- Richard G Wilkinson, senior research fellowa
Fiscella and Franks claim that the relation between societal mortality and income inequality is an ecological fallacy, but in fact their data makes a useful contribution wholly consistent with that relation. Their data shows that individual income levels largely account for the income inequality relation in American counties. But the relation between individual income and mortality is primarily an effect of relative income.1 2 As inequality defines relative income, what the paper suggests is that the relation between population mortality and inequality reflects the health implications of a person's relative income rather than any effects of wider societal processes.
The nub of the problem is the assumption that mortality is affected by absolute levels of income–as if $20 000 bought a given amount of health regardless of the incomes of the rest of society. But in practice how much health it buys depends on whether it makes you rich or poor compared with the rest of society. The reason that the affluence of the United States does not lead to mortality rates as low as in many poorer countries is not because the United States has more people in absolute poverty. Indeed, at the state level the association between mortality and inequality remains after absolute poverty is controlled for.3 The problem is that the United States has more relative poverty, and this is likely to affect health, mainly–as this paper shows–through individual relative income.
The major health inequalities that accompany socioeconomic differences across the social hierarchy are now widely recognised. What the income distribution relation tells us is that health is worse when there is greater inequality across this social gradient. Rather than hinging on some other dimension of inequality operating between neighbours or within the local community, the effect of income inequality is almost certainly tied up with the central sociological processes of social stratification.
That absolute income levels are no longer important in the developed world is shown by the lack of a strong association between mortality and average incomes among developed nations or American states.2 3 4 5 The weak correlation with median state income disappears completely after income distribution within states is controlled for.1 In contrast, within countries or states income is closely related to mortality because it measures differences in socioeconomic status.2
In smaller areas of analysis, however, the situation is reversed: between small areas median income is related to mortality, whereas income distribution within them is not.6 7 8 That is because the salient inequality lies in the societal system of social stratification.5 In the smallest areas the social heterogeneity, which makes inequality important, is lost. At the extreme, in small, socially homogeneous, one class neighbourhoods, the inequalities that matter would all be between, not within, neighbourhoods. Differences in their median community income would do all the work in relation to mortality, while inequality within them would do none. Individual income matters because it is an indicator of position in relation to the societal system of social stratification. After all, Harlem's appalling health does not result from the inequalities within Harlem but from its deprivation in relation to the United States outside Harlem.9
At the intermediate level of American counties used by Fiscella and Franks, some of the social heterogeneity that accounts for the explanatory power of inequality lies within them and some between them; so income inequality explains less variation in mortality here than it does at the state or national level, but more than it would in smaller areas. Thus what their analysis really shows is that the effect of the inequality attributable to the residual social heterogeneity within counties can be explained by the effects of peoples' incomes relative to the wider society.
This kind of analysis cannot finally tell us whether there are also wider knock-on effects to the rest of society from increased relative deprivation–perhaps arising from the higher levels of homicide, accidents, and alcohol related deaths associated with less egalitarian societies.9 10 11 That would be possible only if the relation between individual income and health could be measured independently of its determination in a particular society with its particular level of income inequality.
What is important, however, is that national mortality rates can be lowered by redistributing income. Average health is improved not by simply redistributing a given amount of health (by redistributing the current stock of health producing goods), but by reducing the psychosocial burden of relative deprivation. Fiscella and Franks's interpretation of their results implies a steeply curved relation between absolute income and mortality. If such a relation existed then economic growth without redistribution would rapidly narrow health inequalities. As animal models have shown, subordinate social status has health consequences even when the physical environment and diet are invariant.12 13
I am grateful to members of the International Centre for Health and Society for opportunities to discuss these issues.