Letting the Gini out of the bottle? Challenges facing the relative income hypothesis

The relative income hypothesis interprets statistical associations between income inequality and average health status at the population level, as evidence that income inequality has a deleterious psychosocial effect on individual health. An alternative explanation is that these, population-level associations, are statistical artefacts of curvilinear, individual-level relationships between income and health. Indeed, provided the cost-benefit ratio of health-enhancing goods and services vary, the law of diminishing returns should produce curvilinear, asymptotic relationships between income and health at the individual level, which create ('artefactual') associations between income inequality and health at the population level. However, proponents of the relative income hypothesis have argued that these relationships are unlikely to be responsible for the associations observed between income inequality and average health status amongst high-income populations. In these populations, the individual-level relationships between income and health would be nearer their asymptotes, where a shallower slope should ensure that income inequality has little (if any) 'artefactual' effect on average health status. Yet this argument was based on analyses of population-level data which underestimated the slope and curvilinearity of underlying, individual-level relationships between income and health. It is therefore likely that (at least some part of) the population-level associations between income inequality and average health status (amongst low-, middle- and high-income populations) are 'artefacts' of curvilinear, individual-level relationships between income and health. Nevertheless, it is also possible that income inequality is somehow (partly or wholly) responsible for the curvilinear nature of individual-level relationships between income and health. Likewise, it is possible that income inequality alters the height, slope and/or curvilinearity of these relationships in such a way that income inequality has an independent effect on individual health. In either instance, the 'artefactual' effect of curvilinear relationships between income and health at the individual level would simply reflect the mechanism underlying the relative income hypothesis. © 2002 Elsevier Science Ltd. All rights reserved.