Skilled or educated? Educational reforms, human capital and earnings *

We use OECD-PIAAC data to estimate the earnings effects of both years of education and of numerical skills. Our identification strategy exploits differential exposure to educational reforms across birth cohorts and countries. We find that education has the strongest earnings effect. A one standard deviation increase in years of education raises earnings by almost 22 percentage points (corresponding to a return to education above 7 percentage points), which compares with a lower percentage points return to an equivalent increase in numerical skills. Our results suggest that the same set of unobservables drives the accumulation of both formal years of education and numeracy skills. OLS estimates underestimate returns to human capital, consistently with the idea that educational reforms favour the human capital acquisition of abler children from disadvantaged parental backgrounds. When we consider numerical skills alone education reforms cannot identify any significant effect of skills on wages, however, when we jointly consider schooling and skills as endogenous factors in a recursive structure we find a significant role for skills in determining wages.