The Causality and Characterization of the Widowhood Effect

Abstract

Researchers from a variety of fields have noted a sharp rise in mortality for widows soon after the death of their spouse, a relationship that has often been called the widowhood effect. Because of assortative mating, married couples tend to share many of the same lifestyle characteristics, so this result may reflect correlation rather than a causal relationship. In this dissertation, I attempt to decipher whether the widowhood effect reflects a causal relationship. The key innovation in the dissertation turns on the notion that some causes of death reveal more information about the surviving spouse than others. In the extreme, if a cause of death was randomly assigned, then these types of deaths could be used to identify the death of a spouse does in fact raise mortality of the surviving spouse. In practice, we cannot specify what causes of death are randomly assigned, but instead, we can identify those that are uncorrelated with observed characteristics. Specifically, I use data from the National Longitudinal Mortality Survey and the National Health Interview Survey Multiple Cause of Death supplement to create longitudinal datasets of married couples, aged 50 to 70. I initially use this sample to identify those causes of death that are predicted by socio-economic status (income, occupation and education) and those that are not. I refer to these two types of deaths as informative and uninformative causes of death, respectively. If the heightened mortality of surviving spouses is subject to an omitted variables bias, in single-equation models, I should find a greater excess mortality for informative deaths than for uninformative ones. If omitted variable bias is not a serious concern, I should see little difference between the two types of widows. In Cox proportional hazard models, I find for men the death of a spouse from an uninformative cause has only a slightly smaller impact on mortality than a death from an informative cause. The findings suggest a 30 percent increase in male mortality as a direct result of becoming a widow. I do not find similar evidence for women; in fact, the results show no marriage protection effect.