Optimal Welfare-To-Work Programs with Worker Profiling

Profiling has a pivotal role in welfare-to-work programs in classifying jobless workers according to their abilities and assigning them to suitable labor-market policies. This analysis designs an optimal profiling policy, by embedding dynamic learning about a recipient’s ability within principal-agent framework. In optimal profiling, a certain proportion of low-skilled workers may be persuaded that they are in fact high-skilled and referred to delegated search, together with actual high-skilled workers (positive type II error). This occurs whenever the government prefers that overly optimistic low-skilled workers search for jobs (with low incentive costs) rather than referring them to passive labor-market policies. On the other hand, a high-skilled worker is never classified as being low-skilled, nor she is referred to a passive policy (no type I error). In the US, an optimal profiling strategy would generate annual savings that range from around $0.6 million (South Dakota) to $201.1 million (California)