The mathematics of sex and marriage, revisited

We analyze the problem of modeling marriages in a two-sex model of population dynamics. We first deal with the problem of incomplete and inconsistent census data and then use a simulator to compare the performance of a variety of marriage functions in modeling births and couples during the ten-year period between consecutive U.S. censuses. Unlike most empirical methods for comparing marriage functions based on goodness of fit, the differences in the projections of the various functions in our method are of the same magnitude (or even smaller) than the errors between the projected and real data. We observe that for the population of the United States, the harmonic mean function frequently found and used in the literature is a quite poor performer when compared with many other functions in the family we use.