income inequality and health: acritical review of the ... · the theoretical literature was used to...

10.1177/1077558703257169ARTICLEMCR&R 60:4 (December 2003)Macinko et al. / Income Inequality and Health

Review

Income Inequality and Health:A Critical Review of the Literature

James A. MacinkoNew York University Steinhardt School of Education

Leiyu ShiJohns Hopkins Bloomberg School of Public Health

Barbara StarfieldJohns Hopkins Medical Institutions

John T. Wulu, Jr.Bureau of Primary Health Care, HRSA/DHHS

This article critically reviews published literature on the relationship between incomeinequality and health outcomes. Studies are systematically assessed in terms of design,data quality, measures, health outcomes, and covariates analyzed. At least 33 studiesindicate a significant association between income inequality and health outcomes, whileat least 12 studies do not find such an association. Inconsistencies include the following:(1) the model of health determinants is different in nearly every study, (2) incomeinequality measures and data are inconsistent, (3) studies are performed on differentcombinations of countries and/or states, (4) the time period in which studies are con-ducted is not consistent, and (5) health outcome measures differ. The relationshipbetween income inequality and health is unclear. Future studies will require a more com-prehensive model of health production that includes health system covariates, sufficientsample size, and adjustment for inconsistencies in income inequality data.

Keywords: income inequality; health determinants; social epidemiology; healthinequalities

Over the past decade, there has been an ongoing debate over the role ofincome inequality as a determinant of population health. The debate contin-

Medical Care Research and Review, Vol. 60 No. 4, (December 2003) 407-452DOI: 10.1177/1077558703257169© 2003 Sage Publications

407

at WELCH MEDICAL LIBRARY on September 7, 2010mcr.sagepub.comDownloaded from

http://mcr.sagepub.com/

ues, with evidence both for and against what has become known as the rela-tive income hypothesis—the proposition that the distribution of incomewithin societies (independent of absolute levels of income or wealth) is animportant determinant of that population’s health. This document presents acritical examination of the public health and social science literature describ-ing the relationship between income inequality and health status. It reviewsthe numerous empirical studies in the area, assesses the strength of studydesigns and data sources, presents a summary of the major theories describ-ing the observed relationship, and proposes a comprehensive conceptualframework and specific considerations that should be taken into account forfuture studies.

NEW CONTRIBUTION

This review differs from previous efforts (see Wagstaff and van Doorslaer2000; Lynch, Smith, et al. 2000) because it looks at a comprehensive range of lit-erature, both within and outside of the public health field. This review exam-ines nearly 50 studies on income inequality and health, including the follow-ing: (1) studies aiming to demonstrate or refute the relationship betweenincome inequality and health, (2) theoretical and empirical literature on thepathways through which income inequality could influence health, and (3)social science literature on the production of income inequalities within andbetween countries. This review also emphasizes the role of the health systemas a potential mediator of the relationship, a factor largely unexplored in theexisting literature. Finally, this review proposes a comprehensive conceptualframework for ordering and understanding the relationships among political,economic, social, and biological factors proposed as causes, consequences, orconfounders of the relationship between income inequality and health out-comes. Such a framework is intended not only to organize this review but alsoto suggest a more conceptually sound model for exploring the relationshipbetween income inequality and health outcomes.

METHOD

This integrative literature review was conducted between October 2001and January 2002. It utilized methods identified by Shi (1996) and Fink (1998).The search used the following electronic databases: PubMed, OVID (in-

408 MCR&R 60:4 (December 2003)

This article, submitted to Medical Care Research and Review on April 22, 2002, was revised and ac-cepted for publication on October 23, 2002. This study was partially funded by a grant (T32 HS00029) from the Agency for Healthcare Research and Quality.



