Gender gap variation across assessment types : Explanations and implications

Using Swedish population data, we document that girls outperform boys by a third of a standard deviation in school grades, whereas a gap of similar magnitude but opposite sign persists in SAT scores in the sample of non-randomly selected test takers. We establish that grades capture different attributes than SAT scores, which accounts for much of the variation in gender gaps. A model of SAT participation illustrates how women's greater participation-driven by traits not rewarded by higher scores-leads to their negative selection on observed and unobserved traits. We explore the quantitative importance of this mechanism and conclude that while selection is important, it fails to account for a substantial fraction of the gender gap in SAT scores, suggesting the possibility that the SAT penalizes women