"What is the effect of exposing motivated youth to firm management in practice? To answer this question, we place young professionals for one month in 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. Within the treatment group, we match individuals and firms in batches using a deferred-acceptance algorithm. We show how this allows us to identify heterogeneous treatment effects by firm and intern. We find striking heterogeneity in self-employment effects, but almost no heterogeneity in wage employment. Estimates of marginal treatment effects (MTE) are then used to simulate counterfactual mechanism design. We find that some assignment mechanisms substantially outperform random matching in generating employment and income effects. These results demonstrate the importance of treatment heterogeneity for the design of field experiments and the role of matching algorithms in intervention design."

"We conduct a business plan competition to test whether survey instruments or panel judges are able to identify the fastest growing firms. Participants submitted six- to eight-page business plans and defended them before a three- or four-judge panel. We surveyed applicants shortly after they applied and one and two years after the competition. We use follow-up surveys to construct measures of enterprise growth and baseline surveys and panel scores to construct measures of enterprise growth potential. We find that a measure of ability correlates strongly with future growth, but that the panel scores add to predictive power even after controlling for ability and other survey variables. The survey questions have more power to explain the variance in growth. Participants presenting before the panel were given a chance to win customized management training. Fourteen months after the training, we find no positive effect of the training on growth of the business."