cluding all sociological, political science, psychology, education, and econom-ics journals), and dissertation abstracts. Each database was searched for arti-cles that contained the phrase “income inequality AND health” in either thetitle or abstract. The search was not limited by date but only included articleswritten in English, French, or Spanish. This process revealed 327 potentialarticles. In addition, individual journals that publish frequently on the topicwere each searched from 1990 to the present. This revealed an additional 30relevant articles. All articles were then culled to identify another 18 references.Finally, to ensure that the search parameters were wide enough, an additionalsearch for the terms income inequality and health AND inequality was conductedthrough the Library of Congress catalog, which, after inspection of titles anddescriptions, revealed 17 relevant books. After inspection of abstracts andevaluation of methodologies, 47 key articles were retrieved and abstracted(see the appendix). Only articles published in peer-reviewed journals and rep-resenting empirical studies or literature reviews were abstracted, althoughother articles, books, and editorials are discussed throughout this document.The theoretical literature was used to supplement empirical studies todevelop the conceptual framework and assess the adequacy of the empiricalmodels reviewed, especially in terms of the plausibility of results and theinclusion of a minimum set of covariates used in multivariate models.

CONCEPTUAL FRAMEWORK

To organize this review and to attempt to integrate the many strands of lit-erature examining the relationship between income inequality and health, wepropose a multidimensional conceptual framework, as presented in Figure 1.The framework is based on the work of Starfield and Shi (1999) and others buthas been modified based on findings of the literature reviewed here. Thus, theframework is an outcome and a method of organizing this review.

Figure 1 describes distal and proximal determinants of health at the macroand micro levels. It organizes these determinants into four main columns.Beginning at the left, Panel Arepresents national political, cultural, and histor-ical factors that produce income inequality and influence all other healthdeterminants. Examples include systems of governance, legal and institu-tional factors governing union formation and health financing, cultural atti-tudes toward welfare and charity, and historical patterns of social relations.

Panel B shows income inequality. It is important to note that previous stud-ies have positioned income inequality in different places within the frame-work. Our review of the sociology and economics literature suggests that it ismost appropriately viewed as a consequence of political and cultural factors,and a potential modifier of macro- and microlevel health determinants. The

Macinko et al. / Income Inequality and Health 409



literature reviewed here does not suggest that there is any direct effect ofincome inequality on health outcomes. Instead, income inequality is hypothe-sized to work through one or more health determinants.

Panel C shows macrolevel determinants of health. In the literaturereviewed here, none of these factors (with the possible exception of socialcohesion) is hypothesized to have a direct impact on health, but all are none-theless important precursors to the factors presented in Panel D. Macrolevelhealth determinants include the physical environment (e.g., air and water pol-lution, community or neighborhood characteristics, and infrastructure), mac-roeconomic environment (e.g., employment, gross national product [GNP],cost of living), social relations and cohesion (e.g., social class structure, com-munity-level social capital), and health and welfare services (e.g., primarycare services, educational institutions, welfare programs). Note that theframework does not attempt to describe the relationships among factorswithin each panel.

Panel D describes microlevel health determinants. These include expo-sures and risk factors (e.g., exposure to pollutants and external causes of inju-ries, and unhealthy behaviors such as drinking and smoking), economicresources (e.g., income), social resources (e.g., social networks, coping


C.1Physical

Environment

C.2Macro

Economics

C.3 SocialRelations/Cohesion

C.4 Health& OtherServices

A.1 National Politics & Policies

A.2National Culture &

History

Inequality

D.1Exposures

D.2EconomicResources

D.3Social

Resources

D.4Access/Useof Services

E.1Health

Outcomes

B.1 Income

FIGURE 1 A Simplified Conceptual Framework of the Relationship betweenIncome Inequality and Health Outcomes



abilities), and access and use of services (e.g., ability to use health services,obtain medicines, and join Women, Infants, Children [WIC] programs when-ever needed). It is clear that these factors are often interrelated (e.g., availabil-ity of income can help determine whether one will use health services), but wemake no attempt to tease out the relationships among the separate variables.

Finally, Panel E represents health outcomes. These include life expectancy,mortality, self-rated health, morbidity, mental health, and others. It is likelythat a more detailed look at the causes of these outcomes would necessitatemodifications to the conceptual framework; however, for the purposes of thisreview, we do not make such distinctions.

The framework presented above organizes this review. First, this reviewdescribes the literature examining simple associations between incomeinequality and health. Then, it examines studies that propose a pathwaybetween income inequality and health outcomes. It then reviews criticism ofthese approaches and presents alternative explanations for the observed rela-tionship between income inequalities and health. Finally, we present an analy-sis of the empirical evidence and provide suggestions for future studies onthis topic.

