Wioletta Grzenda , Warsaw School of Economics
A frequency approach is commonly used to model fertility intentions and the realisation of these intentions. In addition, two states are generally considered: having children and childlessness. In this study, we are expanding this approach by proposing the use of discrete choice models to analyse intentions measured on a broader than dichotomous scale, and we propose to use extended Cox models to examine fertility decisions. An important aspect of the study is to take into account both the variability of the features over time and their varying impact on the studied phenomenon. In this work, multinomial logit models and survival models were considered in the Bayesian approach. This approach made it possible to analyse and compare the posterior distributions of parameters. As a result, it was possible to indicate statistically significant differences in the impact of the examined factors on the transition from the state of childlessness to the state of having one child and on the transition from the state of having one child to the state of having two children. Markov Chain Monte Carlo (MCMC) methods were used to determine the posterior expected values. A comparison of the fertility intentions of the surveyed women with the actual events of the births of children revealed major differences between these intentions and their realisation. The questions what factors and to what extent influenced the fact that Polish women declared intention to give birth to a child, and what factors influenced the fact that they actually did so, were also answered.
Presented in Session 32. Methods for Fertility and Life Course Analysis. Session within the Project of Excellence of the Department of Statistical Sciences of Padova University.