Attempt limits are widely used in standardized tests for personnel selection and licensing. Using data from the U.S. Bar Exam, I show that stricter limits increase passing rates, consistent with limits acting as a sorting device: stronger candidates attempt the exam, while weaker ones withdraw. However, whether these caps are welfare-improving is not obvious.
To study this question, I develop a principal–agent model in which the principal seeks to maximize the probability of hiring a strong candidate while minimizing the probability of hiring a weak one. The model reveals a trade-off: caps deter weaker candidates but also prevent some stronger ones from eventually succeeding, so governments strongly averse to human capital loss may oppose them.
I explore how the optimal policy depends on the interaction between candidate sensitivity to attempt limits, government aversion to human capital loss, and the test’s information structure. When the test produces conclusive evidence for weak candidates, the government introduces an attempt limit if and only if candidate sensitivity to the limit is sufficiently high, regardless of its aversion to human capital loss. Conversely, when the test produces conclusive evidence only for strong candidates, an attempt limit can be optimal for any level of candidate sensitivity, provided that the government’s aversion to human capital loss is sufficiently small—the lower the sensitivity, the smaller the required aversion.
Finally, when the government can also design the test signal, it prefers to tailor the test to produce conclusive evidence for strong candidates when both aversion to human capital loss and candidate sensitivity to attempt limits are small.