Manufacturing Skills and Skill Gaps following Volatility and High Unemployment Andrew Weaver Massachusetts Institute of Technology IST/Lisbon June 25, 2014 (joint with Paul Osterman) What are the Issues? Background High and persistent unemployment Firms complain they can’t find skilled workers Questions Does mismatch/gap exist between employer demands and the supply of skills in the marketplace? If so, is it a simple/mechanical result of inadequate worker skills, or are other more complex factors to blame (cyclical demand, corporate strategy, communication among economic actors, etc.)? 2 Relation to Uncertainty and Industrialization Patterns Loss of domestic mfg. raises questions about trajectory of industrial growth, economic development, job quality Policymakers need to understand this issue in order to foster economic growth and improve economic outcomes for workers Common skill-biased technical change (SBTC) narrative leads to focus on supply-side labor market frictions If problem is just structural/skills gap: long-term ed. attainment and worker behavior If other factors matter, SBTC narrative may be misleading and other interventions may be necessary Exhortations to increase STEM education may not solve the problem Institutional approaches may be required: making connections with local labor market intermediaries, solving coordination/communication failures, etc. 3 Presentation Goals Set boundaries on incidence of skill gaps Demonstrate simple skill mismatch story is inadequate Point to importance of intermediaries and institutions in addressing challenges in skill supplies 4 Shortcomings of Existing Research Takes place at very abstract level without direct measurement Unemployment-vacancy indices (Sahin et al. 2012; Canon, Chen and Marifian 2013) Are sensitive to changes in firm strategy (recruitment, wages) Are sensitive to cyclicality Vague measure: hides mechanism (geography? skills?) Only measure inter-industry mismatch (Modestino 2010; Lazear and Spletzer 2012) Supply-Demand indices (Estevau and Tsounta 2011; Rothwell 2012) Use education as proxy Distorts demand measurement: college-educated barista Ignores within-education variation in skills Proxy on both sides: any regional or intra-industry variation generates mismatch 5 Manufacturing Puzzle For manufacturing, important facts are inconsistent with skill gap claims Deloitte and the National Association of Manufacturers (2011) report survey results: 600,000 unfilled jobs due to lack of qualified workers 74% of manufacturers report lack of skilled production workers had significant negative impact If demand exceeds supply for high-skilled manufacturing workers, we would expect wages to increase 6 Manufacturing Wage Trends Community College Wage Premium by Industry Sector 1.35 Ratio of AA to HS Wages 1.30 1.25 1.20 Mfg. Premium Non-Mfg. Premium 1.15 1.10 1.05 2000 2001 2002 2003 2004 2005 2006 Year 2007 Source: CPS MORG (NBER) data. 7 2008 2009 2010 2011 Approach To answer questions about skill and mismatch, it’s necessary to gather direct evidence on skill demands: What skills do employers demand? Which establishments demand high levels of skill? Do establishments, particularly those with high skill demands, have trouble finding workers with these skills? To really narrow in on skills, important to focus on industry/industry sector Paul Osterman and I designed and administered a nationally representative survey of manufacturing plants to answer these questions We conducted extensive fieldwork to identify critical factors relating to institutions, regional eco-system 8 Our Survey Administered in late 2012, early 2013 Random sample—Dun & Bradstreet database