Active labor market policy evaluations: a meta-analysis

This paper presents a meta-analysis of recent microeconometric evaluations of active labor market policies. Our sample consists of 199 program estimates drawn from 97 studies conducted between 1995 and 2007. In about one-half of these cases we have both a short-term impact estimate (for a one-year post-program horizon) and a medium-term estimate (two-year horizon). We characterize the program estimates according to the type and duration of the program, the characteristics of the participants, and the evaluation methodology. Heterogeneity in all three dimensions affects the likelihood that an impact estimate is significantly positive, significantly negative, or statistically insignificant. Comparing program types, subsidized public sector employment programs have the least favorable impact estimates. Job search assistance programs have relatively favorable short-run impacts, whereas classroom and on-the-job training programs tend to show better outcomes in the medium-run than the short-run. Programs for youths are less likely to yield positive impacts than untargeted programs, but there are no large or systematic differences by gender. Methodologically, we find that the outcome variable used to measure program effectiveness matters. Evaluations based on registered unemployment durations are more likely to show favorable short-term impacts. Controlling for the outcome measure, and the type of program and participants, we find that experimental and non-experimental studies have similar fractions of significant negative and significant positive impact estimates, suggesting that the research designs used in recent non-experimental evaluations are unbiased.