Explaining gender gap variation across assessment forms

In Sweden, females outperform males on compulsory and high school GPAs by a third of a standard deviation, while males outperform females on the Swedish SAT by the same magnitude. We establish that GPAs capture different attributes and skills compared to SAT scores. Differences in motivation and effort explain up to 60 percent of the female advantage in GPAs, while cognitive skills explain 40 percent of the male advantage in SAT scores. The latter is accounted for by differential self-selection into taking the SAT. Our findings imply large effects of the choice of university admission criterion on admitted students' characteristics.