For a deeper dive, my co-author Pelin Akyol breaks down the approach, findings, and implications of our study in this podcast episode with Matt Nolan:
tvhe.substack.com/p/pelin-akyo...

Our counterfactuals show: optimal exam design depends on which part of the ability distribution you care about. Time limits matter.

Also: sorting within gender may better reflect true ability—due to differences in signal production.

We model signal quality as a function of ability, question difficulty, and time constraints. The estimation reveals: gender differences aren’t due to risk preferences. They come from differences in how signals are produced under time pressure.

To understand these patterns—especially the gender gap—we estimate a structural model of how students approach multiple-choice questions. Students receive noisy signals for each option. If no option stands out enough, they skip. Skipping is shaped by confidence and risk aversion.

The results? Three big takeaways:

1. More time = higher scores & less skipping (unsurprising)
2. Effects are concentrated in “within reach” questions—low performers gain on easy Qs, high performers on hard Qs
3. Score gains are larger for males, driven by greater reductions in skipping. 🤔

Motivated by the non-monotonic findings of a natural experiment in Türkiye (which reduced time pressure in a national exam), we designed our own experiment: a multiple choice test with negative marking and differential time limits, administered to high school students.