Dynamic screening (2204.13392v1)

Abstract: We study dynamic screening problems where elements are subjected to noisy evaluations and, in every stage, some of the elements are rejected while the remaining ones are independently re-evaluated in subsequent stages. We prove that, ceteris paribus, the quality of a dynamic screening process is not monotonic in the number of stages. Specifically, we examine the accepted elements' values and show that adding a single stage to a screening process may produce inferior results, in terms of stochastic dominance, whereas increasing the number of stages substantially leads to a first-best outcome.