Partnership and mortality in mid and late life: Protection or selection?

The main goal of this paper is to address how different partnership statuses impact the likelihood of death among mature adults and elderly persons in Spain circa 2012 using a massive new dataset of administrative registers linked to census data. First, gross and net effects of having a partner on mortality risks of partnered and non-partnered persons are evaluated; then the characteristics and the importance of selection and protection effects of marriage and partnership with regard to the likelihood of death are assessed. We make use of exact matching methods in order to avoid the selection bias associated with the non-random assignment of persons to different partnership statuses. Protection effects decline gradually with age, but always remain positive. Selection effects show a far more pronounced decline with age leading to a pattern in which selection is much stronger than protection during the mature adult ages, but then disappear entirely and even become negative as people age. While both sexes show similar patterns, the protection effect is slightly higher among men while the selection effect is much higher among women, especially before 65 years of age.