تاثیر نابرابری درآمدی بر سلامت عمومی دارای امتیاز : مدارک و شواهد از یک مطالعه ملی

Previous studies report a strong negative association between income inequality and population health at the aggregate level. However, it is still in hot debate whether this ecological association indicates a genuine, causal effect of income inequality on health, as asserted by the Wilkinson hypothesis, or it simply reflects a nonlinear effect of individual income on health, as suggested by the absolute income hypothesis. Drawing data from the 2005 round of the World Values Survey, I analyze the relationship between individual income, income inequality, and self-rated general health in a multilevel framework. Results show no independent detrimental effect of country income inequality on individual self-rated general health. In contrast, self-rated general health is strongly associated with absolute material conditions both at the individual and at the country level. Therefore, this study gives more evidence to the absolute income hypothesis, i.e., the strong ecological association between income inequality and population health is more likely a reflection of the nonlinear effect of individual income on health rather than a genuine effect of income inequality.

Health inequality, both between and within societies, has become a major concern among studies of social inequality in recent decades. It is well documented that within almost all societies, there is a socioeconomic gradient with respect to health, and higher socioeconomic status (SES) is always associated with better health status (Feinstein, 1993 and Robert and House, 2000). In fact, not only do people at the top of the SES gradient enjoy better health than those at the bottom, but also health deteriorates continuously as the ladder of the SES hierarchy goes lower for all levels of SES (Marmot, 2003 and Robert and House, 2000). Despite the persistent SES-health gradient within societies, international inequality of health is only weakly associated with the level of economic development, especially among developed countries (Preston, 1975, Wilkinson, 1992 and Wilkinson, 1996). In contrast to the weak association between population health and national wealth, a substantial number of ecological studies show that country life expectancy is negatively associated with societal income inequality, even after controlling for the level of per capita income (De Vogli et al., 2005, Flegg, 1982, Ram, 2006, Rodgers, 1979, Wilkinson, 1992 and Wilkinson, 1996). Specifically, life expectancy in more unequal societies is shorter than in more egalitarian societies. In addition to life expectancy, similar associations are also found between income inequality and other indicators of population health, such as infant mortality, age- and cause-specific mortality, mean age at death, self-rated health, population height, and the homicide rate (Flegg, 1982, Hsieh and Pugh, 1993, Qi, 2011, Rodgers, 1979, Waldman, 1992 and Wilkinson, 1996). Moreover, the association between income inequality and health has also been observed in certain within-country studies, such as between states and metropolitan areas in the United States (Kaplan et al., 1996, Kawachi and Kennedy, 1997, Kennedy et al., 1996 and Wolfson et al., 1999). The observation of an ecological negative correlation between income inequality and population health has attracted a lot of attention from various disciplines of social sciences over the last two decades, and it still remains one of the most controversial topics in social studies of health (Lynch et al., 2004, Mackenbach, 2002, Wilkinson, 2002 and Wilkinson and Pickett, 2006). Two competing theories dominate the debate on explaining the ecological association between income inequality and population health: The Wilkinson hypothesis claiming that income inequality has a genuine detrimental impact on health, and the absolute income hypothesis that emphasizes a nonlinear effect of individual income on health and argues in favor of an instance of ecological fallacy. Nevertheless, these two hypotheses cannot be adequately evaluated by ecological studies, and the debate calls for an examination of the independent effect of income inequality on health by controlling for individual SES markers simultaneously in a multilevel framework. Drawing data from a large cross-national study project, the 2005 round of the World Values Survey (WVS), this study aims to examine the “income inequality-health” association and test these alternative hypotheses. Compared with previous studies of this kind, the current study fills the gap of the existing literature in several key aspects: First, although there are a large number of within-country studies that have taken the multilevel approach, cross-national studies that are cognizant of this point are still limited. Second, in the existing international multilevel studies, with a few exceptions (Jen et al., 2009 and Mansyur et al., 2008), the number of countries included is usually very small and only restricted to a subset of the European countries, hence calling for further investigations with a much larger and diverse sample of countries. Third, for the two available studies with a large number of countries (N > 30), the findings are puzzling. Both Mansyur et al. (2008) and Jen, Jones, and Johnston (2009) report a significantly positive effect of income inequality on individual self-rated health, which is not in line with any known explanations. And finally, international comparability of both income inequality measures and self-rated health is crucial for a valid empirical assessment of the relationship between income inequality and health, but has been largely overlooked in previous studies. This study uses an adjusted income inequality measure that is more comparable, and the two appendices ( Appendix B and Appendix C) discuss these problems in further detail.

