Commentary: Theory in the fabric of evidence on the health effects of inequalities in income distribution

Miquel Portaa,b, Carme Borrellc and José L Copeteb

a Institut Municipal d'Investigació Mèdica (IMIM-IMAS), Barcelona, Spain.
b Universitat Autònoma de Barcelona, Barcelona, Spain.
c Institut Municipal de Salut Pública (IMSP), Barcelona, Spain.

Prof. Miquel Porta, Institut Municipal d'Investigació Mèdica, Universitat Autònoma de Barcelona, Carrer del Dr. Aiguader 80, E-08003 Barcelona, Spain. E-mail: mporta{at}imim.es

Theory comes first. Thus, Rodgers' second paragraph1 begins: ‘Theory: Let us suppose that at the individual level there is a relationship between income and life expectancy.’ Then Rodgers expresses his seemingly simple idea in a (slightly less) simple diagram. By drawing the graph, the scholar has left the pier of ‘pure’ concepts; but he is not yet sailing the open sea of quantitative analysis (where theoretical concepts are the guiding stars, of course). He further notes: ‘the relationship in Diagram 1 is defined for an individual.‘1 What about the empirical data? That is a bit like the breeze you need to fill the sails, isn’t it ... Well, the author realizes that ‘in practice, data for studying this type of relationship are available only at the aggregate level’. This—we might call it—’minor practical problem’ entails, he writes, ‘the need to formulate a relation between life expectancy at community or national level, and the incomes of the individuals composing the society concerned'. It seems as if only then does a new hypothesis—radically new, it will turn out—occur to him: ‘the mean life expectancy is a function not only of the mean income level, but also of the distribution of income.’ This seminal idea—does it become apparent to him only by ‘thinking the diagram‘? Surely, from his diagram ‘it is clear that there will be a tendency for greater dispersion of income to be associated with lower mean life expectancy‘.1 The diagram is not semi-theoretical and semi-quantitative, but both theoretical and quantitative—yet, it is still untested.

By the way, what would have happened if data at the individual level had been available to Rodgers? Such is often the case, nowadays ...2

So, theory comes first, then the data; and the nature of available data—once carefully, not gratuitously thought out—gives shape to a new dimension of the theory, which is then empirically tested. At the end of the paper Rodgers seems to conceal his amazement when concluding: ‘The most striking result is the consistent significance of the income distribution variable.‘1 This is the ‘very robust conclusion’ that holds across a variety of statistical models: greater income inequality is associated with higher mortality. We comment on this immediately below. Also most appealing to us today, is the inner structure of Rodgers' paper; in particular, the way theory, hypothesis formulation, model specification, statistical analyses, and conclusions are intertwined in a coherent fabric. Theory does not clothe the data: it weaves in with the data. Hence the unique texture of the evidence knitted by Rodgers.1

How many public health journals would publish Rodgers' article today? We fear many would not. We do not think it is a trivial question. The structure of the paper and its format faithfully reflect the process of inquiry and thinking, and the careful dialogue between theory and data. So also does the parsimonious explanation of methods and results, the tables, the omission of many quantitative results, the interspersed comments ... It is all quite unconventional by today's editorial standards. Not only editorial standards, in fact: unconventional by many standards. It is at odds with most papers we write and read, with the predominant ways of working and thinking ... At odds, perhaps, with the place of theory in our current weaving of epidemiological evidence. By the way, how much thought do you feel the diagram deserves?