INCOME INEQUALITY AND HEALTH

The health and social science literature is replete with studies of the impactof income, poverty, and social policies on the health of individuals (see over-views by Rodriguez-Garcia and Goldman 1992; World Bank 1993; Basch1990). Numerous studies have supported the hypothesis that individualsocioeconomic status (Antoft, Gadegaard, and Lind 1974; Blaxter 1987;Hollingsworth 1981; Mackenbach et al. 1997; Marmot et al. 1991) or its compo-nents of income (Blackburn 1994; Grant 1977; Salkever 1975; Wilkins, Adams,and Brancker 1989), education (Christenson and Johnson 1995), and occupa-tion (Chandola 1998; Robinson 1984) have profound effects on an individual’shealth. Similar results have been found for social class (Antonovsky 1967;Bunker 1983; Dougherty, Pless, and Wilkins 1990; Gregory and Piche 1983;Soderfeldt, Danermark, and Larsson 1989).

In the mid-1970s, researchers began to doubt whether national income con-tinued to play a role in determining population health within industrializedcountries (Fuchs 1974; Preston 1976). It appeared that at a certain level ofdevelopment, additional increases in income had little effect on increasingnational life expectancy (Preston 1976). This coincided with Omran’s epide-miological transition theory that stated that as countries move towardincreased levels of economic prosperity, two things can be expected to hap-pen. First, an epidemiological transition will take place: the cause of deaths




will shift from infectious to chronic diseases. Second, a demographic transi-tion will take place: the burden of mortality will become increasingly made upof the old rather than the very young, and overall life expectancy will increase(Omran 1971).

Although the literature search revealed several early studies on the distri-bution of health within countries (see Antoft, Gadegaard, and Lind 1974;Mathers 1974; Smith and Kaluzny 1974), it was with the 1980 publication ofBritain’s Black Report on health inequalities that intense research efforts beganto focus less on aggregate income and health measures and more on the distri-bution of economic and health benefits across different social and economicgroups within and across societies (Black et al. 1980). Consequently, researchon variations in health status among people with different socioeconomic sta-tus has intensified over the past two decades.

An unexpected and important finding has been that better health outcomesappear to be positively correlated not only with absolute levels of income but,in some cases, even more strongly correlated with the equitable distribution ofincome within society. Wilkinson (1994, 61) summarized this position by stat-ing that “mortality rates are no longer related to per capita economic growth,but are related instead to the scale of income inequality in each society.”

The earliest studies on the relationship between income inequality andhealth stressed associational measures between health outcomes and levels ofincome distribution. Referring to our framework in Figure 1, these studiesassess the association between income inequality (B.1) and a variety of healthoutcomes (E.1) and occasionally control for other health determinants such asGNP (C.2) or health system resources (C.4).

Rodgers (1979) found that in multiple regressions using 56 countries, thelevel of income inequality (as measured by the Gini coefficient) was a signifi-cant predictor of life expectancy. The study estimated that differences in lifeexpectancy between an egalitarian and nonegalitarian country might be asmuch as 5 to 10 years. This relationship held for all countries studied but wasless pronounced among a sample of developing countries with a GNP percapita of less than US$1,000.

Further cross-sectional ecological studies by Waldman (1992) conductedwith 1960 and 1970 data from 56 countries showed that a rise of 1 percent ofnational income held by the richest 5 percent of a given population couldresult in an increase in the infant mortality rate (IMR) of about 2 infant deathsper 1,000 live births. Consistent with Rodgers (1979), this effect was somewhatreduced in developing countries but also significant, even after controlling forGDP per capita, total number of physicians and nurses per 1,000 population,percentage of the country that was urban, levels of primary school enrollment,female literacy, and the gross reproductive rate.




