On the survival of some unstable two-sided matching mechanisms

Abstract

In the 1960s, three types of matching mechanisms were adopted in regional entry-level British medical labor markets to prevent unraveling of contract dates. One of these categories of matching mechanisms failed to prevent unraveling. Roth (1991) showed the instability of that failing category. One of the surviving categories was unstable as well, and Roth concluded that features of the environments of these mechanisms are responsible for their survival. However, Unver (2001) demonstrated that the successful yet unstable mechanisms performed better in preventing unraveling than the unsuccessful and unstable category in an artificial-adaptive-agent-based economy. In this paper, we conduct a human subject experiment in addition to short- and long-run artificial agent simulations to understand this puzzle. We find that both the unsuccessful and unstable mechanism and the successful and unstable mechanism perform poorly in preventing unraveling in the experiment and in short-run simulations, while long-run simulations support the previous Unver finding.