This project will use an important new Census dataset, merged with Department of Labor data on participants in their services and training programs, to evaluate the impacts of Workforce Investment Act (WIA)-funded programs on the labor market outcomes of workers. The investigators will address the following substantive questions: What are the impacts of WIA services and training on various worker labor market outcomes? Are some services (e.g., "core" services, "intensive" services, or training) more effective than others, especially relative to their costs? Do the impacts vary across categories of recipients (i.e., adults, dislocated workers, and youth) or across demographic groups (such as race/gender) within categories? Does WIA participation vary systematically across groups, and does receipt of particular services vary across groups as well for those who participate? Do the various WIA services have impacts on the kinds of jobs and firms in which these workers become employed, as well as on their overall employment and earnings? The investigators will address the following methodological questions: Which non-experimental estimators of training program impacts based on administrative data, if any, generate credible estimates? What types of administrative data are required to generate credible impact estimates; can we define a "minimal administrative dataset"? How much does placing participants and non-participants in the same local labor markets matter and, if it matters, how local is "local" and what methods for taking account of local labor markets work best? How much do estimates based on recently developed dynamic treatment effects methods differ from those produced by traditional approaches in their magnitudes and interpretations?
To address these questions, the investigators will use data from the Longitudinal Employer Household Dynamics Program (LEHD) at the Census Bureau. This powerful new dataset is based on quarterly Unemployment Insurance (UI) records on workers' employment and earnings histories and ES202 records on the characteristics of the employing firms in various states, which have been linked to other Census data (including the 2000 Decennial Census of Population) on households and employers. The investigators plan to integrate LEHD data on individuals with data on WIA participants from the Workforce Investment Act Standardized Record Data (WIASRD) in up to 9 states. The investigators will use these data to estimate a variety of models with postprogram earnings and employment, as well as job characteristics (in some cases), as dependent variables. The investigators will also use the LEHD data in conjunction with earlier data from experimental evaluations of the Job Training Partnership Act (JTPA) to establish the broader credibility of various non-experimental approaches; and they will use variation across states and local areas in expenditures and in service type frequency as possible instrumental variables in the non-experimental evaluations as well. Finally, they will have data on TANF and UI recipients as well as Employment Service (ES) registrants merged with the LEHD data in each state; they can use these data to estimate impacts for additional sub-samples of workers and to generate comparison groups for our non-experimental estimators.
Answering these questions will provide researchers with important substantive information on the effectiveness of a large federal program on which little up-to-date knowledge currently exists. The answers will also have important methodological implications regarding the design of non-experimental evaluations. The data to be used here are probably the best combination of administrative and survey data ever assembled for the purpose of evaluating a social program; the sample sizes are huge, enabling the investigators to estimate impacts for various subpopulations quite precisely; and the covariates on both the household and employer sides are very rich.
Broader Impacts: The research results could influence overall funding levels and the allocation of these funds across service types and/or categories of participants in future reauthorizations of the Workforce Investment Act, and also how One-Stop offices across the country use their WIA funds. The project should also heavily influence both research and practice in the area of econometric program evaluation. The investigators expect to learn a great deal about the credibility of different non-experimental evaluation estimators applied to administrative datasets and about the extent to which such work can provide plausible alternatives to costly and time-consuming experimental evaluations.