Wilkinson (1992, 33) made the claim that few countries having achieved alife expectancy of 70 years or more had a GNP of less than US$5,000 per yearand that beyond that threshold “there is little systematic relation betweengross national product per head and life expectancy.” This led Wilkinson toposit what has become known as the relative income hypothesis. According toWilkinson,

the sense of relative deprivation, of being at a disadvantage to those better off,probably extends far beyond conventional boundaries of poverty. A shift in em-phasis from absolute to relative standards indicates a fall in the importance of di-rect material circumstances relative to psychosocial influences. The social conse-quences of people’s differing circumstances in terms of stress, self-esteem, andsocial relations may now be one of the most important influences on health.(p. 34)

More sophisticated research designs offered support for the relative in-come hypothesis. For example, several studies suggested that the relationshipheld for units of analysis smaller than countries. One study found a stronggradient in mortality related to the gap between the rich and poor within Eng-lish wards/counties (Ben-Schlomo, White, and Marmot 1996). The authorssuggested that the effect of relative income is primarily an ecological orcontextual phenomenon.

Kaplan et al. (1996) found a statistically significant correlation between thepercentage of income received by the less well-off and all-cause mortalityamong U.S. states. Variations in states’ income inequality were also signifi-cantly associated with homicides, crime, low birth rate, educational attain-ment, disability, expenditures on health care, imprisonment, and lack ofmedical insurance.

Kennedy, Kawachi, and Prothrow-Stith (1996) found that inequality in thedistribution of income explains a significant proportion of cross-state variancein several causes of mortality in the United States, independent of poverty andsmoking. The size of the gap in income between rich and poor—distinct fromthe average income level of the poor—was related to mortality within U.S.states. They used the Robin Hood Index (the percentage of income of the rich-est 50 percent of the population necessary to transfer to the poorest 50 percentof the population to have total income distributed equally between the twopopulation halves) as an alternate measure of income inequality. This mea-sure is thought to be less sensitive to outliers in the highest and lowest incomecategories.

Lynch et al. (1998) suggested that excess mortality within U.S. metropolitanstatistical areas (MSAs) attributable to income inequality could be as high as




64.7 to 95.8 deaths per 100,000, depending on the income inequality measureused. The authors concluded that if U.S. states with the highest incomeinequality could redistribute their income to the level of those U.S. states withthe lowest income inequality, then the high income inequality states couldreduce their overall mortality by as much as 139.8 deaths/100,000. This mor-tality difference was similar in magnitude to the combined loss of life fromlung cancer, diabetes, motor vehicle accidents, HIV, suicide, and homicide in1995.

Several studies present evidence that the relative income hypothesis alsoholds for health outcomes other than mortality. For example, Diez-Roux, Link,and Northridge (2000) found that state inequality was associated with severalcardiovascular disease risk factors (body mass index, hypertension, andsedentarism). Income inequality was also found to be associated with higherabdominal weight gain among a representative sample of U.S. men (Kahn etal. 1998).

Soobader and LeClere (1999) showed that income inequality exerts an inde-pendent adverse effect on self-rated health at the county level. Kahn et al.(2000) found that women in the lowest income categories were more likely toreport depressive symptoms or fair/poor health. Among women in the lowestincome quintiles, those women who lived in states with higher incomeinequality were more likely to report depressive symptoms or fair/poorhealth status than comparable women who live in states with lower incomeinequality, even while adjusting for individual characteristics such as age,marital status, education, race/ethnicity, and household size. The relation-ship was consistent for women in the three lowest income quintiles.

Multilevel designs have been used to measure the individual and contex-tual effects of income inequality on population health. For example, Kennedyet al. (1998) found that inequalities in income at the state level had an inde-pendent, negative effect on an individual’s risk of reporting fair or poorhealth. Inequalities in statewide income distribution were found to be associ-ated with self-rated fair or poor health, even while controlling for individualincome and other risk factors. Lochner et al. (2001), using a prospective multi-level design, also found that individuals living in states of higher incomeinequality had a 12 percent increased relative risk of mortality, even afteradjusting for age, gender, race/ethnicity, marital status, and annual income.

Studies on non-U.S. populations include those by Duleep (1995), who con-firmed the relationship between income inequality and mortality for a sampleof the Organization for Economic Cooperation and Development (OECD) andformer Soviet states. Humphries and van Doorslaer (2000) found that income-related health inequality in Canada was between that of the United Kingdomand the United States.




Chiang (1999) provided evidence from Taiwan that income distributionmay become a more important determinant of mortality than absolute incomeas a country develops economically. In 1976, absolute income (GNP) was apowerful and significant predictor of under-5 mortality in Taiwan. By 1995,the distribution of income (Gini) had become a more significant predictor ofmortality in Taiwan than GNP per capita.

