lagged values of the dependent variable and predetermined variables as instruments.” Aid is considered endogenous and its lagged value4 as well as lagged inflation5 are used as instruments. Our results based on GMM estimation are discussed in the following section. DATA AND RESULTS We use annual data from 1975 to 2006 for our eleven countries (Botswana, Cameroon, Côte d'Ivoire, Ethiopia, Ghana, Kenya, Nigeria, Senegal, South Africa, Tanzania and Togo). The series used are real and nominal GDP, domestic investment, exports, imports, FDI, money supply (M2), polity index, consumer price index (CPI) and aid. Growth and GDP per capita are based on real GDP (GDP per capita is expressed as a natural log). Domestic investment, sum of exports and imports, FDI and money supply are expressed as percentages of nominal GDP and aid is expressed as a percentage of nominal GNI. CPI is used to compute inflation which is an instrument for aid. The polity index ranges from 10 to 10 with the former being strongly autocratic and the latter being strongly democratic. All data except the polity index is from the World Bank Indicators database from the World Bank. The polity index series is from Integrated Proceedings of the 2012 Pennsylvania Economic Association Conference 67 Network for Societal Conflict Research database from the Center for Systemic Peace. We estimate three growth equations as shown in equations (1), (2) and (3). The results are displayed in table 3. The Hausman J-test is a test for over-identifying restrictions. Rejection of the null would indicate that the equation is misspecified. As can be seen from the results in table 3, the GMM estimated equations are correctly specified.6 The estimation results for equation 1 show that FDI, political stability and aid are statistically significant determinants of growth. When an interaction term is added for polity index and aid (equation 2), initial GDP, FDI, money supply and aid are statistically significant but the individual impact of the polity index is not statistically significant. When the interaction term for aid and trade is included (equation 3) trade, FDI, aid and the interactive variable for trade and aid are statistically significant determinants of growth. A comparison of the results suggests that these interactive terms do not add significant information to the aid-growth relation and thus, we keep our first specification. Based on the results, we can conclude is that aid is associated with higher growth for our sample. This differs from Alvi, Mukherjee and Shukralla (2008) but is similar to the conclusion reached by Armah and Nelson (2008) and Gomanee, Girma and Morrissey (2005). Both studies focus on SSA as we do, although Gomanee, Girma and Morrissey (2005) use a different estimation technique and Armah and Nelson (2008) have different specifications than ours. While the econometric analysis is supportive of increasing aid to SSA, one concern about the estimation is that the impact of aid is assumed to be uniform for all countries in our sample. To more fully explore the econometric results in the context of our countries we examine macroeconomic trends of the main determinants of growth. Through this we aim to shed light on effectiveness of aid for the specific countries in our sample. CASE STUDY ANALYSIS The above econometric work shows the importance of FDI flows, political regimes and aid in growth. Aid in turn is affected by the inflationary environment. We thus examine these macroeconomic trends to more fully explore our econometric results. Figures 1 through 11 map out GDP per capita, aid as a percentage of GNI, inflation and FDI as a percentage of GDP data for our sample countries. Information about political trends is presented in table 4. Botswana’s GDP per capita has been on a steady upward trajectory since 1975 (figure 1A) and is currently the most prosperous country in our sample based on GDP (PPP) per capita (table 1). Botswana was a major recipient of aid in the 1970s and 1980s but has tapered off as the country has become prosperous (figure 1B). Currently, Botswana receives the least aid from the OECD DAC in our sample (table 2). For most of the three decades, Botswana’s inflation rate has been in the double digits (figure 1C) but has been relatively stable which has probably helped the aid-growth relation. GDP has also been helped by FDI in the 1980s (figure 1D). While FDI declined in the 1990s, there has been some recovery in the 2000s (figure 1D). Botswana’s polity index has ranged from 6-8 over the sample period reflecting a stable democratic environment (table 4). This coupled with significant aid and a moderate level of FDI flows in the 1970s and 1980s, have contributed to Botswana’s overall success. It has led to dependence on aid today (figure 1B and table 2). Cameroon’s GDP per capita was increasing in the 1970s but saw a decline in the mid-1990s and has steadied at a lower level compared to the highs of the 1980s (figure 2A). This suggests that aid has not been as effective which can partially be explained by the volatile inflationary environment (figure 2C). GDP has been further hampered by unstable FDI flows (figure 2D). Moreover, the political environment is not supportive of higher growth as Cameroon has suffered from strongly to weakly autocratic regimes in the sample period (table 4). After the declining rates in the 1980s (which are likely related to rising gross national income) aid has once again increased and given an improved inflationary environment (currently its inflation rate of 3.4% in 2011 is the lowest among the countries in our sample (table 1) could be beneficial. But without an increase in FDI flows and no change in the political climate, rising aid may not be sufficient to help Cameroon’s growth prospects. Côte d’Ivoire’s GDP per capita is on a downward trajectory (figure 3A). It is the only country in our sample with a negative GDP growth rate (table 1). Aid had jumped considerably in the 1990s, but has reduced in the 2000s (figure 3B). The high inflation levels seen in the 1970s may have hindered the aid-growth relationship although that has stabilized considerably in the 2000s (figure 3C). With higher FDI flows since the 1990s (figure 3D) and a shift to a more democratic regime in 2000 (table 4), Côte d’Ivoire could have reversed the trend. Unfortunately, the move to a more democratic system was short-lived (table 4). Overall, Côte d’Ivoire’s growth forecast looks weak. Ethiopia’s GDP per capita has been steady but very low for the entire period (figure 4A). Aid has been substantial for most of the period (figure 4B) and the country is one of the major recipients of aid from OECD DAC (table 2). There is considerable volatility in inflation observed in Ethiopia (figure 4C) with high levels as well as several deflationary periods. At 28.8% it has the highest inflation rate of our sample countries (table 1). This may explain why despite rising aid, GDP remains low. The one bright spot for Ethiopia is that after years of practically non-existent levels, Proceedings of the 2012 Pennsylvania Economic Association Conference 68 FDI flows have picked up in the 2000s although they remain at very modest levels (figure 4D). A move to a more democratic regime in the early 1990s was also a step in the right direction. Unfortunately, the country has a very weak form of democratic government (table 4) which is not as beneficial for growth. there is some stability in the 2000s (9C and 9D). In addition, aid has been very modest for South Africa which suggests that it is not a prime determining factor of GDP. With improvements in FDI flows and inflation rates and a strong and stable democratic regime it is not surprising that South Africa has emerged as a member of the BRICS. Ghana’s GDP per capita is on an upward trajectory for the last three decades (figure 5A) and in terms of GDP per capita (PPP) is currently ranked only below South Africa in our sample (table 1). After modest levels in the 1970s and early 1980s, aid has been increasing for the country (figure 5B). Ghana’s inflation rate has been falling but still remained high and has been steady in the 2000s (figure 5C). FDI flows have increased in the 1990s and 2000s although they remain a small percentage of GDP (figure 5D). Since the 2000s, Ghana’s political regime has become more democratic (table 4). Improvements (modest as they may be) in all these factors lead to brighter prospects for Ghana’s future. In Tanzania’s case aid, inflation and FDI flows have improved since the 1990s (figures 10B, 10C and 10D). This may explain the improvement in GDP (10A). Despite this, Tanzania has not been as successful as some of the other countries with less aid. This may be explained by its political system which has remained autocratic for the entire sample period. Kenya’s GDP per capita has been somewhat volatile over the sample period (figure 6A). Aid has declined in the late 1990s and 2000s (figure 6B). The volatility in inflation may have hurt the aid-growth relation while the stability in the 2000s has helped Kenya (figure 6C). The shift to a more democratic regime in the 2000s (table 4) was also beneficial. One problem is the low level of FDI flows since the 1970s (figure 6D). Nigeria’s GDP per capita has been steady over the sample period (figure 7A). Nigeria had several years of low aid flows (7B) and that coupled with a high inflationary environment did not help GDP. In recent years, Nigeria has become a major recipient of aid (figure 7B) and was a top three recipient of OECD DAC aid in 2010 (table 2). This could be beneficial given that inflation rates have stabilized in the 2000s (figure 7C). FDI has seen some highs and lows but appears to have stabilized in the 2000s as well (figure 7D). After some instability, Nigeria is back to a more democratic regime although it is only moderately democratic (table 4). Overall, the 2000s have some modest improvements which could help Nigeria. Senegal’s GDP per capita has been steady for the sample period (figure 8A). The relatively high levels of aid were hindered by a volatile inflationary environment (figures 8B and 8C) which explains Senegal’s GDP trajectory. FDI flows were also volatile in earlier decades (figure 8D) further hurting GDP for the country. The 2000s may see some improvement because aid remains strong and both inflation and FDI has stabilized (figures 8B, 8C and 8D). Moreover, Senegal has become strongly democratic in the 2000s (table 4). South Africa has had a relatively high GDP per capita (figure 9A). After some instability in inflation rates and FDI flows, Togo’s GDP per capita shows a stagnating economy (figure 11A). This is despite relatively high aid in the 1970s and 1980s (figure 11B). There is some volatility in the inflationary environment including deflationary episodes (figure 11C) which may explain the lack of success of aid. Moreover, FDI flows have not always been reliable, although there is improvement in the late 1990s and early 2000s (figure 11D). The political system has contributed to Togo’s struggles with growth (table 4). CONCLUSION In this paper we examine the effectiveness of aid in promoting growth for countries in SSA. Using panel data in a GMM framework for our sample of eleven countries, we find that aid is beneficial to growth. However, there are complicating factors to this general conclusion. Our econometric analysis also shows that political regimes and FDI flows are statistically significant contributors to growth. The importance of regimes cannot be over-emphasized. Donors are more reluctant to aid autocratic regimes. But even if funding continues, aid is not beneficial if it lines the pockets of potentially corrupt autocratic leaders. Moreover, aid flows in these cases may be more “projectsupport” than “budget-support”. If aid is in the form of the latter, governments can invest where they perceive the greatest need based on a comprehensive overview of the requirements of the country. Fungible aid is more flexible and can be more beneficial. However, fear of more autocratic governments may lead to a shift to “project-support” aid which while helpful in some ways, may not be as beneficial to overall growth. Thus, an improved political climate may be as important as aid flows in promoting growth. Another benefit for countries is from FDI flows. There are many benefits associated with FDI flows. Poorer countries may be more reliant on flows from abroad than from domestic sources. Secondly, FDI flows may also come with technical expertise and “good business practices”. However, foreign investors, particularly foreign direct investors require Proceedings of the 2012 Pennsylvania Economic Association Conference 69 increased stability (both economic and political). Moreover, they need incentives. The kind of commitment needed to attract and maintain FDI flows requires political stability and good use of aid flows. In turn, this could help improve aid flows and lead to better use of aid for development projects to support foreign investment. countries. It also clarifies why some countries have had successes with promoting growth through aid. What we can thus conclude is that an over-reliance on aid as a driver of growth will not be beneficial to countries. It can ultimately only be one piece rather than the entire strategy to promote growth. If a country pursues the latter it does show why aid has lead to dependencies and has been harmful to TABLES Table 1: Macroeconomic Indicators (2011 estimates) Country GDPPC PPP Real GDP Growth Unemployment Rate Inflation Rate Public Debt Botswana $ 16,300 6.2% 7.5%1 7.8% 20.3% Cameroon $ 2,300 3.8% 30% 3.4% 16.2% 2 5.2% 65.8% Côte d'Ivoire $ 1,600 -5.8% NA Ethiopia $ 1,100 7.5% NA Ghana $ 3,100 13.5% 28.8% 42.3% 3 8.8% 38.7% 4 11% Kenya $ 1,700 5.3% 40% 11.0% 48.5% Nigeria $2,600 6.9% 21% 10.8% 17.6% 5 Senegal $ 1,900 4.0% 48% 3.4% 33.2% South Africa $ 11,000 3.4% 23.9% 5.0% 35.6% Tanzania $ 1,500 6.1% NA 11.1% 36.9% $ 900 3.8% NA 4.5% NA Togo Notes: NA indicates data is not available. 1 Unemployment rate for Botswana is a 2007 estimate. 2 There are no statistics on unemployment for Côte d'Ivoire, but it is estimated that it could be as high as 40-50%. 3 Ethiopia’s unemployment rate is a 2000 estimate. 4 Ghana’s unemployment rate is a 2008 estimate. 5 Senegal’s unemployment rate is a 2007 estimate. Source: CIA World Factbook. Proceedings of the 2012 Pennsylvania Economic Association Conference 70 Table 2: Official Development Assistance (ODA) from OECD DAC Countries (2010) Country ODA (millions) Major Donors Botswana $ 157 U.S., EU Institutions Cameroon $ 538 France, Germany Côte d'Ivoire $ 848 France, IDA Ethiopia $ 3,539 IDA, U.S. Ghana $1,694 IDA, U.S. Kenya $ 1,631 U.S., IDA Nigeria $ 2,069 IDA, U.S. Senegal $ 931 France, IDA South Africa $ 1,032 U.S. EU Institutions Tanzania $ 2,961 IDA, U.S. $ 421 France, Switzerland Togo Notes: ODA figures are for aid in 2010 from OECD DAC countries. IDA refers to International Development Association (World Bank Fund). GDP per capita (PPP) and real GDP growth data are 2011 estimates. Sources: OECD, World Bank. Table 3: GMM Results for Growth Variable GMM (1) GMM (2) * GMM (3) GDPPC 0.66 (0.23) 0.84 (0.03) 0.56 (0.22) DI 0.01 (0.88) 0.01 (0.83) 0.01 (0.59) Open 0.02 (0.21) 0.02 (0.22) 0.08* (0.03) FDI 0.42* (0.00) 0.38* (0.01) 0.36* (0.01) MS -0.08 (0.36) -0.07 (0.38) -0.06 (0.39) PSI 0.20* (0.00) 0.10 (0.29) 0.08 (0.54) Aid 0.15* (0.04) 0.19* (0.01) 0.64* (0.02) 0.02+ (0.16) 0.02 (0.20) Aid*PSI -0.01* (0.08) Aid*Open J test (p-value) 0.98 0.95 0.69 Notes: J-Test is test for over-identification of instruments. Rejection of the null hypothesis indicates that the instruments are not appropriate and the model is misspecified. *, ** and *** indicates that the variables are statistically significant at 5%, 10% and 15% respectively. + indicates that while the variable is not statistically significant at usual levels of significant, it is still an important determinant of growth. Proceedings of the 2012 Pennsylvania Economic Association Conference 71 Table 4: Polity Index (1975-2006) Country Max Min Trends Botswana 8 6 Botswana is one of the most consistently democratic regimes in our sample with an increasing polity score from 6 to 8. Cameroon -4 -8 In the early 1990s, Cameroon went from being relatively strongly autocratic (-8) to relatively weakly autocratic (-4) Côte d'Ivoire 4 -8 In 2000, Côte D'Ivoire shifted to a democratic government since 2003 has been classified as being in an interregnum period. Ethiopia 1 -8 Between 1991 and 1994, Ethiopia attempted to transition from an autocratic government to a more democratic government. However, since 1995 until the end of the sample, Ethiopia’s score has not exceeded 1 which is an extremely weak form of democracy. Ghana 8 -7 Ghana has had long history of autocratic governments. There was a brief period of democracy in the late 1970s but that quickly reversed to an autocratic government which remained in place until the 1990s. By the 2000s, Ghana’s government has become more democratic. Kenya 8 -7 Kenya’s trajectory has been somewhat similar to Ghana. After extremely autocratic governments for most of the sample period, there is a shift to a more democratic government in the 2000s. Nigeria 7 -7 Nigeria shifted to a more democratic government in 1978 which lasted for a few years but went back to a more autocratic regime by 1984. Again, there was a shift in 1998 and currently Nigeria’s polity score is 4 (weak to moderately democratic) Senegal 8 -6 Senegal started off with an autocratic regime which weakened in the 1980s and become strongly democratic in the 2000s (scores of 8 since 2000). South Africa 9 4 Apartheid notwithstanding, South Africa has had a democratic regime for the entire sample period (the only country besides Botswana). Since 1994 (end of apartheid), South Africa has had a strongly democratic regime (polity score of 9). Tanzania -1 -6 Tanzania has had an autocratic government for the entire period. Since 1995 the country has had a weakly autocratic government (polity score of -1) Togo -2 -7 Following a strongly autocratic regime until 1990, Togo has had varying degrees of weak autocratic regimes since then, with a current polity score of 4. Source: Data from Center of Systemic Peace, Integrated Network for Societal Conflict Research database. Proceedings of the 2012 Pennsylvania Economic Association Conference 72 Figure 1B: Botswana's Aid as a Percentage of GNI 14 5000 4500 4000 3500 3000 2500 2000 1500 1000 500 0 12 10 Percentage 6 4 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 1975 2005 2003 2001 1999 1997 1995 1993 1991 1987 1989 1985 1983 1981 1979 1977 0 Figure 1C: Botswana's Inflation Rate Figure 1D: Botswana's FDI as a Percentage of GDP 20 18 16 14 12 10 8 6 4 2 0 15 5 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 -5 1977 0 1975 Percentage 10 1975 Percentage 8 2 1975 U.S. dollars Figure 1A: Botswana's Real GDP Per Capita -10 -15 Proceedings of the 2012 Pennsylvania Economic Association Conference 73 Figure 2B: Cameroon's Aid as a Percentage of GNI 7 35 6 30 5 25 4 2005 2003 2001 1999 1997 1993 1995 2005 2003 2001 1999 1997 1995 1993 1991 -1 1989 0 1987 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 1975 -10 1 1985 5 0 2 1983 10 3 1981 15 1979 20 1977 Percentage 40 -5 1991 Figure 2D: Cameroon's FDI as a Percentage of GDP 1975 Figure 2C: Cameroon's Inflation Rate Percentage 1975 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 1975 0 1989 200 1987 400 1985 600 1983 Percentage U.S. dollars 800 1981 1000 10 9 8 7 6 5 4 3 2 1 0 1979 1200 1977 Figure 2A: Cameroon's Real GDP Per Capita -2 Proceedings of the 2012 Pennsylvania Economic Association Conference 74 Figure 3A: Côte D'Ivoire's Real GDP Per Capita Figure 3B: Côte D'Ivoire's Aid as a Percentage of GNI 25 1000 20 800 15 Percentage 600 400 10 5 200 25 3 20 2 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 0 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 -1 1985 0 1983 5 1981 10 1 1979 15 1977 Percentage 4 -5 1979 1975 2005 2003 2001 1999 1997 1993 1991 1989 1987 1985 1983 1981 1979 1977 1975 1995 Figure 3D: Côte D'Ivoire's FDI as a Percentage of GDP 30 1975 Percentage Figure 3C: Côte D'Ivoire's Inflation Rate 1977 0 0 1975 U.S. dollars 1200 -2 -3 Proceedings of the 2012 Pennsylvania Economic Association Conference 75 Figure 4A: Ethiopia's Real GDP Per Capita Figure 4B: Ethiopia's Aid as a Percentage of GNI 25 160 140 20 100 Percentage U.S. dollars 120 80 60 40 15 10 5 20 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1975 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 1975 Figure 4C: Ethiopia's Inflation Rate 1977 0 0 Figure 4D: Ethiopia's FDI as a Percentage of GDP 7 40 6 5 1 Proceedings of the 2012 Pennsylvania Economic Association Conference 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 -1 1983 0 1981 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 2 1979 -20 1979 -10 1977 0 3 1977 10 4 1975 Percentage 20 1975 Percentage 30 76 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 1975 0 -1 Proceedings of the 2012 Pennsylvania Economic Association Conference 2005 2003 2001 1999 40 1997 60 1995 4 1993 120 1991 5 1989 140 1987 Figure 5C: Ghana's Inflation Rate 1985 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 0 1981 0 1983 2 1979 50 1981 100 1977 250 1979 150 1975 200 U.S. dollars 300 1977 80 Percentage 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 1975 U.S. dollars 350 1975 Percentage Figure 5A: Ghana's Real GDP Per Capita Figure 5B: Ghana's Aid as a Percentage of GNI 18 16 14 12 10 8 6 4 Figure 5D: Ghana's FDI as a Percentage of GDP 100 3 2 20 1 0 77 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 1975 -0.2 Proceedings of the 2012 Pennsylvania Economic Association Conference 2005 2005 2003 2001 2003 1999 1989 1987 1985 1983 2001 0 1997 0.2 1999 0.4 1995 0.6 1997 0.8 1993 1 1995 1.2 1991 1.4 1993 Figure 6D: Kenya's FDI to GDP 1991 1989 1987 1985 Figure 6C: Kenya's Inflation Rate 1981 340 1983 360 1979 12 1981 420 1977 14 1979 380 1975 400 Percentage 440 1977 50 45 40 35 30 25 20 15 10 5 0 Percentage 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 1975 U.S. dollars 460 1975 Percentage Figure 6A: Kenya's Real GDP Per Capita Figure 6B: Kenya's Aid as a Percentage of GNI 18 16 10 8 6 4 2 0 78 Figure 7B: Nigeria's Aid as a Percentage of GNI 12 500 450 400 350 300 250 200 150 100 50 0 10 Percentage 8 6 4 Figure 7C: Nigeria's Inflation Rate 2005 2001 2003 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1975 1977 0 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 2 1975 U.S. dollars Figure 7A: Nigeria's Real GDP Per Capita Figure 7D: Nigeria's FDI as a Percentage of GDP 10 80 70 8 Percentage 50 40 30 20 6 4 2 10 Proceedings of the 2012 Pennsylvania Economic Association Conference 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 -2 1975 0 0 1975 Percentage 60 79 Figure 8A: Senegal's Real GDP Per Capita Figure 8B: Senegal's Aid as a Percentage of GNI 18 16 500 14 400 12 Percentage 300 200 10 8 6 4 100 2 5 30 4 2005 2003 2001 1999 1995 1997 1993 1991 1989 1987 1985 1983 3 Percentage 20 15 10 5 2 1 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 -1 1977 0 0 1975 Percentage 25 -5 1981 Figure 8D: Senegal's FDIas a Percentage of GDP 35 -10 1979 1975 2005 2003 2001 1999 1997 1995 1993 1991 1987 1989 1985 1983 1981 1979 1977 1975 Figure 8C: Senegal's Inflation Rate 1977 0 0 1975 U.S. dollars 600 -2 Proceedings of the 2012 Pennsylvania Economic Association Conference 80 Figure 9A: South Africa's Real GDP Per Capita Figure 9B: South Africa's Aid as a Percentage of GNI 0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 4000 3500 2500 35 6 30 5 25 4 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 2005 2003 2001 1999 1997 1995 -1 1993 0 1991 0 1989 5 1987 2 1 1985 10 3 1983 15 1981 Percentage 7 20 1981 1975 Figure 9D: South Africa's FDI as a Percentage of GDP 40 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 Percentage Figure 9C: South Africa's Inflation Rate 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 1975 0 1979 500 1979 1000 1977 1500 1975 2000 1977 Percentage U.S. dollars 3000 -2 Proceedings of the 2012 Pennsylvania Economic Association Conference 81 Figure 10A: Tanzania's Real GDP Per Capita Figure 10B: Tanzania's Aid as a Percentage of GNI 35 30 300 25 250 20 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1975 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 0 1981 0 1979 5 1977 50 Figure 10C: Tanzania's Inflation Rate Figure 10D: Tanzania's FDI as a Percentage of GDP 40 7 35 6 30 5 Percentage 25 20 15 10 5 4 3 2 1 Proceedings of the 2012 Pennsylvania Economic Association Conference 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 0 1975 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 0 1975 Percentage 10 1983 100 15 1981 150 1979 200 1977 Percentage 350 1975 U.S. dollars 400 82 Figure 11A: Togo's Real GDP Per Capita Figure 11B: Togo's Aid as a Percentage of GNI 18 400 16 350 14 300 12 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 Figure 11D: Togo's FDI as a Percentage of GDP 12 10 Percentage 8 6 4 2 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 -2 1975 2005 2003 2001 1999 1997 1995 1993 1991 1989 1987 1985 1983 1981 1979 1977 0 1975 45 40 35 30 25 20 15 10 5 0 -5 -10 1975 2005 2003 2001 1999 1997 1995 1993 1991 1989 1985 1987 1983 1981 1979 1977 0 1975 2 0 1985 4 50 Figure 11C: Togo's Inflation Rate Percentage 6 1983 100 8 1981 150 10 1979 200 1977 Percentage U.S. dollars 250 -4 ENDNOTES 1 While the unemployment rate is not available, it is estimated to be between 40-50% in 2011. 2 Since 2011, South Africa is part of the BRICS. 3 World Bank Annual Meetings Report, 2011. 4 Lagged values have been used as an instrument of aid by Alvi, Mukherjee and Shukralla (2008) and Armah and Nelson (2008) and others. Also, Gomanee, Girma and Morrisey (2005) who do not use the same estimation technique because they find that instruments are not necessary, also use lagged aid as a variable in their growth regression and note that “this can be interpreted as an instrument (in the spirit of Hansen and Tarp).” 5 Armah and Nelson (2008) use a lagged policy variable among others as an instrument for aid. This variable includes a budget variable, trade openness and inflation. Since our paper focuses on globalization as a separate indicator of growth we leave it out here. In addition, we do not have budget surplus/deficit data and thus our policy variable is based solely on inflation. 6 The Sargan test which is a special case of the J-test is often reported in studies. We conducted Sargan tests for each specification and the results (not reported here) reinforce the conclusion of appropriate instruments. Proceedings of the 2012 Pennsylvania Economic Association Conference 83 REFERENCES Ali, A. M., H. S. Isse, and W. Peek. 2009. A sensitivity analysis approach on the effect of foreign aid on growth. Journal of Applied Business and Economics. 10(3): 97-110. Alvi, E., D. Mukherjee, and E. K. Shukralla. 2008. Foreign aid, growth, policy and reform. Economics Bulletin. 15(6): 1-9. Armah, S. and C. Nelson. 2008. Is foreign aid beneficial for Sub-Saharan Africa? A Panel Data Analysis. Working Paper. Barro, R. J. 1991. Economic growth in a cross-section of countries. The Quarterly Journal of Economics. 106(2): 407-43. Burnside, C. and D. Dollar. 1997. Aid, policies and growth. American Economic Review. 90(4): 847-68. Center of Systemic Peace, Integrated Network for Societal Conflict Research database. Clemens, M.A, S. Radalet, R. Bhavnani, and S. Bazzi. 2004. Counting chickens when they hatch: The short-term effect of aid on growth. Center for Global Development Working Paper No. 44. Collier, P. 2007. The bottom billion why the poorest countries are falling behind and what we can do about it? Oxford University Press. Dalgaard, C-J., H. Hansen, and F. Tarp. 2004. On the empirics of foreign aid and growth. Economic Journal. 114(496): 191-216. Djankov, S., J. G. Montalvo, and M. Reynal-Querol. 2006. Does foreign aid help? Cato Journal. 26(1), Winter issue. Easterly, W. 2003. Can foreign aid buy growth? Journal of Economic Perspectives. 17(3): 23-48. Gomanee, K., S. Girma, and O. Morrissey. 2005. Aid and growth in Sub-Saharan Africa accounting for transmission mechanisms. United Nations University-World Institute for Development Economics Research (UNU-WIDER) Research Paper No. 2005/60. Hansen, H. and F. Tarp. 2001. Aid effectiveness disputed. Journal of International Development. 12(3): 375-98. Hansen, H. and F. Tarp. 2001. Aid and growth regressions. Journal of Development Economics. 64: 547-70. OECD, International Development Statistics Database. Radelet, S. 2006. A primer on foreign aid. Center for Global Development Working Paper No. 92. Rajan, R. and A. Subramanian. 2008. Aid and growth: What does the cross country evidence really show? Review of Economics and Statistics. 90(4): 643-65. Williamson C. R. 2008. Foreign aid and human development: The impact of foreign aid to the health sector. Southern Economic Journal. 75(1). World Bank, World Development Indicators Database. Proceedings of the 2012 Pennsylvania Economic Association Conference 84 EOECONOMIC ANALYSIS OF ALTERNATIVE WAYS TO REFORM MEDICAL MALPRACTICE Tracy C. Miller Grove City College 100 Campus Dr. # 3018 Grove City, Pa 16127 ABSTRACT It is generally believed that malpractice liability raises the cost of health care. A number of reforms to the malpractice liability system have been considered and some have been implemented in selected states. In spite of its problems, however, malpractice liability plays an important role in promoting safe and high quality health care. Rather than imposing uniform limits on malpractice liability, a better approach is to combine a system of privately developed best practice guidelines with legal reforms facilitating binding early offers from providers and contractual agreements between patients and providers to limit liability. INTRODUCTION One cause of high medical costs not addressed by health care reform is malpractice liability. The high cost of malpractice liability has resulted in a variety of state laws being passed to cap damage payments, or make it more difficult to find a health care provider liable. It is hard to dispute the assertion that the malpractice liability system needs to be reformed. Meaningful reform, however, should not reduce or eliminate malpractice liability for patients who value the protection it provides enough to pay the resulting higher prices for health care. Instead reform should make it possible for patients and physicians to make contractual agreements to limit liability and make it easier for defendants to settle negligence claims by offering immediate payments for actual damages. Malpractice is defined as “unreasonable” or negligent treatment of a patient by medical personnel responsible for health care (Wright, Godfrey, & Eberhardt, 1988). Evidence of damage or injury is necessary. Malpractice purportedly raises health care costs in two ways-- through the cost of insurance premiums that physicians pay to protect themselves from the risk of large judgments, and through the extra costs of diagnosis and treatment incurred by doctors to defend their decisions in case they are sued. The central elements of malpractice are based on common law, which gradually evolved as part of legal practice, beginning in England in the eleventh century (Wright, Godfrey, & Eberhardt, 1988). Negligence is defined as “the omission to do something which a reasonable man, guided by those ordinary considerations which ordinarily regulate human affairs, would do: or the doing of something which a reasonable and prudent man would not do” (Harrison, Worth, & Carlucci, 1985, p. 42) In Australia, USA and the UK, a doctor is liable if he or she is more likely than not to have acted negligently. In the US and UK compliance with customary practice is an acceptable defense (Kessler, Summerton and Graham). Australia takes a different approach- judges are the ultimate arbiter of the standard of negligence. Some states in the US allow judges to reject customary practice as a defense in some circumstances. A malpractice crisis arose in the mid-1970s, exemplified by insurance unavailability for some physicians and high costs for others (Wright, Godfrey, & Eberhardt, 1988). Similar crises arose in the 1980s and in the early 2000s. In response to the crisis in the 1970s, most state legislatures responded by enacting tort reform laws. Reforms that have been implemented include patient compensation funds, limits on noneconomic damages, statutes of limitations, limitations on contingent fees, and absolute dollar limits on compensation that a patient could recover for a single injury or death. In some states, providers are liable for all damages up to the limit. In other states, a patient compensation fund is responsible for recoveries above a specified amount, such as $100,000. Arbitration systems & pretrial screening panels were also adopted with the purpose of directing the handling of claims away from courts and eliminating frivolous claims. THE CASE AGAINST LIMITS ON MALPRACTICE AWARDS Laws to limit malpractice liability have been challenged in a number of state courts, and some of the laws have been struck down. Some judges reject compensation limits on the grounds of equal protection- they give unequal advantage to health-care providers (Wright, Godfrey, & Eberhardt, 1988). Courts have also argued that caps are regressive. In striking down a $350,000 cap on noneconomic damages in Wisconsin in 2005, the court noted that the cap had the greatest impact on severely injured children (Nelson, Morrisey, & Kilgore, 2007). Since economic damages include lost income, they are lower for children, so caps on noneconomic damages reduce the incentive of filing lawsuits on behalf of injured children. Courts have also thrown out damage caps on the grounds that evidence is inconclusive about their effectiveness. Proceedings of the 2012 Pennsylvania Economic Association Conference 85 In addition to the distributional impacts of malpractice liability and laws to limit it, malpractice liability impacts health care in three ways. It influences the incentives of health care providers to make mistakes, to practice defensive medicine, and to provide treatment and procedures of questionable medical value. Limiting liability can save costs by reducing defensive medicine but it can lead to increases in the frequency of errors and in the cost of excessive medical care, sometimes referred to as offensive medicine (Avraham, 2011). Some would argue that because most health care providers have liability insurance, the threat of being sued has little impact on physician behavior and health care outcomes. Although malpractice insurance protects physicians from legal expenses, through experience rating of premiums, it creates incentives for physicians to practice safer medicine. Insurers reward physicians for passing certification tests, adopting new techniques, or purchasing safer equipment, while penalizing physicians with more claims filed against them, with substance abuse problems, or with other personal characteristics that might increase their risk of being negligent. The conventional wisdom is that malpractice insurance has less experience rating than other lines of insurance (Danzon, 2000, Sloan, 1990). Svorny (2011) disputes the conventional wisdom, noting that experience rating is hard to observe because different insurance companies specialize in insuring providers with different levels of risk. Thus individual companies may not charge different premiums to physicians in the same specialty, but those whose risks are higher pay more. Some companies only insure physicians with spotless records. Physicians whose applications are rejected by lowpriced insurance companies must turn to surplus-lines carriers and pay between 150 and 500 percent of the prices in standard markets (Svorny, 2011, p. 7). CRITICAL ASSESSMENT OF THE US MALPRACTICE SYSTEM Compared to identifying the negligent party in auto accidents and workplace injuries, medical negligence is hard to identify because of the fact that patients seeking treatment are already ill or injured and because treatment has different effects on different patients (Danzon, 2000). This opens the door for decisions that reflect cognitive biases of judges and juries. Evidence suggests that courts compensate injured victims when their harm is large, even if negligence is not involved (Avraham, 2011). Critics of malpractice point out that a very small percentage of victims of negligence file malpractice cases. Malpractice cases that are filed have very high transactions cost as a share of payouts to victims. According to Studdert, Mello, and Gawande (2006), for every dollar spent on patient compensation, 54 cents went to administrative expenses. Malpractice liability contributes significantly to health care costs by raising the cost of practicing medicine, particularly in selected specialty areas. Obstetricians and gynecologists bear some of the highest liability costs, which have resulted in some giving up their practice and others moving away from high-risk areas. Obstetricians can be held liable for injuries sustained at birth for up to 21 years (Wright, Godfrey, & Eberhardt, 1988). Caps result in high costs of a different kind. With pregnancies, caps resulted in less care being taken but may result in more procedures, particularly if procedures are risky, as is the case with Caesarian sections, as well as induction and stimulation of labor (Currie & Macleod, 2008). Caps on noneconomic damages also reduce the incentive for filing lawsuits on behalf of children, married women and the elderly, for whom economic damages are generally smaller, thus resulting in their families bearing the costs of medical injuries that they sustain. Although it has its flaws, a careful assessment reveals that the malpractice system works better than many critics believe. When cases aren’t filed, it is often due to small damages, either because the patient fully recovers or is very old (Svorny, 2011). In evaluating a sample of cases raised against a large hospital between 1977 and 1989, Farber and White (1991) found that most of the cases that go to trial, which are a small proportion of the total, are won by defendants, but that the vast majority of cases were dropped, dismissed or settled out of court. Their results imply that the discovery process in malpractice cases serves the purpose of enabling poorly informed plaintiffs to gather information about whether negligence is likely. This results in a substantial percentage of cases being dropped by plaintiffs or dismissed by the judge. Plaintiffs received damage payments in one fourth of the cases where the hospital rated care quality as good, but in 89 percent of the cases where hospitals rated care quality as bad (Farber & White, 1991). The high transactions cost of malpractice cases do not seem so large when compared to the losses from injuries that may be prevented as doctors and hospitals seek to take precautions to keep their insurance premiums down. Farber and White (1991) noted that the expected liability of the hospital in their study was 25 times as high in cases involving negligence (where the hospital reported bad care) than in cases that it did not (reported by the hospital as good care). Thus a strong incentive exists to provide high quality care so as to avoid liability. Thus limits on liability may increase negligence and could even increase the number of malpractice cases. Proceedings of the 2012 Pennsylvania Economic Association Conference 86 If most victims of malpractice do not file lawsuits, as asserted by Localio et al. (1991), then high awards may be warranted. If the probability of being sued is low even if a physician is guilty of negligence, then his incentive to be careful may be too low unless the cost when he is found guilty is considerably greater than the damage caused by his actions. SUGGESTIONS FOR REFORM Proposals to reform medical liability have included a variety of measures. A common approach is to limit compensation to loss of earning capacity & reasonable medical expenses and related monetary costs. This would eliminate a major share of compensation that goes for pain, suffering, or punitive damages. Another possible reform is to eliminate or reduce contingent fees. Lawyer fees are contingent if they depend on case outcomes. Courts in the UK and Australia do not allow contingent fees, but they do allow conditional fees that vary with the outcome of the case (Kessler, Summerton and Graham, 2006). Some states have set limits on contingent fees. Eliminating contingent and conditional fees may discourage lawyers from filing suits where claims are weak or frivolous. At the same time, contingent fees enable some risk averse patients to file valid claims that they might not choose to file if they had to pay hourly fees regardless of the outcome of the case (Danzon, 2000, p. 1375). The problem with the system of medical liability is not that it results in very large payouts for those guilty of negligence. The problem is the cost of defending cases where the physician was not negligent. This includes the costs of defensive medicine whose primary purpose is to protect the physician from liability if he or she is sued. Although, caps on malpractice awards may reduce the incidence of cases where the plaintiff has a low probability of winning, caps will also discourage some cases being brought where there was negligence. Defensive practices added an estimated $500,000 per year of life saved for heart disease patients according to one study (Kessler and McClellan, 1996). Doctors practice defensive medicine because they bear “uninsured, nonmonetary costs of liability” (Kessler, Summerton and Graham, 2006). These include lost time, harm to their reputations and the emotional impacts of responding to claims. Health insurance reimbursement encourages more defensive medicine, since doctors and patients bear less of the cost. Rather than putting caps on malpractice liability, defensive medicine can be limited by other means, such as by managed care organizations not reimbursing unnecessary defensive diagnoses and treatment. One promising reform that could reduce spending on defensive medicine is to make evidence of compliance with written clinical practice guidelines an acceptable defense against negligence. Medical guidelines have been developed by a variety of organizations over the years, with a dramatic increase in the number of guidelines being produced beginning in the 1990s (Avraham, 2011, p. 8). The Agency for Health Care Policy and Research played an important role in this. Although guidelines seem to have had limited effect on physician behavior so far, if implemented with adequate incentives they have the potential to reduce the use of costly defensive medicine and provide physicians grounds for an affirmative defense if they are sued. Existing guidelines, many of which have been developed by government agencies, are problematic. Government agencies do not have an incentive to keep guidelines up-to-date so that they are consistent with the latest evidence about best practices. Outdated guidelines could themselves give health care providers cover for not using the most up-to-date procedures for treating a particular patient. Nevertheless, guidelines could play a beneficial role in discouraging defensive medicine if they were designed by private firms. With appropriate changes in legal infrastructure, a private regulation regime could permit firms to develop guidelines and be held accountable for the consequences of those guidelines in terms of promoting safe treatment of patients (Avraham, 2011). Another way to save costs is to make it easier to settle cases quickly. In some cases, both plaintiffs and defendants could benefit by a reform that trades off reduced compensation for greater access to compensation. One way to accomplish this would be to reform liability rules to favor binding early offers made by defendants. This approach was proposed by a 2006 Department of Health and Human Services study (Hersch, O'Connell, & Viscusi, 2006). It involves a statute that would give a health care provider the option to offer an injured patient, within 180 days after a claim is filed, periodic payments of the patient’s net economic losses as they accrue. The statute would provide criteria for how much compensation would be required in a way similar to workers’ compensation statutes (O'Connell, 2007). The plaintiff would have a choice between accepting the early offer or declining it in favor of litigation. If he or she declined the early offer, however, the plaintiff could only win if gross negligence were proven beyond a reasonable doubt. The advantage to the plaintiff of accepting the early offer would be reduced uncertainty, delays, and transactions costs compared to full-scale tort claims. While a reform facilitating cases to be settled with early offers will reduce the number of cases resulting in large payouts for noneconomic damages, the impact on the incentive to file lawsuits is unclear. Fewer lawyers will earn the high fees from large payouts, but contingent fee lawyers will collect some fees in a larger percentage of cases. Proceedings of the 2012 Pennsylvania Economic Association Conference 87 Damage caps clearly reduce the benefits to some plaintiffs of filing a lawsuit, while early offer reform creates the opportunity for gains to both the defendant and the plaintiff compared to the status quo. A statute to reform liability to include an early offer reform was first proposed in the 1980s (O'Connell, 2007). It is not as appealing to defendants as rules limiting claims and is opposed by the plaintiff’s bar because it would reduce spending on litigation. If enough consumers, health care providers and insurance companies can be convinced that early offer reform could potentially benefit everyone but the lawyers, perhaps it will be implemented in some states. Early offer reform preserves the essential feature of the malpractice liability system so that patients are fully protected against the costs they bear due to provider negligence. Although it could have similar effects to caps on noneconomic damages in ordinary negligence cases, it permits plaintiffs to reject early offers and proceed with lawsuits seeking noneconomic damages in cases involving gross negligence. In addition to damage caps in some states, liability of physicians is also limited by a variety of other provisions. Government employed physicians are shielded from malpractice claims by the Federal Torts Claims Act (Svorny, 2011). Joint Underwriting Associations, which operated in 13 states in 2007, act as insurers of last resort for physicians who cannot obtain insurance in private markets. By not exposing physicians to the full costs of liability for their actions, such arrangements increase the risk of patients being harmed. much that patient values the protection he or she receives. Instead some have argued that courts should honor contracts that patients could agree to in advance that limit how much liability the health care provider would incur (Cannon, 2010). This would make is possible for some patients to pay less for health care in exchange for a smaller award in the case of negligence. Under our current system where third parties pay most medical expenses, individual patients would not benefit much from such agreements. Thus an important step to improving the malpractice system is for patients to pay a greater share of their own health care expenses. Then many people could benefit directly from reforms that allow negotiation between patients and health care providers over malpractice limits. Such negotiated limits should not apply to gross negligence, which is clearly provable in only a small percentage of cases (Hersch, O'Connell, & Viscusi, 2006). Many would object to patients negotiating away the right to hold health care providers liable because of concerns that uninformed patients would be taken advantage of. It is unlikely that courts would ever uphold any and all contracts to limit liability. Nevertheless, a menu of limitations established by third parties, such as health insurance providers, could play an important role in reducing the costs of health care in exchange for less protection in the event of negligence. As Cannon points out, “we have struck a balance that demands greater protection from simple negligence than many patients would prefer, that is uniform and inescapable”. Because the existing liability system raises the cost of health care for everyone, some get too much protection from negligence and not enough health care. CONCLUSION Liability of health care providers combined with malpractice insurance could play an even bigger role than it does now in promoting safe, efficient, and low cost health care. This could be accomplished by eliminating state licensing requirements and relying on companies that provide liability insurance to certify that health care professionals are competent and do quality work. Liability insurers have incentives to do a more comprehensive job of regularly evaluating the quality of care and the competence of physicians than do licensing boards. “Malpractice underwriters review physicians annually”, evaluating claims histories, investigating physicians for behavior that increases their risks of negligence and providing information and advice to the medical community (Svorny, 2011). Without licensing, arbitrary education and training standards would no longer restrict the supply of health care providers. Instead liability insurance companies would provide the information patients need to choose physicians that provide safe and high quality care. One problem with malpractice liability is that the court system, which is a government monopoly, gives about the same level of protection to every patient regardless of how Rather than capping malpractice damages for everyone, moving toward a free market in health care along with reforms to our legal system would enable individual patients and their families to choose in advance the level of protection from risk that they are willing to pay for. This would make health care more affordable while permitting those who value a high level of protection against negligence the freedom to pay more for such protection. Liability for malpractice gives health care providers an incentive to pursue measures to reduce risk. This is the cost poorly informed patients pay for safer medicine and the right to be compensated in case of negligence. The best way to reform liability for medical malpractice is to permit mutually beneficial arrangements, such as binding early offers and agreements between patients and providers to limit liability. Reforms such as caps on liability take away an important protection that many patients value. Additional reforms such as requiring the loser to pay all court costs also make sense. Nevertheless, the key is to preserve the protection against negligence of the existing system, while opening the door for alternative arrangements that would make it possible to Proceedings of the 2012 Pennsylvania Economic Association Conference 88 reduce health care costs for those who cannot afford the high costs of the existing system. REFERENCES Avraham, R. (2011). Clinical Practice Guidelines: The Warped Incentives in the U.S. Healthcare System. American Journal of Law and Medicine, 37(1), 7-40. Cannon, M. (2010, November 12). Reforming Medical Malpractice Liability Through Contract. Retrieved November 9, 2011, from Cato Institute. Congressional Budget Office. (2004, January 8). Limiting Tort Liability for Medical Malpractice. Retrieved May 9, 2012, from Congressional Budget Office. Currie, J., & Macleod, W. B. (2008, May). First do no harm? Tort Reform and Birth Outcomes. Quarterly Journal of Economics, 123(2), 795-830. Danzon, P. M. (2000). Liability for Medical Malpractice. In A. J. Culyer, & J. P. Newhouse, Handbook of Health Economics,Volume 1 (pp. 1339-1404). Amsterdam: Elsevier. Farber, H., & White, M. (1991). Medical Malpractice: An Empirical Examination of the Litigation Process. The RAND Journal of Economics, 22(1), 199-217. Harrison, L., Worth, M., & Carlucci, M. (1985). The development of the principle of medical malpractice in the United States". Perspectives in Biology and Medicine, pp. 41-72 cited in Wrght, Fedorvich and Close. Hersch, J., O'Connell, J., & Viscusi, K. (2006, June 5). Evaluation of Early Offer Reform of Medical Malpractice Claims: Final Report. Retrieved May 25, 2012, from Department of Health and Human Services: http://aspe.hhs.gov/daltcp/reports/2006/medmalcl.pdf Kessler, D., & McClellan, M. (1996). Do doctors practice defensive medicine. Quarterly Journal of Economics, 111(2), 353-90. Kessler, D., Summerton, N., & Graham, J. (2006). Effects of the medical liability system in Australia, the UK, and the USA. Lancet, 368, 240-246. Localio, A. R., Lawthers, A., Brennan, T., Laird, N., Hebert, L., Peterson, L., et al. (1991). Relation between Malpractice Claims and Adverse Events Due to Negligence. New England Journal of Medicine 325, 245-251. Nelson, L., Morrisey, M., & Kilgore, M. (2007). Damages Caps in Medical Malpractice Cases. Milbank Quarterly, 85(2), 259-286. O'Connell, J. (2007). Binding Early Offers versus Caps for Medical Malpractice Claims. Milbank Quarterly, 85(2), 287296. Sloan, F. (1990, May). Experience Rating: Does It make Sense of Medical Malpractice Insurance? American Economic Review, 80(2), 128-133. Studdert, D., Mello, A., & Gawande, T. (2066). Claims, Errors, and Compensation Payouts in Medical Malpractice Litigation. New England Journal of Medicine, 354, 20242033. Svorny, S. (2011). Could Mandatory Caps on Medical Malpractice Damages Harm Consumers. Washington, DC: Cato Institute. Wright, J. E., Godfrey, G. M., & Eberhardt, T. C. (1988, September-October). Analyzing Caps on Malpractice Monetary Awards, Part I: Historical and Social Perspectives. Nursing Economics, 6(5), 257-260. Proceedings of the 2012 Pennsylvania Economic Association Conference 89 DOES A FIRM’S DIVIDEND INITIATION AFFECT ITS RISK? Henry F. Check, Jr. Penn State Lehigh Valley (retired) 4014 Autumn Ridge Road Bethlehem, PA 18017 John S. Walker Department of Business Administration Kutztown University of Pennsylvania Kutztown, PA 19530 Karen L. Randall 1159 Hagues Mill Road Amber, PA 19002 ABSTRACT The Jobs and Growth Tax Relief Reconciliation Act of 2003 (JGTRRA) was signed into law by President Bush on May 28, 2003. Under the JGTRRA corporate dividends are no longer taxed as ordinary income; instead, they are taxed preferentially like capital gains and in some cases not taxed at all. This study examines the stock performance of a sample of 18 firms which initiated cash dividends immediately after JGTRRA was enacted. We find that the risk of these firms’ stock, as measured by the variance of returns, decreased in absolute terms after dividends were initiated. Since this was a period of generally declining stock market volatility, we also compared our sample firms’ variances to the market as a whole and found that a majority of the firms’ risk declined relative to the market as well. Finally, we compared the firms’ systematic risk pre- and post-dividend initiation, as quantified by the Capital Asset Pricing Model’s (CAPM) beta measure, and found that a majority of the firms’ systematic risk changed. When systematic risk (i.e., the firm’s beta) changed, roughly half of the firms in our sample experienced higher betas and half saw lower betas post dividend initiation. The application of this study is to suggest to corporations dividend policy strategies which would reduce their cost of capital, and to suggest to investors equity-screening strategies which would increase their rate of return. This work is especially timely as the sunset provision of JGTRRA will restore the ordinary income treatment of dividends after December 31, 2012, unless extended by Congress, and corporations with huge market capitalizations (specifically Apple and Google) are considering initiating cash dividends. INTRODUCTION Companies decide to pay dividends for various reasons and have a number of constraints on those payments. For example, firms are precluded from retaining earnings indefinitely; if they can’t find viable investments and continue to retain earnings, they could face an accumulated earnings tax. Thus, even high-growth companies should eventually pay a dividend when their need for capital diminishes. For companies paying dividends, or initiating dividend payments, the amount of dividends paid to shareholders might be limited by loan covenants that require debt servicing and/or liquidation prior to dividend payments. Standard textbook discussions of dividend theory examine dividend decisions using various models. For example, Graham, Smart, and Megginson (2010) discuss (1) the agency cost/contracting model, (2) the signaling model, and (3) the catering model of dividend payout. The agency cost/contracting model recognizes conflicts that can arise between a firm’s managers and shareholders. The model explains the desirability of dividend payments as a way to help resolve conflicts between managers who want to retain earnings to fund perquisites and negative NPV (“pet”) projects and shareholders who want to receive a portion of the firm’s earnings in order to reduce agency costs. The signaling model explains that dividends and share repurchases are a way for management to “signal” the marketplace that the firm is sufficiently “strong” to utilize cash flow in this manner. Healy and Palepu (1988) find that dividend-initiating firms experience higher earnings growth than industry peers within the first year following initiation. These higher earnings translate into abnormal returns for investors. Their findings are consistent with the signaling model, as the results indicate that investors can interpret announcements of dividend initiations as managers’ forecasts of future earnings changes. Proceedings of the 2012 Pennsylvania Economic Association Conference 90 The catering theory suggests that firms will pay dividends when shareholders are paying a premium for stocks that are paying a dividend. Conversely, the theory suggests that firms will not pay a dividend when shareholders are paying a discount for shares that are paying a dividend. Baker and Wurgler (2004a, 2004b) provide analysis and data that support this more recent theory on dividend policy. While research on the catering theory is more current, it is closely related to what other finance textbooks term the “clientele effect.” Brigham and Ehrhardt (2008) observe that different groups of stockholders prefer different dividend payout policies; for example, certain shareholder groups (clienteles) may prefer higher payouts. While catering theory appears to be one of the most recent additions to the various theories on dividend policy, Graham, Smart, and Megginson (p. 494) observe that “it does not describe why dividends are paid in the first place.” The focus of our research is not to explain why a firm decides to initiate a dividend. Instead, we look at the impact dividend initiation has on a firm’s risk profile using a combination of three different risk metrics and a statistical/econometric methodology that is not used in prior research looking at dividend initiation, to the best of our knowledge. We look at firms that initiated dividend payments shortly after the passage of the Jobs and Growth Tax Recovery Reconciliation Act of 2003 (JGTRRA). We find that a firm’s total risk is reduced as a result of dividend initiation, which can be one benefit of dividend initiation. However, we find inconsistency in the impact to a firm’s systematic risk. Because an investor is capable of diversifying away firm specific risk but not systematic risk, changes in systematic risk are more noteworthy. RELEVANT RESEARCH The scholars who first come to mind when you think about seminal work on dividend policy are Lintner (1956) and Modigliani and Miller (1958 and 1959). Lintner uses the term “progressive, continuing partial adaptation” to describe firms’ behavior to increase dividends by only a fraction of what the firm’s current financial performance will support. If we apply this conservative approach to dividend initiation, then a firm’s management will only decide to initiate a dividend when the free cash flow is sufficient to maintain that dividend for the foreseeable future. This view is consistent with signaling theory as management’s decision about increasing the firm’s dividend conveys valuable information to investors. Modigliani and Miller (1958 and 1959) introduce two historic irrelevancy propositions and argue that the value of the firm is independent of capital structure and dividend policy. Their papers apply, many have since argued, unrealistic specific assumptions, such as, “…all capital market participants, inside managers and outside investors alike, have the same information about the firm’s cash flows” (p. 105, Miller, 1988). One benefit of M&M’s work is that it gives future researchers a reference point. Our research finds that dividend initiation reduces total firm risk, and investors can learn this information at the time that a firm announces that it will begin paying a dividend. Thus, material, asymmetric information about risk can exist prior to the dividend initiation announcement. Past researchers who have looked at the impact to risk from dividend initiation include Venkatesh (1989) and Dyl and Weigand (1998). Venkatesh is one of the first to look at the volatility of returns after initiation and finds decreases in the post-initiation volatility as measured by standard deviation. He feels that the decrease is only in the firm-specific volatility as opposed to systematic volatility, which he shows by examining firm betas. He finds the betas to be stable while the standard deviations are unstable. He explains that risk is reduced post dividend initiation because investors can now observe dividend declarations, not just earnings announcements. Consequently, after initiation, investors give less weight to information cues other than dividends and earnings, whereas in the pre-dividend initiation period, they may have reacted to other information more strongly. Nearly ten years later, Dyl and Weigand find that dividend initiation leads to a decrease in total risk and systematic risk in the year following the dividend announcement. They view a firm’s dividend initiation announcement as providing investors valuable information about the firm’s risk and, thus, term this the “risk information hypothesis.” According to this view, the initiation of dividends is a signal to markets that a firm’s earnings and cash flows have become fundamentally less risky, and therefore beta is lower. Firms are expected to have fewer surprises after dividend initiation. If the earnings volatility decreases, then the volatility of returns will decrease as well. The D&W study finds lower betas over the three year period following dividend initiation. We believe our research makes an important extension to Venkatesh’s and D&W’s research as we consider changes to a firm’s risk relative to the risk of the overall stock market. D&W only looked at total and systematic risk. Grullon, Michaely, and Swaminathan (2002) examine the relationship between changes to dividends and changes in systematic risk. While D&W and our research look at dividend initiation, GM&S exclude companies that are initiating dividends. Nevertheless, we believe it is noteworthy to report that GM&S find that systematic risk declines around the time of dividend increases. In contrast to the consistent pattern of lower total and relative risk that we observe, we find inconsistent changes to systematic risk in the case of dividend initiation. The focus of our research is to consider the impact of dividend initiation to a firm’s risk. Looking at price behavior Proceedings of the 2012 Pennsylvania Economic Association Conference 91 pre- and post-initiation and the wealth effect to shareholders are logical extensions to our research. Jin (2000) finds that there is an “apparent heterogeneity” in how stock prices react to dividend initiation announcements. Her research observes that there can be both benefits and costs to dividend initiation. For example, initiation of dividends can help to mitigate agency costs. In contrast, dividend initiation could be a signal that management expects fewer growth opportunities for the firm. A more extensive discussion of the literature, which is beyond the score of this paper, finds that the impact from dividend initiation to share price and stockholders wealth is inconsistent. Yet, it is relevant to mention these inconsistencies as our analysis of systematic risk also produced inconsistent results. Our research provides some new insights into the impact to risk from dividend initiation, but it also adds empirical results to the extant literature that are consistent with some and inconsistent with other researchers’ findings. Moreover, our study focuses on firms that initiated dividend payout soon after the Jobs and Growth Tax Relief Reconciliation Act of 2003 (JGTRRA). We have not proven or asserted any significance to the timing of firms’ decisions to pay dividends to the JGTRRA. However, we reference the JGTRRA as part of our timeline because our dataset contains firms that implemented their dividend initiations shortly after the passage of this act. METHODOLOGY AND RESULTS Our process began with an Internet search for firms which had initiated cash dividends in the 2003–2004 time frame. We found 25 such firms. We then obtained daily and monthly stock prices from Yahoo! Finance. (Clayton, Jahera and Schmidt (2008) has certified Yahoo! Finance as a data source suitable for academic research.) That search resulted in seven firms being dropped from our sample because they were no longer publicly traded. Four firms were acquired by other firms (American Power Conversion Corporation, Phelps Dodge Corporation, Reebok International Limited, and Tektronix), two firms were taken private (Clear Channel Communications, Inc. and Harrah’s Entertainment), and one firm was broken up and sold off (Cendant Corporation). Table 1 shows our 18 sample firms. We also obtained from Yahoo! Finance daily and monthly S&P 500 Index prices, and daily and monthly 10-year Treasury yields for our market and risk-free return measures, respectively. The initial observation for each firm was the first day that firm was publicly traded or January 1, 1997, whichever date was later. Final observations were the last day the firm was publicly traded or December 31, 2008, whichever was earlier. This time frame brackets the dividend initiation dates for our sample firms and avoids much of the market turmoil of the recession. Dividend initiation dates were obtained as part of the original Internet search. We first tested whether each firm’s total risk decreased following the initiation of dividends. From the stock prices, which were adjusted for dividends and stock splits, stock returns and variances were calculated. The hypotheses tested are: In words, the null hypothesis says that the risk of the firm before dividend initiation is either the same as or less than the risk of the firm after dividend initiation; the alternate hypothesis says that the risk of the firm before dividend initiation is greater than the risk of the firm after dividend initiation. This is an F-test where the before/after variance ratio for each firm is compared to a critical F value. The critical F value is different for each firm as the degrees of freedom vary with the number of before/after dividend initiation observations. Proceedings of the 2012 Pennsylvania Economic Association Conference 92 Table 2 presents the results of this hypothesis test using daily data. Notice every firm exhibited a statistically significant decrease in risk. Table 5 in the Appendix shows comparable results using monthly data. constructed by subtracting and adding a margin of error, E, to the after and before risk ratios to obtain the LCLs and UCLs. The endpoints of the confidence intervals are calculated as: For clarity, the confidence interval method of hypothesis testing is shown in Figure 1. We used a 95% confidence coefficient to construct our confidence intervals. If the upper limit of the post-dividend 95% confidence interval is less than the lower limit of the pre-dividend 95% confidence interval, we can reject the null hypothesis of the postdividend variance ratio being greater than or equal to the predividend variance ratio, meaning that the firm is less risky relative to the market after dividend initiation. We next test whether each firm’s total risk per unit of market risk decreases following the initiation of dividends. As done with the stock price data, we calculate market returns and variances. This enabled us to then calculate the ratio of firmto-market variance for each firm before and after the dividend initiation announcement. The hypotheses tested are: In words, the null hypothesis says that the risk of the firm relative to the market before dividend initiation is either equal to or less than the risk of the firm relative to the market after dividend initiation. The alternate hypothesis says that the risk of the firm relative to the market before dividend initiation is greater than the risk of the firm relative to the market after dividend initiation. Figure 1 – Confidence Interval Method of Hypothesis Testing Table 3 presents the results of this process using daily data. For example, consider the first company, Analog Devices. The firm’s LCLbefore (9.74) is greater than the UCLafter (6.59) confirming that the confidence interval for relative risk prior to dividend initiation is higher than the risk interval after initiation. Note that all sample firms evidenced a statistically significant decrease in relative risk. Table 6 in the Appendix shows comparable results using monthly data. The second set of hypotheses is tested using the confidence interval approach. The endpoints of the confidence intervals are denoted as LCL for the lower confidence limit and UCL for the upper confidence limit. The confidence intervals are Proceedings of the 2012 Pennsylvania Economic Association Conference 93 positive alpha means a stock is under priced; a negative alpha means a stock is overpriced. In detail, our dummy variable regression model is: The above discussion and the first two sets of hypotheses tested pertain to total risk. However, the relevant risk measure for investors is systematic risk, the risk that remains after an investor’s portfolio is fully diversified. One measure of systematic risk is the Capital Asset Pricing Model’s (CAPM) beta measure. Beta is calculated as the slope of a linear regression where the dependent variable is the excess return of the stock and the independent variable is the excess return of the market. Excess returns are returns which have the risk-free rate of interest netted out. This regression line is referred to as the Characteristic Line. The hypotheses tested are: Table 4 summarizes the results of the dummy variable regression using daily data. Table 7, in the Appendix, presents these results more completely. Notice that 13 of the 18 sample firms exhibited a statistically significant change in systematic risk; six decreased and seven increased. Tables 8 and 9 in the Appendix show comparable results using monthly data. In words, the null hypothesis says that the stock’s systematic risk is unchanged; the alternate hypothesis says that the stock’s systematic risk has changed. Changes in betas can be tested in a variety of ways such as a difference of means, a change in the correlation coefficient, and the Chow Test that compares sums of squares. We utilized a dummy variable regression method since that allowed us to test the intercept, alpha, as well as the slope, beta, simultaneously. Nonzero alphas are especially interesting to investors since they represent a “free lunch.” A Proceedings of the 2012 Pennsylvania Economic Association Conference 94 trading days, different starting and ending days of the week, and different seasons, and might involve different ends of quarters and ends of years. Our results from the monthly data in no way contradict the daily data results. They are merely less significant. We find that systematic risk, as quantified by beta, exhibited a statistically significant change for 13 of our 18 sample firms. It is somewhat perplexing that the change observed was a nearly equal number of increases and decreases (seven up, six down). To our knowledge this effect has not been previously published. A priori it might be expected that beta should decrease when a firm initiates dividends since that initiation could be taken as a signal to investors that the firm is confident enough in its future cash flows to commit themselves to continuing dividend payments. However, the initiation of dividends could alternatively be a signal that the firm is moving from the growth phase to the maturity phase of its lifecycle. The payment of dividends might mean that the firm can no longer reinvest earnings in projects that would earn more than their rate of return on equity. SUMMARY AND CONCLUSIONS The analysis above demonstrates quite conclusively that initiating dividends reduces a firm’s total risk and affects its systematic risk. These results are consistent with prior research that examined other timeframes. However, we believe that we are the first to examine these effects on a relative basis. Just as we judge a firm’s stock performance relative to other firms in its industry or to the market as a whole, the riskiness of a firm’s return should be compared to a suitable benchmark. Simply demonstrating a decrease in risk following dividend initiation leaves open the possibility that the market as a whole was becoming less volatile during the period studied. This is in fact what occurred during the period covered by our study. Thus it becomes necessary to examine relative risk. We confirm that the decrease in total risk caused by initiating dividends is larger than the decrease in market risk that occurred independently and concurrently. That relative decrease was statistically significant in all 18 firms in our sample using daily stock price data. The changes in the risk measures that we studied—i.e., absolute risk, relative risk, and systematic risk—were all stronger using daily data compared to monthly data. We attribute this to the smaller sample sizes and to the additional variability introduced by the somewhat arbitrary assigning of months. Different months contain different numbers of We posit that the change in beta caused by a dividend initiation is affected by the timing of that initiation. If the market feels that the timing is appropriate, beta will decrease. If the market feels that the initiation is early or late, beta will increase. Such an increase could reflect the market’s judgment that sub-prime projects had been accepted or additional fertile opportunities for diversification of earnings sources were not being cultivated. This effect is an opportunity for future research. An event study methodology could be used to compare the change in beta, decrease or increase, with subsequent changes in the firm’s earnings. REFERENCES Baker, Malcolm, and Jeffrey Wurgler. 2004a. “A Catering Theory of Dividends,” Journal of Finance, vol. 59, no. 3 (June):1125–1165. Baker, Malcolm, and Jeffrey Wurgler. 2004b. “Appearing and Disappearing Dividends: The Link to Catering Incentives,” Journal of Financial Economics, vol. 73, no. 2 (August):271–288. Brigham, Eugene F. and Michael C. Ehrhardt. 2008. Financial Management: Theory and Practice. 12th ed. Mason, OH: Thomson Higher Education. Clayton Ronnie J., John S. Jahera, Jr., and Bill Schmidt. 2009. “Estimating Capital Market Parameters: CRSP Versus Yahoo Data.” In Advances in Investment Analysis & Portfolio Management, Vol. 3 (pp. 173–193). Edited by Cheng F. Lee and Alice C. Lee. New Jersey and Taipei: Center for Pacific Basin Business, Economics, and Finance Research, Airiti Press, Inc. Proceedings of the 2012 Pennsylvania Economic Association Conference 95 Dyl, Edward A., and Robert A. Weigand. 1998. “The Information Content of Dividend Initiations: Additional Evidence,” Financial Management, vol. 27, no. 3 (Autumn):27–35. Graham, John R., Scott B. Smart, and William L. Megginson. 2010. Corporate Finance: Linking Theory to What Companies Do. 3rd ed. Mason, OH: South-Western Cengage Learning. Grullon, Gustavo, Roni Michaely, and Bhaskaran Swaminathan. 2002. “Are Dividend Changes a Sign of Firm Maturity?,” Journal of Business, vol. 75, no. 3 (July):387– 424. Healy, Paul M., and Krishna G. Palepu. 1988. “Earnings Information Conveyed by Dividend Initiations and Omissions,” Journal of Financial Economics, vol. 21, no. 2 (September):149–175. Jin, Zhenhu. 2000. “On the Differential Market Reaction to Dividend Initiations,” Quarterly Review of Economics and Finance, vol. 40, no. 2 (Summer):263–277. Lintner, John. 1956. “Distribution of Income of Corporations Among Dividends, Retained Earnings, and Taxes.” American Economic Review, vol. 46, no. 2 (May):97–113. Miller, Merton H. 1988. “The Modigliani–Miller Propositions After Thirty Years,” Journal of Economic Perspectives, vol. 2, no. 4 (Fall):99–120. Modigliani, Franco, and Merton H. Miller. 1958. “The Cost of Capital, Corporation Finance and the Theory of Investment,” American Economic Review, vol. 48, no. 3 (June):261–297. Modigliani, Franco, and Merton H. Miller. 