Travail de recherche/Working paper
Résumé : We extend the analysis of (Echenique, Lee, Shum, and Yenmez, 2013) by looking at the testable revealed preference restrictions for stable aggregate matchings with non-transferable utility. We rephrase their revealed preference test in terms of a bipartite graph. From this, we obtain an simple condition that verifies whether a given matching is rationalisable. Next, for a matching that is not rationalisable, we show how to find the minimal number of matches that needs to be removed in order to restore rationalisability. This produces a goodness-of-fit measure that indicates how close a matching is to being rationalisable. We also show that the related problem of finding the minimal number of types that we need to remove in order to obtain rationalisability is NP-hard. We provide two illustrations to demonstrate the usefulness of our results.