Statistical profiling as a targeting tool: Can it enhance the efficiency of active labor market policies?

Digitization has spurred interest in the potential of statistical profiling to improve the targeting of active labor market policies. Despite growing adoption, empirical evidence on the effectiveness of such profiling in program allocation is scarce. We evaluate a semi-automated statistical profiling model in Austria that aims to target policies based on predicted reemployment prospects (low, medium, high). Our analysis shows that a reallocation of resources from low-chance to medium-chance segments, as envisaged by the Public Employment Service, would not yield the desired efficiency gains. Employment programs have a stronger impact on jobseekers with low job prospects than on those with medium prospects, and training programs are not consistently less effective in the low-chance segment either. Our findings suggest that the focus should remain on the most disadvantaged, both from an efficiency and an equity perspective. They caution against relying on overly coarse profiling and stress the need for nuanced targeting strategies.