1959. “The Cost of Capital, Corporation Finance and the Theory of Investment: Reply,” American Economic Review, vol. 49, no. 4 (September):655–669. Venkatesh, P. C. 1989. “The Impact of Dividend Initiation on the Information Content of Earnings Announcements and Returns Volatility,” Journal of Business, vol. 62, no. 2 (April):175–197. Proceedings of the 2012 Pennsylvania Economic Association Conference 96 Proceedings of the 2012 Pennsylvania Economic Association Conference 97 Proceedings of the 2012 Pennsylvania Economic Association Conference 98 Proceedings of the 2012 Pennsylvania Economic Association Conference 99 Proceedings of the 2012 Pennsylvania Economic Association Conference 100 THE BUREAU OF MOTOR FUEL TAXES COMPLIANCE STRATEGY: PENNSYLVANIA DEPARTMENT OF REVENUE Thomas O. Armstrong* Daniel Meuser James Dehnert Nic Banting Commonwealth of Pennsylvania Department of Revenue, Harrisburg, PA 17128 ABSTRACT In August 2011, the Pennsylvania Governor’s Transportation Advisory Commission’s Final Report stated that Pennsylvania’s transportation infrastructure is in need of repair and reinvestment. The Motor License Fund provides state financial resources for the transportation infrastructure, primarily for the Commonwealth’s highway program. The Fund’s revenues consist of fuel taxes, license and registration fees, fines and penalities. The Motor License Fund, budgeted at $2.4 billion for FY 2012-13, is projected a nominal increase on average for five years of only 1.05%, where the Report argues for initiatives to enhance Motor License Fund revenue collections. Initiatives for improving tax compliance which enhance revenues without increasing the current tax burden on compliant taxpayers can advance economic efficiency. The Bureau of Motor Fuel Taxes has proposed the Compliance Strategy which will provide the technological infrastructure to increase revenue, reduce costs, and advance tax adminstration of the Commonwealth’s motor fuel laws. The Strategy focuses on automation of tax registration, processing and risk management, while educating the taxpayer base of their responsibilities. The Strategy, consisting of 11 components, is projected to take about five years to implement with a significant return on investment. Evasion of motor fuel taxes is an economic problem necessitating the implementation of the Compliance Strategy to reduce evasion and provide added revenue to meet on-going transportation challenges. I. INTRODUCTION The optimal taxation literature argues that tax reform is of critical economic importance where one of its objectives is tax simplicity (Armstrong, 2002). Tax simplicity is the constructing of tax systems with relatively low compliance costs and administrative burdens which elicit a high degree of voluntary compliance, therefore a low degree of tax evasion (Armstrong and Brehouse, 2001). Motor fuels tax evasion occurs when businesses deliberately fail to comply with their tax obligations (Franzoni, 1999). Fuel tax evasion becomes greater due to the complexity of the fuel distribution system and differences in state taxes imposed upon fuel. The greater the tax evasion the less financial resoures supporting Pennsylvania’s transporation infrastructure in need of repair and reinvestment (Transportation Funding Advisory Commission, August 2011). Generally, the motor fuel tax is an indirect tax that is collected from citizens by vendors selling or distributing the fuel. From Denison and Eger (March/April 2000), most fuel tax evasion occurs after the tax is collected from individuals and before the tax is remitted to the state. Figure 1 provides the key elements of the petroleum process. Crude petroleum is produced domestically or imported to the refinery. Pennsylvania has minor reserves of crude oil. Although Pennsylvania is credited with drilling the first commercial oil well in 1859, the State’s current production is minimal, with output derived primarily from stripper wells that produce less than 10 barrels per day (Department of Community and Economic Development, October 22, 2009). Most petroleum is imported into Pennsylvania which processes the crude into diesel and gasoline products. Pennsylvania is the leading petroleum refining State in the Northeast. Pennsylvania’s large-scale petroleum refineries are located along the Delaware River near Philadelphia and process crude oil shipped from overseas. These refineries supply the Northeast markets. Proceedings of the 2012 Pennsylvania Economic Association Conference 101 Figure 1: Petroleum Distribution System Domestically Produced Crude Oil Imported Crude Oil Refinery Imported Fuel Bulk Terminals at the "rack" Exported Fuel Distributor Wholesaler Bulk Stations Retailer Consumers: Private, Public Source: Denison and Eger (March/April 2000) The rack is the point of first distibution of the refined fuel. The rack is commonly associated with the terminals of the manufacturer, refiner, or producers. The rack is a mechanism of delivering fuel into a means of transport other than by pipeline or vessel. From the rack, the refined petroleum is sold to the distributor wholesaler or exported. Wholesalers sell fuel to retailers where it is sold to consumers. The fuel tax can be levied at several points in the distribution process, with the tax incidence being passed down the chain to the end consumer (Denison and Eger, March/April 2000). Fuel tax is generally collected and reported at one of three points in the distribution chain: terminal rack/import, wholesale level, and retail level. The fuel tax is administered like a sales tax where the wholesaler in Pennsylvania is responsible for collecting and remitting the tax to the government. Pennsylvania’s liquid fuels and fuels tax rates are the gas tax at 31.2 cents per gallon and diesel tax at 38.1 cents per gallon. For fuel taxes, there are users who are exempt from tax, and there are specific uses that are exempt. For Pennsylvania, there are tax exemptions for political subdivisions when such liquid fuels and fuels are delivered to this Commonwealth, a political subdivision, a volunteer fire company, a volunteer ambulance service, a volunteer rescue squad, a second class county port authority or a nonpublic school not operated for profit (75 Pa.C.S. § 9004(e)(4)). A political subdivision is any county, city, borough, incorporated town, township, school district, vocational school district and county institution district (1 Pa.C.S. § 1991). As the point of taxation moves up the distribution chain, fuel tax exemption becomes more difficult to administer where refunds become more common for tax exempt fuel tax users (Weimer, et.al., 2008). Eliminating refunds for exempt uses substantially reduces the potential for refund fraud. Federal motor fuel taxes are collected at the rack. The IRS provides Terminal Control Numbers (TCN), if a terminal is registered as a part of the taxable fuel bulk delivery system. Currently, Pennsylvania has 700 registered wholesale distributor licensees and 68 registered TCNs (racks). Over the years, the point of taxation has moved up the distribution chain where the Federation of Tax Administrators (FTA) as of January 27, 2012, reported 23 states tax at the rack or first receiver below the rack.1 By imposing the tax at a higher point in the distribution chain offers a potential for decreasing fraud, mainly by reducing the number of taxpayers. Loss of motor fuel tax represents a significant loss of funding to every state, especially since each state Proceedings of the 2012 Pennsylvania Economic Association Conference 102 receives a share of federal-aid highway dollars based on their consumption. in Pennsylvania (federal tax 24.4 cents plus PA state tax 38.1 cents), this could mean $5,000 additional profit to the distributor. To understand the magnitude and significance of the evasion, it is necessary to understand the profit collected by the distributor. The profit on a gallon of fuel is usually a few cents. If taxes can be evaded, the profit can be as much as 45 cents per gallon higher (24.4 cents federal diesel tax per gallon plus 20 cents average State tax). Thus, one truckload of fuel, 8,000 gallons, could potentially yield about $3,600 in additional profits if both Federal and State diesel taxes are evaded (45 cents x 8,000 gallons). As mentioned in Pennsylvania, the tax per gallon of diesel is 38.1 cents and 31.2 cents per gallon of gasoline. Considering 62.5 cents diesel fuel tax per gallon Tax evasion incentives increase when states have differential tax rates. Table 1 provides gas and diesel tax rates of Pennsylvania and surrounding states. The Table shows Pennsylvania’s fuel rates are higher than the surrounding states. This results in Pennsylvania having greater difficulty to track fuel to its ultimate destination into the Commonwealth where the fuel would be taxed in one state but evaded for payment to Pennsylvania. Better tracking movements of fuel using electronic filing and payments will reduce evasion and enhance compliance. TABLE 1: STATE MOTOR FUEL TAX RATES State Gasoline* Diesel Fuel* Delaware 23.00 22.00 Maryland 23.50 24.25 New Jersey 14.50 17.50 New York 25.80 24.05 Ohio 28.00 28.00 Pennsylvania 31.20 38.10 Virginia 17.50 17.50 West Virginia 33.40 33.40 * Federation of Tax Administrators, January 2012. The purpose of the paper is to provide information on the proposed Bureau of Motor Fuel Taxes Compliance Strategy. The Compliance Strategy consists of 11 components. As each component is implemented, improved tax compliance will enhance motor fuel revenues over and above the nominal average increase of 1.5 percent over five years; thereby, helping to support Pennsylvania’s transportation resource needs. Section 2 provides a brief background of the Bureau of Motor Fuel Taxes. Section 3 details the 11 components of the Compliance Strategy where some components can be implemented administratively while other components require legislative statutory authority. The final Section concludes the paper. II. BUREAU OF MOTOR FUEL TAXES From Armstrong and Meuser (2011), the Bureau of Motor Fuel Taxes is responsible for licensing and bonding activities for applicants within registration guidelines for the various Motor Fuels and Motor Carrier Road Tax/International Fuel Tax Agreement (IFTA).2 Furthermore, the Bureau directs the collection programs for delinquent accounts for these various taxes. In addition, the Bureau conducts investigations/cursory audits of suspect files; files criminal complaints for violations of tax acts; and issues and files citations for infractions of tax laws. The Bureau is budgeted to collect $2.4 billion for FY 2012-13 for the Motor License Fund--one of the Commonwealth’s special funds. The Liquid Fuels and Fuels Tax, the state’s excise tax, is a permanent trust fund tax of 12 cents per gallon or a fractional part is imposed on all liquid fuels and fuels used or sold and Proceedings of the 2012 Pennsylvania Economic Association Conference 103 delivered by distributors in Pennsylvania. The tax is imposed on the ultimate consumer, but the distributor is liable for collecting and remitting the tax. Payments and reports are due on or before the 20th of the month following the month of tax collection. Returns filed annually are approximately 7,000 by paper (80%), e-TIDES (18%), or Telefile (2%).3 Payment for the Liquid Fuels and Fuels Tax are by credit card (2%), EFT (80%), or by paper check (18%).4 The Commonwealth received for FY 2010 was about $752 million. On the same tax return as the state excise tax, the Oil Company Franchise Tax is imposed on all taxable liquid fuels and fuels on a cents-pergallon equivalent basis and is remitted by distributors of liquid fuels and fuels. The rate is 19.2 cents-per-gallon on all liquid fuels and 26.1 cents-per-gallon on all fuels used or sold and delivered by distributors within the Commonwealth. Payments and reports are due from distributors on or before the 20th of the month following the month of fuel sales. Returns filed annually were about 750 by paper (75%), e-TIDES (23%), or Telefile (2%). Payment for the Oil Company Franchise Tax is by credit card (2%), EFT (80%) or by paper check (18%). The Commonwealth received for FY 2010 was about $1.3 billion. The Alternative Fuels Tax is imposed on alternative fuels used to propel vehicles on public highways. Alternative fuels include natural gas, compressed natural gas, liquid propane gas and liquefied petroleum gas, alcohols, gasoline-alcohol mixtures containing at least 85% alcohol by volume, hydrogen, hythane, electricity, and any other fuel not taxable as liquid fuels or fuels. Each alternative fuel is converted to a gasoline gallon equivalent where the basis of this conversion is statutorily set at 114,500 Btu. The tax rate applied to the gasoline gallon equivalent equals the current liquid fuels tax and oil company franchise tax applicable to one gallon of gasoline.5 Alternative fuels dealer-users are required to remit this tax. Reports and payments are due on or before the 20th of each month for fuel sold or used in the preceding month. Returns filed annually are approximately 960 for by paper (100%). Payment for the Alternative Fuels Tax is by EFT (80%) or by paper check (20%). The Commonwealth received for FY 2010 was about $550,000. Pennsylvania joined the International Fuel Tax Agreement (IFTA) effective January 1, 1996. This agreement provides for base state reporting of fuel taxes for operators of qualified motor vehicles used in interstate operations.6 Qualified motor vehicles operated in Pennsylvania intrastate activities only are subject to fuel taxation under the Motor Fuels Road Tax. The Motor Fuels Road Tax is equivalent to the rate per gallon in effect on liquid fuels, fuels, or alternative fuels, plus an oil company franchise tax component. IFTA payments and reports are due on or before the last day of April, July, October, and January for the quarter ending the last day of the preceding month. Motor Carriers Road Tax reports are filed annually. Almost 67,000 returns are filed annually by paper (100%). Payment is by EFT (10%) or by paper check (90%). The Commonwealth received revenues of over $42 million for 2010. Table 2 provides the Motor License Fund five year projections. For the current fiscal year, the FY 2012-13 is budgeted at $2.4 billion and is projected a nominal increase on average for five years of only 1.05%. Table 2: Motor License Fund Five Year Revenue Projections* Fiscal Year Revenues 2012-13 budget 2,433,560,000 2013-14 estimated 2,474,680,000 2014-15 estimated 2,493,140,000 2015-16 estimated 2,512,100,000 2016-17 estimated 2,537,620,000 Average Annual 1.05% *Governor's Executive Budget 2012-2013, C2.6 Proceedings of the 2012 Pennsylvania Economic Association Conference 104 In August 2011, the Pennsylvania Governor’s Transportation Advisory Commission’s Final Report stated that Pennsylvania’s transportation infrastructure is in need of repair and reinvestment where projected increases over time fall short of funding requirements. The Report argues for initiatives to enhance Motor License Fund revenue collections. The Bureau of Motor Fuel Taxes has proposed the Compliance Strategy which will provide the technological infrastructure to increase revenue, reduce costs, and advance tax adminstration of the Commonwealth’s motor fuel laws. III. of staff reductions and compliance service and enforcement enhancements, revenue agencies have considerable incentives to automate these processes through greater use of technology. From the Organisation for Economic Cooperation and Development, OECD, (January 28, 2009), the key benefits of accelerating the slow pace of technology innovations are the following: • • • • • COMPLIANCE STRATEGY For Pennsylvania to maximize receipt of motor fuel taxes, the Bureau of Motor Fuel Taxes Compliance Strategy has been developed to enhance enforcement resources and use those resources more efficiently. The underlying theme of the strategy is to compile data electronically into a single data source so that instantaneous verification and validation that taxes have been paid can occur. Paper-based processes associated with tax returns and payments processing have consumed significant usage of a revenue agency resources (Armstrong and Meuser, 2012). With an economic climate Faster collection of government revenue Improved data accuracy and elimination of reverse workflows Reduced paperwork for taxpayers Faster crediting of tax refunds Faster capture of taxpayer data for a range of administrative purposes. The data currently required by the Bureau of Motor Fuel Taxes due to technology and process limitations is not to be maximized for tax collection efforts. Therefore, a compliance gap will exist for fuel taxes where, on average, it is estimated at six percent at all state levels.7 Figure 2 below provides a description of the fuel landscape to verify purchases and cross match into a single data source. Figure 2: Complete Verification of Fuel Landscape Distributor Sales Distributor Purchases Other States Imports The fuel landscape, shown in Figure 2, highlights all the sources of fuel transactions to which data can be verified. The flow of fuel from the terminals at the rack to retail, in-state to out-of-state, from data reported to data gathered through inspection and investigation. The sources of fuel transaction can be linked to Figure 1, the Petroleum Distribution System. The Compliance Strategy focuses on consolidating this data into a single data source for detailed analysis. Transporter Report Single Data Source FuelCAP Case Terminal Operator Reports Fuel transactions can be traced from the terminal (rack) all the way to consumption using the BOL (Bill of Lading) number. The BOL number remains consistent no matter who handles the fuel. The goal of the compliance strategy is to capture the fuel transactions from all those that handle the fuel and then reconcile their records. For instance, a load of fuel will be reported on a distributors disbursement schedule when it is sold, the same load will be reported on the buyers receipt schedule and then again on the transporters monthly report. By capturing and Proceedings of the 2012 Pennsylvania Economic Association Conference 105 storing electronically all this data it will make it easier to identify unreported loads of fuel. Ultimately the goal is to increase the Motor Fund annual revenue by cost reductions which are easily measurable and will be defined throughout the document or revenue increases which can be realized by several methods and approaches: • • • • • • • • Increase compliant revenue Collect outstanding delinquent liabilities Identify true delinquencies through accurate audit and examination Reduce administrative costs Reduce enforcement costs Reduce the number of cases that go through the appeal/legal process Reduce the number of taxpayers Preventative enforcement The Motor Fuel Tax Compliance Strategy is comprised of 11 components each designed to lower administrative costs or increase motor fuels revenue: A. Mandate Liquid Fuels and Fuels Tax EFile B. Distributor Classification Clarification C. Reduce Mandatory E- Payment Threshold from $20,000 to $500 D. Increase LFFT Bond Based on Taxpayer Risk E. Institute IFTA bonding based on Taxpayer Risk F. Operation AAA (Avoid Audit Assessment) G. Automated LFFT Audits/Examination CROSS-MATCHING H. Mandating E-File Transporter Reports I. Mandating IFTA E-File J. Online licensing – LFFT, IFTA, Transporters, Alt Fuels (includes refund applications) K. Move Point of Taxation to the rack Below is a brief description of each of the compliance strategies as proposed by the Department of Revenue’s Bureau of Motor Fuel Taxes. A. Mandate Liquid Fuels and Fuels Tax E-file One of the methods to evade motor fuel taxes is the filing of false information. Paper filing is not only costly to administer, but is more difficult to detect fuel tax evasion due to false information provisions. Currently, approximately 30% out of the 700 required monthly filers utilize the Department’s E-Tides filing system. This mandate would take that number 100%. It would no longer be acceptable for taxpayers to file paper tax returns that often exceed thousands of pages in transactions. Many of the large distributors have avoided sending their tax returns electronically to Pennsylvania despite doing the same for the feds and other states. It is unlikely to cause resistance but merely close a loophole for the Bureau. The data would be uploaded to the Department and then easily accessible to the Department’s examiners and auditors. Weimer, et.al., (2008) report that many states have moved to require all fuel taxpayers file their returns electronically. Currently the business process does not allow the release of the necessary funds from Revenue to the Pennsylvania Department of Transportation (PennDOT) until the tax return and payment are processed. This can be lengthy when an electronic payment is made but the paper tax return is waiting to be processed. This process can take approximately 30 days; however, if both electronic payment and electronic tax return are received timely this would only take 48 hours. All Liquid Fuel and Fuels Tax (LFFT) transactional tax data will now be electronically captured, detailed line items will be available to easily examine and queried. This will now give the Bureau of Motor Fuel Taxes the platform by which to provide detailed analysis while increasing the efficiency of our auditors and examiners. In addition, this may also cause voluntary compliance revenue increases under the premise that all data is now instantly accessible to auditors and data accuracy can now be verified on a large scale. There is a projected $100,000 per year cost savings as a result of: • • • • • Identify true delinquencies through accurate audit and examination, Reduce administrative costs, Reduce enforcement costs, Reduce the number of cases that go through the appeal/legal process, Preventative enforcement, Proceedings of the 2012 Pennsylvania Economic Association Conference 106 • • Accurate PennDOT cost analysis, and Reduce audit costs. B. Distributor Classification Clarification One method of reducing tax evasion is to reduce the number of taxpayers responsible for remitting the tax which permits more thorough audits without increasing auditing resources (Denison and Eger, March/April 2000). There are approximately 600 class 1-5 distributors. These companies are permitted to buy tax free fuel and sell to other registered distributors. Tax evasion most frequently occurs when a distributor purchases tax free fuel and collects the tax on the sale but fails to accurately report it to the department. If the current classification language was clarified and rewritten to exclude many unqualified existing companies; then by having less companies in this category, tax evasion can be reduced. The form that determines Distributor eligibility is Rev 543 and has not been revised since 2004. This form is policy set forth by the Secretary of Revenue. If the department adopted a less subjective approach to issuing class 1 licenses and reduced number of filers by 20%, the risk of tax evasion would be significantly lowered. The Bureau conducts approximately 100 pre-license investigations each year on new companies applying for a license. It is estimated that each pre-license investigation cost’s $1,000 in enforcement and administrative resources. Existing guidelines for Distributor classification are ambiguous and lead to companies filing multiple applications and cause an unnecessarily long licensing process and increased costs. Taxpayer service and resolution is also estimated at $1,000 per year in resource costs. By reducing the number of filer’s employees would spend less time servicing, examining and auditing these companies. The auditing department would increase its percentage of filers that are audited each year. If accounts were reduced from 600 to 500 the department would realize $100,000 per year cost savings as a result of: • • • Reduction in number of taxpayers, Identify true delinquencies through accurate audit and examination, Reduce administrative costs, • • • Reduce enforcement costs, Reduce the number of cases that go through the appeal/legal process, and Preventative enforcement. C. Reduce Mandatory E-Payment Threshold from $20,000 to $500 As stated previously, the Department of Revenue only releases the funds to PennDOT when both the tax return and payment have been processed. Paper checks take longer to process and can often delay the LFFT funding process. Many hours are spent researching illegible and incomplete checks. Checks sent in error often require taxpayer calls and hours wasted with resubmission processes. In addition, the risk of payments being lost, moved or incorrectly applied is greatly reduced if the payments are electronic. The cost of storage and processing is lower with electronic payments than paper checks. This initiative will reduce mandatory electronic payment threshold from $20,000 to $500. Currently, about 70% of LFFT payments are made via EFT (ACD Debit or ACH Credit) – by moving this threshold it would be far closer to 100%. The result is almost 100% of payments would be processed 48 hours after the timely due date. The projected savings is about $5,000 per year as the result of the following: • • • Identify true delinquencies through accurate audit and examination, Reduce administrative costs, and Reduce enforcement costs. D. LFFT - Increase Bond Based on Taxpayer Risk States impose bonding requirements on the wholesaler to provide additional protection against fuel tax fraud. Bonding enhances protection where bonds in most states are at least equal to the value of three months of their tax remittance due. As mentioned by Denison and Eger (March/April 2000), three months is ordinary sufficient time for detection of fradulent activity. In Pennsylvania, a bond amount is calculated by taking the number of gallons sold x tax rate x 1.5. For new companies estimated gallons are used and then recalculated each March. In the Proceedings of the 2012 Pennsylvania Economic Association Conference 107 event that a distributor defaults on a tax payment or is caught in fraudulent activity, under common practice it takes a minimum of three months before a license is suspended or cancelled. By changing the bond calculation from 1.5 to 3, the Commonwealth’s interests are further protected. Now, taxpayer risk can be defined as likelihood that the taxpayer will result in an account receivable due to inability to pay or tax evasion. High risk accounts share consistent traits and can usually be identified at the time of licensing.8 Only companies with low risk of evasion or default will have the privilege of buying tax free fuel. This will reduce the opportunities for companies to evade tax by paying their tax owed upon purchase. E. IFTA – Institute Bonding Based on Taxpayer Risk Unlike LFFT, bonding is not currently implemented for the IFTA; however, the statute allows for it. Currently, a loophole in the law allows a company to run up liabilities under their Employee Identification Number (EIN), refuse to pay, and then reopen a new account under the same name, owner, address but different EIN – typically a wife’s, child’s or other relative. Qualified accounts and those who refuse to adhere to the laws would be identified as high risk; thereby, requiring a bond based on taxpayer risk. The Commonwealth’s interests would be protected by the number of dubious accounts being reduced. If a company becomes fraudulent and unresponsive, the bond will be recalled to cover outstanding tax liabilities. If no bonds are instituted, the Commonwealth continues to run the risk of evasion and increased accounts receivables. This concept would not change the licensing process for all filers; however if it was discovered upon licensing that the owner, address, or company has a previous liability with the Bureau, a bond would be exercised due to the obvious risk. Further risk calculations could include fleet size, non-compliance with other states, record keeping compliance and outstanding liens against owners. The expected increase in revenue is projected at $1 million per year attributed to: • • • • • • Identify true delinquencies through accurate audit and examination, Reduce administrative costs, Reduce enforcement costs, Reduce the number of cases that go through the appeal/legal process, Preventative enforcement, and Reduce accounts receivables. F. Operation AAA (Avoid Audit Assessment - Enforcement & Education) Nagan (1990) proposes that one of the methods to reduce tax noncompliance is to increase visibility. Increasing visibility can occur with enhancing education and outreach to motor fuel taxpayers. Currently, there are 1,200 new accounts being created and 1,200 accounts cancelled each year. It is very important to ensure that all tax records are accurately kept by all those susceptible to audit. Information needs to be provided to taxpayers including what data is required and when it must be filed. Many of the audit findings are reoccurring and can be avoided with effective education and enforcement by the Bureau of Motor Fuel Taxes staff. The purpose of Operation AAA is to educate high audit risk accounts and heavily enforce the records keeping requirements of the Bureau. From the Bureau’s perspective, understanding the taxpayer base means that the Bureau will have to condense and simplify our record keeping requirement instructions. In addition, Motor Fuels must also have a comprehensive system of calling and visiting high audit risk accounts. The final piece is visiting already audited companies to ensure compliance to avoid repeat audit. At present in excess of 95% of the IFTA accounts receivable listing is a result of IFTA audits. By reducing the IFTA accounts receivables, the Bureau will be able to allocate more time to preventative enforcement. Furthermore, by showing a history of education and enforcement with the taxpayer if the audit was to come under appeal, there would be opportunity for the company to claim they were not properly informed of their tax responsibilities. Proceedings of the 2012 Pennsylvania Economic Association Conference 108 Operation AAA is expected to gain about $150,000 per year as the result of the following: • • • • • Identify true delinquencies through accurate audit and examination, Reduce the number of cases that go through the appeal/legal process, Preventative enforcement, Reduce accounts receivables, and Citation Issuance. G. Automated LFFT Computer Audits/Examination CROSSMATCHING Weimer, et. al., (2008) report that most states agree that all information should be filed electronically for cross checking of fuel reports. Currently, less than 1% of transactions are verified between distributors on a monthly basis. By moving towards an automated crossmatch system, all transactional disparities will be instantly recognizable. This will not only generate accurate liabilities but will indirectly increase the accuracy of filers who will soon learn that every transaction is being scrutinized. This component exponentially increases our ability to identify that all taxes have been paid. It will give a holistic view of all the fuel transactions in Pennsylvania that can be verified by independent sources. This is the single most important element of the Compliance Strategy. This is contingent on 100% LFFT E-file (A) and 100% Transporter Efile (I) and is not possible until that has been completed. Once the tax data is stored electronically, three main types of evasion can be investigated on an automatic and uniform level: 1. 2. 3. Sales between distributors can be reconciled, i.e. receipt on one taxpayer schedule is a disbursement on another (basically an audit). Imports can be reconciled with the schedule data received from other states.9 Data gathered at the retail level via Fuel Composition and Accountability Program (FuelCAP) cases thereby can be reconciled against the distributor sales. According to a leading fuel tax system company (ACS) which implements a comparable system ‘Vista’; “The Michigan Department of Treasury implemented VISTA/FT in 2003 with the goal of increasing motor fuel tax collections by $4M. The system provided the state with tools which enabled them to exceed their prediction of collection by more than $20M in increased revenues. Their 2004 goal was $12.1M and they realized more than $20M in increased collections, and an additional $11M in compliance improvements. Michigan’s published results include increased revenue of $32.5M in the first year of operation, with an additional $15.9M obtained in the second year. As of the end of 2009 Michigan’s Return on Investment (ROI) has been well over $25 to $1.” Once implemented, it is expected to generate over $50 million over 2 years. This is the result of: • • • • • • • • • Increase in compliant revenue, Increase delinquent collections, Reduce accounts receivable, Identify true delinquencies through accurate audit and examination, Reduce administrative costs, Reduce enforcement costs, Reduce the number of cases that go through the appeal/legal process, Preventative enforcement, and Reduce audit costs. H. Mandate E-file IFTA IFTA requires trucks operating in interstate commerce to report their mileage in all member jurisdictions to their home state or province (hereafter referred to as states). They also report all tax paid on fuel by state. The tax due to each state is calculated and compared to the tax paid to each state and appropriate payments or refunds by state are made (Weimer, et. al., 2008). With approximately 11,000 filers, IFTA is the larger of the two taxpayer bases for Bureau administration responsibility. The administration process is 100% paper filing and paper payments. This initiative would implement a new submissions system and process enabling the user to input the states traveled in, the miles and the mpg and tax due would be calculated. The system would also enable online payment through ACH debit, credit or credit card. If a refund was due then it Proceedings of the 2012 Pennsylvania Economic Association Conference 109 could be direct deposited into the taxpayer’s account. IFTA only contributes less than 2% of the overall income for the Bureau however it absorbs approximately 70% of the Bureau’s administrative and enforcement resources. By moving towards electronic submission many of these resources could be deployed towards Liquid Fuels and Fuels Tax operations which contribute over 97% of the Bureau’s revenue source. Mandating e-filing of IFTA returns and payments would reduce administration and enforcement costs and provide accurate data for reporting, thereby automated examination and audits would then be possible. A significant stakeholder outreach program would be necessary to ensure accurate compliance. The expected increase in revenues as the result of the e-file mandate corresponding with an electronic submission platform is about $100,000 per year. The expected increase can be attributed to the following: • • • • • • • • Reduce enforcement costs, Reduce administrative costs, Identify true delinquencies through accurate audit and examination, Reduce the number of cases that go through the appeal/legal process, Preventative enforcement, Reduce accounts receivables, Citation Issuance, Increase compliant revenue, and Increase delinquent collections. I. Mandate E-File Transporter Reports • There are approximately 600 Fuel Transporters registered in Pennsylvania. These reports list all the fuel transported in and out of Pennsylvania listing both the buyer and seller. Many of department’s resources are consumed by the manual administrative processes ensuring that the 600 companies file on time and keep their account information up to date. After the mandate is complete and automation is deployed, the resources can be focused on analyzing data discrepancies. This is often the missing link in many fuel tax investigations. The line item data is not currently captured however would be essential to the FuelCAP program, computerized audits, and examination. FuelCAP (Fuel Accountability and Composition) program is an enforcement tool whereby the Bureau verifies timely and accurate submission of liquid fuel and fuel taxes in order to prevent evasion schemes and the associated resources needed in the recovery. This is achieved by gathering raw data at the retail/end user levels and following the physical trail of fuel back to suppliers to form a complete picture of all associated transactions. By mandating E-filing transporter reports, this data source would be used to reconcile fuel tax transactions to ensure the tax has been paid. By capturing all this data electronically and combined with component ‘G’ this will now become a key data source for independent verification that taxes were paid. The result of this initiative is an expected increase in revenues of about $500,000 per year. The gains would include the following: • • • • • • • • • J. Reduce enforcement costs, Reduce administrative costs, Identify true delinquencies through accurate audit and examination, Reduce the number of cases that go through the appeal/legal process, Preventative enforcement, Reduce accounts receivables, Citation Issuance, Increase compliant revenue, and Increase delinquent collections. Online Licensing (LFFT, ALT Fuels, Transporters and IFTA) This component of the Compliance Strategy would consist of a public facing interface whereby the taxpayer can fill in an application for a license or refund online for Liquid Fuel and Fuels Tax (LFFT), Alternative Fuels (ALT Fuels), transporters and IFTA. Currently, 100% of applications for licenses and refunds are still received as paper forms. If this data would be received in an electronic format, security and completeness of checks can be automated. Error checks would ensure that data is valid therefore increasing the accuracy of enforcement efforts. In addition, not only would the department be able to reduce costs via automation, the taxpayer would see the turnaround time for a refund or a license application reduced. Clerical and administrative resources can be reassigned to Proceedings of the 2012 Pennsylvania Economic Association Conference 110 more in depth analysis of the newly captured data. Clean, fast accurate data – fewer wasted mailings, expedited license issuance and increased customer service. and the oil company franchise tax, which at the present time is 19.2 cents per gallon.10 • It is expected once the online licensing component has been implemented, approximately $100,000 per year is expected in increased revenues. In addition, other additional impacts are the following: • • • Reduce administrative costs, Reduce enforcement costs, and Increase delinquent collections. K. Move Point of Taxation to the Rack The incentive for tax evasion increases when competitors can undercut prices and attract customers from compliant competitors. This tax evasion is more likely to occur when fuel tax is levied at the wholesale distribution or retail level (Denison and Egar, March/April 2000). The Department of Revenue collects taxes at the wholesale distributor level. The point of taxation for both liquid fuels and fuels is on the sale of product from a registered distributor to a nonregistered distributor. Imposing the liquid fuel tax at a higher point in the distribution chain, to the terminal, or “rack” as it is commonly referred to, will offer the potential of reducing instances of fraud, mainly by decreasing the number of taxpayers to collect fuel taxes by the Bureau. Weimer, et. al. (2008) mentions taxing at the terminal rack is generally widely accepted as one key measure a government can take towards increasing motor fuel excise compliance. This component of the legislative initiative would move the liquid fuel and fuels point of taxation from the registered distributor level to the terminal, or rack, which is a higher point in the fuel distribution system. The rack is a mechanism of delivering fuel into a means of transport other than pipeline or vessel. This changes the imposition and collection of taxes from the wholesale level to the rack, resulting in the reduction of collection points. Collecting the tax at the rack does not change the motor fuel tax rates, only how and where it is collected. Pennsylvania’s liquid fuels and fuels tax rates are: • Gas tax is 31.2 cents per gallon, which is the sum of the fixed liquid fuels tax of 12 cents per gallon (state excise tax), Diesel tax is 38.1 cents per gallon, which is the sum of the fixed fuels tax of 12 cents per gallon (state excise tax), and the oil company franchise tax, which at present time is 26.1 cents per gallon. At the federal level, the point of taxation for gasoline was moved to the terminal rack in 1986. In 1990, administrative regulations were tightened requiring the gas tax imposition at the point of the import, removal from the terminal or refinery, or the point of sale of any unregistered entity. The point of taxation was also moved to the rack at the federal level for diesel in 1994. The IRS provides Terminal Control Numbers (TCN), if a terminal is registered as a part of the taxable fuel bulk delivery system. Currently, Pennsylvania has 700 registered wholesale distributor licensees and 68 registered TCNs (racks). Taxing at the rack means the tax is due from the consumer, but it is remitted to Pennsylvania by the supplier, or in rare cases, by the distributor. The tax is pre-collected by the supplier before going all the way down the supply chain. Imposing the liquid fuel tax at a higher point in the distribution chain offers a potential for eliminating fraud, mainly by reducing the number of taxpayers. Loss of motor fuel tax represents a significant loss of funding to every state, especially since each state receives a share of federal-aid highway dollars based on their consumption. Weimer, et.al., (2008) report that many states have seen increased revenues after moving the point of taxation up the fuel supply chain. All purchases from the rack (in Pennsylvania) would include the Pennsylvania tax. Weimer, et.al., (2008) state that taxation at the rack makes the most sense for states, including Pennsylvania, with refineries and terminals. In instances where gallons were exported and reported to another state, Pennsylvania would refund the tax paid. Additionally, political subdivisions, currently exempt under the current statute such as school districts, municipalities, and townships would be exempt from paying the tax or included in the refund process. By Proceedings of the 2012 Pennsylvania Economic Association Conference 111 exempting users of fuel; and thus not having to pay the tax, reduces the potential for refund fraud Weimer, et.al., (2008).11 Following the passage of legislation, it is estimated that 24 months would afford the time necessary for the Department of Revenue and the Bureau of Motor Fuel Taxes to successfully implement such an extensive change to our motor fuel tax law.12 Therefore, if legislation on Title 75 of the Vehicle Code were passed by July of 2013, it is a fair assumption that the Department could be fully prepared to “go live” no later than July of 2015 with electronic reporting, electronic processing, and employee training/redeployment in place by that juncture. The resulting outcomes of moving the point of taxation to the rack is the following: • • • • • • • The Pennsylvania point of tax would be consistent with the point of taxation of the federal government, which is at the rack. Stakeholders already reporting to the federal government would already have those processes in place when reporting to Pennsylvania. Pennsylvania would see a decrease in our number of licensed distributors, and tax return processing would see a significant drop in bureau work hours. The tax examination and refund verification process could be automated The if all data was electronic.13 automation would require a new tax processing and submission engine. Mandate that all payments must be electronic and all refunds will be issued by electronic fund transfer.14 Mailing costs would decrease as the result of the elimination of mailing monthly tax returns to associated taxpayers.15 Licensing and bonding hours will be reduced. It is expected that an increase in revenues once all facets are fully implemented to be about $20 to $25 million annually, though quantifying such an increase is very difficult to assess. Additional expected results are the following: • Reduction in number of taxpayers, • • • • • IV. Identify true delinquencies through accurate audit and examination, Reduce administrative costs, Reduce enforcement costs, Reduce the number of cases that go through the appeal/legal process, and Preventative enforcement. CONCLUSION Denison and Egar (March/April 2000) classify four basic categories to evade motor fuel taxes: 1) failure to file information, 2) filing of false information, 3) filing of false exemptions, and 4) failure to pay assessed taxes. The Department of Revenue Bureau of Motor Fuel Taxes Compliance Strategy addresses the four categories by providing outreach and education to taxpayers to address their tax obligations; providing greater electronic methods for filing and payment along with cross-matching ability for the Bill of Lading sales reciepts; and moving the point of taxation at a higher level of the distribution chain to the rack. Components of the Compliance Strategy are under consideration by the administration. ENDNOTES * The authors would like to thank Sally Fishel and discussant for their assistance and comments. The conclusions do not necessarily reflect the positions of the Pennsylvania Department of Revenue. All possible errors are the author’s. 1. Based on the data reported by the FTA Motor Fuel Tax Uniformity Committee E-Commerce Subcommittee Survey, January 27, 2012; seven states point of taxation are position holder at the rack, nine states are exchange receiver at the rack, seven states are first receiver below the rack, four states are importation into the state (first receipt into storage), and 23 states point of taxation are at the distributor stage. 2. IFTA, International Fuel Tax Agreement, is an agreement between states that simplifies the reporting of fuel taxes by interstate haulers by establishing a uniform system for administering and collecting taxes. Motor carriers choose a base jurisdiction in which to register and file a single return with a single payment to their base jurisdiction. The base jurisdiction processes the IFTA return for net fuel taxes and forwards fund Proceedings of the 2012 Pennsylvania Economic Association Conference 112 to, or requests refunds from, each jurisdiction. By 1996, all 50 states and 9 Canadian provinces were IFTA members (Weimer, et.al., 2008). 3. E-TIDES, Electronic Tax Information and Data Exchange System, is an Internet filing system that allows electronic filing of returns, payments and/or extension requests. Motor Fuel Taxes, Corporation Taxes, Sales and Use Tax, and Employer Withholding Taxes use e-TIDES. The taxpayer can also purchase software to file online. Telefile is the process where taxpayers and businesses can file simple tax returns over the telephone. 4. EFT (electronic funds transfer) is a method of payment whereby the taxpayers will electronically pay from their bank account to settle their Commonwealth of Pennsylvania liability. This can be initiated by the taxpayer or Alternative Fuel Ethanol Methanol Propane/LPG E-85 M-85 Compressed Natural Gas (CNG) Liquefied Natural Gas (LNG) Electricity the Department depending on the agreed terms. Credit Card payments are taxpayer initiated form of electronic payment, the difference from EFT being that it is not directly linked to a bank account but instead the taxpayers credit card. 5. The point of taxation for the Alternative Fuels tax is at the retail or end-user level for alternative fuels. This is due to the fact that most alternative fuels have many uses other than as a fuel for propelling a vehicle on the public highways; therefore, until they are placed into a vehicle they do not qualify as an alternative fuel. Alternative fuels are taxed at the rate of the Commonwealth’s Liquid Fuels and Fuels Tax, plus the Oil Company Franchise Tax using a gasoline gallon equivalent calculation using a BTU conversion factor for each alternative fuel. The Department publishes revised tax rates each December for the following calendar year. 2012 Tax Rates for Alternative Fuels Amount Equivalent Rate of to One Gallon of Conversion Gasoline @ (BTU/gal of 114,500 BTU per alternative fuel) gallon 76,400 1.4999 56,560 2.024 83,500 1.371 80,460 1.423 65,350 1.752 Tax Rate per Gallon of Alternative Fuel $0.208 $0.154 $0.228 $0.219 $0.178 29,000@3,000 PSI 3.948 $0.079 66,640 1.718 $0.182 3,412BTU/KWH 33.558 KWH 0.0093/KWH 6. Qualified motor vehicles include those used, designed, or maintained for the transportation of persons or property which: a) have two axles and a registered or gross weight greater than 26,000 pounds, or b) have three or more axles regardless of weight, or c) are operated as a vehicle combination exceeding 26,000 pounds. 7. Currently, there is no tax gap analysis for Pennsylvania’s Motor Fuels taxes. 8. With over 100 new companies applying for licenses every year the Commonwealth is constantly exposed to threat of potential tax evasion. Identified factors for high risk taxpayers that increase the likelihood of tax evasion are the following: • Home state: If the company is out of state, enforcement and collection powers/efforts are substantially reduced. It is also very difficult to conduct a thorough pre-license interview over the phone as there would be no on site visit. • Years of residency: If the company is based in PA, how long have they been in existence? Companies new to PA that has no record of compliance are obviously a higher risk. • Value of company’s physical assets: If a company has zero physical assets, if a liability is incurred there will be nothing to lien or leverage. • Credit Score of applicant: Credit score is frequently used by insurance companies, banks and the majority of financial institutions to evaluate the risk of potential members. Proceedings of the 2012 Pennsylvania Economic Association Conference 113 • • • • • Principals–criminal activity in background check: The Department has the ability through the Pennsylvania Office of the Attorney General to perform background checks on the principals before licensing. The issue was particularly relevant when a Distributor’s principal ended up on the Interpol watch list after they left the country. Number of years in business: The high rate of attrition means new companies possess a higher risk of default. Outstanding liabilities in another state: Established liabilities maybe subjective without knowing all the circumstances but if activity is shown in another state then it would be prudent to check with our sister agency in that state for flagrant activity. Only retail distribution: Smaller distributers who are simply selling to a few retail outlets should not qualify for a license. Allowing these companies to purchase tax free fuel leaves us more susceptible to tax evasion. How many times have they applied? Companies who continually apply after multiple license denials should be flagged as potential high risk account due to the constant manipulation of data on their application. 9. Bootlegging is fuel purchased in a state with a low-tax rate, such as New Jersey at $0.145 without filling out the proper export documentation; it is then exported to a border state with a higher rate, Pennsylvania at $0.312, and sold at retail stations without remitting the tax in the high-tax state. Import and export information can be analyzed more efficiently with electronic reporting. Furthermore, false claim of exports can be detected through fuel tracking systems with electronic reporting. 10. The total state and federal taxes per gallon of gasoline in Pennsylvania is the Liquid Fuels Tax at $0.12; Oil Company Franchise Tax at $0.192; Underground Storage Indemnification Fund at $0.011; and the Federal Excise Tax at $0.18. 11. Diesel fuel dyed red is not subject to highway use tax. If used on public roads and did not pay the federal and state diesel fuel tax, tax evasion is the result. Fuel to be used for exempt or off-road use is currently dyed red per IRS standards. The dye is usually added at the terminal with dye injectors that should be temper proof. 12. Rack legislation would require an overhaul of Title 75 of the Vehicle Code, Chapter 90, perhaps as much as 50% of the current statutory revisions would either need to be amended or replaced. Furthermore, regulations will be reviewed and possibly repealed or replaced in order to be consistent with the new law. Other related chapters of Title 75 will be reviewed in order to ensure no incompatibility issues. For example, enforcement statutes may or may not need review and revision such as Chapter 94. 13. Daisy chain tax evasion schemes occur where several fallacious purchases of fuel occur without remitting tax payments where one of dummy companies purchasing the fuel, known as the burn company, dissolves along with any tax liability. Weimer, et. al., (2008) report that states that have moved the point of taxation to the rack deter this scheme. 14. Moving the point of taxation to the rack may result in fuel previously not taxed would now be taxed and subject to a refund. The Title 75 statutory changes would attempt to maintain exemptions within the statute. 15. Pennsylvania’s current fuel tax is on the distributor; thus payment to the state is delayed, allowing for interest free use of the funds, known as the float. Title 75 changes would address this issue to maintain the float; thereby not increasing costs to the distributors. Proceedings of the 2012 Pennsylvania Economic Association Conference 115 REFERENCES Armstrong, Thomas O. and Daniel Meuser 2011. "Increasing Innovation in Tax Administration Collections: Pennsylvania Department of Revenue," Pennsylvania Economic Association Proceedings, 80-92. Armstrong, Thomas. Proceedings. 1-10. 2002. “State Taxation Reform Proposals for Pennsylvania,” Pennsylvania Economic Association IBID and Jason R. Brehouse. 2001. "State Tax Simplification: State Nuisance and Obsolete Provision Tax Repeals, Including Proper Placement within the Tax Reform Code," Pennsylvania Economic Association Proceedings, 171-175. Department of Community and Economic Development, Commonwealth of Pennsylvania. October 22, 2009. “Pennsylvania’s Initiative for Energy and the Environment Plan for Support and Investment,” Harrisburg, PA. Denison, Dwight V. and Robert J. Eger III. March/April 2000. “Tax Evasion from a Policy Perspective: The Case of the Motor Fuels Tax,” Public Administration Review. 60;2, 163-172. Franzoni, Luigi Alverto. 1999. “Tax Evasion and Tax Compliance,” Encyclopedia of Law & Economics. Edward Elgar and the University of Ghent. OECD. January 28, 2009. Forum on Tax Administration, “Tax Administration in OECD and Selected Non-OECD Countries: Comparative Information Series”, 1-215. Pennsylvania Office of the Budget. February 7, 2012. “Governor’s Executive Budget 2012-13,” Harrisburg, PA. Transportation Funding Advisory Commission-Final Report. August 2011. Pennsylvania Governor’s Transportation Funding Advisory Commission. Weimer, Mark; Balducci, Patrick; Fathelrahman, Eihab; Whitmore, Susan; and Anthony Rufolo. 2008. “National Cooperative Highway Research Program: Identifying and Quantifying Rates of State Motor Fuel Tax Evasion,” Transportation Research Board, Washington, D.C. Proceedings of the 2012 Pennsylvania Economic Association Conference 116 A PANEL STUDY ANALYSIS OF ECONOMIC GROWTH IN SOUTH EAST ASIA Tai McNaughton Clarion University of Pennsylvania Clarion, PA 16214 ABSTRACT Identifying factors relevant to economic growth has been the centerpiece of economic development and studies in economic history for decades. Rostow (1960) represents one of the first authors to develop a progressive discussion of how economic growth occurs in all economies by identifying five stages of growth. These views sparked a great deal of research in economic development to determine underlying factors influencing economic growth. This paper will look at several economies in Southeast Asia simultaneously to identify the variables that have been determining economic development in the region. By identifying these variables, future growth can be extrapolated. ] Theories of how to achieve growth have differed throughout the years. An export led growth hypothesis is one of the ways that is said to create regional growth. The economies of South East Asia are well known for using a growth policy that focuses on export expansion. Furthermore, these countries also show a lot of potential for growth. They could even have the underlying factors to become just as advanced as Japan. Therefore, it is very crucial that these countries grow to reach their full potential. However, literature varies on the actual effectiveness of export expansion policies. This study attempts to determine the variables that determine growth in the economies of Cambodia, Indonesia, Laos, Malaysia, Philippines, Singapore, Thailand, and Vietnam. This study shows that increases in Foreign Direct Investment, Gross Capital Investment, and Life Expectancy will result in an increase in the standard of living across all of the countries is South East Asia. The volume of Exports of Goods and Services is statistically insignificant in describing growth while Manufacturing has a negative relationship with growth. foreign direct investment in Southeast Asia and other developing countries over the past one hundred years. FDI is said to be desirable by other countries because it not only is a growth enhancer due to capital accumulation, but it also helps boost productivity and output growth. Hsiao (2006) concluded after looking at the results of other studies that there are positive effects of “FDI on transitional and long-run economic growth through capital accumulation and technical or knowledge transfers” (1083). An article by Thomsen (1999) confirms this as he agrees that FDI leads export oriented growth. He also suggests that FDI should be judged by quality instead of quantity. However, Hsiao (2006) points out that FDI inflows could be attracted to growing economies and therefore FDI and economic growth can run bi-directionally. Manufacturing According to the article by Rasiah and Hing (2006), attempts to analyze the effects of industrialization on growth could not be completed thoroughly without breaking down the different forms of industrialization to actually show what helps growth the most. According to an article by Jongwanich (2010), within a country it is important to distinguish the difference in exports to tell how much growth will occur. Jonhwanich (2010) breaks exports down into “total merchandise exports, manufacturing exports, and exports of machinery and transport equipment” (21). This allows for comparison among the different types of exports. A study completed in 1967 showed that in manufacturing there was a positive and strong elasticity of change with growth (Rasiah, 2009). Due to government intervention, manufacturing soon became a huge industry across the resource rich Southeast Asia countries. Rasiah and Hing (2009) also make it a point to label the countries of Southeast Asia with transitional economies. LITERATURE REVIEW Exports of Goods and Services Foreign Direct Investment According to Rasiah and Hing (2009), foreign ownership led export-oriented industrialization in Southeast Asia. In the basic understandings of economics, it is said that industrialization drives increasing returns. Much like Rasiah and Hing’s article that dealt with foreign direct investment (FDI), an article by de Mello Jr and Fukasku (2000) examined the actual relationship between foreign direct investment and foreign trade. They detail the huge increase in An article by Trindade (2005) shows the effects of exportoriented growth and export enhancing policies in two of the countries in my sample of eight. In an article by Ghirmay, Grabowski and Sharma (2001), the way growth and exports are connected were analyzed. They stated that there have been many different studies where the goal was to figure out whether export expansion leads to growth or how exports affect economic growth. According to Ghirmay, Grabowski, and Sharma (2001), a part of economic theory leads to the Proceedings of the 2012 Pennsylvania Economic Association Conference 117 belief that “export expansion is believed to promote economic growth via two paths: by improving efficiency in the allocation of productive resources and by increasing the volume of productive resources through capital accumulation” (689). They concluded that a time series study of a country by country approach would be more effective than a study of how growth is effected across different countries. There is a casual effect between export growth and output growth by looking at studies where growth is regressed on exports. A lot of studies have been carried out to prove the belief that export growth improves economic growth by improving economic efficiency (Ghirmay, 2001). This has been one of the major underlying beliefs for trade promotion. However, some economists differ from this belief and instead feel that there has to be a minimum level of development before benefits from trade are actually shown to help the economy. This is where Ghirmay (2001) comes to the conclusion that exports are leading to accumulation rather than actually improving the efficiency in an economy. Life Expectancy and Education Booth (2004) feels that globalization and trade open a country up to the outside world and therefore improve the situation the country is in. Booth shows this by referencing the mass amounts of growth over the past 100 years in Southeast Asia and adds in the incredible increases in life expectancy rates and literacy rates. She then compares this to other countries that were just as undeveloped as these Asian countries. Thompsen (1999) confirms the importance of education as he talks about education in an investment in human capital sense. DATA AND METHODS A time series, panel study regression was run in TSP with the purpose to see if Economic Growth measured in PPP Converted GDP Per Capita at constant prices is dependent on Foreign Direct Investment as a percent of GDP, Real Gross Capital Formation, Average Life Expectancy at birth, Volume of exports of Goods and Services, and Manufacturing as a percent of GDP. Due to incomplete Education data it was not used in this study. The natural log of PPP converted GDP Per Capita as well as the natural log of Gross Capital Formation were also substituted in but produced the similar final results. All of these variables were collected for the South East Asian countries of Cambodia, Indonesia, Laos, Malaysia, Philippines, Singapore, Thailand, and Vietnam. The data for these countries begins in 1980. Myanmar and Brunei did not have enough data in order for those two countries to be included into the data analysis. The following is the regression model actually used: (1) Y (PERGDP) = β0 + β1 (FDI) + β2 (GROSS CAP FORMATION) + β3 (LIFEEXP) + β4 (VOLUME EXPORTS) + β5 (MANU) Where: Economic Growth (Y) = f(FDI, Gross Capital Formation, Life Expectancy, Volume of Exports, and Manufacturing) When using time series data across countries in a panel study, it must be determine whether differences across countries can be captured by differences in the constant term β0 (fixed effects approach) or whether the constant terms for each country are randomly distributed (random effects approach). This approach to panel estimation is outlined in Greene (2000). Table 1 lists the variables, how they were measured, their source, and the years they were collected for. The graphs included visually show the distributions of the data across the various countries. It should be noted that a few of the variables do have data that goes back further in time than 1980. While the years are not included in the study, it is important to show the long-term trend to help with the qualitative forecast located at the end. Graph 1 shows the Economic Growth levels, measured in GDP PPP per capita. Each of the eight countries’ data is represented. In comparison, Singapore has achieved a higher level of growth than all 7 other countries. Graph 2 shows the level of Foreign Direct Investment, net inflows. Singapore has a higher level of FDI percent of their GDP. There is no clear trend line across all of the countries. Graph 3 shows the volume of Savings or Gross Capital Formation volume. There is an overall increasing trend in gross capital formation across all countries. Graph 4 shows the variable for health across all countries in the region. The overall trend is an increase in life expectancy as time increases. Graph 5 shows the Percentage of GDP that comes from the Exports of Goods and Services. There is no clear trend in the region. Graph 6 shows the Percentage of GDP that comes from Manufacturing. There tends to be a trend of increasing Manufacturing percentage as time goes on with a recent decrease. EMPIRICAL RESULTS Table 2 is a correlation matrix for PPP converted GDP per capita. A correlation matrix shows the relationship between two variables. The closer a number is to positive 1, the more highly correlated the two variables are. Proceedings of the 2012 Pennsylvania Economic Association Conference 118 Life Expectancy and FDI are both highly correlated with PPP converted GDP per capita. Also, Manufacturing is expected to have a negative relationship with GDP per capita, PPP. Model 7 and Model 8 shows mostly the same pattern in results as Model 5 and Model 6.The only difference is that Volume of Exports of Goods and Services affects the Natural Log of GDP negatively. However, while possibly economically significant, this is not statistically significant. Regression Models CONCLUSIONS/POLICY RECCOMENDATIONS Table 3 shows the fixed effect results from the panel regressions. Each model has the coefficient with corresponding p-value below. In Model 1 and Model 2, the dependent variable is economic growth measured in PPP GDP per capita and Gross Capital Formation is at 2000 constant rates. In Model 3 and 4, the dependent variable is the natural log of PPP GDP per capita and Gross Capital Formation is measured with the natural log of 2000 constant rates. The Fixed Effects results showed interesting results. However, due to the results of the Hausman test, the Random Effects models better explain the development of the given region. Table 4 shows the Random Effects results from the panel regressions. Each model has the coefficient with corresponding p-value below. In Model 5 and Model 6, the dependent variable is economic growth measured in PPP GDP per capita and Gross Capital Formation is at 2000 constant rates. In Model 7 and 8, the dependent variable is the natural log of PPP GDP per capita and Gross Capital Formation is measured with the natural log of 2000 constant rates. In regression analysis, coefficients show how much of an effect that a given variable has on the dependent variable. The coefficient tells how much the dependent variable will increase with each increase in the independent variable, or conversely shows how much the dependent variable will decrease with each increase in the independent variable. These are denoted by positive coefficients and negative coefficients respectively. Also important in regression analysis is p-values. A p-value shows the probability that a correlation between two variables is indeed important instead of being random. If pvalues are low, the more probability that the independent variable has an effect on the dependent variable. Model 5 shows that Manufacturing affects the standard of living negatively. Also, the volume of exports of goods and services does now have an effect on the development. Therefore, Model 6 shows the results without the Exports of Goods and Services. An increase in Manufacturing still results in a decrease in the standard of living. All of the other variables are statistically significant and affect the standard of living positively. This study produced interesting results. First, the Volume of exports of goods and services was completely statistically insignificant. This goes against previous literature as well as theories supporting export expansion as a way of increasing growth. Also, each increase in the percent of GDP that comes from Manufacturing actually decreases the standard of living in the region. Not surprisingly, larger net inflows of FDI, Gross Capital Formation and Life Expectancy all affect the standard of living positively. A major caveat for this study would be the data. Limited data for several of the countries, to the point where they could not be included, was further exacerbated by the lack of reporting in the remaining countries. There was also a lack of Educational data for the region. Furthermore, there was heteroskedasticity that could not be corrected for. According to the results, if countries in this region want growth, policies toward increasing inflows of FDI, Gross Capital Formation, and Life Expectancy are all pertinent. By not focusing on export expansion in the region and instead concentrating on these proven growth enhancers will result in economic growth. Unfortunately, export expansion is currently believed to be one of the main growth enhancing strategies that is promoted. This was proven unsuccessful in the region and therefore will be limited in the future. QUALITATIVE FORECAST A lack of historical data made it impossible to do a quantitative forecast. However, a qualitative forecast is plausible. Due to the very nature of Life Expectancy, increasing Life Expectancy will result in growth. Life Expectancy is not ever expected to decrease. Therefore, to assume that it will decrease in the future (and decrease growth simultaneously) is unfounded. In fact, the only decrease that happened in the historic data can be attributed to a major natural disaster in Cambodia. This is shown in Graph 4. Furthermore, the coefficients for Life Expectancy in the regression model outputs are largely positive implying that this will increase growth as time goes on. Growth will come from taking measures to increase Life Expectancy. Foreign Direct Investment inflows present great potential for the region. Singapore, the most developed country in the region, almost has yearly the largest percentage of their GDP Proceedings of the 2012 Pennsylvania Economic Association Conference 119 coming from FDI over every other country from 1970 on. This is shown graphically in Graph 2. FDI will be attracted to this region due mainly to the success that Singapore has already had. The Exports of Goods and Services do nothing for the region. Exports are shown in Graph 5. They fluctuate from year to year and have no pattern. Due to export expansion promotion, they will continue to try to increase their exporting. This has been proven to not help with growth and should not continue to take place. TABLES AND GRAPHS Table 1: Variable Description Variables (Y and Xs) Economic Growth (PCRGDP) FDI (FDI) Savings (REALK) Health (LIFE) Exports (VEXP) Manufacturing (MFG) Measure Source Years PPP GDP per capita(constant 2005 international $) World Bank 1980-2010 Foreign Direct Investment, net inflows (% of GDP) Gross capital formation (constant 2000 US$) Life Expectancy at Birth in years Volume of Exports of Goods and Services Manufacturing as a Percentage of GDP World Bank 1980-2010 World Bank 1980-2010 World Bank 1980-2010 Gapminder 1980-2010 World Bank 1980-2010 Table 2: Correlation Matrix PCRGDP FDI REALK LIFE VEXP MFG YEAR PCRGDP 1.00000 0.73534 0.17118 0.72475 -0.025344 0.24163 0.12257 FDI REALK LIFE VEXP MFG YEAR 1.00000 -0.11194 0.55627 0.13654 -0.017440 0.15483 1.00000 0.32569 -0.16012 0.63156 0.23271 1.00000 -0.077926 0.52272 0.22081 1.00000 -0.17698 -0.061674 1.00000 0.10964 1.00000 Table 3: Fixed and Random Effects Variable FDI REALK LIFE VEXP MFG R-squared Model 1 294.519 .093 .811520E-07 .031 1099.81 .000 2.32371 .814 -330.055 .004 .911734 Model 2 296.071 .092 .817391E-07 .028 1096.06 .000 -330.801 .004 .911723 Model 3 .542660E-02 .042 .214970 .000 .067133 .000 -.144041E-03 .691 ,015458 .000 .992051 Model 4 .531747E-02 .045 .215991 .000 .067103 .000 .015478 .000 .992047 Proceedings of the 2012 Pennsylvania Economic Association Conference 120 Table 4: Fixed and Random Effects 2 Variable FDI Model 5 325.276 .001 .758159E-07 .037 1116.80 .000 2.28777 .880 -325.031 .000 -64798.7 .000 .625298 REALK LIFE VEXP MFG C R-squared Model 6 326.473 .001 .763696E-07 .034 1113.34 .000 -325.841 .000 -64536.9 .000 .625657 Model 7 .57437E-02 .042 .209041 .000 .068366 .000 -.157203E-03 .713 .015419 .000 -1.61027 .000 .626445 Model 8 .56168E-02 .045 .215991 .000 .068318 .000 .015441 .000 -1.63557 .000 .625410 Graph 1: Economic Growth Economic Growth Cambodia 50000 Indonesia 40000 Lao PDR 30000 Malaysia Philippines 20000 Singapore 10000 Thailand Vietnam 0 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 GDP per capita, PPP (constant) 60000 Proceedings of the 2012 Pennsylvania Economic Association Conference 121 Graph 2: Foreign Direct Investment Foreign Direct Investment, net inflows 25 Cambodia Percent of GDP 20 Indonesia 15 Lao PDR 10 Malaysia Philippines 5 Singapore 2009 2006 2003 2000 1997 1994 1991 1988 1985 1982 1979 1976 1973 -5 1970 0 Thailand Graph 3: Savings 8E+10 7E+10 6E+10 5E+10 4E+10 3E+10 2E+10 1E+10 0 Cambodia Indonesia Lao PDR Malaysia Philippines Singapore Thailand 1960 1964 1968 1972 1976 1980 1984 1988 1992 1996 2000 2004 2008 Volume (constant) Gross Capital Formation Vietnam Proceedings of the 2012 Pennsylvania Economic Association Conference 122 Graph 4: Health 90 Cambodia 80 Indonesia 70 Laos 60 Malaysia 50 40 Philippines 30 Singapore 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 Live Expectancy (Number of Years) Health Thailand Graph 5: Volume of Exports of Goods and Services Volume of Exports of Goods and Services Percent of GDP 200 Cambodia 150 Indonesia 100 Laos 50 Malaysia 0 -50 1980 1987 1994 2001 2008 Philippines Singapore -100 Graph 6: Manufacturing, Value Added Manufacturing, Value Added 35 Cambodia 30 Indonesia 25 Lao PDR 20 Malaysia 15 Philippines 10 Singapore 5 Thailand 2008 2004 2000 1996 1992 1988 1984 1980 1976 1972 1968 1964 0 1960 Percent of GDP 40 Vietnam Proceedings of the 2012 Pennsylvania Economic Association Conference 123 REFERENCES Booth, A. (2004). Linking, de-linking and re-linking: Southeast Asia in the global economy in the Twentieth Century. Australian Economic History Review, 44(1), 35-51. de Mello, L. r., & Fukasaku, K. (2000). Trade and foreign direct investment in Latin America and Southeast Asia: Temporal Causality Analysis. Journal Of International Development, 12(7), 903-924. Ghirmay, T., Grabowski, R., & Sharma, S. C. (2001). Exports, investment, efficiency and economic growth in LDC: An empirical investigation. Applied Economics, 33(6), 689-700. Greene, William H. (2000). Econometric analysis, 4th ed. Prentice Hall Inc. Upper Saddle River, NJ. Jongwanich, J. (2010). Determinants of export performance in East and Southeast Asia. World Economy, 33(1), 20-41. Rasiah, R., & Hing, A. (2009). Industrializing Southeast Asia. Journal Of The Asia Pacific Economy, 14(2), 107-115. Rostow, W. W. (1960). The stages of economic growth: A non-communist manifesto. Cambridge: Cambridge University Press. Hsiao, F. T., & Hsiao, M. W. (2006). FDI, exports, and GDP in East and Southeast Asia--Panel data versus time-series causality analyses. Journal Of Asian Economics, 17(6), 1082-1106. doi:http://dx.doi.org/10.1016/j.asieco.2006.09.011 Thomsen, S. (1999). Southeast Asia: The role of foreign direct investment policies in development. Trindade, V. (2005). The big push, industrialization and international trade: The role of exports. Journal Of Development Economics, 78(1), 22-48. doi:http://dx.doi.org/10.1016/j.jdeveco.2004.08.006 Proceedings of the 2012 Pennsylvania Economic Association Conference 124 The Dodd-Frank “Wall Street Reform” Act of 2010: Is our Financial System More Stable Now? Adora D. Holstein Department of Economics and Legal Studies Robert Morris University Moon Township, PA 15108 ABSTRACT This study maps key provisions of the Dodd-Frank Act against regulatory gaps and systemic vulnerabilities revealed by the recent financial crisis. Issues of regulatory capital arbitrage have been addressed, but limits on the use of short-term funding for long-term assets, reducing runs on money market funds, and privatizing Fannie Mae and Freddie Mac still need to be worked out. A systemic oversight council now closely monitors, regulates, and limits the growth of large, systemically important banks and nonbanks. The ‘living will’ required has the potential of making complex institutions more transparent, and easier for the FDIC to liquidate. However, no periodic fee is assessed on nonbanks to build up a fund that will preclude the use of taxpayer’s money. Moreover, the Volcker Rule provision falls short of restoring the Glass-Steagal separation of banking and investment banking. Ultimately, enforcement of new rules will depend on the regulatory philosophy of the oversight council members, and adequate funding of regulatory agencies so they can keep up with financial innovations. INTRODUCTION The subprime crisis drastically changed the financial landscape in the United States. Of the five largest investment banks, only two survived (Goldman Sachs and Morgan Stanley), but along with large consumer finance companies (American Express and CIT) they obtained Fed authorization to become bank holding companies (BHCs).1 The federal government took substantial ownership stake in the world’s largest insurance company (AIG), and took two government sponsored enterprises (Fannie Mae and Freddie Mac) into conservatorship. There is a general understanding that the root cause of the financial crisis was the bursting of a housing bubble fueled by lax underwriting and securitization of subprime mortgages. Yet asset bubbles (most recently the dot.com bubble) have burst in the past but did not push the U.S. and the global economy at the brink of another depression. The financial crisis of 2008 revealed how vulnerable the U.S. financial system was to systemic risk. With the Fed’s aggressive use of new emergency lending facilities and Congressional authorization of the Troubled Assets Relief program, fewer banks failed during this crisis than during the S&L crises of the 1980s and the Great Depression. In July 2010, the Dodd–Frank Wall Street Reform and Consumer Protection Act (DFA) was signed into law by President Obama. It has been hailed as the most comprehensive reform of financial regulation in the U.S. It was designed to (1) strengthen safeguards for consumers and investors, (2) provide better tools for limiting risk in major financial institutions and markets, and (3) help create a stronger, more resilient financial system that is less vulnerable to crisis, fraud and abuse and more efficient in allocating financial resources (Senate Banking Committee 2010). As expected, there is an ongoing debate about whether DFA, known to many as “Wall Street Reform,” is excessive and doubts about how well it can prevent another financial crisis abound. The objectives of this paper are: (1) to review the views of academic researchers as well as financial regulators on what critical regulatory flaws and vulnerabilities of the financial system led to the housing bubble, spread the negative repercussions throughout the financial system, and ultimately imposed costs on the real sector of the economy; (2) to map the provisions of DFA against these critical regulatory flaws and financial system vulnerabilities; and (3) to determine what flaws and vulnerabilities, if any, are not addressed adequately or at all in the DFA. To start, I will review the literature on the market failures that justify government intervention in the financial sector of the economy and the difference in regulation of banks and nonbank financial institutions. Against this backdrop, I will then review the academic literature as well as reports of evaluations done by regulatory agencies on what caused the housing bubble and the financial crisis that followed once it burst. This is followed by a mapping of the provisions of DFA against these critical regulatory flaws and financial system vulnerabilities. Following an evaluation of DFA’s potential for making the financial system more stable, the paper closes with some concluding comments. LITERATURE REVIEW Despite the reliance of the U.S. on the market system to determine the prices of most goods and services, financial institutions are subject to varying degrees of government Proceedings of the 2012 Pennsylvania Economic Association Conference 125 regulation. The need for government intervention in the finance sector of the economy and the extent of such intervention has many rationales in the economics literature. Among these are negative externalities (Pigou 1932), efficient markets hypothesis (Fama 1970), and asymmetric information (Ackerloff 1970), which brings about the problems of adverse selection and moral hazard. Prior to the establishment of the Federal Deposit Insurance Corporation (FDIC) in 1934, the failure of a bank created uncertainty about which banks will fail next and resulted in bank runs. Depositors panic and rush to close their accounts when they are unable to distinguish the good banks from the bad banks (a problem of adverse selection). This problem arises from the fact that bank insiders have better information about the soundness of the bank’s financial position than its depositors (a problem of asymmetric information). Banks perform a vital function of transferring funds from savers to borrowers who can use it for productive investments. When there is uncertainty about the financial soundness of banks, the resulting drop in prices of their stocks easily gets transmitted across the stock market. A banking crisis ultimately affects the whole economy adversely, because a drop in equity capital among banks will mean a reduction in loans that banks can extend to consumers and businesses, imposing a negative externality to people who lose their jobs or whose net worth shrink. The importance of infusing liquidity into the banking system during a financial crisis was recognized by Pigou (1932) eight decades ago when he said that “…other things being equal, the actual occurrence of business failures will be more or less widespread, according [to whether] bankers' loans, in the face of crisis of demands, are less or more readily obtainable.” Hence, regulation of banks in the U.S. has been relatively more intrusive than the regulation of nonbank financial institutions like brokerdealers and mutual funds. Instead of exercising due diligence in determining what risks their banks are taking and moving their deposit accounts elsewhere if these are above acceptable levels, most depositors rely on the protective cover not only of FDIC insurance, but also of various federal regulators who ensure that banks maintain a minimum level of liquidity, adequate capital commensurate with the market risk of their loan and investment portfolio, and orderly liquidation if they fail. In contrast, regulation of the securities industry has leaned more towards ensuring disclosure of information so that investors can effectively exercise market discipline should they perceive mispricing: selling securities that are overpriced and buying securities that are underpriced. This regulatory stance is predicated on the efficient markets hypothesis (EMH), which states that asset prices are always and everywhere correct as all available information is processed by rational market actors (Fama 1970). Thus, investors and stockholders of investment banks, insurance companies, mutual funds, and other nonbank financial institutions exercise their own due diligence or pay for expert advice to analyze bond ratings by credit rating agencies, and independently audited, financial statements and other information that the Securities and Exchange Commission (SEC) requires them to disclose to the public. Sarbanes-Oxley (2002) has added another assurance by requiring that financial statements be attested to by the Chief Executive Officer of public corporations. There is also an assumption that these public entities are subject to oversight by a Board of Directors. Economists would caution that certain forms of government intervention in the finance industry have unintended consequences. For example, the safety net created by banks’ access to the Fed’s discount window and the FDIC’s insurance fund creates a moral hazard that depositors will no longer have the incentive to monitor the risks that their banks take. In addition, as the FDIC have been observed to use the purchase and assumption approach to resolving a bank that is too big to fail, instead of paying off its depositors up to the insured limit, yet another moral hazard problem arises. The assurance that its liability to all depositors can be transferred to the FDIC and a potential buyer increases management’s incentive to take risks. As Bullard et al. (2009) point out, systemic risk can become more of a problem for the economy when firms who should fail are not allowed to fail. The causes of the financial crisis have been the subject of many studies in academia as well as the private sector. Acharya and Richardson (2009) estimated that losses from underlying mortgages accounted for less than half of the total asset write-downs that financial institutions made since the third quarter of 2007, the rest resulted from amplifiers such as: 1) Excessive leverage. Investment banks loaded up on debt to increase returns on equity when asset prices were rising. They also increased their exposure to product leverage through complex derivatives which needed only a slight deterioration in the value of underlying assets for losses to escalate rapidly. Lastly, they overindulged in liquidity leverage, using structured investment vehicles (SIVs) or relying too much on wholesale markets to exploit the difference between borrowing cheap short term money and investing in higher-yielding long-term assets. The combined effect was that falls in asset values cut deeply into equity and triggered margin calls from lenders. The drying up of liquidity had an immediate impact because debt was being rolled over so frequently; 2) Derivatives like collateralized debt obligations (CDOs), credit default swaps and interest rate swaps, Proceedings of the 2012 Pennsylvania Economic Association Conference 126 which created ‘synthetic’ exposures to subprime mortgages without investors actually having to own them, and asset backed-commercial paper (ABCP), which was the major source of short-term funds for SIVs; 3) Concentration of counterparty risk. The decision by the Fed to facilitate JP Morgan’s acquisition of Bear Sterns had less to do with the size of Bear’s assets than with its central role in markets for credit-default and interest rate swaps; and 4) Fair-value accounting rules which require financial institutions to mark down the value of mortgagebacked securities (MBS) and other derivative securities to current market prices. Downward price movements trigger the need to unwind investments, further depressing prices. Other studies echo the above assessments. Hellwig (2009) claims that a faulty financial infrastructure that allowed mortgage-backed instruments to allocate risk in the real estate market, along with structured investment vehicles and a lack of capital at financial institutions combined to create a downward spiral of problems. Morgan (2009) adds to this by saying that the crisis was due to the failure by market participants to process information, reliance on flawed risk management models, and inadequate interaction between central banks and other regulators. Rajan (2009) puts the blame on unsustainably low interest rates set by monetary policymakers and financial innovations that led to the housing boom and subsequent defaults Shin (2011) focused on the role of U.S. branches and subsidiaries of foreign (mainly European) banks in spreading the crisis outside the U.S. by borrowing or raising wholesale short-term funding through U.S. money market funds, then lending it back as they used these funds to buy long-term private-label mortgage-backed securities and structured products. Reports issued by U.S. and international regulatory authorities after their own investigations also mirror those in the academic literature. In its March 2008 report, the President’s Working Group on Financial Markets 2 (PWG 2008) identified instances when existing regulatory policies failed to mitigate risk-taking among financial institutions: 1) Supervisory authorities did not insist on appropriate disclosures of potential exposure to off-balance sheet vehicles; 2) Existing capital requirements encouraged the securitization of assets through facilities with very low capital requirements; and 3) Existing regulation failed to provide adequate incentives to maintain capital and liquidity buffers sufficient to absorb extreme system-wide shocks without taking actions that tended to amplify shocks. The Bank for International Settlement’s Committee on the Global Financial System (BIS 2009) issued a report in April 2009 noting that what set the stage for the crisis was a surge in leverage at banks and investment banks in the U.S. and Europe, but in in less visible ways such as: 1) risk embedded in structured credit products, which made traditional measures of balance sheet leverage less meaningful; 2) assets held in highly leveraged off-balance sheet vehicles; and 3) funding by off-balance sheet vehicles and by some large financial institutions of a growing amount of long-term assets with short-term liabilities in wholesale markets, a mismatch in maturity that increased their exposure to liquidity risk. Similarly, the Government Accountability Office (GAO 2009) reported the following findings from their investigation: 1) the traditional measure of leverage, total assets to equity on the balance sheet, did not fully capture risk exposure arising from the use of derivatives; 2) deleveraging by selling these derivatives caused prices to spiral downward during times of market stress and exacerbated the financial crisis; 3) the SEC relaxed the net capital rule on the five largest broker-dealers in 2004 and did not do an independent assessment of their risk-management models; and 4) multiple regulators were responsible for individual markets or institutions, but no one has clear responsibility to assess the potential effects of the buildup of system-wide leverage or the collective effect of institutions’ deleveraging activities. The Securities and Exchange Commission (SEC) Chairman Mary Schapiro testified before the Financial Crisis Inquiry Commission in January 2010 that the crisis resulted from many interconnected and mutually reinforcing causes, including: 1) The rise of mortgage securitization (a process originally viewed as a risk reduction mechanism) and its unintended facilitation of weaker underwriting standards by originators and excessive reliance on credit ratings