Non-technical summary
Marital dissolution is associated with many changes in the lives of family members. To fully understand these changes researchers need access to surveys that collect data on the same individuals over time, usually referred to as panel surveys. Such data enable examination of marital dissolution as a process that occurs over time with couples going through a number of stages including preparing to separate, separation, legal divorce and repartnering. We know that marital dissolution has a big impact in relation to economic wellbeing with evidence that women fare much worse than men in these areas following marital breakdown. In this paper we use panel data from Australia to examine the changes in income and wealth for women and men in relation to marital dissolution.
We use statistical models suited for data with repeated observations on the same individuals. Panel surveys aim to interview the same people on a regular basis, but often fail to reach all members as some may refuse to be re-interviewed or are unable to be contacted. Weights can be used to rebalance the participating sample when other sample members do not participate. Panel surveys also need to devise rules about which members of the households to follow when children are born or households break up and sample members form new households. The rules developed to respond to these events have important implications for modelling techniques and the reliability of results. Many researchers ignore these issues but in our paper we explicitly examine how they affect our results.
We examine four stages of marital dissolution. We find that women experience much greater financial loss than men, but lower educated women experience higher personal income gains during the period of separation and divorce than higher educated women. Most of these gains are sustained through subsequent repartnering. Our results are broadly robust to variations in sample design and non-response. Nevertheless, there are some differences that alter the conclusions when the sample is weighted or restricted to the original sample.