Socioeconomic determinants of health: Health inequalities: relative or absolute material standards?
BMJ 1997; 314 doi: https://doi.org/10.1136/bmj.314.7080.591 (Published 22 February 1997) Cite this as: BMJ 1997;314:591- Richard G Wilkinson, senior research fellow (R.G.Wilkinson{at}sussex.ac.uk)a
Introduction
That mortality in developed countries is affected more by relative than absolute living standards is shown by three pieces of evidence. Firstly, mortality is related more closely to relative income within countries than to differences in absolute income between them. Secondly, national mortality rates tend to be lowest in countries that have smaller income differences and thus have lower levels of relative deprivation. Thirdly, most of the long term rise in life expectancy seems unrelated to long term economic growth rates. Although both material and social influences contribute to inequalities in health, the importance of relative standards implies that psychosocial pathways may be particularly influential. During the 1980s income differences widened more rapidly in Britain than in other countries; almost a quarter of the population now lives in relative poverty. The effects of higher levels of relative deprivation and lower social cohesion may already be visible in mortality trends among young adults.
The existence of wide-and widening-socioeconomic differences in health shows how extraordinarily sensitive health remains to socioeconomic circumstances. Twofold, threefold, or even fourfold differences in mortality have been reported within Britain, depending largely on the social classification used.1 2 3 This series will illustrate some of the most important mechanisms involved in the generation of these differences.
Fundamental to understanding the causes of these differences in health is the distinction between the effects of relative and absolute living standards. Socioeconomic gradients in health are simultaneously an association with social position and with different material circumstances, both of which have implications for health-but which is more important in terms of causality? Is the health disadvantage of the least well off part of the population mainly a reflection of the direct physiological effects of lower absolute material standards (of bad housing, poor diets, inadequate heating, and air pollution), or is it more a matter of the direct and indirect effects of differences in psychosocial circumstances associated with social position-of where you stand in relation to others? The indirect effects of psychosocial circumstances here include increased exposure to behavioural risks resulting from psychosocial stress, including any stress related smoking, drinking, eating “for comfort,” etc; most of the direct effects are likely to centre on the physiological effects of chronic mental and emotional stress.
Evidence from three sources suggests that the psychosocial effects of social position account for the larger part of health inequalities. If valid, this perspective would have fundamental implications for public policy and for our understanding of the pathways through which socioeconomic differences have an impact on human biology.
Income within and between societies
Despite the difficulty of disentangling material from social influences on health, it is possible to look at the relation between income and health in population groups where income differences are, and are not, associated with social status. Social stratification exists within rather than between societies. Therefore, while income differences among groups within the developed societies are associated with social status, the differences in average per capita incomes between developed societies are not. We may therefore compare the association of income and health within and between societies.
Within countries there is a close relation between most measures of health and socioeconomic circumstances. As an example, figure 2) uses data from 300 685 white American men in the multiple risk factor intervention trial to show the relation between mortality and the median family income in the postcode areas in which they lived.4 Among black men in the trial, larger mortality differences are spread over a smaller income range.5 In Britain, there are similar gradients in mortality and sickness absence among men and women.6 7
The regular gradients between income and mortality within countries contrast sharply with the much weaker relation found in the differences between rich developed societies. Figure 3) shows the cross sectional relation between life expectancy and gross domestic product per capita for 23 members of the Organisation of Economic Cooperation and Development (OECD) in 1993. Using data from the OECD countries reduces the influence of extraneous cultural differences by restricting the comparison to developed, democratic countries with market economies. Currencies have been converted at “purchasing power parities” to reflect real differences in spending in each country. The correlation coefficient of 0.08 shows that life expectancy and gross national product per capita are not related in this cross sectional data. Excluding government expenditure makes little difference: the correlation with private consumer's expenditure per capita is only 0.10.
