the implementation of WIA. In contrast, Hollenbech et al. and Hollenbech only have data on participants who successfully exit WIA and their data covers the period shortly after WIA was implemented. This is problematic for several reasons. First, Hollenbech et al. and Hollenbech are unable to observe earnings when an individual is actively engaged in training, so they miss the opportunity cost to participants of participating in the program. These costs can be significant for the least successful participants who spend a longer time in WIA because they struggle to find a job. Second, during the period immediately after WIA was implemented, local WIA officials were actively managing their caseload in to improve their performance measures and only allowed individuals to exit WIA once they were considered a success (Decker, 2011). Finally, neither Hollenbech et al. nor Hollenbech conduct the appropriate specification tests to determine if their comparison sample is indeed comparable to their treated sample. Taken together, these factors suggest that the estimated effects in both Hollenbech et al. and Hollenbech represent upwardly biased estimates of the effect of WIA.
The most comprehensive evaluation of TAA is Decker and Colson (1995), who find that TAA training has no positive impact on participants in the three years after they first filed for unemployment insurance. These finding are comparable to the Heinrich et al. (2008) finding for WIA dislocated workers. The major drawback of the Decker and Colson study is that their data consist of TAA participants from the late 1980s and there have been important changes in TAA since then. A more recent evaluation (Reynolds and Palatucci, 2011) finds positive earnings effects for TAA participants. A major limitation of this more recent study, however, is that the authors form a comparison sample from the Current Population Survey’s (CPS) dislocated workers survey. As has been extensively documented previously (i.e., Heckman, Lalonde and Smith, 1999) it is important to compare treated and untreated individuals in the same local labor market, something that is not possible given the limited geographic identifiers in the CPS. Without this detailed geographic information, the results reported in Reynolds and Palatucci are not reliable. Until better analysis is available, the results in Decker and Colson (1995) and Heinrich at al. (2008) are most reliable in suggesting that the TAA training program does not have a significant impact on participant wages.