Fertility in cohabiting and married unions. A multi-process analysis of five countries

Mathew J. Creighton, Universitat Pompeu Fabra
Gosta Esping-Andersen, Universitat Pompeu Fabra
Roberta Rutigliano, Universitat Pompeu Fabra

We examine how novel partnership forms influence the likelihood of having children. Based on GGS data for Austria, Germany, Italy, Norway, and Spain (for Spain the 2006 fertility and family survey) we focus both on first and second births. These nations represent quite well the major variations within Europe in terms of fertility levels and welfare state support to working mothers and families. As a first step, we estimate Kaplan Meier survival curves for both cohabiting and married couples. As expected, we find that cohabitation is generally associated with fewer births. But the impact is not straightforward. Fertility differentials are larger in some countries than others, and for second births the effect is actually the opposite in two cases. Our analyses of Spain yield surprising results. Here cohabitation has grown rapidly over the past two decades, and Spain now boasts rate that lies between Austria and Germany, suggesting that it is no longer confined to an ‘avant-garde’ population. In addition, cohabitation in Spain is unexpectedly stable and, in terms of fertility behavior, is closer to Norway than to any Mediterranean 'model' (e.g. Italy). In Spain fertility behavior among cohabiting couples is basically similar to married couples – at least as far as first births are concerned. The second step utilizes a multi-process approach to assess the relationship between selection into partnership and fertility taking into account selection and the correlation among the two different processes. We estimate two simultaneous processes: transition to partnership and transition to childbirth. In this way we account for the competing risks of the transition to marriage and cohabitation. Further, by estimating the covariance of both partnership and fertility we are able to account for how changes in sample composition over time affects the association between independent variables.