In this study, I use the 2005 round of the WVS data to test the Wilkinson hypothesis by modeling the association between individual income, income inequality, and self-rated general health in a multilevel framework. My analysis provides little support for an independent detrimental effect of income inequality on self-rated general health when individual SES markers are controlled. Moreover, the results suggest that absolute wealth, both at the individual and at the country level, are more significant predictors of individual health than is the relative distribution of income. Therefore, it is likely that the strong ecological association between income inequality and population health is mainly a reflection of the nonlinear effect of individual income on health, as indicated by the absolute income hypothesis (Gravelle, 1998 and Lynch et al., 2004). These findings are robust to possible outlier countries and the choice of model forms. My main findings are generally consistent with Mansyur et al. (2008) and Jen et al. (2009), except that the puzzling positive effect of income inequality on individual health found in their study is only observable for non-OECD countries in this study. Yet, the lower comparability of the self-rated health in those non-OECD countries may play a role in the observed positive effect in my study. Compared to those two studies, I use the data collected over a short time period, rather than relying on pooled data over a long time period. Therefore, the temporal comparability is less an issue for my findings. Moreover, I use an adjusted income inequality measure that is theoretically more comparable than the measures used in those two studies. Like many other cross-national studies, this study has several limitations and the results should be viewed with caution. First, although self-rated general health does contain valuable information on true health status and some of this information is even beyond the measurement of objective indicators (Murray & Chen, 1992), the reporting behavior is also likely to be affected by individual characteristics such as age and gender, as well as contextual factors such as the average health status and culture. My ecological analysis of the relationship between self-rated general health and life expectancy for the studied countries (Appendix C) throws some light on the international comparability of self-rated general health. Nonetheless, this is still a field in progress and more extensive examinations of the cross-population comparability of health indicators are necessary. Second, the absolute income measures used in this study are household income deciles instead of the real values. Due to the way that the WVS collected information, it is impossible to construct the absolute values of income. Since the deciles may vary significantly from one country to another, the unit effect of income deciles on health is less interpretable. For instance, the regression results regarding the cross-level interaction between household income deciles and Gini index is subject to equally plausible explanations. Nonetheless, the income deciles should be able to capture key differences in the levels of absolute income and the main findings reported here are unlikely to be affected by the scale of income measure. Third, although the number of countries analyzed in this study is much larger than many previous studies, it is still comparatively small in the multilevel framework, especially when the samples are divided into OECD and non-OECD countries. As a result, I only estimated a random-intercept model that assumes all the effects of individual variables on health are fixed across country, which may be too strong to be realistic. When possible, it should be checked how these assumptions affect the model results in future studies. Finally, the survey quality of the WVS may vary significantly across participating societies, such as how the sample is drawn, how the fieldwork is conducted, and how high the response rate is. Due to the lack of detailed documentations, the potential impact of these variations is not controlled in the current study. Since the WVS is one of the earliest and largest international survey projects, all the three available large cross-national studies (including mine) have exclusively used the WVS data. Therefore, before we confidently disprove the Wilkinson hypothesis, it is important to validate the current findings by examining further data from other large international surveys such as the International Social Survey Project (ISSP) and the World Health Survey.