The Marriage Myth

Abstract

An extensive literature suggests that marriage confers health benefits to men and women; however the literature also suggests that the healthy are more likely to marry. Using a new econometric strategy we find strong evidence that unobservable factors are significantly correlated with both health and relationship choice. Controlling for this correlation, marriage is not better for health in all cases. Rather, cohabitation benefits the health of men and women over 45 while being never married is no worse for health. Divorce harms the health of younger men while becoming widowed harms the health of younger women. Our results offer a more nuanced picture of the causal effect of relationships on health that can inform policymakers in light of changing social norms and economic constraints associated with marriage.

Keywords: endogenous dummy variables, mixed logit, dynamic panel data, health and marriage

JEL Classification: C33, C35, I1, J12