Data on changes over time between countries show a weak but non-significant relation. During 1970-93 the correlation between increases in life expectancy and percentage increases in gross domestic product per capita among OECD countries was 0.30, suggesting that less than 10% of the increases in life expectancy were related to economic performance. Though the recent rise in national mortality in eastern Europe suggests that time lags may be short, the period used here allows for the possibility of longer lags.8
As figure 3) uses data for whole countries, the contrast between it and the strong relation shown in figure 2) cannot arise from sampling error. A strong international relation is unlikely to be masked by cultural factors: not only are the international comparisons confined to OECD countries, but the picture is supported by comparisons among the 50 states of the United States, where cultural differences are smaller. The correlation reported between age adjusted mortality and median incomes in the states was -0.28.9 As with the international comparisons, social stratification mainly occurs within rather than between American states.
Income and mortality are so strongly related within societies that this relation cannot be assumed to exist between developed societies but has somehow become hidden. Its robustness within societies shows not merely in mortality data but in measures as diverse as medically certified sickness absence among civil servants and prescription items issued per head of population in relation to local rates of unemployment.7 10 However, the contrast in the strength of the relation within and between societies would make sense if mortality in rich countries were influenced more by relative income than by absolute material standards.
Income distribution
A second source of evidence that relative income has a powerful influence on health comes from analyses of the relation between measures of income inequality and mortality both among developed countries11 and among the 50 states of the United States.12 Cross sectional data and data covering changes over time both show that mortality tends to be lower in societies where income differences are smaller, even after average incomes, absolute poverty, and a number of other socioeconomic factors have been controlled for. This relation has now been shown independently on over a dozen different datasets and has been reported absent only once.11 The most plausible explanation is that mortality is lower in more egalitarian societies because the burden of relative deprivation is reduced.
The weak association between mortality and median (absolute) incomes of the 50 American states disappears when the distribution of income within each state is controlled for.9 The correlation coefficient drops from -0.28 to -0.06, suggesting that absolute income is unrelated to mortality in the United States. Unfortunately, further exploration of the international relation between income distribution and mortality will depend on taking account of the differences in response to income surveys in different countries. Response rates vary by more than 30%, and as non-responders are concentrated particularly among the rich and poor, high non-response leads to smaller reported income differences.13 14
Epidemiological transition
The third reason for thinking that health is influenced more by relative than absolute income centres on the epidemiological transition. Although absolute material standards remain important in less developed countries, there are indications that the epidemiological transition represents a stage in economic development after which further improvements in material standards have less influence on health. Not only do the infectious diseases of poor countries give way to degenerative diseases as the main causes of death, but the transition also coincides with a flattening of the curve relating life expectancy to gross domestic product per capita.11 15 In addition, several of the so called “diseases of affluence” (including coronary heart disease, stroke, hypertension, obesity, and duodenal ulcers) reverse their social distribution to become more common among poor people in affluent societies, reflecting that the majority of the population has risen above a minimum threshold level of living.11 16 When those who are less well off cease to be thin, obesity ceases to be associated with social status.
A theory of health and social position?
If the association between health and socioeconomic status within societies-at least in the developed world-is not primarily the direct effect of material standards, then some might think it resulted simply from differential social mobility between healthy people and unhealthy people. However, many research reports show that this is not the major part of the picture,17 18 19 20 and social selection is entirely unable to account for the relation between national mortality rates and income distribution.
This pushes us-inexorably though perhaps reluctantly-towards the view that socioeconomic differences in health within countries result primarily from differences in people's position in the socioeconomic hierarchy relative to others, leaving a less powerful role to the undoubted direct effects of absolute material standards. If health inequality had been a residual problem of absolute poverty it might have been expected to have diminished under the impact of postwar economic growth, and it would tend to distinguish primarily between the poor and the rest of the population-rather than running across society, making even the higher echelons less healthy than those above them (see figure 2).
