On the Survival of Some Unstable Two-Sided Matching Mechanisms : An Experimental and Computational Investigation of the Stability Hypothesis

We conduct a laboratory experiment and parallel computational adaptive agent simulations on unraveling of contract dates in two-sided matching markets inspired by the entry-level professional labor markets for medical interns in Britain. Experimental design consists of decentralized and mixed games, which adopt centralized mechanisms that have "successfully" or "unsuccessfully" been used in Britain to stop unraveling. The experiment uses an unsuccessful and unstable priority mechanism, a successful and unstable linear programming mechanism, and a successful and stable deferred acceptance mechanism (firm optimal stable mechanism). We study computational adaptive agent markets to model the learning in the experiment and to conjecture about robustness of experimental results. We find that the linear programming mechanism is not as successful as the deferred acceptance mechanism in preventing early contracts in the experiment although it performs better than the priority matching mechanism in terms of overall inefficiency