Mortality, lifestyle and socio-economic status

https://doi.org/10.1016/j.jhealeco.2007.03.001Get rights and content

Abstract

This paper uses the British Health and Lifestyle Survey (1984–1985) data and the longitudinal follow-up of May 2003 to investigate the determinants of premature mortality in Great Britain and the contribution of lifestyle choices to socio-economic inequality in mortality. A behavioural model, which relates premature mortality to a set of observable and unobservable factors, is considered. A maximum simulated likelihood (MSL) approach for a multivariate probit (MVP) is used to estimate a recursive system of equations for mortality, morbidity and lifestyles. Health inequality is explored using the Gini coefficient and a decomposition technique. The decomposition analysis for predicted mortality shows that, after allowing for endogeneity, lifestyles contribute strongly to inequality in mortality, reducing the direct role of socio-economic status. This contradicts the view, which is widely held in epidemiology, that lifestyles make a relatively minor contribution to observed socio-economic gradients in health.

Section snippets

The economic framework

The purpose of this paper is to investigate whether or not lifestyle choices have a direct and significant influence on the risk of mortality and to assess their contribution to the observed socio-economic gradient of mortality. It is well documented that individual socio-economic status is correlated with health and that health inequalities in the population are associated with socio-economic differentials (see e.g., van Doorslaer and Koolman, 2004). However, lifestyles might capture part of

A model of health and lifestyles

We use a dynamic programming approach to define the economic problem.4

Data and empirical model

This paper uses data from the first wave of the Health and Lifestyle Survey. The HALS was carried out between Autumn 1984 and Summer 1985, in two home visits (the second one by a nurse). The questionnaire was designed and piloted by a study team at the University of Cambridge School of Clinical Medicine and funded by the Health Promotion Research Trust. The sample design permits inference about the British population, aged 18 and over in 1984–1985.

In May 2003, 19 years later, the data has been

Variables and sample

The lifestyle variables indicate whether the individual is a non-smoker, a prudent consumer of alcohol, eats breakfast, sleeps the “optimal” numbers of hours, is not obese, and did sufficient physical activity in the last fortnight12

A multivariate probit model for mortality

Our model consists of a recursive system of equations for lifestyles, morbidity and mortality. Its most important feature is that the random components of the lifestyle equations are allowed to be freely correlated with the random component of the mortality equation. If there are unobservable individual characteristics, influencing both individual's healthy behaviours and their probability of death, the model is able to take them into account.

Endogeneity can arise with the inclusion of

Results for the multivariate probit models

First, following Wilde's (2000) result on identification of multiple equation probit models, we tried to estimate a multivariate model where each equation has the same regressor matrix. However, the MSL of the multivariate probit does not converge to a global maximum. Therefore, following Schmidt (1981) and Maddala's (1983) approach, we changed the specification of the model by setting some exclusion restrictions.

We compared four different sets of exclusion restrictions and used information

Results from a decomposition analysis of total health Inequality

In this section we compute the Gini coefficient to give a robust measure of health inequality. Since our mortality indicator is a binary variable, we use predicted mortality, that is the linear index for death predicted from probit models of the mortality equation, to analyze total health inequality, van Doorslaer and Jones (2003), dealing with an ordered categorical dependent variable for SAH from the Canadian National Population Health Survey, used predicted health from ordered probit or

Conclusion

We propose a simple behavioural model where the economic agent maximizes lifetime utility. A value function is used to relate future utility to survival probability. Health investments decisions are assumed to influence longevity. For the empirical model we use the British HALS (1984–1985) data and the longitudinal follow-up of May 2003 to investigate the determinants of premature mortality risk in Great Britain.

We relate the risk of mortality to a set of observable and unobservable factors.

Acknowledgements

The authors wish to thank Paul Contoyannis, Fabrice Etilé, Martin Forster, Angel López, Chiara Monfardini and Nigel Rice for their suggestions and comments. Data from the Health and Lifestyle Survey (HALS) were supplied by the ESRC Data Archive.

References (42)

  • M.G. Marmot et al.

    Inequalities in health—specific explanation of a general pattern?

    The Lancet

    (1984)
  • M.G. Marmot et al.

    Health inequalities among British civil servants: the Whitehall II study

    The Lancet

    (1991)
  • M. Marmot et al.

    Social inequalities in health: next questions and converging evidence

    Social Science Medicine

    (1997)
  • S. Morris et al.

    Inequity and inequality in the use of health care in England: an empirical investigation

    Social Science and Medicine

    (2005)
  • J. Mullahy et al.

    Air pollution, cigarette smoking, and the production of respiratory health

    Journal of Health Economics

    (1990)
  • J. Mullahy et al.

    Employment, unemployment, and problem drinking

    Journal of Health Economics

    (1996)
  • A. Wagstaff et al.

    On decomposing the causes of health sector inequalities with an application to malnutrition inequalities in Vietnam

    Journal of Econometrics

    (2003)
  • J. Wilde

    Identification of multiple equation probit models with endogenous dummy variables

    Economic Letters

    (2000)
  • Adda, J., Lechene, V., 2001. Smoking and Endogenous Mortality: Does Heterogeneity in Life Expectancy Explain...
  • Adda, J., Lechene, V., 2004. On the Identification of the Effect of Smoking on Mortality. CeMMAP working papers...
  • R. Auster et al.

    The production of health: an exploratory study

    Journal of Human Resources

    (1969)
  • Cited by (0)

    View full text