The theories and mechanistic findings of Richard Wilkinson and others

Naturally, the most interesting aspect of Rodgers' paper is the main finding itself: greater inequality in income distribution is associated with higher mortality.1 Through an intense, and at times difficult, process of inquiry developed over the past few decades, a diversity of theories and data have provided a rich body of evidence on the impact upon health indicators of the shape of the income distribution. Much of the evidence, but by no means all,2 supports the idea that the level of population well-being depends not only on the absolute income of a society, but also on the distribution of income across society. At an aggregate level, among less developed nations, a relationship frequently exists between increase in GNP per capita and improvement in health. This relationship may not hold for developed societies, where the income distribution may matter more. The paper by Rodgers1 was one of the earliest studies, if not the earliest, to empirically show that. (The history of political science would, no doubt, provide the ideological and social roots for Rodgers' work; unfortunately, the analysis of such historical foundations is beyond our expertise, as we further acknowledge later.) Subsequently, the work of Richard Wilkinson was fundamental in defending this relative income hypothesis.3–6 Wilkinson's work is mainly based on studies showing a relationship between income inequality and life expectancy at the cross-national level. This evidence has expanded in the last 5–7 years with studies focusing on income inequalities within countries, mainly in the US (states and metropolitan areas).2,7–10

The importance of relative income on health has been found not only in ecological designs but also in multilevel studies, where individual data as well as aggregated data have been integrated. For example, in a cross-sectional multilevel study, Kennedy et al.11 detected a relationship between income inequality and self-perceived health, after controlling for personal characteristics and absolute income. A relationship between health and income distribution has also been observed using different measures of income distribution.12

On the other hand, not all studies using individual and aggregate data found relationships between income inequality and health indicators.2 For instance, a recent aggregate study using data from 16 countries did not find a relationship between income inequality and life expectancy, although it did find a relationship with infant mortality.13 This study also found inconsistent associations between indicators of mortality and social capital. Other studies using both individual and aggregate data found different results: one found a relationship between income inequality and mortality,14 while another did not observe this relationship.15

As in other branches of science (including public health and the social sciences),16 research on health and social inequalities has increasingly dived into ‘black boxes‘; i.e. it has focused on causal processes and pathways, mediators, effect modifiers, and mechanisms of all sorts. As an example, Kawachi, Kennedy and Wilkinson17,18 point out that the mechanisms by which the societal distribution of income could affect health can be conceptualized in three categories:

(1) By influencing investment in human capital. As income inequality increases, the interests of the rich and the poor diverge. The greater the gap in income, the greater the disparity in interests. This often implies pressure from the upper social classes for lower taxes. It may then happen that social spending (education, health, infrastructure) diminishes, leading to diminished opportunities for the poor.

(2) Through social processes, such as disruption of the social fabric. Societies with more income inequality may have less social cohesion and social capital. The diminution of the social fabric is then related to poor health through several possible mechanisms (e.g. social cohesion may influence health behaviours in a community, social capital may facilitate increasing access to local services and amenities).

(3) Through psychosocial processes. The perceived widening of the gap leads to frustration, with possible health consequences. Other psychosocial processes that can be involved are hopelessness, sense of control over life, job insecurity, etc.

The focus on mechanisms in no way diminishes the importance of theory. Mackenbach, for instance, has questioned whether educational achievement should be treated as a confounder or as an intermediary between income inequality and mortality. He has also stressed the importance of coherent integration of theoretical concepts and empirical analyses.2

Some criticisms: the ‘whys‘, ‘whats' and ‘hows'

Muntaner and Lynch19,20 have offered a critical appraisal of the income inequality and social cohesion model, which can be summarized as follows.

1. Income inequality
The theory is in practice restricted to developed countries. It thus neglects the impact of international economic relations on the level of income inequality. Yet, world income inequality affects within- and between-country income inequality. Another issue that the theory should address is the social mechanisms that generate income inequalities; theories of social stratification and class analysis try to explain how class positions in a society generate income inequalities. Usually high income is associated with capitalist classes. The model by Wilkinson et al. does not include the central issue of what produces economic inequality in the first place. Neo-Marxian models proffer social mechanisms that aim to explain why and how income inequality occurs.21,22

2. Social cohesion
According to some critics,23–25 this concept needs to be better defined. Social cohesion may have both good and bad consequences on health (e.g. in fascist societies, social cohesion has been used as a mean of social control).19 It is also necessary to analyse the relation between class and social cohesion, as well as the political aspects related to income inequality and social cohesion. For instance, whilst countries belonging to the former Soviet Union have suffered from a breakdown of social cohesion, to understand such breakdown the change from a socialist state to a capitalist one must be analysed.19,24 Furthermore, models that emphasize social cohesion as a potential determinant of population health have implications for social and health policy. For ease, policy-makers will prefer the call to ‘increase social cohesion’ in order to reduce health inequalities, rather than proposals to reduce income inequalities through taxation and state transfers. Some views of the role of social cohesion make the community ultimately responsible for their health.19 Thus, again, policy-makers as well as scholars choose specific threads of theory and data shreds to weave their policies and papers ...

