Unemployment and Mortality: A Longitudinal Study in the United States, 1968-1992

Background. Studies have consistently found evidence of an association between unemployment and all-cause mortality. The association appears to follow a gradient, with more or longer unemployment spells associated with higher mortality rates than fewer or shorter (or no) unemployment spells. The association persists after simultaneously adjusting for potential confounders measured at baseline. Studies have found that the local unemployment rate modifies the relationship, with two of four studies finding that the mortality rate for the unemployed group was higher in areas or periods with low unemployment rates. Methods. I examined the relationship between unemployment and mortality in 2868 male household heads followed for up to 25 years and 2676 male household heads followed for up to 16 years as part of the Panel Study of Income Dynamics. I used annual measures of unemployment as time-varying variables in Cox regression analyses and controlled for annual measures of potential confounders. I looked for an overall association, for a gradient in the association, and for variation in the association according to the local unemployment rate and the time period. Results. Men who were unemployed one or more times on the days of the annual survey had a higher hazard of death while in the labour force than men who were working on the days of the survey (hazard ratio 3.23 [1.61-6.48]) and had a higher hazard of death while in the labour force or retirement than men who were working or retired (hazard ratio 3.47 [1.41-8.56]). Men who experienced longer unemployment spells died earlier while in the labour force or retirement than those who experienced shorter (or no) unemployment spells (hazard ratio 1.03 [1.00-1.05] for a one week change in the duration of unemployment). No clear relationship emerged between the number of unemployment spells and mortality. Men who were unemployed on the day of the survey and lived in an area with a low local unemployment rate tended to die earlier while in the labour force than men who were working on the day of the survey or who lived in an area with a high unemployment rate or both (hazard ratio 3.50 [0.78-15.67]). Discussion. Unemployment spells and their duration increase the hazard of dying. These associations persisted after adjusting for annual measures of potential confounders. The study provides suggestive support for targeting assistance at unemployed men in areas with low unemployment rates. Future research should explore the behavioural and biological mechanisms through which unemployment could affect mortality and examine possible ameliorating factors.