by investors; 2) A wide-spread view that markets were almost always self-correcting and an inadequate appreciation of the risks of deregulation that, in some areas, resulted in weaker standards and regulatory gaps; 3) The proliferation of complex financial products, including derivatives, with illiquidity and other risk characteristics that were not fully transparent or understood; 4) Perverse incentives and asymmetric compensation arrangements that encouraged significant risk-taking; Proceedings of the 2012 Pennsylvania Economic Association Conference 127 5) Insufficient risk management and risk oversight by companies involved in marketing and purchasing complex financial products; and 6) A siloed financial regulatory framework that lacked the ability to monitor and reduce risks flowing across regulated entities and markets. The assessment of the Basel Committee on Banking Supervision 3 (2010) is that the crisis was caused by: 1) excess liquidity, resulting in too much credit and weak underwriting standards; 2) excess leverage; 3) inadequate liquidity buffers; 4) too little capital of insufficient quality; 5) major shortcomings around risk management, corporate governance, market transparency, and the quality of supervision; 6) procyclical deleveraging process; and 7) interconnectedness of systemically important, too-bigto fail financial institutions. A study by the Financial Stability Oversight Council (2010) created by the Dodd-Frank Act, concluded that the bursting of the housing bubble exposed certain vulnerabilities in the U.S. financial infrastructure, namely: (1) interconnectedness of large, complex financial institutions due to exposure to a common counterparty in subprime MBS-related derivatives transactions; (2) risk concentration in subprime MBS; and (3) lack of transparency about which institutions were interconnected and the extent of their exposure to subprime MBS risk caused the markets to seize up. Most recently, Fed Chairman Bernanke gave a lecture on “The Fed and the Financial Crisis” at George Washington University in March 2012, in which, looking back, he noted that the decline in house prices and the associated mortgage losses were key triggers of the crisis, but that the effects of those triggers were amplified by the following vulnerabilities in the financial system (Bernanke 2012): 1) borrowers and lenders took on too much debt or leverage; 2) banks and other financial institutions failed to adequately monitor and manage their risk exposures to subprime mortgages; 3) financial institutions relied excessively on short-term funding, such as commercial paper; 4) the increased use of exotic financial instruments concentrated risk; 5) gaps in the regulatory structure left important firms without strong supervision; and 6) failures of regulation and supervision, including consumer protection, and insufficient attention paid to the stability of the financial system as a whole. In the next two sections, I will discuss the provisions of DFA according to two categories of regulatory flaws and financial system vulnerabilities revealed by the financial crisis of 2008. The first section will discuss the DFA provisions that address problems that contributed to excessive leverage that fueled a housing bubble. This is followed by a discussion of the DFA provisions that address problems that spread the risk throughout the financial system. DFA PROVISIONS PERTAINING TO PROBLEMS THAT LED TO THE HOUSING BUBBLE Provisions of the Dodd-Frank Act (DFA) are mapped to the regulatory flaws and financial system vulnerabilities revealed by the financial crisis of 2008 in Table 1. A more detailed discussion of each problem that led to the housing bubble which burst in the Spring of 2006 is outlined below followed by a discussion of relevant provisions of DFA that address such problem. This discussion owes much to a book edited by Acharya and Richardson (2009) and an article written by Crotty (2009), both of which provide a detailed description of the roles played by credit rating agencies , unregulated ‘shadow banks’ (e.g. mortgage brokers, finance companies, off-balance conduits, investment banks, insurance companies, hedge funds, and money market funds), lightly regulated derivatives (e.g. collateralized debt obligations and credit default swaps) , as well as compensation schemes that were skewed towards excessive risk-taking. No credit risk retention for securitized loans By selling mortgage loans to investment banks and government-sponsored enterprises (Fannie Mae and Freddie Mac) for securitization into mortgage backed securities (MBS), banks were able to augment deposit funds with capital market funds to extend a larger amount of mortgage loans. Securitization, however, enabled banks to pass on 100% of the credit risk, thus removing their incentive to observe strict underwriting standards. Moreover, with the growing demand for high-yielding subprime mortgage-backed securities, banks were able to increase the volume of loans originated through their mortgage brokers or finance company affiliates. Because these affiliates were not regulated at the federal level, and state regulation was less stringent, and unevenly enforced compared to federal regulation, lax underwriting standards and fraudulent lending practices gave rise to high-risk creative loan contracts (e.g. no income documentation or Alt-A loans, interest only payments, negative amortization). Title X of DFA created a new Bureau of Consumer Financial Protection (BCFB) within the Federal Reserve to regulate mortgage brokers and finance companies at the federal level for the first time. It also consolidated the responsibilities previously handled by various federal Proceedings of the 2012 Pennsylvania Economic Association Conference 128 agencies pertaining to consumer complaints against mortgage-related businesses. The BCFB is empowered to act fast without waiting for Congress to pass a law prohibiting bad lending practices that harm consumers. In addition, Section 1141 of DFA amended the Truth in Lending Act to prohibit creditors from making mortgage loans without regard to a consumer’s repayment ability. To ensure that banks and bank holding companies (hereafter referred to as banking organizations) would have “skin in the game” or have something to lose if they do not observe strict underwriting standards, DFA requires that capital be reserved against no less than five percent (5%) of mortgage loans sold for securitization. Securitizers are also required to disclose information that would enable investors to assess the quality of underlying loans. Regulatory arbitrage through off-balance sheet conduits As part of the securitization process, a bank sets up and administers an off-balance sheet conduit, a shell company that issues asset-backed commercial paper (ABCP) and sells it to investors, among the biggest of which are money market mutual funds. This short-term funding is then used by the conduit to purchase long-term mortgage-backed securities (MBS) other collateralized debt obligations (CDOs) as assets. Banks often became the servicers of subprime mortgage loans that they have sold, earning fees for collecting the monthly payments from homebuyers and passing on a proportionate share of these collections to ABCP investors. Securitizing subprime mortgage loans and moving the MBS into off-balance sheet conduits enable the banks to reduce the capital required to meet capital adequacy requirements, thus increasing its leverage. As administrator of the conduit, the bank provides it with: a) liquidity enhancement (LE): the bank insures the conduit against liquidity risk, making a commitment to provide a backup line of credit by repurchasing the conduit’s non-defaulted assets in case it cannot rollover maturing commercial paper. This saves capital charge because the capital requirement on mortgage loans held on balance sheet is at least three times that required for on-balance sheet liquidity enhancement facilities; and b) credit enhancement (CE): the bank insures the conduit against credit losses from the MBS and other assets held. The capital requirement for credit enhancements is larger than for liquidity enhancements, but the sum of capital requirements on these two enhancements is still lower than that for on-balance sheet mortgage loans. According to Acharya and Schnabel (2009), a bank’s risk exposure to the conduit varies across three types of conduits: a) Fully supported conduit: LE covers 100% of ABCP outstanding and CE covers 100% of all assets held, i.e. with full recourse to the bank’s balance sheet; b) Partially supported conduit: LE covers 100% of ABCP outstanding but CE only covers an average of 7~10% of total assets; c) Structured investment vehicle (SIV): LE and CE cover an average of 25% of assets. Although only the backup liquidity line and credit enhancement were on the balance sheet and subject to a capital charge, the bank is exposed to the risk of default and potential illiquidity of the MBS collateral. During the financial crisis, the limitations on LE and CE were largely ineffective as large financial institutions like Citibank were forced to take back assets or extend more CE, thus effectively providing full recourse to its balance sheet. DFA authorized the revision of capital adequacy requirements on banking organizations consistent with the Basel III international capital standards, which were established to incorporate lessons learned from the financial crisis (see Table 2 for increases in risk-weights assigned to certain assets). Federal bank regulators have proposed that effective January 1, 2015, all banks and bank holding companies comply, on a consolidated basis, with the following minimum capital ratios: (a) baseline total capital to risk-weighted assets ratio of 8% (unchanged); (b) tier 1 capital to risk-weighted assets ratio of 6% (increased from 4%); (c) leverage ratio or tier 1 capital to average consolidated assets (not risk-weighted) of 4% (unchanged but with stricter definition of tier 1 capital); (d) new common equity tier 1 capital ratio of 4.5%; and (e) new capital conservation buffer or common equity to riskweighted assets of no less than 2.5% (see Table 3). The last two requirements in the list will mean that banks will have to keep common equity of at least 7% of riskweighted assets. The common equity tier 1 capital ratio is a new minimum requirement designed to ensure that banking organizations hold high-quality regulatory capital that is available to absorb losses in times of market stress. In addition, a bank that has a capital buffer below 2.5% will be subject to restrictions that no more than 60% of retained earnings (or less depending on how low the buffer is) can be paid out as capital distributions and discretionary bonus payments. To remove the incentive to reduce capital charges through off-balance sheet activities, DFA also requires a banking organization to hold capital, with 100% risk weight, against off-balance sheet guarantees, repurchase agreements and other repo-style transactions, financial standby letters of credit, forward agreements, and other derivative contracts. In addition, effective January 1, 2018, large, complex banks (with over $50 billion in assets) designated as systemically important will be required to maintain supplemental tier 1 capital of no less than 3% of its total leverage exposure, including unconditionally cancellable commitments such as liquidity and credit enhancements (exposure is calculated at10% of the Proceedings of the 2012 Pennsylvania Economic Association Conference 129 notional amount), and all other off-balance sheet exposures (exposure is calculated at 100% of the notional amount). This is intended to discourage the acquisition of excess leverage and to act as a backstop to the risk-based capital requirements. Finally, DFA attempts to remove future regulatory arbitrage opportunity by prohibiting a bank from converting its charter in order to get out from under an enforcement action, unless both the old regulator and new regulator approve of such conversion). Loose oversight of derivatives and hedge funds During the financial crisis, lack of transparency about the extent of exposure to subprime MBS by financial institutions caused the markets to seize up. The structuring of CDOs backed by subprime mortgages became so opaque, and the valuation of these products became too complex for many investors to exercise due diligence, and for directors and stockholders of financial institutions to exercise market discipline. Opaqueness of off-balance sheet conduits also made it difficult for banks to assess which banks would be hit by losses and how much if confidence in the ABCP market made it difficult for conduits to issue new commercial paper, the so-called ‘rollover risk’. This caused the ABCP market to freeze up in the Fall of 2008. Derivatives like CDOs, ABCP, and credit default swaps, as well as the hedge funds and private equity funds that invested in these derivatives were not subject to disclosure requirements by the SEC, and were instead loosely regulated by the CFTC. This regulatory stance was predicated on the fact that investors in derivatives were sophisticated investors who can do their own due diligence or are able to pay for management fees charged by hedge funds and private equity firms. Credit default swaps, which in effect, insured investors in derivative products backed by subprime mortgages, enhanced the marketability of these products not only in the U.S. but in other countries as well, thus fueling the housing bubble. DFA requires the standardization of credit default swaps and trading of many over-the counter derivatives in central clearinghouses or exchanges. It empowers the SEC and CFTC to (a) impose capital and margin requirements on swap dealers and major swap participants, and (b) collect and publish data to increase transparency of derivative products, their trading volumes, and counterparties. To incentivize centralized clearing, derivative and repo-style transactions cleared through central counterparties, instead of over-the-counter through bilateral transactions, will have reduced capital requirements. Large foreign and domestic banking organizations operating in the U.S. that have been designated as systemically important will be required to maintain a supplementary tier 1 capital of no less than 3% of its potential exposure for each derivative contract to which it is a counterparty, and to disclose the exposure amount and other counterparties to such derivative contract. Over-reliance on credit rating agencies Investors in the U.S. and abroad relied on credit rating agencies (CRAs) in assessing the investment risk of complex CDOs and ABCP, rather than doing their own due diligence. The financial crisis revealed that CRAs (a) gave certain mortgage-related securities higher ratings than were deserved, and were too slow in issuing downgrades once the credit markets slumped, (b) did not adequately differentiate the risk of MBS from a corporate bond of similar rating, (c) engaged in conflict of interest when investment banks hired them as consultants and earned fees for providing advice to arrangers of structured financial products on how CDO tranches can obtain desired credit ratings, and (d) did not make their track record and methodologies used in arriving at credit ratings publicly available. Despite the fact that CDOs were new financial products, models used by CRAs relied on historical default rates that proved to be too low. The inaccuracy of ratings on structured financial products contributed significantly to the mismanagement of risks by financial institutions and investors. For example, banks are restricted to invest only in investment grade fixed-income securities, but as long as a CDO tranche backed by subprime mortgage loans got AAA ratings banks did not violate any regulation by investing in it. Holding a AAA CDO tranche as an investment involved a lower capital charge than holding mortgage loans on their balance sheets. The consequence of this is not trivial. While securitization was hailed as a way to transfer risk outside of the banking system, about half of AAA-rated MBS remained within the banking system, 30% held by banks as on-balance sheet investments, and another 20% by offbalance sheet ABCP conduits and SIVs (Acharya et al. 2009). Likewise, money market mutual funds which are known to guarantee a stable net asset value became the largest holders of AAA-rated asset-backed commercial paper. On August 9, 2007, the ABCP market froze when BNP Paribas suspended calculation of the NAV of two of its money market mutual funds that had invested in ABCP. This led banks to take a large number of ABCP conduits and SIVs back on their balance sheets, causing interbank lending rate to rise sharply. This happened again the days immediately following Lehman’s bankruptcy. Moreover, federal bank regulators based capital requirements against securities held by banks and brokerdealers on ratings by credit rating agencies. By basing capital requirements on risk-weighted assets, regulators did not catch the build-up of excessive leverage made possible Proceedings of the 2012 Pennsylvania Economic Association Conference 130 by faulty credit ratings and regulatory arbitrage through the use of off-balance sheet conduits. According to an IMF (2008) data, total bank assets almost doubled from about $8 trillion in 2004 to about $15.5 trillion in mid-2007, but risk-weighted assets only grew from $4 trillion to $5.5 trillion. DFA created a new Office of Credit Ratings within the SEC to: (a) increase transparency by requiring CRAs to publicly disclose their rating methodologies and ratings track record, and (b) to remove conflicts of interest issues by requiring the use of third-party sources of data, not just the institution being rated, prohibiting compliance officers from working on ratings, methodologies , or sales, and requiring a one-year look back review when a rating agency employee goes to work for an obligor or underwriter of a security or money market instrument subject to the agency’s rating. The SEC has been empowered to register an agency for providing bad ratings over time. By nullifying SEC Rule 436 (g) which exempted credit ratings provided by CRAs from being considered part of the registration statement for securities, investors will be able to sue CRAs for reckless failure to conduct a reasonable investigation of the facts or to obtain an analysis from an independent source. The capital adequacy requirements have also been modified so that if a banking organization sells mortgage loans for securitization but later invests or holds every tranche of the securitization, its overall capital charge would be greater than if the bank held the underlying mortgage loans. This is intended to discourage regulatory capital arbitrage through securitization, and ensure that securitization really transfers the risk outside of the banking system. To deal with the problem of regulators relying on security ratings by credit rating agencies, section 939A of DFA requires federal agencies to remove references to credit ratings from regulations and replace credit ratings with appropriate alternatives. The implementation rulings issued since July 2010 include alternative measures of creditworthiness for determining required risk-based capital for residential mortgages, securitization exposures, and counterparty credit risk. To assess a bank’s risk exposure to mortgage loans on single-family homes valued up to $1 million that are not government guaranteed, bank regulators have proposed that starting in 2015, the loan-to-value (LTV) ratio and loan terms be used. Under this proposal, the property value in the LTV ratio will be based on an appraisal that conforms to appraisal regulations. Loans are to be classified into two categories: Category 1 would be for loans that do not have loan terms associated with higher credit risk as the subprime foreclosure crisis revealed (e.g. undocumented income, balloon payment, negative amortization, over 30-year maturities, adjustable-rate mortgages (ARM) with an annual interest rate cap above 2% and life of loan cap above 6%). In addition, Category 1 loans can only include those that are first lien loans, not 90 days or more past due, and in the case of an ARM loan, the borrower qualified accounting for interest rates going up by the maximum 6% life-of-loan cap, real estate taxes, and mortgage insurance. Loans that do not satisfy these criteria will be classified in Category 2. Category 1 loans will have lower risk weights, and within each category, the higher the LTV ratio, the higher is the risk weight (see Table 4). For securitization exposures, banking organizations are required to satisfy specific due diligence requirements instead of relying exclusively on credit ratings. This includes conducting and documenting an analysis of the risk characteristics of the underlying assets (property types, occupancy rate, average LTV ratio, industry and geographic diversification) prior to acquisition, and at least every quarter thereafter, relevant information regarding the performance of the underlying credit exposures (e.g. default rates, prepayment rates, loans in foreclosure). If the banking organization is not able to demonstrate a comprehensive understanding of a securitization exposure to the satisfaction of its primary federal supervisor, it will be required to assign a risk weight of 1,250% to the exposure. For a minimum required capital ratio of 8%, this would mean putting up capital equal to the exposure amount. In June 2012, bank regulators proposed alternative approaches for determining the risk exposure to securitization. A banking organization that chooses to apply another approach will be required to assign a 1,250% risk weight to all securitization exposures. There is a 20% floor on the risk weight for securitization exposures, and varies up to 100% depending on the weighted average risk of the underlying mortgage loans or other assets. Irrational (herding) behavior of market participants Herding behavior is said to occur when investors buy or sell to imitate what other market participants are doing rather than on the basis of their own assessment of the asset’s fundamentals. Some argue that even if more information were available about subprime MBS, CDOs, and ACBP, such herding behavior would still have caused some market investors to underestimate the risk of these financial products or overestimate the duration of the housing boom. While no regulation can ever be designed to prevent irrational or excessive risk-taking behavior among homebuyers and investors, many of the DFA provisions already discussed are designed to increase transparency in the securitization and derivatives markets. Hopefully, the previous experience of market loss will motivate market participants to use enhanced information to analyze the fundamentals of investments. As for Proceedings of the 2012 Pennsylvania Economic Association Conference 131 nonbank financial institutions that take on too much risk, DFA has restricted the Fed to extending access to their liquidity facilities, but not bailing out failing institutions. Instead, as will be discussed in the next section, DFA subjects large, complex financial institutions to stricter prudential standards and an orderly liquidation process in case of bankruptcy. Compensation schemes rewarded excessive risk-taking Executives, traders, loan managers, and brokers were rewarded with bonuses proportional to the trading volumes or returns they have generated in a specific year, without imposing any loss on them when risky activities that were initially profitable, later led to losses. DFA has a ‘clawback’ provision which requires a publiclytraded company that restates its earnings because of serious accounting errors to seek repayment of compensation from any past or present executive officer in excess of what would have been paid under the restatement. Also if a large financial company is put into receivership, the FDIC can take back the last two years of compensation from any senior executive or director, past or present, who is substantially responsible for the failure of the firm. In addition, a ‘Say on Pay’ provision gives stockholders of financial institutions a right to a non-binding vote on executive pay and golden parachutes, thus giving them a chance to disapprove where they see misguided incentive that increase risk-taking beyond what they find acceptable. To promote transparency, DFA directs the SEC to clarify disclosures relating to compensation, including requiring public corporations to provide charts that compare their executive compensation with stock performance over a five-year period. Last, but not the least, systemically important financial institutions (the designation criteria for which will be discussed later) will be subject to restrictions on the amount of bonuses they can pay out of retained earnings if they are not able to maintain a required capital conservation buffer. This will be further explained below. DFA PROVISIONS ON SYSTEMIC RISK The financial crisis revealed the interconnectedness of large financial institutions on one hand, and the financial and real sectors of the economy, on the other. The phases of the run on shadow banks and traditional banks in 2007 and 2008 following the surge in foreclosures on homes financed by subprime mortgage loans are outlined in Acharya et al. (2009). In early 2007, Countrywide Financial and hundreds of nonbank mortgage lenders specialized in subprime and Alt-A mortgages (loans with no income documentation) collapsed when their source of wholesale financing dried up. Some were merged into large banking institutions to which they were affiliated under a Financial Holding Company. This was followed by the collapse of an entire system of structured investment vehicles and asset- backed commercial paper conduits when investors realized the assets were illiquid and ‘toxic’. Large banks took back assets in off-balance sheet conduits back to their balance sheets, and started to mark down their holdings of CDOs and subprime mortgages that they could no longer securitize. The next phase was the collapse of hedge funds hedge funds such as those managed by Bear Sterns in June 2007, and independent broker-dealers starting with Bear Sterns in March 2008, when the repurchase agreement or ‘repo’ market, which was their main source of short-term funding for purchases of subprime mortgage-backed securities, unraveled. A run on hedge funds, can easily happen because investors could redeem their investments after a short ‘lock-up’ period. Although private equity funds have longer-maturity financing, a refinancing crisis also became possible once primary brokers failed or cut back financing. Thus, the Fed’s most radical postDepression action was to set up a Primary Dealer Credit Facility to provide liquidity to primary dealers after the collapse of two hedge funds at Bear Sterns. Next came the run on the three trillion dollar money market mutual fund industry, which invested funds into commercial paper backed by AAA tranches of CDOs. The Fed had to extend temporary guarantees to money market mutual fund investors to unfreeze the market for commercial paper, which is a main source of short-term funding for corporations. The drying up of liquidity also led to AIG being unable to post significant additional collateral when its credit default swap obligations ballooned. Likewise, many monoline bond insurers lost their AAA ratings and eventually failed. After Lehman’s bankruptcy in September 2008 and acquisition of Merrill Lynch by Bank of America, there was a run on the three other big investment banks, and interbank lending froze as banks started to hoard cash. Thus, the runs on nonbank financial institutions ultimately threatened depository institutions that are the main source of funding to consumers and businesses in the real sector of the economy. The flaws in the financial infrastructure of the U.S. financial system are discussed below with the corresponding provisions of DFA that are intended to correct these. No one regulator to monitor and regulate systemic risk Banks lobbied for many years to repeal the Glass-Steagal Act, which was passed after the Great Depression, to separate the business of banking from that of investment banking. Banks argued that they could not compete with Proceedings of the 2012 Pennsylvania Economic Association Conference 132 large foreign banks, which could engage in securities underwriting, insurance, and in the case of Japan and Germany, also could own equity interest in nonfinancial corporations under so-called “universal banking” system. The Gramm-Leach- Bliley Act (GLBA) enacted in November 1999 allowed the creation of financial holding companies under which umbrella, a bank could engage in securities underwriting, insurance, and real estate operations. This led to consolidation in the financial industry which gave rise to large, complex financial institutions. However, GLBA provided for regulation by function, subjecting such institutions to multiple regulators (see Figure 1). There was no one regulator to monitor, identify, and act on early warning signs of risk to the stability of the financial system. Moreover, regulation by function led some financial institutions to shop for the least restrictive regulators. Some examples are discussed below: a) GLBA allowed banks to be affiliated with mortgage brokers and finance companies under a financial holding company umbrella. The lax underwriting practices of these affiliates were not subject to oversight by the bank regulators. Most of the subprime loans that eventually got securitized and held by banks in their books as investments or in off-balance sheet conduits were originated by these affiliates; b) Faced by the possibility of regulation in European countries in 2004, the five largest U.S. investment banks with foreign subsidiaries (the ‘Big Five’) requested an exemption from the SEC’s net capital rule, which limited their debt –to-equity ratio to 8:1. The SEC Commissioners yielded to this request, in exchange for Fed oversight of the financial holding company each belonged to. An investigation by the Government Accountability Office (GAO 2009) requested by Congress found that this resulted in the Big Five’s debt to equity ratio increasing three to four times just before the Fall of 2008, without sending warning signals to any of the functional regulators. In addition, the GAO report revealed that the SEC relied on the internal risk-management models of the holding companies without doing its own independent review; and c) AIG, the largest insurance company in the world, marketed credit default swaps as derivatives, when in fact these were insurance products that obligated AIG to pay CDO investors in the event of default by subprime borrowers. Derivatives were loosely regulated by the CFTC, while insurance products were strictly regulated by the New York State Insurance Commission. A Frontline (2009) investigation reports that in 2004, CFTC Chairman Brooksley Born recommended increased reporting requirements for derivatives but the Fed Chairman, Treasury Secretary, and the Economic Adviser to then President Bush strongly opposed her. This highlighted the need for one regulator to provide a backstop if one regulator did not effectively do its job and to monitor emerging threats to the whole financial system. The Council also noted that the Basil II capital requirements failed to provide adequate incentives for financial institutions to maintain capital and liquidity buffers sufficient to absorb extreme system-wide shocks without taking actions that tended to amplify shocks. Upon enactment in July 2010, Section 622 of DFA created a new Financial Stability Oversight Council (hereafter, the Council) within the U.S. Treasury, with ten voting members: Treasury Secretary (as Chair) along with the heads of Fed, FDIC, SEC, CFTC, FHFA, NCUA, OCC, the new BCFP, and an insurance industry expert. A minimum of 2/3 or 7 votes are required for any Council decision. As systemic regulator, the Council is empowered to designate Systemically Important Financial Institutions (SIFI) for orderly liquidation by the FDIC in case of failure, but otherwise, for higher prudential standards by the Fed (stricter examination, higher capital, liquidity, and reporting requirements, limits on growth, activities, and credit exposures). DFA gave the Council broad authority to require information from SIFIs for the purpose of comprehensive monitoring to: (a) identify systemic risks that could arise from the material financial distress or failure, or ongoing activities of such companies, or that could arise outside the financial services marketplace; (b) promote market discipline, by eliminating expectations on the part of shareholders, creditors, and counterparties of such companies that the U.S. government will shield them from losses in the event of failure; and (c) respond to emerging threats. A financial institution that received TARP assistance and subsequently de-banks is automatically designated as a SIFI under the so-called “Hotel California” provision of DFA. Other candidates are banks, bank holding companies (BHCs), and nonbank financial institutions with $50 billion or more in assets, including a foreign bank or broker-dealer operating in the U.S. if the home country has not adopted, or made progress towards adopting, appropriate regulation to mitigate systemic risk. The Council may also request information from any financial market utility or financial institution for purposes of determining whether the utility or financial institution’s payment, clearing, and settlement activities are systemically significant. The Paperwork Reduction Act analysis portion of the Notice of Proposed Ruling estimated that about 124 companies (37 FDICinsured banks, 26 U.S. BHCs, and 98 subsidiaries of foreign-owned banks) could be designated as SIFIs. The Council uses a three-stage process and metrics to designate a SIFI: Proceedings of the 2012 Pennsylvania Economic Association Conference 133 Stage 1. A pool is selected from U.S. and foreign nonbanks with consolidated assets of $50B that also exceeds a threshold test on one quantitative metric pertaining to short-term debt and leverage ratios, loans and bonds outstanding , and amount of exposure to credit default swaps and other derivatives; Stage 2. Further evaluation of the potential threat posed by the Stage 1 pool to financial system stability; Stage 3. Council determines whether the company’s material financial distress would pose a threat to U.S. financial stability based on (a) transmission channels through which the company’s distress would affect the broader economy (e.g. interconnectedness, liquidity risk and maturity mismatch, substitutability, size, extent of regulatory scrutiny), and (b) whether resolution of this nonbank would pose a threat to U.S. financial stability (Polk and Wardell 2011). While keeping insurance regulation at the state level, DFA also created a new Federal Insurance Office (FIO) within the Treasury that is authorized to collect data to monitor insurance companies and recommend an insurer, including its affiliate/s, for designation by the Council as a SIFI. A financial market utility (FMU) can also be designated as systemically important if the Council determines that its failure or a disruption to its operations could threaten the stability of the U.S. financial system by way of spreading significant liquidity or credit problems among financial institutions or markets. DFA defines an FMU as “any person that manages or operates a multilateral system for the purposes of transferring, clearing, or settling payments, securities, or other financial transactions among financial institutions or between financial institutions and the person. An FMU designated by the Council as systemically important would become subject to additional reporting, examination and enforcement action by its primary federal regulator, as well as compliance with risk management standards governing its ability to complete timely clearing and settlement of financial transactions, its policies and procedures relating to participant or counterparty default, capital and financial resource requirements, and margin and collateral requirements. In July 2012, the Council designated eight FMUs as systemically important: The Clearing House Payments Com LLC on the basis of its role as operator of the Clearing House Interbank Payments System, the Depository Trust Co., the Chicago Mercantile Exchange, Inc., CLS Bank International, Fixed Income Clearing Corp., ICE Clear Credit, LLC, National Securities Clearing Corp., and The Options Clearing Corp. An Office of Financial Research (OF was created, within the Treasury, to operate a Data Center and a Research and Analysis Center to support the functions of the Council. The heads of the OFR and FIO serve as advisors or non-voting members of the Council, along one state insurance commissioner, one state banking supervisor, and one state securities commissioner, who are appointed to two-year terms by state regulators. Interconnectedness and “Too Big To Fail” The financial crisis revealed weaknesses in risk management that made the largest financial institutions so interconnected due to exposure to a common counterparty in subprime mortgage-related derivatives transactions. The decision by the Fed to facilitate JP Morgan’s acquisition of Bear Sterns in March 2007 had less to do with the size of Bear’s assets than with its central role in markets for credit default and interest rate swaps. Banks and broker-dealers experienced difficulties in assessing counterparty risk, aggregating exposures across business lines, valuing instruments when markets became illiquid, managing contingent liquidity facilities, and managing liquidity risk, especially with respect to holdings of asset-backed commercial paper (ABCP) to fund off-balance sheet SIVs (Burns 2008). Moreover, while banks were subject to orderly liquidation by the FDIC, there was no orderly liquidation process for nonbank financial institutions. Values of subprime mortgages and the derivatives based on these fell too fast during the financial crisis and could not be easily valued once the market froze. This was incompatible with the long period needed to unwind a business under the Bankruptcy Code. To unfreeze financial markets, the Fed was left with no choice but to extend emergency loans to nonbanks, and Congress eventually authorized the use of taxpayer’s money to bailout financial institutions. To enhance transparency in how SIFIs are interconnected, the OFR will create publicly accessible databases of financial companies and financial instruments, including financial transaction and position data, which identifies counterparties, subject to confidentiality restrictions. To enhance transparency in the derivatives market, in particular, DFA incentivizes the clearing of derivative contracts through a centralized counterparty by reducing the capital requirement compared to those traded over the counter. Central clearinghouses are subject to disclosure, capital and margin requirements. To promote financial stability and address the perception that large financial institutions are “too big to fail,” Section 165 of DFA establishes limits on growth of financial institutions, which in effect, also limits industry concentration. These are (a) prohibition of a merger or consolidation or acquisition that would result in a company’s consolidated liabilities exceeding10 percent of the aggregate consolidated liabilities of all financial companies, and (b) Fed approval of a SIFI’s acquisition of Proceedings of the 2012 Pennsylvania Economic Association Conference 134 a financial company with $10 billion or more in consolidated assets. Management and director interlocks between unaffiliated SIFIs are prohibited, except if temporary in case of a merger, acquisition or consolidation. A SIFI will be subject to an annual stress test by the Fed, and is also required to do its own stress test at least once a year. The results of these stress tests are to be made public to increase transparency. In addition, SIFIs are required to maintain a countercyclical capital buffer to build up an extra layer of capital or reduce leverage in times of excessive aggregate credit growth that, in the judgment of bank regulators, elevates system-wide risk. DFA empowers the Fed & FDIC to set standards for risk management and crisis avoidance. The FDIC can conduct a special examination of a SIFI deemed not to be in a financially sound condition or recommend enforcement action against a BHC to the appropriate Federal banking agency if its conduct poses a risk to the Deposit Insurance Fund. DFA gives the FDIC back-up authority if such agency does not act. Every year, each SIFI is required to submit a Resolution Plan or ‘living will’ specifying remediation efforts for going concerns under conditions of severe financial distress, or in case of failure or insolvency, how it will be reorganized or liquidated under the Bankruptcy Code within a reasonable period of time and in a manner that substantially mitigates serious adverse effects on the financial system. In the case of foreignowned banks, the required resolution plan is limited to its U.S. operations. The Council prescribes a minimum content for this Resolution Plan, among which are: (a) mapping of the interconnectedness of the domestic and foreign core business lines and critical operations that, if disrupted, would materially affect the funding or operations of the firm; (b) unconsolidated balance sheet and a consolidating schedule for all entities that are subject to consolidation, detailing liabilities, collateral management, of-balance sheet exposures, trading and derivatives activities, material hedges and hedging strategies mapped to a legal entity, major counterparties of the company and the likelihood that their failures would result in the firm’s material financial distress or failure, and trading, payment, and settlement systems; (c) strategic plan for rapid and orderly resolution in the event of material financial distress or failure, including funding, liquidity, capital needs and available resources for each core business line and critical operation, strategies to maintain operations and funding of its material entities and ensure that any bank subsidiary will be adequately protected from the activities of any nonbank subsidiary. The first set of Resolution Plans were submitted in July 2012 by the largest eight banking organizations with $250B or more in assets , namely: Bank of America Corp, Barclays PLC, Citigroup Inc., Credit Suisse Group AG, Deutsche Bank AG, Goldman Sachs Group Inc., JP Morgan Chase & Co., Morgan Stanley & Co LLC, and UBS AG. The middle-tier nonbanks (with $100~$250 billion in assets) are required to submit theirs in July 2013, while the lowest –tier nonbanks (with $50~$100 billion in assets) have up to the end of 2013. In the event of change in conditions or circumstances that render the resolution plan ineffective until revisions are made, the FDIC and the Fed can require an interim update. Once received, both the Fed and the FDIC will review the plan for credibility, i.e. the resolution strategies are based on information and data that are verifiable and employ reasonable projections assuming (a) financial distress at a time when the entire financial system would also be under stress, and (b) no extraordinary support by the U.S. or any other government entity would be available to prevent the company’s failure. If the Fed and FDIC jointly determine that the plan is not credible or is deficient, the SIFI will be required to fix the deficiencies and resubmit a plan within 90 days. If the deficiencies are not remedied, the FDIC and the Fed will impose more stringent capital, leverage or liquidity requirements. If within two years following the imposition of such requirements, the revised resolution plan is still deficient, the Fed and FDIC can either restrict growth, activities, and operations or mandate divestitures. Triggers for remediation such as capital levels, stress test results, and risk-management weaknesses will be used by the FDIC and the Fed to identify a SIFI that will subject to early remediation requirements (similar to the FDIC’s Prompt Corrective Action authority). Remedial actions will include restrictions on growth, capital distributions, executive compensation, and sale of assets and common stocks. Extension of FDIC and Fed safety net to nonbanks The runs on shadow banks (mortgage brokers, finance companies, money market mutual funds, hedge funds, broker-dealers, and investment banks) eventually led to a deleveraging process that caused banks to mark down their holdings of subprime mortgage-related assets, and liquidity to dry up as large depositors withdrew cash while interbank lending froze. The FDIC had to increase the insurance limit to $250,000, and eventually, Congress agreed to pass the Trouble Assets Relief Program (TARP), authorizing the use of taxpayer funds to augment the FDIC insurance fund and to lend cash to banks throughout the country. While over 85% of TARP funds have been repaid, there was strong public outcry to prevent future use of taxpayer’s money to extend the Fed and FDIC safety net for banks to nonbank financial institutions. Section 619 of DFA, also known as the Volcker Rule, was designed to protects taxpayers and depositors by Proceedings of the 2012 Pennsylvania Economic Association Conference 135 prohibiting insured banks from using deposits and the bank’s capital to a) engage in proprietary trading (trading to profit from near-term price movements), and (b) lend or extend guarantees, and invest more than 3% of tier 1capital on hedge funds and private equity funds. The proposed rule to implement the Volcker Rule lists the following as permitted or non-proprietary trading activities: (1) market making-related activity or holding an inventory of securities to meet reasonably expected near term demands of clients, customers or counterparties; (2) riskmitigating hedging on a portfolio basis (i.e. conducted at a level above that of a specific trading desk) in recognition of a bank’s need to reduce the risk they face from exposures to interest rate, exchange rate, and other risk factors; (3) underwriting that is essential for facilitating equity and debt issuance for raising capital; (4) transactions on behalf of customers such as purchase, sale, acquisition, or disposition of securities and other instruments; (5) transactions in government securities; (5) investments in small business investment companies, public welfare investments and certain qualified rehabilitation expenditures under federal or state tax laws; and (7) certain offshore activities and insurance activity. In contrast, banks are prohibited from so-called ‘bright line’ proprietary trading, which involves the use of the bank’s capital to benefit from future price movements, and in addition, has one or more of the following additional characteristics: (1) the trading activity is organized and conducted for the sole purpose of generating profits from trading strategies; (2) the level of trading is not commensurate with formal market making responsibilities or customer exposure; (3) there is physical and/or operational separation from market making and other operations having customer contact; (4) trades with or use the services of sell side analysts or brokers/dealers; (5) receives or utilizes research or soft dollar credits provided by other broker-dealers; and (6) the compensation structure is similar to those of hedge fund managers and other managers of private pools of capital. As for hedge funds and private equity funds, a bank is permitted to organize and offer such a fund only if: (1) it is part of the bank’s trust, fiduciary, or investment advisory services to customers; (2) the bank’s ownership interest one year after a fund is established or the aggregated investments of the bank in such funds does not exceed 3% of the bank’s Tier 1 capital; (3) the bank does not guarantee, assume, or insure the obligations or performance of the fund; (4) no bank director or employee has an ownership interest in the fund unless he or she is directly engaged in providing services to the fund; and (5) the bank does not share the same name, or variation of the same name, with the fund. If a bank engages in any activity or investment that functions as an evasion, if not a violation, of the Volcker Rule restrictions, regulatory agencies are empowered to ask the bank to terminate such activity or subject it to additional capital requirements and quantitative limits on activities. Full compliance by banks is scheduled for July 21, 2014 with possible three one-year extensions up to or July 21, 2017, to give banks up to five years to divest inventories of illiquid mortgage-backed securities affected by the implementation of the Volcker Rule prohibitions. So far, the writing of the final rule to implement the Volcker Rule has faced repeated delays as regulators grappled with the complexity of the proposal and widespread discontent from the financial industry, reflected in thousands of comment letters from banks and lobbyists concerned about loss of profits, jobs, international competitiveness, and market liquidity. Issuance of the final rule was pushed from September 2011 to July 2012. For permitted trading activities, a bank holding company or bank that has aggregated trading assets and trading liabilities equal to 10% or more of its tier 1 capital and has quarter-end total assets of $1 billion or more will be required to put up a so-called market risk capital to be considered adequately capitalized. No small bank holding companies or banks would be subject to this rule. The minimum required ratio of total qualifying capital (the sum of tier 1 capital and tier 2 capital, net of all deductions) to the sum of credit risk-weighted assets and market riskequivalent assets is 8%. The market risk-equivalent assets is calculated as the new market risk measure multiplied by a factor of 12.5, where the new market risk measure is expanded to include three new market risk components, namely: stressed value-at-risk, incremental risk, and comprehensive risk. There is also a quarterly requirement to publicly disclose the aggregate amount of on-balance sheet and off- balance sheet securitization positions by exposure type and the aggregate amount of correlation trading positions. In addition, for each material portfolio covered by the market risk capital rule, a bank is required to disclose the high, low, and mean value at risk (VaR)based measures over the reporting period, a comparison of VaR-based measures with actual results and an analysis of important outliers, incremental and comprehensive risk capital requirements, the bank's valuation policies, procedures, and methodologies for covered positions (including, for securitization positions, the methods and key assumptions used for valuing such positions), characteristics of its internal models, the approach used by the bank to determine liquidity horizons, and a description of the stress tests applied to each market risk category. ANALYSIS In the previous two sections, I presented provisions of DFA that address the failures of regulation and flaws in the financial system that weakened market participants’ Proceedings of the 2012 Pennsylvania Economic Association Conference 136 exercise of market discipline. Many of the provisions of DFA are steps in the right direction and have the potential to make the financial system more stable. In the discussion below, I will focus on remedies that I find weak or lacking. No credit risk retention for securitized loans Taking oversight of mortgage brokers and finance companies out of state regulators and into uniform regulation at the federal level by the newly created Bureau of Consumer Financial Protection has the potential of preventing the lax underwriting and fraudulent lending that led to the massive growth of subprime mortgage loans. Although many critics blame the Community Reinvestment Act for encouraging banks to lend to lower income Americans, this legislation did not authorize banks to give up sound underwriting practices, nor allow their affiliate mortgage brokers and finance companies to do the same. As in any regulatory agency, the effectiveness of the BCFB depends on the regulatory philosophy of its Chairman, or the President that appoints him/her to that position. If the leadership takes on the position that markets are self-correcting and industry self-regulation is sufficient, the authority to act on bad lending practices given to it may not be exercised. Likewise, a Congress that believes in the same regulatory philosophy could simply starve the agency of funds, hence qualified staff, so that it could not implement its rules. In addition, the requirement that banks and securitizers disclose as much information about the underlying mortgages of a structured product is a good way to increase transparency for investors, but it does not prevent some investors from relying on apparent market evaluation by a ‘herd’ of investors instead of a fundamental analysis of the investment based on information that has been made available for their use. Finally, I find the required minimum capital reserve set at 5% of the credit risk or future loss from defaults of mortgage loans sold to be too low of a potential loss from bad underwriting practices to be an effective disincentive. Regulatory arbitrage and off-balance sheet conduits The revisions of capital adequacy requirements do have the potential to discourage the use of off-balance sheet conduits to reduce capital charge. Given that the baseline capital requirement is only 8%, or 11% on banks with over $50 billion in assets, the question remains: why allow such off balance sheet commitments or transactions at all? For as long as some transactions can be reported off the books, the temptation to hide risky transactions remains. Moreover, making the required capital on bank investment in investment-grade tranches of CDOs greater than that on mortgage and other loans banks originate, could discourage or reduce such investment, but not completely eliminate it. If the benefit of securitization is really to allow banks to transfer the risk outside of the banking system, it would seem that prohibiting such an investment altogether makes more sense. It will also simplify regulatory oversight of banks in that bank examiners will not have to verify that the investment-grade classification of such investment is appropriate. To distinguish prime from subprime mortgage loans, the use of loan-to-value ratios and loan characteristics, instead of credit ratings, seem easy to implement and appropriate. If, hypothetically, a subprime borrower with an outstanding Category 2 loan of $200,00 finds the appraisal value of the house to be lower, say, $140,000 (i.e. the borrower is ‘under water’), the LTV ratio is 143%. The risk-weight is 200%, which would mean the capital charge will be calculated on twice the outstanding balance ($400,000). A 10.5% capital charge (85 plus the capital buffer of 2.5%) means that if the borrower who is now under water and has no incentive to keep the house were to walk away, bank stockholders would absorb the first $6,300 of the $60,000 loss from a foreclosure sale of this house at the appraised value. Provided the probability that a 30% housing market correction is only 10.5% (about once in 10 years), the FDIC insurance fund and bank creditors are protected from this loss. Perhaps, the risk-weights ought to also account for the difference in real estate market volatility, assigning a higher weight to the ten largest metropolitan areas tracked by the Case-Shiller Home Price Index. Loose oversight of derivatives and hedge derivatives Transparency will indeed be enhanced by to finally subjecting derivatives to disclosure, capital and margin requirements, and hedge funds and private equity fund to registration requirements. The requirement and incentives provided to trade credit default swaps and other derivatives that can be standardized through centralized clearinghouses will also increase transparency as far as counterparty exposures and also reduce counterparty risk. Establishing a separate clearinghouse for each type of derivative may, however, increase counterparty risk as well if there are not enough participants in that particular market. Duffle and Zhu (2011) recommend that credit default swaps, for example, be traded with interest rate swaps and other derivatives that are already traded in a centralized clearinghouse, instead of a clearinghouse dedicated to credit default swaps only. Over-reliance on credit rating agencies While capital adequacy requirements on baking organizations are no longer dependent on credit ratings, market investors are still dependent on them. DFA imposes new disclosure requirements that will enhance Proceedings of the 2012 Pennsylvania Economic Association Conference 137 transparency of CRAs rating methodologies and track record. Yet, DFA does not provide market investors with increased choice. Some argue that there is a need to relax entry barriers to increase the number of competing credit rating agencies, or have a public rating option since credit rating has an element of a public good in it, or subject a credit rating agency to a penalty if future default on a security rated turns out to be more than what was implied by its rating (Hunt 2011). Irrational (herding) behavior of market participants Here is one where losses borne by market participants and longer experience in the market can be far better deterrents than government regulation, except, if there is fraud involved or publicly disclosed information is not consistent with generally accepted accounting principles. Compensation schemes rewarded excessive risk-taking Provisions of DFA intended to correct this problem is probably the most criticized in Wall Street (see Earle, 2011; Gold et al. 2011), but a closer examination of the provisions show that these are quite lame. For example, the stockholder’s ‘say on pay’ is embodied in a non-binding vote. Moreover, the ‘clawback’ provision is limited to taking back executive compensation that was based on inaccurate financial statements, but does not provide incentives for financial companies to design compensation schemes that would not reward traders for short-term gains from excessively risky trades that later result in significant losses. Also, for a failed institution that is under FDIC receivership, DFA caps what the FDIC can take back to the last two years’ compensation from past or present senior executive or director, who is substantially responsible for the failure of the firm, instead of giving the FDIC the discretion to determine that amount in proportion to the loss borne by stakeholders of the institution and the Deposit Insurance Fund. No one regulator to monitor and regulate systemic risk In composition, the Financial Stability Oversight Council resembles the Clinton era President’s Working Group, which was supposed to be a forum for heads of various federal financial regulators to coordinate oversight and consult with each other. This group was in existence throughout the build-up of the housing bubble and chose not to regulate the derivatives market. It failed to detect or act on early warning signs mainly because the more influential members believed in less regulation. So, the two-year old Council is susceptible to the same thinking. What makes the Council potentially more effective is that it is supported by an Office of Financial Research which will support it in terms of collecting and analyzing measures of systemic risk, and DFA has taken many institutions that used to be in the shadow banking system under federal regulation and subject to disclosure requirements. It is also empowered to designated large, complex financial institutions as systemically important so that they can be subject to stricter regulation and monitoring by the Fed and FDIC. The fact that SIFIs are subject to higher prudential standards, restrictions on growth, and closer scrutiny should be an incentive to divest certain operations in order to fall below the $50 billion asset cut-off. The fact that no major divestiture in the finance industry has occurred since DFA was enacted suggests that SIFI designation is not enough of a disincentive. As I said earlier, how well early warning signs can be detected, and how well data, including the annual resolution plans, can be analyzed largely depends on the quality of OFR analysts hired, and ultimately how competitive the salaries that the government can pay them are compared to what the private sector pays. A bleak outlook on the Council’s ability to obtain early warning signs in the future is cast by the fact that JP Morgan Chase’s trading loss made public in May, escaped risk management oversight not only of its top management and the Fed staff embedded in the company’s headquarters. It seems more cost-effective to break-up the large systemically important financial institutions by (a) ending the bank holding company status of Goldman Sachs and Morgan Stanley, and (b) repealing the Gramm-LeachBliley Act which allowed the comingling of banking, investment banking, and insurance activities in one financial holding company or banking organization. Interconnectedness and “Too big to fail” The Resolution Plans do function as some kind of a ‘living will’ in that it requires systemically important financial institutions to map the interconnectedness of their lines of business and operations, identify counterparties to significant transactions, and their transactions, as well as financial institutions that might possibly acquire them in times of financial distress. Although there are now limits on the growth of financial institutions designated as systemically important, at their existing size, they are still “too big to fail”. If financial markets should ever freeze up again as happened in 2007~2008, no one can convince me, and I am sure, the executives of systemically important financial institutions, that the Fed or the Treasury, through Congress, would not bail them out. The moral hazard will always be there for as long as that perception exists. Aside from the break-up of large financial institutions mentioned earlier, reducing the interconnectedness of these large institutions will reduce the urgency of a bailout. There is nothing in DFA that credibly reduces current size because the two surviving large investment banks are now bank holding companies within the Fed’s safety net. Proceedings of the 2012 Pennsylvania Economic Association Conference 138 Moreover, the Volker Rule not only stops short of going back to the Glass-Steagal separation of banking from investment banking and insurance businesses, but will engage federal bank regulators in a ‘cat-and-mouse’ game with banks in discriminating between permissible and prohibited trading activities. A study by the Financial Stability Oversight Council completed in January 2011 identified difficulties in implementing the Volcker Rule, some of which are listed below: a) Banking entities can engage in prohibited proprietary trading through inconsistent or incomplete hedging in the context of their market making activities; b) A combination of permitted activities might be used to circumvent the proprietary trading restrictions; c) Discerning the source of a market maker’s profit is challenging, especially for less-liquid financial instruments; d) Measuring the revenue that is attributable to the bidask spread is difficult and not consistently observable especially in illiquid markets. Moreover, to the extent that market makers need to assume some risk to facilitate customer transactions, a portion of trading revenues will always derive from underlying market movements; e) Market makers often deal with other intermediaries in order to manage their overall risk exposure and maintain market continuity. Distinguishing appropriate volumes of inter-dealer trading for market making versus proprietary trading may be a challenge; and f) Measures of “near term” trading accounts and “shortterm” price movements are dependent on market instrument characteristics. What constitutes trading in the near term, therefore, may depend on the characteristics and the trading volume of the particular market. As a result, a trading account would not preclude illiquid financial instruments, such as swaps. g) It would no doubt be difficult and costly to enforce the Volcker Rule, and violations would be costly to adjudicate. A lot depends on the regulatory agency’s budget, independence, integrity, and ability to get ahead of financial innovation and regulatory arbitrage. A second best solution to breaking up SIFIs or going back to the Glass-Steagal separation of banking, investment banking, and insurance businesses, is to establish an FDIC-like insurance fund that can be tapped to bailout a nonbank in distress. The institution of deposit insurance in 1934 significantly reduced episodes of banking crises in the U.S. to two: the S & L crisis of the 1980s and the financial crisis of 2008. Acharya et al. (2011) proposed periodic assessment on nonbanks to build up a fund that can be used to bail out or liquidate a failing nonbank institution. It will serve as some kind of a Pigouvian tax so that the negative externality that the potential failure of a nonbank imposes on the economy can be internalized. I believe such an assessment was proposed when the shape of Wall Street Reform was being deliberated in Congress, but a few Republicans whose votes were needed to pass DFA asked that it be struck down. Another idea that has been presented is to require financial institutions to maintain a certain amount of contingent capital like bonds that would convert to preferred stock or common stock in times of market stress (Coffey 2011). CONCLUSION AND AREAS FOR FURTHER REGULATORY ACTION The Dodd-Frank Act is indeed a good faith effort on the part of Congress and the Obama administration to establish centralized monitoring of systemic risk, close previous gaps in regulation, and increase transparency of complex financial products to enable market participants to exercise market discipline. It addresses many of the key flaws and vulnerabilities that led to the subprime crisis. However, its effectiveness in preventing another financial crisis depends on how adequately the regulatory agencies involved will be funded, the regulatory philosophy of the top leadership of these agencies, which will determine how strictly the provisions will be enforced, and how well the regulated institutions can innovate their way around the new rules. Moreover, the work of financial reform is not complete. While bank regulators have proposed new capital requirements on banks that address credit and market risks, they have not proposed liquidity thresholds. Likewise, work still needs to be done on putting limits so that reliance by financial institutions on short-term funding of longer term assets can be reduced. Lastly, privatization of Fannie Mae and Freddie Mac, which were put under federal conservatorship in 2008, still has to be worked out. It is important that the moral hazard inherent in the perception of government guarantee be removed. At the very least, once privatized, these institutions have to be designated as SIFIs and subjected to the orderly liquidation authority of the FDIC. In its second annual report to Congress, the Financial Stability Oversight Council identified financial structure vulnerabilities that have not been addressed in DFA. One is the risk of runs on money market funds (MMF) due to two core characteristics: (1) lack of mechanism to absorb a sudden loss in the value of a portfolio security without threatening the stable $1.00 NAV; (2) a “first mover” advantage, which gives investors the incentive to redeem as soon as a threat to the value or liquidity of the MMF is perceived. The Council expressed support for either of two recommendation made by SEC Chairman Shapiro: (a) mandatory floating NAV to reflect the true market value of underlying assets; and (b) a capital buffer to absorb losses, possible combined with a restriction to reduce the incentive to exit the fund. A restriction to reduce the incentive to exit MMFs would make these funds less attractive compared to bank deposits. Rather, than impose such a restriction, Proceedings of the 2012 Pennsylvania Economic Association Conference 139 adopting the floating NAV to make these funds more transparent and (not or) requiring a capital buffer would limit risk-taking by fund managers and promote market discipline among investors. repo market, in which MMFs are lenders. The Council also recommended that the CFTC and SEC keep pace with competitive and technological developments in financial markets, such as high speed, automated trading activities and cyber security threats. Another related area where the Council recommends action is the elimination of intraday credit usage in the tri-party Figure 1: Regulatory Structure: November 1999~July 2010 Financial Holding Company National bank State member bank Commercial paper funding corporation State nonmember bank National thrift Non-U.S. commercial bank Special financing entity State thrift Bank holding company Asset management company National bank U.S. federal regulators Broker/ dealer Non-U.S. investment bank Other CFTC OTS State regulator FDIC SEC Non-U.S. regulator Federal Reserve OCC Futures commission merchant Futures commission merchant National bank Life insurance company (broker/ agent/ underwriter) U.S. securities broker/ dealer / underwrite r Mortgage Broker/ Finance Company Non- U.S. securities broker/ dealer / underwrite r Unregulated SRO Source: Government Accountability Office (GAO 2009) Note: The Dodd-Frank Act would add the Financial Stability Oversight Council on top of the Financial Holding Company in this chart, and subject the mortgage brokers and finance companies to federal regulation by the new Consumer Financial Protection Bureau. It also abolished the OTS and transferred its regulatory authority to the FDIC and the OCC (see Legend below). Proceedings of the 2012 Pennsylvania Economic Association Conference 140 Legend: CFTC Commodity Futures Trading Commission FDIC Federal Deposit Insurance Corporation Federal Reserve Board of Governors OCC Office of the Comptroller of the Currency OTS Office of Thrift Supervision SEC Securities and Exchange Commission SRO Self-regulatory organization (financial industry associations) Table 1: Dodd-Frank Act Provisions and Problems that Led to the Housing Bubble and Systemic Risk Problem DFA Provision Securitization • Banks are required to put up capital against no less than 5% of the subprime mortgage loans sold so that they would have “skin Banks had no incentive to observe strict underwriting in the game”; standards because banks could originate subprime • Securitizers are required to disclose information that would mortgage loans, then pass on 100% of the credit risk of enable assessment by purchasers of mortgage backed securities such loans by sale to investment banks and governmentof the quality of underlying loans. sponsored enterprises (Fannie Mae and Freddie Mac). New Bureau of Consumer Protection to: Shadow Banking: mortgage brokers & finance companies • consolidate the responsibilities previously handled by OCC, No Federal regulator of mortgage brokers & finance OTS, FDIC, the Fed, NCUA, HUD, FTC pertaining to consumer companies; regulation by States were uneven and lax. complaints against mortgage-related businesses; • act fast without waiting for Congress to pass a law prohibiting Loose prosecution of mortgage lending fraud. bad lending practices harming consumers. Shadow Banking: off-balance sheet SIVs Banks moved MBS off-balance sheet into SIVs. Only the backup liquidity line was on the balance sheet and subject to a capital charge, although 100% the bank was exposed to the risk of default and potentially illiquidity of the MBS collateral. Credit Rating Agencies • Loose oversight of credit rating agencies who engaged in a conflict of interest by earned consulting fees to advise investment banks on the optimal way to securitize MBS to get desired credit ratings; • Lack of transparency about credit rating methodologies and track record. Derivatives Loose Oversight of Complex or Opaque Derivatives like credit default swaps, asset-backed commercial paper, and collateralized debt obligations led to lack of transparency as to risk of the financial product and counterparty risk exposure. No single regulator of systemic risk Banks are now required to put up capital of no less than 5% of the risk exposure to SIVs. New Office of Credit Ratings within SEC to: • prohibit compliance officers from working on ratings; • require a 1-year look back review when an employee is hired by an underwriter of a previously-rated security; • require public disclosure of rating methodologies and ratings track record; • require use of 3rd party sources not just the institution being rated. • Trading of standardized derivative products in central clearinghouse or exchanges will be required; • SEC and CFTC to collect and publish data to increase transparency of different derivative products, volumes, and counterparties. • SEC and CFTC to impose capital & margin requirements on swap dealers and major swap participants. New Financial Stability Oversight Council (Council) within the U.S. Treasury • Treasury Secretary as chairman with 9 other voting members: Heads of FED, FDIC, SEC, CFTC, FHFA, NCUA, OCC, BCFP, Insurance Expert • has broad authority to monitor & designate Systemically Important Financial Institutions for stricter regulation by the Federal Reserve Board, and in case of failure, orderly Proceedings of the 2012 Pennsylvania Economic Association Conference 141 Transfer of FDIC & Fed Safety Net for Banks to Nonbanks Failure of nontraditional banking operations during the Financial Crisis was a drain to the FDIC Insurance Fund, the Fed’s Emergency Lending Facilities, and Treasury. Moral Hazard & Externality Problems of “Too Big To Fail” • No orderly liquidation process for SIFIs • Financial asset values fall too fast or cannot be easily valued in times of market stress; incompatible with the long period needed to unwind a business under the Bankruptcy Code ; Reliance on emergency loans from the Fed or taxpayer bailout loans from the Treasury. liquidation by the FDIC. Volcker Rule • Protects taxpayers and consumers by limiting banks’ ability to gamble with customer deposits. • Prohibits Insured Banks from the following activities: • Proprietary trading - trading activity in which it acts as a principal in order to profit from near-term price movements [speculative trading]; • Investing, making loans, purchasing assets, extending guarantees to hedge funds and private equity funds, and affiliate/s for which the banking entity acts as investment manager or advisor or sponsor. Annual Living Will & Orderly Liquidation • Fed & FDIC to: set standards for risk management and crisis avoidance; • require the annual submission by each SIFI of a 0Resolution Plan specifying remediation efforts for going concerns under conditions of severe financial distress or orderly liquidation by the FDIC in case of insolvency or failure. Table 2 Changes in Asset Risk-weight Determination Credit Exposures to: Foreign sovereigns, banks, public sector entities Changes in Asset Risk-weight Determination More risk-sensitive treatment using the Country Risk Classification measure produced by the Organization for Economic Co-operation and Development. Risk weight is assigned according to loan terms and loan-to-value ratio instead of credit ratings. Risk weight changed from one risk weight of 35% to weights of 35%~100% (see Table 4). Increased risk-weight from 100% for all commercial real estate exposure to 150% risk weight for certain credit facilities Highly volatile commercial that finance the acquisition, development or construction of real property. real estate Replaces the current external credit ratings-based approach with a formula-based approach for determining a Securitization securitization exposure’s risk weight based on the weighted average risk of underlying assets and exposure’s relative position in the securitization’s structure. Raises the credit conversion factor for most short-term commitments (extension of credit or purchase assets, contingent liabilities, repo-style transactions, standby letters of credit, forward agreements) that cannot be cancelled Off-balance Sheet items unconditionally: Maturity of 1 year or less: raised from 0 to 20%; Maturity over 1 year: 50% Credit enhancement or warranty on assets sold/transferred to 3rd parties: 100% Revises the measure of the counterparty credit risk of repo-style transactions. Removes the cap of 50% risk weight for derivative contracts. Derivative Contracts Transactions cleared through Provides preferential capital requirements for derivatives and repo-style cleared with central counterparties that meet specified standards. Also requires that a clearing member of a central counterparty calculate a capital requirement for its a central clearinghouse default fund contributions to that central counterparty. Residential mortgage Table 3 Proposed Capital Ratios to be Applied to Risk-weighted Assets Aspect of Proposed Requirements Proposed Minimum Capital Ratios Baseline total capital ratio Unchanged at 8 % Tier 1 capital ratio Increased from 4% to 6 % New common equity tier 1 capital ratio 4.5% Unchanged at 3% banks with CAMEL rating of 1; 4% otherwise, but modified by more stringent definition of tier 1 capital. New supplementary leverage ratio required of SIFIs. New common equity tier 1 capital of at least 2.5% must be maintained above the 8% minimum risk-based capital requirement to avoid restrictions on capital distributions and certain discretionary bonus payments. Leverage ratio (Equity to total assets) Capital Conservation Buffer Proceedings of the 2012 Pennsylvania Economic Association Conference 142 Table 4 – Proposed Risk Weights (percentages of 8% baseline capital requirement for Residential Mortgage Exposures Loan-to-value ratio Category 1 Risk-weight Category 2 Risk-weight 60% or less 60 ~ 80% 80 ~ 90% 90% or more 35% 50% 75% 100% 100% 100% 150% 200% 3 ENDNOTES 1 Bear Sterns was acquired by Goldman Sachs, Lehman Brothers filed for bankruptcy and Merrill Lynch was purchased by Bank of America in 2008. 2 The President’s Working Group includes the Treasury Secretary, and the chairmen of the Fed, SEC, and Commodities Futures Trading Commission (CFTC). It was created by President Clinton in 1998 in the aftermath of the 1987 stock market crash, and has since then served as an inter-agency coordinator for financial market regulation and policy issues. The BCBS is a committee of banking supervisory authorities, which was established by the central bank governors of the G–10 countries in 1975. It currently consists of senior representatives of bank supervisory authorities and central banks from the United States, and 25 other countries (Argentina, Australia, Belgium, Brazil, Canada, China, France, Germany, Hong Kong, India, Indonesia, Italy, Japan, Korea, Luxembourg, Mexico, the Netherlands, Russia, Saudi Arabia, Singapore, South Africa, Sweden, Switzerland, Turkey, and the United Kingdom). REFERENCES Acharya, Viral and M. Richardson. ed. 2009. Restoring Financial Stability: How to Repair a Failed Financial System. NJ: John Wiley & Sons. _____, T. Philippon, M. Richardson, and N. Roubini. 2009. The Financial Crisis of 2007-2009: Causes and Remedies. A Bird’s Eye View, in V. Acharya and M. Richardson. ed. Restoring Financial Stability. _____ and Phillip Schnabel.2009. How Banks Played the Leverage Game, in V. Acharya and M. Richardson. ed. Restoring Financial Stability. _____, T. Cooley, M. Richardson, R. Sylla, and I. Walter. 2011. The Dodd-Frank Wall Street Reform and Consumer Protection Act: Accomplishments and Limitations. Journal of Applied Corporate Finance, 23(1), 43-56. Akerlof, George. 1970. “The Market for "Lemons": Quality Uncertainty and the Market Mechanism,” Quarterly Journal of Economics, 84 (3): 488-500. Basel Committee on Banking Supervision. 2010. Basel III and Financial Stability. Speech by Stefan Walter, Secretary General, at the 5th Biennial Conference on Risk Management and Supervision, Financial Stability Institute, Bank for International Settlements, Basel, November 3~4, 2010. Bank for International Settlements. 2009. The Role of Valuation and Leverage in Procyclicality. Committee on the Global Financial System. Publication No. 34, Bank for International Settlements, April 2009. http://www.bis.org/pub/cgfs34.pdf Bernanke, Ben. 2012. The Federal Reserve and the Financial Crisis. Lecture delivered at George Washington University, March 2012. Bullard, James; Christopher Neely, and David Wheelock. 2009. Systemic Risk and the Financial Crisis: A Primer. Federal Reserve Bank of St. Louis Review, September-October 2009, 91(5): 403-17. Coffee Jr., John. 2011. Systemic Risk After Dodd-Frank: Contingent Capital and the Need for Regulatory Strategies Beyond Oversight. Columbia Law Review, 111(4):795-847. Crotty, James. 2009. Structural causes of the global financial crisis: a critical assessment of the ‘new financial architecture’. Cambridge Journal of Economics, 33(4):563-580. Davis Polk and Wardwell, LLP. 2011. Council Releases Proposed Rules on Designation of Systemically Important Nonbank Financial Companies. Client Memorandum. October 17, 2011. Dodd–Frank Wall Street Reform and Consumer Protection Act .2010. Public Law 111-203. July 21, 2010. Duffle, Darrell and Haoxiang Zhu. 2011. Does a Central Clearing Counterparty Reduce Counterparty Risk? Review of Asset Pricing Studies, April 2011. Proceedings of the 2012 Pennsylvania Economic Association Conference 143 Earle, J. E. (2011). The Dodd-Frank Act: Immediate and Longer-Term Impacts on Executive Compensation. Benefits Law Journal, 24(1): 58-69. Fama, Eugene (1970). “Efficient Capital Markets: A Review of Theory and Empirical Work, “Journal of Finance. May 1970, pp. 383-417. Federal Deposit Insurance Corporation. 2012. “Regulatory Capital Rules: Regulatory Capital, Implementation of Basel III, Minimum Regulatory Capital Ratios, Capital Adequacy, Prompt Corrective Action, and Transition Provisions.” Notice of Proposed Ruling, June 7, 2012. Financial Stability Oversight Council. 2011. The Dodd-Frank Act After One Year. Annual Report. July 2011. Retrieved from: www.ustreasury.gov _____. 2012. The Dodd-Frank Act After Two Years. Annual Report. July 2012. www.treasury.gov _____. 2011. Study and Recommendations on Prohibitions of Proprietary Trading and Certain Relationships with Hedge Funds and Private Equity Funds. Completed pursuant to section 619 of the Dodd-Frank Act, January 2011. Frontline. 2009. The Warning. www.pbs.org/wgbh/pages/ frontline/warning/interviews/ Government Accountability Office. 2009. Financial Regulation: Financial Crisis Highlights Need to Improve Oversight of Leverage at Financial Institutions and Across System. GAO-09-739. Hellwig, Martin.2009. Systemic Risk in the Financial Sector: An Analysis of the Subprime Mortgage Financial Crisis. De Economist.157:129–207. Hunt, John Patrick. 2011. Credit Rating Agencies and the Worldwide Credit Crisis: The Limits of Reputation, the Insufficiency of Reform, and a Proposal for Improvement. http://ssrn.com/abstract=1267625. Morgan, Jamie. The Limits of Central Bank Policy: Economic Crisis and the Challenge of Effective Solutions. Cambridge Journal of Economics, 33(4):581-608. Rajan, Raghuram G. The Credit Crisis and Cycle-Proof Regulation. Federal Reserve Bank of St. Louis Review, SeptemberOctober 2009, 91(5):397-402. Pigou, Arthur. 1932. The Economics of Welfare. 4th ed. London: Macmillan. President’s Working Group on Financial Markets. 