Does selection bias cause us to overestimate gender differences in competitiveness?

Abstract

Experimental evidence suggests there is a substantial difference in the willingness of men and women to compete that could help explain the gender gap in labor market outcomes. The use of volunteer samples, however, raises a question about whether self-selection into experiments biases the estimated difference in competitiveness. To address it, we first measure the willingness of 1,145 individuals to compete in a classroom experiment. We then identify among them the subset of ‘lab volunteers’ by observing who accepts an invitation to participate in lab experiments. To test for the existence of selection bias, we compare the gender gap among lab volunteers to that in the population from which they were recruited. We find that selection causes us to overestimate the gender gap in competitiveness by 16 percentage points in absolute terms and, in relative terms, by a factor of 2 to 3 depending on the econometric model. We also show that selection causes us to significantly overestimate the gender gap in risk attitudes and the tendency of low-performing men to select into competition. We present evidence men and women select differently into the lab, and discuss the implications of our findings for future research.