Objectives: Mortality risk in old age is associated with both marital status and living arrangements (with whom a person is living). This study analyses how the association between living arrangements and mortality risk varies by age and gender. Data and methods: We use a dataset extracted from the Belgian national register for the population aged 60 years and over, from 1 January 1991 to 31 December 2010, that includes more than three million persons and describes their living arrangements. First we compute and compare single-year probabilities of death for each living arrangement. Then, in order to recompose the living arrangement trajectories, we construct multistate life tables using annual probabilities of the transitions between successive living arrangements. Results: Our results confirm the protective effect of living with a spouse, and also show that at older ages living alone becomes advantageous. A crossover occurs in the mortality risks of those living alone and those living with others (not with a spouse), but large differences appear between the genders. Institutional living arrangements are largely detrimental for survival compared with private living arrangements for both genders. Discussion: The variation of mortality risk by living arrangements is gender-specific and decreases by age in relative terms. Among possible explanations, the selective role of health is paramount, as well as the presence of a spouse, who is the primary caregiver in most cases. According to this perspective, the gender gap in healthy life expectancy and the age difference between spouses contribute to explaining the gender differences. Our findings are highly relevant to social policies and the social discourse pertaining to the future needs of the elderly within the context of population ageing.