The trail left by Rodgers' paper in the academic literature

The influence of Rodgers' paper1 appears to have been considerable. To assess it, there can be no substitute for scholarly reading—witness the contributions of our colleagues in these pages of the International Journal of Epidemiology, or the anthology edited by Ichiro Kawachi et al., which Rodgers' paper opens.17 Beyond the scholarly literature, the influence of Rodgers' work on social organizations and policies would be harder—though no less important—to assess (see below).

We also thought it might interest readers of this debate to know the papers that have used or otherwise mentioned Rodgers' article.1 We thus searched for citations to it made by papers included in the Social Sciences Citation Index (SSCI) data base (1981–2001) and in the Science Citation Index (SCI) data base (1980–2001). We are aware of the limitations that this approach has, but it seemed useful, nonetheless ...26 We found 70 such papers, including just one of his own (see Appendix on web-site http://ije.oupjournals.org). The vast majority appeared in journals included in the SSCI. Figure 1Go shows the number of citations received by Rodgers' paper1 since its publication. As can be seen, references have continued in recent years. Social Science and Medicine is the journal that has published most papers citing Rodgers' (9 papers), followed by the British Medical Journal (6), the American Journal of Public Health (also 6) and Social Indicators Research (4). Authors citing Rodgers' article most frequently were: RG Wilkinson (7 papers), JB Williamson (4), JW Lynch (4) and, with 3 papers each, GA Kaplan, I Kawachi, BP Kennedy, JM Mellor and JD Milyo (some of them were co-authors of the same paper).



View larger version (15K):
[in this window]
[in a new window]
 
Figure 1 Number of citations received by Rodgers' paper (Population Studies 1979;33:343–51) over the years

Sources: Institute of Scientific Information, Inc.: Web of Science, Social Science Citation Index (1981–2001) and Science Citation Index (1980–2001).

 
Questions we cannot answer

The influence of Rodgers' work surely stemmed away from the scholarly literature and has expanded beyond it. Unfortunately, this is but one among several important issues that we are unable to address. However, they seem important enough to formulate, in the hope that they will be addressed by more knowledgeable colleagues in the near future. So, how were these and other findings by Rodgers received by the academic, political and economic communities? In particular, what influence did his research have on social policies in different countries and international organizations? Who was GB Rodgers, what did he do before and after leaving the International Labour Organization (ILO) World Employment Programme? Was the ILO interested in income inequalities as determinants of human mortality, or was this Rodgers' more personal endeavour? What was the political, economic and scientific context in which the analyses of Rodgers and others arose, and how were they mutually linked?

Answers to these and related questions would certainly help us today to understand the process through which theory, hypothesis formulation, model specification, statistical analysis and conclusions were intertwined to produce a unique fabric; a multi-layered tapestry of the relationships between health and income distribution.

References

1 Rodgers GB. Income and inequality as determinants of mortality: An international cross-section analysis. Popul Stud 1979;33:343–51.[ISI]

2 Mackenbach JP. Income inequality and population health [editorial]. BMJ 2002;324:1–2.[Free Full Text]

3 Wilkinson RG. Income and mortality. In: Wilkinson RG (ed.). Class and Health: Research and Longitudinal Data. London: Tavistock, 1986, pp. 88–114.

