Superiority of Instantaneous Decisions in Thin Dynamic Matching Markets (2206.10287v2)

Abstract: We study a dynamic matching procedure where homogeneous agents arrive at random according to a Poisson process and form edges at random yielding a sparse market. Agents leave according to a certain departure distribution and may leave early by forming a pair with a compatible agent. The primary objective is to maximize the number of matched agents. Our main result is to show that a mild condition on the departure distribution suffices to get almost optimal performance of instantaneous matching, despite operating in a thin market. We are thus the first to provide a natural condition under which instantaneous decisions are superior in a market that is both sparse and thin. This result is surprising because similar results in the previous literature are based on market thickness. In addition, instantaneous matching performs well with respect to further objectives such as minimizing waiting times and avoiding the risk of market congestion. We develop new techniques for proving our results going beyond commonly adopted methods for Markov processes.