Curbing cream-skimming: evidence on enrolment incentives

Can enrolment incentives reduce the incidence of cream-skimming in the delivery of public sector services (e.g. education, health, job training)? In the context of a large government job training program, we investigate whether the use of enrolment incentives that set different 'shadow prices' for serving different demographic subgroups of clients, influence case workers' choice of intake population. Exploiting exogenous variation in these shadow prices, we show that training agencies change the composition of their enrollee populations in response to changes in the incentives, increasing the relative fraction of subgroups whose shadow prices increase. We also show that the increase is due to training agencies enrolling at the margin weaker members, in terms of performance, of that subgroup.