Small effects of selective migration and survival in retrospective studies of fertility

Gunnar Andersson, Stockholm University

In this paper, we assess the accuracy of fertility estimates that are based on the retrospective information that can be derived from an existing cross-sectional population. Swedish population registers contain information on the childbearing of all people ever living in Sweden and thus allow us to avoid any problems of selectivity by virtue of survival or out-migration when we estimate fertility measures for previous calendar periods. We calculate two types of fertility rates for each year in 1961-1999: (i) rates that are based on the population that were living in Sweden at the end of 1999 and (ii) rates that also include information on people who had died or emigrated before the turn of the twentieth century. We find that the omission of information on individuals who have emigrated or died, as the situation would be in any demographic survey, most often have negligible effects on our fertility measures. However, first-birth rates of immigrants gradually become more biased as we move back in time from 1999 so that they increasingly tend to over-estimate the actual fertility of that population.