2008. “Policy Statement on Financial Market Developments”. www.ustreas.gov/press/releases/reports/pwpolicystatemkturmoil_03122008.pdf. Shin, Hyun Song. 2012. Global Banking Glut and Loan Risk Premium. IMF Economic Review, 60: 155-192. Proceedings of the 2012 Pennsylvania Economic Association Conference 144 GOVERNMENT POLICY AND RESULTANT EFFECT ON NICHE INDUSTRIES: THE CASE OF USPS “EVERY DOOR DIRECT MAIL” Brenda Ponsford, Ph.D., J.D. Clarion University William R. Hawkins, M.A., M.A. U.S. House of Representatives Committee on Foreign Affairs Subcommittee for Oversight and Investigations ABSTRACT The USPS has introduced a new service for direct mail advertisers called “Every Door Direct Mail”. The concept is that the advertiser identifies the target neighborhoods and the USPS carrier simply puts a piece in every mail drop on his route. Yet, for something so easy to conceive and implement, the impact is harmful to the private mailing service industry that was created to deal with past USPS bureaucracy and inefficiency. This paper examines the interaction of inefficiency and economic principles, entrepreneurial response to burdensome regulations, and the impact on private firms when a state enterprise suddenly become entrepreneurial itself. The United States Postal Service (USPS) has introduced a new service for direct mail advertisers called “Every Door Direct Mail” (EDDM). The concept is extremely simple. The advertiser identifies the neighborhoods it wants to target, and its printed material is delivered with the day's mail to every address for less than 14.5 cents apiece. The advertiser doesn't need to know names or street addresses; meaning no labeling is needed. The USPS carrier simply puts a piece in every mail drop on his route. Yet, for something so easy to conceive and implement, the impact is revolutionary on the private mailing service industry that was created to deal with past USPS bureaucracy and inefficiency. Direct mail is the most heavily used direct marketing medium, and its popularity continues to grow despite postage increases. The mail is the third largest advertising medium after newspapers and television because it can be directed to a specific target audience and sent cheaply using bulk mailing rates. The largest single expense of direct mailing is postage, and can account for one-third of the total cost of a direct mail campaign. Other costs include designing, writing, printing and packing the mailing --- and buying mail lists. A four-color postcard is the simplest and most common medium and is also the least expensive to mail. Other forms of direct marketing such as brochures, newsletters, catalogs, packets or offers with return envelops can raise costs considerably. For the USPS, direct mail advertising, called “standard mail” by the USPS but also called “junk mail” by many outside the industry, is the second largest source of revenue behind First Class mail. It generates 26 percent of revenue compared to 51 percent from First Class (USPS 2011). Advertising, both First Class and Standard combined, generated $20 billion of the USPS $67 billion of total revenues in 2010 (30 percent). Paul Vogel, president and chief marketing officer for the USPS has stated, "We believe it could be a billion-dollar product for the Postal Service by 2016 (Liberto 2012.) The USPS estimates the new direct mail “saturation” program will bring the struggling agency some $750 to $800 million in 2012. Its simplicity is attractive to advertisers, especially small business; but it also provides the USPS the opportunity to wipe out a class of private middle men who have made direct mail profitable by working to overcome past USPS bureaucratic practices that made using direct mail more difficult and expensive than it needed to be. Until the advent of EDDM, the USPS refused to delivery any mail that did not have a correct mailing address. As the USPS (2012a) has stated in its promotion of EDDM, “ In the past, if a company mailed books or fliers to a city route, specific addresses had to be imprinted on the pieces. However, for rural routes that wasn’t necessary. All that had to be printed was “Postal Customer.” Now, under EDDM, all mailings can be distributed by Postal carriers without specific addresses because both city and rural routes are treated the same. Letter carriers deliver supplements along with the day’s mail to every door that businesses want to reach. Proceedings of the 2012 Pennsylvania Economic Association Conference 145 The typical small business that wanted to do direct mail advertising found it difficult to compile an address list over any substantial area. This created a market for specialty firms to go to the expense of compiling address lists, and keeping them up to date. Such firms would then sell printed address labels for designated zip codes to advertisers. This was a practical example of the principle of economy of scale. The larger mailing service firms would not just compile and sell lists in their own communities, but would have data bases for entire regions or even the entire country. It also became an example of the principle of barriers to entry because of the expensive process using obsolete methods by which the USPS “helped” the mailing services keep their lists up-to-date. If a mailing service managed to collect 90 percent of the addresses on a carrier route, the USPS would provide the missing 10 percent to complete the list. The process for doing this was cumbersome and primitive. The mailer’s list would be sent to the local post office in the form of cards, one for each address. The cards would be the old “don’t fold, spindle or mutilate” punch cards that could be read by a computer. These obsolete cards were convenient, however, because the list was handled at the local level by each individual letter carrier. The carrier would first put the cards in the order that he walked his route, then add cards for the missing addresses. The cards for the missing addresses would be filled out by hand by the carrier. The deck of cards would be shipped back to the mailer where staff would make new punch cards for the new addresses and run the complete deck of punch cards through the computer to generate an undated, carrier routed list from which labels could be printed for sale. Shipping a box of cards to every carrier in the covered area several times a year was an expensive and time consuming process, especially for the larger, national mailers. There was also a need to maintain secure warehouse space to store the decks of cards. To qualify for discounted commercial postage rates, address lists need to be updated as frequently as ever 90 days. There is no USPS charge for the manual ALSS other than the cost of shipping the cards to the proper post offices. Placing decks into the hands of carriers was necessary because until fairly recently there was no central list of addresses in the USPS at any level above the carrier. Only the carrier knows who was on his route and how he chose to walk it. The Address List Sequencing Service has not changed in decades. It is currently described on the USPS (2012b) website as follows: Address List Sequencing is a service provided to mailers who have mailing lists containing at least 90 percent—but not more than 110 percent— of the total possible deliveries in a 5-digit ZIP Code™ area. The Postal Service receives cards from the mailer containing mailing addresses and sequences the addresses in delivery order. The mailer can then place mail in sequence and present it to the Postal Service as carrier-ready mail. The Postal Service does not compile mailing lists: Address List Sequencing is designed primarily for use in high-density or saturation mailings to a particular area for mailers who already maintain a list. The USPS also offers a service that can check the accuracy of mailer software in properly coding address lists (CASS). Mailers must pass with a minimum score of 98.5 percent to be certified by the USPS and thus be eligible for discounted postal rates. Established mailing service firms have opposed any modernization of the USPS process than would lower costs and thus make it easier for new firms to enter the industry or for advertisers to compile mailing lists on their own. There has been a fear among mailers for two decades that the USPS would compile and digitize in some central manner all the address information for a given zip code. In other words, create accurate lists that could compete with the lists compiled by mailers. The USPS could then sell this information itself to advertisers, cutting out the middle men. What would be logical than for the USPS to know to whom it was delivering and to provide this information as part of its service? But the USPS is prevented by law from providing any address lists to the public (Title 39 of the U.S. Code). It can only due updates of listed that are 90 percent compiled by a mailer. The USPS can now do Electronic Address Sequencing. Lists are sent to the National Customer Support Center (NCSC) in Memphis, TN. Mailers must still compile the initial list themselves. It is expected that the mailer will have already used the ALSS process in putting their list together and will have certified it via the CASS process before submitting it to EAS. Files are limited to 500,000 records per file. Customers are limited to submitting three files per day. First time customers using the EAS process may submit only one ZIP Code on their first submission to ensure their file meets processing requirements. There is a fee charged by the USPS for the EAS service. To protect their lists, mailers sell one-time use labels to customers. To prevent the labels from being copied and resold, mailers “seed” their lists with fake addresses unique to their business. Material received at a seeded address will be reported to the mailer who will then know that its list is being used by someone else without authorization. The USPS has aided the security of mailers by refusing to update lists found to have seeded addresses from list owners other than Proceedings of the 2012 Pennsylvania Economic Association Conference 146 the firm that has submitted the list for updating. As the USPS (2012c) states: If a seed address is identified in the qualification process, the CDS List Owner and the mailer will be notified. All ZIP Codes and/or address groups containing seed addresses will be disqualified. Customers will not receive any address information from the NCSC for any ZIP Codes and/or address groups disqualified due to the presence of seed addresses. By dropping the need for address information to use the new EDDM service, the mail service industry is effectively out of the mass direct mailing business. For over forty years entrepreneurs built an industry that only existed because of bureaucratic behavior by the USPS, reinforced by lobbying by the private mailing service industry against modernization. But under the pressure of being made a quasiprivate business itself and facing massive deficits as its legal monopoly in First Class mail has been undercut by email and online bill paying and its package service was forced to compete with private hyper-efficient firms like FedEx, the USPS has had to become more entrepreneurial itself. It has circumvented the legal ban on selling address lists in competition with private mailing service by devising an EDDM that does not use addresses at all. Rather than marvel at the innovation of the USPS, the better question is why were addresses needed at all for advertisers who wanted their material to be delivered to every stop? The resources of an entire industry were wasted on doing something that really never needed to be done. REFERENCES Liberto, Jennifer 2012. “Postal Service: We Need More Junk Mail” CNN Money, March 21, http://money.cnn.com/2012/03/20/smallbusiness/postalservice-junk-mail/index.htm USPS 2011. Postal Service Revenue: Structure, Facts, and Future Possibilities, U.S. Postal Service Office of Inspector General, Report Number: RARC-WP-12-002, October, p. 2. http://www.uspsoig.gov/foia_files/RARC-WP-12-002.pdf USPS 2012a. “Hit the Target with Every Door Direct Mail,” USPS website, http://www.delivermagazine.com/2012/01/hit-the-targetwith-every-door-direct-mail/ USPS 2012b. Address Quality Services, United States Postal https://www.usps.com/business/address-qualityService, services.htm USPS 2012c. Electronic Address Sequencing (EAS) User Guide, USPS, January 2012. https://ribbs.usps.gov/eas/documents/tech_guides/EAS_USE R_GUIDE.PDF Proceedings of the 2012 Pennsylvania Economic Association Conference 147 FOREIGN AID EFFECTS ON GROWTH IN LATIN AMERICA Tai McNaughton Clarion University of Pennsylvania Clarion, PA 16214 ABSTRACT Although Latin America has had much potential for growth, the countries still remain burdened by low growth and high corruption rates. To combat the low growth rates, foreign aid was given to the region. To find the true effect of aid on the region, several regressions were estimated where the dependent variable is income. Due to the controversial nature of aid, the initial hypothesis was that foreign aid could either incite growth or hinder it in the Latin American region. The empirical evidence brought forth from this panel study suggests that foreign aid negatively effects growth in the region of Latin America. This paper expands on what previous literature has done by not only adding new important variables to the empirical pool of evidence, but also by accounting for country effects to capture factors that are unique to each Latin American country. Once random effects are accounted for, foreign aid injections are shown to negatively impact growth. Latin America has had stagnating growth throughout the past century. The countries in this region have been receiving Foreign Aid to try and help reinvigorate the growth process. It has been generally accepted that giving a country Foreign Aid helps boost growth. But several economists, including Easterly (2003), have not taken this statement as a given truth. This study attempts to see the true effects of Foreign Aid on growth in Latin America. More years, countries, and control variables are added to this panel data study than previous ones. Furthermore, Latin America has typically been studied from the inequality and corruption perspective. Few studies of the region have shown the effects of aid. Throughout the literature, it was found that Aid is affected by corruption and inequality. For these reasons, the initial assumption was that Foreign Aid could either incite or hinder growth it in the Latin American region. For this study, Foreign Aid was regressed on a growth measure while controlling for other important regional indicators of growth. The random effect results suggested that Easterly (2003) was correct when he stated that foreign aid negatively impacts growth. measure of economic success. However, Angeriz also states the importance of using a per capita measure of income for measuring economic success, as it can be a better indicator of a standard of living for the individuals of a country. Crespi (2012) also uses per capita income as a measure of growth. Andres’ (2011) study used the natural logarithm of GDP per capita as their proxy for standard of living. In a study by Deutsch (2011), the standard of living in the countries of Latin America was approximated, using a durable goods index instead of a traditional variable measure used for standard of living. Deutsch concluded that “the standard of living increases with the size of the city, the education level of the individual and his/her age” (584). Foreign Aid Foreign Aid effects are very controversial. Easterly (2003) disagrees with the widely promoted statement that aid acts as a catalyst for growth. Mosley (1987) suggests that there is a micro-macro paradox, where aid is effective at the microeconomic level but not the macroeconomic level. Furthermore, the effectiveness of aid at the microeconomic level has been proven to positively affect growth (World Bank, 2008). However, Bornschier (1995) found that aid has a short term effect on increasing the rate of economic growth. Recently, Arndt (2010) finds that foreign aid, over the long run, has a positive effect on growth in developing countries. Contrastingly, Rajan and Subramanian (2008) concluded that there was no systematic effect of aid on growth. Most often the explanations for this are for political reasons as “aid inflows can weaken governance, for example, by increasing the returns to corruption and/or increasing rent seeking activities.” (Arndt, 2010, p. 3) Corruption is a huge problem facing the Latin American region (Andres, 2011). Corruption has been measured by the ICRG and the Freedom from Corruption Index. Furthermore, the study by Rajan (2008) confirms this as it was shown that some of aid is directed at consumption and activities that are not necessarily for growth. LITERATURE REVIEW Investment Standard of Living According to Angeriz (2011), Gross Domestic Product (GDP) is widely used, and is a popular indicator as a Investment in education and infrastructure are both needed for development (Astorga, 2011). As a proxy for education, Deutsch (2011) used the number of years of Proceedings of the 2012 Pennsylvania Economic Association Conference 148 schooling. Investment in education can also be used to measure education. As a proxy for education, Astorga (2011) used the investment share of GDP as well as literacy rates. According to Puryear (2011), education investment is “widely agreed to be one of the most powerful tools for reducing inequality” (p. 124). Furthermore, stagnating taxes caused a lack of funds for education, as well as for new infrastructure, which furthers inequality in the region (Astorga, 2011). According to Astorga (2011), low levels of investment in infrastructure and human capital are at the root of the low productivity performance in Latin America. Crespi (2012) agrees as it is stated that productivity is needed for developing countries to grow. According to Astorga (2011), The most successful period in the Latin American countries economies were led by “public provision of economic infrastructure and human capital” (p. 220). Unfortunately, a according to Crespi (2012), low productivity growth is said to be the biggest challenge facing Latin America. On the positive side, Latin America has recently started increasing its spending in the education sector (Puryear, 2011). Even with the increase in spending, it is still ranked below the average for educational spending in the world. However, spending does not account for quality. The governments of Latin America have been focusing their efforts on increasing enrollment while failing to address the quality of the education. Sadly, even though access to education has increased, the low quality of education is said to be failing to significantly contribute to increasing human capital. Inequality According to a study by Goni (2011), social equity should be at the top of this region’s development agenda. Furthermore, he states that high inequality is absolutely detrimental to development and prosperity as income inequality often translates to higher poverty levels. It is said that income inequality leading to poverty is said to be an underdevelopment trap. A proxy for income inequality is measured most often by the Gini Coeffcient (Andres, 2011). Ethnic minorities form the majority of disadvantaged groups in these societies (Deutsch, 2011). Family background is highly associated with opportunity deprivation. According to Goni (2011), high inequality leads to conflict, political instability, and increased social tensions which all hinder growth. According to Goni (2011), the fiscal system in developing countries is said to be of little to no help in combating inequality. In Latin America there is weak fiscal redistribution as cash transfers are used and there are low levels of tax collection which are said to be below the international norm. However, transfers are said to help more with fiscal redistribution, and the role of tax is relatively minor. Tax collection volumes in Latin America have been increasing steadily for years, but still are not where they should be due mainly to inefficiencies including evasion and concessions. The study by Goni concluded that Latin Americas failing fiscal redistribution can be explained by a low volume of resources being collected and transferred, tax collection being too regressive, and transfers are poorly targeted. Possibly more importantly, a study by Andres (2011) concluded that tax corruption leads to the stunting of the education and health programs whose funding is being taken away for personal gains. Macro-Economic Variables According to Andres (2011), like most models of Latin America, variables that are also included are ones such as openness of the economy measured in trade, FDI, and inflation. While most economists agree that FDI boosts growth, Bornschier (1995) suggests that FDI over the long run decreases the rate of growth. Perez insinuates that exchange rate appreciation will help the country develop faster. He goes further to state that exchange rate policies are needed to promote expansion (Perez, 2010). Most economist agree that openness to international trade will result in greater productivity but Astorga (2011) points out that trade integration is negatively correlated with growth in this region. Also, different sectors were used for growth during different times in history. However, not one type of export sector is dominant across the region. Separate regions of Latin America produce different exports, some export of labor while other regions export commodities (Perez, 2010). DATA AND METHODS Both pooled least squares and country effects were estimated with EVIEWS software with the purpose to see if Economic Growth is dependent on these eleven variables: Foreign Aid, Health, Population, Exchange Rates, Trade, Savings, FDI, Education, Inflation, Inequality, and Corruption. The natural log was taken of the measures for Economic Growth, Foreign Aid, Population, Exchange Rates, and Inflation to make the data more comparable across the sample of countries. The education variable was prorated for all of the countries. Measures for the quality of education, infrastructure, as well as the measure of taxes were sporadic at best, completely missing at worst. Therefore, these variables were not included in this study. There are twenty countries included in this study of Latin America. The countries are Argentina, Belize, Brazil, Proceedings of the 2012 Pennsylvania Economic Association Conference 149 Chile, Columbia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Guatemala, Guyana, Haiti, Honduras, Jamaica, Mexico, Nicaragua, Paraguay, Peru, Uruguay, and Venezuela. The following is the regression model used: k LNGDPit = β + β LNOFFDEVit + ∑ β i xkit + ε it 0 k =3 1 (1) LNGDPit : The natural log of GDP PPP per capita for each cross section ( i ) over time ( t ) β : Intercept 0 β LNOFFDEVit : Slope ( β ) of the natural EMPIRICAL RESULTS 1 1 log of Official Development Assistance for each cross section ( i ) over time ( t ) k ∑ β i xkit : Specific explanatory variable ( k ) with k =3 slope ( β i ) from a specific cross section ( i ) over time(t) with independent variables denoted as ( x ) ε it Savings: With a higher percentage of gross savings there should be a higher economic growth rate. FDI: With a higher level of FDI there should be a higher economic growth rate. Education: With more years completed there should be a higher economic growth rate. Inflation: With a lower inflation rate there should be a higher economic growth rate. Inequality: With a lower Gini coefficient there should be a higher economic growth rate. Corruption: With a higher level of Freedom from Corruption there should be a higher economic growth rate. : Random error term( ε ) for each country( i ) over time( t ) Where: Economic Growth (Y) = f(Foreign Aid, Health, Population, Exchange rates, Trade, Gross Savings, FDI, Primary Education, Inflation rates, GINI coefficients, and Freedom from Corruption) Table 1 lists the variable, the variable measure, its source, and whether the natural log of the measure was taken for the study. After adjustments, all the variables begin at the year 1993. Table 2 shows the mean, standard deviation, minimum, maximum and number of observations for each variable. Table 3 shows the correlation matrix. A correlation matrix shows the relationship between two variables. Hypotheses Foreign aid: With an increase in foreign aid there should be a higher or lower economic growth rate. Health: With an increase in life expectancy there should be a higher economic growth rate. Population: With a larger population there may be a higher or lower economic growth rate. Exchange Rates: With an increase in exchange rate there should be a higher economic growth rate as a country’s currency is appreciating. Trade: With an increase in trade there should be a higher economic growth rate. Table 4 shows the results from pooled least square, fixed effect, and random effect estimations where the dependent variable is economic growth measured by the natural log of PPP GDP per capita. These regressions show the coefficients for each variable followed by the p-value directly underneath. Also reported are the number of countries included, number of observations, the adjusted R2, and the Durbin Watson Statistic. The pooled least squares estimations suggest that official development assistance does not contribute to growth. Also, the coefficients for trade, the natural log of inflation and the Gini coefficient were negative. Life expectancy, the natural log of the exchange rate, gross savings, primary education, and freedom from corruption were all positive. These were all expected results. Furthermore, the significance of the natural log of official development assistance, life expectancy, the natural log of population, the natural log of the exchange rate, trade, primary education, the natural log of inflation, and freedom from corruption were all statistically significant at the .15 level. Surprisingly FDI and the Gini coefficient were not statistically significant. However, their economic significance is confirmed. Pooled least square approaches show the overall results of the region. However, when using time series data across countries in a panel study, it must be determined whether differences across countries can be captured by differences in the constant term β0 (fixed effects approach) or whether the constant terms for each country are randomly distributed (random effects approach). This approach to panel estimation is outlined in Greene (2000). According to the country fixed effect approach, the natural log of official development assistance, life expectancy, the natural log of population, the natural log of exchange rates, trade, FDI, and the natural log of the inflation rate effect growth positively while only life expectancy, the natural log of the exchange rate, trade and the natural log of the inflation rate are statistically significant. Savings, primary Proceedings of the 2012 Pennsylvania Economic Association Conference 150 education, the GINI coefficient, and freedom from corruption all effect growth negatively. However, primary education and freedom from corruption are statistically insignificant. Furthermore, due to the results of the Hausman test, the random effect approach best explains the growth of the region. The random effects results suggest that official development assistance, savings, FDI, the natural log of inflation, the GINI coefficient and freedom from corruption effect the growth negatively in the region. None of these variables are statistically significant. Furthermore, life expectancy, the natural log of population, the natural log of the exchange rate, trade, and primary education effect growth positively. Trade and primary education are also statistically insignificant. education, the natural log of inflation and inequality were all consistent with previous literature. On the contrary, savings, FDI, and freedom from corruption conflicts with the literature. Gross savings effecting growth negatively could mean that this economy needs to consume more at this point for growth to occur. Furthermore, at this point in time, FDI does not seem to help with growth. A possible explanation is that an MNE is financing in the host country therefore driving up the interest rate and crowding out domestic savings. The negative coefficient on freedom from corruption implies that more corruption results in more growth due to the way that freedom from corruption is measured. This could have implications that corruption actually helps growth in the region or that there is curvature in the data. CONCLUSIONS/POLICY RECCOMENDATIONS Table 5 shows the expected coefficient sign results with the actual sign results for pooled least squares (Pooled), the fixed effect approach (Fixed), and the random effect approach (Random). Positive signs are denoted as (+), negative signs are denoted as (-), and variables that could have either sign in the region are denoted as (+/-). Due to the controversial nature of Official Development Assistance and Population’s effect on growth, either sign could have been expected. Also, variables that have the same sign as the expected sign are said to have economic significance. An example would be for the resulting sign of life expectancy and savings coefficients in the pooled least squares column. Robustly, several coefficient signs were consistent across all three models. First, the natural log of population and the natural log of exchange rates were consistently positive across all models. Population increases having a positive effect on growth in this region backs development theories for increases in population. In line with previous literature, exchange rate appreciation had a positive effect on growth. Increases in exchange rates relates to increases in purchasing power. Increasing purchasing power allows for more items to be bought with less money therefore inciting people to purchase and consume more. Also consistent across the models was that inequality effects growth negatively. This also backs theories for decreasing income inequality. Because the random effects model better explains this set of data, the other models are no longer concentrated on in this study. According to the random effects approach, the Latin America region has negative effects of foreign aid on growth. This is consistent with theories that oppose aid. Furthermore, life expectancy, the natural log of population, the natural log of the exchange rate, trade, primary Like most studies of developing countries, this one was limited due to insufficient data for several variables and countries. There was also a lack of data reporting for the countries in the Latin American region. Furthermore, the data was also shown to have first order autocorrelation for which it was not corrected. According to the random effect results, foreign aid, measured in official development assistance, was shown to effect growth negatively in the Latin American region. Therefore giving aid with the preconceived notion that it will help the region grow is off base. This backs up theories by Easterly (2003) as he states that aid agencies have been failing as they have been expecting to give money to countries in exchange for sustained growth. Aid has been empirically shown to fail at stimulating growth in the Latin American region. Aid is given under the assumption that it will be used to reinvest in the nation therefore inducing growth. As suggested by Easterly (2007), if the end results of aid were changed to not only be growth enhancing, these large lump sums of money could be used to help the members of a country in other ways. This includes improving the human securities of each and every individual of the country. Reform is needed on the policies of aid to actually help the poor in every country. Other variables that positively affect growth should be increased through policy as well. Life Expectancy can be increased through advancements in medicine. Requiring the spending of money to invest in life saving equipment and techniques would benefit the region. Also, children’s health should be increased. Healthier children can go to school longer and live longer. This could have implications on increasing the population which was also shown to affect growth positively. Education needs to be expanded further. The region is doing an excellent job of increasing Proceedings of the 2012 Pennsylvania Economic Association Conference 151 primary education enrollment. It cannot regress if growth is wanted. Incentives toward enrollment and completion should be expanded. FURTHER RESEARCH For future studies, I would like to get a proxy for the quality of education, infrastructure, and taxes whether this is through doing the research or finding a measure. More data could mean better results. It could also be important to see if secondary schooling and secondary education quality would have a greater impact on growth. Also, I would correct for autocorrelation and possible heteroskadasticity. Furthermore, a reordering of my variables for inequality and freedom from corruption could give different results. Increasing trade and the exchange rate will result in more growth. Generally with a steadier inflation rate, more stable government and economic system, and better terms of trade, the result is exchange rate appreciation. Discentives toward savings, FDI, and inflation could help with growth in Latin America as well. Also, decreasing inequality could have a lasting impact by increasing growth. Restructuring to get rid of such wide disparities in income should be considered. TABLES Table 1. Variable Description Variable Measure Source Growth World Bank Y World Bank Y Gapminder World Bank Unctadstats N Y Y World Bank World Bank Unctad N N N Barro&Lee IMF World Bank Heritage N Y N N Foreign Aid Health Population Exchange Rates Trade Savings FDI Education Inflation Inequality Corruption GDP per capita, PPP (constant 2005 international $) Net official development assistance received (constant 2009 US$)-WB Life Expectancy at Birth(years) Population, Total Exchange Rate crossrates between the United States and Individual Countries Trade (% of GDP) Gross savings(% of GDP) Inward and Outward FDI stock annual (% of GDP) Primary Education Completed Inflation, average consumer prices GINI Index Freedom from Corruption Index Table 2. Descriptive Statistics LNPGDP LNOFFDEV? LIFE? 8.562323 18.97561 69.2634 Mean Standard 0.572299 1.100375 5.408678 50.906 Minimum 6.90407 13.28788 Maximum 9.572384 21.20087 79.188 620 Count 609 593 LNPOP? 15.93171 1.498078 11.89136 19.08824 620 LNEXCH? TRADE? -1.11307 64.5953 5.237882 38.72677 -8.76785 11.54567 24.67823 280.361 620 595 GRSAV? 16.20325 7.581806 -24.0038 40.6815 589 Natural Log (Y/N) FDI? PREDU? 20.63873 22.76774 20.30434 11.95348 0.065406 2.9 60.5 165.3109 611 620 LNINFL? GINI? FRCO? 4.423169 51.88986 34.43125 4.489624 5.442963 14.68867 10 -6.90776 34.42 26.57375 70.81 79 599 278 320 Table 3. Correlation Matrix Proceedings of the 2012 Pennsylvania Economic Association Conference 152 LNPGDP?NOFFDEV LNPGDP? 1 1 LNOFFDE -0.34108 LIFE? 0.709122 -0.36828 LNPOP? 0.470559 0.344469 LNEXCH? 0.073015 0.121607 TRADE? -0.37574 -0.21454 GRSAV? 0.423355 -0.07988 FDI? 0.082025 -0.16464 PREDU? 0.201109 -0.41609 LNINFL? 0.211682 0.037989 GINI? -0.19988 0.405963 FRCO? 0.546838 -0.38323 LIFE? 1 0.083543 -0.29835 0.013314 0.329464 0.385739 0.414463 0.196784 -0.41089 0.554528 LNPOP? LNEXCH? TRADE? GRSAV? 1 0.211349 -0.69334 0.300639 -0.18881 -0.37933 0.3841 0.370313 -0.0749 1 -0.3406 -0.01629 -0.2956 -0.06036 -0.00398 0.077676 -0.06484 1 -0.13415 0.404718 0.30193 -0.14323 -0.09689 -0.06444 1 0.045229 0.060186 0.203784 0.079114 -0.07743 FDI? PREDU? LNINFL? GINI? 1 0.095219 1 0.134612 0.086443 1 0.036342 -0.19483 0.31907 1 0.173973 0.182202 -0.02023 -0.20691 FRCO? 1 Table 4. Pooled, Fixed, and Random Results-Dependent Variable: LNGDP Pooled Fixed Random C 4.425475 2.335439 1.238492 0 0.3878 0.0266 -0.130692 0.009262 -0.007915 0 0.3464 0.3888 0.064815 0.047262 0 0 0 0.141731 0.189705 0.136844 0 0.3113 0 0.01327 0.09483 0.038852 0.0227 0 0.0003 0.001452 -0.002536 -0.001779 0.5599 0.1262 0.2027 -0.003922 0.001757 0.000325 0 0.011 0.5414 0.000667 0.000273 -0.000211 0.3859 0.4933 0.5681 0.002214 -0.001091 0.001627 0.1383 0.7229 0.4304 -0.005564 0.063822 -0.002586 0.0934 0.0178 0.711 -0.002386 -0.007908 -0.002635 0.5238 0.0048 0.2637 0.004631 -0.000649 -8.49E-05 0.0002 0.3714 0.9027 19 19 19 Observations 188 188 188 Adjusted R2 0.841022 0.976953 0.944908 D-W stat 0.419615 0.479277 0.265063 LNOFFDEV? LIFE? LNPOP? LNEXCH? GRSAV? TRADE? FDI? PREDU? LNINFL? GINI? FRCO? Countries Proceedings of the 2012 Pennsylvania Economic Association Conference 153 Table 5. Sign of Coefficient Variable Expected Sign Resulting SignPooled - Resulting SignFixed + Resulting SignRandom + LN of Official Development Life Expectancy +/+ + + - LN of Population +/- + + + LN of Exchange Rates + + + + Savings + + - - Trade + - + + FDI + + + - Primary Education + + - + LN of Inflation - - + - Inequality - - - - Corruption + + - - Proceedings of the 2012 Pennsylvania Economic Association Conference 154 REFERENCES Andres, A. R., & Ramlogan-Dobson, C. (2011). Is corruptions really bad for inequality? Evidence from Latin America. Journal of Development Studies, 47(7), 959-976. Angeriz, A., Arestis, P., & Chakravarty, S. P. (2011). Inequality adjusted growth rates in Latin America. Cambridge Journal Of Regions, Economy And Society, 4(1), 49-62. Arndt, C., Jones, S., & Tarp, F. (2010). Aid, growth, and development: Have we come full circle? Journal of Globalization and Development, 1(2), Astorga, P., Berges, A. R., & FitzGerald, V. (2011). Productivity growth in Latin America over the long run. Review of Income and Wealth, 57(2), 203-223. Bornschier, V., Chase-Dunn, C., & Rubinson, R. (1995). Cross-national evidence of the effects of foreign investment and aid on economic growth and inequality: A survey of findings and a reanalysis. In C. Chase-Dunn (Ed.), The historical evolution of the international political economy. Volume 2 (pp. 48-80). Elgar Reference Collection. Library of International Political Economy, no. 9. Crespi, G., & Zuniga, P. (2012). Innovation and productivity: Evidence from six Latin American countries. World Development, 40(2), 273-290. doi:http://dx.doi.org/10.1016/j.worlddev.2011.07.010 Deutsch, J., & Silber, J. (2011). An ordinal approach to the study of intergenerational opportunities for standard of living: The case of Latin America. Journal of Economic Inequality, 9(4), 579-604. doi:http://dx.doi.org/10.1007/s10888-011-9177-0 Easterly, W. (2003). Can foreign aid buy growth? Journal of Economic Perspectives, 17(3), 23-48. doi:http://dx.doi.org/10.1257/089533003769204344 Easterly, W. (2007). Was development assistance a mistake? American Economic Review, 97(2), 328-332. doi:http://dx.doi.org/10.1257/aer.97.2.328 Greene, William H. (2000). Econometric analysis, 4th ed. Prentice Hall Inc. Upper Saddle River, NJ. Goni, E., Lopez, J., & Serven, L. (2011). Fiscal redistribution and income inequality in Latin America. World Development, 39(9), 1558-1569. doi:http://dx.doi.org/10.1016/j.worlddev.2011.04.025 Mosley, P. (1987). Overseas aid: Its defense and reform. Wheatshead Books: Brighton. Perez Caldentey, E., & Vernengo, M. (2010). Back to the future: Latin America's current development strategy. Journal Of Post Keynesian Economics, 32(4), 623-643. Puryear, J., & Ortega Goodspeed, T. (2011). How can education help Latin America develop? Global Journal of Emerging Market Economies, 3(1), 111-134. Rajan, R.G., and A. Subramanian (2008). Aid and growth: What does the cross-country evidence really show? The Review of Economics and Statistics, 90(4): 643-65. World Bank (2008). Annual review of development effectiveness 2008: Shared global challenges. Independent Evaluation Group, available at: http://go.worldbank.org/U2T30HQKG0 Proceedings of the 2012 Pennsylvania Economic Association Conference 155 OFFSETS IN THE DEFENSE TRADE: COUNTERING COMPARATIVE ADVANTAGE INERNATIONAL BUSINESS Brenda Ponsford, Ph.D., J.D. Clarion University William R. Hawkins, M.A., M.A. U.S. House of Representatives Committee on Foreign Affairs Subcommittee for Oversight and Investigations ABSTRACT The United States has a comparative trade advantage in military equipment. Its foreign customers, however, demand offsets against the cost of imports that prevent the American economy from fully realizing its gains from trade. The Commerce Department has reported that from 2000 to 2010, the value of offset agreements equaled 75.7 percent of the $70 billion in reported military sales to foreign governments. Requirements that American producers buy imports, transfer technology, share work or invest in foreign projects raise a variety of issues for American defense firms and the U.S. government to maximize benefits to the national industrial base. Nobel laureate Paul Samuelson (1964) was challenged by the mathematician Stanislaw Ulam to "name me one proposition in all of the social sciences which is both true and non-trivial." It was several years later that he thought of the correct response: “comparative advantage”. 10 Every student of business and economics knows the famous wine and cloth example of this theory presented by David Ricardo in his 1817 work On the Principles of Political Economy and Taxation. A more concise statement come from the Financial Times (2012) “The idea that a country or region should specialise in making and exporting goods and services that it can produce most efficiently. In turn, the country should import goods and services that it has a comparative disadvantage producing. This should in theory lead to an increase in overall trade.” 11 Americans are told that under the theory of comparative advantage they must accept a loss in competitiveness and employment in a number of consumer goods categories, from electronics to clothing to automobiles, and should concentrate their efforts in other areas of comparative strength. As Samuelson noted in his reply to Ulam, “the thousands of important and intelligent men who have never been able to grasp the doctrine for themselves or to believe it after it was explained to them.” This is particularly true when the United States has been running massive trade deficits for over a decade. The shift of capital and labor between industries to bring the economy back into balance has not been very evident. One of the few bright spots in the U.S. trade accounts has been aerospace, both civilian and military. In 2011, the U.S. ran an overall trade deficit in goods of $726.3 billion, but a surplus in aerospace trade of $42 billion. U.S. arms exports for 2011 again lead the world at $10 billion. 12 For the period 2006-2011, U.S. arms exports totaled $46.5 billion. Yet, when the United States has a clearly demonstrated comparative advantage in a major sector, such as armaments and military equipment, its customers demand offsets against the cost of imported U.S. defense systems that prevent the American economy from fully realizing its gains from trade. Offsets in defense trade encompass a range of industrial compensation arrangements required by foreign governments as a condition for the purchase of defense articles and services from American suppliers. This mandatory compensation can be directly related to the purchased defense article or service or it can involve activities or goods unrelated to the defense sale (indirect offset). In either case, the intent of the offset is to force U.S. exporters to give back a portion of their proceeds from the sale. According to the Commerce Department, during 1993-2010, 52 U.S. firms reported entering into 763 offset-related defense export sales contracts worth $111.59 billion with 47 countries. The associated offset agreements were valued at $78.08 billion. This amounted to a 70 percent give back on the value of the contracts. 13 Direct offsets can involve co-production, licensed production or subcontractor production of all or part of Proceedings of the 2012 Pennsylvania Economic Association Conference 156 the U.S. origin defense product being exported, moving manufacturing from the U.S. to an overseas site. Investment arising from the offset agreement can establish or expand a subsidiary or joint venture in the foreign country. It can be either a direct or indirect offset. Technology transfers occur from moving production overseas or in providing assistance along with capital in creating subsidiaries or joint ventures. Indirect offsets often involve various types of commercial counter trade arrangements including barter, that specifies the exchange of selected goods or services, and counter-purchase - an agreement by the exporter to buy (or to find a buyer for) a specific value of goods from the importer. considers offsets to be economically inefficient and trade distorting, and prohibits any agency of the U.S. Government from encouraging, entering directly into, or committing U.S. firms to any offset arrangement in connection with the sale of defense goods or services to foreign governments. U.S. prime contractors generally see offsets as a reality of the marketplace for companies competing for international defense sales. Several U.S. prime contractors have informed BIS that offsets are usually necessary in order to make a defense sale.” 15 The hands off policy of the U.S. government leaves American firms to face the demands of foreign governments on their own. In the U.S. executive branch, responsibility over defense industrial policy is shared between several departments. The Office of Management and Budget is the interagency coordinator for preparing the annual report for Congress on offset policy. Other agencies involved in the process include the Departments of Commerce, Defense, Labor, State, and the Treasury, and the Office of the U.S. Trade Representative. In 1984, the Congress amended the Defense Production Act, adding Section 309 addressing offsets in defense trade. It required the President to submit an annual report assessing the impact of offsets on the U.S. defense industrial base to Congress. When the DPA was amended in 1992, the Secretary of Commerce was authorized to develop and administer the regulations necessary to collect offset data from U.S. defense exporters. The Secretary of Commerce delegated this authority to the Bureau of Industry and Security (BIS). The 2007 BIS report attempts to counter critics who allege offsets cost American jobs 16. “BIS has developed an estimate of employment impact related to offsets by using U.S. aerospace-related employment and value added data collected by the U.S. Department of Commerce, Bureau of the Census. Based on BIS’ calculations, it appears that 2002-2005 defense export sales had a net positive effect on employment in the defense sector during the four-year period (an annual average of 16,085 work years), says the report. 17 But BIS notes that “These calculations are based on the supposition that this value represents 100 percent U.S. content in all exports, which is not necessarily an accurate assumption.” The calculation also does not take into account the negative impact on U.S. jobs from indirect offsets that by definition affect non-defense sectors. The DPA was last reauthorized in 2003. The amended DPA required the President to designate a chairman of an Interagency Team to consult with foreign nations and domestic firms on the adverse effects of offsets. The Interagency Team was formally established in August, 2004, chaired by the Defense Department, with members representing the departments of State, Commerce, Labor and the USTR. In 2009, the BIS amended its offset reporting regulation to require that companies assign the appropriate North American Industry Classification System (NAICS) code(s) to each offset-related defense export sales contract and offset transaction reported. Prior to 2009, BIS only required classification by broad industry descriptions. The change to NAICS classification reporting has allowed BIS to gather more accurate information to better assess the economic impact of offsets on the U.S. industrial base. The BIS 2007 annual report to Congress acknowledged, “Some legislators in the U.S. Congress have raised concerns about the effects of offsets on the U.S. industrial base, because most offset arrangements involve purchasing, subcontracting, and co-production opportunities for U.S. competitors, as well as transferring technology and know-how.” 14 The BIS dropped this statement in its 2012 report, even though Congressional concern remains. The BIS stated in both 2007 and 2012, “The official U.S. Government policy on offsets in defense trade states that the Government The top four industry sectors reported by industry during 2009-2010 were aircraft manufacturing (NAICS 336411); other guided missile and space vehicle parts and auxiliary equipment manufacturing (NAICS 336419); radio and television broadcasting and wireless communications equipment manufacturing (NAICS 334220); and military armored vehicle, tank, and tank component manufacturing (NAICS 336992). These four categories represented 58.0 percent of all defense export sales contracts reported during 2009-2010 based Proceedings of the 2012 Pennsylvania Economic Association Conference 157 on quantity and 70.3 percent of the defense export sales contracts based on value, totaling $9.8 billion for the four sectors Using NAICS data for both exports and offsets, the BIS found that exports by prime defense contractors during the 2009-2010 period created 45, 576 jobs in American manufacturing; but offsets cost the domestic manufacturing base 23,022, over half the number of jobs created by exports. This is a case of the glass being half empty or half full. If offsets help prime contractors land export deals, the net effect on the U.S. economy is positive. But if offsets are seen as an unfair trade policy by foreign governments meant to lessen the benefit to America from its comparative advantage in defense goods, then the BIS data confirms this effect as well. 18 In the American industries of aircraft parts and aircraft engine parts, total employment is reduced by offsets, the result of the transfer of work from American factories to foreign factories as part of direct offset deals. Many foreign governments demand that some of the work generated by the production of the equipment being purchased be “shared” with the importing country. Shifting parts manufacturing to the overseas client is an easy way to fulfill this demand. Consider the F-35 Joint Strike Fighter (JSF) program. The U.S. planned a total of 2,456 aircraft for the Air Force, Marine Corps, and Navy. The F-35 Lightning II is a fifth-generation fighter-bomber with stealth characteristics, making it the world’s most advanced tactical fighter. The export market for this warplane is expected to be around 1,000 aircraft. Eight countries (Australia, Canada, Denmark, Italy, Netherlands, Norway, Turkey and the United Kingdom) are participating in the JSF development program to facilitate sales while avoiding messy offset agreements. Israel and Singapore have also joined as “security cooperation participants.” But this approach has not been as smooth as hoped. Foreign partners have disagreed over work shares and the transfer of proprietary technology. Denmark, Italy, the Netherlands, Norway, and Turkey expressed dissatisfaction in 2003-2004 with the type and quantity of the work their companies had been awarded on the F-35. These countries threatened to reduce their participation in the program, or to purchase European fighters instead of the F-35. 19 Italy and the UK have lobbied for F-35 assembly facilities to be established in their countries. The British are the lead foreign partner to the U.S. in the development program with a number of UK firms, such as BAE and Rolls-Royce participating. The F-35 tail section will be built in the UK. Yet, the UK has also announced that it will greatly reduce purchases of the aircraft, perhaps only buying 40. In July 2010, Lockheed and the Italian firm Alenia Aeronautica reached an agreement to establish an F-35 final assembly and checkout facility at Cameri Air base, Italy, beginning in 2014. It was also reported that South Korean companies could bid for work on the F-35 if South Korea purchases the aircraft. In November 2009, it was reported that the Confederation of Danish Industries had demanded that the Danish government secure subcontract guarantees with Lockheed regarding Danish work on the F-35 program before the Danish government decides to buy the plane. Some foreign partners in the F-35 program have argued that the United States has been too cautious regarding the transfer of JSF technologies. 20 The motive for direct offsets is to build the capabilities of domestic defense industries and the technological base of the importing country. Murad Bayar, Undersecretary of National Defense for Turkey made the standard case at an offsets conference held in Ankara in 2009. “Local design and development” and co-production are favored, with an emphasis on manufacturing final products rather than parts or components. Turkey wants goods it can export as well as provide on a long-term basis into the supply chains of foreign prime contractors. Turkey is looking to increase industrial output, not cash savings. Local production will be more costly, but worth it to improve local capabilities and create jobs. According to Bayar, the offset agreement counts for 40 percent of the factors that will decide which foreign systems to buy. 21 Countries of all sizes and levels of development demand direct offsets and local production. In the UK, the Ministry of Defense will ask prime contractors to outline what they will do to increase small business participation in their supply chains for government contracts. 22 India’s offset policy, which applies to procurements valued at $65.12 million or higher, requires foreign contractors to invest 50 percent of the deal's value in the Indian defense industry. 23 Even Azerbaijan, whose defense industry was only founded in 2005, has an informal policy that requires many basic defense purchases to include domestic industrial participation. 24 Not all countries have a defense sector robust enough to Proceedings of the 2012 Pennsylvania Economic Association Conference 158 make full use of offsets. And even in larger countries, the need to provide benefits to the civilian sector for political reasons has lead to the increasing use indirect offsets to leverage defense purchases into commercial sectors. Developing countries with less industrialized economies generally pursue indirect offsets to help create profitable businesses and to build their domestic infrastructure. According to the most recent published BIS data, in 2010, direct offsets accounted for 33.10 percent of the actual value of reported offset transactions. Indirect offsets accounted for 63.11 percent of the actual value of reported offset transactions. During 1993-2010, direct offsets accounted for 40.22 percent of the actual value of the reported offset transactions, with indirect offsets accounting for 59.04 percent. 25 In order of value, offsets imposed in 2010 amounted to the following: Indirect purchases of goods, $1.2 billion; Technology transfer, $874.8 million; Subcontracting, $605.6 million; licensed production, $380.3 million; overseas investment, $185.3 million; and Coproduction, $148.3 million. 26 In the commercial sector, the use of offsets is considered an illegitimate distortion of normal trade practices. This view is embodied in the Agreement on Government Procurement (GPA) which was signed in Marrakesh on 15 April 1994 — at the same time as the Agreement Establishing the World Trade Organization. The use of offsets, defined as measures to encourage local development or improve the balance-of-payments accounts by means of domestic content, licensing of technology, investment requirements, counter-trade or similar requirements, is explicitly prohibited in the GPA. However, national security is exempt from the GPA, as it should be because it is too important to be left in a “free trade” environment or subject only to commercial considerations. 27 Only 37 countries have joined the GPA, 27 of who are members of the European Union. None of the so-called developing nations have done so. Of the 28 countries on which detailed information was compiled by the Interagency Team, 16 used offsets from military procurement to aid civilian sectors. And India and Taiwan were thought to be trending towards the use of offsets for civilian purposes. All used both direct and indirect offsets. 28 So the legitimate concern over national security is being used as leverage to extort unfair advantages in the commercial sector which would be illegal if attempted openly. The U.S. attempt to improve the “transparency” of government procurement during the Doha Round of global trade talks failed even before the overall negotiations broke down last year. The WTO General Council Decision of August 1, 2004 dropped this issue from the Doha agenda. According to the USTR, “a number of WTO Members remained concerned that a transparency agreement could lead to market access commitments.” 29 The aversion to expanding even slightly the GPA came even though the agreement has plenty of loopholes. Article V allows states to take into account “the development, financial and trade needs so that they can continue to restrict procurement to “promote the establishment or development of domestic industries....[and] support industrial units so long as they are wholly or substantially dependent on government procurement.” They are also free to negotiate “mutually acceptable exclusions from the rules on national treatment.” Governments have long used procurement to support domestic industry, particularly public infrastructure, national defense and other strategic sectors. It is both good politics and sound economics for money taken out of the economy through taxes to be plowed back into the economy via procurement. Government spending is also a vital stabilizer during the business cycle, as the United States is again demonstrating with its rescue packages aimed at the financial system and the auto industry, and the large number of proposals for new stimulus packages and infrastructure programs. The drawn out recovery from the deep world recession will only intensify the competitive pressures that finally sank the troubled Doha Round over issues of market access and support for strategic agricultural and industrial sectors. Among the findings of the InterAgency Team from its dialogues with European governments (France, Germany, Italy, Spain, Sweden and the United Kingdom) in 2007, “There is no common national view. There are differences in views between national defense sectors and government departments/agencies” 30 In many countries, the acquisition of the best available military capabilities, the priority of defense ministries, has taken second place behind the desire to develop broader industrial policies favored by non-defense ministries to benefit the national economy or influential special interests. To the extent that offsets are the “deal maker” in bids for weapons systems, especially indirect offsets, then governments are indicating a willingness to sacrifice defense capabilities for other economic benefits This division of interests can become Proceedings of the 2012 Pennsylvania Economic Association Conference 159 particularly marked in regions where the threat of major war is thought to be remote. Compare, for example, offset policies in Europe to those used in regions such as the Middle East and Africa were imminent conflict is a more serious concern. Sacrificing military capability in a war zone would not be rational, so choosing the best weapons to purchase has top priority. Despite some annual fluctuations, the average offset percentage demanded by the 25 European countries involved in offset activity during the 14-year (19932006) reporting period was 97.7 percent of the export contract values. These percentages reached a peak of 153.3 percent in 2003. In 2006, the European average offset percentage increased from 83.7 percent in 2005 to 85.5 percent. Many European governments require a minimum of 100 percent offsets on purchases of foreign defense systems. Of the 313 offset agreements with Europe during the 14-year period, 206 (65.8 percent) had offset percentages of 100 percent. Another 27 agreements specified offset percentages of more than 100 percent, including two for which the offset percentage was at least 200 percent. In sum, 74.4 percent (by number) of offset agreements with Europe featured offset percentages of 100 percent or more during 1993-2006. Unfortunately, the BIS stopped reporting data by country or even geographical region after 2006, releasing information only in terms of global totals. The 18 non-European countries surveyed by the BIS for the same 1993-2006 period accounted for 34.2 percent of offset agreements (by value), but more than half (52.1 percent) of the value of reported U.S. defense export contracts. The non-European countries’ average offset requirement for the 14-year reporting period was 46.7 percent of contract value. Offset demands on U.S. defense firms by the Middle East and Africa region ranged between 35.3 percent and 55.3 percent, with an average percentage during the 14-year period of 43.7 percent of the export contract values. The Middle Eastern and African average offset percentage increased from 43.2 percent in 2005 to 50.3 percent in 2006. From an American perspective, it is clear that the main objective of foreign offset demands is to transfer production capacity and technology from the United States to industrial centers overseas. Co-production transactions have involved the construction of major facilities overseas for the assembly of entire defense systems, such as aircraft, missiles, or ground systems. Co-production arrangements of this kind generally impose a high cost on the foreign government, including up-front construction and tooling costs and increased unit costs for limited production runs. Some countries negotiate with prime contractors for production or assembly contracts related to future sales to third countries of the defense systems or system components in order to justify these investments. U.S. prime contractors sometimes develop long-term supplier relationships with overseas subcontractors based on short-term offset requirements. These new relationships, combined with mandatory offset requirements and obligations, can endanger future business opportunities for U.S. subcontractors and suppliers, with possible negative consequences for the domestic industrial base. Other kinds of offsets can increase research and development spending and capital investment in foreign countries for defense or nondefense industries. Such offsets can also help create or enhance current and future competitors for U.S. subcontractors and suppliers, and in some cases prime contractors. 31 The final report issued in 2006 by the Interagency Team concluded that there were several adverse effects from the use of offsets. The first was obvious, direct offsets reduce the near-term benefits of the sale by reducing the amount of domestic work supported in the United States. Offsets are not free; estimates indicate that they increase the price of defense equipment by as much as 15 to 30 percent. Certain types of offsets distort the ability of the provider to fulfill the offset requirement in accordance with best business practices, especially those demanded for political reasons or as a source of foreign aid or economic assistance. Indirect offsets may create business incentives for prime contractors to place future defense work in foreign countries that would otherwise be performed by U.S. domestic subcontractors. 32 Other major military equipment exporters have found similar problems. Patrick Colas des Francs, chief executive of the French firm Coges, recently said, “Offsets are a threat to small and medium-sized companies. This is a real problem.” About 10 large companies dominate the French land arms industry, but about 4,000 small and medium-sized companies depend on them for sub-contracting work. Demands by countries such as Brazil and India for the transfer of technology to set up local assembly of the Rafale may be acceptable to large companies such as Dassault, which builds the fighter jet, Colas des Francs argued. “Offset deals not only take production work away from French subcontractors and suppliers, but also provide accelerated access to knowledge and skills that allows companies in the client country to compete in world Proceedings of the 2012 Pennsylvania Economic Association Conference 160 markets,” he said. 33 The 2011 BIS report takes an even-handed approach to the question of the impact on the U.S. industrial base of offsets, "Defense export sales can be an important component of U.S. defense contractors’ revenues and further U.S. foreign policy and economic interests. Exports of major defense systems can also lower overhead and unit costs for the Department of Defense (DOD); and help sustain production facilities, workforce expertise, and the supplier base to support current and future U.S. defense requirements. Exports also promote interoperability of defense systems between the United States and friends and allies and contribute positively to U.S. international trade account balances. However, offset agreements and associated offset transactions can negate some of the potential economic and industrial base benefits accrued through defense exports if the offset activity displaces work that otherwise would have been conducted in the United States and/or if competitors are established in foreign countries” 34 A paper on offsets was published in April, 2008 by the Economic Policy Institute, a think tank that is very influential in the Democratic Party. It was written by Owen E. Herrnstadt, director of the Trade and Globalization Department of the International Association of Machinists and Aerospace Workers. 35 The paper argues that offsets have reached such a high level that they need active oversight, including a review process that can limit or block such transactions. Herrnstadt suggests an expanded role for the U.S. Export-Import Bank, an agency created to assist in financing the export of U.S. goods and services. The Ex-Im Bank has as its objective “to contribute to maintaining or increasing employment of United States workers,” and it has Congressional authority to conduct economic impact analysis and reviews of “exports of capital goods and services (e.g., manufacturing equipment, licensing agreements) that will result in the foreign production of exportable goods.” Since many offset agreements result in the substitution of foreign production for American export goods, the Ex-Im Bank already has the basic authority that can be expanded to change U.S. policy from benign neglect to proactive supervision and control. The Ex-Im Bank is due for reauthorization by Congress this year. Another approach would be to build on the Committee on Foreign Investment in the U.S. (CFIUS) model. CFIUS is a multi-agency committee, co-chaired by the Treasury and Homeland Security. It was created in 1988 to analyze foreign acquisitions of privately-owned American entities. It has the authority to modify or block any deal that is deemed to pose a risk to national security and the definition of what constituted such a danger was expanded by the Foreign Investment and National Security Act of 2007. Recreating and expanding the role of the Interagency Team, led by the Defense Department and including the State and Commerce Departments plus the office of the USTR, into an oversight committee with the power to modify or block offset agreements could be done fairly quickly. It would provide a flexible organization that could deal with issues on a case by case basis as an alternative to a more restrictive legislation. Proceedings of the 2012 Pennsylvania Economic Association Conference 161 ENDNOTES 1 Paul .A. Samuelson , 1964. "The Way of an Economist," in P.A. Samuelson, ed., International Economic Relations: Proceedings of the Third Congress of the International Economic Association, Macmillan: London, 1969 pp. 1-11. 2 Financial Times Lexicon, 2012. http://lexicon.ft.com/Term?term=comparative-advantage. 3 Paul Holtom, Mark Bromley, Pieter D. Wezeman and Siemon T. Wezeman, Trends in International Arms Transfers 2011, Stockholm International Peace Research Institute, March, 2012. The world’s second largest arms exporter was Russia at $7.9 billion in 2011 and $35.7 billion over the 2006-2011 period. France was the third largest exporter in 2011 at $2.4 billion, but Germany was third for the 2006-2011 period at $14.4 billion. http://armstrade.sipri.org/armstrade/html/export_toplist.php 4 U.S. Department of Commerce, Bureau of Industry and Security, Office of Strategic Industries and Economic Security, Offsets in Defense Trade: Sixteenth Study, January 2012, p. 3. 5 U.S. Department of Commerce, Bureau of Industry and Security. Office of Strategic Industries and Economic Security, Offsets in Defense Trade: Twelfth Study, December, 2007, p. iii. See also Defense Production Act Amendments of 1992 (Pub. L. 102-558, Title I, Part C, §123). http://www.bis.doc.gov/defenseindustrialbaseprograms/osies/offsets/default.htm. 6 Offsets in Defense Trade: Sixteenth Study, 2012. p. 1. 7 William Matthews 2007. “Dodd Seeks Law Overhaul to Protect High-Tech Base” Defense News, January 29, p. 14. Sen. Christopher Dodd (D-CT) had claimed that offsets cost the defense industry 10,000 jobs annually, with even more jobs lost in the manufacturing supply chain. Connecticut is home to several major defense contractors, including jet engine maker Pratt & Whitney, helicopter manufacturer Sikorsky, submarine builder Electric Boat and aircraft producer Northrop Grumman, 8 Offsets in Defense Trade: Twelfth Study, 2007. p. v. 9 Offsets in Defense Trade: Sixteenth Study 2012. p. 13. 10 “Norway Signs Industrial Partnership with Eurofighter Consortium,” Defense Daily, January 29, 2003. Joris Janssen Lok, “Frustration Mounts Among JSF Partners,” Jane’s Defence Weekly, March 24, 2004; Thomas Dodd, “Danish Companies Consider Quitting JSF Programme,” Jane’s Defence Weekly, January 9, 2004. Tom Kingston, “Unsatisfied Italy May Cut JSF Participation,” Defense News, May 10, 2004. Lale Sariibrahimoglu, “Turkey may withdraw from JSF program,” Jane’s Defence Weekly, November 10, 2004. 11 Jeremiah Gertler 2012. F-35 Joint Strike Fighter (JSF) Program, Congressional Research Service, Feb. 16, p. 17. 12 From notes taken by William Hawkins at the SMi Group conference Offsets 2009, Ankara, Turkey, January 27, 2009. 13 “Britain’s new industrial participation policy protects SMEs and sets $1.6m threshold” Counter-Trade & Offsets, Feb. 13, 2012. 14 Vivek Raghuvanshi 2010. “Indian State Seeks Special Zone for Defense Firms,” Defense News, April 22, India's central state of Gujarat is seeking approval to set up a special economic zone in which foreign firms could operate to fulfill offset requirements. 15 “Azerbaijan goes global in search for industrial participation partners” Counter-Trade & Offsets, Dec. 12, 2011. Proceedings of the 2012 Pennsylvania Economic Association Conference 162 16 17 Offsets in Defense Trade: Sixteenth Study, 2012. p. 5. Ibid, pp. 23-24 18 . The GPA allows, “the protection of essential security interests relating to the procurement of arms, ammunition or war materials, or to procurement indispensable for national security or for national defense purposes.” The more general security exemption which places “the traffic in arms, ammunition and implements of war and to such traffic in other goods and materials as is carried on directly or indirectly for the purpose of supplying a military establishment” outside the regular rules of international commerce is Article XXI of the 1994 General Agreement on Tariffs and Trade. 19 Ibid, Appendix F. 20 2005 Trade Policy Agenda and 2004 Annual Report, Office of the U.S. Trade Representative, Washington, DC, March 2005, p. 45. 21 Report of the Interagency Team on Consultation with Foreign Nations on Limiting the Adverse Effects of Offsets in Defense Procurement, December 17, 2007, p. 8. 22 Offsets in Defense Trade: Sixteenth Study, p. 6. See also General Accountability Office (GAO) report on offset activities, “Defense Trade: U.S. Contractors Employ Diverse Activities to Meet Offset Obligations,” December 1998 (GAO/NSIAD-99-35), pp. 4-5. 23 U.S. Department of Commerce, Bureau of Industry and Security (BIS), Offsets In Defense Trade, Eleventh Study, Washington, D.C., February, 2007, Appendix H: Interagency Team Final Report on Consultation with Foreign Nations on Limiting the Adverse Effects of Offsets in Defense Procurement, p. 5-2. 24 Pierre Tran, “Smaller French Firms Feel Offset Heat,” Defense News, Feb. 19, 2012 25 Offsets In Defense Trade, Eleventh Study, 2007. p. 6. 26 Owen E. Herrnstadt, Offsets and the lack of a comprehensive U.S. policy: What do other countries know that we don't? Economic Policy Institute, Briefing Paper #201, April 17, 2008. REFERENCES Are embedded in the endnotes due to the unusual nature of many of the secondary sources. Proceedings of the 2012 Pennsylvania Economic Association Conference 163 ACADEMIC PERFORMANCE IN GRADUATE MANAGERIAL ECONOMICS Rod D. Raehsler Department of Economics Clarion University Clarion, PA 16214 ABSTRACT Using an ordered probit model, final grades for graduate managerial economics are analyzed. It is found that the most important factors influencing final grades for students are scores on the GMAT as well as grades earned in undergraduate sections of business statistics and principles of microeconomics. For the GMAT, it is found that scores achieved on the quantitative portion are more relevant than scores measuring verbal ability. INTRODUCTION How often do we, as economics faculty, hear students describe their trepidation (and sometimes outright fear) of enrolling in an economics course? If I have learned anything during my many years of teaching experience it is that most students simply do not like economics courses. This, of course, is especially true among students who find critical thinking and quantitative analysis to be particularly challenging. For that reason, I have enjoyed teaching the managerial economics course in the MBA program at Clarion University. Through self-selection, students in that course are well-prepared and come into the course with a much more positive attitude on learning important economic concepts that will help them make sound business decisions in the future. That was, at least, the case until just the past two years. Over the past two years I have casually noticed a decline in attentiveness and attitude among graduate students taking the graduate managerial economics course (ECON 510) in either a traditional or online environment. My colleagues have noticed a similar trend. This paper represents an initial attempt to develop an explanation for this attitudinal shift. While the determination of student perception of the course will not be accomplished in this analysis, I will take a look at factors influencing academic performance in the course. All educators are interested in determining or verifying factors that influence student academic performance. The identification of new factors that positively influence this performance can often signal a need for changes in an academic curriculum whereas the empirical establishment of linkages between an important sequence of courses and disciplines that are historically expected can support claims that are often made to students and the maintenance of existing curricula. A great deal more attention has been paid to the empirical analysis of possible determinants of academic performance since the work of Spector and Mazzeo (1980). In their seminal work, they employed a logit model to examine the performance of students in intermediate macroeconomics. The subsequent literature is extensive, including work on principles of economics courses by Becker (1983), Borg et al (1989), Park (1990), Watts and Bosshardt (1991), Becker and Watts (1996, 1999, and 2001). Papers on intermediate economics or econometrics are relatively scarce but include Specter and Mazzeo (1980), Ramonda et al (1990), Becker and Greene (2001), and Yang and Raehsler (2005). While the literature in accounting is mostly limited to gender-related study (Mutchler et al, 1987; Lipe, 1989; Tyson, 1989; Ravenscroft and Buckless, 1992), other studies do exist that focus on income tax courses, CPA exams or other related accounting topics (Murphy and Stanga, 1994; Grave et al, 1993; Brahmasrene and Whitten, 2001). In the literature of finance, Berry and Farragher (1987), among others, were the first to survey introductory financial management courses. Since then there have been papers on introductory finance courses (Ely and Hittle, 1990; Paulsen and Gentry, 1995; Chan et al, 1996; Cooley and Heck, 1996; Sen et al, 1997; Chan et al, 1997; Nofsinger and Petry, 1999; VanNess et al, 1999) but no use of more advanced specifications modeling academic performance in any of the existing educational literature. For example, VanNess et al, (1999) employed ordinary least squares and ordered probit models to discover that part-time instructors typically assigned higher grades, cumulative grade point average was a significant explanatory variable for the final grade, and students majoring in either economics or finance had higher average academic performance in the class. In a similar fashion, Sen el al.(1997) identified grade point average, gender, and performance in prerequisite courses as important indicators of final grades in introductory finance. In their work, females Proceedings of the 2012 Pennsylvania Economic Association Conference 164 tended to outperform males in the sample while grade point average and prerequisite course grades had a positive influence on performance in the finance course. Interestingly, Didia and Hasnat (1998) discovered that female students did not display a grade disadvantage and transfer students may not fare worse than traditional students. A few studies on higher level or graduate finance courses were completed by Rubash (1994), Mark (1998), and Trine and Schellenger, (1999). Trine and Schellenger, (1995) find identify six factors as significant at either ten percent or two percent levels with a coefficient of determination measure of 0.15 in a stepwise multiple regression analysis. Additional work to identify factors in greater detail and with greater precision has not been accomplished in the financial education literature. This paper concentrates on the academic performance in the graduate managerial economics course (ECON 510), a required course for all students in the MBA program at Clarion University. The organization of this paper is as follows: Section II introduces an ordered probit model; Section III presents estimated results of the ordered probit model; and Section IV examines marginal probabilities from changes in GMAT scores and the statistically important dummy variables. The conclusion is presented in Section V. DATA DESCRIPTION AND THE ORDERED PROBIT MODEL In this study, students enrolled in ECON 510from the fall semester of 2006 through the spring semester of 2012. As discussed above, ECON 510 is a required core course for all MBA students in the college and prerequisites consist of two principles of economics courses (macroeconomics and microeconomics) along with the two business statistics courses (ECON 221 and ECON 222) as well as two beginning accounting courses. Academic performance in the principles of microeconomics course is considered here as that course is particularly relevant to material taught in ECON 510 (some material is actually identical). The ECON 222 is the only business statistics course included in this analysis since that course is the first that presents regression analysis critical for success in the ECON 510 course. The lengthy time period chosen for this analysis encompasses a total of 185 graduate MBA students in the potential sample. Students who have not yet taken ECON 510 or who received GMAT waivers are not included in this analysis thereby reducing the useable sample to 104 students. When quantitative and verbal scores are used from the GMAT, the sample size shrinks to 100. This is still a sufficient sample size to produce a relevant statistical result. Variation in grading was minimal throughout this sample as the same two instructors taught the course with identical grading methods and material used during each term. This and the large sample size helps in providing robust empirical results using the ordered probit model in this study. It is well known in the econometric literature that when dealing with qualitative measures such as grades or success and failure, the standard ordinary least squares (OLS) regression technique can produce spurious probability estimates (probabilities that are negative or exceed unity) and negative variances (Greene, 2003). To overcome these shortcomings, a binary probit or logit model, which produces consistent probability estimates, provides a better explanation for two outcomes. When more than two outcomes exist, a model of multiple choices such as a multinomial logit or a similar probit model is often used. However, a multinomial logit or probit model suffers from the “independence from irrelevant alternatives” problem. In other words, odds ratios between outcomes i and j are to be independent of all other alternatives; an extremely restrictive condition placed on the model for most types of data analysis. As a result, we opt for the ordered probit model which takes into account the ordinal nature in the dependent variable (Zavoina and McElvey, 1975). The use of the ordered probit model is justified in the multiple-category case (final letter grades in this study), which is considered ordinal in nature. This phenomenon is particularly true in a state-supported university where “curving” the final letter grade and examinations is more common. As a consequence, the distance between various letter grades (the dependent variable Y) is not a fixed interval. In other words, the difference between an A and a B is not the same as that between a B and a C (and so forth). Very often, a greater degree of grading leniency is given to students performing at the lower end of the grading scale and educators find it more difficult to fail students for a wide variety of reasons (budget conditions at academic institutions, student complaints, and even the human nature of instructors). This greater leniency given to low performers compared to high performers in class makes the distribution of different letter grades generally ordinal in nature. Stated more succinctly, a final grade of A is better than a B, which is better than a C, and so on, however, the measured difference between an A and B and a B and C are not necessarily the same. Proceedings of the 2012 Pennsylvania Economic Association Conference 165 This being the case consider a latent regression equation in matrix form: rather than the composite GMAT scores (variable names are QU and VE respectively). y* = x′β + ε ..............................................................(1) Y = βo+ β1ENRi + β2YEXPi + β3UGPAi + β4QUi + β5VEi + β6M1i + β7M2i + β8S1i + β9S2i + εi ………(6) Or in linear form for the first model: EMPIRICAL RESULTS Y = βo + β1 ENRi + β2EXPi + β3UGPAi + β4GMATi + β5M1i + β6 M2i + β7S1i + β8S2i + εi .....(2) Where y* = unobserved latent variable of letter grades y = 0 = C or below if y* ≤ 0……………………......(3) y = 1 = B if 0 < y* ≤ µ …………………………….(4) y = 2 = A if y* ≥ µ …………………………...