Matching Dynamics with Constraints

TL;DR

This paper explores uncoordinated matching markets with local constraints, proposing a unified framework for various stable matching variants, and analyzing the existence and complexity of reaching stable states under different preference structures.

Contribution

It introduces a general class of coalition formation games with constraints, unifies several stable matching variants, and provides polynomial improvement sequences and complexity results.

Findings

Polynomial sequences to stable states in consistent instances

Exponential lower bounds when consistency conditions are violated

NP-hardness of reaching specific stable states

Abstract

We study uncoordinated matching markets with additional local constraints that capture, e.g., restricted information, visibility, or externalities in markets. Each agent is a node in a fixed matching network and strives to be matched to another agent. Each agent has a complete preference list over all other agents it can be matched with. However, depending on the constraints and the current state of the game, not all possible partners are available for matching at all times. For correlated preferences, we propose and study a general class of hedonic coalition formation games that we call coalition formation games with constraints. This class includes and extends many recently studied variants of stable matching, such as locally stable matching, socially stable matching, or friendship matching. Perhaps surprisingly, we show that all these variants are encompassed in a class of "consistent" instances that always allow a polynomial improvement sequence to a stable state. In addition, we show that for consistent instances there always exists a polynomial sequence to every reachable state. Our characterization is tight in the sense that we provide exponential lower bounds when each of the requirements for consistency is violated. We also analyze matching with uncorrelated preferences, where we obtain a larger variety of results. While socially stable matching always allows a polynomial sequence to a stable state, for other classes different additional assumptions are sufficient to guarantee the same results. For the problem of reaching a given stable state, we show NP-hardness in almost all considered classes of matching games.