The Gender Gap in Math: What are we Measuring?

Standardized assessments are widely adopted to determine educational and economic opportunities. These standardized assessments use exclusively, or in large part multiple-choice questions. But multiple-choice exams may not be adequate to compare students' competencies across genders. In this paper, I show that female students get lower marks when randomly assigned to exams with a larger proportion of multiple-choice questions. Specifically, a 10 percentage point increase in the proportion of multiple-choice questions widens the gender difference in mathematics performance by 0.026 standard deviations in favor of men, an effect that represents about 50% of the overall gender gap. Moreover, the proportion of multiple-choice questions has negative spillovers to other open-ended questions on the same exam. I show that women put differentially less effort than men in tests that contain a larger proportion of multiple-choice questions. I provide suggestive evidence that these results are driven by women's lower confidence and by the self-stereotypes that women face in traditionally male-type domains.