Designing Matching Mechanisms under Constraints: An Approach from Discrete Convex Analysis

In this paper, we consider two-sided, many-to-one matching problems where agents in one side of the market (hospitals) impose some distributional constraints (e.g., a minimum quota for each hospital). We show that when the preference of the hospitals is represented as an M-natural-concave function, the following desirable properties hold: (i) the time complexity of the generalized GS mechanism is O(|X|^3), where |X| is the number of possible contracts, (ii) the generalized Gale & Shapley (GS) mechanism is strategyproof, (iii) the obtained matching is stable, and (iv) the obtained matching is optimal for the agents in the other side (doctors) within all stable matchings. Furthermore, we clarify sufficient conditions where the preference becomes an M-natural-concave function. These sufficient conditions are general enough so that they can cover most of existing works on strategyproof mechanisms that can handle distributional constraints in many-to-one matching problems. These conditions provide a recipe for non-experts in matching theory or discrete convex analysis to develop desirable mechanisms in such settings.