Two Centuries of Marriage and Mortality in the United States: An Analysis of New Family History Data

Sven E. Wilson, Brigham Young University

This paper estimates the relationship between marital status and mortality using a large sample of family histories from the United States. The data under analysis contain birth, death and marriage dates on a sample of family-linked individuals. Using both linear models of life expectancy and parametric and semi-parametric hazard models, the analysis finds strong and consistent benefits of marriage that are highest in early adulthood, but diminish steadily over the life course. The effects of being never married and widowed are similar, and protective effects of marriage exist for both men and women, though they are stronger for men. The marriage effect is also shown to be relatively stable across birth cohorts. Sibling-level random effects are incorporated into both the linear models and the parametric hazard model. In general, the random effects have little impact on the coefficient estimates.