Editor—Ross et al report that income inequality in the state or province and metropolitan area is associated with mortality in the United States but not in Canada.1 They offer two explanations for this discrepancy. Firstly, this association is observed only at levels of inequality present in the United States, not the lower levels observed in Canada. Secondly, the actual association of income inequality with mortality is modified by the social and political characteristics specific to a place.
There is, however, another possible explanation: confounding of the association of income inequality with health at the state level (or metropolitan area level) in the United States. This is not the same as confounding at the individual level by, for example, personal income, which varies between states.2 State level confounding may occur when characteristics of the states are correlated with income inequality, remain associated with mortality within strata of states by income inequality, and hence cause a spurious association of income inequality with mortality. Two possible candidates are the extent of rurality and the welfare policies of the states. Both vary between states, are plausibly related to population health, and are probably correlated with the geographically biased distribution of income inequality.
The United States provides a rich natural experiment to study the association of income inequality with health, and several studies have now replicated the association of state level income inequality with health using different data sets.3,4 But the use of different data sets in these studies is akin to reanalysing one cross sectional study of the same 50 people by using different measures of exposure and outcome; it is not akin to separate studies of a different 50 people each time. Thus, a spurious association of income inequality with health due to state level confounding will remain for each new analysis of the same natural experiment.
It seems likely and plausible that income inequality is associated with health.5 Instead of examining the possibility of state level confounding, however, we propose two research strategies. Firstly, potential state level confounders are included as covariates in analyses. This will not be without difficulty, however, owing to the high probability of (multi)collinearity of ecological variables and challenging theoretical considerations—for example, causal ordering of ecological variables, such as income inequality and welfare policies. Secondly, many different natural experiments should be analysed to look for a consistent association of income inequality with health. From this perspective, the results from Canada and the United States are just two separate natural experiments to which we want to add results from many more natural experiments.
References
- 1.Ross N, Wolfson M, Dun J, Berthelot J-M, Kaplan G, Lynch J. Relation between income inequality and mortality in Canada and in the United States: cross sectional assessment using census data and vital statistics. BMJ. 2000;320:898–902. doi: 10.1136/bmj.320.7239.898. . (1 April.) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Blakely T, Woodward A. Ecological effects in multi-level studies. J Epidemiol Community Health. 2000;54:367–374. doi: 10.1136/jech.54.5.367. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Kaplan G, Pamuk E, Lynch J, Cohen R, Balfour J. Inequality in income and mortality in the United States: analysis of mortality and potential pathways. BMJ. 1996;312:999–1003. doi: 10.1136/bmj.312.7037.999. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Kennedy B, Kawachi I, Glass R, Prothrow-Stith D. Income distribution, socioeconomic status, and self-rated health in the United States: multilevel analysis. BMJ. 1998;317:917–921. doi: 10.1136/bmj.317.7163.917. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Kawachi I, Kennedy B, Wilkinson R, editors. Society and population health reader: income inequality and health. New York: The New Press; 1999. [Google Scholar]