4 Wilkinson RG. Income distribution and mortality: a ‘natural’ experiment. Sociol Health Illn 1990;12:391–412.[CrossRef][ISI]

5 Wilkinson RG. Income distribution and life expectancy. BMJ 1992; 304:165–68.[ISI][Medline]

6 Wilkinson RG. Unhealthy Societies. The Afflictions of Inequality. London: Routledge, 1996.

7 Kaplan GA, Pamuk E, Lynch JW, Cohen RD, Balfour JL. Income inequality and mortality in the United States: analysis of mortality and potential pathways. BMJ 1996;312:999–1003.[Abstract/Free Full Text]

8 Kennedy BP, Kawachi I, Prothrow-Stith D. Income distribution and mortality; cross sectional ecological study of the Robin Hood Index in the United States. BMJ 1996;312:1004–07.[Abstract/Free Full Text]

9 Ben-Shlomo Y, White IR, Marmot M. Does the variation in the socioeconomic characteristics of an area affect mortality? BMJ 1996;312:1013–14.[Free Full Text]

10 Lynch JW, Kaplan GA, Pamuk ER et al. Income inequality and mortality in metropolitan areas of the United States. Am J Public Health 1998;88:1074–80.[Abstract]

11 Kennedy BP, Kawachi I, Glass R, Prothrow-Stith D. Income distribution, socioeconomic status and self-rated health: a US multi-level analysis. BMJ 1998;317:917–21.[Abstract/Free Full Text]

12 Kawachi I, Kennedy BP. The relationship of income inequality to mortality: does the choice of indicator matter? Soc Sci Med 1997;45:1121–27.[CrossRef][ISI][Medline]

13 Lynch J, Davey Smith G, Hillemeier M, Shaw M, Raghunathan T, Kaplan G. Income inequality, the psychosocial environment and health: comparisons of wealthy nations. Lancet 2001;358:194–200.[CrossRef][ISI][Medline]

14 Lochner K, Pamuk E, Makuc D, Kennedy BP, Kawachi I. State-level income inequality and individual mortality risk: a prospective multilevel study. Am J Public Health 2001;91:385–91.[Abstract]

15 Osler M, Prescott E, Gronbaek M, Christensen U, Due P, Engholm G. Income inequality, individual income, and mortality in Danish adults: analysis of pooled data from two cohort studies. BMJ 2002;324:1–4.[Free Full Text]

16 Porta M, Álvarez-Dardet C. Epidemiology: bridges over (and across) roaring levels. J Epidemiol Community Health 1998;52:605.[Free Full Text]

17 Kawachi I, Kennedy BP, Wilkinson RG (eds). Income and Inequality. The Society and Population Health Reader. Vol. I. New York: The New Press, 1999.

18 Kawachi I. Income inequality and health. In: Berkman LF, Kawachi I (eds). Social Epidemiology. New York: Oxford University Press, 2000, pp. 76–94.

19 Muntaner C, Lynch J. Income inequality, social cohesion and class relations: a critique of Wilkinson's neo-Durkheimian research program. Int J Health Serv 1999;29:59–81.[ISI][Medline]

20 Muntaner C, Lynch J, Oates G. The social class determinants of income inequality and social cohesion. Int J Health Serv 1999;29: 699–732.[ISI][Medline]

21 Wright EO. Clases. Madrid: Siglo XXI de España Editores, 1994.

22 Wright EO. Class Counts. Comparative Studies on Class Analysis. New York: Cambridge University Press, 1997.

23 Lynch J, Muntaner C, Due P, Davey Smith G. Social capital. Is it a good investment strategy for public Health? J Epidemiol Community Health 2000;54:404–08.[Free Full Text]

24 Muntaner C, Lynch J, Davey Smith G. Social capital, disorganized communities and the Third Way: Understanding the retreat from structural inequalities in epidemiology and public health. Int J Health Serv 2001;31:213–37.[ISI][Medline]

25 Whitehead M, Diderichsen F. Social capital and health: tip-toeing through the minefield of evidence. Lancet 2001;358:165–66.[CrossRef][ISI][Medline]

26 Porta M. The bibliographic ‘impact factor’ of the Institute for Scientific Information: how relevant is it really for Public Health journals? J Epidemiol Community Health 1996;50:606–10.[ISI][Medline]