Learning Management through Matching: A Field Experiment Using Mechanism Design

We place young professionals into established firms to shadow middle managers. Using random assignment into program participation, we find positive average effects on wage employment, but no average effect on the likelihood of self-employment. We match individuals to firms using a deferred-acceptance algorithm, and show how this allows us to identify heterogeneous treatment effects by firm and intern characteristics. We find striking heterogeneity in self-employment effects, and show that some assignment mechanisms can substantially outperform random matching in generating employment and income effects. These results demonstrate the potential for matching algorithms to improve the design of field experiments.