Virtual Seminar in Economic Theory

Speaker: Sulagna Dasgupta (Chicago)

Title: Optimal Test Design for Knowledge-based Screening

Format: Virtual

 

Abstract: A principal seeks to maximize the chances of passing a good quality agent by testing his knowledge of a subject matter, modeled as a binary state. Higher quality agents possess more precise information. I show that the optimal test takes a "pick the correct answer'' form, with at most three options. While the efficient test only rewards correctness, without regard to the degree of ex-ante obviousness of the answer, screening leads to under-rewarding of "obvious'' answers and over-rewarding of "counterintuitive'' ones, when there is such an ex-ante obvious answer. When the principal can pick the question however, she picks one with no ex-ante obvious answer, so that the optimal test, once again, rewards only correctness. In particular, in this case the optimal test is the simple True-False test ubiquitous in the real world, regardless of all other details.

Organisers: Julia and Francesco

For more information, please contact the organisers.