BACKGROUND: The past 30 years have seen major changes in the living arrangements of older people and in long-term care policy in England and Wales. Co-resident family care and institutional care can be alternatives for seriously disabled older people, so changes in the availability of either may affect demands for the other. AIMS: To analyse changes in older people's living arrangements in three successive decades in England and Wales and compare the subsequent mortality of older people living with relatives, those living alone or with a spouse, and those living in institutions. METHODS: Cross-sequential analysis of household circumstances at the beginning and end of three decades using multinomial logistic regression of data from the Office for National Statistics Longitudinal Study. Poisson regression analysis of subsequent mortality. RESULTS: Risks of moving to an institution in 1991-2001 relative to living alone or in a couple were lower than in 1981-1991 for women and men and also lower in comparison with living with relatives for women. Chances of living with relatives rather than alone or in a couple showed a downward trend over time. Institutional residents had higher mortality than those living alone/in a couple or those with relatives. This excess was greater in 2001-2005 than in previous periods; even so, 26% of male and 36% of female institutional residents in 2001 survived 3 years. CONCLUSION: Policy changes are important influences on use of institutional care. The health status of older people living with relatives and those in institutions is not equivalent.