n=903 36% response rate Focus on “core” production workers (Ben-Ner and Urtasun 2013, Osterman 1995)—62% of estab. employment Concrete skill questions: does this job require reading complex technical manuals? algebra? geometry? etc. Defined skill gaps as prolonged core worker vacancies (> 3 mos.) 9 Basic Skill Demands Basic Skill Demands for Core Production Jobs All Establishments Basic reading (ability to read basic instruction manuals) 75.6% Basic writing (ability to write short notes, memos, reports less than one page long) 60.5% Basic math (ability to perform all of math categories below) 74.0% Addition and subtraction Multiplication and division Fractions, decimals, or percentages 10 Require basic reading, writing, and math 42.4% Require use of computers several times per week or more frequently 62.3% Ability to use word processing software or ability to search Internet for information 41.7% Interpersonal/Problem-Solving/Soft Skill Demands Percent of Establishments Citing Interpersonal, Problem-Solving, and Other Soft Skills as Very or Moderately Important for Core Jobs 11 Very Important Very or Moderately Important Cooperation with other employees Ability to evaluate quality of output 81.2% 71.0% 99.3% 95.8% Ability to take appropriate action if quality is not acceptable 76.3% 97.7% Ability to work in teams Ability to learn new skills 64.2% 50.1% 91.1% 89.3% Ability to independently organize time or prioritize tasks 45.6% 84.4% Ability to solve unfamiliar problems 38.8% 83.0% Ability to critically evaluate different options 35.7% 74.1% Ability to initiate new tasks without guidance from management 35.2% 80.9% Extended Skill Demands Extended Skill Demands for Core Production Jobs All Establishments Extended reading (docs > 5pg.; trade jrn.; tech. docs) 52.6% Extended writing (>1pg.) 22.1% Extended math (ability to perform any of three math 38.0% categories below) Algebra, geometry, or trigonometry 31.5% Probability or statistics 13.6% Calculus or other advanced mathematics 7.4% Extended computer 41.9% Use CAD/CAM 28.4% Use other engineering or manufacturing software 29.2% Ability to write computer programs (such as program a CNC machine for a new piece, etc.) 18.6% Unique skill 12 25.9% Skill Gap Evidence Vacancies 90.0% Percent of Establishments 80.0% 70.0% 76.3% 64.9% 60.0% 50.0% 40.0% 30.0% 17.4% 20.0% 10.2% 10.0% 7.6% 7.4% 5.6% 10.6% 0.0% No vacancies More than zero, less More than 5%, less than More than 10% than or equal to 5% or equal to 10% Vacancy Measure as a Percent of an Establishment's Core Employees 13 Any Vacancies Long-Term Vacancies What Skill Demands are Associated with Hiring Difficulties? Demands for higher level reading, math, and unique skills are significant predictors of long-term vacancies Computer and soft skills/problem-solving/initiative skills are not So is this relationship between skill demands and hiring problems an automatic/mechanical one? Examine high skill-demanding establishments 14 Which Establishments Demand High Skills? Establishments that demand extended skills are characterized by: high-tech cluster membership high-performance work organization (TQM/self-managed team) frequent process (not product) innovation more foreign competition If the simple skill mismatch story is accurate, these establishments should have significantly higher levels of hiring difficulties 15 Long-Term Vacancies: Estab. Characteristics Models High-tech Above-avg. tech. TQM pct. Self team pct. Product innovation Process innovation Industry cluster Part of larger firm More foreign comp. 16 Pct. LT vac.-- Pct. LT vac.