……(5) and where µ is the threshold value by which expected letter grades in ECON 510 are determined. Individual variable definitions are as follows: UGPA = undergraduate grade point average on a 4.0 scale. YEXP = years of previous work experience. ENR = ‘1’ for taking the course online and ‘0’ for taking the course in a traditional delivery mode (live). GMAT = Score on the GMAT examination. M1 = ‘1’ denotes a grade of C in principles of microeconomics (minimum to take ECON 510). A ‘0’ implies an A or B. M2 = ‘1’ denotes a student received a B in principles of microeconomics; M2 = 0 implies he or she received a letter grade of other a C or A. It is to be pointed out that the letter grade A is used as the reference group. S1 = ‘1’ denotes a student received a C in the second business statistics course (ECON 222), ‘0’ for other letter grades (A or B since a minimum of C is required). S2 = ‘1’ denotes a student received a B in ECON 222, ‘0’ for other letter grades. A grade of A is the reference group. ε i = Normally distributed residual. A second model used in this analysis considers the use of quantitative and verbal scores on the GMAT Before looking at empirical estimates of the structural models, it is worthwhile to look at whether university MBA programs in the region require managerial economics to see if results developed here have a broader impact. Table 1 presents a summary of this type of information for programs across eight states in this region. There is considerable variation in the percentage of MBA programs requiring managerial economics across states ranging from 80 percent in Maryland to only 29.73 percent in New York. Pennsylvania, interestingly, has the second lowest percentage of MBA programs requiring managerial economics at 40 percent. It is worth noting that the cumulative numbers show that less than half of the MBA programs across eight states in this region require students to take managerial economics. These percentages fall (to an overall value of 42.22 percent) when observing AACSB accredited programs in the same region. This counters a popular claim that managerial economics is necessary for graduate programs to retain or secure accreditation. Most schools either have managerial economics as an elective course or have a required economics course that concentrates on macroeconomic issues and forecasting macroeconomic variables rather than the microeconomic bent found in the managerial course. Table 2 below provides mean values for important variables used in the analysis. The GMAT sample refers to that part of the analysis focused on using composite GMAT scores as an explanatory variable while the VE/QU sample refers to the sample using the verbal and quantitative scores from the GMAT in the model. Information from Table 2 shows that the average grade in graduate managerial economics over the sample period is slightly above a B (an A is scored as ‘2’ and a B scored as ‘1’ with C or lower assigned a score of ‘0’). The undergraduate grade point average is above 3.33 (as would be expected for a graduate program) while the average number of years of experience in the workplace a little below 1.5 years. The enrollment variable (ENR) shows that only around 23 percent of students in the sample took the course online (assigned the ‘1’ for the dummy Proceedings of the 2012 Pennsylvania Economic Association Conference 166 variable). The principles of microeconomics foundation grade dummies (M1 and M2) show that approximately 23 percent of students received a C, 34 percent received a B, and 43 percent earned an A in that course. Likewise, the business statistics dummy variables (S1 and S2) show that about 19 percent of students in the sample earned a C, 40 percent earned a B, and the remaining 41 percent earned an A in the second undergraduate statistics course. The average GMAT score turned out to be 484.808 with average verbal and quantitative scores at 26.75 and 29.95 respectively. These are at the lower end of the historic GMAT score distributions (national averages for the GMAT, verbal examination, and quantitative section are 540, 29, and 37 respectively), something to be expected of a smaller MBA program at a public institution. This will be an important point when viewing the empirical results of the ordered probit model estimation. It may be that important variables predicting academic performance in graduate managerial economics might differ across various types of MBA programs. Table 1 already eluded to the wide variation across MBA programs with respect to requiring this particular course. Estimates of the ordered probit models needed to calculate expected final grades and the marginal impact of important explanatory variables on expected grades in ECON 510 are derived using TSP version 4.5 (2002). The initial estimates of the GMAT sample using equation (2) are shown in Table 3. After eliminating the statistically insignificant explanatory variables using a stepwise approach, the optimal ordered-probit model to be used in the comparative statics analysis can be determined. This model is presented in Table 4. Note that the undergraduate grade point average, type of enrollment (online or traditional), years of work experience, and undergraduate grades in principles of microeconomics do not significantly help explain the distribution of final grades in ECON 510. Given that this course utilizes both microeconomics and applied business statistics, it is somewhat disconcerting that final grades in principles of microeconomics are not significant. Fortunately, this result changes when GMAT scores are replaced with verbal and quantitative test results. Variables that matter the most in helping to forecast grades in ECON 510 using the results from Table 4 are the composite GMAT scores and final grades in the undergraduate business statistics course that focused on regression analysis. It is also important to point out that the threshold variable (µ) is statistically significant (p- value is 0.000). This result verifies that the ordered probit model is appropriate to use with this data set. Rather than relying solely on the composite GMAT score as a predictive variable, I also ran the ordered probit model by incorporating more specific measures of verbal and quantitative (or mathematical) performance on the GMAT. Table 5 shows the estimated values for the full model described by Equation (6). Performing a stepwise procedure on the original ordered probit mode, I found that two similar models were very close in terms of predictability and variable significance. One model included dummy variables identifying grades in undergraduate business statistics while the second model incorporated grades in the principles of microeconomics course. When dummy variables for both were included, the predictive ability of the model declined significantly. This implies that there is a significant correlation between the undergraduate grades in business statistics and the beginning microeconomics course. Table 6 below shows that the quantitative score along with grades in business statistics are important in predicting final grades in ECON 510. Table 7 shows that the quantitative score along with final grades in principles of microeconomics are important explanatory variables. Even though the results for the two models are similar, both are retained in the analysis because the first supports the previously established importance of undergraduate business statistics while the second identifies the principles of microeconomics as an important course. In both cases, the threshold variable is statistically significant thereby identifying the ordered probit model as appropriate for this analysis. APPLICATIONS OF THE ORDERED PROBIT MODEL One of the primary advantages in employing an ordered probit analysis to final grades is that it allows for the development of expected probabilities and a more complete discussion of the marginal effects of important explanatory variables on final grades for a course. The use of a cumulative normal distribution in developing these probabilities represents a distinct advantage over a linear estimation model. The value of probabilities for an average student to receive a particular letter grade in ECON 510 can be found using the relevant average values of explanatory variables provided in Table 2 for each of the equations below: Proceedings of the 2012 Pennsylvania Economic Association Conference 167 Prob [y = 0 or C] = φ (–β′x) …………….….……..(7) Prob [y = 1 or B] = φ [µ – β′x] – φ (–β′x) ……...…(8) Prob [y = 2 or A] = 1 – φ(µ – β′x) ………………..(9) ∂ Prob [Y = 1 or B] / ∂ GMAT = [ø(–β′x) –ø(µ –β′x)]*(π½Μ 2) ………………..(11) = (0.0885 – 0.6179) * 0.004156 = -0.002200 For example, with the GMAT sample the value for β’x is found by plugging the average values into the optimal version of Equation (2) which can be found in Table 3. When this is done, the value of β’x turns out to be 1.35 with the threshold value of µ equal to φ is the cumulative normal 1.65. In each formula density function. For instance, the probability for a typical student to receive a C (or less) or B or A using the GMAT-focused data can be calculated as p (y = 0, C or below) = φ (-1.35) = 0.0885 p (y = 1 or B) = φ (1.65 – 1.35) – φ (-1.35) = 0.6179 – 0.0885 = 0.5294 p (y = 2 or A) = 1 – φ (1.65 – 1.35) = 1 – 0.6179 = 0.3821 This means that a typical student has an 8.85 percent chance of earning a C, a 52.94 percent chance of a B, and a 38.21 percent chance of earning an A in ECON 510. An identical procedure can be employed on the second data set that focuses on verbal and quantitative scores using the optimal results for equation (6) provided in Table 6 and Table 7. Results of all calculations are presented in Table 8. Since scores from the GMAT play an important role in helping determine final grades in ECON 510, the marginal probabilities as described in Greene (2003) are calculated using the first data set and results: ∂ Prob [Y = 0 or C] / ∂ GMAT = – ø(–β′x)*(π½Μ 2) ..(10) = –ø(-1.35)*(0.004156) = -0.0885*0.004156 = -0.000368 ∂ Prob [Y = 2 or A] / ∂ GMAT = ø (µ – β′x)]*(β2) …………………………..(12) = 0.6179 * 0.004156 = 0.002568 Where β2 is the slope on GMAT for the optimal model. Equations (7) through (12) can also be run on the second model using the quantitative scores (QU) in place of the GMAT composite scores. Results of the calculations above show that a one point increase in the GMAT composite score results in a 0.0368 percent decrease in the probability of a student receiving a C in ECON 510. Likewise, that same increase leads to a probability decrease of 0.2200 percent and a probability increase of 0.2568 percent of receiving a B and A respectively in ECON 510. Therefore, a student scoring fifty points higher than average (for the sample) on the GMAT (a score around 534) will have a 12.84 percent higher chance of earning an A as a final grade in ECON 510. A summary of these marginal probabilities is provided in Table 9. As described earlier, results from Table 8 show that expected grades in ECON 510 are relatively high. Regardless of which sample is used, most students earn B’s while relatively few students receive C’s or less (less than nine percent for each). Expectations for A’s are very high for the typical student ranging from 38.21 percent to 38.59 percent. Table 9 indicates that higher GMAT scores and, in particular, higher scores on the quantitative section lead to significant increases in grade expectations. For example, only a one point increase in the quantitative portion score will increase the probability of earning an A in ECON 510 by between 2.7 and 2.8 percent. Using the dummy variables S1, S2, M1, and M2 along with equations (7), (8), and (9) it is possible to calculate the marginal impact of earning higher (or lower) letter grades in undergraduate business statistics and principles of microeconomics on expected grades in ECON 510. Table 10 below, for example, shows that using the GMAT model a student earning an A in business statistics increases their probability of getting an A in ECON 510 by a Proceedings of the 2012 Pennsylvania Economic Association Conference 168 significant margin (28.14 percent). Obtaining a C in business statistics, however, will decrease the probability of that student earning an A in ECON 510 by 25.23 percent. C in principles of microeconomics should focus on earning a B or C rather than expect an A in ECON 510. CONCLUSION The same analysis can be done using the data set that incorporated quantitative scores in place of the composite GMAT score. Results of this analysis using this data along with the same dummy variables for grades in business statistics are presented in Table 11. It is easy to see that the results are very similar. For variety in the discussion, suppose a student earns a B in business statistics. One can see that the probabilities of earning a B or C in ECON 510 for this student increase (10.46 percent and 8.52 percent respectively) while the probability of earning an A falls by 18.98 percent. Clearly, students earning an A in business statistics have a distinct advantage when it comes to competing for the top grade in ECON 510. The final part of this analysis uses dummy variable for grades earned in the principles of microeconomics course rather than business statistics. Using the results shown in Table 12 it is easy to see that earning an A in principles of microeconomics is just as important as earning an A in business statistics. The fact that the optimal model does not include all dummy variables simultaneously indicates that the majority of students earning an A in one course also earned an A in the other undergraduate foundation course. A student earning an A in principles of microeconomics has a 24.36 percent higher chance of earning an A in ECON 510 while a student earning a Even though this represents a very preliminary inquiry into factors determining final grades in ECON 510, a few important results can be emphasized. First, GMAT scores and, in particular, scores on the quantitative portion of the GMAT appear to matter greatly in determining student expectations on final grades in graduate managerial economics. This might be an important consideration if programs begin to look at other entrance criteria or providing waivers of the GMAT for entrance into the MBA program. Results also show that grades in two undergraduate foundation courses, business statistics and principles, significantly affect grade expectations in ECON 510. In fact, earning an A in both courses dramatically increases the probability of a student earning an A in the graduate managerial economics course. It is interesting that the undergraduate grade point average does not factor in, however, this might be due to the fact that the undergraduate major of the student was not included as a variable (a more labor intensive data collection activity). This needs to be done as high grades in one field might not be equivalent in effect as higher grades in a more challenging field (such as economics and finance). Dummy variables for academic major and gender do need to be included in future work. Proceedings of the 2012 Pennsylvania Economic Association Conference 169 Table 1: Managerial Economics Requirements Across MBA Programs __________________________________________________________________________________________ All Schools AACSB Accredited Schools Number Percent Number Percent Number Requiring Requiring Number Requiring Requiring of Managerial Managerial of Managerial Managerial State Schools Economics Economics Schools Economics Economics __________________________________________________________________________________________ MD 10 8 80.00 5 4 80.00 WV 6 4 66.67 2 2 100.00 DE 5 3 60.00 2 1 50.00 VA 20 11 55.00 11 6 54.55 NJ 17 9 52.94 9 5 55.56 OH 27 14 51.86 15 6 40.00 PA 40 16 40.00 21 8 38.10 NY 37 11 29.73 25 6 24.00 All 162 76 46.91 90 38 42.22 __________________________________________________________________________________________ Table 2: Mean Values of Variables _________________________________________________________________ GMAT VE/QU Variable Sample Sample Y (ECON 510 Grade) 1.279 1.290 UGPA 3.332 3.333 YEXP 1.471 1.520 ENR 0.221 0.230 GMAT 484.808 -------VE -------26.750 QU -------29.950 M1 0.231 0.230 M2 0.337 0.340 S1 0.192 0.200 S2 0.404 0.390 n 104 100 _________________________________________________________________ Proceedings of the 2012 Pennsylvania Economic Association Conference 170 Table 3: Estimates of Ordered Probit Model (Model 1) ________________________________________________________________ Variable Coefficient Standard Error t ratio p value Constant -0.772684 1.543 -0.501 0.617 ENR 0.047591 0.297 0.160 0.873 UGPA 0.197797 0.380 0.521 0.603 YEXP 0.023949 0.042 0.564 0.573 GMAT 0.004007 0.002 2.418** 0.016 M1 -0.407402 0.321 -1.269 0.204 M2 -0.429951 0.290 -1.481 0.139 S1 -0.507629 0.357 -1.420 0.156 S2 -0.399440 0.282 -1.418 0.156 µ 1.68721 0.202 8.337* 0.000 _________________________________________________________________ Dependent variable = letter grade for graduate managerial economics Number of observations = 104 Likelihood ratio = 21.7253 (p value = 0.005) Log likelihood function = -89.5556 Scaled R-squared = 0.1984 * = significant at 1% ** = significant at 5% *** = significant at 10% Table 4: Optimal Model Estimation of Ordered Probit Model (Model 1) _______________________________________________________________ Variable Coefficient Standard Error t ratio p value Constant -0.318299 0.822 -0.387 0.699 GMAT 0.004156 0.002 2.584** 0.010 S1 -0.743270 0.323 -2.299** 0.022 S2 -0.514727 0.268 -1.918*** 0.055 µ 1.65309 0.199 8.310* 0.000 _________________________________________________________________ Dependent variable = letter grade for graduate managerial economics Number of observations = 104 Likelihood ratio = 17.4640 (p value = 0.001) Log likelihood function = -91.6862 Scaled R-squared = 0.1611 * = significant at 1% ** = significant at 5% *** = significant at 10% Proceedings of the 2012 Pennsylvania Economic Association Conference 171 Table 5: Estimates of Ordered Probit Model (Model 2) _________________________________________________________________ Variable Coefficient Standard Error t ratio p value Constant -0.974964 1.532 -0.636 0.617 ENR -0.044372 0.304 -0.146 0.873 UGPA 0.525767 0.405 1.299 0.603 YEXP 0.030606 0.042 0.723 0.573 VE -0.010437 0.021 -0.487 0.016 QU 0.045893 0.016 2.908* 0.016 M1 -0.374835 0.333 -1.127 0.204 M2 -0.472975 0.299 -1.582 0.139 S1 -0.467198 0.362 -1.291 0.156 S2 -0.331975 0.290 -1.146 0.156 µ 1.73684 0.213 8.147* 0.000 _________________________________________________________________ Dependent variable = letter grade for graduate managerial economics Number of observations = 100 Likelihood ratio = 24.1941 (p value = 0.004) Log likelihood function = -83.7887 Scaled R-squared = 0.2279 * = significant at 1% ** = significant at 5% *** = significant at 10% Table 6: Optimal Model Estimation 1 of Ordered Probit Model (Model 2) _________________________________________________________________ Variable Coefficient Standard Error t ratio p value Constant 0.420441 0.507 0.830 0.407 QU 0.044018 0.015 2.875* 0.004 S1 -0.809875 0.323 -2.509** 0.012 S2 -0.503465 0.268 -1.879*** 0.060 µ 1.66952 0.205 8.157* 0.000 _________________________________________________________________ Dependent variable = letter grade for graduate managerial economics Number of observations = 100 Likelihood ratio = 17.4782 (p value = 0.001) Log likelihood function = -87.1467 Scaled R-squared = 0.1675 * = significant at 1% ** = significant at 5% *** = significant at 10% Proceedings of the 2012 Pennsylvania Economic Association Conference 172 Table 7: Optimal Model Estimation 2 of Ordered Probit Model (Model 2) _________________________________________________________________ Variable Coefficient Standard Error t ratio p value Constant 0.404717 0.489 0.827 0.408 QU 0.045934 0.015 3.015* 0.003 M1 -0.644362 0.304 -2.119** 0.034 M2 -0.719322 0.273 -2.630* 0.009 µ 1.67537 0.205 8.182* 0.000 _________________________________________________________________ Dependent variable = letter grade for graduate managerial economics Number of observations = 100 Likelihood ratio = 18.5771 (p value = 0.000) Log likelihood function = -86.5973 Scaled R-squared = 0.1775 * = significant at 1% ** = significant at 5% *** = significant at 10% Table 8: Probabilities of Grade Values in Graduate Managerial Economics _________________________________________________________________ Grade Model 1 Model 2 Model 3 C 0.0885 0.0838 0.0823 B 0.5294 0.5303 0.5318 A 0.3821 0.3859 0.3859 _________________________________________________________________ Model 1: GMAT model with business statistics dummies Model 2: QU model with business statistics dummies Model 3: QU model with principles of microeconomics dummies Table 9: Marginal Effect of GMAT Score or Quantitative Score on the Probabilities of Final Grades in Graduate Managerial Economics _________________________________________________________________ QU QU Grade GMAT Model 1 Model 2 A +0.002568 +0.027031 +0.028208 B -0.002200 -0.023343 -0.024428 C -0.000368 -0.003689 -0.003780 _________________________________________________________________ Model 1: QU model with business statistics dummies Model 2: QU model with principles of microeconomics dummies Proceedings of the 2012 Pennsylvania Economic Association Conference 173 Table 10: Marginal Effect of Grades in Business Statistics on the Probabilities of Final Grades in Graduate Managerial Economics (GMAT Model) _________________________________________________________________ Grade in ECON 222: ECON 510 Grade A B C A +0.2814 -0.1893 -0.2523 B -0.1712 +0.1007 +0.0938 C -0.1102 +0.0886 +0.1585 _________________________________________________________________ Table 11: Marginal Effect of Grades in Business Statistics on the Probabilities of Final Grades in Graduate Managerial Economics (QU Model) _________________________________________________________________ Grade in ECON 222: ECON 510 Grade A B C A +0.3077 -0.1898 -0.2747 B -0.1922 +0.1046 +0.1038 C -0.1155 +0.0852 +0.1709 _________________________________________________________________ Table 12: Marginal Effect of Grades in Principles of Microeconomics on the Probabilities of Final Grades in Graduate Managerial Economics (QU Model) _________________________________________________________________ Grade in Microeconomics: ECON 510 Grade A B C A +0.2436 -0.0595 -0.2295 B -0.1502 -0.0703 +0.1046 C -0.0934 +0.1298 +0.1249 _________________________________________________________________ Proceedings of the 2012 Pennsylvania Economic Association Conference 174 REFERENCES Becker, E. William, Jr. 1983. Economic education research: Part III, statistical estimation methods. Journal of Economic Education. 14 (3): 4-15. Berry, T. D. and E. J. Farragher. 1987. A survey of introductory financial management courses. Journal of Financial Education. 13(2): 65-72. Borg, O. Mary, Paul M. Mason, and Stephen L. Shapiro. 1989. The case of effort variables in student performance. Journal of Economic Education. 20(3): 308-313. Chan, Kam C., Connie Shum, and Pikki Lai. 1996. An empirical study of cooperative instructional environment on student achievement in principles of finance. Journal of Financial Education. 25(2): 21-28. Chan, Kam C., Connie Shum, and David J. Wright. 1997. Class attendance and student performance in principles of finance. Financial Practice and Education. 7(2): 58-65. Cooley, Philip, and Jean Heck. 1996. Establishing benchmark for teaching the undergraduate introductory course in financial management. Journal of Financial Education. 22 (Fall): 1-10. Didia, Dal and Baban Hasnat. 1998. The determinants of performance in the university introductory finance course. Financial Practice and Education. 8(1): 102-107. Ely, David P. and Linda Hittle. 1990. The impact of math background on performance in managerial economics and basic finance courses. Journal of Financial Education. 19(2): 59-61. Estrella, A. 1998. A new measure of fit for equations with dichotomous dependent variables. Journal of Business and Economics Statistics. (April): 198-205. Graves, O. Finley, Irva Tom Nelson, and Dan S. Dcines. 1993. Accounting student characteristics: Results of the 1992 federation of schools in accountancy (FSA) survey. Journal of Accounting Education. 11(2l): 221-225. Greene, W. H. 2003. Econometric Analysis, 5th Ed., Upper Saddle River, NJ: Prentice Hall. Leiber, J. Michael, B. Keith Crew, Mary Ellen Wacker, and Mahesh K. Nalla. 1993. A comparison of transfer and nontransfer students majoring in criminology and criminal justice. Journal of Criminal Justice Education. 4(1): 133151. Liesz, Thomas J. and Mario G. C. Reyes. 1989. The use of “Piagetian concepts to enhance student pPerformance in the introductory finance course. Journal of Financial Education. 18(2): 8-14. Lipe, Marlys G. 1989. Further evidence on the performance of female versus male accounting students. Issues in Accounting Education. 4(1): 144-152. Marks, Barry. 1998. An examination of the effectiveness of a computerized learning aid in the introductory graduate finance course. Financial Practice and Education. 8(1): 127-132. Murphy, P. Daniel and Keith G. Stanga. 1994. The effects of frequent testing in an income tax course: An experiment. Journal of Accounting Education. 12(1): 27-41. Mutchler, Jane F., Joanne H. Turner and David D. Williams. 1987. The performance of female versus male accounting students. Issues in Accounting Education. 2(1): 103-111. Proceedings of the 2012 Pennsylvania Economic Association Conference 175 Nofsinger, John and Glenn Petry. 1999. Student study behavior and performance in principles of finance. Journal of Financial Education. 25(Spring): 33-41. Park, Kang H. and Peter M. Kerr. 1990. Determinants of academic performance: A multinomial logit approach. Journal of Economic Education. 21(2): 101-111. Paulsen, Michael B. and James A. Gentry. 1995. Motivation, learning strategies, and academic performance: A study of the college finance classroom. Financial Practice and Education. 5(1): 78-89. Raehsler, Rod D., Tony R. Johns, Chin W. Yang, and Ken Hung. 2011. Academic preparation, gender, and student performance in operations management: An ordered probit analysis. Clarion University Working Paper. Raimondo, Henry J., Louis Esposito, and Irving Gershenberg. 1990. Introductory class size and student performance in intermediate theory courses. The Journal of Economic Education. 21(4): 369-381. Ravenscroft, Susan P. and Frank A. Buckless. 1992. The effect of grading policies and student gender on academic performance. Journal of Accounting Education. 10(1): 163-179. Rubash, Arlyn R. 1994. International Finance in an International Environment. Financial Practice and Education. 4(1): 95-99. Sen, Swapan, Patrick Joyce, Kathy Farrell, and Jody Toutant. 1997. Performance in principles of finance courses by students with different specializations. Finanical Practice and Education. 7(2): 66-73. Spector, C. Lee and Michael Mazzeo. 1980. Probit aAnalysis and economic education. Journal of Economic Education. 11(2): 37-44. Trine, A. D. and M. H. Schellenger. 1999. Determinants of student performance in an upper level corporate finance course. Academy of Educational Leadership Journal. 3(2): 42-47. Tyson, Thomas. 1989. Grade performance in introductory accounting courses: Why female students outperform males. Issues in Accounting Education. 4(1): 153-160. VanNess. β. F., VanNess R.S. and R. Kamery. 1999. The effect of part-time instruction on grades in principles of finance. Financial Practice and Education. (Fall/Winter): 105-110 Watts, Michael and William Bossbardt. 1991. How instructors make a difference: Panel data estimates from principles of economic courses. Review of Economics and Statistics. 73(2): 336-340. Yang, C. W. and R. D. Raehsler. 2005. An economic analysis on intermediate microeconomics: An ordered probit model. Journal of Economics Educators. 5(3):1-11. Zavoina R., and W. McElvey. 1975. A statistical model for the analysis of ordinal level dependent variables. Journal of Mathematical Sociology. 4(1): 103-120. Proceedings of the 2012 Pennsylvania Economic Association Conference 176 DECOMPOSING RECENT MONEY SUPPLY CHANGES WITH IMPLICATIONS FOR CURRENT FED POLICY Richard Robinson and Marwan El Nasser School of Business, E336 Thompson Hall SUNY at Fredonia, Fredonia, NY, 14063 ABSTRACT In response to the financial crisis of 2008, the Federal Reserve radically increased the monetary base. Banks responded by increasing excess reserves rather than increasing bank loans, and the public responded with a substantial flight to liquidity in the form of currency and demand deposits. As a result, the money-supply multipliers substantially decreased so that the actual money supply measures grew more moderately than the base. This paper presents decompositions of the money-multiplier changes, as they have occurred since the onset of the financial crisis, into changes in the currency-to-deposit ratios, and change in the reserve-to-deposit ratio. By doing so, possible nearterm increases in the multipliers are simulated so that the possibility of full, or partial restoration to their pre crisis levels are assessed. In conjunction with the equation of exchange, Fed policy possibilities for controlling inflation over various horizons are simulated. This analysis further illustrates the Fed’s exit dilemma from its financial-crisis policy. INTRODUCTION One conventional view of Federal Reserve policy during the post 2008 financial crisis is that its open-market operations were merely “pushing on the string of a liquidity trap.” In fact, the Fed began lowering its Federal Funds rate target in September 2007. Over the following year, it lowered this target rate from 5.25 percent to 2 percent. Post the financial panic of mid-September, 2008 (the collapse of Lehman Brothers), it further dropped the target rate to approximately .25 percent. As indicated by Blinder (2010), it became apparent at this time that lowering “liquidity premiums,” and “risk premiums” became the essentially needed task of monetary policy.1 Since the riskless overnight rate became essentially zero, the Fed needed to more directly flatten both the risk-free yield curve and the risk-structure of yields. The policies the Fed utilized to accomplish this narrowing of spreads has been termed quantitative easing, i.e. QEI was initiated in late 2008, QEII in late 2009, and operation twist in 2011. Essentially this latter policy involved purchasing very large amounts of targeted securities: i.e. mortgage-backed securities, agency debt securities (which allowed agencies to further support the mortgage market), commercial paper, money-market mutual fund securities, and medium and longer-term Treasury bonds. Essentially this quantitative easing involved the Fed directly purchasing one of the risky or less-liquid assets from financial markets, and paying for these purchases by either liquidating its portfolio of T-bills or increasing the monetary base. The former involved no net increase in the Fed’s assets, but the latter resulted in an increase. Total Fed assets skyrocketed from approximately $.91 trillion on September 3, 2008, to approximately $2.2 trillion on November 12, 2008. Almost all of this increase was absorbed by the banking system as excess reserves which increased from approximately $2 billion in August, 2008, to $.77 trillion in December, 2008. The early stages of the quantitative easing policy was “ad hoc, reactive and institution based” (Blinder, 2010, p. 469). Initially, in late 2008, the Fed developed its policy “on the fly,” often on short-term notice, by acquiring assets as a result of rescue operations, e.g. the Maiden Lane MBS portfolio from AIG. The problem of this early financialcrisis period, however, was that the resulting increase in the monetary base was not loaned by the banking system. The increase remained as excess reserves, as shown in Table 1a. As a result, the money multiplier collapsed, also shown in Table 1a. Because of radical increase in the monetary base, however, on net the money supply did increase at relatively modest rates during this period and up to the present time (March of 2012). These rates are also shown in Table 1b. As bank intermediaries ceased their loan function, the economy sank into the Great Recession. The problems posed for Fed policy now concern what might happen when banks resume their loan activity, especially with this very large existing monetary base. In addition to the problem with excess reserves, the public responded to the financial crisis with a substantial flight to liquidity in the form of currency and demand deposits. As a result, the money-supply multipliers substantially decreased.2 The monetary policy dilemma the Fed must now confront is that if the money multipliers rapidly increase towards their pre-crisis levels then a radical increase in the money supply and consequent significant inflation will surely result. This is a problematic fact that looms over Fed policy, a problem examined here. Proceedings of the 2012 Pennsylvania Economic Association Conference 177 This paper presents decompositions of the growth in M1 and the changes in its multiplier. This decomposition allows insights into the future course and effects of monetary policy. For example, any increase in the M1 multiplier that results from a switch from excess reserves to bank loans will have a different effect on the economy than an increase in the multiplier caused by a reduction in currency held by the public. Consequently we decompose changes in the money multiplier, as they have occurred since the onset of the financial crisis, into various factors such as the changes in the currency-to-deposit ratio, changes in the other-checkable-deposit ratio, and changes in the total-reserve-to-deposit ratio. In light of this decomposition, possible near-term increases in the multiplier are simulated so that the possibility of full or partial restoration of the multiplier to its pre-crisis level is assessed. This examination includes explorations of the options the Fed has for handling the excess reserves, and simulations of the money supply expansion process, which together with simulations of the equation-of-exchange, allow explorations of possible scenarios for inflation. This simulation analysis further illustrates the Fed’s exit dilemma from its financial-crisis policy. OPTIONS FOR THE FED’S EXIT STRATEGY The Fed currently (3/2012) holds $.950 trillion in MBS and Agency Debt Securities, and $1.66 trillion in US Treasury securities.3 The problem is the liquidation of its holdings of securities while also reducing excess bank reserves, which are approximately $1.6 trillion. These reserves are eligible for loans by banks, and if they are loaned quickly, then inflation will surely result. The simulations provided in the next section indicate the problems the Fed must face in controlling inflation as a result of its current portfolio. Bernanke (2010) stated the options the Fed has for a planned “exit strategy.” 1. Redeeming or selling securities in conventional (traditional) open-market operations. 2. Passively redeeming MBS, agency debt, and Treasuries as they mature, and not purchasing replacement securities. 3. Selling reverse repurchase agreements so as to maintain the prices of these securities. 4. Increasing the interest rates it pays on reserve deposits so that banks will not loan these reserves out. 5. “Offer to depository institutions term-deposits which … could not be counted as reserves.” (p.8) With respect to the first option, i.e., the selling of securities in conventional open-market operations, we note that because the Fed’s current portfolio of securities is largely longer-term Treasuries, and mortgage-backed securities, both of which it acquired to support different segments of capital markets, it must be reluctant to sell these securities at least until after these particular market segments have experienced robust and sustained restoration. Over the previous two years, the Fed has not used open market operations to reduce the monetary base, but if markets warrant, this is certainly the most viable policy option. The Fed did sell the Maiden Lane MBS portfolio during 2011 although it sterilized this sale by purchase of long-term Treasuries. With respect to the second Fed option, i.e., the passive allowance of redemption of securities without repurchasing similar securities, we note that as of September, 2011, the Fed has done the opposite by purchasing similar replacements to those that matured and were redeemed. This action maintained the monetary base rather than allowing it to contract. Still, this policy option has potential for decreasing the base if and when credit conditions warrant. With respect to the third option, the selling of reverserepurchase agreements, note that if this option is utilized, it would allow the Fed to maintain the prices of securities involved in that if market conditions for the securities the Fed sells on a repo basis deteriorate, repurchase could restore the price level. Of course, since the monetary base has been stable over 2010 to early 2012, this option remains to be utilized to any substantial amount. With respect to the fourth option, i.e. the payment of interest on reserve deposits, this can only be an effective option when credit conditions are poor, and the interest rates the Fed must pay are low, such as its current .25 percent. (Blinder, 2010, examines this difficulty.) It must be kept in mind that the interest paid on these reserves becomes an addition to the monetary base, so that as an instrument for controlling the base, this could only work when competitive rates are low. If credit conditions heat up and interest rates rise so that banks could earn perhaps 6 percent or higher on loans, then for the Fed to pay higher than this to keep banks from loaning out their excess reserves would mean substantial increases in the very base its controls are aimed at. This only delays and exacerbates the inevitable problem. The fifth option, the conversion of excess reserves into time deposits with appropriate interest paid, has the same problem as the fourth option. The interest is added to the same reserves that are to be controlled, so that the problem of liquidating excess reserves is exacerbated and merely delayed. It is interesting to note that Bernanke did not mention increasing reserve requirements, which certainly appears to be a viable option under the current circumstances of extremely high excess reserves. It appears that increasing Proceedings of the 2012 Pennsylvania Economic Association Conference 178 the required-reserve-to-deposit ratio would absorb at least a portion of the excess, and as a consequence, ameliorate a portion of this problem. The possible reasons why this might not be considered as a current option include: • The Fed might fear that the announcement effects of changes in reserve requirements could mistakenly indicate an over tightening of credit when conditions do not warrant this reaction. • An increase in reserve requirements might have differential impacts on banks where those that have already begun to loan, something the Fed surely wants, are hurt relative to those who have preferred to sit on their current position. • Note that Chairman Bernanke is a scholar of the Great Depression of the 1930s (see Bernanke, 1983, and 2005), when reserve requirements were doubled over a six month period in 1936, and credit markets collapsed so as to cause a second massive decline in real output during the Great Depression (the first being between 1930 and 1933). Surely the Fed is likely to be reluctant to repeat this historical experience.4 After considering these three reasons, only moderate increases in the required reserve ratio are acceptable, an increase which we do consider in the simulations presented below. If the loan markets do heat up, then the money multiplier will rise rapidly towards its pre-crisis level, and the Fed will have to reverse its quantitative easing policy, and reduce the monetary base. To not do so would result in substantial increases in the money supply, and consequent inflation. In order to indicate the extent of the problem, and the tight path the Fed must follow to manage and avoid this inflation, below we provide simulations of changes in the money-multiplier, the base, money velocity and real economic activity, as measured over various time horizons, and then the consequent inflation rates that follow as an endogenous result. MULTIPLIER DECOMPOSITION c= currency in the hands of the public o= other checkable deposits demand deposits demand deposits In equation (2), L is a simplified money multiplier in that the variable “reserves” includes both required and excess reserves, and M1 includes currency in the hands of the public, plus demand deposits, plus other checkable deposits.6 Equation (3) gives the total differential of (2), and (4) gives this differential in the form of instantaneous percentage changes, i.e. dln(L)/dt where t is time. dL = (dc + do)[ dL/dt Lt = dc dt [ 1 L 1+c+o 1+c+o - - dr dt ] – (dr + dc)[ 1 r+c [ ] - 1 r+c ]+ do dt [ 1 L r+c 1+c+o ] ] (3) (4) Equation (4) suggests the time-series model (5) where βL, βr, βc and βo are all changes through time, and ut is a random error term that reflects any imperfections in the simplified multiplier measurements. βL Lt = β1βct + β2βrt + β3βot + ut where β1 = β2 = β3 = 1 r+c 1 1+c+o 1 1+c+o <0 (5) - 1 r+c = β2 + β3 >0 Model equation (5) can be estimated from time-series data. This allows the decomposition of the money multiplier into changes in the currency ratio (βc), changes in the reserve ratio (βr), and changes in the other-checkable-deposits ratio (βo). EMPIRICAL ESTIMATION The money supply process for M1 is described by equation (1) where L is the money multiplier as given by (2), and Base consists of the reserves of the banking system plus currency in the hands of the public.5 M1 = L•Base (1) 1+c+o (2) L= r+c where r= reserves demand deposits Data Analyzed To estimate equation (5), monthly data for M1, the Monetary Base, and Total Reserves were gathered from series H.3, www.federal reserve.gov, 12/2006 to 3/2012, seventy-five monthly observations. All data was Not Seasonally Adjusted, and the reserve data was also not adjusted for required reserves. Similar monthly data for Currency in the Hands of the Public, and Demand Deposits, were gathered from series H.6, Not Seasonally Adjusted for the same months. Table 2a presents the basic statistics for the data analyzed. Table 2b presents the Proceedings of the 2012 Pennsylvania Economic Association Conference 179 means and standard deviations for the relevant variables as measured over the time period 8/2008 to 3/2012. The statistics show that although there was a flight to currency during the period 8/2008 to 3/2012, the coincident increase in demand deposits was greater so that the currency-to-deposit ratio actually declined. The reservesto-deposit ratio, however, increased by approximately 1,345 percent, so that this more than counteracted the effects on the multiplier of a decrease in c, and as a result, the net effect decreased the multiplier. Equation (5) indicates that the effects of βc on the multiplier must be less than βr, and the large positive magnitude of βr must account for the almost 50 percent decline in the multiplier over the 2008 – 2012 period. Table 2a also shows that the Base increased substantially (213 percent) over the 8/2008 to 3/2012 period. This Base increase offset the Multiplier decline to result in an increase in M1 by 60 percent over this period. Equation (6) presents the decomposition of this decline in M1. %βM1= %βMult.+ %βBase + (%βMult.)(%βBase) (6) Table 3 presents the OLS estimate of equation (5), the percentage change in the multiplier, with the constantintercept term suppressed in this regression. Since the OLS estimation of equation (5) indicates that both β1 and β3 are insignificantly different from zero, then equation (7) was also estimated. The results are reported in Table 4. The F-ratio for inclusion of βc is F = .67. As a result, the OLS analysis that the hypothesis that βc and βo contribute explanatory power to the model is rejected at 99% significance level, i.e. the joint hypothesis that β1 = β3 = 0 cannot be rejected. Lt = β2βrt + ut We conclude, from the combination of the mean and OLS analysis that increases in the total-reserve ratio and increases in the other-checkable-deposits ratio have effects of similar, but opposite in direction, magnitudes on the money multiplier. Changes in the currency ratio have a smaller, but occasionally significant magnitude, effect. Implications for Monetary Policy .5983 = -.4898 + 2.1329 - (.4898)(2.1329) βL The effects of changes in the total reserve ratio (βr) clearly had an impact of substantial size. The mean of the calculations of β2 shows the largest absolute magnitude among the coefficients; βr clearly had a substantial inverse effect on the multiplier. The impact of βo on the multiplier is clearly positive; an increase in “other-checkabledeposits” as a ratio of demand deposits increases the multiplier. In addition, the t-statistics shown in Table 5 indicates that the means statistics for β1, β2, and β3 are significantly different from 0 for each. Also, the ranges for β2 and β3 are entirely positive, although the range for β1 includes both negative and positive values. Also, at the means for these coefficients we do have β1 = β2 + β3 as established by (5). (7) Calculated Effects of c, o and r on the Multiplier The OLS empirical analysis indicates that changes in the currency ratio βc, and the other-checkable-deposits ratio βo, have little to no effects on the money multiplier, i.e. β1 ≈ β3 ≈ 0. Table 5, however, presents descriptive statistics of β1, β2, and β3 as based upon monthly calculations using the formulae of equation (5) and monthly calculations of c, r and o. The mean estimates align closely with the OLS estimates, but the range for each coefficient is fairly wide. For example, the minimum value for β1 is -.18 which indicates that βc might have had a significant inverse impact on the multiplier over some sub-period, although in total over the entire period investigated, its effects often have little to no impact. The estimate of equation (5) has interesting implications for the future of monetary policy in that it allows an estimation of the change in the money-multiplier given a return to normal levels with respect to excess reserves. Table 6 presents the percentage changes in the multiplier (L) given returns of r, o and c to their former levels on 8/2008. Table 2a shows their levels on the selected dates of 8/2008 and 3/2012, and the time difference values of βr = +1.94, βc = -1.20, and βo = -.44 as calculated. These reversals are presented in Table 6. Using the mean values for β1, β2, and β3 presented on Table 5, the percentage changes in L are computed and presented. As shown, if all three variables (r, c and o) return to their pre-financial crisis levels, the multiplier will increase by 65.31 percent. The flight to liquidity involving the increase in currency ratio (c) and other-checkable-deposits ratio (o), add only slightly more than 5% to the multiplier, but the reduction in the total reserves-to-deposit ratio to its pre-crisis level increases the multiplier slightly more than 60%. During 2006, and before the onset of the financial crisis, the reserve-to-deposit ratio hovered about its approximate long-term average of .14. Post the crisis, and because of the radical increase in the excess reserves, this reserve-todeposit ratio has hovered above 2.0. A return to “normality” of a ratio of .14, i.e. if the banking system loans the excess reserves as is typically the case during economic growth periods, then the money supply increases by approximately 60 percent assuming no change in the monetary base. It should be obvious that a 60 percent Proceedings of the 2012 Pennsylvania Economic Association Conference 180 increase in the money supply M1 over a period of even 5 years or less is unacceptable for the course of Federal Reserve monetary policy. A 60 percent increase over a period of even a decade would likely be unacceptably large because of consequent inflation. Simulated Inflation Results under Various Fed Options The purpose of quantitative easing is to purchase private sector assets to such a substantial amount so as to reduce term- and risk-premiums. Anderson, et al. (2010) examined these monetary policies across international central banks. They point out that quantitative easing policies have been effective only when the central banks are credible with respect to unwinding their balance-sheet adjustments so as to prevent consequent inflation. Bullard (2010) reinforces the importance of this credibility for Fed policy with respect to the current need for unwinding excess reserves. Note that not all economic analysts find the Fed’s policy as credible. (See Meltzer, 2010, as an example.) Its credibility depends upon two characteristics: • The central bank’s political authority (perhaps independence from other governmental influence) and willingness to handle the inflation problem. • The central bank’s power to resolve the problem. The latter depends upon fundamental economic timesequence characteristics, i.e., is the increase in the monetary base that results from quantitative easing capable of being unwound, and can the money multiplier be controlled? To examine these issues, we utilized timesequence simulations to illustrate the extent of the Fed’s problem, and possible resolutions of these difficulties. The monthly measures of the multiplier for M1 indicate it was stable at 1.66 for the months of October, 2007 through August, 2008, the year prior to the financial crisis. Starting in September, 2008, the multiplier began to collapse to its low level of .82 in January of 2012, a decline of slightly over 50 percent.7 Table 7 shows the required annualized geometric growth rate if the multiplier restores to 1.66 over three different time-horizons: a 3 year period, a 5 year period, and an 8 year period. Given these calculations, and given scenarios for changes in the monetary base presented in Table 8, various changes in M1 can be simulated as presented in Table 9. To complete the simulations, consider the equation of exchange presented by equation (8), and its associated percentage change presented by (9a), and (9b) for small percentage changes. Emerging from a recession is usually associated with an increase in the velocity of money, i.e., %ΔV > 0. Assume for the sake of this simple analysis that %ΔV = 0, a rather favorable scenario for Fed policy in controlling inflation. In a similar fashion, emergence from a recession generally means that the slack economy allows real GDP to grow at a 3-4.5 percent rate, at least for the first year or two of recovery. Over the years simulated by Tables 7, 8 and 9, allow a 3 percent annual growth rate in real GDP, and allow it to extend throughout the entire time horizon simulated. This is a most generous scenario for controlling inflation, which is consequently simulated by the figures presented in Table 10. M1•V = P•Q (8) where V is the velocity of circulation, P is the average price index, Q is real GDP %ΔM1 + %ΔV + (%ΔM1)(%ΔV) = %ΔP + %ΔQ + + (%ΔP)( %ΔQ) %ΔM1 + %ΔV ≈ %ΔP + %ΔQ (9a) (9b) We can consider the inflation rates presented by Table 10 as the minimum rates we can expect over the various horizons assuming the multiplier for M1 is allowed to restore to its pre-crisis level of 1.66. Changes in the velocity of circulation are likely to be positive, however, rather than no change. The average growth rate in real GDP is not likely to equal 3 percent on average since it is currently below that figure. In addition, for reasons indicated below, the growth rate in the monetary base may actually be positive, perhaps significantly so. Since inflation rates in excess of 6 percent per year are certainly not desirable, it appears that Federal Reserve policy must be aimed at controlling the multiplier, preventing it from restoring to its pre-crisis level. The simulations indicate that actual reductions in the base by amounts greater than 3 percent per year, and over horizons of at least 8 years, are likely to be the only option the Fed has for controlling inflation. In fact, price stability occurs if the base is reduced by 50 percent over any of the horizons analyzed, provided velocity does not increase. Moderate inflation (3.26%) results from reductions on onequarter of the monetary base spread over ten years, but this again assumes a stable velocity and a real GDP growth rate of three percent. Increases in velocity, however, will require greater reductions in the base for the purpose of controlling inflation. To what extent might an increase in the required reserve ratio be useful for controlling the money multiplier? If the reserve to deposit ratio decreases from its current value of 2.08 down to .28, rather than to its pre-September, 2008, level of .14 (a doubling of the current required reserve ratio), then for the simulations presented in Table 6, βr = -1.80. For the purpose of the Table 6 calculations, this Proceedings of the 2012 Pennsylvania Economic Association Conference 181 moderate increase in the reserve-to-deposit ratio still results in the money multiplier increasing by approximately fifty-six percent rather than sixty percent. Increasing the required reserve ratio, therefore, has only a potential of moderately controlling the restoration of the money multiplier and consequent inflation. This is at first review a perplexing result, but this moderation of the Fed’s traditional tools for controlling the money supply is the natural consequence of the enormous flight to liquidity of the 2008 financial crisis. highly likely that substantial inflation will result. This occurs because a mobilization of excess reserves into bank loans will increase the money multiplier by upwards of 50 percent. This translates into a similar increase in the money supply. Given this, simulations indicate that a reduction of ¼ of the monetary base over a ten-year horizon results in a 3.26 percent annual inflation rate provided velocity of money does not increase (the demand for money does not drop). This latter contingency is unlikely during a robust recovery. If velocity increases by whatever percentage per year, then this percentage increase will be an addition to the inflation rate. If we have no change in the monetary base over the decade that follows the initial stages of loan restoration, i.e. if the Fed decides to continue its support of the Treasury bond and MBS markets for these long-term securities, then avoiding inflation rates of more than 5 percent per year is unlikely. The conclusion is unmistakable. CONCLUSION The money multiplier and money supply simulations presented above show some startling evidence concerning future monetary policy, i.e. if the Fed does not liquidate at least some of its holdings of Treasuries and MBS, and provided that bank loans do return to normal levels, it is Table 1a: Money Supply Components (in $billions except for Mult. 1)1 3/2006 3/2008 3/2010 3/2012 M1 $1,383.4 $1,387.7 $1,712.3 $2,220.7 Mult. 1 1.734 1.683 0.825 0.838 Monetary $798.0 $824.4 $2,074.6 $2,650.4 $43.9 $45.0 $1,186.9 $1,608.0 $1.5 $2.6 $1,120.3 $1,509.7 $42.4 $42.3 $65.6 $98.3 Base 2 Total Reserves Excess Reserves Required Reserves 1 2 See www.federalreserve.gov for sources of money supply data, “Series H.3.” The “Monetary Base” consists of bank reserves plus currency in the hands of the public. Table 1b: Annualized Growth Rates in M1, 2008 to 2011 Year 2008 2009 2010 2011 Annualized M1 Growth Rates 14.12% 7.67% 10.52% 18.09% Proceedings of the 2012 Pennsylvania Economic Association Conference 182 Table 2a: Basic Statistics for M1 and Its Multiplier Month M11 Base1 3/2012 8/2008 %β 5 2,239.2 1,400.9 +59.83 2,654.54 847.30 +213.29 Total Reserves1 1,606.48 44.13 3,540.34 Demand Deposits1 Currency2 Other Check.3 771.8 305.6 +152.55 1,033.1 774.8 +33.23 430.2 305.7 40.73 Mult.4 R c o .8435 1.6745 -48.98 2.08 0.14 1,344 1.34 2.54 -47.24 0.56 1.00 -44.28 1 Denoted in $billions. “Currency in the Hands of the Public” in $billions. 3 “Other Checkable Deposits” in $billions. 4 Multiplier = M1/Base = L in equation (2). 5 %β is the percentage change over the period 8/2008 through 3/2012. 2 Table 2b: Mean Averages and Standard Deviations Over Period 8/2008 to 3/2012 M1 Base Demand Deposits $426.6 Currency Multiplier r c o $1,526.4 Total Reserves $674.7 Means $1,617.9 $836.9 1.2343 1.314 2.0877 .8642 Std. Dev. 265.8 694.9 618.3 142.3 88.4 0.4066 1.080 0.3945 .1568 Table 3: OLS Estimates of Model Equation (5)1 Coefficient OLS S.E. of Est. Estimate β1 -.0453 .0737 β2 -.2492 .0252 β3 +.2701 .2077 R2 = .6066 F= p = .000 36.50 1 The constant-intercept term was suppressed in the OLS estimation. t-statistic pvalue .46 .000 .198 .000 -0.62 -9.90 +1.30 Table 4: OLS Estimates of Model Equation (7)1 Coefficient OLS S.E. of Estimate Est. β2 -.2551 .0247 R2 .5969 1 The constant-intercept term was suppressed in the OLS estimation. t-statistic pvalue .000 -10.32 Table 5: Descriptive Statistics Using Monthly Observations from Equation (5), 1/2006 to 3/2012 Coefficient Mean SE Mean t-statistic Minimum Maximum Range β1 -.0515 .0116 -4.44 -.1845 +.0793 .2638 β2 -.3099 .0082 -37.79 -.4212 -.2078 .2134 β3 +.2584 .0046 +56.17 +.2090 +.3552 .1461 Proceedings of the 2012 Pennsylvania Economic Association Conference 183 Table 6: Impacts on %βL of Restoration of r, c and o to 8/2008 Levels βL/L Variable %βL Change βr = - βr(Mean β2) = (-1.94)(-.3099) = 1.94 +.6012 βc = βc(Mean β1) = (+1.20)(-.0515) = +1.20 -.0618 βo = βo(Mean β3) = (+.44)(+.2584) = +.44 +.1137 60.12% -6.18% 11.37% Total = +.6531 Total = 65.31% Table 7: Annualized Growth Rates for Restoration of the M1 Multiplier from .82 to 1.661 1 N Year Horizon Growth Rate Per Year 3 Year 26.50% 5 Year 15.15% 8 Year 9.22% 10 Year 7.31% The multiplier was .82 in January, 2012, and 1.66 in July, 2008. Annualized growth rates were calculated by g = (1.66/.82)1/N - 1. Table 8: Annualized Growth Rates of Monetary Base with Overall Reductions as Stated Over Various Horizons Horizon 3 Years Base Reduced by ½ Over Horizon -20.63% Base Reduced by ¼ Over Horizon -9.14% 5 Years -12.94% -5.59% 8 Years -8.30% -3.53% 10 Years -6.70% -2.84% Proceedings of the 2012 Pennsylvania Economic Association Conference 184 Table 9: Annualized Growth Rates for M1 Given Assumptions for Growth Rates in the Multiplier as Presented in Table 7.1 Horizon 3 Years No Change in Base 26.50% 5 Years 15.15% 8.72% .0025% 8 Years 9.22% 5.36% .0015% 10 Years 7.31% 4.26% .0013% 1 Base Reduced by 1/4 14.94% Base Reduced by ½ .4000% %ΔM1 = %ΔMult + %ΔBase + (%ΔMult)( %ΔBase) Table 10: Inflation rates Given the Simulated Data of Tables 7 8, and 9, Stable Velocity, and 3% Growth Rates in Real GDP.1 Horizon 1 3 Years No Change in Base 23.50% Base Reduced by ¼ 13.94% Base Reduced by ½ -.6000% 5 Years 12.15% 7.72% -.0075% 8 Years 6.22% 4.36% -.0085% 10 Years 4.31% 3.26% -.0087% Using (3b), Inflation Rate = %ΔM1 - .03 since %ΔV = 0. REFERENCES Anderson, Richard G., Charles S. Gascon, and Yang Liu. 2010, “Doubling Your Monetary Base and Surviving: Some International Experience,” Federal Reserve Bank of St. Louis Review, November/December, p. 481-505. ______________ 2010b, “Federal Reserve’s Exit Strategy,” Testimony before the Committee on Financial Services, US House of Representatives, Washington, DC, March 25, 2010. www.federalreserve.gov/newsevents/testimony/bernanke2 0100325a.htm. Bernanke, Ben S. 2009, “The Crisis and Policy Response,” The Stamp Lecture at the London School of Economics, January 13, London, England. See www.federalreserve.gov/newsevents/speech/bernanke2009 0113a.htm. Blinder, Alan S. 2010, “Quantitative Easing: Entrance and Exit Strategies,” Federal Reserve Bank of St. Louis Review, November/December, p. 465-479. ______________ .1983, “Nonmonetary Effects of the Financial Crisis in the Propagation of the Great Depression,” American Economic Review, (73 (3): p. 257276. ______________. 2005, Essays on the Great Depression. Princeton University Press, Princeton, NJ. ______________ 2010a, “Federal Reserve’s Exit Strategy,” Testimony before the Committee on Financial Services, US House of Representatives, Washington, DC, February 10, 2010. www.federalreserve.gov/newsevents/testimony/bernanke2 0100210a.htm. Brown, E.C. 1956, “Fiscal Policy in the Thirties: A Reappraisal,” American Economic Review, (December). Bullard, James. 2010, “Three Lessons for Monetary Policy from the Panic of 2008,” Federal Reserve Bank of St. Louis Review, May/June, p. 155-163. Curdia, Vasco and Michael Woodford. 2010, “The Central Bank Balance Sheet as an Instrument of Monetary Policy,” presented at the 75th Carnegie-Rochester Conference on Public Policy, “The Future of Central Banking,” April 1617, 2010, www.carnegie-rochester.edu/april10pdfs/Curdia%20Woodford.pdf. Proceedings of the 2012 Pennsylvania Economic Association Conference 185 Federal Reserve. 2009, Detailed Information about the Federal Reserve’s Balance Sheet and Monetary Policy can be found at “Release H.4.1” http://www.federalreserve.gov/monetarypolicy/bst.htm Friedman, Milton and Anna Schwartz. 1963, A Monetary History of the United States, 1860-1960, Princeton University Press, Princeton, New Jersey. Hanson, A.H. 1941, Fiscal Policy and Business Cycles, W.W. Norton & Co., New York Hession, Charles H. and Hyman Sardy. 1969, Ascent to Affluence: A History of American Economic Development, Allyn and Bacon, Inc. Boston, Mass. Meltzer, Allan. 2010, “The Fed’s Anti-Inflation Exit Strategy Will fail,” Wall Street Journal, January 27, 2010. Mishkin, Frederic. 2010, The Economics of Money, Banking & Financial Markets, 9th edition, Addison Wesley, Imprint of Pearson, Boston, Mass. Robinson, Richard and Marwan ElNasser. 2010, “Escaping the Trap: Prospects for Federal Reserve Policy During the Recovery,” Financial Decisions, Volume 22, Number 1. Taylor, John B. 2010, “An Exit Rule for Monetary Policy,” Working Paper, Stanford University, February 10, 2010, www.stanford.edu/~johntay/House%20FSC%2010%20201 0.pdf. ENDNOTES 1 The history presented here is largely taken from Blinder (2010), Anderson, et al. (2010), Bullard (2010), Taylor (2010), and Curdia and Woodford (2010). 2 This decrease in the money multiplier did not result in a decrease in the money supply since the Fed radically increased the monetary base. 3 As of 3/21/2012. See www.federalreserve.gov, “Release H.4.1.” 4 See Robinson and El Nasser (2009), and also Friedman and Schwartz (1963), for a review of this monetary policy explanation of the 1936-37 experience. Also see Hession and Sardy (1969) for the economic history of this 1936-37 experience. For various fiscal policy interpretations of this downturn see Hanson (1941), and Brown (1956). 5 See Table H.6, at www.federalreserve.gov for definitions of M1 and M2. 6 Since L = M1 = Base currency+demand deposits+other checkable deposits currency+reserves then dividing both numerator and denominator by demand deposits gives (2). 7 The M1 multiplier reached .90 in May of 2009, and has continued a slight declining path since then. Proceedings of the 2012 Pennsylvania Economic Association Conference 186 WORLD OIL PRICES: ECONOMIC IMPACT AND ECONOMETRIC FORECAST Carrie R. Williams Department of Economics Undergraduate Clarion University of Pennsylvania Clarion, PA 16214 ABSTRACT Early in any principles of economics course students examine how oil shocks and changes in the price of crude oil can impact markets and the macroeconomy. A rise in the price of oil will cause the supply of most manufactured goods to decline leading to a price increase in those markets. Together this can lead to an overall increase in aggregate prices and unemployment known to economists as stagflation. Minimal empirical research has been done to assess the precise relationship between changes in oil prices and important macroeconomic variables. This research aims to explore this relationship by evaluating oil price movements in the world market and important linkages to macroeconomic variables in the United States beginning with data in 1919. Analysis will be qualitative and quantitatively based through the use of simple correlation analysis, multivariate regression analysis, and within sample forecasting models. This will help identify the models with the lowest root mean squared error (RMSE) to operate out-of-sample forecasts and produce rational predictions of oil prices for the months of March 2012 to February 2013. By analyzing results of these models, this study is able to elaborate on the relevance macroeconomic variables have on the accuracy of various forecasting models when examining a highly dependent variable such as oil prices. INTRODUCTION With the shock of West Texas Spot Price of Oil reaching record highs in April of 2011 at a rate of $110.040 a barrel, economists, politicians, and citizens alike are anxious about the coming months. After an almost $15 spike in oil prices from February to March of 2011, concerns were reconciled with a slow decline of prices returning to the $80 a barrel range in the fall of 2011. Concurrently, growth rates were reported at 2.3 percent at the time of the oil price shock and dropped to below 0.5 percent for the two consecutive quarters of the declining oil prices. Oil prices and growth rates began to climb at the start of the July to September quarter. Although the economy has shown respectable improvement in the past quarters, the nation is continually affected by the increase in the price of oil. Companies in attempts to maintain profits with increasing production costs are struggling to avoid cutting wages or worse, worker layoffs. Citizens whom are making better wages are being forced to spend their income on high gasoline prices and increased prices of goods and services such as fuel efficient vehicles. For example consumer spending, which is primarily focused on automobile sales, rose at a rate of 2.1 percent in the fourth quarter according to an article published by Associated Press (2012). When it comes to economic growth, the numbers for the last quarter of 2011 reflected the highest growth rates since the spring of 2010 at 3 percent. Unfortunately, growth rates for the start of 2012 are expected to slow once again. Will this mean oil prices will remain above $100 a barrel in the next year or will prices fall once again as they did in 2011? This study looks to examine through various forecasting models which of these outcomes will most likely take place. According to patterns in the data, forecasting models that incorporate macroeconomic variables will provide the best estimate of the next year’s oil prices. Models being evaluated include linear trend, quadratic trend, two Auto-regressive Integrated Moving Average (ARIMA) models, two Ordinary Least Squares (OLS) models, and two Vector Autoregressive (VAR) models. Previous literature primarily focuses on VAR models; therefore, this study will examine the six models; the ones with the lowest RMSE when performing within sample forecasts will be used to make the best fitting out-of-sample predictions of oil prices for the next twelve months. LITERATURE REVIEW When it comes to research in the field of macroeconomic variables and oil prices, many researchers have focused on predicting GDP growth rather than oil prices. Researchers such as Hamilton (2008, 2009) have shown that predicting GDP growth in correlation to oil prices is difficult due to the volatility of oil prices. Decreases in oil prices have Proceedings of the 2012 Pennsylvania Economic Association Conference 187 been found to have significantly less of an impact on GDP growth than those of increasing periods. Studies have shown that 10 out of 11 recessionary periods in United States history have been preceded closely by sharp increases in oil prices. (Hamilton 2005) According to Kilian, in order to determine this relationship a nonlinear and linear approach to forecasting should be taken to avoid biased results. Research that has been done on forecasting oil prices and that have provided linkages to macroeconomic variables have focused solely on the VAR models. They provide effects of high oil prices as being consumer spending increases, reduced economic output in the short term resulting in worker layoffs or wage cuts, a negative effect on the purchasing power of the U.S. dollar. (EIA 2012) Aligning with previous studies decreases in oil prices have not been successfully linked to the same responses in the economy. This study looks to further examine these effects through incorporation of a variety of forecasting models and comparing their results to historical patterns as well as their correlation to prominent macroeconomic variables. Wu and McCallum (2012) make the clear distinction between oil futures prices and oil spot prices; yet, include both price categories in their forecast model. This study does not investigate this difference in prices, but instead utilizes spot prices in order to better capture interest rates such as Hotelling’s (1929) research proved effective. This is further supported by research done by Cologni and Manera (2005) where oil price shocks were found to have an unexpected impact on interest rates. As a result, this study looks to further support the relationship oil prices have with the macroeconomy. VARIABLE INPUTS West Texas Spot Price of Oil: This variable is measured in dollars per barrel. West Texas Intermediate (WTI) also called as Texas light sweet is defined as a type of crude oil used as a benchmark in oil pricing and is the underlying commodity of New York Mercantile Exchange's oil futures contracts. WTI is light crude, lighter than Brent crude. It contains about 0.24% sulfur and is rated as sweet crude, again sweeter than Brent. WTI is refined mostly in the Midwest and Gulf Coast regions in the U.S., since it is high quality fuel and is produced within the country. (Oil and Gas IQ 2012) Real Gross Domestic Product: Real GDP is gross domestic product in billions of constant dollars. In other words, it is a nation's total output of goods and services, adjusted for price changes. (Investor Glossary 2012) Housing Starts: The number of residential buildings (single-family and multi-family) construction units in thousands begun during a given time period (usually one month) based on the number of building permits issued. Housing starts are sensitive to interest rate changes and reflect the household sectors willingness to invest in new construction. It is a key indicator of business-cycle activity. More specifically, it is one of the 12 leading economic indicators tracked by the Bureau of Economic Analysis. (Economic Glossary 2012) Consumer Price Index: The Consumer Price Index (CPI) is a measure of the average change over time in the prices paid by urban consumers for a market basket of consumer goods and services with 1982-84 as the base year. (BLS 2011) Index of Industrial Production: An industrial production index (IP) is an index covering production in mining, manufacturing and public utilities (electricity, gas and water), but excluding construction with 2007 as the base year. (OECD 2001) To see graphical representations of the above variables for the past fifteen years, refer to Figures 14. Graphically, the data shows similar time series structure to oil prices in the GDP, CPI and IP datasets, but differing trends in housing starts compared to oil prices. This will be essential in the conclusions of the various forecast models. STRUCTURE OF FORECASTING MODELS Each model is initially evaluated as a within sample forecast. To elaborate, within sample forecasts use the model to predict oil prices for March 2011 to February 2012, values of which are known prior to running the forecast. Examining the root mean squared error, RMSE of each within sample forecast demonstrates how accurate the forecast model was in predicting the actual values of oil prices. The following formula is used to determine the RMSE, with the difference between the forecasted value and the actual value represented by Fi minus Ai. ∑π(πΉπ −π΄π )2 οΏ½ π (1) As one would expect, the models with the smallest difference between the forecasted and actual values will have the lowest RMSE. The models with the Proceedings of the 2012 Pennsylvania Economic Association Conference 188 lowest RMSE will presumably output the closest prediction of oil prices for the unknown or out-ofsample forecast of March 2012 to February 2013. Trend Models: When using trend models to forecast future oil prices, economic variables are not included either directly or indirectly into the analysis. Future prices are reliant upon the linear or quadratic trend resulting from past prices, not past prices themselves. As the graphical representation shows in the Appendix, oil prices do not follow a strictly linear or quadratic trend; therefore, the RMSE is expected to be insignificant for both linear and quadratic trend within sample forecasts. Linear Trend Model: A linear equation (2) is used to forecast the oil prices over time, Pt as a function of a constant, C and time, t with εt being the normally distributed error term with zero mean and constant variance. Pt = C + βt + εt (2) Quadratic Trend Model: Here, a quadratic equation with oil prices over time is forecast using the trend model with an additional t2 term, which allows for concavity in the trend to occur. Pt = C + βt + βt2 + εt (3) linkage is further evaluated and considered when making conclusions on the resulting predictions. ARIMA Model (1,2,0) ARIMA Model (1,2,2) Structural Ordinary Least Squares (OLS) Approach: After running simple correlation matrices (see Tables 1-3) with the macroeconomic variables described in the input variable section, the strength of the relationship to oil prices is determined on a scale. Strong correlation exists when variables report a value between +/- 0.70 and 1, otherwise weak to no correlation applies. Oil prices are shown to be strongly correlated with consumer price index (CPI), industrial production (IP), and real gross domestic product (GDP). This can be further described visually when comparing the graphs of the individual variables in Figures 1-4. Through lagging those correlated macroeconomic variables by 12 months, this approach provides a forecast of oil prices with a direct linkage to the variables in question. To further explain, the association can be made on how influential last year’s CPI, IP, and GDP were on today’s oil prices. The two models chosen for this study are as follows with a constant, C and time, t: OLS1 (lagCPI, lag IP, lag GDP): Pt = C + β1CPIt-12 + β2IPt-12 +β3GDPt-12 (5) Univariate Auto-regressive Integrated Moving Average (ARIMA) Models: This type of forecast is formulated to be correlated with past oil prices as well as past shocks or errors; where the autoregressive, AR portion is a function of past oil prices and the moving average, MA indentifies errors and inconsistencies. Pt = β0 +β1Pt-1+…+ βpPt-p + θ1εt-1 + θ2εt-2 +…+ θqεt-q + εt (4) The differing factor, I represented by d in ARIMA (p,d,q) completes the model by removing the nonstationarity of the generalized ARMA model. Using the Box-Jenkins procedure, possible combinations of p, d, and q are identified through comparisons in the output diagrams of the Autocorrelation function and the Partial Autocorrelation function. Once combinations are categorized, the model with the lowest RMSE is used for further analysis. The two models listed below were chosen as the optimal cases for study. Economists believe macroeconomic variables can be indirectly linked to this model through interest rates projected through past oil prices. This indirect OLS2 (OLS1 + time): Pt = C + β1CPIt-12 + β2IPt-12 +β3GDPt-12 + t (6) Vector Autoregressive (VAR) Models: The final model type assesses the forecast with the same macroeconomic variables as the OLS approach, but the variables are not calculated and examined in the same format. Here, each variable is dependent upon the number of determined lags of each of the other variables along with itself. For example: Pt = C + β1Pt-1 +α1CPIt-1 + θ1IPt-1 + λ1GDPt-1 + t +t2 +εt (7) VAR lag 1 and VAR lag 9 were determined as being the best suited for this study through within sample forecasts to identify the lowest RMSE. A lag of 1 represents one month and without fail a lag of 9 represents 9 months or ¾ of a year. Being that the macroeconomic variables are not only lagged, but also reflective of the interdependence of these variables in the economy, expectations for significant results are high. Proceedings of the 2012 Pennsylvania Economic Association Conference 189 EMPIRICAL RESULTS The following table illustrates each model and its RMSE found through within sample forecasts. The results show the ARIMA and VAR models provided the lowest RMSE, while both trend models were significantly greater, as expected. In addition to high RMSE results, exempting the trend models from further analysis was confirmed by the lack of correlation to the macroeconomic variables of this study. However, the remaining models with relatively lower RMSE results, either directly or indirectly incorporated the macroeconomic variables and therefore out-of-sample forecasts were computed using ARIMA (1,2,0), ARIMA (1,2,2), OLS1, OLS2, VAR lag 1 and VAR lag 9. TABLE 4: RMSE RESULTS Model RMSE Trend 49.5413 Quadratic Trend 33.321 ARIMA (1,2,0) 8.7986 ARIMA (1,2,2) 8.8007 OLS1 (lagCPI, lagIP, lagGDP) 26.2274 OLS2 (OLS1 + time) 26.436 VAR lag 1 9.0285 VAR lag 9 9.8585 The results of the out-of-sample forecasts illustrated sweeping differences across the six models. The oil prices predicted by all six models are detailed in the table as well as graphically in Table 5 and Figures 510. The results show similarities in the increasing trend of oil prices in the ARIMA (1,2,0), ARIMA (1,2,2) and VAR lag 1 models. Yet, the difference in the rate at which they increase and the final price at which the year will conclude is inconsistent. Alternative conclusions can be seen when examining the remaining models, where decreases in the prices of oil occur. CONCLUSION As discussed above, this study found the ARIMA (1,2,0), ARIMA (1,2,2) and VAR lag 1 to be similar; nonetheless, the VAR lag 1 model has the most potential of the three. Both ARIMA models have significant increases in oil prices over the next year with about a 20% increase in ARIMA (1,2,0) and an overwhelming increase of about 130% in ARIMA (1,2,2). With past trends in oil prices and the state of the current economy, these models do not incorporate the macroeconomic variables to the extent necessary. With the support of past research, increases in oil prices have been shown to coincide with increases in economic growth rate. Using the ARIMA (1,2,2) model, oil price increases such as this would coincide with above average increases in the economic growth rate, an unsustainable prospect for a developed economy such as the United States. According to research in the field of economic development, a developed country will not grow at rates greater than about 5% annually. Consequently, oil prices and the economic growth rate of the United States would stabilize or decrease at some undisclosed period of time. With the current slowing of the economic growth predicted to occur at the end of the first quarter of 2012, we might expect to see prices of oil to decreases or remain stable rather than increase. Therefore, the ARIMA models are not chosen as the most accurate models for predictions of oil prices in the next year based primarily of poor macroeconomic integration and unrealistic results for an economy such as the United States. The models with the strongest predictions of oil prices are the VAR lag 1 and VAR lag 9, since both models incorporate the interdependence of the macroeconomic variables and qualified as the second model type in terms of accuracy. These two models are similar in structure, but a lag of 1 is a one month analysis compared to a lag 9 that is three quarters of the entire year. The difference in the lag length yields notably different predictions for the coming year. Beginning with the VAR lag 1 model, an increase of about 0.62 percent over the next year is a more honest prediction than that of either ARIMA models. Being that this model allows for the next months prices and values for other economic variables to be predicted through the previous months values, this model should accurately forecast the relationship oil prices have with the entire economy. Since CPI and IP drive more than 70% of economic growth, these macroeconomic variables are vital to the continued growth of the economy (Zhang 2012). Increases in oil prices typically increase the costs of goods and services and increase production costs for producers, especially in oil related industries, which can easily decrease the rate at which the economy grows. With increases as gradual as this model predicts, oil prices would not have that dynamic of an effect on economy growth. Even though the economy experienced an unexpected 3% growth rate last quarter, growth rates are predicted to fall to around the 2.5% range for the next quarter, according to Zhang (2012). If the growth rates continue to stay at or around 2 percent, Proceedings of the 2012 Pennsylvania Economic Association Conference 190 we should expect this model to be the best forecast of where oil prices will be within the next year. FIGURE 2: CONSUMER PRICE INDEX 150 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 U.S. Dollars 1982-84 Base Year 200 Year FIGURE 3: INDEX PRODUCTION OF INDUSTRIAL Index of Industrial Production 110 60 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Overall, the oil price forecasts depict various possibilities for the future, but only one prediction will be accurate. As the results have shown, models with the ability to portray the macroeconomic variables and their linkage to oil prices will have the highest probability of being accurate. Whether studies focus on the ARIMA model or the VAR models, oil prices rising are inevitable. Consumer Price Index 2007 Base Year Finally, the VAR lag 9 model offers a hopeful future in the eyes of the consumers, but not necessarily for the economy in terms of growth initially. Being able to incorporate oil price shocks and errors throughout a longer period of time allows for a more accurate statement of the economy. Here, we notice the prices of oil decline instantly and continue to decline until October 2012, where minor increases occur to meet the final price of $103.40 per barrel in February 2013. A decrease followed by an increase in oil prices parallels the historical pattern (Figures 1-4) as well as incorporates the possible recovery of the economy in the near future. This model incorporates historical trends in both oil prices and the macroeconomic variables, has shown accuracy in predicting within sample oil prices and corresponds well with the reality of the current economy; therefore, the VAR lag 9 is the optimal model in which to predict oil prices for March 2012 to February 2013. Year FIGURE 4: HOUSING STARTS FIGURE 1: WEST TEXAS SPOT OIL PRICES 3000 2000 1000 0 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 In thousands Housing Starts Year Proceedings of the 2012 Pennsylvania Economic Association Conference 191 TABLES 1-3: CORRELATION MATRICES FOR OLS MODELS Table 1: P LagCPI P LagIP 1 LagCPI 0.80879 1 LagIP 0.77598 0.96905 1 Table 2: P LagCPI LagIP LagHS P LagCPI LagIP 1 0.77627 0.75039 -0.2486 1 0.97049 1 LagHS 1 Table 3: P P LagCPI LagIP LagRGDP LagCPI LagIP LagRGDP 1 0.80758 1 0.7763 0.969 1 0.81512 0.9872 0.99212 1 FIGURE 5: ARIMA (1,2,0) TABLE 5: FORECAST RESULTS 12-Mar 12-Apr 12-May 12-Jun 12-Jul 12-Aug 12-Sep 12-Oct 12-Nov 12-Dec 13-Jan 13-Feb ARIMA (1,2,0) 104.13619 106.07203 107.99265 109.92229 111.8519 113.78487 115.71992 117.65753 119.59753 121.53998 123.48486 125.43216 Out-of-Sample Forecast Results ARIMA (1,2,2) VAR lag 1 VAR lag 9 110.63516 102.64344 102.75974 121.37798 103.03889 101.35284 133.3922 103.43644 99.50581 146.09218 103.83614 98.78431 159.16216 104.23808 98.04998 172.43188 104.64233 98.33978 185.8096 105.04893 98.93201 199.2458 105.45798 100.10676 212.71381 105.86952 101.08132 226.19923 106.28362 102.17473 239.69432 106.70035 102.92626 253.19489 107.11975 103.40427 OLS1 74.07549 74.68449 73.78339 74.71326 72.0032 71.16884 72.40003 71.41613 71.33563 72.13512 71.12897 70.7487 Proceedings of the 2012 Pennsylvania Economic Association Conference OLS2 74.17126 7479378 73.88547 74.81686 72.07515 71.23466 72.46633 71.4733 71.39623 72.17596 71.16102 70.78098 192 130 110 90 Month West Texas Spot Oil Price (dollars per barrel) West Texas Spot Oil Price (dollars per barrel) ARIMA (1,2,0) FIGURE 8: VAR LAG 9 VAR lag 9 102 97 Month ARIMA (1,2,2) 190 90 Month FIGURE 9: OLS 1 West Texas Spot Oil Price (dollars per barrel) West Texas Spot Oil Price (dollars per barrel) FIGURE 6: ARIMA (1,2,2) OLS 1 74 69 Month VAR lag 1 105 100 Month FIGURE 10: OLS 2 West Texas Spot Oil Price (dollars per barrel) West Texas Spot Oil Price (dollars per barrel) FIGURE 7: VAR LAG 1 OLS 2 80 75 70 65 Month Proceedings of the 2012 Pennsylvania Economic Association Conference 193 REFERENCES (2001,). Industrial Production Index. Economic Outlook. Retrieved Apr 2, 2012, from OECD (2011). US Business Cycle Expansions and Contractions. National Bureau of Economic Research, Consumer Price Index. Bureau of