Gender differences in leaving questions blank on high-stakes standardized tests

Around the world, multiple-choice tests are widely used as part of high-stakes examinations. To counteract lucky guessing, many of them have instituted a penalty for wrong answers. In this paper, we use administrative data from Turkish college admissions test to study the heterogeneity in gender differences in tendency to leave questions blank across subjects, difficulty levels, and stakes. Exploiting the tracking system and using the resulting variation in the effective guessing penalty across different test sections, we find that female test-takers skip significantly more questions than male test-takers in the quantitative track while we do not find a significant difference in other tracks. Among quantitative track students, the gender gap is larger in Math and when questions are more difficult while it reverses in Literature. We also find that self-assessment is related to skipping behavior and explains part of the gender gap. Male test-takers are more likely than female test-takers to report that they are good at Math, Science, and Social Science after conditioning on their number of correct answers in the corresponding test sections. This gender gap, consistently with the one in skipping behavior, reverses when it comes to Literature. Differently from previous literature, our findings suggest that the magnitude and the sign of the gender gap in answering questions under uncertainty is context dependent.