On Estimation of a Two-Sided Matching Model

This paper provides an assessment of the implementation of a canonical two-sided matching model using marriage data from the National Longitudinal Survey of Youth (1979-1998, NLSY). Results indicate that agents tend to be patient in the marriage market, and their attractiveness is more sensitive to permanent income (education) than transitory income (wage). The average marriage premium is about 1.3-1.6 per cent. Results also show that using non-random samples of spouses distorts preference parameters, that matching outcomes are sensitive to singlehood utility specifications, and that segregation in sorting could limit heterogeneity in agents' behavior. The proposed algorithm - solving for agents' marriage sets based on analytic equations and feeding the solution into the likelihood estimation - does not demonstrate sufficient flexibility to match singlehood duration data