-- LTV--Logit-- LTV--Logit-RF RF+wage RF RF+wage Pct. LT vac. LTV--Logit -0.01 (0.007) -0.001 (0.006) 0.000 (0.000) 0.000 (0.000) 0.002 (0.007) 0 (0.007) -0.052 (0.038) -0.019 (0.033) 0.001 0.000 0.001 (0.001) 0.019 (0.038) 0.005 (0.038) -0.014** (0.007) -0.001 (0.006) 0.002 0.000 0.002 (0.000) 0.001 (0.007) 0.002 (0.007) -0.017** (0.007) -0.001 (0.006) 0.003 0.000 0.003 (0.000) 0.003 (0.007) 0.003 (0.007) -0.068* (0.037) -0.024 (0.033) 0.004 0.000 0.004 (0.001) 0.018 (0.039) 0.013 (0.038) -0.072* (0.039) -0.026 (0.034) 0.005 0.000 0.005 (0.001) 0.022 (0.040) 0.021 (0.039) 0.017*** (0.006) 0.003 (0.007) 0.002 (0.007) 0.119*** (0.032) 0.024 (0.037) 0.019 (0.038) 0.014** (0.006) 0.003 (0.006) 0.001 (0.007) 0.013** (0.006) 0.001 (0.007) 0.001 (0.007) 0.117*** (0.032) 0.032 (0.037) 0.027 (0.039) 0.117*** (0.033) 0.031 (0.038) 0.025 (0.039) Long-Term Vacancies: Red. Form Cont’d County pop. density County unemp. rate (2011) Pct. change in core emp. last 2 yrs. 0.000 0.000 0.000 0.000 (0.000) (0.000) (0.000) (0.000) -0.148 -0.129 -0.795 -0.695 (0.122) (0.124) (0.711) (0.721) 0.016 0.016 (0.024) (0.025) -0.011*** -0.012*** (0.004) Standardized division wage Low wage R-Squared N 0.003 0.005 (0.003) (0.017) 0.121*** 0.18 (0.032) (0.160) 0.036 0.034 0.050 0.073 0.040 0.038 783 784 766 738 766 738 Source: PIE Manufacturing Survey. * p<0.10, ** p<0.05, *** p<0.01 17 (0.004) Summary of Results No widespread problem with skill gaps It is worth paying attention to the minority of establishments reporting difficulties Skills are important Extended math is important Extended reading is surprisingly prominent Unique skills: may reflect internal training decline However, many establishment characteristics associated with higher skill demands (high-tech, HPWS, process innovation) are not associated with hiring difficulties This implies no simple/mechanical relationship between higher skill demands and hiring problems: other factors mediate relationship 18 What’s Going On? Skills are critical, but skill gap formulation is not necessarily the best way to frame the issue American skill production system has been changing Decline in mfg. establishment size (Holmes 2011; Henly and Sanchez 2009) Small firms provide less internal training (Lynch and Black 1998) External training actors like community colleges are more important than they once were But system is disaggregated More potential for coordination failures and underinvestment in public goods 19 Intermediaries/Institutions are Important Rochester story Kodak Monroe Community College Rochester Regional Photonics Cluster (RRPC) Addressed coordination failure Intermediaries and institutions are critical for matching supply and demand, as well as coordinating increases in skill demands and supplies Challenge coming from volatility/uncertainty: Simple SBTC story says high returns to education/skills will provide should provide incentive for supply side of labor mkt. However, volatilty may destroy the very institutions and intermediaries necessary to raise skill levels on both supply and demand side 20 Thank You Andrew Weaver weaver55@mit.edu 21 Our Survey 22 Production in the Innovation Economy (PIE) project Administered in late 2012, early 2013 Dun & Bradstreet database Establishment approach: Bloom and Van Reenen 2007, Lynch and Black 1998 Manufacturing establishments, excluding baking, printing, and publishing Random sample, stratified by estab. size (>10 emp.) Targeted plant managers (identified appropriate person) $10 incentive n=903 36% response rate Survey Design Focus on “core” production workers (Ben-Ner and Urtasun 2013, Osterman 1995)—62% of estab. employment Battery of concrete skill questions (>30) Examples Reading Basic: Does this job require reading basic instruction