Glass ceilings in research: Evidence from a national program in Uruguay

Female researchers have lower probability than male researchers of being accepted into the largest national research support program in Uruguay. Age, scientific productivity, teaching activities and previous applications explains 5.2 percentage points of the 7.1 point gender acceptance probability gap. The remaining 1.9% can be attributed to gender discrimination. This phenomenon is stronger at the top 2 levels (out of 4) of the program evidencing glass ceilings. Results are robust to issues of simultaneity (research productivity affecting probability of being accepted and vice versa), joint determination and correlation of variables and productivity effects at early stages of career development. The paper tests four hypotheses that are likely to produce a glass ceiling in any R&D incentive schemes: male overrepresentation in the initial setup of the program, male overrepresentation on evaluation committees and two types of field-level effects (a pure composition effect without discrimination and differentiated discriminatory effects by fields). We show evidence of gender bias in the initial setup of the program and bias in the gender structure of committees. Nevertheless, these hypotheses have little quantitative power to explain the glass ceiling. The pure field composition effect is also not important. We find solid evidence of glass ceilings in the three areas where women are most active: health-related sciences, natural sciences and humanities. On the other hand, we find no such effects in social sciences, agricultural sciences or engineering.