Abstract:
This paper proposes a simple solution to the independence of irrelevant alternatives (IIA) problem in Choo and Siow (2006) model, overcoming what is probably the main limitation of this approach. The solution consists of assuming match-specific rather than choice-specific random preferences. The original marriage matching function gets modified by an adjustment factor that improves its empirical properties. Using the American Community Survey, I show that the new approach yields significantly different results affecting the qualitative conclusions of the analysis. The proposed solution to the IIA problem applies to other settings in which the relative "supply" of choices is observable.