A number of papers .nd that changes in schooling are not correlated with changes in per capita income. Two non-competing interpretations that have been given are that the social return on schooling is close to zero and the measurement error of changes in schooling is high. This paper shows that the lack of significance of schooling is threefold. First, there is a problem of a proper definition of the way in which years of schooling should enter in a production function. Second, collinearity between physical and human capital stocks seriously undermines the ability of educational indicators to display any significance in panel data estimates. And third, failure to cope with measurement error and endogeneity produces biased estimates. As opposed to the earlier empirical literature, the social return of schooling is positive and significant, but no Lucas-type externalities are observed.