Resit exams are a bad idea

What should be the best design for important exams? One attempt and you are out? Resit allowed once? Twice? Maximum number of attempts over all exams? If you ever have to go through a meeting about this type of rules, you will be surprised how opinionated people are about this. I do not think it is because they can base their views on hard evidence, but rather that they like the system they (successfully) went through themselves in their studies.

Peter Kooreman asks whether allowing students to resit on failed exams within the same academic year makes them learn more, the ultimate objective of an exam. His exercise is theoretical and looks at a student who does not like working but wants to pass. The probability of passing an exam depends on the effort. If there is only one exam, the student provides more effort than for the first exam in a set of two chances. For the second exam, the effort should be equivalent to the lone exam. Thus with two exams, the probability of passing is higher. Effort is lower, and likely much lower. After all, some students get through the first exam with luck and little study while the others get seriously about it only one the second exam.

This short study misses a couple of important ingredients, though. The first is that it assumes that students are risk neutral. My casual empiricism tells me that students are quite nervous about exams, indicating quite a bit of curvature. Also, students have a clear preference for passing exams earlier than later. Risk aversion increases the distance between the two exam schemes. Impatience reduces it and could change the ordering. Finally, it would be great to see some empirical evidence. But the latter is likely asking for a bit too much here.