ECONOMICS 515 ECONOMETRICS: THEORY AND APPLICATIONS Professor Thornton W, 6:30-9:10 Office and Phone: Email: Website: Office Hours: Required Text: Optional Text: Other: Winter 2016 Pray-Harrold 408 Pray-Harrold 703-C, 487-0080. jthornton@emich.edu. http://people.emich.edu/jthornton. T, Th, 12:15-2:00; W, 4:30-6:30; and by appointment. William Griffiths, R. Carter Hill, and George Judge, Learning and Practicing Econometrics, John Wiley & Sons, 1993. Peter Kennedy, A Guide to Econometrics. This text is recommended but not required. Stata Guide and Assignments and Topic Outlines. These can be downloaded from the Econ 515 homepage on my website. Course Objective The purpose of this course is to expose you to advanced econometric methods. The focus of the class is on five types of statistical models: classical and general linear regression models, regression models involving systems of equations, panel data regression models, binary discrete choice regression models, and duration models. You will learn methods for estimating and testing hypotheses about the parameters of these models, as well as a basic understanding of model specification issues. The emphasis of this class is on the practical application of econometrics, and understanding the important relationship between economic and statistical models in empirical research involving economic phenomena. Grade Your grade in the class will be based on 3 take-home projects, a number of class preparation assignments, and a final exam. The purpose of the take-home projects is to allow you to use data sets to implement the methods and techniques covered in the lectures, and to illustrate the link between economic theory and statistical analysis. These three projects are tantamount to “takehome exams.” You should not consult other members of the class when doing these assignments. The final exam will cover end of course material not included on the take-home projects. The class preparation assignments require you to run the sample programs provided in Stata Guide and Assignments, bring the computer output to class, and participate in discussing the results obtained. Each computer assignment has a weight of 25%. The final exam has a weight of 15%. The class preparation assignments have a weight of 10%. Statistical Program The statistical program used in this class is Stata. The program is available on the computers in the social science lab, seventh floor, Pray-Harrold Hall. You may purchase a copy of the Stata/IC statistical program (recommended for this class and future use) for $198; an annual license for $125; or a six month license for $75. Purchasing this program or a license to use is not required. If you are interested in getting a copy of Stata that you can use on your home computer, go to: http://www.stata.com/order/new/edu/gradplans/student-pricing/. Homepage To access the homepage for this class, go to my website: http://people.emich.edu/jthornton. Click on the link for Econ 515. The Econ 515 homepage can be used to obtain a variety of information including the syllabus, handouts, assignments, self-teaching guides for statistical packages, data for the self-teaching guides and assignments, lecture outlines, and articles available on the web. Outlines of Topics and Related Readings The following is an outline of topics to be covered and a list of related reading assignments. Readings from the Kennedy book are optional. 1. The Classical Linear Regression Model A. B. C. D. E. F. G. The data generation process. Assumptions of the classical linear regression model. The classical linear regression model in matrix format. The least squares estimation rule. The maximum likelihood estimation rule. Properties of estimators when applied to the classical linear regression model. Measures of goodness of fit for the classical linear regression model. Readings: Griffiths, Hill, and Judge. Chapters 5, 6, 8, 9. Classical liner regression model lecture outline. Kennedy. Chapters 3, 5. Class Preparation Assignment: Stata Guide, pages 2-8. 2. Hypothesis Testing: Testing Restrictions on Model Parameters A. Fixed value restrictions, linear restrictions, and nonlinear restrictions. B. Maintained hypotheses, testable hypotheses, nested hypotheses, and non-nested hypotheses. C. The general procedure for hypotheses testing. D. Small sample tests: t-test, F-test. E. Large sample (asymptotic) tests: Asymptotic t-test, Likelihood ratio test, Wald test, Lagrange Multiplier test. Readings: Griffiths, Hill, and Judge. Chapters 7, 10, 14. Hypothesis testing lecture outline. Kennedy. Chapter 4. Class Preparation Assignment: Stata Guide, pages 9-12. 3. The General Linear Regression Model A. Nonspherical errors and sources of nonspherical errors. B. Assumptions of the general linear regression model. C. Properties of the OLS estimator when applied to the general linear regression model with nonspherical errors. D. The generalized least squares (GLS) estimator. E. Properties of the GLS estimator when applied to the generalized linear regression model with nonspherical errors. F. Practical problems with using the GLS estimator. G. The feasible generalized least squares (FGLS) estimator. H. Properties of the FGLS estimator when applied to the generalized linear regression model with nonspherical errors. I. General linear regression model with heteroscedasticity. J. Tests of heteroscedasticity. Readings: Griffiths, Hill, and Judge. Pages 477-482. General linear regression model lecture outline. Kennedy. Chapter 8. Class Preparation Assignment: Stata Guide, pages 12-14. 4. The Seemingly Unrelated Regressions Model A. Introduction to equation systems. B. Assumptions of the seemingly unrelated regressions model. C. Estimation procedures for the seemingly unrelated regressions model: Ordinary least squares, generalized least squares, feasible generalized least squares, iterated generalized least squares, direct maximum likelihood. D. Testing hypotheses in the seemingly unrelated regressions model: Approximate Ftest, Wald test, likelihood ratio test, Lagrange multiplier test. E. Cross equation restrictions in the seemingly unrelated regressions model. Readings: Griffiths, Hill, and Judge. Chapter 17. Seemingly unrelated regressions model lecture outline. Class Preparation Assignment: Stata Guide, pages 14-15. 5. The Simultaneous Equations Regression Model A. B. C. D. E. F. Sources of correlation between a right-hand side variable and the error term. Definitions and basic concepts for simultaneous equations models. Under-identified, exactly identified, and over-identified equations. Identifying a structural equation. Rank and order conditions for identification. Single equation methods of estimation: ordinary least squares, indirect least squares, instrumental variables, two-stage least squares, limited information maximum likelihood. G. Systems methods of estimation and estimator properties: three-stage least squares, iterated three-stage least squares, full-information maximum likelihood. H. Hypothesis testing in the simultaneous equations regression model. Readings: Griffiths, Hill, and Judge. Pages 581-584. Chapters 18, 19. Simultaneous equations regression model lecture outline. Kennedy. Chapters 9, 10. Class Preparation Assignment: Stata Guide, pages 15-17. 6. Regression Models for Panel Data A. B. C. D. E. One way error component regression models. Fixed and random effects models. Two-way error component regression models. Fixed and random effects models. Hypotheses testing in panel data regression models. Readings: Griffiths, Hill, and Judge. Chapter 17 appendix B. Panel data models lecture outline. Class Preparation Assignment: Stata Guide, pages 18-21. 7. Binary Discrete Choice Models A. B. C. D. Introduction to binary discrete choice models. The linear probability model. The probit model. The logit model. Readings: Griffiths, Hill, and Judge. Chapter 23. Binary discrete choice models lecture outline. Kennedy. Chapter 15. Class Preparation Assignment: Stata Guide, pages 21-22. 8. Duration (Survival) Analysis A. B. C. D. E. F. Parametric duration models. Weibull regression model. Semiparametric duration models. Cox proportional hazards regression model. Nonparametric duration models. Life-table and Kaplan-Meier estimators. Readings: Duration models for time to event data lecture outline. Class Preparation Assignment: Stata Guide, pages 22-24. 9. Analysis of Experimental Data A. Randomized controlled and natural experiments. B. Difference estimator. C. Difference estimator with control variables. D. Difference in differences estimator. Readings: Analysis of experimental data lecture outline.