Second-order statistical discrimination

The low representation of female workers in elite jobs is sometimes attributed to a tail effect: If the human capital distribution exhibits less variation among females than among males, then even with comparable average human capital there will be fewer females in the right tail than males. This paper offers an explanation for why the human capital distribution might have this property. We show that the belief that the female human capital distribution has a lower variance than the male distribution can be self-fulfilling, in that it provides individuals with incentives to invest in human capital such that the resulting distribution exhibits exactly this characteristic. If this happens, fewer females are employed in high-end jobs (a “glass ceiling” effect). The average productivity of female workers may at the same time be higher than that of male workers.

► We build a statistical discrimination model based on second moment of skills. ► In equilibrium, the human capital of women has a lower variance than that of men. ► At the same time, women can have higher average human capital than men. ► The model can explain the low representation of female workers in elite jobs. ► It complements biological explanations of second moment gender differences.