Extended Data Table 4.

Gender differences in attribution

The sample consists of 17,929,271 potential article authorships and 3,203,831 potential patent inventorships. The top panel is estimated on the sample of potential article authorships and the bottom panel is estimated on the sample of potential patent inventorships. The dependent means are 3.18% and 1.31%, respectively. Specification (1) includes none of the control variables discussed above and estimates the gender gap to be 1.97 and 1.50 percentage points for articles and patents. Specifications (2-5) gradually introduce controls for days worked, PI status, publication month, job title, field, and team (which subsumes field). The observations are weighted by the inverse number of teams per employee times the inverse number of potential articles or patents per employee. Each coefficient is tested against the null hypothesis of being equal to 0 using a two-sided t-test. We do not adjust for multiple hypothesis testing. Standard errors are clustered by team and employee and are in parentheses. Statistical significance indicated by * p < 0.10, ** p < 0.05, *** p < 0.01.