manuals? Extended: Does this job require reading complex technical documents or manuals? Any document that is longer than five pages? etc. Math Basic: Does this job require mathematical operations involving multiplication and division? Extended: Does this job require mathematical operations involving probability and statistics? Algebra, geometry, or trigonometry? etc. 23 Survey Design (2) Concrete questions re skill gaps Defined skill gaps as prolonged core worker vacancies (> 3 mos.) Background data on establishment and workforce 24 Industry Employment/financial trends Innovation Training Age structure Sex composition etc. Hypotheses H1: If skill gaps are a widespread problem in manufacturing, then long-term vacancies will be a widespread problem Benchmark from Deloitte survey: 74% of mfg. firms suffered from lack of skilled production workers H2: If demand for higher skills mechanically leads to hiring problems, the establishments characterized by the highest skill demands should experience greater problems Higher level skill demands (math, reading, etc.) should be associated with hiring problems If high-tech or other types of establishments have higher skill demands, they should have more severe hiring problems 25 Analysis of Establishments with Hiring Difficulties Two dependent variables: Long-term vacancies as a percentage of total core workers (OLS) Indicator for long-term vacancies (Logit) First estimate models with skill variables as regressors, then with high-skill establishment characteristics as regressors Reduced form controls Supply: county unemployment rate (2011), county population density Demand: change in core workers over past two years Wage measures (mgmt. strategy): standardized by Census geographic division (2011 to avoid simultaneity) All models control for establishment size 26 Long-Term Vacancies: Skill Models Any extended skill Pct. LT vac. LTV 0.016*** (0.006) 0.073** (0.035) Extended reading Extended writing Extended math Extended computer Unique skill New skills Evaluate quality R-Squared/Pseudo R2 0.025 N 869 * p<0.10, ** p<0.05, *** p<0.01 27 0.023 870 Pct. LT vac.-detailed skills LTV-detailed Pct. LT vac.-detailed, red. form LTV-detailed, red. form 0.012** (0.006) -0.002 (0.007) 0.017*** -0.006 0.007 (0.006) 0.013** (0.006) -0.002 (0.006) 0.001 (0.006) 0.465*** (0.178) -0.12 (0.205) 0.531*** -0.187 -0.11 (0.179) 0.442** (0.177) 0.263 (0.172) -0.252 (0.187) 0.011* (0.006) -0.003 (0.007) 0.019*** -0.006 0.007 (0.006) 0.015** (0.006) -0.003 (0.006) 0.002 (0.006) 0.513*** (0.189) -0.143 (0.215) 0.608*** -0.197 -0.062 (0.185) 0.553*** (0.185) 0.224 (0.180) -0.282 (0.192) 0.052 831 0.053 832 0.082 778 0.060 778 Implications Demanding high skill levels is not necessarily a ticket to trouble A wider range of institutional policies responses may be relevant Targeted policies may have the potential to affect hiring outcomes even holding current worker skill levels constant Institutional relationships (e.g., between firms and community colleges or labor market intermediaries) Incentives for firm-level human resource/training policy Policies that reduce risk of mfg. career for job applicants 28 Further Research Agenda PIE Survey Institutional and other factors that mediate skill and hiring problems Community college density: determinants and implications for economic growth LPNs and job ladders Social entrepreneurship and philanthropic capital markets 29 Extended Skill Demands: Logit Analysis High-tech industry Above-average tech. TQM pct. Self team pct. Frequent product innovation Frequent process innovation Industry cluster Part of larger firm More foreign competition Pseudo R-Squared N 30 * p<0.10, ** p<0.05, ***p<0.01 Any Extended Skill Extended Reading Extended Math Extended Computer Unique Skill Extended Writing 0.172*** (0.032) 0.019 (0.032) 0.001*** 0.000 0.001 (0.001) 0.288*** (0.038) 0.039 (0.035) 0.001 0.000 0.002*** (0.001) 0.026 (0.039) 0.001 (0.033) 0.0003 0.001 0.001** (0.001) 0.143*** (0.043) 0.059* (0.036) 0.001*** 0.000 0.0005 (0.001) 0.006 (0.037) 0.087*** (0.032) 0.001* 0.000 0.001** 0.000 0.025 (0.036) -0.018 (0.030) -0.0001 0.000 0.001 0.000 0.032 0.006 -0.06 -0.025 -0.014 0.038 (0.037) (0.041) (0.040) (0.041) (0.038) (0.034) 0.078** 0.081** 0.076** 0.125*** 0.036 0.02 (0.037) 0.073** (0.030) -0.025 (0.035) 0.059* (0.035) 0.066 804 (0.040) 0.062* (0.034) 0.01 (0.039) 0.062 (0.040) 0.082 797 (0.037) 0.066** (0.032) -0.089** (0.038) 0.112*** (0.040) 0.043 796 (0.040) 0.093*** (0.034) -0.088** (0.039) 0.119*** (0.041) 0.061 795 (0.036) 0.095*** (0.031) -0.039 (0.036) 0.027 (0.037) 0.038 800 (0.034) -0.023 (0.029) 0.029 (0.033) -0.019 (0.033) 0.013 792 Size Distribution 40.0% 35.0% 30.0% 25.0% 20.0% 15.0% 10.0% 5.0% 0.0% <20 20-99 PIE 31 100-249 250-499 County Business Patterns 2010 500+ Weighting and Validation Small establishments somewhat more likely to respond than large estab. For all descriptive statistics we use size weights based on the employment-weighted proportion of establishments of various size classes in the Census Bureau’s County Business Patterns (CBP) data Validate aggregate workforce data with CPS: close match 32 Validation with CPS 33 PIE CPS (2012) Hourly wage 16.95 16.49* Union 18.1% 13.7%* Female 26.7% 26.6% Age 30 or less 20.6% 21.3% Age 31-40 27.5% 22%* Age 41-55 35.8% 38.8%* Age 56 plus 16.1% 17.9%* *=significant differences at 95 percent level or higher. Geographic Distribution PIE Survey West 18% South 26% 34 Northeast 21% Midwest 35% Geographic Comparison 40.0% 35.0% 30.0% 25.0% 20.0% 15.0% 10.0% 5.0% 0.0% Northeast PIE survey 35 Midwest South County Business Patterns 2011 West Industry Distribution NAICS 36 311 312 313 314 315 316 321 322 323 324 325 326 327 331 332 333 334 335 336 337 339 NAICS 3-Digit Industry Food mfg. Beverage and tobacco product mfg. Textile mills Textile product mills Apparel mfg. Leather and allied product mfg. Wood product mfg. Paper mfg. Printing and related support activities Petroleum and coal products mfg. Chemical mfg. Plastics and rubber products mfg. Nonmetallic mineral product mfg. Primary metal mfg. Fabricated metal product mfg. Machinery mfg. Computer and electronic product mfg. Electrical equip, appliance, and comp. mfg. Transportation equipment mfg. Furniture and related product mfg. Miscellaneous mfg. + other PIE Survey Pct. of Estab. 4.5% 1.2% 0.6% 2.4% 1.3% 0.1% 3.4% 1.4% 3.6% 0.8% 6.4% 6.0% 3.4% 3.4% 22.2% 12.0% 7.7% 3.1% 5.4% 3.0% 8.4% Industry Comparison NAICS PIE Survey CBP Pct. of Pct. of Estab. Estab. 2011 PIE-CBP 311 312 313 314 315 316 321 322 Food mfg. Beverage and tobacco product mfg. Textile mills Textile product mills Apparel mfg. Leather and allied product mfg. Wood product mfg. Paper mfg. 4.5% 1.2% 0.6% 2.4% 1.3% 0.1% 3.4% 1.4% 8.6% 1.7% 0.8% 2.1% 2.4% 0.4% 4.7% 1.5% -4.1% -0.5% -0.2% 0.3% -1.2% -0.3% -1.3% -0.1% 323 324 325 326 327 331 332 333 Printing and related support activities Petroleum and coal products mfg. Chemical mfg. Plastics and rubber products mfg. Nonmetallic mineral product mfg. Primary metal mfg. Fabricated metal product mfg. Machinery mfg. Computer and electronic product mfg. Electrical equip, appliance, and comp. mfg. Transportation equipment mfg. Furniture and related product mfg. Miscellaneous mfg. + other 3.6% 0.8% 6.4% 6.0% 3.4% 3.4% 22.2% 12.0% 9.4% 0.8% 4.4% 4.3% 5.2% 1.6% 18.8% 8.1% -5.8% 0.0% 2.0% 1.6% -1.8% 1.8% 3.4% 3.9% 7.7% 4.4% 3.3% 3.1% 5.4% 3.0% 8.4% 2.0% 3.9% 5.6% 9.3% 1.1% 1.5% -2.6% -0.9% 334 37 NAICS 3-Digit Industry 335 336 337 339 Occupational Wage Comparison Average Hourly Wages by Selected Manufacturing Occupations 2008 2011 % change Production occupations 15.87 16.74 5.5% Machinists 18.17 19.51 7.4% Industrial Engineering Technicians 22.89 24.42 6.7% Mechanical Engineering Technicians 23.74 24.92 5.0% Industrial Engineers 35.47 37.56 5.9% Source: BLS Occupational Employment Statistics. 