Introduction

1. The Impact of a Temporary Help Job: An Analysis of Outcomes for Participants in Three Missouri Programs Carolyn J. Heinrich LaFollette School of Public Affairs University of Wisconsin-Madison Peter R. Mueser University of Missouri-Columbia Department of Economics Kenneth R. Troske University of Kentucky Department of Economics and Center for Business and Economic Research October 2006

2. Introduction • Employment in temporary help service (THS) firms increased from less than 0.5% in 1982 to over 2.5% by 2004 • Growth was even more dramatic among the most disadvantaged • Increasingly used as a tool to help those facing difficulties obtaining employment • Complementary with “work first” programs

3. Introduction (continued) Definition of THS • Temporary help service provider hires worker • It then contracts with firm for firm to “use” worker • Worker activity occurs at firm site • Temporary help service receives payment from firm • Temp help service provider pays wages, taxes, benefits, etc. • Firm has no legal employment relationship with worker • THS is classified as an industry even though work site is in other industry

4. Introduction (continued) • Two views of temp help • Previously dominant view of temp help • less job stability • fewer fringe benefits • lower wages • Alternative view • path to permanent and stable employment • access to informal training and screening • consistent with “work-first” strategy

5. Introduction (continued) • Our analysis looks at effects of holding a THS job for those entering three federal programs in Missouri in two different years—1997 and 2001: • Temporary Assistance for Needy Families (TANF): Single mothers with very low incomes • Job Training—Job Training Partnership Act (JTPA) in 1997 and Workforce Investment Act (WIA) in 2001: Low income adults & displaced workers • Employment Exchange (“job service”) Anyone seeking a job • For each program, individuals are likely to be facing employment difficulties • But level of job skills differs across program • As does the severity of the employment shock

6. Literature • Empirical studies confirm that temp help services jobs • pay lower wages • offer fewer work hours • shorter in tenure • less likely to provide fringe benefits (e.g. health insurance, pensions)

7. Literature (continued) • Causal impact? • Most studies that look at impact find small or no negative effects of holding a temp help service job on eventual employment success (Lane et al. 2002; Heinrich et al. 2005; Anderson et al. 2002; Segal and Sullivan 1997; Booth et al. 2000) • One exception is Autor and Houseman (2005) who find that working in temp help does not lead to eventual employment success

8. Our contribution • We examine whether any other industry serves a similar role as THS • We look at 3 classes of workers who differ by their level of market disadvantage: Do effects differ? • We look at 7 industry groups: How do other specific industries compare with temporary help? • We look at how temp help workers succeed: Role of job changes in helping temp help workers succeed? • We look at whether the effect of temp help varies across the business cycle • We perform diagnostics to test whether results are likely spurious

9. Our findings • Temporary help industry serves a unique role as a transitional industry • Earnings are lower than in most other industries • Within 2 years, earnings have largely caught up • Still, those with initial jobs in some other industries are doing better (often manufacturing) • The catch up for temp help workers depends on moving to a better job

10. Our findings (continued) • Results strikingly similar for participants in different programs and for men and women • Benefits of a job in an alternative industry are slightly larger during a downturn, but basic patterns are similar (2001 vs 1997) • Effect estimates are not likely to be spurious

11. Data • Participants entering program in calendar year 1997, and 2001 • Focus much of the discussion on 1997 results • Age at least 18 but less than 65 • Program info from Missouri state administrative sources • Earnings/employment from the Unemployment Insurance (UI) “wage record data” for both Missouri and Kansas

12. Missouri

13. Population: 5.70 m (2003) • Land area: 178,415 sq km • Cities: • Kansas City metro area: 1.12 m • St. Louis metro area: 2.05 m • 4 smaller metro areas Switzerland population: 7.17 m Portugal population: 10.05 m

14. Population: 5.70 m (2003) • Land area: 178,415 sq km • Cities: • Kansas City metro area: 1.12 m • St. Louis metro area: 2.05 m • Columbia metro area 149,000 Missouri is a very “typical” of US states in terms of income, industry, age, race, politics.

