Disappointing returns to education in Ghana: A test of the robustness of OLS estimates using propensity score matching

This paper uses propensity score matching to test the robustness of OLS returns to education estimates in Ghana, a country for which other researchers have found low returns to primary school. Matching provides a way to test for common support, which would fail when there are not enough similar non-educated individuals with whom to compare educated individuals. Matching also relaxes the typical linear functional form used in OLS estimates. Common support issues do not seem to bias OLS results, and matching estimates are not statistically different from OLS results, indicating that inflexible functional forms do not bias estimates. These results indicate that, at least in this application, OLS provides reasonable estimates of education returns.