Selection Bias in Heterogeneous Returns to College in Highly Stratified Educational Systems in Europe

• In Event: 585 - Regular Session. Social Stratification and Inequality: Casual Inference

Abstract

Information on economic returns to college education has indisputable value, yet most existing estimates in Europe unrealistically assume the homogeneity of treatment effects across the population or the ignorability assumption. In this paper we relax those assumptions by using the newly developed local instrumental variable method for heterogeneous treatment effects due to unobservables presented in general form in Heckman, Urzua, and Vytlacil (2006). The method is applied to the measurement of earnings returns to college education for 28-to-38-year-olds in Austria, Czech Republic, Germany, Poland, and Slovakia (and the United Kingdom as well, as an additional point of reference) based on pooled EU-SILC data from 2005 and 2011, when data on parental background were also collected. Our results reveal positive self-selection, or ability bias, across countries and for both men and women in most, though not all, modeled conditions. In a world of essential heterogeneity, we find that the marginal returns to education for those who achieved some kind of college degree are substantively larger than the estimated returns for those who did not go to college, as well as for estimates generated from a Mincer-type model using the same data. We suggest, though cannot demonstrate, that the highly stratified nature of these educational systems create mechanisms for students to be more realistic about their educational and economic prospects, which may contribute to the size of positive self-selection estimated.

Authors

• Michael Lee Smith, Institute for Social and Economic Analyses

• Shu-Ling Tsai, Academia Sinica