When Scale and Replication Work: Learning from Summer Youth Employment Experiments

Two sources of treatment heterogeneity can undermine the scale-up and replication of successful human capital interventions: variation in the treatment itself and changes to the population served. This paper combines two new summer youth employment experiments in Chicago and Philadelphia with previously published evidence to show how repeated study of an intervention as it scales and changes contexts can guide decisions about public investment. Results show that these programs generate consistently large proportional decreases in criminal justice involvement, even as administrators recruit additional youth, hire new local providers, find more job placements, and vary the content of their programs. Using both endogeneous stratification within cities and variation in 62 new and existing point estimates across cities uncovers a key pattern of individual responsiveness: impacts grow linearly with the risk of socially costly behavior each person faces. Identifying more interventions that combine this pattern of robustness to treatment variation with bigger effects for the most disconnected could aid efforts to reduce social inequality efficiently.