different programs provide essentially the same services means excess administrative costs and less efficient provision of services. One obvious way to increase efficiency is by creating a more integrated job training system. Holzer (2008a, b) discusses the U.S. training system and possible approaches to a new system including reducing the fragmentation of programs.
Effectiveness of Government Sponsored Training Programs
One obvious question is whether government sponsored training programs are effective in the first place. Holzer (2008b) details the decline in funding for training programs relative to the size of the economy over the past decades (until the influx of funding through the ARRA). He relates this to the “work first” philosophy originally embedded in WIA and to what he sees as an incorrect but “widespread perception” that training has not been shown to be “cost‐ effective at raising future earnings of participants.” (p. 9).
While there are a number of different ways to measure and assess effectiveness, the program evaluation literature typically focuses on assessing whether the estimated benefit to the average participant (often referred to as the average treatment effect) exceeds the costs of providing the training, including both the direct cost of the program itself and the indirect cost of potential lost earnings, if any, while undertaking training. When measuring benefits, most studies focus on two outcome measures: labor market earnings and the probability of being employed after participation.
There is a large and important empirical literature on training but relatively little work done on an ongoing basis to examine the effectiveness of current job training programs, and there is substantial variability in the amount and quality of such evaluation across the various programs. Many claims about the benefits of these programs come from government‐ produced reports that rely on simplistic performance measures that are subject to manipulation by program administrators and that do not allow analysts to estimate the causal impact of the program (Barnow, 2011). This brief review focuses on studies that exploit