Who Marries Whom and Why

Using a transferable utility model of the marriage market, the paper derives a statistic to measure the systematic welfare gain to any two types marrying each other relative to them not marrying. This statistic also defines a non-parametric marriage matching function which can be used to (1) decompose the distribution of marital matches into the share attributable to preferences and that attributable to the avail-ability of different types of participants and (2), to predict changes in who marries whom as the distributions of types of participants change. This marriage matching function is estimated with data from the 1970 US Census and Vital Statistics. The estimated model is able to largely predict marital behavior by age and education in 1980. It over predicts the marriage rate of young adults for all education groups. Our analysis suggests that the gains to marriage for young adults fell substantially from 1970 to 1980. ∗ We thank Pierre Andre-Chiappori, Angelo Melino and Adonis Yatchew, and seminar participants at Bosto