Life events and life satisfaction: Estimating effects of multiple life events in combined models

How do life events affect life satisfaction? Previous studies focused on a single event or separate analyses of several events. However, life events are often grouped non-randomly over the lifespan, occur in close succession, and are causally linked, raising the question of how to best analyze them jointly. Here, we used representative German data (SOEP; N = 40,121 individuals; n = 41,402 event occurrences) to contrast three fixed-effects model specifications: First, individual event models in which other events were ignored, which are thus prone to undercontrol bias; second, combined event models which controlled for all events, including subsequent ones, which may induce overcontrol bias; and third, our favored combined models that only controlled for preceding events. In this preferred model, the events of new partner, cohabitation, marriage, and childbirth had positive effects on life satisfaction, while separation, unemployment, and death of partner or child had negative effects. Model specification made little difference for employment- and bereavement-related events. However, for events related to romantic relationships and childbearing, small but consistent differences arose between models. Thus, when estimating effects of new partners, separation, cohabitation, marriage, and childbirth, care should be taken to include appropriate controls (and omit inappropriate ones) to minimize bias. How do different life events (e.g., marriage and childbirth) affect life satisfaction? To answer this question, past studies focused on a single event or separate analyses of several events. In reality, however, life events are often grouped together as they happen over the lifespan, occur in close succession, and are linked through common causes. The current paper aims to analyze life events jointly using representative German data (SOEP; N = 40,121 individuals; n = 41,402 event occurrences). We compare three different models: First, models with each life event by itself (other events are ignored but might still bias results through undercontrol bias). Second, combined models which controlled for all other life events regardless of when they occurred (these may also introduce bias, namely, overcontrol bias). Third, the model we favored which only controlled for any preceding (but not succeeding) life events. In this preferred model, the events of new partner, cohabitation, marriage, and childbirth had positive effects on life satisfaction, while separation, unemployment, and death of partner or child had negative effects. The choice of model made little difference for employment- and bereavement-related events. However, for events related to romantic relationships and childbearing, small but consistent differences arose between models. Thus, when estimating effects of new partners, separation, cohabitation, marriage, and childbirth on life satisfaction, care should be taken to include appropriate controls (and omit inappropriate ones) to minimize bias that potentially occurs due to other life events.