In this paper we carry out a meta evaluation of the international evaluation literature regarding active labour market policies (ALMPs). Many attempts have been made to estimate the net impact of reintegration measures on the individual job entry chance. So, the question is what conclusions can be drawn from the literature. How big is the net effect of reintegration measures? Our paper differs from earlier meta analyses of ALMP impact evaluations in the fact that we try to make inferences about the size of the net effect. To that end we analyze the size distribution of the net impact estimates resulting from the international evaluation literature. In our analysis we distinguish between different types of reintegration measures. Furthermore, we carry our regression analyses in which we explain the estimated effect found in individual studies from the type of measure, the labour market context and characteristics of the evaluation method used.
A problem with the analysis is that different studies measure different things. Using the theoretical framework of hazard models, we analyze to what extent the different approaches can be related to each other and whether it makes sense to combine the different studies in one meta analysis.
The results indicate that on average the net impact of reintegration measure is fairly small. As a result of reintegration job entry chances are probably not increased by more than 3 percentage points on average. The case is most convincing for training and counseling with average net effects ranging from 5.7 to 9.7 percentage points. The pos¬itive results for training are surprising. However, our sample hardly contains experi¬mental evaluations of training, which are usually seen as the most reliable evaluations. Overall we find that the net effect estimate tends to be smaller if an experiment is used for the evaluation. Hence, the results for training may be too rosy. If we account for the method used train¬ing comes out less well, while sanctions (which are treated as one type of measures) perform better. Wage subsidies and job creation come out poorly. We also find that the net impact of ALMPs depends on the socio-economic situation: it tends to be lower when unemployment is low than during a recession period.
We see this paper as a first step and want to exploit the literature further. Recently, a lot of studies using a more developed methodology for non-experimental evaluation have appeared. It will be interesting to see to what extent inclusion of these studies in the meta analysis will alter the results.