Estimating men’s fertility from vital registration data with missing values

Comparative perspectives on men’s fertility are still rare, in part because vital registration data are often missing paternal age information for a substantial number of births. We compare two imputation approaches that attempt to estimate men’s age-specific fertility rates and related measures for data in which paternal age information is missing for a non-negligible number of cases. Taking births with paternal age information as a reference, the first approach uses the unconditional paternal age distribution, while the second approach considers the paternal age distribution conditional on the maternal age. To assess the performance of these two methods, we conduct simulations that mimic vital registration data for Sweden, the United States, Spain, and Estonia. In these simulations, we vary the overall proportion and the age selectivity of missing values. We find that the conditional approach outperforms the unconditional approach in the majority of simulations and therefore should be generally preferred.