Long-run patterns in the spousal correlation of lifespan

Abundant research has demonstrated that substantial spousal concordance in health exists across multiple measures and populations and that the magnitude of this association is similar to spousal correlations in education and socioeconomic status. This strong spousal concordance in health results from both sorting in the marriage market and from behavioral and environmental risk factors shared by the couple. In this paper we test the hypothesis that the spousal correlation in lifespan–a summary measure of health–is similarly concordant. We create a new data set to investigate this hypothesis. We link about 15 million couples drawn from the US censuses of 1880–1940 to genealogical data available in family trees to obtain lifespan information for spouses. We document spousal correlations across time and compare spousal correlations in lifespan to opposite-gender sibling correlations in lifespan using a sample of 26 million opposite-gender sibling pairs. Surprisingly, and in contrast to what other health measures show, we find that the spousal lifespan correlation (0.063) is relatively small, though only a little smaller than the sibling correlation (0.076), which captures both genetic and early life environmental risk factors. However, we also find that the spousal correlation roughly doubles in size across cohorts (from around 0.05 for the 1880 birth cohort to 0.10 for the 1920 birth cohort). The fact that the spousal and sibling correlation track each other closely across cohorts suggests a growing importance of local risk factors that are shared by both spouses and siblings in determining lifespan, an interpretation that is also consistent with our other findings which suggest that the duration of marriage also increases spousal correlations. For the last decade, however, the growth in spousal correlation outpaces that of siblings, suggesting either increased assortative mating or an increase in the relative importance of couple-specific risk factors.