Finally, in order to obtain similar outcome measures for individuals in both the treatment and comparison samples, states will be required to provide longitudinal, individual‐ level, unemployment insurance wage record data that can be matched to the treatment and comparison sample data for several years prior to treatment as well as several years after treatment. The data prior to treatment will be used to ensure that treatment and comparison individuals have comparable labor market dynamics. The longitudinal post‐treatment data will be used to estimate both short‐run and long‐run treatment effects.
At first blush—at least to non‐economists—these data requirements might seem like overkill. But this is essential information to zero in on the vital questions of which training programs work and for whom. And ultimately, clear answers to these questions are needed to ensure that the most effective assistance can be given to people facing job loss and dislocation and to ensure that taxpayer dollars are used in the most effective way.
These data will allow researchers to examine both the overall benefits of the job training system on an on‐going basis and assess the effectiveness of its component parts. For example, what are the benefits of the services currently provided in the ES program? How does the cost‐benefit ratio of ES compare with WIA intensive services? Also, there has been no research studying the return to training at community colleges that is funded through job training programs. Further, there has been very little work examining whether the returns to training vary over the business cycle, something that would be quite useful to know since the ARRA stimulus package included a significant increase in funding for these programs. Other important research topics include estimating the effect of the job training programs on the marginal participant, as well as estimating the entire distribution of effects.
The impact the program has on the average participant is interesting for a new program or for considering shutting down an existing program. To decide whether to expand or contract a program, however, the impact of the program on the marginal participant is the relevant statistic. Knowing something about the distribution of returns shows whether there are a set of participants who receive a significant benefit from training even if the typical participant receives relatively small benefits. Such information would allow training resources to be most