individual‐level micro data and utilize a methodology and data that are appropriate for estimating the causal impact of a program.
Of the primary federal job training programs discussed above, WIA has been the subject of the most scrutiny, although unlike its predecessor program, JTPA, there has never been a random assignment evaluation of WIA.8 However, there have been two meta‐analyses showing that, when done correctly, results from non‐experimental evaluations closely match the results from experimental analyses (see Card, Kluve and Webber, 2009 and Greenberg, Michalopoulos and Robbins, 2006).
The most rigorous evaluation of WIA to date are the studies by Heinrich et al. (2008), Hollenbeck et al. (2005) and Hollenbeck (2009). For participants in the adult worker program, Heinrich et al. find that WIA increases quarterly earnings by approximately $400 per quarter for the 16 quarters after beginning the program. They also find that WIA increases the probability of being employed by approximately 8‐10 percentage points per quarter. Given the cost of the program of around $4,000 per participant, these estimates suggest that the benefits accruing to participants in the WIA adult program exceed the costs of the program.
Heinrich et al. find appreciably smaller benefits for participants in the WIA dislocated worker program once they take into account nonrandom selection into the program. For dislocated workers, the benefits are negative for up to 10 quarter after entry and never significantly exceed zero. The estimated increase in the probability of being employed is positive and significant toward the end of the sample period but never exceeds 3‐4 percentage points. Based on these estimates, it does not appear that the benefits of the dislocated worker program exceed the costs for the average participant.
The Hollenbech et al. (2005) and Hollenbeck (2009) studies tend to find larger effects for both the adult program and the dislocated worker program. The primary reason for the different findings lies in the data used in these studies. Heinrich et al. have data on all WIA participants starting when they enter the program and have data starting several years after
A randomized evaluation of WIA has been funded and is beginning to be implemented (see Bellotti et al., 2009). 11