15. Basic Model Reference quarter

16. Control Variables • Background: • age, age2 • years of education, high school, college • nonwhite • St. Louis central area • Kansas City central area • suburban, small metro, nonmetro

17. Control Variables (continued) • Prior labor market experience • proportion of previous 8 quarters working • working all previous 8 quarters • no work in previous 8 quarters • total earnings in prior year • total earnings two years prior • prior industry • Quarter of entry (1997:1-1997:4 or 2001:1-2001:4) • Unemployment in county in outcome quarter

18. Industry code • One industry in quarter • temporary help services (THS) • manufacturing • retail trade • service (but not THS) • other • Multiple industries • including THS • not including THS

19. Dependent variables • Basic model • Earnings in outcome quarter (quarter 9) • includes zeros • Difference-in-difference (quarter 9 earnings) – (quarter -9 earnings) • OLS • Interpretation is as prediction of “expected earnings”

20. Implicit Assumptions of the Analysis • We assume that, conditional on the control variables, industry in reference period is not associated with outcome earnings • Is this reasonable? • We think so: • Extensive list of control variables including prior work history and prior industry • Previous paper (Heinrich et al., 2005) controlled for selection and it didn’t matter

21. Implicit Assumptions of the Analysis • Also: • Determinants of industry choice from logit model reveals very little difference in type of industry • Very similar results with very different samples • Diagnostics suggest that effects estimates are not likely to be spurious

22. Employment in Industries 1997: Females

23. Employment in Industries 1997: Males

24. Who Gets a THS job? • MNL predicting industry in reference quarter (quarter 1 after program entry) • Dependent variable: THS, THS and other, other industry, no job (excluded)

25. Who Gets a THS Jobs? • Nonwhites • Those living in metro areas “Race and place matter”

26. Who Gets Temporary Help Jobs? • Nonwhites • Those living in metro areas Why? • Employers can screen nonwhites cheaply • Temp help jobs require labor market scale

27. Predicting Quarterly Earnings 1997 TANF Females Reference quarter industry Reference quarter earnings

28. Predicting Quarterly Earnings 1997 Training & Employment Exchange Females 2. Mean earnings 8 quarters later

29. Predicting Quarterly Earnings 1997 Training & Employment Exchange Males

30. Predicting Employment Probability • For both men and women employment in any sector in the reference period is strongly positively associated with the probability of employment eight quarters later relative to not having a job. • Once we control for characteristics there is very little difference between workers in the temp help sector and other sectors in the probability of employment in the future.

31. Transitions between sectors 1997 Transitions between sectors: Temporary help is easy to leave. Temporary help jobs often lead to manufacturing jobs

32. Analysis for 2001 • We redo our analysis for individuals entering the three programs in 2001 • 1997 was a period of growth • Unemployment around 3-4 percent in 1997-1998. • Between 1997-1999 employment grew by 4.4 percent • 2001 was a period of contraction • Unemployment was over 5.5 percent • Between 2001-2004 employment declined by 1.5 percent.

33. Analysis for 2001 • Did the role of THS change? • No: • THS is still unique: THS employment increases with program entry more than any other industry • Growth in temp help is somewhat less strong, however, especially for TANF participants

34. Predicted Quarterly Earnings 8 Quarters Later: Program Entry in 1997 and 2001 Females

35. 2001 Results • Results for Men are similar

36. 2001 Results • Temporary help still unique in its role as a transitional industry • Earnings, employment and transition results from 2001 follow a strikingly similar pattern to the 1997 result • Impact of Temporary Help jobs is slightly less beneficial during a recession

37. Correlated Errors: Robustness Check • What if unmeasured factors are correlated with reference quarter industry and outcome earnings? • Altonji, Elder and Taber (2005) suggest using measured controls to suggest how large the bias of unmeasured factors may be • We implemented their methods in 3 ways

38. Robustness Checks: Summary ▲ One test implies coefficient is not spurious ▲▲ Two tests imply coefficient is not spurious ▲▲▲ Three tests imply coefficient is not spurious Coefficients in red not statistically significant

39. Robustness Checks: Summary Coefficients in red not statistically significant

40. Robustness Checks: Summary • In most cases, in order for estimated coefficient to be spurious • error term needs to be related to THS employment in a very different way than observed controls • This seems implausible • All estimated impacts are unlikely to be spurious

41. Conclusions • We have investigated temp help jobs obtained following an employment “crisis” • Temporary help industry serves a unique role as a transitional industry • Earnings are lower than in most other industries • Within 2 years, earnings have largely caught up

42. Conclusions • Still, those with initial jobs in some other industries are doing better (often manufacturing) • The catch up for temp help workers depends on moving to a better job • Results strikingly similar for participants in different programs and for men and women • Benefits of a job in an alternative industry are slightly larger during a downturn, but basic patterns are similar

43. Conclusions • Obtaining a temporary help job is clearly better than having no job • We see no evidence that a strategy of waiting for a “better” job yields any benefits. • These results do not differ across our three programs • Heterogeneity of our sample suggests that our results are general.