Need for a theory
A theory is needed which unifies the causes of the health inequalities related to social hierarchy with the effects of income inequality on national mortality rates. At its centre are likely to be factors affecting how hierarchical the hierarchy is, the depths of material insecurity and social exclusion which societies tolerate, and the direct and indirect psychosocial effects of social stratification.21
One reason why greater income equality is associated with better health seems to be that it tends to improve social cohesion and reduce the social divisions.11 Qualitative and quantitative evidence suggests that more egalitarian societies are more cohesive. In their study of Italian regions, Putnam et al report a strong correlation (0.81) between income equality and their index of the strength of local community life.22 They say, “Equality is an essential feature of the civic community.” Kawachi et al have shown that measures of “social trust” provide a statistical link between income distribution and mortality in the United States.23 Better integration into a network of social relations is known to benefit health.24 25 This accords with the emphasis placed on relative poverty as a form of social exclusion, and with the evidence that racial discrimination has direct health effects.26 However, social wellbeing is not simply a matter of stronger social networks. Low control, insecurity, and loss of self esteem are among the psychosocial risk factors known to mediate between health and socioeconomic circumstances. Indeed, integration in the economic life of society, reduced unemployment, material security, and narrower income differences provide the material base for a more cohesive society. Usually the effects of chronic stress will be closely related to the many direct effects of material deprivation, simply because material insecurity is always worrying. However, as Hogarth's Gin Lane shows, even absolute poverty has often killed through psychosocial and behavioural pathways.
Pathways
In terms of the pathways involved in the transition from social to biological processes, there is increasing interest in the physiological effects of chronic stress. Social status differences in physiological risk factors among several species of non-human primates have been identified. Animals lower in the social hierarchy hypersecreted cortisol, had higher blood pressure, had suppressed immune function, more commonly had central obesity, and had less good ratios of high density lipoproteins to low density lipoproteins-even when they were fed the same diet and social status was manipulated experimentally.27 28 Among humans, lower social status has also been associated with lower ratios of high to low density lipoproteins, central obesity, and higher fibrinogen concentrations.29 In experiments in which social status was manipulated, subordinate monkeys “received more aggression, engaged in less affiliation, and spent more time alone than dominants … they spent more time fearfully scanning the social environment and displayed more behavioral depression than dominants.”30 Loss of social status resulting from being rehoused with more dominant animals was associated with fivefold increases in coronary artery atherosclerosis.31
Although research has shown that psychosocial factors are related to both morbidity and mortality, differences in reporting make international comparisons of morbidity unreliable. Nevertheless, because patterns even of self reported morbidity are predictive of mortality rates, we can probably assume that mortality differences indicate differences in objectively defined morbidity.32 33 Although no obvious patterns have emerged from attempts to assess international differences in the extent of inequalities in self reported morbidity when people are classified by education or social class, across countries there is a close relation between the extent of inequalities in income and in self reported morbidity.34 35
Relative poverty and mortality
Although Britain had a greater increase in inequality during the 1980s than other developed market economies,36 the proportion of the population living in relative poverty (below half the average income) may-for the first time in two decades-have decreased slightly during the early 1990s. It now stands at almost one in four of the whole population (incomes after deducting housing costs).37 Among children the proportion is almost one in three. Particularly worrying is the likely increase in the proportion of children emotionally scarred by the tensions and conflicts of family life aggravated by living in relative poverty. During 1982-92 there were no improvements in national mortality rates among young men (aged 20-40) and smaller improvements among younger women (aged 15-24) than at most other ages.38 Socioeconomic differences in mortality are at their maximum at these ages, and the national trends are likely to be partly a reflection of the increased burden of relative deprivation. Among young men, deaths from suicide, AIDS, violence, and cirrhosis increased. These causes suggest that the psychosocial effects of relative deprivation are unlikely to be confined to health. As in the international data, where death rates from accidents, violence, and alcohol related causes seem to be particularly closely related to wider income inequalities, the predominance of behavioural causes may reflect changes in social cohesion.9 13
The papers in this series are intended to illustrate some of the processes which give rise to the relation between relative deprivation and health. What comes out of several of them may not have been so different had the subject been crime, drug misuse, or poor educational performance. Important aspects of the evidence suggest that the rest of society cannot long remain insulated from the effects of high levels of relative deprivation.
Acknowledgments
Funding: Paul Hamlyn Foundation.
Conflict of interest: None.