Labor Statistics. Retrieved Apr 9, 2012, from U.S. Department of Labor Economic Effects of High Oil Prices. Retrieved Apr 13, 2011, from U.S. Energy Information Administration (2011, Dec 7). Oil Price Developments: Drivers, Economic Consequences and Policy Responses. from OECD Hamilton, J. (2008). Understanding Crude Oil Prices. National Bureau of Economic Research Working Paper Series, (2012). U.S. Economy at a Glance: Perspective from the BEA Accounts. Bureau of Economic Analysis, Hamilton, J. (2009). Causes and Consequences of the Oil Shock of 2007-08. National Bureau of Economic Research Working Paper Series, (2012). U.S. GDP Revised Up to 3% in Q4. TradingEconomics.com, (2012, Mar 29). Final 4th-Quarter GDP Report Shows Solid Growth. Associated Press In Economic Definition of Housing Starts. Retrieved Apr. 9, 2012, from http://glossary.econguru.com/economic-term/housing starts In Economic Forecasting. Retrieved Apr. 6, 2011, from http://www.investopedia.com/terms/e/economicforecasting.asp In IQ Glossary. (chap. West Texas Intermediate) Retrieved Apr. 10, 2012, from http://www.oilandgasiq.com/glossary/west-texasintermediate-(wti) In Introduction to ARIMA: nonseasonal models. Retrieved Apr. 1, 2012, from http://www.duke.edu/~rnau/411arim.htm Hooker, M. A. (1996). What Happened to the Oil Price-Macroeconomy Relationship?. Science Direct, 38 (2), pp. 195-213. Hotelling, H. (1929). Stability in Competition. The Economic Journal, 39 (153), pp. 41-57. Hudson, E. A, and D. Jorgenson. (1974). U.S. Energy Policy and Economic Growth 1975-2000. Bell Journal of Economics and Management Science, Hui, G, and K. Kliesen. (2005). Oil Price Volatility and U.S. Macroeconomic Activity. Federal Reserve Bank of St. Louis, pp. 669-84. Kennedy, A.. (Executive Producer). (2011, Dec. 9 ). [Television broadcast]. CBS News. Kilian, L. Not All Oil Price Shocks Are Alike: Disentangling Demand and Supply Shocks in the Crude Oil Market. The American Economic Review, 99 (3), pp. 1053-1069. In Investor Glossary. Retrieved Apr. 6, 2012, Montagne, R.. (Executive Producer). (2011, Mar. 31 ). [Television broadcast]. National Public Radio. Brown, S. P.A., and M. Yucel. (2002). Energy Prices and Aggregate Economic Activity: An Interpretative Survey. Science Direct, 42 (2), pp. 193-208. OFarrell, R. (2011). Are Goldman Sachs’ Predictions Accurate?. Insider Monkey, Cohan, P. (2011, Feb 23). What Do Rising Oil Prices Mean for U.S. Economic Growth?. Daily Finance Roubini, N, and B. Setser. (2004). The Effects of the Recent Oil Price Shock on the U.S. and Global Economy. Cologni, A, and M. Manera. (2005). Oil Prices, Inflation and Interest Rates in a Structural Cointegrated VAR Model for the G-7 Countries. Selected Works, Shenk, M. (2012, Apr 2 ). Oil Rises Most in Six Weeks on U.S. Manufacturing. Bloomberg News Williams, J. L, and F. Alhajji. (2003). The Coming Energy Crisis?. Oil and Gas Journal, Proceedings of the 2012 Pennsylvania Economic Association Conference 194 Wu, T, and A. McCullum. (2005). Do Oil Futures Prices Help Predict Future Oil Prices?. Economic Research and Data, Zhang, M. U.S. GDP Growth Rate Forecast: 3% in Q4 on Consumer Spending, Q1 Likely Weaker. International Business Times, Proceedings of the 2012 Pennsylvania Economic Association Conference 195 ADVANTAGEOUS SELECTION IN HEALTH INSURANCE Paul Sangrey, Grove City College, 1025 Bells Hill Road, Landisburg, PA 17040 ABSTRACT Advantageous selection refers to people with higher levels of risk buying lower levels of insurance because of the correlation between having a high level of risk and the willingness to live with a high level of risk. It interacts with adverse selection to create an ambiguous correlation between health status and purchasing insurance. Advantageous selection can exist because people who have a lower desired level of risk will take actions that reduce their overall risk and have some control over their own level of risk. Advantageous selection has been demonstrated empirically by Wang, Huang, and Tzeng; by Eldridge, Koc, Onur, & Velamuri; and by Fang, Keane, and Silverman. Starting with Akerlof’s seminal article (1970), information asymmetries and their resulting effects have become a major focus of economic analysis. Health Economics is particularly notable in this area because of the large number of information asymmetries and the importance of their effects. Perhaps the most important of these is adverse selection in the insurance market – people who have a higher level of risk than the public information would predict taking advantage of the asymmetric information and purchasing more insurance because they face a relatively lower price. In recent years, researchers have begun to investigate the possibility of advantageous selection. Advantageous selection refers to people with higher levels of risk buying lower levels of insurance because of the correlation between having a high level of risk and the willingness to live with a high level of risk (de Meza & Webb, 2001). It has been most conclusively demonstrated in certain non-health insurance markets. However, advantageous selection is present in the health insurance market and has a significant dampening effect on adverse selection as evidenced by the ambiguous correlation between health status and insurance ownership. When deciding how much to charge for insurance, the insurance company looks at the average risk and calculates its cost accordingly. If the insured group did not differ materially from the general population, then the average risk for the insured population would be the same as the population as a whole, and so the actuarially fair premium would be the same. However, the adverse selection literature has demonstrated that people who choose to buy insurance are not a random sample drawn from the population (Rothschild & Stiglitz, 1976). The adverse selection model is built on four assumptions. First, insurance purchases are voluntary. Second, risks are heterogeneous. Third, heterogonous risks are lumped together and charged the same price. Fourth, prospective purchasers know more about their own risk than the insurance company does (Hemenway, 1990, p. 1065). risk in the insurance population is higher than the population as a whole forcing the insurer to charge a higher price and, thereby, leading to a large number of relatively low risks dropping out of the market. This analysis is relatively noncontroversial, and highly influential. For example, this is the reason behind the mandate in the Patient Protection and Affordable Care Act of 2010 (Marcus, 2012). In doing so, the bill would force everyone to become part of the insurance pool, thereby making the average risk of an insured person equal to the average risk of the general population. However, this is not the only way in which the insured population differs from the general population. People who choose health insurance do so because they are willing to give up a certain level of current wealth in order to reduce their risk. In other words, they reduce their expected income but increase their expected utility (Folland, Goodman, & Stano, 2001). However, the level at which people optimize their total utility by substituting certain income for expected income differs from person to person. As a result, the preferred level of insurance coverage will differ between people simply because of tastes and preferences and be entirely unrelated to their expected loss. Insurance minimizes the difference between the risk being faced by the insuree and the insuree’s utility-maximizing level of risk by decreasing the expected loss, and thereby, the level of risk born by the actor. If the utility-maximizing level of risk is equal to or above the expected risk the person in question will not purchase insurance. In adverse selection, for people with a higher expected loss the expected risk is higher, and so the desired level of insurance is higher. However, if people with a higher expected loss, and thereby a higher expected level of risk, also have a higher utility-maximizing level of risk, they may not choose to purchase insurance anyway, and so the adverse selection problem would be smaller or even nonexistent. This leads people who have higher risks to buy more insurance than people with lower risks. Therefore, average Proceedings of the 2012 Pennsylvania Economic Association Conference 196 As in Figure 1, a person with a desired risk of DR1, in other words, a utility-maximizing level of risk of DR1, and an expected loss of EL1 would purchase insurance. However, if the desired risk rose to DR2, the person would no longer be interested in buying insurance, unless the expected loss also rose to at least EL2. This tendency would not have any effect, however, if expected loss and desired risk were independent of each other, and would actually lead to greater levels of adverse selection if they were negatively correlated. If they were positively correlated, especially significantly so, it would have a significant dampening effect on the level of adverse selection in the market. Postulating such a correlation rests upon two assumptions. First, people who have a lower desired level of risk will take actions that reduce their overall risk. Second, people have some control over their own level of risk. If these two conditions hold, people who buy more insurance will take actions to reduce their expected loss in areas where they have some control over that risk, thereby reducing the expected loss seen by the insurance company, and the actuarially fair premium. Moral hazard and adverse selection effects will also exist contemporaneously, and so the net effect could be positive or negative, but that does not deny the existence of an advantageous selection effect. The first of these two statements is not that shocking at first glance. People who prefer a lower level of risk will presumably both buy insurance to lower their level of risk and take actions to reduce the actual risk of a loss. However, as Ehrlich and Becker (1972) demonstrated, insurance and self-protection can be both complements and substitutes. Donder and Hindriks (2006) argue that the only equilibrium that will exist in equilibrium is a separating equilibrium, where insurance companies offer only a few contracts. In order to get the risk avoidant people in one group, they actually offer them either more insurance than they would prefer or less than they would prefer, in such a way that the people choose a greater quantity than they would if they could determine the quantity. Essentially, insurance companies limit the possibilities of insurance that the potential insurees have, and use their decision in that situation to discover information about them. The insurance companies then use that information to optimize their decisions. This forces people to buy insurance to reduce their risk to below the optimal level given the cost of the insurance. Using the Ehrlich and Becker’s result, Donder and Hindricks argue this can go to the point where people who were originally more risk avoidant may actually engage in a lower level of self-protection than people who were originally more risk avoidant because they have a higher level of insurance, and this reduces their risk below the optimal level. However, even if insurance companies offer heavily risk avoidant people more insurance than they would want in order to separate people by their risk level, in any insurance market it is not possible for the person to place all of the risk onto the insurance company. This is especially true in the health insurance market. People almost always have to pay the time costs associated with the medical expense. They also have to deal with the pain of being sick. Furthermore, a person rarely insures against the entire monetary loss. Therefore, the insuree still faces a certain level of risk no matter what level of insurance they have purchased, and a heavily risk avoidant person would attempt to minimize this risk. In doing so, they would necessarily limit the risk faced by the insurance company. There may be a few individuals who were originally somewhat risk avoidant and the level of insurance they are forced to take to create the separating equilibrium as described by Donder and Hindricks (2006) reduces their level of risk to such a great extent that they would substantially reduce their efforts expended on selfprotection. However, this would only apply to the people who marginally belong to the heavily risk avoidant group. Therefore, the majority of heavily risk avoidant people would engage in more self-protection than people who are not very risk avoidant, which would make them better risks, thereby reducing the actuarially fair premium to insure them. The second proposition, that people have some control over their own level of risk seems harder to prove, and for some kinds of healthcare costs, such as genetic diseases, may not apply. However, this is not the case at all for most forms of health care. The most obvious examples are areas such as HIV/AIDS. If someone is highly risk avoidant, they will be less likely to engage in risky sexual behavior or to share needles. Likewise, people who are risk avoidant are less likely to smoke, use illegal drugs, or binge drink, thereby reducing their risk and the risk of the insurance company. More importantly, someone who is risk avoidant will invest in a greater level of health capital through exercise, good nutrition, and preventative care. This reduces their risk of a loss as well. In fact, this even applies in areas such as genetic diseases where the patient has no control over their risk of contracting the disease but still has a substantial degree of control over the strength of their immune system and, therefore, the associated expense of fighting it. This last argument is a little less strong for investing in human capital is inherently risky, as evidenced by psychological studies that have demonstrated that higher educated people have lower levels of risk aversion (Shaw, 1996). This would have a mitigating effect on people investing in health capital to reduce their expected loss. They would be substituting one form of risk for another. However, many ways of investing in health capital, such as good nutrition and good exercise, protect the investor from a variety of diseases. In a sense, a person is able to diversify Proceedings of the 2012 Pennsylvania Economic Association Conference 197 his investment across a number of different diseases, thereby reducing the risk of investing in health capital. Furthermore, certain risky health behaviors such as smoking, drug use, and certain sexual behaviors have significant nonhealth losses associated with them as well. As a result, someone who was risk-avoidant would engage in a lower amount of these behaviors than would be caused by just looking at their desire to avoid the health loss. This leads to an even greater level of advantageous selection, and an even further decreased level of risk seen by the health insurance company. However, even though advantageous selection is theoretically possible, its existence as a statistically significant phenomenon is an empirical question, not a theoretical one. Some strong evidence exists for it in certain insurance markets. Most importantly, Wang, Huang, and Tzeng (2009) performed one of the better studies testing for advantageous selection. They focused their research on the commercial fire insurance market in Taiwan, not in health insurance. However, the theoretical model is general and applies to almost, if not all, forms of insurance, and so if it is demonstrated to be valid empirically in one insurance market, it is more likely to be so in other insurance markets, such as health insurance. Wang et al. conducted research on a certain type of company in the two largest cities in Taiwan – class-A was the technical term used. They used data on the actual incidence of fires, not just insurance claims, in order to count all of the fires, not just those that were reported to the insurance companies. They also conducted information on protection activities that was not observable by the insurance companies. They measured this using data on whether the fire safety equipment worked. The insurance companies knew about the existence of the equipment but did not have information regarding whether it actually worked. They also collected data on informal fire defense organizations within the firm, which were relatively common and normally undergo twiceyearly training and have responsibilities that reduce the risk of a fire (Wang et al., 2009). They found that, “(1) firms that purchase market insurance have a greater tendency to make an effort to engage in selfprotection; (2) firms that engage in self-protection are less likely to suffer a fire accident; and (3) firms with commercial fire insurance have a lower chance of suffering a fire accident than those without such insurance,” (Wang et al., 2009, p. 18). The first conclusion applies directly to the health insurance market. If someone who purchases fire insurance also engages in self-protection, then by analogy someone who purchases health insurance would engage in activities that reduce one’s risk of getting sick. The second conclusion also seems reasonable for the reasons described previously: people have some control over their risk of getting sick by limiting certain risky activities and investing in health capital. The third conclusion, however, may not be valid in the health insurance market. If the risk of realizing a loss because of a fire has a lower correlation with information the insuree has than the risk of realizing a loss because of a health emergency has, then the information asymmetry will be larger in the health insurance market. This will lead to greater adverse selection effects, which could very well offset any advantageous selection effects. To further substantiate the statistically significant existence of advantageous selection in the health insurance market, one can look at some studies that found a negative correlation between buying health insurance and risk in the health insurance market. Fang, Keene, and Silverman conducted research in the Medigap insurance market. The Medigap insurance market is ideal for testing for advantageous selection for two reasons. First, the market is heavily regulated. Specifically, in 47 states, there are only ten standardized Medigap plans, and in a designated enrollment period the insurer cannot deny coverage, place conditions on a policy, or charge more for pre-existing conditions (Fang, Keane, & Silverman, 2006, p. 305). This essentially strengthens the information asymmetry because the insurance company is not allowed to use some of the information available to it, and so one would expect to cause a great deal of adverse selection. Secondly, the Medigap program is closely linked to the Medicare program. As a result, the researchers had access to the Medicare Current Beneficiary Survey (MCBS), which combines survey data with Medicare administrative records, and the Health and Retirement Study (HRS), which is a longitudinal study covering large sample of the people eligible for Medigap insurance. Between the two sources they had access to data on health expenditures, diagnoses, treatments, and risk aversion (Fang et al., 2006, p. 305). They found that medical expenditures for senior citizens with Medigap coverage were about $4,000.00 less than those without. Furthermore, once they adjusted for health, which the insurance companies could not use in pricing the plans, people with Medigap spent about $2,000.00 more than those without. In other words, the reason that people with Medigap plans spent $4,000.00 less on average, rather than $2,000.00 more on average was because they were healthier (Fang et al., 2006, p. 306). Because the insurance companies could not use the insured’s health status in pricing, essentially the insurance companies did not know the health status of the patients, which, as mentioned above, would be expected to lead to substantial adverse selection. However, the likelihood of moral hazard combined with the observed advantageous selection points to a substantial advantageous selection effect. This test was for correlation, and not causation, and so it does not directly argue that the observed correlation was because Proceedings of the 2012 Pennsylvania Economic Association Conference 198 of people taking actions to mitigate their risk, as the fire insurance study did. However, the correlation has to be explained, and this requirement does substantially support the theoretical model because the theoretical model predicts what the empirical study has found. Evidence for advantageous selection has also been found in the insurance market in Australia. Australia, like many European countries, has both a universal public health system funded through taxation and a privately-funded health system (Eldridge, Koc, Onur, & Velamuri, 2011, p. 3). The private hospitals specialize in elective procedures, and are primarily fed by private insurance companies. The public hospitals tend to deal with a large majority of all emergency services(Eldridge et al., 2011, p. 5). As a result, large, unpredictable expenses are usually covered through the public health system. So the people who purchase private health insurance are not insuring against those kinds of risks. However, the private health market allows for certain procedures, which would be difficult or impossible to obtain the public insurance market, and so private health insurance still reduces the level of risk seen by the insuree. As in all insurance markets, people with a greater level of risk avoidance relative to the cost will participate in the private insurance market; while those will a lower level of risk avoidance relative to the cost will not. Several different studies have found evidence of advantageous selection including those conducted by Ellis and Savage using 2001 data from the National Health Survey (NHS) and conducted by Buchmueller, Fiebig, Jones & Savage using 2005 NHS data (Eldridge et al., 2011, p. 8). Eldridge et al. (2011)used 2005 NHS data in their study and said that “Those who report being in good health are more likely to be insured relative to those in poor health – by 17% and 18% in the two specifications respectively.” Again, this is a correlational study, not a causal one, and only demonstrates that it is empirically possible for people to be healthier if they have insurance. As mentioned above, private health insurance companies specialize in elective procedures. Unless that specialization results in people participating in the private insurance market receiving significantly more useful preventative care, the people choosing the health insurance would have had to be healthier when they chose the health insurance. In other words, there would have to be some level of advantageous selection. However, not all empirical studies of health insurance have found evidence of advantageous selection. At best, the results have been ambiguous. On the other end of the spectrum, when Harvard started to subsidize its health care plans with a fixed payment, as opposed to a proportional payment of some kind, adverse selection set into place a death-spiral that resulted in the eventual closing of the most generous plan (Cutler & Zeckhauser, 1998). In this market, there was strong evidence for a positive correlation between risk and purchasing insurance. However, most of the studies have found only a minimal amount of adverse selection, less than would be expected based on an educated guess using a model like the one developed by Rothschild and Stiglitz’s (1976) seminal article. For example, Cardon and Hendel (2001) used data from the 1987 National Medical expenditure survey, which provided relatively accurate consumption data and information on all policies offered to each employee. They found that the link between health insurance choice and health care consumption to be mostly explained by observables. Neither Finkelstein, McGarry and Sufi (2005) in their analysis of the long-term care market nor Chiappori and Salanie (2000) in their study on the French automobile insurance market found strong evidence for a statistically significant positive correlation between purchasing insurance and having a higher level of risk. Essentially, as Cohen and Siegelman (2010, p. 77) argue, “The studies yield different findings about the existence of the coverage-risk correlation and…there is good basis for expecting the existence of…a coverage-risk correlation to vary across markets, and, indeed, even across segments of the same market.” This heterogeneity in the market is evidence of the phenomenon of advantageous selection. For adverse selection only causes a positive correlation, and the moral hazard phenomenon also can only cause a positive correlation between consumption of healthcare and buying insurance (Cohen & Siegelman, 2010, p. 39). Since neither can explain a negative correlation, one would expect a relatively large positive correlation. However, this is not the case. Instead, the correlation is rather varied, which points to the correlation actually being a net effect with the variance being caused by the heterogeneity of the two effects. This other effect is advantageous selection. There are likely other determinants involved as well, but that does not negate the existence of advantageous selection. One would expect the advantageous selection to vary across markets, as mentioned above. The control people have over their level of sickness depends on the type of sickness being faced by the person. For example, there is a large difference between the control someone has over genetic factors and nutrition factors. As a result, one would expect a greater level of adverse selection in areas where the person has less control over their level of sickness, and less of a level in areas where they do not. Essentially, the insured population is not a random sample drawn from the population, and it does not differ just because of adverse selection caused by asymmetric information. Rather, it is more risk avoidant than the general population at a statistically significant level. As a result, it takes actions to Proceedings of the 2012 Pennsylvania Economic Association Conference 199 avoid the risk they face, thereby, reducing the risk faced by the insurance company. This conclusion can have significant policy implications. For if adverse selection is not a particularly significant problem, then trying to solve it using governmental mandates of some kind may not be necessary, in fact, in some cases, they actually may make the actuarially fair premium more expensive. In Donder and Hindrick’s model (2006, p. 9), under certain conditions establishing a social insurance plan in which participation if required was actually a Pareto inferior action, in other words, it makes everyone worse-off utility-wise. However, this field of advantageous selection is a relatively new field. Not much research has been done, and most of the research that has been done has been theoretical. As a result, more empirical studies attempting to directly demonstrate advantageous selection would be useful. A study such as the Wang et al. study would be particularly useful because it does a better job at establishing causality than strictly correlational studies such those conducted by Eldridge et al. and Fang et al. This would better establish that it is actual advantageous selection, and not something such as unobservables simply being unimportant, that causes the level of the correlation to be small. In the years since Akerlof’s (1970) article, the research in information, especially as it pertains to health economics has come a long way, but it has a long way to go. However, information asymmetry is not the only thing going on in the health insurance market. There is also a psychological phenomenon where people who prefer a lower level of financial risk also prefer a lower level of other kinds of risk, such as risk of pain and risk of lost-time. As a result, people in the insurance market would be lower risks, if this was the only phenomenon occurring. However, adverse selection and moral hazard also occur. As a result, the a priori net effect is ambiguous. The empirical evidence regarding the correlation between risk and insurance level has also been ambiguous. Certain studies have found evidence of a positive correlation, some of essentially no correlation, and some of a negative correlation. The existence of a positive or ambiguous correlation has broad policy implications because it raises strong concerns about the wisdom of engaging in Herculean efforts to try to bring everyone into the insured pool. Premiums will likely not have to be much higher if people have choice regarding their participation in the pool. Furthermore, it explains why insurance companies have been able to engage in community rating and other policies, which would be expected to cause significant adverse selection effects. FIGURE 1: DESIRED INSURANCE Proceedings of the 2012 Pennsylvania Economic Association Conference 200 REFERENCES Akerlof, G. A. 1970. The Market for "Lemons": Quality Uncertainty and the Market Mechanism. The Quarterly Journal of Economics, 84, 488-500. Cardon, J. H., & Hendel, I. 2001. Asymmetric Information in Health Insurance: Evidence from the National Medical Expenditure Survey. The RAND Journal of Economics, 32, 408-427. Chiappori, P. A., & Salanie, B. 2000. Testing for Asymmetric Information in Insurance Markets. Journal of Political Economy, 108, 56-78. Cohen, A., & Siegelman, P. 2010. Testing for Adverse Selection in Insurance Markets. Journal of Risk and Insurance, 77, 39-84. Cutler, D. M., & Zeckhauser, R. J. 1998. Adverse Selection in Health Insurance. Forum for Health Economics & Policy, 1(2), 1-31. de Meza, D., & Webb, D. C. 2001. Advantageous Selection in Insurance Markets. The RAND Journal of Economics, 32, 249-262. Hemenway, D. 1990. Propitious Selection. The Quarterly Journal of Economics, 105(4), 1063-1069. Marcus, R. 2012, March 20). 116 Billion Reasons to Be for the Individual Mandate Retrieved April 6, 2012, from http://www.washingtonpost.com/opinions/116billion-reasons-to-be-for-the-individualmandate/2012/03/20/gIQAFt1LQS_story.html Rothschild, M., & Stiglitz, J. 1976. Equilibrium in Competitive Insurance Markets: An Essay on the Economics of Imperfect Information. The Quarterly Journal of Economics, 90, 629-649. Shaw, K. L. 1996. An Empirical Analysis of Risk Aversion and Income Growth. Journal of Labor Economics, 14, 626-653. Wang, K. C., Huang, R. J., & Tzeng, L. Y. 2009. Empirical Evidence for Advantageous Selection in the Comercial Fire Insurance Market. The Geneva Risk and Insurance Review, 34, 1-19. Donder, P. D., & Hindriks, J. 2006. Does Propitious Selection Explain why Riskier People Buy Less Insurance. Louvain-la-Neuve, Belgium: Department of Economic Sciences of the Catholic University of Louvain: Institute of Economic and Social Research. Ehrlich, I., & Becker, G. S. 1972. Market Insurance, SelfInsurance, and Self-Protection. Journal of Political Economy, 80, 623-648. Eldridge, D., Koc, C., Onur, I., & Velamuri, M. (2011). The Impact of Private Hospital Insurance on Utilization of Hospital Care in Australia: Evidence from the National Health Survey Working Paper Series: Victoria University of Wellington, School of Economics and Finance. Fang, H., Keane, M. P., & Silverman, D. 2006. Sources of Advantageous Selection: Evidence from the Medigap Insurance Market Working Paper Series: National Bureau of Economic Research. Finkelstein, A., McGarry, K., & Sufi, A. 2005. Dynamic Inefficiencies in Insurance Markets: Evidence from Long-Term Care Insurance. The American Economic Review, 95, 224-228. Folland, S., Goodman, A. C., & Stano, M. 2001. The Economics of Health and Health Care (6th ed.). Boston: Prentice Hall. Proceedings of the 2012 Pennsylvania Economic Association Conference 201 THE EFFECT OF RELIGION AND EMPOWERMENT OF WOMEN ON FERTILITY Denae A. Heath Cameron D. McConnell Clarion University of PA 840 Wood Street Clarion, PA 16214 ABSTRACT This study uses regression analysis to determine the associations between women empowerment, religions, and the fertility rates in developed and developing countries. The authors test the hypothesis that factors of women’s empowerment and the health of the woman affect fertility rates in a country, more so in the underdeveloped countries. The countries looked at in this study include a variety of religions and are: Afghanistan, Argentina, Bolivia, Chad, India, Mongolia, Switzerland, United States, and Yemen. This study provides empirical evidence that is representative of women's empowerment and indicators that are used for determining the development of a country for the period 1960-2010. INTRODUCTION Throughout the world, there are many differences between each country in regard to the levels of fertility, the empowerment of women, and the religious beliefs (including the extent to which the religion is practiced). Previous research has shown a correlation of religion on fertility and how fertility rates are impacted by the degree to which women are empowered, but all of the variables that affect fertility have been presented separately. This research was conducted with the idea that, as underdevelopment is linked to a lack in women’s empowerment, it is also linked to high fertility rates. Religion fits right in with this idea as well; based on the fundamentals of the religion, some discourage women from obtaining a higher status and others have very strict rules about products that may reduce pregnancies, such as contraceptives. Other factors to be considered are the availability of advanced healthcare systems and access/ affordability to surgical procedures that prevent further reproduction. The religion’s stance on abortion is yet another aspect that affects a country’s overall fertility rate. There have been many papers written about all of these factors, but none tying them all together and few encompassing the array of countries that are focused on in this paper. The hypothesis that is tested here is that factors of women’s empowerment and the health of the woman affect fertility rates in a country, more so in the underdeveloped countries. Also, the religion that the majority practices in a given country has an effect on fertility rates and an indirect effect on the development stage of the country. LITERATURE REVIEW Religion’s Effect on a Country One researcher that agrees about religion’s effect on economic growth is McQuillan (2004), who claims that an important variable to be considered concerning human fertility is one’s religion. Some economies are more affected by religion than others, for instance in countries where Catholicism and Middle Eastern religions are prominent. He continues, stating that religion affects fertility rates the greatest when the church has a strong influence on its members. Many religions have already revised their rules regarding abortion and contraceptive use, causing some of the outlooks of the church not to have as great of an impact on its members. With the societies of today’s world, many members of churches aren’t as avid about the rules of their religion compared to yesteryears. In addition, Wonsub (2011) also found that through observed tests, it is shown that explanatory power of both religious division and separation is substantial enough to argue that they strongly influence the economic growth of a society. On the other hand, through McCleary’s (2003) results for given religious beliefs, the overall effect from greater church attendance is to reduce economic growth. These effects include: resources used up by the religious sector, the social-capital aspect, and the influence of organized religion on laws and regulations. The Economics of Population and Development Population is an important key in determining how economies grow in less developed countries. Gobin and Simon (1992) state that the higher the population density, the faster economic growth will be in those types of countries. Furthermore, they claim that there is no correlation between the population growth rate and economic growth, as well as total population size and economic growth. Simon (1997), in a separate study years later, disagrees, claiming that, a high population growth rate in less developed countries with high fertility decreases per capita income, decreasing economic growth in the short run. This eventually negatively affects output per worker, thus decreasing economic growth overall. Simon also finds that there is a correlation between total population size and economic growth, specifically that if population decreases, the economy will decline. They both concur that a lower population growth rate helps economic Proceedings of the 2012 Pennsylvania Economic Association Conference 202 performance in the long run (Simon, 1997) and has a positive effect on per capita income (Gobin, 1992). Fertility’s Relation to a Country’s Status When it comes to fertility and its impact on economic growth, there are differences that can be seen when looking at developed countries versus less developed countries. Aarssen (1995) believes wealthier countries ought to have lower fertility rates countries than countries with less wealth. Moreover, Aarssen (1995) argues that there is more of a focus on the time and money needed to raise a child along with the presence of sterility due to Western lifestyles. In addition to Aarssen’s (1995) look at changes in society and how they’ve affected fertility, Moe (1998) studies about which societies give more importance to family planning. Moe (1998) concludes that as economies begin to progress and become more developed, the expenditures necessary for raising a child are deemed more important than the utility of a child until the parents are economically stable. Women’s Empowerment and Fertility Singh (1994) agrees with Aarssen’s (1995) outlook on high fertility rates and brings forth that the lack of empowerment of women is a consequence of a country being less developed. He also incorporates women’s education, labor force participation, and family planning services available to societies with a decrease in fertility rates, mostly attributable to economies that are more developed. Singh’s (1994) thoughts directly relate to the purpose of the research conducted. DATA AND METHODS The independent variables used in this research include female population, the female labor participation rate, female health, female education, the income of the country (per capita), and the human development index (HDI) of the country. Presented in Table I is the data collected, all of which were extracted from the World Bank, except for the maternal mortality rate and the HDI- which were extracted from the International Human Development Indicators, Online. For the maternal mortality ratio, the natural log was used in the panel regression and proration was necessary throughout some of the variables. The dependent variable for this research is the fertility rate, total births per woman (FERT) for the years ranging from 1960- 2010; which was extracted from the World Bank as well. The countries that were used in this research were chosen based on their level of gender inequality, development level, and the majority religion that was practiced. Though the first two considerations are expressed through our data, the third factor, religion, is not due to a lack of accessible data. In Table II, you will find a chart expressing each of our chosen countries and the religion that the majority practices within them. By using this pie-chart, the intention is to show a well rounded set of data that will hopefully prove our hypothesis and incorporate religions of the world. In Table III, the descriptive statistics for the data are shown. It is clear that the data used was of a large variety and the means are quite reflective of the range of data encompassed. Presented in Table IV, in table form, is a correlation matrix. As one can see, there is a strong negative correlation between fertility and female life expectancy. The same can be said for fertility with HDI and there is a strong positive correlation between maternal mortality and fertility. In terms of female life expectancy, there is a strong negative correlation between it and maternal mortality. There is a strong positive correlation between female life expectancy and GDP per capita PPP as well as with female life expectancy and HDI. EMPIRICAL RESULTS Displayed in Table V is the linear relationship between fertility rates and variables that portray women empowerment and economic development/ growth. As predicted, female population is positively associated to fertility throughout all models. Similar to previous research, female life expectancy (Models 1-5), female secondary education (Models 2-5), and female labor participation (Models 3-6) are all negatively associated with fertility. This was expected and assists to prove the hypothesis here. Maternal mortality was positively associated with fertility in two out of its three included models. GDP per capita PPP was, as anticipated, positively associated to the fertility rates and HDI was negatively associated with fertility. Both of these variables were only used in separate single models. Given the information previously presented, most of the results from the data is economically significant and proves that there is a solid relationship between our variables and fertility rates. When looking at female population, female life expectancy, female labor participation, GDP per capita PPP, and HDI all compared to fertility, they are all very statistically significant, with p-values of 0.0000. As for female secondary education, Models 3 and 4 are 0.0000, making them statistically significant, whereas Model 2 is at 0.0626 and Model 4 has a p-value of 0.0051; all of these values have a great statistical significance, but aren’t as significant compared to the other variables. Maternal mortality in Models 5 and 6 have statistically significant p-values (0.0000), but Model 4 has a p-value of 0.9641, making it statistically insignificant. Overall, all of the models can be considered statistically significant, but Model 4 does raise questions based on the pvalue of maternal mortality. Proceedings of the 2012 Pennsylvania Economic Association Conference 203 CONCLUSION At a confidence level of 95% it can be concluded that all of the variables from all models, with an exception of two, have a significant effect on fertility. This is due to their calculated values of t being greater than their respected t-value compared to the number of observations. One of the exceptions found in this research is female secondary education in Model 2. The t-value is relatively close to being seen as significant but persists to show no relationship. The second exception, maternal mortality, doesn’t have a t-value in the proximity of the necessary value to be considered significant, proving it to have no relationship with fertility, but only in Model 4. This finding was odd, seeing that the relationship and statistical significance can be proven for this variable in the other models. The results show proof to support our thesis that the negatively associated variables, in relation to fertility, prove that fertility decreases while female population, female life expectancy, and female labor participation all increase. It can be said when looking at these variables that as women are empowered, fertility rates will decline. This leads to a negative relationship between fertility and GDP as well as fertility and HDI, allowing for a conclusion to be drawn that as women are empowered, economic development and growth occur. This can be said for all researched countries, so it’s prevalent for all religions. Some refinements that should be included are that the data availability for the selected countries was limited. In terms of religion, not many existed, making it difficult to incorporate that into the panel regression. Access to family planning sources and use of contraceptives were also factors that should have been considered, but as stated before, the data was unrecorded in different countries/ years. In the case of India, one thing to be separately considered is the preference for a son by most families. This causes an increase in their fertility rate (with families trying longer to obtain this goal) and female infanticide in India. Selective abortions also occur, as do child trafficking/ kidnapping. An inequality index was also difficult to obtain for this research purpose. Statistically, autocorrelation was present due to the DurbinWatson test. Ceteris paribus, the results are still considered significant due to the correlations and the levels of p-values (as with the t-value tests). Though the variables were considered significant in this research, many variables still exist that would further prove the effects on fertility and assist in implementing better policies to support growth in each country. POLICY RECOMMENDATION In terms of policies, the lower developed countries are encouraged to take action in promoting women and their empowerment. Things of this nature include the access to contraceptives and necessary procedures, the ability to enter the labor force, a reduction in inequality, and an equal distribution of income. More advanced healthcare and education would provide for development to progress and for an overall better economy (not just in terms of financial factors, but with factors that determine development as well). Something that policy cannot affect is one’s religion and the religious actions taken due to beliefs. Depending on the faith one has in his/ her religion, it will affect the degree to which a policy will affect a country. With high faiths, many policies may not work, but as one digresses from his/ her belief in the religion, policies will have a much greater effect. Proceedings of the 2012 Pennsylvania Economic Association Conference 204 TABLES Table I: Variable Female Population Female Participation Rate Female Education Female Health Pregnant Women’s Health Income Development Measure Population, female (% of total) Labor participation rate, female Percentage of female students, total secondary Life expectancy at birth, female Maternal mortality ratio GDP per capita PPP Human development index (HDI) Abbreviation Source Years POPF World Bank 1960-2010 LABPARF World Bank 1960-2010 FEMSTUSEC World Bank 1960-2010 LEFEM World Bank 1960-2010 MM International HDI Indicators World Bank International HDI Indicators 1960-2010 GDPPCPPP HDI 1960-2010 1960-2010 Table II: Mongolia Afghanistan, Chad, Yemen Argentina, Bolivia, Switzerland, USA India Proceedings of the 2012 Pennsylvania Economic Association Conference Buddhism Christianity Hinduism Islamic 205 Table III: Std. Error Mean Median Std. Dev. Skew 0.048 0.965 0.583 0.779 0.071 0.367 FERT 4.814 0.109 4.937 2.319 POPF 50.093 0.054 50.589 1.149 LABPARF 47.900 0.983 55.800 16.158 FEMSTUSEC 39.559 0.749 46.308 13.210 LEFEM 60.634 0.662 60.045 14.041 MM 5.072 0.135 5.500 1.761 11094.979 880.966 3140.408 13789.292 0.589 0.014 0.577 0.217 GDPPCPPP HDI 1.127 0.086 Min. Max. Count 1.380 9.223 450 47.854 51.561 459 15.900 67.800 270 7.598 57.345 311 30.813 84.400 450 1.946 7.496 171 568.552 43710.279 245 0.198 0.908 239 Table IV: FERT POPF LABPARF FEMSTUSEC LEFEM MM GDPPCPPP FERT 1 POPF -0.3782 1 LABPARF -0.4736 0.5811 1 FEMSTUSEC -0.7581 0.4916 0.2914 1 LEFEM -0.8957 0.5687 0.3268 0.7810 1 0.8665 -0.6856 -0.3762 -0.7430 -0.9574 1 GDPPCPPP -0.7007 0.5342 0.2535 0.4038 0.8182 -0.8715 1 HDI -0.9017 0.7967 0.5504 0.7839 0.9607 -0.9407 0.8728 MM HDI 1 Table V: Variables Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Fertility (Y) -4.66515 (-2.04618) 0.0413 0.390323 (8.123735) 0.0000 -0.16615 (-42.1478) 0.0000 -1.39998 (-0.70786) 0.4796 0.326968 (7.694058) 0.0000 -0.1628 (-30.7498) 0.0000 -0.00865 (-1.86909) 0.0626 -24.0761 (-9.95835) 0.0000 0.858058 (15.85043) 0.0000 -0.18354 (-37.8345) 0.0000 -0.01765 (-4.39036) 0.0000 -0.04358 (-12.9503) -26.2727 (-9.50902) 0.0000 0.903599 (14.98068) 0.0000 -0.16907 (-15.1059) 0.0000 -0.0358 (-7.03295) 0.0000 -0.0481 (-13.3438) -37.4928 (-16.959) 0.0000 1.019908 (22.19574) 0.0000 -0.1441 (-14.7503) 0.0000 -0.01491 (-2.84615) 0.0051 -0.05121 (-17.0361) -28.7122 (-5.25879) 0.0000 0.683518 (6.392881) 0.0000 Female Population Female Life Expectancy Female Secondary Education Female Labor Participation Proceedings of the 2012 Pennsylvania Economic Association Conference -0.03708 (-6.30985) 206 0.0000 Maternal Mortality 0.0000 -0.00288 (-0.04507) 0.9641 GDP per capita PPP 0.0000 0.512605 (8.188415) 0.0000 0.000037 (7.856456) 0.0000 HDI 0.0000 0.636408 (5.449248) 0.0000 -5.03059 (-5.38343) 0.0000 9 Countries 9 9 9 9 9 Included Observations 50 40 30 19 19 19 450 307 250 162 150 156 0.962469 0.036084 0.974993 0.046088 0.847937 0.030161 Total Pool (Balanced) Observations 0.826943 0.901789 0.949987 Adjusted R2 0.010635 0.011832 0.027212 D-W Stat Listed first is the coefficient, then the (t-statistic), then the p-value REFERENCES Aarssen, Lonnie W. "Why Is Fertility Lower In Wealthier Countries? The Role Of Relaxed Fertility-Selection." Population And Development Review 31.1 (2005): 113-126. EconLit. Web. 22 Feb. 2012. International Human Development Indicators Online, 2011. Maps of World. World Religion Map. 20 February 2012 <http://www.mapsofworld.com/world-religion-map.htm>. McCleary, Robert J. Barro and Rachel M. "Religion and Economic Growth." Harvard University (2003): 53. 22 Feb. 2012. McQuillan, Kevin. "When Does Religion Influence Fertility?." Population And Development Review 30.1 (2004): 2556.EconLit. Web. 22 Feb. 2012. Moe, Karine S. "Fertility, Time Use, And Economic Development." Review Of Economic Dynamics 1.3 (1998): 699718.EconLit. Web. 22 Feb. 2012. Simon, Julian L. "Population Growth May Be Good For Ldcs In The Long Run: A Richer Simulation Model." The economics of population: Key modern writings. Volume 1. 176-204. Elgar Reference Collection. International Library of Critical Writings in Economics, vol. 78., 1997. EconLit. Web. 22 Feb. 2012. Simon, Julian L., and Roy Gobin. "The Relationship Between Population And Economic Growth In Ldcs." Population and development in poor countries: Selected essays. 180-198. Princeton:, 1992. EconLit. Web. 22 Feb. 2012. Singh, Ram D. "Fertility-Mortality Variations Across Ldcs: Women's Education, Labor Force Participation, And ContraceptiveUse." Kyklos 47.2 (1994): 209-229. EconLit. Web. 22 Feb. 2012. Wonsub, Eum. "Religion and Economic Development- A Study on Religious Variables Influencing GDP Growth over Countries." University of California, Berkeley (2011): 30. 22 Feb. 2012. World Bank, World Development Indicators Online, 2010. Proceedings of the 2012 Pennsylvania Economic Association Conference 207 THE EFFECT OF HUMAN DEVELOPMENT LEVEL ON THE RELATIONSHIP BETWEEN ENERGY CONSUMPTION AND ECONOMIC GROWTH: A CHAOS THEORY APPROACH Dr. Ezatollah Abbasian Department of Economics Bu-Ali Sina University Hamedan, Iran Email: abbasian@basu.ac.ir Phone No: (+98)9125025902, Fax: (+98)8118272084 Maysam Nasrindoost Department of Economics Bu-Ali Sina University Hamedan, Iran Email: m.nasrindoost@gmail.com ABSTRACT This paper studies the effect of human development level on the complexity of the relationship between energy consumption and economic growth. Energy intensity is a key variable that could be used as an indicator of the change on socio-economic structures, monetary structures or environment-economy relations. It could be used as an indicator of the appearance of new structures too. Standard economic analysis takes a continuous, deterministic and predictive approach, which encourages the search for predictive policy to correct environmental problems. This is actually what happens with the relationship between economic growth and energy consumption. The point is that nonlinear interdependent relationships can be described via chaos methods, which help us understand patterns and structures of behavior. Using the energy intensity and human development indices in 139 world countries, the paper's results show that low human development is correlated with most likely chaotic energy intensity time series. Economic instability and inappropriate government policies are recognized as the main causes of results. Furthermore, some implications of chaotic energy intensity series are pointed out. INTRODUCTION Recently, the issue of the dematerialization of developed economies (the reduction of material as well as energy intensities over time) has gained popularity in the field of ecological economics. The so-called hypothesis of “intensity of use” was first put forward by Malenbaum (1978) and states that income is the main factor that explains the consumption of materials. That is, during the process of economic development, countries would tend to increase consumption of energy and materials at the same rate as growth in income, until one defined level of income is reached. Beyond that level, however, we have to expect a delinking between economic growth and the consumption of materials. The study of relationship between energy consumption and economic growth dates back to the late 1970s, especially after two world energy crises, from literature's perspective. This group of studies emerged by the work of Kraft and Kraft (1978) who investigated the relationship between GNP and energy consumption. Currently, there are four hypotheses about energy consumption and economic growth relationship (Payne; 2008). The first one states that energy consumption affects the economic growth as a complementary factor for the labor and capital inputs. Therefore, “growth hypothesis” states a unidirectional causal relation from energy consumption to economic growth. According to the second hyp othesis, there is a unidirectional causal relation from economic growth to energy consumption, “thrift hypothesis”. It may be true because economic growth is affected by energy consumption for some reasons (Ramos-Martin and Ortega-Cerda; 2003) such as: a structural change in the economy (for example, shifting from high energy intensity sectors to lower intensity ones); an improvement in energy efficiency (for example, technological improvements) and, changes in consumption patterns (for example, a change in consumer’s culture). According to the third hypothesis which is “feedback hypothesis”, there is an interaction between energy consumption and economic growth. Finally, in the “neutral hypothesis”, there is no relation between energy consumption and economic growth. One of Proceedings of the 2012 Pennsylvania Economic Association Conference 208 the most popular reasons for this neutrality, is because of little energy consumption share of GDP 36. In here, feedback hypothesis is the most disputable among others and more relevant to our objectives: because feedback is one of the fundamental properties of chaotic series. While energy consumption and economic growth relationship is unidirectional (or there is not any relationship), the economic policies will take place only in one step, but existence of feedback causes more complexity in economic policies decision making. “Rebound effect” is an example of such complexity 37. Because of requirements to consider the feedback effect in the relationship between energy consumption and economic growth, the question which demands to answer in this paper is: how the existence of such a relationship is related to the performance of nations in social and economic aspects? To answer this question we classify the challenges in this respect into three categories: 1. 2. 3. As it is recently argued, policy makers are confused about the use of studies about energy consumption and economic growth (see Chontanawat et Al.; 2008). This may occur due to various results of the studies, even in one country (see Lee and Chang; 2007). Fatai et Al. (2004) showed that energy consumption has a significant positive effect on economic growth in India and Indonesia, but Masih and Masih (1996), reached a completely different results for the same countries. One of the most probable reasons for this might be methodological differences in definitions, specification of variables and econometric models. More of done studies in this area didn't consider nonlinearity and dynamic properties of the energy consumption and economic growth relationship. Lee and Chang (2007), express that nonlinearity assumption has shed more light on the study of such relationships. Ramos-Martinand and Ortega-Serda (2003), believe that nonlinear processes include both external shocks and internal causes within the system. Adrangi et Al (2001), found some evidences of nonlinear behavior in the energy price (which probably transmits to energy consumption). Seifritz and Hodgkin (1991) showed strong evidences of nonlinear properties of energy consumption, by analysis of overall world energy information collected over last 150 years. We can see similar ideas in the other papers such as: Homilton (1996), Mork et Al. (1994) and Balk et Al. (1999). Unruh and Moomaw (1998), states that all of historical events and shocks in the energy price are visible in dynamic studies. Some economists believe that the relationship between energy consumption and economic growth is weak, so that without more efforts, we can agree with neutral hypothesis. One of the most important reasons for this is assumed as fossil fuels become scarcer, their price will rise, which in turn will trigger technological changes and substitutions that improve energy efficiency. Then, the role of energy in economic activities will decline (see Dias et Al.; 2006). Therefore, in order to answer the paper's question, we need to use an appropriate method which take cares about three above mentioned concerns in it. Such method must be able to investigate the existence of feedback in the energy consumption and economic growth relationship, as: 1. Study the relations without any need to determine econometric specification. 2. Consider nonlinear and dynamic aspects of energy consumption and economic growth relationship. 3. Focus on the quality of energy consumption as well as quantity of it in correlation with economic growth 38. Thus, in this paper we do three mentioned tasks as follows: • To tackle the challenges 1 and 2: The simultaneous use of chaos theory and energy intensity index satisfies such properties. Energy intensity is one of the most important energy efficiency indices. It yields quantity of energy used (usually in British Thermal Units, BTU) for 1$ gross domestic production. So, the quality of energy consumption and economic growth relation are determinates of energy intensity index. On the other hand, chaos theory is based on theory of nonlinear growth with feedback. Adrangi et Al (2001) states that chaotic dynamic is necessarily nonlinear and may be able to explain a richer array of time series behavior. Also, Ramos-Martinand and Ortega-Serda (2003) believe that evaluation of energy intensity under the complex (or chaotic) system framework will be helpful to a better understanding of the relationships. Therefore, when the behavior of energy intensity series is chaotic, energy consumption and economic growth have an interaction with each other 39. • To tackle the challenge 3: For investigating the relationship between energy consumption and economic growth, we take Human Development Index (HDI) into account. On the one hand, HDI was introduced by economists such as Mahbub ul Haq Haq and Amartya Sen in 1990s to aggregate two new dimensions to the comparison of the performance of nations. In the other hand, energy use and social, political and economic development of nations represent a phenomenon with a relationship of dependence that is the influence of each one on the other (Dias et Al.,2006). There are some studies which consider the role of HDI in quality energyeconomic growth relationship. For example, Martinez and Ebenhack (2008) found a strong relationship between HDI Proceedings of the 2012 Pennsylvania Economic Association Conference 209 values and energy consumption in the majority of the world countries. Also, Dias et Al.(2006) found that when energy use was associated with HDI, it was possible to find opportunities to put into practice the energy conservation concepts, by looking for new conditions to achieve development at a less intensive energy use. More important of others in this respect, Chontanawat et Al. (2008) studied “growth hypothesis” in a sample countries which includes more than 100 countries by considering their human development index (HDI). They found that growth hypothesis is valid for 69 percent of high, 42 percent of medium and 35 percent of low HDI countries. So, we can conclude that simultaneous attention to non-liner and dynamic aspects of the relationship between energy consumption and economic growth as well as considering to the social effects of human behavior on such relation is a new incentive to study in this paper. In the next section of this paper, the Kolmogorov Entropy test as preferable test for investigation of chaotic behaviors is introduced. The empirical results are presented in the third section and finally discussions and conclusions will be made in section four. KOLMOGOROV ENTROPY TEST There are numerous tests for investigating the chaotic behavior. All of them are based on distinguishing between one or more properties of chaotic behavior. The most popular tests for chaos are: BDS, Lyapunov Exponent, Correlation Dimension and Kolmogorov Entropy tests. BDS tests the existence of nonlinear structure in residuals of a nonlinear model (such as GARCH). In other words, it tests chaos indirectly. For this reason, BDS test is not sufficient to recognize the existence of chaotic behavior. Lyapunov Exponent and Kolmogorov Entropy tests are based on “sensitive dependence on initial conditions” property (or “butterfly effect”). While Lyapunov Exponent calculates the average degree of divergence of nearby orbits in phase space, Kolmogorov Entropy measures whether nearby trajectories separate as required by chaotic structure. Finally, Correlation Dimension is based on “strange attractor” property of chaotic series. In this paper, we consider Kolmogorov Entropy test for some reasons: firstly, as Adrangi et Al (2001) argues, among the chaos tests, Kolmogorov Entropy test probably represents the most direct test for chaos. Secondly, this test has a high similarity with Lyapunov Exponent test. So, Kolmogorov Entropy test is preferred to all other tests for our purpose. Consider X. and X. + βX. as two points in the state space. Both of them make an orbit in the state space by use of system equations. These orbits are functions of time because system variables are so. If one of them is assumed to be the reference orbit, the difference between these two orbits (βX.) is a function of time, too. Also, due to “sensitive dependence on initial conditions”, βX. is a function of initial values. So, we have: βX = βX (X., t). This is illustrated in Fig (1). As we can see, difference between orbit 1 and orbit 2 is not equal in any pair of time points (for example t0 and t1), necessarily. Also, at the same time (for instance t2), the difference between orbits 1 and 2 is not the same as it between orbits 3 and 4, necessarily. It is true because initial values are different. Then, time and initial value are two factors that affect the difference between orbits. βX = βX (X., t) decreases asymptotically along with time in a system containing attracting fixed point. It is true because orbits converge to a special point in such systems. But, βX = βX (X., t) will be erratic in a chaotic system. In other words, divergence in the initial conditions makes orbits different such that their differences after “t” time steps are our criteria to distinguish chaotic series from others. Grassberger and Procaccia (1983) calculate the average divergence rate by use of correlation integral concept as: K2 = Limε→∞ Limm→∞ Lim N→∞ Ln ( Cm,N(ε) / Cm+1,N(ε) ) (1) Cm,N(ε) is correlation integral in “m” dimension, with N observation and ε usually is 0.5, 1, 1.5 or 2 standard errors of data. Due to “sensitive dependence on initial conditions”, orbits in a chaotic series change rapidly and then K2 must be positive. Our methodology in this paper is calculating the K2 for energy intensity data. We calculate it with different assumptions about parameter m, (m=2,...,14) and in 139 world countries. In the other word, we run the Kolmogorov test 14*139=1946 times and classify these results by HDI ranking of countries. Details will be illustrated in next section. DATA AND EMPIRICAL RESULTS The level of human development is calculated annually in 177 countries in three groups (high, medium and low human development countries) by World Bank (WB). HDI is based on three other indices: 1) A long and healthy life 40, 2) Access to knowledge 41 and 3) A decent standard of living 42. These three dimensions are standardized to values between 0 and 1, so that the geometric mean of these values is equivalent to HDI value in the range 0 to 1. According to World Bank report in year 2008, about 39 percent of world countries have high, 48 percent medium and 13 percent low HDI. On the other hand, energy intensity time series are available from Energy Information Administration (EIA) for the 1980- Proceedings of the 2012 Pennsylvania Economic Association Conference 210 2005 periods. EIA publishes official world's overall energy estimates and sometimes analyses and forecasts them. The unit of energy intensity is BTU per 1$ GDP. Energy intensity data does not exist for all of 177 HDI ranking countries, so we test the chaotic behavior in energy intensity time series in 139 world countries. Among our sample countries, there are 38 percent of the sample countries with high HDI, 47 percent with medium HDI and 15 percent with low HDI. It is concluded that there is no serious error associated with the selection of countries, when we compare the composition of countries in our sample with one in the mentioned WB ranking. By calculating the Kolmogorov Entropy test statistic, for example, for Iranian energy intensity time series is equal to 0.139 when dimension and epsilon, (ε), are assumed 5 and 0.5*S.E. respectively. Therefore, under such assumptions, Iranian energy intensity time series is chaotic. We can repeat above process for all of 139 sample countries. The result is showed in Fig (2) by HDI ranking. As Fig (2) shows, based on the study assumption, energy intensity is chaotic in the most of low HDI countries. Also, this process can be repeat for different dimensions (m=1, 2, ..., 14), while epsilon, (ε), is the same as before. We summarized these results in table (1) and Fig (3) by HDI ranking. As Fig (3) shows, with all of the assumed dimensions, the highest percentage of the chaotic energy intensity is correlated to low HDI countries and vice versa. In other words, there is a negative correlation between human development level and chaos in energy intensity variable. Finally, there are some chaotic energy intensity implications: • Complexity in the relationship between energy consumption and economic growth is more in low HDI countries. • Feedback in relationship between energy consumption and economic growth is stronger in low HDI countries. • In the low HDI countries, slight errors in study of energy intensity leads to an increased difference in results (butterfly effect). • The relationship between energy consumption and economic growth most probably must be nonlinear in low HDI countries. • Chaos in the energy intensity time series of countries is mostly of the low-dimension type. According to these consequents, some policy recommendations are emerged for low HDI countries: 1. 2. DISCUSSION AND CONCLUSION Using the energy intensity and human development indices in 139 world countries and chaos theory approach, this study demonstrates that low human development is correlated with more chaotic energy intensity time series. Details of such a result can be debated from two aspects: causes and consequents. From cause's aspect of view, one of the most important reasons for chaos in the economic variables in these countries is the low socio-economic development level. The economic instability sheds some light on this claim. Considering energy as a strategic commodity makes its price instable. In this case, in the next step, it would causes instability in whole of the economy. Since energy intensity is deeply correlated to quality of energy consumption and economic growth relationship, chaos in energy intensity series is expected. Therefore, chaos in energy intensity series is less probable in the developed countries (which are more economically stable). Furthermore, government policies might also be inappropriate from time and functions viewpoints in the low HDI countries as well. Once more, it causes chaos in energy intensity series. Fig (4) summarizes these factors. 3. 4. 5. Policy makers should pay attention to the quality of energy use in the policies designs which are related to energy. For example, longevity and educational level of people in these countries are more important factors in policy design and should be taken into account. The relation between energy consumption and economic growth is more probably non-liner and dynamical in these countries. Then, it is better to determine policies which emerge from non-liner and dynamic models. Measurement errors in these countries led to more divergence of true policy. Then we need to more precise data to determine energy policies in these countries. Energy policies making in these countries should pay attention to instability in economic conditions. In these countries, energy policies must be strategically and multi-step. It is true because of existence of feedback in energy-economic growth. Proceedings of the 2012 Pennsylvania Economic Association Conference 211 FIGURES AND TABLES m Table (1): Chaotic Behavior Percentage in the Countries’ Energy Intensity for Different Dimensions 2 3 4 5 6 7 8 9 10 11 12 13 14 HDI Low 100 100 90.48 71.43 66.67 61.9 47.62 28.57 28.57 14.29 14.29 9.52 9.52 Medium 100 100 90.77 67.69 47.69 38.46 23.08 15.38 9.23 3.08 1.54 1.54 1.54 High 100 98.11 83.02 64.15 35.85 24.53 18.87 13.21 9.43 5.66 5.66 3.77 1.89 Source: Author's observations Fig (1): Factors that effect on the difference between the orbits Source: Author's observations Series1, Low, 71.43 Percent Series1, Medium, 67.69 Series1, High, 64.15 HDI ranking Fig (2): Chaotic Behavior Percentage in the Countries’ Energy Intensity (m=5) Source: Author's observations Proceedings of the 2012 Pennsylvania Economic Association Conference 212 Percent Low dimensions Medium High Fig (3), Chaotic Behavior Percentage in the Countries’ Energy Intensity for Different Dimensions Source: Author's observations Fig (4): Factors of chaos in energy intensity series Source: Author's observations Proceedings of the 2012 Pennsylvania Economic Association Conference 213 ENDNOTES 1 - See Payne(2008) and Chontanawat et Al.(2006) for comprehensive list of studies in this subject. 2 - As said by Schipper and Grubb (2000): “claims surface that improving energy efficiency improvements do not in fact reduce demand nearly as much as expected (a weak rebound) or that improved efficiency leads to effects that erase most of the expected savings (a strong rebound) or indeed stimulates greater energy use than if no improvements had taken place at all (a backfire) by lowering the cost of energy services and by stimulating economic activity”. 3 Some other economists argue that energy consumption and economic growth relationship is a strong one. They said it is true not due to quantity of energy used but for energy quality. For example it is said that “Energy quality is important to account for in the assessment of E/GDP ratios”. (See Stern; 1993 and Kaufmann; 1992). Also, econometric analysis confirms a strong connection between energy use and GDP when energy quality is accounted for (see Stern; 1993 and Cleveland et Al.; 2000). 4 The inverse of this phenomenon is not necessarily true. 5 As measured by life expectancy at birth. 6 Measured by two indicators: the adult literacy rate and the combined gross enrolment ratio (GER) in primary, secondary and tertiary education). 