Marital status, health and mortality: the role of living arrangement

Paul Boyle, Peteke Feijten and Gereltuya Altankhuyag, University of St Andrews and Gillian Raab and Lin Hattersley, General Register Office for Scotland

[Project number 30054]

We want to investigate the effect of marital status and living arrangement on health and mortality risk. In terms of model effects, we will test the hypotheses that:

1. Those who live alone are more likely to die than those who live with others.

2. Unmarried adults (separated, divorced or widowed) who live with others are no more likely to die than married adults who live with others.

3. Living with other adults is more protective than living with children.

The purpose of the study is to contribute to the body of literature on marital status and health, by adding a new element to the explanation of why marriage is associated with better health.

It is widely known that being married lowers the risk of dying, and in a number of previous studies the protective effects have been shown to be greater for men than women. However, we have begun to investigate whether it is marriage, or living with others, which is protective for health outcomes. Using data from the 2001 Sample of Anonymised Records (SAR), Boyle and Feng (submitted) do indeed show that for men it is not being married, but living with others which reduces the risk of limiting long-term illness (LLTI). Thus, single, separated and divorced men who live with others have no higher risk of LLTI than married men. This study also shows that for women the risk of LLTI is considerably higher for those who live alone, regardless of marital status but, interestingly given previous studies which suggest that men benefit from being married more than women, unmarried women who live with others have slightly worse health outcomes than married women who live with others.

This project will use a similar methodology to that used by Gardner and Oswald (2004) who used the British Household Panel Study (BHPS) to compare marital status and other characteristics in 1991 with subsequent death between 1993 and 2001. However, their study did not explore living arrangements and only included a sample of 3,695 with 741 deaths being recorded in the period. By utilising the ONS LS, we will increase the sample and the number of deaths substantially. We will also examine whether the effects of marital status and living arrangements on subsequent death varies following the 1971 and 1981 Censuses (and, if mortality numbers are large enough, also mortality risk following the 2001 census). These will be separate analyses per 10-year period, to see whether effects change over time. To fully utilize the longitudinal power of the ONS LS, we also intend to analyse mortality risk of 1971 and 1981 ONS LS members over time (up to death or censoring), and assess whether changes in living arrangement between census years affects their mortality risk. Lastly, we will explore the effect of living arrangement on health, by comparing the LLTI for those who were living alone, or not, in 1991 and 2001.

Top of page