Heterogeneity in marginal returns to language training of immigrants

We estimate the effect of language training on subsequent employment and wages of immigrants under essential heterogeneity. The identifying variation is based on regional differences in language training availability that we use to instrument endogenous participation. Estimating marginal treatment effects along the distribution of observables and unobservables that drive individual participation decisions, we find that immigrants with higher gains are more likely to select into language training than immigrants with lower gains. We document up to 15% higher employment rates and 13% wage gains for immigrants with a high desire to participate but the positive returns vanish with increasing resistance to treatment. This pattern of selection on gains correlates with unobserved ability and motivation, promoting investments in education and job-specific skills that yield higher returns when complemented by language capital in the host country.