7 As measured by the GDP per capita expressed in purchasing power parity [PPP] US dollars REFERENCES Adrangi,B.;Chatrath,A.;Dhanda,K.K. and Raffiee,K., 2001. "Chaos in oil prices? Evidence from futures markets". Energy Economics (23), 405-25. Balke, NS; Brown, SPA and Yucel, M., 1999."Oil price shocks and the US economy: where dose the asymmetry originate?". Working paper. Dallas: Federal Reserve Bank of Dallas. Chontanawat,J.; Hunt,L.C. and Pierse,R., 2008."Dose energy consumption cause economic growth?: Evidence from a systematic study of over 100 countries". Journal of Policy Modeling (30), 209-20. Chontanawat,J.;Hunt,L.C. and Pierse,R., 2006."Causality between energy consumption and GDP: Evidence from 30 OECD and 78 non-OECD countries". Surrey Energy Economics Discussion Paper,SEEDS(113),university of Surrey. Cleveland, c.J.,2000.” Energy Quality, Net Energy, and the Coming Energy Transition” Sixth Annual Energy Conference “The Future of Oil as an Energy Source,” sponsored by the Emirates Center for Strategic Studies and Research, 7-8 October, 2000 in Abu Dhabi, United Arab Emirates. Dias,R.A., Mattos,C.R. and Balestieri J.A.P., 2006.”The limits of human development and the use of energy and natural resources”. Energy policy(34),1026-1031. Fatai,K., Oxley,L. and Scrimgeour FG., 2004."Modeling the causal relationship between energy consumption and GDP in New Zealand, Australia, and India. Indonesia, the Philippines and Thailand. Math Comput Simulation (64), 431-45. Grassberger,P. and Procaccia,I., 1983. "Measuring the strangeness of strange attractors", Physical (9D),30-31. Proceedings of the 2012 Pennsylvania Economic Association Conference 214 Hamilton, JD., 1996. "This is what happened to the oil price - macroeconomic relationship" .J Monetary Econ(38),215-20. Kraft, J., Kraft, A., 1978, “On the relationship between energy and GNP”. Journal of Energy and Development 3, 401– 403. Kaufmann, R. K. 1992. A biophysical analysis of the energy/real GDP ratio: implications for substitution and technical change. Ecological Economics, 6: 35-56. Lee,CC and Chang,CP., 2007."The impact of energy consumption on economic growth: Evidence from linear and nonlinear models in Taiwan". Energy (32), 2282-94. Malenbaum, W., 1978, "World Demand for Raw Materials in 1985 and 2000", McGrawHill, New York. Martinez,D.M. and Ebenhack,B.W., 2008,”Understanding the role of energy consumption in human development through the use of saturation phenomena”. Energy policy(36), 1430-1435. Masih, A.M.M., Masih, R., 1996, “Energy consumption, real income and temporal causality: results from a multi-country study based on cointegration and error-correction modeling techniques”. Energy Economics 18, 165–183. Mork,KA; Olsen,O. and Mysen,HT., 1994."Macroeconomic responses to oil price increases and deceases in seven OECD countries". Energy J(15).19-35. Payne, J.E., 2008. “On the dynamics of energy consumption and output in the US”. Applied Energy Journal (in pressing). Ramos-Martin,J. and Ortega-Serda, M., 2003."Nonlinear relationship between energy intensity and economic growth". Paper submitted to the ESEE conference Frontiers2, held in Tenerife, Spain, 12-15 February. Schipper, L. and Grubb, M., 2000. “On the rebound? Feedback between energy intensities and energy uses in IEA countries”. Energy Policy (28), 367-388. Seifritz,W. and Hodgkin,J., 1991."Nonlinear dynamics of the per capita energy consumption". Energy (163)615-20. Stern D. I.,1993,”Energy use and economic growth in the USA: A multivariate approach”, Energy Economics 15, 137-150. Unruh,G.C. and Moomaw, W. R., 1998."An alternative analysis of apparent EKC-type transitions". Ecological Economics (25), 221-29. Proceedings of the 2012 Pennsylvania Economic Association Conference 215 LIVING IN KEYNES’S LONG RUN: THE EFFECTS OF THE OVERUSE OF ECONOMIC STIMULUS David Nugent 3235 Wainbell Avenue Pittsburgh PA 15216 ABSTRACT This paper is a theoretical paper that addresses the issue of the role of short run economic stimulus in long run economic growth. Topics addressed include a general model of economic output and cycles of economic growth and recession, the strategy of employing Keynesian policies of economic stimulus to increase demand during periods of idle capacity and the determination of the appropriate stage in an economic cycle to curtail economic stimulus. Topics also include unintended consequences of economic stimulus. INTRODUCTION In “A Tract on Monetary Reform”, John Maynard Keynes (1923) stated that “In the long run, we are all dead”. In more recent years, the comment has been made that Keynes is dead, and we are living in his long run. Although Keynes might have been joking, and the statement was made in reference to monetary policy at the time rather than the fiscal stimulus that came to be associated with Keynes in later years, the 90 year old statement “In the long run, we are all dead” seems relevant at a time when years of deficit spending has resulted in levels of government debt and government spending that may be difficult to sustain. This paper begins with a general description of economic growth and economic cycles that entail expansion followed by recession. The paper then addresses government action in the form of economic stimulus that may increase aggregate demand and if successful may shorten the duration and severity of a recession. The paper will also address adverse consequences of economic stimulus and the effects of continued economic stimulus after a recession has ended. labor and overall increases in total output of goods and services. Consistent, strong growth tends to be temporary. At some point during an economic cycle, output reaches a peak, and expansion turns into recession. Aggregate demand falls, resulting in falling output, idle factory capacity, office vacancies, less employment of labor and other consequences of economic contraction. As an economic cycle continues, market adjustments take place that eventually result in economic activity reaching a trough, followed by renewed expansion. Reliance on market forces may result in lengthy periods of recession. An alternative may be government intervention in the form of economic stimulus, which entails monetary and fiscal policies. MONETARY POLICY Economics textbooks (Samuelson, 1970; Parkin, 1998) describe monetary policy in terms of actions taken by the Federal Reserve to maintain economic growth and economic stability. Actions include changes in the money supply and setting the discount rate, which is the rate charged by the Federal Reserve for borrowing by banks. If the Federal Reserve were to increase the money supply, the potential result could be economic stimulus that would result if more money were available to people, as well as the avoidance of deflation that might result from an inadequate money supply. If economic growth results in real increases in the production of goods and services, growth in the money supply that roughly matches economic growth is appropriate. ECONOMIC GROWTH AND ECONOMIC CYCLES Consider how economic output occurs in a period of growth. Economic textbooks (Samuelson, 1970; Schiller, 2006) describe factors of production as capital, labor and raw materials that combine to produce valuable goods and services. When an economy is growing, demand for goods and services is typically sufficient for producers to sell their output. If business owners perceive that demand will grow, they will invest in new capital assets to increase productive capacity. Producers of raw materials will similarly increase their output. Increased productive capacity and increased raw material production will result in increased employment of Decreases in the Federal Reserve’s discount rate can serve to stimulate economic growth. If interest rates charged by banks to investors and consumers were to also decrease, the result could be greater investment and greater consumer demand. Conversely, increases in the Federal Reserve’s discount rate can serve to counteract inflationary pressures. ADVERSE EFFECTS OF MONETARY POLICY In general, if the money supply increases faster than the growth in output of goods and services without a corresponding decrease in the velocity, or turnover, of money, the result can be inflation. In “Money Mischief”, Proceedings of the 2012 Pennsylvania Economic Association Conference 216 Milton Friedman stated that “substantial inflation is always and everywhere a monetary phenomenon” (Friedman, 1992, page 193). Examples of substantial inflation include the German hyperinflation of the early 1920s, and the Chinese hyperinflation of the 1930s and 1940s. Milton Friedman suggested that economic turmoil caused by the Chinese hyperinflation contributed to the communist victory over the Chinese Nationalist government. (Friedman, 1992) Schiller (2006) describes hyperinflation that occurred in the United States during the Revolutionary War. The Continental Congress issued Continental dollars in great quantities, giving rise to hyperinflation. “The expression “not worth a continental” became a popular reference to things of little value”. (Schiller, 2006, page 316) Although the United States has not experienced hyperinflation in recent decades, inflation has occurred, tending to be slight to moderate, including several years of double digit inflation in the late 1970s and early 1980s. In 1980, the inflation rate was 13.58%. Between April 1945 and April 2012, the Consumer Price Index rose from 17.8 to 230.1 (Historical Inflation, inflationdata.com). Regarding adverse effects of the Federal Reserve’s policy of reducing the discount rate, it might be argued that reductions in interest rates contributed to the housing bubble that resulted in economic disaster from which the economy has not yet fully recovered. To illustrate how interest rate changes might have contributed to a housing bubble, suppose that a hypothetical family has annual income of $40,000. Further suppose that the family can comfortably spend 25% of their income on interest. That family could spend $10,000 per year on interest (40,000 X .25 = 10,000). If the mortgage interest rate were 10%, the family could borrow $100,000. (10,000 / .10 = 100,000; 100,000 X .10 = 10,000). Suppose that the mortgage interest rate were to fall to 5%, the family could borrow $200,000 (10,000 / .05 = 200,000; 200,000 X .05 = 10,000). Suppose that the mortgage interest rate were to fall further, to 2.5%. The family could borrow $400,000 (10,000 / .025 = 400,000; 400,000 X .025 = 10,000). It seems plausible that large numbers of families, with access to greater and greater mortgage loans, would bid up the prices of houses so that a house that sold for $100,000 when mortgage rates were 10% might rise to $400,000 as mortgage rates fall to 2.5%. It also seems plausible that at some point, home buyers would realize that a house that just a few years earlier sold for $100,000 is not really worth $400,000, and the market values would plummet. The hypothetical family with annual income of $40,000 might find itself with $400,000 of debt and a house worth $100,000. Alternately, consider a similar hypothetical family with annual income of $40,000 that purchased a house for $100,000 when mortgage rates were 10%. Suppose that as interest rates fell from 10% to 5% and then to 2.5%, they learned that the increased market value of their home enabled them to refinance their home for higher loan balances without increasing their interest payments. Suppose that family then spent the money. During the period of rising home values, the family spent tens of thousands of dollars beyond their income. That spending could provide economic stimulus that could have led to substantial economic growth, until the housing bubble burst. Then that family would find itself in the same position as the other hypothetical family, with $40,000 of annual income, $400,000 of debt and a house worth $100,000. FISCAL STIMULUS Much of the economic stimulus employed in recent decades has been in the form of fiscal stimulus that can be traced to theories suggested by John Maynard Keynes in “The General Theory of Employment, Interest and Money”, (Keynes, 1936). Fiscal stimulus typically entails the incurrence of deficit spending by governments. If a government increases spending without increasing tax revenue, or if a government reduces tax revenue without reducing spending, the result may be the stimulus of demand. Economic stimulus through fiscal policy is theorized to arise from a multiplier effect (Parkin, 1998; Samuelson, 1970; Schiller, 2006). A multiplier effect entails increases in spending from a series of transactions arising from an initial expenditure. Consider an illustration that an economics professor might provide to a class. Image that the government hires somebody to cut grass in a park. Suppose that person receives $20 from the government. That person might then go to a farmer and use the $20 to buy food. The farmer might use the $20 to go to a barber and get a haircut. The barber might use the $20 to hire somebody to paint the door of the barbershop. The painter might then spend the $20 to buy something, and the seller will use the $20 to buy something else, and so forth. Although there are additional factors, such as taxes and the tendency for a person’s marginal propensity to consume to be less than 100%, that would limit the multiplier effect, the point is that the initial expenditure of $20 would result in total expenditures and an increase in demand for goods and services that would exceed the initial $20. Along similar lines, a tax cut can theoretically have a multiplier effect. If a tax cut were to result in greater after-tax income, the spending of the greater income could have a multiplier effect in a manner similar to the preceding. Proceedings of the 2012 Pennsylvania Economic Association Conference 217 EVIDENCE FROM WORLD WAR TWO It is often suggested that economic performance during World War Two is evidence of the validity of fiscal stimulus. Data available in reference and economics books (Encyclopedia Britannica Almanac, 2005; Schiller, 2006) seem to be consistent with theories of economic stimulus. United States federal spending increased from $13.7 billion in 1941 to $92.7 billion in 1945. The federal deficit increased from $4.9 billion in 1941, peaked at $54.6 billion in 1943 and decreased to $47.6 billion in 1945. Gross Domestic Product increased from $125 billion in 1941 to $213 billion in 1945. The increase in Gross Domestic Product during the same period of time that spending and deficits also increased seems to be evidence of the validity of fiscal stimulus. In the years since World War Two, fiscal policy has been used in attempts to stimulate the economy, with the result that deficits have been incurred in many of those years. ADVERSE EFFECTS OF FISCAL STIMULUS Potential adverse effects of fiscal stimulus include crowding out effects and effects of increasing debt burden. Economics textbooks (Parkin, 1998; Schiller, 2006) suggest that if the government finances deficits through borrowing, the government borrowing could crowd out private borrowing, resulting in less private investment and slower economic growth. Although current low levels of interest rates suggest that a significant crowding out effect has not yet occurred, conditions could change as the economy improves. If the government were to run a $1.5 trillion deficit during a period of economic expansion and rising interest rates, it seems plausible that a crowding out effect could be significant. The effects of increasing debt burden include the increasing debt service that arises as ongoing deficit spending increases Federal debt. In terms of government budgets, debt service can be defined as “the interest required to be paid each year on outstanding debt” (Schiller, 2006, page 258). If deficits were to occur for a number of years, at some point the debt service might become so substantial that it could place a great burden on the economy. In implementing fiscal policy, excessive debt burden might be avoided if deficits were limited to periods of recession. An inference that can be made regarding theories of fiscal stimulus is that deficit spending is meant to be temporary. The multiplier effect is most effective if there is excess capacity in the economy such that increased aggregate demand will reduce idle capacity and increase employment of labor. However, once labor is fully employed, excess inventory is sold, factories increase output to full capacity, and the economy is growing, stimulus can be curtailed. Further stimulus through deficit spending has the potential to crowd out private investment, unnecessarily increase debt burden, and cause other adverse effects. If fiscal policy were to follow a pattern of incurring budget deficits during recessions, and incurring budget surpluses during periods of economic growth, the reduction in debt during economic expansion would allow for deficits during recessions without creating overwhelming debt burden. The reduction in deficit spending following a crisis can be illustrated by the years following World War Two. Between 1940 and 1945, the national debt increased from $43 billion to $259 billion. During the same period United States gross domestic product (GDP) increased from $100 billion in 1940 to $213 in 1945. (Schiller, 2006) As a percentage of GDP, the national debt in 1945 was 121.6% (259 / 213 = 1.216) In the years following World War Two, federal spending fell. In 1945, federal spending was $92.7 billion and the federal budget deficit was $47.6 billion. By 1948, federal spending fell to $29.8 billion and there was a federal budget surplus of $11.8 billion (Britannica Almanac, 2004). Although federal spending has increased steadily since 1948, and there have been only a few years of surplus, for several decades budget deficits were relatively small. In general, GDP grew faster than the national debt. As a result, the national debt as a percentage of GDP declined. In 1980, the national debt was $914.3 billion, and the GDP was $2.795 trillion (Schiller, 2006). National debt as a percentage of GDP was 32.7% (914.3 / 2795 = 0.327). In the years since 1980, federal deficits have generally been substantial (with the exception of surpluses from 1998 to 2001). Particularly in the last few years, national debt has grown faster than GDP. As a result, national debt as a percentage of GDP has grown. In 2011, the national debt was $14.764 trillion and GDP was $15.094 trillion (Government Debt Chart, usgovernmentspending.com). National debt as a percentage of GDP was 97.8% (14.764 / 15.094 = 0.978). The substantial increase in national debt gives rise to the question of whether the debt burden is becoming too great. An increase in national debt from $914.3 billion in 1980 to $14.764 trillion in 2011 seems to be tremendous. However, the national debt as a percentage of GDP of 97.8% in 2011 is still less than the 121.6% figure in 1945. The post World War Two prosperity that allowed for the reduction of the relative magnitude of national debt as a percentage of GDP suggests that prosperity combined with small deficits may be possible. Proceedings of the 2012 Pennsylvania Economic Association Conference 218 Differences in circumstances and political atmosphere may make recreation of the post World War Two scenario difficult. World War Two deficits began at a time when the national debt as a percentage of GDP was substantially lower than the current level. In 1940, the national debt was $43 billion and the GDP was $100 billion (Schiller 2006). National debt as a percentage of GDP was 43% (43 / 100 = 0.43). It could be argued that the 43% figure in 1940 meant that the national debt had room to grow that it may not have today. The current debate regarding fiscal stimulus includes the suggestion by some economists such as Paul Krugman (2012) that the problem with the economic stimulus is that the magnitude of the stimulus is inadequate, and that deficits comparable to World War Two levels may be warranted. The relative magnitudes of current deficits are smaller than during World War Two. For example, in 1943, the federal budget deficit was $54.6 billion and GDP was $192 billion (Encyclopedia Britannica Almanac, 2005, Schiller, 2006). The federal deficit as a percentage of GDP was 28.4% (54.6 / 192 = 0.284). In contrast, in 2011, the federal deficit was $1.30 trillion and GDP was 15.09 trillion (Historical Federal Receipt and Outlay Summary, taxpolicycenter.org; Government Debt Chart, usgovernmentspending.com). The federal deficit as a percentage of GDP was 8.6% in 2011 (1.30 / 15.09 = 0.086). The apparent stimulus effect of deficits during World War Two suggests that larger deficits today might provide greater economic stimulus. However, the expected effects of such deficits on the debt burden might make such deficits politically impractical. If the federal deficit were increased to 28.4% of the GDP for 2011, the result would be a deficit of $4.286 trillion (15.09 X .284 = 4.286). With national debt as a percentage of GDP close to 100% as of 2011 and likely to exceed 100% by the end of fiscal year 2012, an acceleration of the growth of debt would be difficult to implement. If a fiscal policy had been employed that limited deficit spending to periods of recession, and if the result had been a substantially lower ratio of national debt to GDP, the government would have been in a better position to increase borrowing. Years of deficit spending during periods of expansion may effectively limit the ability to increase borrowing at a time when deficit spending may be warranted to stimulate the economy. A concern that might arise if World War Two levels of deficit spending were proposed is that the spending increases meant to be temporary might become permanent. A fiscal policy that would call for reduction in spending when the economy improves may be inconsistent with actual spending patterns. In contrast to the period following World War Two when spending reductions largely entailed cuts in defense spending from World War levels, current spending appears to be very difficult to cut because it entails social spending that many voters do not want to see cut, as well as necessary ongoing levels of defense spending. ADDRESSING ISSUES OF DEBT BURDEN Concerns about the increasing level of national debt as a percentage of GDP raises the question of what level of debt is too much. The issue regarding debt burden is not necessarily that the federal government will be unable to pay its debts, but rather whether the debt service can be handled comfortably. Currently, interest rates are so low that interest payments are relatively modest. In 2011, interest on the national debt was approximately $250 billion (United States Federal Budget, wikipedia.org) and the national debt was $14.764 trillion (Government Debt Chart, usgovernmentspending.com). This suggests an average interest rate of 1.7% (250 / 14,764 = 0.017). This level of interest does not seem to be an excessive burden. Perhaps the concern should be what could happen if ongoing deficits were combined with the level of interest rates that occurred in the late 1970s and early 1980s. For example, in March 1980, the interest rate on 3 month Treasury Bills was 15.526% (Treasury Securities, hsh.com). To illustrate a debt burden that could result in a difficult level of debt service, suppose that over a period of years, the national debt were to rise to 200% of GDP. Suppose that over the same period, expansion of the money supply were to give rise to substantial inflation, and that the Federal Reserve were no longer able to suppress interest rates, which were to rise to an average rate of 15% on Federal debt. The result would be interest payments of 30% of GDP (15% X 200% of GDP). Such a level of debt service could be a tremendous hindrance to the economy. POSSIBLE ACTIONS TO AVERT ADVERSE EFFECTS To avoid a debt burden that may hinder economic growth and may make increased deficit spending difficult at a time when fiscal stimulus is genuinely warranted, policy makers may want to consider seeking a fiscal policy comparable to the years following World War Two. Between 1945 and 1980, the national debt as a percentage of GDP fell not because of debt repayment, but rather because the economy grew faster than the national debt. If fiscal policy were adopted that would result in growth of the national debt that were less than the economic growth rate, the national debt as a percentage of GDP would decline. To illustrate, consider a deficit that would maintain the status quo. Suppose that the national debt and GDP were both $15 trillion, resulting in a ratio of national debt to GDP of 100%. Proceedings of the 2012 Pennsylvania Economic Association Conference 219 Further suppose that the economy were to grow at 4% per year. The GDP would rise by $600 billion (15 trillion X .04 = 600 billion). If the budget deficit were 4% of the national debt, the resulting deficit of $600 billion (15 trillion X .04 = 600 billion) would maintain the ratio of 100% (15.6 trillion / 15.6 trillion = 1.00). Alternately, suppose that the national debt and GDP were both $15 trillion, and the economy were to grow at 4% ($600 billion increase in GDP). However, suppose the budget deficit were 1% of the national debt. The resulting deficit would be $150 billion (15 trillion X .01 = 150 billion). Because the rate of growth of the national debt is less than economic growth, the national debt as a percentage of GDP would fall. The national debt would increase to $15.15 trillion (15 billion + 150 billion = 15.15 trillion). GDP would increase to $15.60 trillion (15 trillion + 600 billion = 15.60 trillion). National debt as a percentage of GDP would decrease to 97.1% (15.16 / 15.60 = 0.971). If such a fiscal policy were followed during periods of economic growth, over a period of years, reduction in the ratio of national debt to GDP would make increased budget deficits during recessions more palatable. CONCLUSIONS Economic activity tends to occur in cycles. Periods of growth are followed by periods of recession. During periods of recession, governments (and government agencies) attempt to reduce the severity of a recession through monetary and fiscal policies. Monetary policy entails actions by the Federal Reserve to maintain economic growth and economic stability. Decreases in the Federal Reserve’s discount rate may encourage investment. Expansion of the money supply may provide money needed for transactions as the economy grows. However, excessive expansion of the money supply can result in inflation. Fiscal stimulus entails attempts to stimulate economic growth through budget deficits. If government spending were increased while tax revenue remains the same, the result could be a multiplier effect that could increase aggregate demand during a recession. Similarly, if taxes were reduced while government spending remained the same, the resulting increase in after-tax income could result in increased aggregate demand. During a recession, increased aggregate demand can result in increased employment of labor and idle productive capacity. However, when full employment of labor and productive capacity is reached, continued fiscal stimulus can have adverse effects. out effect may not be an issue at current depressed levels of economic output, but it may become an issue if trillion dollar deficits occur during economic expansion. Debt burden in the United States is not so much a current problem, but rather is a potential problem. Although the national debt has grown substantially in recent years, as a percentage of GDP, the national debt is, as of 2012, is still less than the level at the end of World War Two. As a result of relatively low interest rates, the debt service (interest expense) on the national debt is currently manageable. Potential problems with debt burden may arise if the national debt as a percentage of GDP continues to grow. Reduction of debt burden does not necessarily mean that the government must have a budget surplus. If the relative size of the budget deficit were reduced such that the growth rate in the national debt were less than the growth rate of the economy, the national debt as a percentage of GDP would decline. Although high levels of spending and debt may not necessarily mean economic ruin, high levels of spending may have adverse implications for economic growth. Spending that begins as part of a fiscal stimulus package tends to become permanent, making prospects for reducing government spending dim. In 2009, Newsweek Magazine published a cover article titled “We Are All Socialists Now” (Meacham and Thomas, 2009) which stated that in the United States, total government spending was expected to be 39.9% of GDP in 2010, compared with 47.1% in the typical European country. The implication seems to be that government control of an economy does not necessarily mean that the government owns the means of production. Rather, it seems to suggest that if government has a claim on a large portion of output, that government has substantial control over the economy. As government spending in the United States approaches European levels, the United States may experience European style slow economic growth. If high levels of government spending hinder economic growth, continued expansion of government spending as a percentage of GDP may further hinder economic growth. Perhaps diminished economic prospects are the long run effect of decades of employing Keynesian theories of fiscal stimulus as a rationale for budget deficits and government expansion. Slow economic growth may be the price of fiscal stimulus. Adverse effects of fiscal stimulus may include crowding out effects and effects of increasing debt burden. The crowding Proceedings of the 2012 Pennsylvania Economic Association Conference 220 REFERENCES Encyclopedia Britannica 2005, Chicago: Encyclopedia Britannica, Inc. Friedman, Milton, 1992, Money Mischief: Episodes in Monetary History, New York: Harcourt Brace Jovanovich. Government Debt Chart, www.usgovernmentspending.com/spending_chart_1997_201 7usb_hof, retrieved 5/21/2012. Historical Inflation, www.inflationdata.com/inflation/inflation_rate/historicalinfla tion.aspx, retrieved 6/16/2012. Keynes, John Maynard, 1923, A Tract on Monetary Reform, London: Macmillan. Keynes, John Maynard, 1936, The General Theory of Employment, Interest and Money, London: Macmillan. Meacham, Jon, and Evan Thomas, 2009, “We Are All Socialists Now”, Newsweek, February 16, 2009 Issue, from www.thedailybeast.com/newsweek/2009/02/06/we-are-allsocialists-now.html, retrieved 5/24/2012. Parkin, Michael, 1998, Macroeconomics, New York: Addison-Wesley. Samuelson, Paul A., Economics, New York: McGraw-Hill. Schiller, Bradley R., The Macro Economy Today, New York: McGraw-Hill Irwin. Treasury Securities 1980-1989, www.hsh.com/indices/tsec80s.html, retrieved 5/19/2012. United States Federal Budget, http://en.wikipedia.org/wiki/2011_United_States_federal_bu dget, retrieved 5/22/2012. Krugman, Paul, 2012, End This depression Now!, New York: W. W. Norton & Company. Proceedings of the 2012 Pennsylvania Economic Association Conference 221 DISCUSSANT COMMENT ON LIVING IN KEYNES'S LONG RUN: THE EFFECTS OF THE OVERUSE OF ECONOMIC STIMULUS Michael J. Hannan, Ph.D. Office of the Provost/VPAA Edinboro University of Pennsylvania Edinboro, PA 16444 In general the paper sought to examine the role of short run economic stimuli on long run economic growth. However, the paper primarily presented basic macroeconomic concepts such as the definition and role of fiscal and monetary policy and potential economic implications of such policy. Much was presented at a level expected in a principles of macroeconomics course and nearly no relevant or recent literature was cited. Throughout the paper, I attempted to keep in mind the main premise stated in the abstract, which focused on how short run stimuli would impact long run economic growth - this is a good topic. Instead, the paper discussed two major concepts. The first concept was the crowding out effect, resulting from an ever-rising debt burden leading to higher interest rates that would crowd out private investment. While this is a basic concept, there was no literature cited or empirical work performed to show that such an effect was expected. The paper seemed to ignore the existence of world money markets and the fact that crowding out would not necessarily occur provided that other global lenders were willing to invest in U.S. federal obligations. The second concept focused on how high government spending as a percent of GDP would lead to greater government control of the economy and hinder economic growth. This was rationalized from data showing the figures of 39.9% for the U.S. and 47.1% in Europe. Since European growth was characterized as slow, it was reasoned that as U.S. government spending approached this level, the country would also suffer slow economic growth in the future. No literature was cited regarding this hypothesis and the author unfortunately ignored other economic, historical, political and demographic differences between the U.S. and Europe in making this assertion. Slow economic growth in Europe is due to many factors beyond government spending as a share of GDP. There were a number of other issues within the paper that will not be discussed here. Instead, I offer a few suggestions for more targeted research. First, as mentioned earlier, this is an interesting topic and if one has an interest in the crowding out effect, I would recommend developing a thorough review of the literature on this topic. With historically large U.S. budget deficits and the European debt crisis, it is a good time to be looking anew at crowding out possibilities and there should be recent literature in this area. While interest rates are at very low historical levels, there is concern about the amount and type of assets held on the Federal Reserve balance sheet and whether this may eventually lead to higher inflation and higher real interest rates. Second, in a number of places in the paper the author attempts to draw parallels between current deficit and debt conditions with those that existed during WWII. If this is an area of interest, the author should consider a thorough and focused comparison between conditions in the two periods recognizing differences in the sizes of the economies, Federal Reserve policies, industrial composition of the economy, globalization of financial and goods markets, among related variables. Proceedings of the 2012 Pennsylvania Economic Association Conference 222 KANTIAN MARKETS, BOYCOTTS, AND EFFICIENCY Richard Robinson School of Business, E336 Thompson Hall SUNY at Fredonia, Fredonia, NY, 14063 ABSTRACT The welfare efficiency of competitive markets certainly requires that participants conform to moral duties. These include negative duties such as those against fraud, deception and coercion, and also positive duties such as those that favor beneficence. Mainstream Western philosophical notions of these duties are examined here where product, capital and internal labor markets are shown to be capable of pricing conformance to these duties through both formally- and informally-organized boycotts. A definition of Kantian markets is provided here which properly categorizes and explores those duties shown to be necessary for any claim of market efficiency. Through this exploration, the so called “Adam Smith” problem concerning the morality of the invisible hand is resolved. JEL: D63, D64. Keywords: Boycotts, Ethics INTRODUCTION Market philosophers have long argued about “the Adam Smith problem,” i.e. the seemingly apparent contradiction between Adam Smith’s ethical philosophy as expressed in The Theory of Moral Sentiments (1759), and the egotistical agent of the invisible-hand as expressed in Wealth of Nations (1776). In attempting to resolve this apparent conflict, Samuel Fleischacker (2004, p. 91) argued that “human beings can pursue even their individual interests together, that even society without benevolence need not be a hostile society, that economic exchange, even among entirely self-interested people is not a zero-sum game.” More recently, Mark White (2008) argued that in the Wealth of Nations, Smith did not make an argument in favor of egoism, although he showed that pursuit of self-interest is sufficient for a hypothetically flourishing society. Smith was only arguing that because agents knew their own self-interests, the economic philosophy of mercantilism was inefficient by its nature, i.e., that markets can operate based on self interest, not that they should operate in this way. The moral behavior of participants still determines whether markets are hostile to society’s interests, and the extent of this hostility. Chambers and Lacey (1996) point out that markets can even price society’s sense of ethics in that through product- and capital-market boycotts, society forces its ethical will on producers. We have a considerable history of product- and capital-market boycotts of those producers who violate society’s sense of propriety. Generally these boycotts have been effective in changing business practices in areas such as child labor, negative environmental impacts, consumer fraud, and the like. A few of these boycotts are reviewed below. In general, economists would argue that pricing away practices that are generally perceived as unethical is an effective method for society to reach an efficient allocation of resources; that if boycotts of this sort are effective, then in general, public welfare must be enhanced. Society wishes to discourage immoral practices, at least what it considers as immoral, and the market can certainly play a role in expressing society’s outrage. Through this mechanism, society must believe it is better off or it would not have supported the boycott to begin with. This is the invisible hand argument supporting the ethical workings of competitive markets. If society is not outraged by some practice that philosophers, or other moralists, consider to be unethical (assuming the public is made aware of this practice), then the democratic discourse of the marketplace has considered this practice and judged it as meeting ethical constraints. Many boycotts do fail. Perhaps we live in a highly ethical society; perhaps we live in an amoral society. Whatever society’s ethical sense, it will be reflected through its markets. In this light, it is worthy to examine: 1. a hypothetical society where markets meet certain generally accepted philosophical standards which are expressed here as Kantian standards of objective ethical criteria, 2. examine to what extent our Kantian criteria helps to understand the market forces that price ethics through different types of boycotts, and 3. judge whether such markets are necessary for welfare efficiency. Shleifer (2004) argues that competition weakens the ethical behavior of firms. He argues that ethics evolved from the needs of social cooperation; that some ethical standards are antiquated, or develop as political indoctrination, or are subject to cultural relativism, and capable of economic versus morality tradeoffs. As an example, Shleifer argues that demands for cheaper shoes overcome society’s objections to asian child-labor. This is a particularly simplistic and dark view of the relation between morality and markets, a view we attempt to refute here. I show that the evidence supports the argument that through boycotts, markets force changes in some business practices society judges as unethical. I organize this empirical evidence around certain boycott Proceedings of the 2012 Pennsylvania Economic Association Conference 223 characteristics, a categorization that I hope allows some understanding of the forces that lead to success. Kant (1785, p. 402-404) argued that the CI merely formalized the ethical decision-making of the common person.3 Any organization of society’s view of ethics must be based upon more substantive philosophy than cultural relativism or political indoctrination. For this reason I organize our analysis around the widely accepted Kantian notions of duties, both perfect (full obligations), and imperfect (partial obligations with practical limits). This view of duty lays the foundation for our analysis of the market pricing of ethics. As already stated, Kant presents three versions of the categorical imperative.4 Kant envisioned these versions, presented below as three “formulas,” as entirely consistent with each other, and in fact he envisioned each as logically necessitated by the others. It is important to keep in mind that these formulas are axiomatic guides for the derivation of applicable and practical maxims: We also show that the Kantian criteria, and the associated duties required of market-participants, have strong implications for welfare efficiency. I argue that the pricing of society’s sense of ethics is essentially the pricing of certain types of Kantian duties, all of which conceivably can be fairly priced by the market. Since these duties are expressed by the practical moral maxims derived from, and in conformance with, Kant’s categorical imperative (CI), we begin with a brief examination. Formula 1 – the formula of autonomy or of universal law: “I ought never to act in such a way that I could not also will that my maxim should be a universal law.” (1785, p. 402) MORAL MARKETS Kant’s Categorical Imperative, Markets and Practical Moral Maxims Kant argued that the very notion of morality requires free will, and this requires individuals who are autonomous in their actions. Unlike the Plato-Socratic approach, or that of Aristotle (1953), or of religious ethics (all perfectionist), Kant did not attempt to objectively establish notions of the good in human characteristics, actions or results. He argued that fundamental moral principles as they apply to society must be adopted by a rational, open and democratic discourse, but this discourse should only be engaged with associated rational reflection.1 Kant argued that from rational free will, a “supreme practical principle,” called the categorical imperative (CI), follows, namely that the moral maxims we personally decide to live by must be those we will that all live by. As a result, through democratic discourse, our moral maxims become universal; they must apply to all. Kant actually presented three interdependent versions of this categorical imperative (CI), as reviewed below. These versions do not stem from abstract laws of nature. They would not exist without the imposition of the reasoning man. They do not transcend our experience as in Socratic philosophy. They are person-centered and proceed from the reasonable (rational) free will.2 These versions of the CI appear to us as obviously true. They are therefore axiomatic, so we can derive practical maxims from them. They need no prior argument to establish them. They stand alone in forming the premises for our supporting arguments in favor of our derived practical maxims. Furthermore, they should not be viewed as unrealistic, or overly formulaic. In fact, Formula 2 – the formula for the respect for the dignity of persons: “Act so that you treat humanity, whether in your own person or in that of any other, always as an end and never as a means only.” (1785, p. 429) Formula 3 – the formula of legislation for a moral community: “All maxims that proceed from our own making of law ought to harmonize with a possible kingdom of ends.” (1785, p. 436) Kant’s first formula, the imperative of universal law, prohibits us from behaving by personal maxims that are applicable only to us, and that are designed only for our convenience. For example, if our business temporarily suffered from financial distress, and we decided that it would be acceptable to commit some fraudulent act to ameliorate our problem, we would violate the imperative of universal law. We could never will this temporarily fraudulent behavior to be universal. That would be equivalent to willing that the foundation of trust upon which our business relations are built be universally destroyed. A maxim of “commit fraud only when we temporarily suffer from financial distress” is unacceptable in Kant’s ethic. In a similar way, Kant’s formula for the respect for the dignity of persons would also be violated by the maxim of fraud described in the above paragraph. Fraud is essentially a lie. It deceives others into serving our own ends, while not allowing others to pursue their personal ends. This example illustrates the consistency of the first two formulas. Indeed, Kant argued that one formula logically follows from, and is necessitated by the other. The motivation for pursuit of the first two formulas lies in the third, the formula of legislation for a moral community. This motivation is rooted in Greek philosophy as developed in the Socratic dialogues, such as Gorgias (see Plato, 1989). In that dialogue, Socrates develops two principles: • “No man does evil voluntarily.” • “It is better to suffer evil than to commit evil.” Proceedings of the 2012 Pennsylvania Economic Association Conference 224 With respect to the first of these propositions, Socrates states that to know what is good, and to choose what is evil, is an absurdity. No person knowingly and willingly selects evil; they select evil only in spite of the fact that it is evil, not knowingly and voluntarily because it is evil. Evil is merely the necessary result of ignorance. This is true because willingly committing evil is self destructive of the reasoning individual, and no one rationally selects self destruction, at least according to Plato and Socrates. (Of course, this assumes that we are all reasoning individuals.) Note that this preserves the notion of free will in that people could still select to perform evil actions, but they would do so only out of ignorance. Reasoned free will is what gives meaning to life, according to Socrates, Plato, and certainly also Kant. By selecting evil, a person therefore destroys that which gives meaning to life. Kant, however, wants to extend this Greek philosophical argument. Kant argues that we seek a kingdom of ends, what in Greek philosophy is termed the good. By kingdom, Kant means “the union of different rational beings in a system by common laws” or maxims. (1785, p. 4.433)5 Through the second formula, duties follow, but they are motivated by the pursuit of this kingdom of ends. These duties are actually derived from the practical ethical maxims formed from the categorical imperative. By harmony, Kant means that these rational beings pursue consistent and coordinated duties aimed ultimately at pursuing this kingdom of ends. Moral actions are therefore those that are motivated by the pursuit of this ultimate good. They cannot be those that serve only the self at the expense of others in this “union of rational beings.” (1785, p. 4.430) Indeed, Kant argued that examination of motivation is the only basis for judging the morality of some action, and Formula 3 provides the only justifiable motivation. Other possible motivations are egotistical, that is they are inherently selfish. Kant’s kingdom of ends refers, of course, to an harmonious overall society, one where reasoning people pursue maxims which they form democratically and therefore find acceptable, and which are derived from the other two formulas of the categorical imperative. Can this kingdom of ends motivation ever be applied to the neoclassical economic-market participant who we generally describe as self interested? This is the apparent contradiction examined here. Certainly one can argue that pursuit of this kingdom of ends is not practical, especially within the neoclassical firm. What else could motivate managers: greed, gluttony, sexual addiction, … ? These pursuits are also not practical in that they can never be ultimately satisfying, but only transitory in satisfaction, if even that. It can be argued that pursuit of harmony, however, has the potential of leading to success in business or other market endeavors, even if perfection in that pursuit is not ultimately achievable. Pursuit of profits may be the goal of the firm, but the motive for market participants behaving ethically should be the pursuit of the kingdom of ends. The practical maxims derived from the CI provide the moral constraints on market behavior. Kantian Markets The notion of a Kantian market is necessarily abstract, but given the review of the categorical imperative presented above, my definition is nonetheless straightforward. It requires two conditions: 1. Kantian markets are ruled by a complete set of practical moral maxims that are consistent with the categorical imperative. 2. The market participants in Kantian markets are all motivated by the pursuit of a moral community. Concerning the first condition, by a complete set of moral maxims we mean a set that is sufficient to operationalize the first two formulas of the categorical imperative. For example, this set of maxims would include those aimed at allowing no personal exceptions to their conformance, i.e., all must comply because all must be allowed to pursue their own ends provided they do not impinge on the freedom of others to pursue their own ends.6 This certainly requires that deception, outright fraud, and coercion are declared in violation, and this would apply to all contracts whether explicit or implicit, i.e., those contracts reached by both formal (legal) and informal agreement.7 The second condition is clearly the most abstract. It is the formula for the pursuit of the kingdom of ends that is the most difficult to apply to markets. We cannot envision legally requiring proper motivation, but we still must specify this requirement to describe our abstract notion of Kantian markets. In this abstract idealistic model, all our market interactions must be constrained by this motivation. Of course, the above two conditions are not independent. The second condition requires the first in that the pursuit of a moral community implies that all agents approve the maxims referred to in the first condition. It must be explained how the kingdom of ends motivation could operate in a market system of largely self-interested transactions. I state “largely self interested” since there are market transactions among more intimately connected agents such as those within families, within firms, and even within friendly communities. We can all envision the motivation of pursuing a moral community to be accepted and active in these latter groups, but we ponder how such motivation could flourish in markets with more disinterested agents and transactions. The answer is explained by Sullivan (1989, p. 214-16). In Kantian analysis, there are two levels of moral community to be pursued: Proceedings of the 2012 Pennsylvania Economic Association Conference 225 1. 2. a civic moral-community of justice in the sense of pursuing all the moral maxims needed for judicially meeting the second formula, i.e., an impersonal community of judicial duty, a community that meets the required judicial duty conditions, but that also fully pursues the kingdomof-ends motivation, i.e., a community of full moral duty. We can envision sub-communities within an overall community of judicial duty that also meet the conditions for being a community of full moral duty. An overall market economy should certainly meet the conditions for a community of judicial duty, and perhaps markets such as the labor market that is internal-to-the-firm could be envisioned as meeting the conditions for a community of full moral duty. It is shown below that the efficiency of an overall economy’s markets requires that the judicial level of duties be a necessary, but not sufficient condition. Other markets, such as the internal-labor-markets within firms, must meet certain necessary additional full moral duties. Notions of Duties and Market Efficiency In Kantian analysis, there are two types of duties: perfect and imperfect duties, which are frequently termed negative and positive duties. (See Sullivan, 1994, Chapter 8, for a review, and in the original see Kant, 1785, p. 4:421 – 431.) Duties stem from the moral law, or in our analysis, they stem from the moral maxims specified as necessary for a community of full moral duty. Perfect duties, often termed negative duties, are absolute prohibitions against actions that violate the second formula, actions involving deception, fraud, coercion and the like. Imperfect duties, often termed positive duties, stem from obligatory actions that have practical limitations, actions such as beneficence. For these duties, we are obliged to perform some actions such as various forms of charity, but for practical survivability reasons, we cannot give all to the poor. Although we are obliged to be charitable, there must be practical limitations to this charity; we are not obliged to make ourselves poor because of these activities. Judicial duties are negative (or perfect) duties; they prohibit actions of fraud and the like. A community of full moral duty clearly requires that all are committed to the second formula (respect for the dignity of others), and its associated moral maxims. The markets that compose our overall economy must be motivated to exhibit these duties or our markets would be, in a broad sense, inefficient and probably considerably underperforming. A community of full moral duty, however, requires that both negative and positive duties be fully obligatory. We could hope that the latter classification of duties should be performed within certain markets where agents are more closely connected, such as the internal labor markets of firms, rather than in the broader more impersonal markets where agents are distant from each other. In Kantian markets, the self worth of agents motivates them to “pursue their own morally permissible welfare and happiness, but also to promote those of others.” (Sullivan, 1994, p. 156.) Markets are an expression of the mutual dependence of their participants, who we assume aim at fulfilling their own needs. Mutual respect requires that these agents treat each other not merely as the means to their own ends, but also by allowing others to pursue their ends, i.e. conditions specified under the second formula. For freely competitive markets, we assume that both sides of a transaction pursue their own ends, but they are also interested in enabling others to achieve their ends, and the closer the market participants, the greater the interest. Positive duties are clearly necessary for promoting the interests of all, but in Kantian analysis, the closer the relationship between agents, the greater the expectation of duties of a positive sort. Market transactions between agents who are closer, rather than more distant, should exhibit these positive duties with practical limitations that would themselves strengthen the closer the relationship is. The obligation for beneficence is stronger the closer the agents, and this closeness is largely determined by the nature of the particular market in question. For example, consider the duty of charity as briefly reviewed above, especially in the context of the competitive firm, in particular the competitive corporation with publicly traded shares. These firms do have charitable expenses that are usually aimed at community relations, or the seeking of positive publicity. In either case, these contributions have bottom line impacts that justify their expenditures. There are other charitable contributions, however, those explained as being required of corporate social responsibility (CSR), that are unrelated to the bottom line, but are said to be required from a social-duty obligation. These CSR expenditures are, however, expected to be impersonal in that the corporation is distant from the intended beneficiaries. The counter argument of Chambers and Lacey (1996) is that distribution of these funds back to the corporate owners (in the form of dividends) allows those closer to intended beneficiaries to make the charitable decisions. Corporate shareholders legally own these revenues, and their charitable expenditures by management pose agency problems unless these expenditures positively affect future revenue. The duty of beneficence is greater among those closer to the beneficiary. We would expect, therefore, a greater extent and degree of effectiveness in charitable giving if these revenues were expended by owners rather than management. Complete market efficiency clearly requires a community of judicial duty. Markets based on fraud and deception could not possibly be termed efficient since these markets would implode in that they would be abandoned by participants. The more participants perceive the probability of Proceedings of the 2012 Pennsylvania Economic Association Conference 226 encountering outright fraud or even partial deception concerning the product, service, or payment, the more participants would leave the market. Furthermore, without the moral motivation of the pursuit of the kingdom of ends, we could not expect agents to fully conform to these negative duties, although fear of retribution, or ostracism if caught, might motivate a considerable degree of conformance. For this reason, the motivation of the third formula is also necessary for many market transactions, but perhaps not all. THE MARKET PRICING OF ETHICS Motivation for Boycotts As reviewed above, Chambers and Lacey (1996) pointed out that markets are capable of pricing ethics through boycotts of both product and capital markets. Boycotts are expressions of moral disapproval and outrage by multiple agents who refuse to either purchase a product, or participate in its production. (A strike is a sort of boycott.) With widespread participation, boycotts are expressions of society’s moral disapproval. We must also point out that labor mobility, and certainly participation in internal labor markets, also are capable of pricing ethical behavior. People change firms if they believe that the internal responsibility assignments, evaluation and reward practices of their firms are unethical, or that abusive or humiliating behavior is practiced. Certainly a firm’s employment reputation influences its ability to engage talented employees, and therefore the firm’s performance. All of these market forces are essentially transfer mechanisms for imposing society’s sense of ethics onto the firm. With respect to these market forces, the relevant question is “What do we mean by society’s sense of ethics?” Our notion of Kantian markets assists in answering this question. To the extent of deviations from the two levels of Kantian markets reviewed above, we might expect boycotts of all three of the markets mentioned: product, capital, and internal labor. In particular, violations of obligations of a community of judicial duty (prohibitions against fraud, deception and the like) should solicit either formally or informally organized boycotts to the extent that society is outraged by the violations, i.e., the greater the contempt for society’s sense of judicial-duty obligation, the greater society’s outrage. Formally organized boycotts have leaders; they manifest reasoned argumentative-communication from these leaders to possible participants (propaganda); and public expressions about the moral violations perceived. Informal boycotts do not have leaders, but do have a variety of communications (letters to the editor, blogs, etc.) that express outrage. Informal boycotts are often so quickly successful that formal organization does not have time to be arranged. The perceived immoral behavior is altered quickly before the boycott can be further organized. Given the closeness of internal employee arrangements, we expect that positive duty violations would solicit reactions from participating employees, perhaps even boycotts. Beneficence is expected among close agents within firms. This is a positive duty, and as such, it has practical limitations. The extent of its practice is actually subject to a marginalist solution, but certainly there is the potential for beneficence to impact the firm’s returns. Hence the motive of pursuit of a community of full-moral duty provides positive duties that are also subject to market forces if violated, and perhaps boycotts. With respect to our duties, both positive and negative, we can conclude that markets are capable of exerting pricing forces. Our notion of Kantian markets, with their associated perfect and imperfect duties, are not unrealistic unless the operational specifications of the duties themselves are overly extreme, and it is unlikely that the public would form unrealistic-impractical notions of duty. Markets would likely only react to violations of duties society judges as ethically necessary and realistically formed. Chambers and Lacey (1996) point out the democratic nature of the firm’s pursuit of profit. As envisioned by Smith (1776), the marketplace is essentially a voting place where the invisible hand moves participants to respond to meet society’s needs. Under conditions of sufficient knowledge about the workings of the firm (especially about any externalities generated), the product market will democratically voice the concerns of the populace about perceived ethical lapses. Firms that pursue maximization of profit as a goal must respond to society’s sense of moral rightness, and this sense forms the constraints on the profit goal. As evidence of this, we examine and classify several successful boycotts in a section below. By “successful” we mean that they ultimately lead to changes in firm behavior, and/or social policy. Judging the Morality of Markets White (2008) points out that a market in which agents follow only negative (perfect) duties would be the impersonal marketplace of Smith’s Wealth of Nations; it would be minimally ethical. In Smith’s The Theory of Moral Sentiments, it is the sympathy for the suffering of others that motivates moral behavior. Indifference to the suffering of others essentially violates Kant’s second formula of respect for the dignity of others, of treating others as an “ends.” Markets are, however, mostly impersonal, and feeling sympathy for the suffering of another agent during an impersonal market transaction is not generally expected. Kant’s analysis of respectful beneficence in such transactions is relevant for these impersonal market transactions: “Since Proceedings of the 2012 Pennsylvania Economic Association Conference 227 our duty is to behave as if our help is either merely what is due him, or but a slight service of love, to spare him humiliation and maintain his respect for himself.” (1798, 4:448-9) Clearly, even impersonal market transactions require more than mere performance of judicial duties if market transactions can be said to be ethical, i.e., we need more than a minimally ethical marketplace. It is argued here that ethical markets require motivation from both Smith’s sympathy for the suffering of others, and Kant’s pursuit of the moral community. The latter motivation is capable of stirring emotions of outrage over violations of established negative duties, and perhaps even violations of established positive duties. In essence, the boycott mechanism needs both motivations: Smithian sympathy, and Kantian outrage. The former is likely to lead to new developments of, or variations in required duties, and the latter is likely to lead to enforcement of the already established duties. When we witness the suffering of others due to some market practice, a suffering that might be addressed by some new norm, we form a hypothetical alleviation for the problem. If others wish to alleviate this suffering by similar means, a boycott of that market practice might be organized. It is the suffering of others that ultimately motivates the boycott and possible new norm. In other situations that involve the flaunting of existing norms, a Kantian outrage over this flaunting might also lead to a boycott. When active in markets, the boycott mechanism transfers these emotions of sympathy and outrage into imposing society’s sense of ethics onto firm behavior. We must realize that boycotts are expressions of reasoned reflection about acceptable and enforceable moral duties. They are formed through public-democratic discourse, and hence they meet the Kantian requirements for forming or enforcing moral maxims. Gauthier (1986, Chapter VIII, p. 84-5) argues that markets are the ideal model for an ethical society. He argues that “in understanding the perfect market as a morally free zone we shall be led back to its underlying, antecedent morality” of mutually agreed upon constraints on behavior. This is a peculiar use of the term “morally free zone.” It appears to perpetuate the notion that perfect markets are essentially amoral, and that only violations of the neoclassical perfectmarket definitional-conditions should be interpreted as unethical. Given our boycott analysis, however, a preferable term in this context would be fully-moral zone, where society’s sense of ethics is fully operationalized through the pricing mechanism of competitive markets. The less competitive the marketplace is, the less the boycott effect could impose society’s ethical sense on the firm. A perfect monopoly would have a decreased motive to respond to society’s outrage. Boycott Classifications and Effectiveness Today’s social media particularly enables the organization of boycotts. For example, Facebook offers a boycott communication page. As a result of instant electronic communication, attempts to organize boycotts are numerous and with lower transactions costs than in the pre-social-media era. Complaints against business practices are easily masscommunicated, and similar complainers are more easily found. A casual review clearly indicates that most of these boycott attempts fail in having any effect on the business in question. In this section, however, we wish to review several of the more classically successful boycotts for the purpose of a possible bifurcation- classification system which might enable a judgment of conditions that are more likely to lead to boycott success. We also review some less than successful boycotts for comparison purposes. The successful product-market boycotts examined here include the California grape boycott in support of the Migrant Farm Workers’ Union during the 1960s and 1970s; the tuna boycott of Bumblebee, Starkist, and Chicken of the Sea in the late 1980s; and the Calvin Klein clothing boycott in 1995. A successful capital-market boycott is also examined, i.e. the South African boycott of the 1960s. The ongoing Swiss company Nestle Foods boycott (due to problems associated with its powdered baby-formula exports to Africa), and the boycott of the French company Danone in 2001 (due to its employee relations) are two less-than-successful attempts reviewed here for comparison purposes. As reviewed above, there are two categories of market interactions: (1) the impersonal transactions of strangers, and (2) the more intimate contacts of those familiar with each other during ongoing transactions. The example I reviewed of the former category is the supermarket purchase, and the example I presented for the latter category is the internallabor market. In addition, I suggested that although both categories require a complete judicial conformance of negative duties, and both require some degree of positive duty, it is the latter category that is more likely to require a higher degree of positive duty given the greater personal contact. There are more extensive personal obligations of beneficence in the internal-labor market. I also suggested that there are two motives for organizing or joining boycotts: (1) Smithian sympathy, and (2) Kantian outrage. The former motivates individuals to attempt to organize new social norms, and this is motivated by perceived sympathy for the suffering of others, a suffering that current social norms do not alleviate. The latter motive (Kantian outrage) stems from concern for the flaunting of the perceived current social norms or laws, the more blatant the flaunting, the greater the outrage, and the more likely a boycott will be organized and successful. Proceedings of the 2012 Pennsylvania Economic Association Conference 228 Table 1 presents a two-by-two matrix that categorizes boycotts into whether the perceived violation is between close or distant relations, and also whether the boycotts appear to be motivated by Smithian sympathy (a desire to develop new norms), or Kantian outrage (a desire to see current norms enforced whether these norms are actual law or generally accepted, but not legal, social standards). We classify six boycotts according to this matrix, four classic and successful boycotts, and two more recent, but unsuccessful boycotts. Note that in Table 1, it is questionable whether boycotts involving distant relations are purely due to perceived violations of negative duties. They might involve some degree of flaunting of a lower-level of positive duty. In a similar way, violations involving those with close relations could involve violations of negative duties as well as positive duties as indicated. Table 1: Boycott Classifications Smithian Sympathy Kantian Outrage Distant Relations Close Relations Tuna South African Nestle Calvin Klein Danone California Grape The Calvin Klein boycott, the tuna boycott, the South African disinvestment boycott, and the Nestle Foods boycott all involved the more distant relations of impersonal market transactions. Although not formerly an organized boycott, a rapid reduction in demand occurred for Calvin Klein’s products in response to the public’s negative outrage concerning the company’s use of pubescent children in sexually provocative advertisements during 1995. These billboard and magazine advertisements were withdrawn quickly. This can be categorized as an informal boycott. The company’s policies changed too rapidly for any organized boycott to be formed. (See www.icmrindia.org/casestudies/catalogue/Marketing/MKTG 084.html for a complete story of these events.) The Nestle boycott concerns its marketing of baby formula in less-than-developed countries, especially Africa. The baby formula, as compared to breast feeding, is criticized for preventing a full development of the infants’ immune system. Although the boycott was formally organized in 1978, publicity concerning the problem began in 1973. Various agreements with the company were reached over the years, but these were judged as quickly violated, and the boycott reorganized. It is currently ongoing. These agreements clearly were attempts at establishing new norms for marketing baby formula, and as such, they clearly involved Smithian sympathy to develop new socially acceptable norms. (See the Wikipedia story on the Nestle boycott.) Both the French company Danone Foods and the California Grape Boycotts involved labor relations. As such, both involved potential violations of close relations involving employees. The Danone boycott came in response to an announcement that the company would layoff approximately 200 employees in March, of 2001. The announcement came as a surprise to those separated from the company. Some local mayors ordered their hospitals and schools to boycott purchases from Danone, and organized a more general public boycott. The impact on Danone’s revenues and security prices was modest, and the boycott was eventually abandoned, but since Danone previously had very good employee relations, and these relations deteriorated, perhaps the boycott had some effect. (See www.paristempo.com/danone1.html for a complete story about these events.) The tuna boycott involved the methods for catching tuna, which are frequently found with dolphin. Netting methods catch both, and in the past, both have been canned as one fish. In 1986 the International Marine Mammal Project and the Earth Island Institute began the boycott. In 1990, Starkist, Chicken of the Sea, and Bumblebee signed agreements to stop purchasing, processing or selling tuna caught in the offending way, and the U.S. Legal Standard for the Dolphin Safe Tuna label was established. This boycott clearly exhibited Smithian sympathy for development of new norms, and also clearly involved offenses of negative duties (perceived broad environmental degradation) of distant transactions. (See the Wikipedia story on the history of the tuna boycott.) The California Grape Boycott began in 1966 with the organization of the United Farm Workers as led by Cesar Chavez. The boycott was aimed at consumers of table grapes with the objective being the establishment of collectivebargaining agreements with the grape growers. The boycott was nationwide, and by 1970, a majority of growers had signed collective agreements. The boycott was generally judged as successful. (See the Wikipedia report titled “Delano Grape Boycott.”) Racial-segregation policies formally began in South Africa in 1948. Boycotts of those companies doing business in South Africa began in the 1960s, along with boycotts of sports teams, university exchanges, and export products. Capital market boycotts also voice society’s demand for moral reform of business behavior, and these were initiated in the late 1960s by university and other foundations. In 1994, multi-racial elections were held and the apartheid policies were entirely ended. These boycotts were clearly initiated from Smithian sympathy, and involved distant relations. (See the Wikipedia story concerning the history of the South African boycotts.) Proceedings of the 2012 Pennsylvania Economic Association Conference 229 CONCLUSION Two of the six boycotts reviewed above, Nestle and Danone, were motivated by Smithian sympathy, but they failed in that the clearness of the norms to be established was murky. It was not entirely clear that the layoffs at Danone (small in magnitude as compared to the size of the firm) should have been prevented, nor that the baby formula problem at Nestle was sufficiently widespread to warrant prohibition. Two of the successful boycotts (the tuna and South African boycotts) were also motivated by Smithian sympathy to establish new norms: new standards for fishing and canning in the former boycott, and new integration norms for the latter. Both established clear and well-defined new norms that were ultimately formed. The California grape and Calvin Klein boycotts were motivated by Kantian outrage over violation of clear norms. The former involved the grape growers refusal to recognize the migrants’ union, a violation of clear expectations for employee relations. The latter involved a company’s exploitation of children in violation of social-sexual norms. In both cases, the norms violated were clear and generally accepted by society. Boycotts of the Smithian sympathy sort are clearly the least likely to succeed. New norms are difficult to form, and success would require a clear need for a well-defined new norm. Boycotts involving Kantian outrage over flaunting a clearly established norm are the most likely to succeed. It is most important to recognize that it is the profit motive that allows boycotts to be particularly effective. In essence, profit seeking allows the market-voting mechanism to function. I argued that formal and informal boycotts are an efficient mechanism for transferring society’s sense of ethics through the marketplace and onto the competitive firm. For this mechanism to work, society needs clarity concerning the duties demanded of market participants, a clarity provided by the Western ethical tradition of Kantian analysis of perfect and imperfect duties. These duties are derived from the practical moral maxims that stem from reflective-reasoned democratic discourse. When we envision the marketplace (product markets, capital markets, and internal labor markets) as essentially a mechanism for voting, the expectation of ethical duty demanded of the participating agents are viewed as voiced. The above definition of Kantian markets voices two separate levels of what is meant by moral markets: the community of judicial duty, and the community of full moral duty. The former notion is applicable for the more impersonal product and capital markets. The latter notion is applicable for markets exhibiting closer relations such as the internal labor market. Boycotts work to enforce ethical notions of both types of communities depending upon the type of market. The motivations for these boycotts are Smithian sympathy and Kantian outrage, the former leads to the development of new market duties, and the latter leads to enforcement of existing norms. This Kantian analysis more properly describes the competitive market’s ability to pursue welfare efficiency than the traditional neoclassical model that restricts the ethical content of markets to violations of the perfect competitive conditions. In this sense, it properly compliments this neoclassical analysis. It properly moves economic notions of competitive markets from the amoral (morally neutral) vision to an ethical vision. REFERENCES Aristotle, 1953, The Ethics, Book II, Penguin Classics, London, UK. Chambers, Donald and Nelson Lacey, 1996, "Corporate Ethics and Shareholder Wealth Maximization," Financial Management, Spring/Summer, p. 93-95. Fleischacker, Samuel, 2004, “On Adam Smith’s Wealth of Nations,” Princeton University Press, Princeton, NJ. Gauthier, David, 1989, Morals by Agreement, Oxford University Press, Oxford, UK. Kant, Immanuel, 1785, Fundamental Principles of the Metaphysics of Morals, in Basic Writings of Kant, edited by Allen W. Wood, The Modern Library Classics, The Modern Library, New York. Proceedings of the 2012 Pennsylvania Economic Association Conference 230 _____________, 1798, The Metaphysics of Morals, edited by Mary Gregor, Cambridge University Press, Cambridge, UK. Plato, 1989, The Collected Dialogue Including the Letters, edited by Edith Hamilton and Huntington Cairns, Bollingen Series LXXI, Princeton University Press. Rawls, John, 2001, Justice as Fairness: A Restatement, Belknap Press of Harvard University Press, Cambridge, MA. Shleifer, Andrei, 2004, “Does Competition Destroy Ethical Behavior?” American Economic Review, Vol. 94, No.2, (May), p. 414 - 418. Smith, Adam, 1776, An Inquiry into the Nature and Causes of the Wealth of Nations, repr inted by Clarendon Press, 1996. _____________, 1759, The Theory of Moral Sentiments, edited by D.D. Raphael and A.L. Macfie, Liberty Fund, Indianapolis (1982). Sullivan, Roger, 1989, Immanuel Kant’s Moral Theory, Cambridge University Press, Cambridge, UK. _____________, 1994, 1997, An Introduction to Kant’s Ethics, Cambridge University Press. White, Mark D. 2008, “Adam Smith and Immanuel Kant: On Markets, Duties, and Moral Sentiments,” http:// ssrn.com/abstract=1318605. _____________ 2011, Kantian Ethics and Economics, Standford University Press, Standford, CA. ENDNOTES 1 Kant (1724 – 1804) wrote at the dawn of the democratic age so his political philosophy was not entirely developed. John Rawls is Kant’s primary modern extender. Rawls (2001) is a more complete presentation of this democratic discourse. 2 By “rational” Kant meant logical, i.e., not self contradictory, but with logical deductions after proper reflection. 3 See White (2011, p.23) for a review of this “common person” analysis. 4 Below, we use Sullivan’s (1994, p. 29) interpretations from the original German language of these imperatives 5 This method of reference to Kant is standard and generally used in the philosophy literature. 6 In Kant’s analysis, this non-impingement of the freedom of others to pursue their own ends is termed the universal principle of justice. 7 Coercion is defined as an application of some type of force that prevents someone from making a decision or acting upon some decision. Proceedings of the 2012 Pennsylvania Economic Association Conference 231 THE ECONOMIC IMPACT OF ALVERNIA UNIVERSITY Tufan Tiglioglu Ph.D. Program Alvernia University Reading, PA, 19611 Lisa Cooper, Bari Dzomba, Rachel Gifford, and Joseph Hess Doctoral Students in Leadership Alvernia University Reading, PA, 19611 ABSTRACT Research has shown that economic impact of colleges and universities on local, regional, and state economies are greater than their direct spending. This paper examines the direct, indirect, and induced economic impact of Alvernia University on the local and regional economies. The university’s total spending on goods, services, payroll expenditures, capital investment, ancillary spending by students, visitors, and employees are analyzed using the IMPLAN input output model quantifying the direct, indirect, and induced economic impact on Berks County, Pennsylvania. Qualitative measures such as employee and student volunteer hours, institution sponsored sporting, and community events are analyzed. This paper supports Alvernia University’s impact on the local and regional economic sustainability and growth. INTRODUCTION Economic impact analysis has been used to analyze the direct, indirect, and induced economic impact that universities and colleges have on the local economy where they are located. Direct effects include transactions directly attributable to the university, such as employee salaries. Indirect and induced effects are those that result from associated university spending, for example, the university purchasing food at a local restaurant and that restaurant hiring additional staff to accommodate purchasing demand. In addition to these quantitative measurements, qualitative measures such as employee and student volunteer hours, as well as institution sponsored events including lectures, concerts, and sporting events should also be considered for a more comprehensive analysis. In a report released in 2009 by the Association of Independent Colleges and Universities of Pennsylvania, it was estimated that the total state wide economic impact of independent higher education institutes was $16,133 million. Of this, $8,259 million was estimated to have come from institutional expenditures, $955 million from construction, $1,814 million from students spending, $5,022 million from faculty and staff spending, and an additional $83 million from visitor spending (Association, 2009). In 2011, the economic impact of the University of Scranton on the city of Scranton and Northeastern Pennsylvania was determined to be $382 million overall, with over $182 direct expenditures coming from the university alone (The University of Scranton, 2011). In another study, Bucknell University was found to have impacted the surrounding six county region of Union, Columbia, Lycoming, Montour, Northumberland, and Snyder by $206 million and the state of Pennsylvania by $263 million during the 2009-2010 academic year (Rousu, 2011). The university spent over $164 million in wages, construction costs, and operating expenses. An economic impact analysis conducted in 2005 indicates that the University of Pennsylvania, which is comprised of 12 graduate and professional schools, provided a $9.6 billion direct and indirect impact on the state of Pennsylvania, as well as a $9.8 billion impact on the 11 county region, and $6.5 billion for the city of Philadelphia (Econsult Corporation, 2006). Similarly, an economic analysis conducted for Penn State University’s 24 campuses on each of Pennsylvania’s counties was conducted in 2008, which reports a total direct and indirect impact of $8.46 billion (Tripp Umbach, 2009). The total economic impact of Dickinson College for Cumberland County in Pennsylvania equals an estimated $150,431,937 (Bellinger at al., 2010). Lastly, Hope College contributed a total impact of $213 million on the city of Holland, Michigan, as reported in the 2011 economic impact study by Stokes (2011). The purpose of this paper is to analyze the direct, indirect, and induced economic impact of Alvernia University on Berks County, Pennsylvania. According to the 2010 U.S. Census, the largest city located in Berks County is Reading, which is also the poorest city in the U.S., with the median household income at $28,197 annually. In 2010, the population of Berks County was 411,442 people, with 88,082 individuals residing in Reading (QuickFacts, 2012). Additional demographic and economic information on Berks County, obtained from the IMPLAN modeling software for the model year 2009, is provided in the Appendix in Table 1. Proceedings of the 2012 Pennsylvania Economic Association Conference 232 In an effort to improve economic competiveness, eight principal stakeholder organizations involved in the promotion of the economic prosperity of the Greater Reading Area and Berks County, analyzed past economic development strategies and recommendations and identified five key issues needing attention: entrepreneurship and innovation, workforce/talent development, industry clustering, sites and infrastructure, and quality of place. By developing these areas, it is hoped that residents would be more prosperous and happier to live in the area, and local businesses would have more opportunity, innovation, and would grow faster (Ride to Prosperity, 2010). Local colleges and universities have the opportunity to support and improve upon these key areas, and Alvernia University is committed to doing so through integrated, community-based, inclusive, and ethical learning. This paper is organized as follows. The first section provides a review of the literature of economic impact analysis methods. The second section provides a description of the methodologies utilized for determining the direct, indirect, and induced impact of Alvernia University on Berks County, Pennsylvania. The third section provides the results of the analysis. The fourth section provides a discussion of the analysis results as well as conclusions. LITERATURE REVIEW Communities possess a solid sense of place and a university's presence within an area serves as a central hub for providing quality of place for the citizens within the communities it serves. From arts and cultural events, to sporting events and lecture series, a university's identity enhances the well-being of the community. The most trusted way for universities to determine the impact of their presence on the local community is to conduct an economic impact analysis. These studies are usually designed as unique research projects to meet local and regional needs (Caffrey & Isaacs 1979). Economic impact analyses are a critical tool for universities to demonstrate the value of their presence in a particular geographic area on local businesses and policies. Since the introduction of university economic impact studies by Caffrey and Isaacs in 1971, a number of studies have been conducted using various methods to quantify the economic benefits of a university in a particular area. The most common methods employed in impact studies include, the American Counsel of Education (ACE) method, the Ryan Short-Cut method, the Input Output method or IMPLAN model, and the skills based approach. The ACE method, developed by Caffrey and Isaacs (1971) is the oldest method of calculating economic impact and is still used extensively by universities today including Bucknell (Rousu, 2011) and Penn State (Tripp Umbach, 2009). The ACE method consists of a series of models created by a review of economic impact previously conducted by several American colleges and universities. The primary objective of their study was to derive models for which data could be obtained from the normal records kept by universities, local governments, and businesses; from statistical publications of the federal government. The ACE method tracks the flow of money from the university through a designated area or community to show the ways in which funds derived from the university benefit the local community (Tripp Umbach, 2009). These models consisted of business models, government models, and individual models (Caffrey & Isaacs, 1971). Specifically, the ACE method examines a university’s impact on local businesses, local individuals, and local government. A university’s impact on local business is measured by the amount of money a university spends on supplies and services. A university’s impact on individuals is assessed by how many individuals are employed by the university, and how those individuals affect the local economy by producing goods and services. The impact on local government is measured by examining how much the university contributes to local government in the form of various taxes and revenue derived from the university (Garrido-Yserte & Gallo-Rivera, 2010). The ACE methodology gathers specific cash flow data from the university’s financial documentation, which is needed to assess the impact on each of the three areas outlined above. Financial documents generally used include tax and income statements, as well as specific documentation related to purchases within the specified geographic area. The ACE method also employs surveys to measure the spending of university employees and students within the local community. Once the data has been gathered a regional economic multiplier is applied to the spending data in order to determine the total economic impact (Tripp Umbach, 2009). The models are limited to use in estimation of shortterm economic impact, i.e. over a given period of time, such as an academic school year. They are not able to assess the ultimate economic impact of the university upon the community and they do not embody considerations such as what a community might have been like without the college (Caffrey & Iassacs, 1979). It is necessary to note that the ACE method provides a built-in understatement: the actual economic impacts are probably greater than the models suggest. Caffrey and Isaacs (1979) reasoned that it is better to err on the side of too little than too much, particularly when a public relations function is being served and it is impractical to account for all the real expenditures of every individual and group associated with the university. The Ryan Short-Cut method, developed in 1981 by J.G. Ryan, is a simplified version of the ACE method. According to Ryan, there are three major problems with the ACE method when employing it for smaller universities and community colleges. First, the economic estimates suggested by the ACE method are inappropriate for small schools. Second, the development and administration of surveys to measure student and employee spending are too time Proceedings of the 2012 Pennsylvania Economic Association Conference 233 consuming and expensive for small colleges and the response rate is often too low to obtain accurate results. Finally, the retail gravity method used by the ACE model, which measures individual non-housing expenditures, has been found to be too mathematically complex. Additionally, it is difficult for small universities and colleges to operationalize a retail sales area and obtain appropriate retail sales information. The Ryan Short-Cut method remedies these problems by replacing survey information with readily available, nationally produced datasets identified for substitution of the retail gravity model. The Ryan Short-Cut method also uses a more conservative multiplier which is more appropriate for smaller institutions (Ryan & Malgieri, 1992). The Input Output method or IMPLAN model is similar to the ACE and Ryan Short-Cut methods, in that it measures the impact of university spending on a number of different industries. The major difference in the IMPLAN model is that it measures the forward and backward linkages in an economy rather than linear cash flow. Forward and backward linkages are examined through inter and intra industry transactions by looking at the direct, indirect, and induced effects of university spending (Caroll & Smith, 2006). The model measures the total annual economic activity that results from inter and intra industry transactions. As mentioned previously, direct effects measure university spending on items such as payroll. Indirect effects measure university spending on supplies and contracts with local vendors, and induced effects measure individual household spending as a result of university payroll. The IMPLAN model uses extensive documentation from the university in the form of financial statements and questionnaires given to students and staff to measure the direct provision of the university by providing jobs and purchasing goods and services for production. The model also quantifies indirect impacts of the university in the form of employee spending as a result of the income they earn from the university. Once accurate financial information is compiled, it is analyzed through the use of IMPLAN commercial software that utilizes input output data for over 500 industries to create industry specific multipliers for states and communities. The data used by IMPLAN comes from federal government sources and is used to create appropriate multipliers to estimate economic impact (Morgan, 2010). IMPLAN has been widely used as a method for economic impact assessment by universities such as the University of Pennsylvania (Econsult Corporation, 2006), Bowling Green State University (Carroll, 2006), the University of Phoenix (CBRE Consulting, 2009), and the University of North Carolina (Walden, 2009). The popularity of the IMPLAN model is due to its ease of use and accuracy in creating multipliers compared to other economic impact analysis methods. Other economic impact methods, such as ACE, involve complex calculations to determine multipliers which may result in multipliers that are too large, thus overestimating the impact of a university on a particular area. The benchmarked economic models and multipliers used by the IMPLAN software allow for greater accuracy in calculating impact and reduce the chance of error (Morgan, 2010). The outputs could result in multipliers that are too large or small thus over or underestimating the economic impact. Though the IMPLAN model has been found to be one of the most accurate models in assessing quantitative economic impact data, it lacks the capability to assess the qualitative impacts universities often have on their communities. Scoble, Dickson, Hanney, and Rodgers (2010) suggest that the inclusion of qualitative measures, such as volunteer hours and social programs can be a valuable addition to any economic impact report. The use of both qualitative and quantitative data provides a thorough understanding of all of the ways a university impacts a community. The skills based approach developed by Elliott, Levin, and Meisel (1988), was created to measure some of the qualitative impacts of a university in order to compliment the quantitative data utilized by the methods previously described. The skills based approach assumes that, in addition to providing employment and increased industry spending to an area, universities are also providing a community with a more educated population. It is reasoned that the highly educated students they produce have greater skill sets than non-educated community members and will therefore earn more substantial incomes, thus contributing more taxes to the local government. In addition to spending more in taxes, university alumni are also expected to spend more money in the local economy through the purchase of goods and services which provides further stimulation to other local businesses (Brown & Heaney, 1997). The weakness of the skills based approach, which has led to much criticism, is that it often overestimates the impact of alumni on the local community. There is often little data available to provide an accurate estimate of the true impact as a result of alumni. Additionally, research suggests that the attainment of a college degree increases the chance of migration of college students, making them less likely to stay in the area where they gained their education. Researchers are not able to accurately predict the rate of migration, resulting in the skills based approach being viewed as a highly subjective method of calculating impact (Brown & Heaney, 1997). The skills based approach is not the only method that has received criticism relating to the reliably and truth of statistics. Siegfried, Sanderson, and McHenry (2008), claim that almost all methods of calculating economic impact are in some way inaccurate since they fail to account for counterfactual information in their analysis. The authors define the counterfactuals as the businesses, industries, or Proceedings of the 2012 Pennsylvania Economic Association Conference 234 people which may have been present in a community if the university never existed. By determining what the community’s economy would look like if a university did not exist in the defined area, researchers are able to more accurately isolate the specific economic benefits that only the university can and does provide to the community (Siegfried, Sanderson, & McHenry, 2008). Based on the information provided in the literature on the various methods of determining local economic impact by universities, the Input Output method was determined to be most appropriate to assess Alvernia University’s impact on Berks County. Alvernia University resides inside the city boundaries of Reading, Pennsylvania, an economic area which is recovering at a pace faster than the rest of the Commonwealth of Pennsylvania and nation. While Reading lacks employment diversity and high-value-added service positions, it is an attractive area to focus on with strengths such as below-average cost of living, lower than average foreclosure rates, and affordable housing and business costs. Prior to the recession, the unemployment rate in the Reading area was under 4%. As of December 2011, the unemployment rate in the Reading area was 7.9%, well under the national average of 8.5% (Moody's, 2012). Loss of manufacturing jobs are the root cause for the decline in employment rates in the area, thus showing that a local university within Reading that is well established and focused on the constituents within the community could provide economic and quality of life opportunities for the residents of Berks County, Pennsylvania. METHODOLOGY The most widely used and accepted methodology for measuring the economic impacts of a university on a local economy is the Input Output method, which is used to describe economic transactions between various sectors in a defined economy within a given time period (Cooperatives, 2012). The IMPLAN model was utilized for this analysis of Alvernia University’s economic impact on Berks County, Pennsylvania during the 2010-2011 academic year. IMPLAN economic modeling software utilizes multiplier models to estimate the direct, indirect, and induced effects of the university on the local area. Direct effects measured by the model include transactions directly attributable to the university, such as employee salaries. Indirect effects assessed include transactions made to the local area by the university on items such as supplies, services, and labor. Induced effects assessed are the re-spending that occurs as a result of the indirect spending. The average of total liabilities and net assets reported in Alvernia University’s publically available 2009 Form 990, approximately $90 million, was inferred as 2010-2011 academic year total operating costs. Additionally, approximately $11 million in construction costs was reported in 2009 and was also inferred for 2010. These data were analyzed by the IMPLAN software, using the ShannonWeaver Index of 0.7992 and Industry 392, private junior colleges, colleges, universities, and professional schools, to determine total direct, indirect, and induced effects, as well as direct, indirect, and induced effects without construction costs. These effects were also analyzed for construction costs alone. In 2011, 2450 students attended Alvernia University’s main campus, with 839 students living on campus. We assume, therefore, that 1,611 students lived off campus, of which approximately 75%, or 1208 students, commute to the university approximately 150 days of the year. We estimate that each of these commuting students spend $11 each day they commute; $2 for beverages, $1 for candy, $3 for gas, and $5 for food. Therefore, $2 million was used to assess the direct, indirect, and induced effects of student spending. Finally, the impact of student volunteer hours during the 2010-2011 academic year was quantified by multiplying the number of hours volunteered by the 2010 Pennsylvania minimum wage rate of $7.25 per hour. RESULTS Total operating expenses of Alvernia University for the 2010-2011 was approximately $90 million. This total amount includes $23 million spent by the university on employee salaries and wages, of which only 80% was likely retained in Berks County after accounting for taxes and other deductions, as well as the small percentage of employees located at satellite offices outside of Berks County. Therefore, the resulting salary/wage estimate is $16 million. The total benefits paid during the 2010-2011 academic year were approximately $6 million. Though Alvernia provides several types of benefits, key benefits included current employee and retirement medical benefits. In addition to paying significant amounts of money toward salaries and benefits which were used by employees throughout the region, the university also had a significant impact on the community in the form of local taxes, amounting to $158 thousand during the 2010-2011 academic year. Additionally, the 2010-2011 total expenditures includes $4 million spent on Aladdin, the food service provider for the campus, as well as other key areas of spending including repairs and maintenance, consulting/professional fees, insurance, library books and materials, equipment and technology, travel and entertainment, utilities, and postage and shipping. Specifically, a total of $549 thousand was spent by Alvernia University on repairs and maintenance. To account for purchase of materials outside of Berks County, only 90% of that total was used. Alvernia University spent $421 thousand in consulting and professional fees. We assume that half (50%) of these services were fulfilled locally. The university also paid $232 thousand in insurance Proceedings of the 2012 Pennsylvania Economic Association Conference 235 during the 2010-2011 academic year, of which only 10% is assumed to have been retained in Berks County. Approximately $106 thousand was spent on library books and materials and $277 thousand was spent on equipment and technology. Only 10% of each is assumed to have been retained in Berks County. During the 2010-2011 academic year, $334 thousand was spent on travel and entertainment expenses of which 35% is assumed to have been spent at local businesses such as hotels and restaurants. Over $1.7 million was spent on utilities of which 20% is assumed to have been retained in Berks County. Finally, approximately $252 thousand was paid by Alvernia University in postage and shipping, 80% of which was likely retained in Berks County. These key areas of spending are outlined in the Appendix, Table 2. million from indirect effects, and $4.5 million from induced effects (see Table 7). The top ten industries impacted by employment are presented in Table 8, and include construction of new nonresidential commercial and health care structures (83 employees), food services and drinking places (5 employees), architectural, engineering, and related services (5 employees), as well as an additional 16 employees from other industries. Tables 8 also illustrates the top ten industries impacted financially by construction. Financially impacted industries include the construction of new nonresidential commercial and health care structures ($11 million), food services and drinking places ($280 thousand), architectural, engineering, and related services ($714 thousand), as well as approximately $1.6 million from other industries. Federal and state grants obtained for local spending, equaling approximately $1 million, is also included in the total expenditures. Alvernia University also provided $83 thousand in local donations. Finally, over the last few years, Alvernia University has been updating existing structures, as well as rapidly expanding its campus. The 2009 Form 990 lists construction costs at $11 million. The impact of student spending in Berks County was also assessed, with the total effect $2.9 million. Direct effects seen were $1.9 million, indirect effects were $603 thousand, and induced effects were $430 thousand. These results are provided in the Appendix, Table 9. As presented in Table 10, the top ten industries impacted financially by student spending included food services and drinking places ($932 thousand), petroleum refineries ($541 thousand), soft drink and ice manufacturing ($226 thousand), and an additional $575 thousand from other industries. Table 10 also lists the top ten industries impacted by employment by student spending, including food services and drinking places (17 employees), wholesale trade business (0.5 employees), real estate establishments (0.4 employees), with an additional 1.6 employees from other industries. For this study, the impact of annual operating expenditures was examined using a total operating budget of $90 million as described above. The resulting total effect was $127 million, with $72 million resulting from direct effects, $25.5 million from indirect effects, and $29 million from induced effects (see Table 3). The top ten industries impacted financially by the university are outlined in Table 4, and includes private junior colleges, colleges, universities, and professional schools ($72 million), real estate establishments ($5.5 million), imputed rental activity for own-occupied dwellings ($4 million), and other industries (total $15 million). Approximately $11,000,000 of the total operating budget was spent on construction. It is important to differentiate operations and construction in economic analysis because operations spending is an annual reoccurrence while construction spending is only made for the duration of the project. Therefore, the impact of university expenditures excluding construction spending (capital investment) was analyzed and the total effect was approximately $111 million (Table 5). The top ten industries impacted financially by the university are outlined in Table 6, and includes private junior colleges, colleges, universities, and professional schools ($63 million), real estate establishments ($4.8 million), other state and local government enterprises ($3.1 million), wholesale trade business ($2.1 million), private hospitals ($1.9 million), food services and drinking places ($1.8 million), and other industries (total $34 million). The total effect of construction was approximately $18 million, with $11 million resulting from direct effects, $2.6 Finally, during the 2010-2011 academic year, over 1200 students contributed 20,623 hours of service to the local community, much of which was through the Holleran Center for Community Engagement at Alvernia University. In order to provide a quantified metric for impact, the volunteer hours are multiplied by the minimum wage rate in 2010, $7.25 per hour. Therefore, Alvernia University’s students provided approximately $150 thousand in free labor to the local community during the 2010-2011 academic year. CONCLUSION This study indicates that Alvernia University has a major economic impact on the local economy of Berks County. The study focused the impact of total operating expenditures, operating expenditures without construction costs, construction costs alone, student spending, and student volunteer hours. The full extent of Alvernia University’s economic impact cannot be realized by analyzing only these areas. However, they do offer insight into role Alvernia University plays in ensuring continued and future economic success of Reading and Berks County through existing and new opportunities for purchasing and employment. Proceedings of the 2012 Pennsylvania Economic Association Conference 236 Alvernia University’s impact on Reading and Berks County is significant, with a total effect of $130 million (effects from total operating expenditures and student spending). This is approximately half of that seen from similar, small universities such as Bucknell ($206 million) and Scranton University ($382 million). It is unclear why such a large difference is seen, however, it is likely that differences in assumptions as well as validity of data sources may play a role. Future studies analyzing Alvernia University’s impact should include other variables such as employee and visitor spending. Few studies include an analysis of the impact of volunteer hours, yet the impact from this alone is significant to a community. Quantifying volunteer hours by multiplying hours by minimum wage only touches on the true impact that such generosity generates. It is even more difficult to quantify the emotional, educational, and spiritual returns such volunteers may provide. Alvernia University was named in the President’s Higher Education Community Service Honor Roll, with Distinction, by the Corporation for National and Community Service (CNCS) and the U.S. Department of Education. The honor recognizes the university’s commitment to the local community through community service. Additionally, Alvernia University sponsored numerous cultural, sporting, and educational events open to the community. university also draws hundreds of students and their families to the area on a regular basis. Each day these students spend money on food, gas, and other items generating over $2 million dollars to the local economy. This indirect effect spending no doubt has created induced effects to the economy in the form of more jobs to produce the food and other services that are demanded by commuting students. Alvernia University is recognized as one of the leading Catholic colleges in the South East region of Pennsylvania. In addition, the university is recognized for their commitment to service learning and community engagement through the Holleran Center for Community Engagement. This study has added to the available literature stating that colleges and universities provide a significant impact on the local economies where they are located, and Alvernia University provides similar benefit. While Alvernia University’s impact is great, to more accurately portray its impact, counterfactual should also be considered. Although it is impossible to know what may have existed on the land Alvernia University currently occupies, it is likely that the land may have been used for housing or other small businesses, similar to what exists on the property surrounding the university. Any alternative options short of a large corporation would be unlikely to produce the magnitude of economic impact that Alvernia University provides to the community. As an institute of higher education, Alvernia University offers a variety of unique benefits to the community. Specifically, it is increasing entrepreneurship and innovation, workforce/talent development, infrastructure, and quality of place, by not only providing well-paying jobs to an otherwise economically depressed area, but also by preparing the next generation for the workforce. In addition to Alvernia University’s ability to provide quality employment to area residents as well as workforce development, the university is able to draw hundreds of people to the area that may not have been attracted otherwise. The university has brought numerous highly skilled professors from around the country and around the world to Berks County, who regularly use their talents and skills to benefit the community. These talented professors often conduct valuable research for community organizations securing grants and other funding to improve the area. The Proceedings of the 2012 Pennsylvania Economic Association Conference 237 APPENDIX Table 1 Berks County Economic Information for 2009 Model Year GRP $15,521,300,942 Total Personal $14,457,4300,000 Income Total Employment 207,069 Number of Industries 300 Land Area (Sq. 859 Miles) Area Count 55 Population 407,125 Total Households 162,503 Average Household $88,967 Income Description Primary battery manufacturing Iron and steel mills and ferroalloy manufacturing Imputed rental activity for owner-occupied dwellings Wholesale trade businesses Private hospitals Monetary authorities and depository credit intermediation activities Management of companies and enterprises Natural gas distribution Offices of physicians, dentists, and other health practitioners * Employment and payroll only (state & local govt, education) Employee Compensation Proprietor Income $8,683,467,143 $961,653,961 Other Property Type Income Indirect Business Taxes $4,802,555,156 $1,073,624,682 Final Demand State/Local Government Federal Government $2,208,570,383 $253,421,577 Capital Exports Imports Institutional Sales Top 10 Industries Employment 4,067 1,822 7,034 7,206 3,024 3,555 525 5,902 11,496 $1,852,405,232 $14,565,006,983 -$14,095,477,693 -$1,229,421,850 Labor Income $248,671,200 Output $2,359,493,000 $167,960,600 $2,030,181,000 $0 $483,771,300 $446,926,800 $1,393,623,000 $1,267,336,000 $952,643,600 $205,105,900 $432,738,200 $55,859,720 $807,553,900 $796,593,900 $768,677,600 $444,181,100 $748,020,100 $607,348,500 $689,956,600 Table 2 Other Key Areas of Spending During the 2010-2011 Academic Year Expense Total Amount Repairs and maintenance Consulting/professional fees Insurance Library books and materials Equipment and technology Travel and entertainment Utilities Postage and shipping Total $548,871 $421,407 $232,190 $106,290 $276,761 $334,211 $1,770,449 $252,245 $3,942,424 Amount Spent in Berks County $493,984 $210,704 $23,219 $10,629 $27,676 $116,974 $354,090 $20,180 $1,257,456 Proceedings of the 2012 Pennsylvania Economic Association Conference 238 Table 3 Direct, Indirect and Induced Effects of Total Operating Costs Impact Summary Employment Direct Effect 928.2 Indirect Effect 161.4 Induced Effect 265.4 Total Effect 1,355 Labor Income $31,915,374 $6,713,608 $10,014,634 $48,643,616 Effect $72,000,000 $25,526,114 $29,365,425 $126,891,540 Table 4 Top 10 Industries Impacted Financially by Total Operating Costs Industry Employment Labor Income Private junior colleges, colleges, 931.8 $32,032,366 universities, and professional schools Real estate establishments 46.5 $555,193 Imputed rental activity for owner0 $0 occupied dwellings Other state and local government 14 $844,005 enterprises Natural gas distribution 2.3 $244,208 Wholesale trade business 13.8 $969,041 Private hospitals 16 $1,010,482 Offices of physicians, dentists, and 15.8 $1,213,906 other health practitioners Food services and drinking places 39.7 $711,032 Animal (except poultry) slaughtering, 3.2 $167,792 rendering, and processing Total Effect $72,263,932 $5,558,677 $4,325,609 $3,607,720 $3,409,117 $2,448,590 $2,179,931 $2,076,665 $2,073,726 $1,474,173 Table 5 Direct, Indirect and Induced Effects of Total Operating Costs without Construction Spending Impact Summary Employment Labor Income Direct Effect 878.9 $28,014,606 Indirect Effect 141.7 $5,893,056 Induced Effect 233.0 $8,790,623 Total Effect 1,253.5 $42,698,285 Effect $63,200,001 $22,406,255 $25,776,317 $111,382,573 Proceedings of the 2012 Pennsylvania Economic Association Conference 239 Table 6 Top 10 Industries Impacted by Total Operating Costs without Construction Spending Industry Employment Labor Income Total Effect Private junior colleges, colleges, universities, and professional schools 882.1 $28,117,299 $63,431,674 Real estate establishments 40.9 $487,336 $4,879,283 Food services and drinking places 34.8 $624,128 $1,819,393 Private hospitals Offices of physicians, dentists, and other health practitioners Other state and local government enterprises Wholesale trade businesses Retail Nonstores - Direct and electronic sales Retail Stores - General merchandise Retail Stores - Food and beverage 14.0 $886,978 $1,913,495 13.9 $1,065,539 $1,822,850 12.3 12.1 $740,849 $850,603 $3,166,776 $2,149,318 9.9 $92,877 $324,915 9.0 8.6 $235,094 $256,379 $444,885 $489,830 Table 7 Direct, Indirect and Induced Effects of Construction Impact Summary Employment Direct Effect 83.2 Indirect Effect 17.9 Induced Effect 41.5 Total Effect 142.6 Labor Income $4,935,133 $1,053,497 $1,563,211 $7,551,841 Effect $11,000,000 $2,623,863 $4,587,999 $18,211,862 Table 8 Top 10 Industries Impacted by Employment by Construction Industry Employment Construction of new nonresidential commercial 83.2 and health care structures Food services and drinking places 5.4 Architectural, engineering, and related services 5.0 Wholesale trade businesses 2.7 Private hospitals 2.5 Offices of physicians, dentists, and other health 2.5 practitioners Real estate establishments 1.9 Retail Nonstores - Direct and electronic sales 1.9 Retail Stores - General merchandise 1.8 Retail Stores - Food and beverage 1.7 Labor Income $4,935,133 Total Effect $11,000,000 $96,000 $422,800 $192,848 $156,589 $188,503 $279,849 $713,941 $487,292 $337,812 $322,477 $23,236 $17,375 $46,882 $49,632 $232,640 $60,783 $88,718 $94,825 Proceedings of the 2012 Pennsylvania Economic Association Conference 240 Table 9 Direct, Indirect and Induced Effects of Student Spending Impact Summary Employment Direct Effect 17.3 Indirect Effect 3.6 Induced Effect 3.8 Total Effect 24.6 Labor Income $384,514 $184,755 $147,612 $716,881 Effect $1,937,551 $603,883 $430,557 $2,971,991 Table 10 Top 10 Industries Impacted by Student Spending Industry Employment Food services and drinking places 17.0 Petroleum refineries 0.1 Soft drink and ice manufacturing 0.3 Chocolate and confectionery 0.3 manufacturing from cacao beans Wholesale trade businesses 0.5 Fruit and vegetable canning, pickling, 0.2 and drying Fluid milk and butter manufacturing 0.1 Imputed rental activity for owner0.0 occupied dwellings Management of companies and 0.2 enterprises Real estate establishments 0.4 Labor Income $317,900 $9,898 $32,448 Total Effect $932,002 $540,892 $226,140 $19,651 $179,391 $37,251 $87,570 $12,176 $84,959 $5,781 $65,473 $0 $58,783 $27,990 $53,782 $4,558 $46,774 REFERENCES Association of Independent Colleges and Universities of Pennsylvania, 2009. Making an Impact: The Economic Impact of Independent Higher Education in Pennsylvania (Retrieved from www.aicup.org) Bellinger, W., Bybel, A., de Cabrol, C., Frankel, Z., Kosta, E., Laffey, T., & ... Wood, M., 2010. The Economic Impact of Dickinson College on Carlisle and Cumberland County, 2010. Retrieved from http://www.eric.ed.gov/PDFS/ED512393.pdf Brown, K. H. & Heaney M. T., 1997. A note on measuring the economic impact of institutions of higher education. Research in Higher Education, 38(2). Retrieved from EBSCOhost Caffrey, J., and Isaacs, H.H., 1979. Estimating the impact of a college or university on the local economy. Washington, DC: American Council on Education. Caffrey, J., and Isaacs, H.H., 1971. Estimating the impact of a college or university on the local economy. Washington, DC: American Council on Education. Carroll, M. C. & Smith B. W., 2006. Estimating the economic impact of universities: The case of Bowling Green State University. The Industrial Geographer, 3(2): 1-12. CBRE Consulting, 2009. University of Phoenix economic impact and benefits: California operations. Retrieved from http://cdn.assets-phoenix.net/content/dam/altcloud/doc/about_uopx/california-economic-impact-report.pdf University of Wisconsin Center for Cooperatives (2010). Implan Methodology. Retrieved from http://reic.uwcc.wisc.edu/print/book/export/html/25/ Proceedings of the 2012 Pennsylvania Economic Association Conference 241 Econsult Corporation, 2006. The University of Pennsylvania economic & fiscal report. Retrieved from http://www.upenn.edu/almanac/volumes/v52/n31/pdf_n31/Penn_Economic_Impact.pdf Elliott, D.S., & Meisel, J.B., 1988. Measuring the economic impact of institutions of higher education. Research in Higher Education, 28(1): 17-33. Garrido-Yserte, R. & Gallo-Rivera, M., 2010. The impact of the university upon local economy: three methods to estimate demand-side effects. The Annals of Regional Science, 44(1): 39-67. doi: 10.1007/s00168-00809243-xMIG IMPLAN (2012). Retrieved from http://implan.com/V4/Index.php Moody's 2012. Moody's Analytics Reading. Retrieved from http://greaterreading.com/media/docs/Reports%20and%20Publications/Moodys%20Reading%20PA%20Q1%202012.pdf Morgan, J. Q., 2010. Analyzing the benefits and costs of economic development projects. Community and Economic Development Bulletin. School of Government. The University of North Carolina at Chapel Hill (7). Paff, L. & D’Allegro, M.L., 2009, In the House! Conducting an Economic Impact Study Internally. Retreived From http://www.bk.psu.edu/Documents/Information/EI_Presentation.pdf QuickFacts, 2012. U.S. Census Bureau, State and County. Retrieved from quickfacts.census.gov Ride to Prosperity: Strategies for economic competitiveness in Greater Reading, 2010. Retrieved from http://www.docstoc.com/docs/80694786/RIDE-TO-PROSPERITY Rousu, M.C., 2011. Bucknell University: The Economic Impact of Bucknell University on the State of Pennsylvania and Six-County Region During the 2009-2010 Academic Year. Retrieved from http://www.bucknell.edu/Documents/External%20Relations/BucknellEconomicImpact2011.pdf Ryan, J.G., & Malgieri, P., 1992. Economic Impact Studies in Community Colleges: The Short Cut Method, Paper No. 48: National Council for Resource Development Resource. Scoble, R., Dickson, K., Hanney, S., & Rodgers, G.J., 2010. Institutional strategies for capturing socio-economic impact of academic research. Journal of Higher Education Policy and Management, 32(5): 499-510. doi: 10.1080/1360080X.2010.511122 Siegfried, J. J., Sanderson, A.R., & McHenry, P., 2008. The economic impact of colleges and universities. Change, 40 (2). Stokes, K. B., 2011. The Economic Impact of Hope College: A greater hope for a greater community. Retrieved from http://digital.ipcprintservices.com/publication/?i=88889&p=1 The University of Scranton Economic & Community Impact to the Greater Scranton Area, 2011. Economic & Community Impact Study. Retrieved from http://www.scranton.edu/about/community-relations/docs/2011EconomicCommunityImpactStudy.pdf Tripp Umbach, 2009. The Pennsylvania State University economic impact statement. Final Report. Retrieved from http://econimpact.psu.edu/downloads/giving_back_full.pdf Walden, M. L., 2009. Economic benefits in North Carolina of the University of North Carolina campuses. Retrieved from http://www.northcarolina.edu/nctomorrow/UNC_Economic_ Impact_-_Walden.pdf Proceedings of the 2012 Pennsylvania Economic Association Conference 242 A STUDY OF FACTORS AFFECTING THE ECONOMIC FEASIBILITY OF THE IMPLEMENTATION OF TORREFACTION TECHNOLOGY BY THE PENNSYLVANIA WOOD PELLET INDUSTRY Robert F. Brooker Dahlkemper School of Business Administration 109 University Square Gannon University Erie, PA 16541 Harry R. Diz School of Engineering and Computer Science 109 University Square Gannon University Erie, PA 16541 ABSTRACT Firms in the Pennsylvania wood pellet industry process wood byproducts into pellets that are burned for home heating purposes. Studies indicate that pellet production exhibits economies of scale, but scale is constrained by the logistics of wood byproduct availability. In this paper, an economic model is used to describe the effects of this constraint. Torrefaction technology, by allowing firms to use alternative feedstocks such as animal and food production byproducts, could allow firms to increase scale with consequent lower average pellet production costs. THE WOOD PELLET INDUSTRY AND TORREFACTION Between 2003 and 2009, the demand for wood pellets for home heating in North America increased at an annual compounded rate of 22%. During the same period, wood pellet production capacity increased at an annual compounded rate of 33%. The growth of wood pellet production capacity in excess of North American demand is attributed to the growth in exports to European countries. Total pellet production in North America in 2008 was approximately 3.5 million tons while production capacity was approximately 4.5 million tons. Spelter and Toth (2009). Annual pellet production capacity in Pennsylvania in 2009 was estimated to be one-half million tons, implying that the industry has the ability to process approximately one million tons of wood residues annually, assuming an average feedstock moisture content of 50%. Pellet production capacity in Pennsylvania is increasing rapidly, and “new plant announcements in the region are made almost weekly.” Ciolkosz and Ray (2009). The feedstock used to produce wood pellets is wood byproducts, primarily sawdust and shavings that are generated by the lumber and furniture industries. Firms that produce wood pellets are called pellet mills. The production process involves three parts: (1) drying, to reduce moisture content of the feedstock to a specified level, (2) milling, to homogenize feedstock particle size, and (3) pelletization. The last step compresses, or densifies, the processed wood feedstock into pellets of uniform size and density. Wood pellets are used in the U.S. primarily for home heating purposes, although other uses such as electric generation are growing in importance, particularly in European nations. Pellet mills employ a technology that exhibits increasing returns to scale, which implies that firms with a larger capacity that produce a larger quantity of pellets per year will have a lower average cost of production. Unfortunately, feedstock supply logistics (transportation and storage costs) constrain supply. The number and scale of sawmills that are within a cost-effective distance from a pellet mill limit available feedstock. Mani, et al (2006) and Pirraglia, et al (2010). Because of its reliance for feedstock on the production of wood byproducts, the wood pellet industry is a captive of sawmills that process timber for the lumber and furniture industries. The pellet mill is like a remora attached to a shark. If the shark doesn’t eat, neither does the remora. If sawmills doesn’t process timber, then the supply of wood byproducts is curtailed and pellet mills go hungry. Torrefaction can ameliorate this problem. Torrefaction adds an additional step to the pelletization process in which dried and milled feedstock is subjected to heat in a low-oxygen environment before being pelletized. Torrefaction chars the processed feedstock, adding value to the pellets that are produced from it by increasing energy density, reducing transportation costs, eliminating biological activity, and enhancing water resistance. In addition, and of key interest here, is that torrefaction increases the variety of potential organic feedstocks for pelletization, reducing the extent to which a pellet mill is in thrall to lumber mill production rates. An economic model of production by the lumber industry that illustrates this point is developed below. It details the Proceedings of the 2012 Pennsylvania Economic Association Conference 243 economic linkages that determine the supply of feedstock to pellet mills. The model is then generalized and used to identify industries with a similar economic structure that could supply suitable feedstock if torrefaction was implemented. The solution values are Q = 150, PQ = 2.975, L = 135, PL = 3.25, W = 15, and PW = 0.5. PL = 10 – 0.05L (1’) PW = 2 – 0.1W (2’) MC = 2.225 + 0.005Q (6’) ECONOMIC MODEL OF FEEDSTOCK SUPPLY The feedstock used to produce wood pellets, primarily sawdust and shavings, is a byproduct of the production of lumber from timber by saw mills. The economic model that describes the supply of feedstock is that of joint production where the output of wood byproducts is proportional to the rate of lumber production. The model, which assumes perfect competition, is defined as follows. The demand for lumber is (1), where L is tons per year and PL is price per ton. The demand for wood residue is (2), where W is tons per year and PW is price per ton. Total sawmill output (Q) is (3). The proportion of Q that is lumber (L/Q) is denoted α so (4) and (5) define the output of L and W in terms of Q. Market supply of Q by sawmills, which is the horizontal summation of individual firms’ marginal costs, is (6). Total revenue per unit of Q is (7), where PQ is the implicit price of Q, and (4) and (5) are used to eliminate W and L. Marginal revenue (PQ) is defined by (8) as the weighted average of the prices of lumber and wood residue based on the proportion that each constitutes in total output. Substituting (4) into (1) and (5) into (2) and then substituting the resulting relationships into (8) yields (9), which defines marginal revenue in terms of Q. Finally, the optimal value of Q is found by setting PQ = MC and solving. PL = PL(L) (1) PW = PW(W) (2) Q=L+W (3) L = αQ (4) W = (1 – α)Q (5) MC = MC(Q) (6) QPQ = LPL + WPW = αQPL + (1 – α)QPW (7) PQ = αPL(αQ) + (1 – α)PW((1 – α)Q) (9) PQ = αPL + (1 – α)PW (8) As a way to provide a baseline for comparative statics analysis, the following numerical example will be used. It assumes that that the proportion L/Q = ∝ = 0.9, demand for L and W are (1’) and (2’), and MC is (6’). Marginal revenue (PQ) in (9’) is derived from (1’) and (2’) as described above. PQ = 0.9(10 – (0.05)(0.9Q)) + (0.1)(2 – (0.1)(0.1Q)) (9’) PQ = 9.2 - 0.0415Q Figure 1 represents the solution graphically. Panel a displays the supply of output by saw mills and the implied demand for output, which is derived from the demand for lumber (panel b) and the demand for wood byproducts (panel c). Notice that, because Q is determined by the intersection of supply and the weighted average of the demand for L and for W, supply in panels b and c is perfectly inelastic. Figure 2 represents the effect of an increase in the demand for lumber (1’’). The solution values are Q = 247, PQ = 3.459, L = 222, PL = 3.895, W = 24.7, and PW = - 0.47. Notice that PW is negative. This is a refutation of the Mae West hypothesis: “Too much of a good thing is wonderful.” A negative price for wood byproducts implies that the quantity produced is such that the value of the marginal product of the quantity produced to buyers is negative. In other words, the sawmill must pay to dispose of the excess. PL = 15 – 0.05L (1’’) Figure 3 represents the effect of a decrease in the demand for lumber (1’’’). The 2008 recession, and the major decline in U.S. construction had this effect. The solution values are Q = 53, PQ = 2.491, L = 47.9, PL = 2.60, W = 5.32, and PW = 1.47. Notice that the effect on PW is far greater than the effect on PL, even though production of both decline by the same proportion. PL = 5 – 0.05L (1’’’) Figure 4 represents the effect of an increase in the demand for wood byproducts (2’’). The solution values are Q = 154, PQ = 3, L = 138.9, PL = 3.06, W = 15.43, and PW = 2.46. Notice that the primary effect of this demand change is on PW, while the change has little effect on PL. PW = 2 – 0.1W (2’’) The implication of this model is that, because the supply of feedstock to the wood pellet industry is determined primarily Proceedings of the 2012 Pennsylvania Economic Association Conference 244 by the demand for lumber and other wood products, the wood pellet industry is subject to wide fluctuations in feedstock supply and price. Further, changes in demand for feedstock elicit negligible changes in quantity supplied and relatively large changes in price. Several researchers have noted in this regard that feedstock cost is a major determinant of the feasibility of pellet production. Mani, et al (2006), Nolan, et al (2010), and Uasaf and Becker (2011). ECONOMICS OF TORREFACTION CONCLUSION The wood pellet industry is growing rapidly. The supply of feedstock is a limiting factor which prevents pellet mills from taking advantage of economies of scale in production. The implementation of torrefaction technology by pellet mills would make it possible to draw on alternative feedstocks, resulting in better opportunities for scale economies. In addition, torrefaction produces pellets that are qualitatively superior to plain wood pellets, increasing their commercial value. The model above describes why the production of wood byproducts is largely unresponsive to changes in the demand for wood pellets and why opportunities to take advantage of economies of scale are constrained by the logistics of feedstock supply. Torrefaction makes it possible to process a variety of non-wood feedstocks into pellets. It does this by charring the feedstock heat to eliminate biological activity and other negative characteristics of the feedstock. Research at Gannon University has identified and analyzed torrifiable feedstocks that are available in quantity in the Northwest Pennsylvania region. These include grape pomace, which is the residue from crushing grapes, horse manure, and waste from the production of dog food. Collectively, these sources could provide approximately 8,000 tons (dry weight) of pellet feedstock annually. Diz, et al in progress. Other research has demonstrated the feasibility of using torrefaction for processing other types of animal manure. Ro, et al (2009). The economic model presented above describes a situation in which an industry jointly produces a primary product (lumber) which is its primary source of revenue, and a byproduct that is relatively low in value. The firm maximizes profit by minimizing the quantity of byproduct produced per unit of its primary product. The byproduct may be valuable in some applications (composite wood products, pulp, wood pellets) but excess supply can drive the price below zero, implying that the firm must pay to dispose of the byproduct. The situation described by this economic model applies to industries other than the lumber industry. For example, experiments conducted at Gannon University have demonstrated that byproducts from other industries can be used in combination with torrefaction to produce pellets that have superior characteristics. Specifically, the research demonstrated that feedstocks could include byproducts from the equine industry (manure), industrial food processing (grape pomace), and animal food production (dog food). These are byproducts that are of value in small quantities (for example, as fertilizer) but require costly disposal in large quantities. Torrefaction would add economic value to these byproducts while avoiding the economic costs of disposal. Further, the increased supply of feedstock would allow pellet mills to operate at a larger scale, yielding cost savings. Proceedings of the 2012 Pennsylvania Economic Association Conference 245 Figure 1a 16 14 12 10 PQ 8 6 4 2 0 0 50 100 150 200 Q 250 300 350 400 Figure 1b 16 14 12 10 PL 8 6 4 2 0 0 50 100 150 200 250 300 350 L Figure 1c 2.5 2 1.5 PW 1 0.5 0 0 5 10 15 W 20 25 30 Proceedings of the 2012 Pennsylvania Economic Association Conference 246 Figure 2a 16 14 12 10 PQ 8 6 4 2 0 0 50 100 150 200 Q 250 300 350 400 Figure 2b 16 14 12 10 PL 8 6 4 2 0 0 50 100 150 200 250 300 350 L Figure 2c 2.5 2 1.5 1 PW 0.5 0 0 5 10 15 20 25 30 -0.5 -1 W Proceedings of the 2012 Pennsylvania Economic Association Conference 247 Figure 3a 16 14 12 10 PQ 8 6 4 2 0 0 50 100 150 200 Q 250 300 350 400 Figure 3b 16 14 12 10 PL 8 6 4 2 0 0 50 100 150 200 250 300 350 L Figure 3c 2.5 2 1.5 PW 1 0.5 0 0 5 10 15 W 20 25 30 Proceedings of the 2012 Pennsylvania Economic Association Conference 248 Figure 4a 16 14 12 10 PQ 8 6 4 2 0 0 50 100 150 200 Q 250 300 350 400 Figure 4b 16 14 12 10 PL 8 6 4 2 0 0 50 100 150 200 250 300 350 L Figure 4c 4.5 4 3.5 3 2.5 PW 2 1.5 1 0.5 0 0 5 10 15 W 20 25 30 Proceedings of the 2012 Pennsylvania Economic Association Conference 249 REFERENCES Ciolkosz, Daniel, and Charles Ray. 2009. Status and Needs of the Wood Pellet Industry in Pennsylvania 2009. The Pennsylvania State University. University Par, PA. 16802. Diz, et al. In progress. Torrefaction of Organic Wastes for the Production of Solid Fuel Pellets. Mani, S., S. Sokhansanj, X. Bi, A. Turhollow. 2006. Economics of Producing Fuel Pellets from Biomass. American Society of Agricultural and Biological Engineers. Vol 22(3): 421-426. ISSN 0883-8542. Nolan, Anthony, Kevin McDonnell, Ger J. Devlin, John P. Carroll and John Finnan. 2010. Economic Analysis of Manufacturing Costs of Pellet Production in the Republic of Ireland Using Non-Woody Biomass. The Open Renewable Energy Journal, 2010, 3, 1-11. Pirraglia, Adrian, R. Gonzalez, and D. Saloni. 2010. Techno-economical analysis of wood pellets production for U.S. manufacturers. BioResources 5(4), 2374-2390. Ro, K.S., K.BG. Cantrell, P.G. Hunt, T.F. Ducey, M.B.Vanotti, A.A. Szogi. 2009. Thermochemical conversion of livestock wastes: Carbonization of swine solids. Bioresource Technology, Volume 100 (2009) 5466-5471. Spelter, Henry and Daniel Toth. 2009. North America’s Wood Pellet Sector. United States Department of Agriculture (2009) Forest Service Forest Products Laboratory Research Paper. FPL–RP–656. Uasuf, Augusto and Gero Becker. (2011) Wood pellets production costs and energy consumption under different framework conditions in Northeast Argentina. Biomass and Bioenergy, Volume 35, Issue 3, March 2011, Pages 1357-1366. Proceedings of the 2012 Pennsylvania Economic Association Conference 250 THE COLLEGE EDUCATED AND THE PUBLIC/PRIVATE SALARY DIFFERENTIAL: CAN OCCUPATION EXPLAIN DIFFERENCES? Mary Ellen Benedict Department of Economics Bowling Green State University Bowling Green, OH 43403 Michael Bajic Department of Economics Bowling Green State University Bowling Green, OH 43403 David McClough College of Business Administration Ohio Northern University Ada, OH 45810 ABSTRACT Using the National Survey of College Graduates, this study examines the public/private salary differential. We find that when employing OLS regressions, public sector workers receive a lower average salary than their private sector counterparts. This differential remains when we divide the public sector into federal and state/local workers; however, federal workers face a relatively small differential. In addition, we find that public sector workers trade salary for job attributes, including benefits, job security, and social responsibility. INTRODUCTION Declining tax revenue resulting from the economic downturn of 2007-2009 exacerbated budget challenges facing all levels of government and rekindled interest in the well-established wage disparity between public and private sector workers. In well publicized instances, newly elected governors in Wisconsin and Ohio faced with large budget deficits pursued legislation requiring public sector employees to pay larger proportions of health and pension benefits. In addition, legislation sought to eliminate collective bargaining rights of public sector workers. Among the most adamant opponents to the legislation, teacher unions mounted effective campaigns to force a recall election (June 5, 2012) of the governor in Wisconsin and to win a ballot initiative in Ohio overturning legislation that secured passage in the Ohio State Assembly. A consequence of the contentiousness of the debate is the tendency of analysts to offer normative prescriptions informed by empirical research that appears to be, in many instances, motivated by ideology. In some sense even academic research proposing neutrality in the debate is vulnerable to charges of ideological bias based on the choice of analytical method, which very much influences the empirical results and consequently informs the policy prescriptions. The present paper employs an empirical analysis that includes the usual human capital variables while controlling for fourteen occupational categories. This research deviates from the conventional analysis by employing data that includes variables reporting the extent to which workers value certain attributes of their job and the satisfaction associated with the work attribute. Using these data, we are able to identify trade-offs associated with the wage premium regardless of whether the premium accrues to workers in the public or private sector. Lastly, this research limits the scope of the analysis to only the most educated workers, those who have at least completed a college degree. The data delineate college graduates and accommodate examination of the public/private wage disparity across a range of college and post-baccalaureate graduates. The paper will proceed with a review of the relevant literature, description of the data, specification of the model, discussion of the results, and concluding comments. LITERATURE REVIEW The observed wage disparity between private and public sector workers is well established. Early studies were motivated by the rapid growth of government employment in the second half of the Proceedings of the 2012 Pennsylvania Economic Association Conference 251 20th century (Smith, 1977; Gyourko and Tracy, 1988). Smith (1977) specified separate wage functions to determine if the underlying wage structure differed in the public sector, in part, due to the lack of sensitivity to traditional market forces. She found that public sector wages exceeded the private sector although the differential varied by the level of government and gender. Subsequent studies found that federal workers enjoyed a ten to twenty percent premium compared to the private sector (Ehrenberg and Schwarz, 1986; Gregory and Borland, 1999). In addition to employment growth in the public sector, an increase in the rate of unionization of public workers motivated research examining the effect of unions on wage disparity (Gyourko and Tracy, 1988; Edwards, 2010; Mohanty, 1994). Two distinct methodological approaches have emerged to examine the observed wage disparity between private and public sector workers. Early empirical research compared private and public sector occupations. After controlling for occupation, Hartman (1983) found that federal workers receive 10 percent less than workers in similar occupations in the private sector. Referring to data from the Bureau of Labor Statistics (BLS), Cauchon (2010) reports that the federal government workers receive higher average salaries compared to the private sector in eight out of ten occupations. Of the 216 occupations reviewed, 180 occupations paid higher salaries to federal government workers compared to 36 occupations that paid higher salaries in the private sector. Although federal workers received, on average, higher salaries, the range of salaries was narrower than the private sector. Cauchon reports that state and local government workers receive 5 percent less pay compared to the private sector; however, when using total compensation, salary and benefits, rather than salary alone as the dependent variable, state and local workers, on average, receive more than private sector workers. A second empirical approach arose from challenges that many government occupations have no private sector equivalent. Keefe (2010) observes that ideally workers performing similar work would be compared but too many public sector occupations have no comparable private sector equivalent. BLS data reveal 124 federal government occupations for which there are no direct equivalents (e.g. air traffic controllers) in the private sector (Cauchon, 2010). To address this challenge, research has compared human capital of private and public sector workers. The perceived benefit of the human capital approach is the ability to compare individuals across occupations (Smith, 1977). Illustrative of this approach using the Current Population Survey (CPS), Keefe (2010) does not control for occupation but does control for traditional human capital variables, organizational size, and personal characteristic to find that state and local government workers are slightly underpaid compared to similar private sector workers. A challenge facing the human capital approach is the observed differences in the distribution of human capital across the public and private sectors. According to a recent report from the Congressional Budget Office (2012), 52 percent of federal workers have earned at least a bachelor’s degree compared to 32 percent in the private sector. The disparity is not unique to the United States. Snowden (2012) illustrates a similar education discrepancy using data from the UK Office for National Statistics (ONS) that reports that 40 percent of government workers earned degrees compared to 25 percent in the private sector. These data suggest either that government work requires more skill or that government hires over educated workers. Snowden (2012) offers potential insight into this question when he reports that that the ONS study identifies 59 percent of public sector employees as high skill or upper-middle skill compared to 49 percent in the private sector. It appears therefore that public sector workers require more skill but that the government may nonetheless hire over educated workers given that the ratio of college educated workers to non-college educated workers exceeds the ratio of high and middle skill workers in the public and private sector. Unobserved worker heterogeneity motivated a large body of research. Gyourko and Tracy (1988) find evidence of unobserved worker heterogeneity across both the public and private sector with a negative self-selection effect in the unionized public sector. In contrast, they find a smaller positive selection effect in the non-union private sector. Lee (2004) asserts that previous research ignores worker heterogeneity and self-selection across sectors. Specifying econometric models to account for heterogeneity and selection bias, Lee finds that earlier OLS estimates of federal government wage differentials were biased upward for men and downward for women. Krueger (1988) employs longitudinal data to find that the premium received by federal workers is smaller than estimates derived from cross-sectional data. Gittleman and Pierce (2011) use multiple data sets both showing public sector workers to be more skilled as evidenced by higher average educational attainment. Using CPS data for individual Proceedings of the 2012 Pennsylvania Economic Association Conference 252 characteristics, they then estimate a compensation mark-up from BLS employer cost data. Gittleman and Pierce find that state employees cost 3-9 percent more and local government workers incur compensation cost 10-19 percent more than similarly skilled private sector workers. EMPIRICAL ANALYSIS Overview of the National Survey of College Graduates The National Survey of College Graduates (NSCG), sponsored by the National Science Foundation, is a “once in a decade opportunity” to examine the educational and career characteristics of United States college graduates. The latest version of the data, the 2003 NSCG, surveyed a random sample of individuals living in the United States, under the age of 76, who had received a bachelor’s degree or higher prior to the new millennium. The public-use sample includes data on 100,042 individuals. The NSCG includes a rich set of information on individual and job-related characteristics. Our analysis includes full-time employed individuals and eliminates those who are K-12 teachers, working in higher education and nonprofits, and those respondents who do not identify their race or if the race category included too few observations. We eliminated K-12 teachers because there were so few observations in the private sector. Respondents from academia and the nonprofit sectors are eliminated due to the unique job duties specific to the organizations in which they work. Additional deletions due to missing data reduced the sample to 50,966 observations, of which 42,137 are individuals in the private sector and 8,829 are in the public sector. Consistent with the literature on public/private differentials and salary regressions more generally, the final dataset includes information on individual characteristics (age, gender, race, US citizenship) and family characteristics (marital status, number of children). Regional controls are also included but not listed in the final tables. The human capital effect is captured with several binary variables representing the type of highest degree (a bachelor’s degree is the omitted category in the base regression) and the number of years since the individual obtained the highest degree, used as a proxy for experience. Current job tenure and employer size, defined by a set of binary variables representing the number of employees in the firm, capture the current job’s human capital development and the possibility that larger employers have the ability to pay more (Brown and Medoff, 1989). Occupation has been aggregated into fourteen occupation categories. The omitted category represents a grouping of non-science occupations, including communications, journalism, criminal justice, leisure and fitness, public affairs, and other fields not listed in the NSCG survey. Research indicates varying differentials between federal and private and state/local and private workers so we control for public sector workers, in general, as well as distinguish between federal and state/local workers. Due to the narrow focus on college-educated workers of the present research paper, we do not consider the unionization issue because unions tend to disproportionately represent less educated workers receiving lower salaries. Appendix 1 lists means and standard deviations for the variables used in the analysis for the total sample and delineated by employment sector. Note that even in this dataset, the public sector is on average more educated than the private sector. Although the percentage of individuals holding professional degrees is similar across sectors, 29 percent of public sector workers hold a Master’s degree as their highest degree compared to 25 percent in the private sector, and a higher percentage of federal sector workers have Ph.Ds. making the public sector sample more educated overall. The distribution of occupations is not dramatically different between the two sectors. Public/Private Differentials under Different Methods and Controls We first examine the private/public differential using what Gittleman and Pierce (2011) call the “people approach,” whereby OLS regressions are used to examine the salary differential having controlled for the major factors that affect salary. Table 1 provides the salary differential under different functional form assumptions for earnings and various control factors. Row 2 presents the unadjusted raw differential between the private sector and the two public sectors. The unadjusted differential between private and federal workers is 19.57 percent; the differential is 48.6 percent between private and state/local workers. The results are consistent with the literature on federal workers, where Falk (2012) found a raw differential of 23 percent for college-educated federal and private workers; however, the unadjusted differential for state and local workers is much larger than the 22 percent found with weekly wages from the CPS (Gittleman and Pierce, 2011). Because the norm is to employ log-linear regression functions and because salary decomposition is based on that functional form, we also estimate the unadjusted differential using natural logs of salary. The Proceedings of the 2012 Pennsylvania Economic Association Conference 253 differential falls to 1.9 percent for federal workers and 45.5 percent for state and local workers (Row 3). When we employ a log-linear regression to control for the usual demographics (Row 4), the estimated differential is 2.25 percent and 25.35 percent for federal and state/local workers, respectively. Thus, even though the analysis is limited to collegeeducated, the controls for race and gender, postbaccalaureate schooling, and post-education human capital development (via general and firm-specific experience) explain a good part of the differential, but the federal/private differential rises, perhaps due to larger returns to the private sector associated with some independent variables. For example, a set of regressions by employment sector indicates that private sector workers see larger returns to Ph.D. degrees compared to either federal or state workers (See Appendix 2). The next row in Table 1 controls for firm size using five binary variables representing categories of increasing number of employees. Brown and Medoff (1989) found empirical support for the notion that larger employers hire better-quality workers, although the wage-firm size premium is still largely unexplained. We include the controls for employer size to account for this possibility. Gittleman and Pierce (2011) suggest that including size controls confounds any public sector effect, but they also indicate that they do not include size controls because their particular dataset does not include the information. Using the CPS 2009, Keefe (2010) controls for size in a comparison between state and local to private sector workers and finds a 15.57 percent and a 9.46 percent differential for state and local workers, respectively. Our employer-size controls increase the public/private salary gap to 13.9 percent for federal workers and 28.19 percent for state/local workers. The expected value of earnings for private sector workers with at least a college degree in Keefe’s sample was only $59,911.66 in 2009 dollars, much less than the nearly $84,000 average salary earned by private sector workers in the 2003 NSCG ($97,696 in 2009 constant dollars). Our larger differentials compared to Keefe may be due to the reduced private sector pay as a result of the most recent recession. The last row in Table 1 includes controls for fourteen occupations (See Appendix 1 for a list of the occupations). These controls reduce the salary differential for federal workers to 6.5 percent and 18.1 percent for state/local workers. While our results do not switch the sign of the differential as occupational controls do in Gittleman and Pierce (2011), we do see that controlling for occupations reduces the differentials by 53 percent for federal workers and over 35 percent for state workers. Thus, occupational choice may affect the differential because returns to particular occupations vary across sectors. For many years, labor economists have analyzed wage and salary differentials using a decomposition method that explains the differential in terms of differences in mean levels of the independent variables, the advantage one group has over another group, and the disadvantage one group has due to external factors, such as discrimination (Oaxaca, 1973). Cotton (1988) provides the basis for the following decomposition: πππΈπππππππ ππ − πππΈπππππππ ππ = ∑ π½π∗ (ποΏ½πππ − ποΏ½πππ ) + ∑ ποΏ½πππ οΏ½π½πππ − π½π∗ οΏ½ + ∑ ποΏ½πππ οΏ½π½πππ − π½π∗ οΏ½ (1) The first term on the right-hand side of the equation captures the effect of the differences in the mean values of the independent variables between private (PR) and the public (PU) sectors. So, for example, if private sector workers have higher levels of firm experience compared to their public sector counterparts, this first term captures that difference. The second term estimates the benefit of being in the private sector, while the third term represents the disadvantage of being in the public sector. The last two components present the degree to which earnings are overestimated in the private sector or underestimated in the public sector. π½* is a proportional average of the coefficients, based on each group’s sample size (i.e. β*= βPR *nprivate/n + βPU *npublic/n, where n is the total number of individuals in the sample). β* is considered the average return for a particular control factor in the regression. Table 1 presents the decomposition under the natural log differential. The estimated federal/private differential is only 1.9 percent using log salaries. The small differential occurs because private sector workers have relatively higher proportions in key independent variables, such as the percentage who are doctors, in computer science or math, or the life sciences, where the returns in the private sector are relatively higher. The differences in independent variables and the returns to the private sector explain 70 percent and 30 percent, respectively, of the differential. There is no estimated disadvantage to the federal sector. On the other hand, the state and local/private salary differential is quite large. There is little advantage to Proceedings of the 2012 Pennsylvania Economic Association Conference 254 working in the private sector (explains about 9 percent of the differential), and the difference between state/local and private workers is largely due to differences in the average values for the independent variables (23 percent) and for the disadvantage facing state and local workers in returns (67 percent). The former is mainly due to employer size (where private sector workers have 21 percent of workers in firms with at least 25,000 employees, while state and local workers only have 13 percent) and with the occupational distribution, where there are more private sector individuals in occupations such as engineering, computer science and mathematics, and health practitioners (doctors). The disadvantage to state and local workers is due to the relatively lower returns state and local workers receive for the independent variables. For example, lawyers in the private sector have an average salary that is 77 percent greater than those individuals in the omitted occupation category (the non-science jobs listed earlier), while state and local lawyers receive only a 25 percent return. Likewise, doctors and computer scientists and mathematicians in the private sector receive a higher return to their respective occupations compared to their state and local counterparts (See Appendix 2 for the related regression coefficients). Thus, it appears that we can explain most of the differential for each public sector. Two final methods are used to examine the differential more closely. The first method examines the pay differential by deciles and/or quintiles, to see whether the differential changes when average pay is relatively low or high. In other words, can we explain the overall differential because public workers are paid less than their private sector counterparts within a particular part of the income distribution? Table 2 presents the average salary and related OLS coefficients for several different percentiles of the salary distribution (using the regression with all independent variables). The results in Table 2 indicate that the federal/private differential is large for the lowest percentile, but federal workers have relatively higher salaries in the 25th percentile. The federal advantage disappears in higher percentiles, but the private sector advantage is very small. On the other hand, the state and local/private differential favors state and local workers in the lowest percentile, indicating that state workers actually have a salary premium in those jobs with relatively low pay. The differential becomes negative but is never more than 2.8 percent (75th percentile). These results indicate that withindistribution salary differences are small, but given the results in Table 1, are large overall. Thus, public sector workers are more likely to be in the lower percentiles relative to the private sector workers, creating a wide differential across the income distribution. A second method is used when the researcher assumes that individuals select into their respective sectors. Here, the model is assumed as follows (Greene 2012): π¦π = π½ ′ π₯π + πΏπΈπ + ππ (2) πΏππ‘ πΈπ∗ = πΎ′π + π’π (3) πΈ[π¦π |πΈπ = 1] = π½ ′ π + πΏ + πππ π(−πΎ’Z) (4) Where log salary is a function of π₯ independent variables and πΈ is a binary variable indicating whether the observation is from the private or public sector. However, we may find that the binary variable does not capture choices made regarding sector, therefore: where πΈπ∗ is the probability of choosing to work either in the public or private sector (a standard normal distribution is assumed for the probability). We do not observe the probability, only the outcome, πΈπ . If one employs OLS regression with πΈ on the righthand-side of the salary equation, then δ in equation (2) will be biased and inconsistent, overestimating the impact of the differential. In other words, the coefficients estimated in Table 1 are too negative because they do not account for the choice one makes in terms of employment sector (which is represented by the correlation between the error term in (2) with the error term in the salary regression). The related salary regression is therefore: We now must include the last term on the right-handside to account for the sector choice, represented by the inverse Mills ratio and estimate the relevant coefficient, λ. Why might we consider one choosing an employment sector? There are two obvious reasons. First, a negative coefficient associated with a public sector binary variable becomes less negative or switches signs when the regression moves from salary to total compensation (Gittleman and Pierce, 2011). Thus, it seems that public sector workers tradeoff salary for other work benefits (pensions, vacations, leave). Second, we might consider a variety of job attributes, where public sector workers have different attitudes about the attribute compared to private sector workers. In our dataset, we can examine whether this is the case with a set of survey questions that inquire Proceedings of the 2012 Pennsylvania Economic Association Conference 255 about the respondent’s level of interest in several job attributes, as listed in Table 3. In particular, public sector workers are more likely to report that benefits, job security, and social responsibility are very important or somewhat important compared to private sector workers. This result is consistent with the finding that business economists trade salary for socially responsible work (Benedict et al., 2006). Because the choice is between the public versus the private sector and the differences above are small between the federal and state/local sectors, we employ a probit model to estimate the probability that an individual selects the public sector. The independent variables include binary variables representing whether an individual ranked job characteristics as important or not important. Controls for race and gender are included because minorities and women may move toward the public sector if they feel less discrimination (further, more minorities and women are situated in the public sector). The salary regression is as before, only with a binary variable representing the public sector, rather than two variables for the sub-sectors. The OLS regression in Appendix 3 indicates that the average differential between the public and private sector is -0.14 (i.e. public workers earn 14 percent less on average than their private sector counterparts). However, when we assume that the individual chooses the employment sector, we find that the differential is reduced to 8.1percent, which suggests that public sector workers tradeoff salary for job attributes by going into the public sector. The coefficient on the inverse Mills ratio is statistically significant and negative, indicating that sector choice is indeed important to the model. These results indicate that a compensating differential exists for public sector workers, in that they are willing to receive relatively lower salaries for job characteristics that are important to them. For example, in the probit regression, caring about benefits, job security, and social responsibility are associated with positive coefficients that are statistically significant. These are the three main job characteristics that seem to be traded off for salaries in the second-stage salary regression. selection model we include individual attitudes regarding job characteristics. We find that public sector workers trade salary for fringe benefits, job security, and social responsibility. The other addition we make to the literature is due to the unique qualities of the NSCG data. Because we can focus only on the college-educated, our results are not confounded by the fact that low-skilled, loweducated public sector workers have a wage premium, while educated workers do not. Earlier studies are unable to delineate among different postbaccalaureate degrees. We find that individuals with a Master’s degree receive a higher return in the federal government sector compared to their private sector counterparts. Further, by distinguishing between federal workers and state and local workers, we find that the differential varies by government sector. State and local workers have lower average salaries compared to their federal and private sector counterparts in all of the estimated regressions. We also find that much of the disadvantage to state/local workers is due to the lower returns they receive compared to private sector workers. This result may reinforce the notion of a compensatory differential, where public sector workers realize higher salaries exist in private sector occupations such as doctors and lawyers, but they choose the public sector because they care about different job attributes. CONCLUSION This paper examines the public/private salary differential for the college-educated. Our findings confirm previous research that controls for human capital, employer size, and occupation reduce the unadjusted differential. In addition, our self-selection model supports earlier findings that the differential is reduced after controlling for selection bias. In our Proceedings of the 2012 Pennsylvania Economic Association Conference 256 TABLE 1 – PUBLIC/PRIVATE PAY DIFFERENTIALS Private Federal State/Local $83,771.29 (65,056.23) $75,434.11 (30,131.30) -19.57% $56,383.94 (28,548.04) -48.6% 1.9% -45.5% 70% 23% % Advantage for Private Sector 30% 9% %Disadvantage for Public Sector 0% 67% 4. Demographic/Regional controls -2.25%*** -25.35%*** 5. Employer Size controls -13.90%*** -28.19%*** 6. Occupation controls -6.50%**** -18.10%*** 1. Average Salary 2. Unadjusted Differential 3. Unadjusted LnSalprivate – LnSalpublic Explained by: % Differences in Averages Data Source: 2003 National Survey of College Graduates. Regression coefficients are reported in this table. Appendix 2 contains the regression estimation. *** indicates statistical significance at the 1 percent level of significance. Rows 5 and 6 are adding additional controls to the previous row. Proceedings of the 2012 Pennsylvania Economic Association Conference 257 TABLE 2 – ESTIMATED SALARY DIFFERENTIAL BY PERCENTILE Table 2. Estimated Salary differentials by Percentiles Percentile 10 25 50 75 90 Average Salary $20,015.33 (9095.64) $38,963.97 (4159.55) $56,448.75 (5920.67) $78,986.64 (7481.93) $107,819.45 (10621.54) Federal -0.218* State 0.117** 0.032*** -0.010*** -0.009** -0.017*** -0.013*** -0.028*** -0.013*** -0.019*** TABLE 3 – ATTITUDES ABOUT JOB CHARACTERISTICS Respondent cares about: Table 3. Attitudes about Job Characteristics Very important Somewhat Somewhat important unimportant Not important at All Advancement Private 46.57% 44.18% 7.14% 2.11% Federal 48.09 45.18 5.35 1.37 State/Local 43.91 48.15 6.65 1.29 Benefits Private 63.95 32.71 2.36 0.99 Federal 69.75 28.85 1.07 0.33 State/Local 74.68 24.32 0.87 0.13 Challenge Private 63.29 33.95 2.26 0.49 Federal 64.07 33.38 2.22 0.33 State/Local 62.40 35.23 2.04 0.33 Independence Private 61.78 35.12 2.78 0.32 Federal 60.36 36.12 3.21 0.30 State/Local 61.84 36.06 1.89 0.21 Responsibility Private 46.67 46.74 5.73 0.86 Federal 45.13 48.37 5.71 0.80 State/Local 46.12 48.71 4.63 0.54 Salary Private 62.07 36.52 1.09 0.32 Federal 57.97 40.52 1.21 0.30 State/Local 58.20 40.53 1.10 0.17 Job Security Private 61.51 33.85 3.60 1.04 Federal 70.79 27.26 1.67 0.27 State/Local 73.18 24.86 1.68 0.29 Social Responsibility Private 36.96 48.50 12.03 2.52 Federal 52.35 40.98 5.74 0.93 State/Local 57.52 37.00 4.51 0.96 Data Source: 2003 National Survey of College Graduates. Probabilities are conditional on the employment sector. Proceedings of the 2012 Pennsylvania Economic Association Conference 258 Proceedings of the 2012 Pennsylvania Economic Association Conference 259 APPENDIX 1 – DESCRIPTIVE STATISTICS Variable Salary LnSalary Female Married Number of Children US Citizen Asian Black Hispanic Highest Degree-MA Highest Degree-PHD Highest DegreeProfessional Years Since Highest Degree Years Squared Current Job Tenure Current Job Tenure Squared Employer Size:100-499 Employer Size: 500-999 Employer Size: 1K-5K Employer Size: 5K-25K Employer Size: 25K+ Computer Science or Math Life Science Chemistry Physical Sciences Total Sample N=50,966 80,388.58 (60965.27) 11.084 (0.742) 0.300 (0.458) 0.784 (0.412) 0.913 (1.146) 0.923 (0.266) 0.142 (0.349) State/Local N=5,186 56,383.94 (28548.04) 10.821 (0.559) 0.445 (0.497) 0.731 (0.444) 0.801 (1.086) 0.970 (0.169) 0.099 (0.298) Federal N=3,643 75,434.11 (30131.30) 11.145 (0.488) 0.314 (0.464) 0.771 (0.420) 0.849 (1.126) 0.985 (0.120) 0.103 (0.304) Private N=42,137 83,771.29 (65056.27) 11.112 (0.773) 0.281 (0.449) 0.791 (0.407) 0.932 (1.154) 0.912 (0.284) 0.151 (0.358) 0.070 (0.255) 0.073 (0.261) 0.261 (0.439) 0.054 (0.226) 0.081 (0.272) 17.815 (10.248) 422.407 (433.502) 8.095 (8.274) 133.983 (246.630) 0.123 (0.329) 0.055 (0.228) 0.124 (0.330) 0.132 (0.338) 0.261 (0.439) 0.156 (0.363) 0.023 (0.149) 0.013 (0.114) 0.011 (0.106) 0.161 (0.367) 0.107 (0.309) 0.285 (0.452) 0.031 (0.173) 0.078 (0.269) 18.266 (9.987) 433.355 (406.891) 9.001 (8.028) 145.450 (220.818) 0.186 (0.389) 0.116 (0.321) 0.258 (0.437) 0.195 (0.396) 0.132 (0.338) 0.079 (0.270) 0.038 (0.190) 0.009 (0.096) 0.018 (0.133) 0.121 (0.327) 0.084 (0.278) 0.300 (0.458) 0.106 (0.308) 0.066 (0.249) 18.277 (10.300) 440.120 (425.929) 10.545 (9.219) 196.150 (277.582) n/a 0.054 (0.226) 0.068 (0.252) 0.254 (0.435) 0.052 (0.223) 0.082 (0.275) 17.720 (10.273) 419.529 (437.261) 7.772 (8.173) 127.197 (245.999) 0.126 (0.322) 0.052 (0.223) 0.118 (0.323) 0.135 (0.342) 0.213 (0.409) 0.169 (0.375) 0.015 (0.123) 0.014 (0.116) 0.008 (0.090) n/a n/a n/a n/a 0.115 (0.319) 0.084 (0.278) 0.014 (0.118) 0.040 (0.196) Proceedings of the 2012 Pennsylvania Economic Association Conference 260 Social Sciences 0.017 0.030 (0.129) (0.170) Engineering 0.171 0.125 (0.376) (0.331) Health Practitioners 0.044 0.014 (0.204) (0.116) Other Health-Related 0.041 0.062 (0.198) (0.241) Science or Engineering 0.040 0.032 Technology (0.197) (0.176) Management 0.189 0.182 (0.392) (0.386) Social-related 0.020 0.129 (0.141) (0.335) Sales or Marketing 0.097 0.006 (0.296) (0.077) Art & Music 0.020 0.008 (0.141) (0.092) Law 0.032 0.056 (0.176) (0.230) Data Source: National Survey of College Graduates. 0.031 (0.174) 0.182 (0.386) 0.022 (0.147) 0.046 (0.209) 0.024 (0.152) 0.200 (0.400) 0.010 (0.098) 0.005 (0.074) 0.009 (0.095) 0.029 (0.169) Proceedings of the 2012 Pennsylvania Economic Association Conference 0.014 (0.118) 0.175 (0.380) 0.049 (0.216) 0.038 (0.191) 0.043 (0.202) 0.189 (0.392) 0.008 (0.087) 0.117 (0.321) 0.023 (0.149) 0.029 (0.169) 261 APPENDIX 2 – REGRESSION OUTPUT Variable CONSTANT Female Married Number of Children US Citizen Asian Black Hispanic Federal State/Local Highest Degree-MA Highest Degree-PHD Highest DegreeProfessional Years Since Highest Degree Years Squared Current Job Tenure Current Job Tenure Squared Employer Size:100-499 Employer Size: 500-999 Employer Size: 1K-5K Employer Size: 5K-25K Employer Size: 25K+ Computer Science or Math Life Science Chemistry Physical Sciences Total Sample 10.194*** (0.021) -0.207*** (0.007) 0.087*** (0.008) 0.033 *** (0.003) 0.041 *** (0.014) -0.068*** (0.011) -0.102*** (0.013) -0.141*** (0.013) -0.065 *** (0.014) -0.181*** (0.011) 0.162*** (0.008) 0.333 *** (0.017) 0.278 *** (0.023) 0.021 *** (0.001) -0.0004 *** (0.00003) 0.014*** (0.001) -0.0002 *** (0.00004) 0.174 *** (0.010) 0.196 *** (0.014) 0.217*** (0.011) 0.227 *** (0.010) 0.273 *** (0.009) 0.498*** (0.012) 0.275*** (0.026) 0.328*** (0.033) 0.345*** (0.035) State/Local 10.326*** (0.050) -0.125*** (0.015) 0.049*** (0.016) 0.007 (0.007) 0.077* (0.046) -0.068** (0.027) -0.047** (0.020) -0.142*** (0.025) n/a Federal 10.401*** (0.079) -0.056*** (0.017) 0.026 (0.019) 0.035*** (0.007) 0.062 (0.069) -0.091*** (0.028) -0.072** (0.024) -0.039 (0.029) n/a Private 10.124*** (0.024) -0.227*** (0.008) 0.097*** (0.009) 0.035 *** (0.003) 0.039 *** (0.015) -0.067 *** (0.012) -0.130*** (0.0176) -0.149*** (0.015) n/a n/a n/a n/a 0.165*** (0.016) 0.236*** (0.047) 0.300*** (0.045) 0.014*** (0.003) -0.0003*** (0.0001) 0.017*** (0.003) -0.0003*** (0.0001) 0.105*** (0.025) 0.173*** (0.028) 0.169*** (0.024) 0.194*** (0.025) 0.169*** (0.028) 0.240*** (0.031) -0.01 (0.043) 0.101 (0.088) 0.096 (0.060) 0.229*** (0.018) 0.322*** (0.032) 0.292*** (0.050) 0.021*** (0.003) -0.0004*** (0.0001) 0.010*** (0.003) -0.0001 (0.0001) n/a 0.153*** (0.009) 0.337 *** (0.020) 0.271 *** (0.027) 0.022 *** (0.001) -0.0005 *** (0.00003) 0.013 *** (0.001) -0.0002 *** (0.00004) 0.180*** (0.012) 0.195*** (0.017) 0.222*** (0.012) 0.227*** (0.012) 0.281*** (0.010) 0.564 *** (0.014) 0.417 *** (0.035) 0.393 *** (0.038) 0.451*** (0.047) n/a n/a n/a n/a 0.028*** (0.029) 0.031 (0.035) 0.203*** (0.029) 0.179*** (0.048) Proceedings of the 2012 Pennsylvania Economic Association Conference 262 Social Sciences 0.396*** 0.130*** 0.177*** 0.492 *** (0.028) (0.046) (0.050) (0.035) Engineering 0.479*** 0.316*** 0.314*** 0.541 *** (0.012) (0.027) (0.026) (0.014) Health Practitioners 0.918*** 0.602*** 0.303*** 1.007 *** (0.027) (0.074) (0.070) (0.031) Other Health-Related 0.285*** 0.066** 0.125*** 0.357*** (0.016) (0.030) (0.036) (0.020) Science or Engineering 0.357*** 0.097** 0.122** 0.431 *** Technology (0.018) (0.043) (0.052) (0.021) Management 0.561*** 0.284*** 0.310*** 0.640 *** (0.010) (0.020) (0.022) (0.012) Social-Related 0.073*** -0.077*** -0.037 0.016 (0.023) (0.024) (0.074) (0.041) Sales or Marketing 0.294*** 0.094 -0.094 0.355 *** (0.012) (0.084) (0.092) (0.013) Art 0.156*** 0.101 0.142* 0.213*** (0.020) (0.069) (0.076) (0.022) Law 0.648*** 0.251*** 0.333*** 0.770*** (0.027) (0.050) (0.076) (0.032) Adjusted R2 0.226 0.233 0.241 0.224 F-Statistic 345.28*** 39.41*** 33.11*** 297.38*** Data source: 2003 National Survey of College Graduates. Statistical significance: ***=0.01, **=0.05, *=0.10. Note that employer size binary variables are not included in the regression for federal workers, because the employer size is always greater than 25,000. Proceedings of the 2012 Pennsylvania Economic Association Conference 263 APPENDIX 3 – SELF-SELECTION MODEL OUTPUT Variable Constant Public Female Married Number of Children US Citizen Asian Black Hispanic Highest Degree-MA Highest Degree-PHD Highest DegreeProfessional Years Since Highest Degree Years Squared Current Job Tenure Current Job Tenure Squared Employer Size:100-499 Employer Size: 500-999 Employer Size: 1K-5K Employer Size: 5K-25K Employer Size: 25K+ Computer Science or Math Life Science Chemistry Physical Sciences Social Sciences OLS 10.190*** (0.021) -0.140*** (0.009) -0.208*** (0.007) 0.087*** (0.008) 0.033*** (0.003) 0.043*** (0.014) -0.068*** (0.011) -0.105*** (0.013) -0.141*** (0.013) 0.162*** (0.008) 0.337*** (0.017) 0.279*** (0.023) 0.021*** (0.001) -0.0004*** (0.00003) 0.014*** (0.001) -0.0002*** (0.00003) 0.170*** (0.010) 0.189*** (0.014) 0.210*** (0.011) 0.223*** (0.010) 0.292*** (0.009) 0.498*** (0.012) 0.282*** (0.026) 0.328*** (0.033) 0.352*** (0.035) 0.396*** Self-Selection 10.226*** (0.022) -0.081*** (0.020) -0.190*** (0.007) 0.073*** (0.008) 0.031*** (0.003) 0.034*** (0.012) -0.053*** (0.009) -0.116*** (0.013) -0.128*** (0.011) 0.155*** (0.007) 0.332*** (0.012) 0.287*** (0.027) 0.020*** (0.001) -0.0004*** (0.00003) 0.012*** (0.001) -0.0002*** (0.0004) 0.155*** (0.010) 0.168*** (0.013) 0.201*** (0.010) 0.216*** (0.010) 0.285*** (0.009) 0.492*** (0.012) 0.286*** (0.017) 0.316*** (0.018) 0.348*** (0.023) 0.385*** Probit Estimates -1.704*** (0.091) 0.224*** (0.014) -0.205*** (0.021) 0.564*** (0.023) 0.211*** (0.024) Proceedings of the 2012 Pennsylvania Economic Association Conference 264 Engineering Health Practitioners Other Health-Related Science or Engineering Technology Management Social-related Sales or Marketing Art & Music Law (0.028) 0.479*** (0.012) 0.920*** (0.027) 0.286*** (0.016) 0.357*** (0.018) 0.562*** (0.010) 0.052 (0.023) 0.294*** (0.012) 0.158*** (0.020) 0.646*** (0.027) (0.024) 0.469*** (0.012) 0.909*** (0.033) 0.268*** (0.018) 0.357*** (0.015) 0.565*** (0.012) 0.046*** (0.023) 0.272*** (0.016) 0.128*** (0.030) 0.623*** (0.032) Advancement 0.017 (0.028) 0.530*** (0.054) -0.169*** (0.047) -0.109** (0.043) -0.068** (0.032) -0.338*** (0.065) 0.331*** (0.043) 0.515*** (0.025) Benefits Challenge Independence Responsibility Salary Security Social Responsibility Adj. R2 F-Statistic λ 0.224 352.13*** -0.043*** (0.010) Χ2 156.26*** Data source: 2003 National Survey of College Graduates. Statistical significance: ***=0.01, **=0.05, *=0.10. Proceedings of the 2012 Pennsylvania Economic Association Conference 265 REFERENCES Belman, Dale and John S. Heywood. 2004. Public wage differentials and treatment of occupational differences. Journal of Policy Analysis and Management. 23, 1: 135-152. Benedict, Mary Ellen, David McClough, and Anita McClough. 2006. The price of morals: An empirical investigation of industry sectors and perceptions of moral satisfaction: Do business economists pay for morally satisfying employment? American Economist. 50, 1: 21-36. Brown, Charles C. and James L. Medoff. 1989. The employer-size wage effect. Journal of Political Economy. 97, 5: 10271059. Cauchon, Dennis. 2012. Federal pay ahead of private industry. USA Today. March 8, 2010. Congressional Budget Office. 2012. Comparing the compensation of federal and private sector employees. Cotton, Jerimiah. 1988. On the decomposition of wage differentials. Review of Economics and Statistics. 70, 2: 236-243. Edwards, Chris. 2010. Public sector unions and the rising costs of employee compensation. Cato Journal. 30, 1: 87-115. Ehrenberg, Ronald E. and J. L. Schwarz. 1986. Public sector labor markets. In Handbook of Labor Economics, Volume 2, ed. O. Ashenfelter and R. Layard. New York: North Holland Press. Falk, Justin. 2012. Comparing the compensation of federal and private-sector employees. Congressional Budget Office, January 2012. Found at: http://www.cbo.gov/sites/default/files/cbofiles/attachments/01-30-FedPay.pdf. Gittleman, Maury and Brooks Pierce. 2011. Compensation for state and local government workers. Journal of Economic Perspectives. 26, 1: 217-242. Greene, William H. 2012. Econometric Analysis, Seventh Edition. New Jersey: Prentice Hall. Gregory, Robert and Jeff Borland. 1999. Recent development in public sector labor markets. In Handbook of Labor Economics, Volume 3C, ed. O. Ashenfeleter and D. Card. New York: Elsevier Science. Gyourko, Joseph and Joseph Tracy. 1988. An analysis of public and private sector wages: Allowing for endogenous choices of both government and union status. Journal of Labor Economics. 6, 2: 229-253. Hartman, Robert W. 1983. Pay and pension for federal workers. Washington, DC: Brookings Institute. Keefe, Jeffrey. 2010. Debunking the myth of the overcompensated public employee: The evidence. Briefing paper 276, Washington, DC: Economic Policy Institute, September 15, 2010. Krueger, Andrew. 1988. Are public sector workers paid more than their alternative wage? In When Public Sector Workers Unionize, ed. R. Freeman and C. Ichniowski. Chicago: University of Chicago Press. Mohanty, Madhu S. 1994. Union premiums in the federal and private sectors: Alternative evidence from job qQueues. Journal of Labor Research. 15, 1: 73-81. Oaxaca, Ronald L. 1973. Male-female wage differentials in urban labor markets. International Economic Review. 14: 693709 Ross, Ron. 2011. Two different worlds: The public private sectors. American Spectator. June 1, 2011. Smith, Sharon P. 1977. Government wage differentials. Journal of Urban Economics. 4: 248-271. Snowden, Graham. 2012. Public and private sector wage gap rises to 8.2%. The Guardian. March 27, 2012. Wetterich, Chris. 2012. Comparing pay of public vs. private employees; pick your study. State Journal-Register. January 22, 2012. Proceedings of the 2012 Pennsylvania Economic Association Conference 266 AUTHOR INDEX Ezatollah Abbasian Thomas O. Armstrong Michael Bajic Nic Banting William Bellinger Mary Ellen Benedict Robert F. Brooker Jay Bryson Henry F. Check, Jr. Lisa Cooper James Dehnert Harry R. Diz Bari Dzomba Marwan El Nasser Rachel Gifford William R. Hawkins Michael J. Hannan Denae A. Heath Joseph Hess Adora D. Holstein David McClough page 208 page 101 page 251 page 101 page 55 page 251 page 243 page 48 page 90 page 232 page 101 page 243 page 232 page 177 page 232 page 145,156 page 222 page 202 page 232 page 125 page 251 Cameron D. McConnell Orhan Kara Tai McNaughton Daniel Meuser Tracy C. Miller Maysam Nasrindoost David Nugent Cristian Pardo Brenda Ponsford Tim Quinlan Rod D. Raehsler Karen L. Randall Richard Robinson Paul Sangrey Joe Seydl Yaya Sissoko Niloufer Sohrabji Tufan Tiglioglu, John S. Walker Carrie R. Williams page 202 page 37 page 117, 148 page 101 page 85 page 208 page 216 page 16 page 145,156 page 48 page 164 page 90 page 177, 223 page 196 page 48 page 66 page 66 page 232 page 90 page 187 Mary Ellen Benedict, Michael Bajic, and David McClough Proceedings of the 2012 Pennsylvania Economic Association Conference 267 Proceedings of the 2012 Pennsylvania Economic Association Conference 268