Extensive research has found that marriage provides health benefits to individuals, particularly in the U.S. The rise of cohabitation, however, raises questions about whether simply being in an intimate co-residential partnership conveys the same health benefits as marriage. Here, we use OLS regression to compare differences between partnered and unpartnered, and cohabiting and married individuals with respect to self-rated health in mid-life, an understudied part of the lifecourse. We pay particular attention to selection mechanisms arising in childhood and characteristics of the partnership. We compare results in five countries with different social, economic, and policy contexts: the U.S. (NLSY), U.K. (UKHLS), Australia (HILDA), Germany (SOEP), and Norway (GGS). Results show that living with a partner is positively associated with self-rated health in mid-life in all countries, but that controlling for children, prior separation, and current socio-economic status eliminates differences in Germany and Norway. Significant differences between cohabitation and marriage are only evident in the U.S. and the U.K., but controlling for childhood background, union duration, and prior union dissolution eliminates partnership differentials. The findings suggest that cohabitation in the U.S. and U.K., both liberal welfare regimes, seems to be very different than in the other countries. The results challenge the assumption that only marriage is beneficial for health.