Kawachi I, Kennedy BP The relationship of income inequality to mortality - does the choice of indicator matter? Social Science & Medicine 1997; 45: 1121-7

The studies linking income inequality and health use different measures of income distribution. How sensitive is the relationship to the measure of income distribution. Income data is usually obtained from surveys, and as McIsaac and Wilkinson report, subject to under-reporting, especially among the rich, and non-reporting among the poor. This study looks at data from the 1990 US Census, and computes many different income distribution measures which were not adjusted for federal taxes since the same rates apply to all states, and would not affect the results. No adjustment was made for household size, since there was no relationship within states of household size on median income, poverty rates, mortality rates or measures of income inequality. A sensitivity analysis was done to see if variations in income distribution through taxes, transfers, and household size could affect the results. The authors found that all income inequality measures were highly correlated with each other. They looked at the correlations of inequality measures with age adjusted mortality by state, and found all were strong. Surprisingly, the two published measures used to establish the relationship between state mortality rates and income distribution turned out to be the ones less strongly correlated than the others, so the finding appears to be robust. Adjusting for median income and poverty rates did not affect the relation between income inequality and mortality. As well, it is not sensitive to taxes, transfers and household size. The authors make the important point that while international data on income from the Luxembourg Income Study suffers from poor response rates especially among the rich and poor, these data have high response rates. The authors point out different sensitivities of the income measures to detecting the effects of economic policies that would produce changes in income distribution. They point out that the Atkinson Index may be the best measure since it can incorporate explicit judgments about the social preference for income inequality. 

Abstract

Ecologic studies in the U.S. and elsewhere in the world have demonstrated that income inequality is strongly related to mortality and life expectancy: the greater the dispersion of income within a given society, the lower the life expectancy. However, these empirical studies have been criticized on the grounds that the choice of indicator may have influenced positive findings. Using a cross-sectional, ecologic design, we tested the relationships of six different income inequality indicators to total mortality rates in the 50 U.S. states. The following summary measures of income distribution were examined: the Gini coefficient; the decile ratio; the proportions of total income earned by the bottom 50%, 60%, and 70% of households; the Robin Hood Index; the Atkinson Index; and Theil's entropy measure. All were highly correlated with each other (Pearson r> or = 0.94), and all were strongly associated with mortality (Pearson r ranging from 0.50 to 0.66), even after adjustment for median income and poverty. Thus, the choice of income distribution measure does not appear to alter the conclusion that income inequality is linked to higher mortality. Furthermore, adjustment for taxes and transfers, as well as household size (using equivalence scales), made no difference to the income inequality/mortality association. From a policy perspective, the alternative income distribution measures perform differently under varying types of income transfers, so that theoretical considerations should guide the selection of an indicator to assess the impact of social and economic policies that address income inequality.

Keywords

• Atkinson Index
• Gini coefficient
• income
• income distribution
• income inequality
• mortality
• population health
• poverty
• regional
• redistribution
• socioeconomic status
• underreporting
• wealth inequality