Income inequality and health: Importance of a cross-country perspective

Abstract

This paper uses a unique dataset-containing information collected in 2006 on individuals aged 40–79 in 21 countries throughout the world to examine whether individual income, relative income in a reference group, and income inequality are related to health status across middle/low and high-income countries. The dependent variable is self-assessed health (SAH), and as a robustness check, activities of daily living (ADL) are considered. The focus is particularly on assumptions regarding an individual's reference group and on how the estimated relationships depend on the level of economic development. Correcting for national differences in health reporting behavior, individual absolute income is found to be positively related to individual health. Furthermore, in the high-income sample, there is strong evidence that average income within a peer-age group is negatively related to health, thus supporting the relative income hypothesis. In middle/low-income countries, it is instead average regional income that is negatively associated with health. Finally, there is evidence of a negative relationship between income inequality and individual health in high-income countries. Overall, the results suggest that there might be important differences in these relationships between high-income and middle/low-income countries.

Introduction

In economics, as in epidemiology, there is an ongoing discussion as to whether income inequality in itself contributes to ill-health. Initially, the empirical support for this claim was the existence of a striking negative correlation between the average health status in a population and various measures of income inequality. Fig. 1 plots the propensity to report ‘very good’ health in our dataset against the national Gini coefficient of incomes for the same year, and a crude image of a negative relationship emerges.

However, it has been argued that the observed relationship could be a statistical artifact reflecting a non-linear relationship between income and health at the individual level (Gravelle, 1998). When there are diminishing returns to income in the production of health, the observed relationship between societal income inequality and population health will, to some extent, be spurious and not causal.

At the same time, there are reasons to believe that income inequality could have an impact on health. For example, sharp differences in income and living standards may lead to stress or reduce people's overall well-being in other ways (Wilkinson, 1996). It has also been suggested that social rank and social networks are important health determinants (Cutler and Lleras-Muney, 2006, Marmot, 2005). If the degree of control relative to the demands under which individuals are operating, or the degree of social interaction between people, is related to the level of income inequality in societies, the latter might be important for health outcomes. If income inequality affects health directly or indirectly, social and economic policies that influence income distributions may have important consequences, which so far have been largely ignored by economists and policy makers (Deaton, 2003, Wilkinson, 1992). A possible exception is health effects related to relative rank, which may be unaffected by redistribution to the extent that it keeps relative ranks constant.

Wagstaff and van Doorslaer (2000) identify no less than five competing explanations for the observed correlation between income inequality and population health, reflecting different relationships at the individual level. It is important to distinguish between these hypotheses as they have different policy implications. A minimum requirement is then to use data on health status and incomes of individuals.

Information on individuals is often available for separate countries and several empirical studies of the relationships between income inequality and health in different (mostly developed) economies have been carried out during the past decade. The evidence from this research is largely contradictory, which might be due to the differences between the countries studied, but also to differences in the choice of methods, dependent variables (subjective or objective health) and interpretation of covariates as confounders or mediators.

The objective of this paper is to examine the association between income inequality and health across middle/low and high-income contexts with a new dataset, The Future of Retirement, which contains information on the health status of individuals in 21 countries across the world. More specifically, we intend to test three of the five hypotheses suggested in the literature: the absolute income hypothesis, stating that individual income, but not income distribution, matters to individual health; the relative income hypothesis, according to which an individual's health is affected by the average income in a reference group; and the income inequality hypothesis, according to which income inequality in a society influences everyone's health. Obviously, the hypotheses are not necessarily mutually exclusive. The two hypotheses not considered here are the deprivation hypothesis and the relative position hypothesis.¹

The contribution of this paper is threefold. First, we introduce a useful dataset, not previously available, which contains consistently collected information at individual level, on a wide range of important topics. Furthermore, information from both developing as well as developed economies is available. This is of considerable importance, not least as it entails substantial variation in the exposure to income inequality, thus increasing the chances of detecting any effects on health (Gerdtham & Johannesson, 2004). Secondly, the countries included allow a comparative analysis of whether relationships are different at different levels of economic development. To our knowledge, this has not been done before with individual-level data. Finally, we assume that people compare themselves with individuals of their own age rather than with people in a particular part of the country. This seems reasonable from an economic point of view and turns out to have a strong explanatory power in a high-income context.

The paper is organized as follows. The next section gives a brief overview of the literature on the three hypotheses and the empirical evidence to date. The third section outlines the methodological considerations underlying our econometric approach and after that we provide an overview of the dataset and imported variables. The following section presents the results of the various specifications we have considered. The final section concludes the paper and identifies open issues for future research.