Use the Table of Summaries to link to the summaries either by author or by risk factor.
Abstract | Bibliography | Introduction | Table of
Summaries
Differential mortality in the UK
Attanasio, Orazio P., and Carl Emmerson, April 2001, National Bureau of Economic Research
(NBER) Working Paper No. 8241. †
The existence of a correlation between health outcomes and socio–economic status is
well established, but it is not clear what are the causal links between health and wealth. It is often
implied that a causal relationship runs from socio–economic status to health outcomes, but the
direction of causation can easily be justified in either direction.
The authors aimed to quantify the relationship between mortality, health status and
wealth and they attempted to control for reverse causation by using the two waves of the British Retirement
Survey (1988/89 and 1994) – a longitudinal database, and by conditioning on initial health status. Both
mortality and morbidity were modeled.
It was found that, even after conditioning on the initial health status, wealth rankings
are important determinants of mortality and morbidity. (The relationship between survival probabilities and
wealth rankings is not linear.) For both genders, there is a positive relationship between wealth and the
probability of survival, regardless of the wealth measure used or the specification of the model. The results
for females are only marginally lower relative to those for males. These results are robust to the wealth
definition used. This would indicate a causal link running from wealth to health, but it does not remove the
possibility of a causal link also in the other direction.
Marital status was not found to be significant for mortality or morbidity, whereas
education was found to be significant for morbidity but not for mortality. For females, marital status has
some effect on mortality but education is insignificant. Neither education nor marital status had a strong
effect on female morbidity.
The study population consisted of individuals aged between 55 and 69 in late 1988 / early
1989. (Their spouses were also interviewed.) The surveys asked many health status questions relating to each
of 13 different areas of disability, in order to construct a health index. They also contained several
measures of wealth (financial, housing and pension wealth) and details of current income, employment status
and job history. Details of the highest education attainment were also obtained. This permitted
experimentation to establish what are the most appropriate indicators of the relationship between
socio–economic status and health. In analyses using wealth rankings, couples and single people were
ranked separately. Other analyses were on an individual basis and males and females were considered
separately.
The authors considered that the variables that reflect the amount of life cycle resources
available to an individual were the most likely to be important in modeling health outcomes. They considered
the value of pension wealth to be particularly useful as, for many, it is directly linked to lifetime
earnings. They did not think current income was a good indicator. The two definitions of wealth used in
analyses were financial and housing wealth and total wealth, i.e. financial, housing and pension wealth.
In addition to age, wealth and health status, controls for marital status and education
were used. Dummy variables for different regions were also tested, but once other variables were controlled
for, no significant effect was found. This suggests that the variation in standardized mortality rates seen
across the regions of Britain is completely explained by the observable characteristics of the resident
population.
A limitation of this study was that the first interview was originally not supposed to be
followed by the second one and therefore attrition (for reasons other than mortality) is particularly
severe, and unlikely to be random. This was addressed by modeling explicitly the process of attrition,
allowing for the possibility that this is correlated with mortality and the evolution of health status.
A similar study, using a different approach and using US data, obtained different
results. Hurd and McFadden found that once initial health status was controlled for, no relationship between
economic variables and health outcomes was found. This would indicate a relationship running from health to
wealth, rather than from wealth to health and highlights the importance of reverse causality. This is despite
the fact that the UK has universal health care and the US does not.
The authors comment that further research is required to determine whether the relevant
concept is relative or absolute wealth. The current study could be extended by constructing an explicit model
of the evolution of wealth.
† This study is based on data from England and Wales.