38 Extended Skill Demand Models (2) High tech Above-avg. tech TQM pct. Self team pct. Frequent prod. innovation (g) Frequent process innovation Industry cluster Part of larger firm (g) More foreign comp. Employment size FE (g-2nd) R-Squared N LR test (Chi2) p-value--prop. odds 39 Brant test p-value * p<0.10, ** p<0.05, Extended Skill Index Pr(index=3) Extended Skill Index Pr(index=3) Gen. Ord. Logit 0.095*** (0.021) 0.016 (0.016) 0.0004** (0.0002) 0.001** (0.0003) -0.012 (0.019) 0.047*** (0.018) 0.034** (0.015) -0.025 (0.018) 0.048** (0.020) x 0.031 778 0.097*** (0.020) 0.0004 (0.015) 0.0004** (0.0002) 0.001*** (0.0002) -0.065*** (0.024) 0.051*** (0.018) 0.035** (0.015) -0.060*** (0.018) 0.049*** (0.018) x 0.042 778 0.158 0.801 0.096 ***p<0.01 Quantifying the Skill Gap Core vacancies as pct. of total estab. emp. Core long-term vacancies as pct. of total estab. emp. Core long-term vac. as pct. of total vac. Core workers as pct. of total estab. emp. 1.1% 0.5% 48.4% 62.0% Est. total PIE vac. pct. if vac. are proportional JOLTS vacancies as pct. of total CES mfg. emp. (Aug. 2011) 1.8% 2.0% JOLTS vacancies as pct. of total CES mfg. emp. (4Q 2012) 2.0% Deloitte implied vacancies as pct. of total CES mfg. emp. (Aug. '11) 5.1% Implied long-term vac./emp. (based on long-term vac. pct. of total vac.) PIE 0.9% JOLTS (4Q 2012) 1.0% Deloitte 2.4% 40 Implied total PIE mfg. vacancies 213,712 Implied PIE long-term mfg. vacancies 103,390 PIE long-term vacancies as pct. of mfg. unemployed (4Q 2012) 11.6% Skill Gap Robustness (1) What if we miss skill gaps because we’re looking at a point in time Hiring funnel (based on attempt to hire in last two years) Hiring Funnel for Core Workers 41 Mean Median 75th-weeks 25th-others Weeks required to recruit and hire applicant (start of process to extension of offer) 5.9 4.0 6.0 Typical number of applications received per open core position 23.8 10.0 5.0 Typical number of interviews conducted per open core position 5.9 5.0 3.0 85.4% 95.0% 80.0% Acceptance rate by applicants who are extended an offer Source: PIE Manufacturing Survey. Skill Gap Robustness (2) Alternative Measures 42 Ever reduced production due to vacancies 17.7% Have vacancy and ever reduced prod. 7.6% Long hire times over past two yrs. (>=3 mos.) 10.9% Long-term vac. OR vac. + reduced prod. 25.6% Long-term vac. OR prior long hire times 34.1% Why Manufacturing? Manufacturing is interesting test case for structural mismatch Arguments about mismatch/spiking skill demands are commonly applied to manufacturing Deloitte and the National Association of Manufacturers (2011) report survey results: 600,000 unfilled jobs due to lack of qualified workers 74% of manufacturers report lack of skilled production workers had significant negative impact Capital intensive / sensitive to technology shocks Key theories: tech. shocks drive mismatch (Brynjolfsson and McAfee 2012, Autor, Levy, and Murnane 2003) STEM skills are important Idea of vacancies in industry with millions of laid off workers implies structural gap Broad sector with a lot of variation (high-tech/low-tech, domestic/export, etc.) 12% of GDP; 70% of industry R&D 43