Each student taking the PISA receives a different set of questions which are of similar context and difficulty. But there is a random variation in the proportion of multiple-choice questions each student gets in their test booklet.
For instance, in 2015, some students received an exam mostly made up of multiple-choice questions (70%), while other students’ exam papers contained only 30% multiple-choice questions.
I exploited this random variation in the proportion of multiple-choice questions to investigate how gender differences in maths performance vary.
What I found
Females performed worse than males on multiple-choice questions — this was especially the case when they received an exam booklet with 60% or more multiple-choice questions.
An increase in the share of multiple-choice questions by ten percetage points (such as from 50% to 60%) increased the gender gap in maths scores by 50% in favour of boys.