Partnership status, health and mortality: selection or protection?

Married individuals have better health and lower mortality than non-married people. Studies show that once we distinguish cohabitants from other non-married groups, health differences between partnered and non-partnered individuals become even more pronounced. Some studies argue that partnered individuals have better health and lower mortality because of the protective effects that a partnership offers (protection); others state that partnered people have better health and lower mortality because healthy persons are more likely to form a union and less likely to dissolve it (selection). This study contributes to this debate by investigating health and mortality by partnership status in England and Wales and analysing the causes of mortality differences. We use combined data from the British Household Panel Survey and the UK Household Longitudinal Study and apply a simultaneous equations hazard model to control for observed and unobserved selection into partnerships. We develop a novel approach to identify frailty based on information on self-rated health. Our analysis shows significant mortality differentials by partnership status; partnered individuals have lower mortality than non-partnered people. We observe some selection into and out of union on unobserved health characteristics; however, the mortality differences by partnership status persist. The study offers strong support for the marital protection hypothesis and extends it to non-marital partnerships.