IJE Advance Access originally published online on January 19, 2005
International Journal of Epidemiology 2005 34(2):293-294; doi:10.1093/ije/dyh412
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Commentary |
Commentary: On form, comparability, and levels in the income and health relationship
Centre for Health Equity Studies, CHESS Stockholm University/Karolinska Institutet, 106 91 Stockholm, Sweden. E-mail: johan.fritzell{at}chess.su.se
One important issue in the income and health debate, relates to the form of the association between them.1–3 The rationale for the interest in this topic is that it involves mechanisms, policy implications, and theory, rather than simply being a methodological, technical question. The article by Mackenbach et al. in this issue of International Journal of Epidemiology adds to our knowledge of this area by investigating the relation between self-rated health and income through the use of cross-national data.4 They generally find a curvilinear relation reflecting decreasing marginal health returns of income. In this short note I want to first discuss some methodological concerns, also discussed by the authors. I will thereafter reflect on some more general, substantive implication of this curvilinear association and its relation to the recent heated debate on the assumed relation between income inequality and health.5,6
Comparative, cross-national research is in many respects a fruitful enterprise with high potentials. It has sometimes been characterized as a quasi-natural experiment. In principle, cross-national comparisons make it possible to study what role a social factor, like a certain policy, has in producing an outcome, since the counterfactual cases, countries without such a policy, can be contrasted. However, outcomes that are believed to be due to country-specific factors are often seen in many different national settings. The cross-national approach is therefore also a fruitful way to establish empirical regularities across countries.7 The study by Mackenbach et al. is a good point of reference. For the general discussion on income, income inequality and health, it is essential to state if the curvilinear relation is generally observed.
At the same time one should be aware that the pitfalls of cross-national research are formidable. One basic problem concerns data comparability. This is much more difficult in a cross-national study due to factors like language, country-specific values, cross-national variation in data collection methods, data quality, and sample frame. In brief, cross-national research has a lot of potentials and a lot of pitfalls.
The study by Mackenbach et al. also highlights the problems. The study, definitely, uses sophisticated methodology but that can never compensate for poor data comparability. What is most worrisome in this study is precisely this point. Income information is not only collected by different methods (a common problem), but the distributions to be compared across countries are based on different income concepts. The history of comparative income distribution research provides a good illustration. Before the advent of the Luxembourg Income Study (LIS), OECD presented a famous, but now obscure, report on how income inequality varied across OECD countries.8 The work put into the report was tremendous, but as discussed by commentators,9 the author had to perform aggregate, secondary analyses on national sources, was unable to analyse the micro-relation between pre- and post-taxes and transfers, and could make no adjustments for household size and structure. No doubt, there are still many difficult comparability problems in LIS, but the general conclusion nowadays is that Sawyer's study was so blurred and erroneous that nobody cites it apart from making similar points as I do here. Comparing pre- and post-welfare state redistribution has a fundamental impact on the degree of inequality. In other words, including taxes and transfers in our measure of income strongly reduce inequality. When examining the relationship between income and health this leads to a less steep slope. Moreover, it is important to realize that the extent of re-ranking as we compare along the income distribution process (from market income to equivalent disposable household income) is substantial, in particular in voluminous welfare states.10 This means that those we observe as having the lowest income are not necessarily the same people if we compare the distribution pre- and post-welfare state redistribution. This obviously could have consequences for the relationship between income and health, although it is more difficult to state exactly how. Now as already mentioned Mackenbach et al. are very well aware of this and other comparability problems. I fully agree with them that the inherent problems in their data lead to a focus more on similarities than differences.
Wilkinson's original income inequality hypothesis11,12 has, over the years, been rephrased so that the key question is whether there is a contextual effect over and above the individual association. Second, the vast majority of all studies deal with contextual variation within one specific nation, namely the US. Both these developments lead me to some critical reflections. First, the question of a direct effect of income inequality, over and above a correctly specified micro relation, is indeed scientifically interesting. However, from a public health perspective it is enough that we find a curvilinear association to state that income inequality has negative health effects.13 The main policy implication of Wilkinson's idea was that reducing income inequality would improve population health. If we find that it is a general feature of the income and health association to be curvilinear this policy implication is still as important and valid as ever before. This statement relies of course on one important assumption, namely that the relation, at least partly, is a causal one.
Second, the study of income inequality in the US raises the fundamental issue of generalization. Is the US an exceptional case or not? Likewise important, and seldom discussed, is the question of whether or not a hypothesis on one aggregation level is valid at a different level of aggregation. As Lieberson (p. 108) put it: ‘If the conceptual or theoretical issue is on a given level of analysis, then the empirical evidence obtained on a lower level will not be relevant for determining the merit or validity of the theory.’14 A lower level analysis could still be of interest if we could make a plausible hypothesis of how causal processes on different levels interact with each other. From this perspective one could go even further and say that it is plausible, to my mind, that the distribution of income at various levels of aggregation are inversely related to each other and thereby could also have opposite health effects. For example, if the context is a small area, i.e. a neighbourhood, a low degree of income inequality could be a sign of high economic segregation which in turn might lead to higher income inequality at a higher level of aggregation.
Of course, this type of discussion becomes meaningful only in relation to empirical findings like those presented by Mackenbach et al. The complexity involved should not stop us from pursuing comparative studies, but rather inspire us to work even harder with theoretically guided research in the future.
 |
References |
|---|
 Top References |
|---|
1 Backlund E, Sorlie PD, Johnson NJ. The shape of the relationship between income and mortality in the United States: Evidence from the National Longitudinal Mortality Study. Ann Epidemiol 1996;6:12–20.[CrossRef][Web of Science][Medline]
2 Ecob R, Davey Smith G. Income and health: what is the nature of the relationship? Soc Sci Med 1999;48:693–705.[CrossRef][Web of Science][Medline]
3 Fritzell J, Nermo M, Lundberg O. The impact of income: assessing the relationship between income and health in Sweden. Scand J Pub Health, 2004;32:6–16.
4 Mackenbach JP, Martikainen P, Looman CWN, Dalstra JAA, Kunst AE, Lahelma E. The shape of the relationship between income and self-assessed health: an international study. Int J Epidemiol 2005;34:286–93.
5 Subramanian SV, Kawachi I. The association between state income inequality and worse health is not confounded by race. Int J Epidemiol, 2003;32:1022–28.
6 Lynch J, Harper S, Davey Smith G. Commentary: Plugging leaks and repelling boarders – where to next for the SS Income Inequality? Int J Epidemiol 2003;32:1029–36.
7 Kohn ML. Cross-National research as an analytic strategy. In: Kohn, ML (ed.). Cross-National Research in Sociology, Newbury Park: SAGE, 1989.
8 Sawyer M. Income distribution in OECD countries. OECD Economic Outlook, Occasional Papers, 1976; 3–36.
9 Atkinson AB. Incomes and the Welfare State. Cambridge: Cambridge University Press, 1995.
10 Fritzell J. The gap between market rewards and economic well-being in modern societies. Eur Sociol Rev 1991;7:19–33.
11 Wilkinson RG. Income distribution and life expectancy. BMJ 1992;304:165–68.
12 Wilkinson RG. Unhealthy Societies: The Afflictions of Inequality. London: Routledge, 1996.
13 Deaton A. Health, inequality, and economic development. J Econ Lit 2003;XLI:113–58.[CrossRef]
14 Lieberson S. Making it Count. Berkeley: University of California Press, 1985.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||