PATHWAYS THROUGH WHICH INCOMEINEQUALITIES INFLUENCE POPULATION HEALTH

Several possible explanations have been offered for how income inequali-ties might affect health. These explanations are summarized in Table 1.

PSYCHOSOCIAL PATHWAYS

The first set of explanations is based on individual and society-levelpsychosocial characteristics, such as levels of interpersonal trust and socialcohesion. The theoretical basis for these explanations comes from the work ofEmile Durkheim. Referring to Figure 1, these studies primarily explore therelationships between income inequality (B.1), macrolevel (C.3) or microlevel(D.3) social factors, and health (E.1).

At the individual level, the psychosocial explanation posits that “cognitiveprocesses of social comparison” work to increase individuals’ levels of stressthat concomitantly lead to poorer health status. Stress, poor social support,and lack of control over one’s work are related to poor health and have agreater effect on those at the bottom of social hierarchies (Wilkinson 1996,1998). This hypothesis is supported by the Whitehall study that shows a con-tinuous social gradient in health, that is, within a social (or economic) hierar-chy, individuals at a given level in a hierarchy tend to exhibit poorer healththan individuals in the next highest level of the hierarchy. This relationshipholds for individuals at each and every level of the hierarchy (Marmot et al.1991).

Interestingly, Adam Smith also discussed the idea of relative income and itsimpact on social functioning and well-being. Sen (1999, 71) reminds us ofSmith’s concern that what counts as a necessity in one society is not absolutebut is determined by social norms and expectations. For example, to be able toappear in public and to carry out one’s social functioning may require a higherstandard of clothing or other forms of conspicuous consumption in a richersociety than in a poorer one, and “the same parametric variability may applyto personal resources needed for fulfillment of self-respect.”




The psychosocial explanation has its correlate at the macro level. Advo-cates of the relative income hypothesis have theorized that income inequalitymay lead to changes in society by creating a climate of mistrust, reduced coop-eration, and decreased propensity to join voluntary organizations. Severalauthors have variously termed this concept social cohesion, social trust, andsocial capital. In spite of difficulties with the definitions and measurement ofthese concepts (see Macinko and Starfield 2001), all suggest a social mecha-nism that is related to psychosocial stresses associated with the status andpower differentials caused by income inequalities. This explanation is basedon the role of social relations and networks in determining individual and


TABLE 1 Explanations for the Relationship between Income Inequalityand Health

Explanation Synopsis of the Argument

Psychosocial (micro):Social status

Income inequality results in “invidious processes of socialcomparison” that enforce social hierarchies causingchronic stress leading to poorer health outcomes for thoseat the bottom.

Psychosocial (macro):Social cohesion

Income inequality erodes social bonds that allow people towork together, decreases social resources, and results inless trust and civic participation, greater crime, and otherunhealthy conditions.

Neo-material (micro):Individual income

Income inequality means fewer economic resources amongthe poorest, resulting in lessened ability to avoid risks,cure injury or disease, and/or prevent illness.

Neo-material (macro):Social disinvestment

Income inequality results in less investment in social andenvironmental conditions (safe housing, good schools,etc.) necessary for promoting health among the poorest.

Statistical artifact The poorest in any society are usually the sickest. A societywith high levels of income inequality has high numbers ofpoor and consequently will have more people who aresick.

Health selection People are not sick because they are poor. Rather, poorhealth lowers one’s income and limits one’s earningpotential.



population-level health (Berkman and Kawachi 2000; Berkman and Syme1979; Durkheim 1951).

For example, Kawachi et al. (1997) found that income inequality withinstates (as measured by the Robin Hood Index) was correlated with groupmembership and lack of social trust. Aggregated rates of social mistrust (thepercentage of respondents who answered that “you can’t be too careful indealing with other people”), perceived lack of fairness (the percentage ofrespondents who answered that “most people would try to take advantage ofyou if they got the chance”), and per capita membership in voluntary organi-zations were associated with total state all-cause mortality while adjusting forpoverty and income inequality. Path analysis suggested that income inequal-ity leads to a lowering of social trust that then leads to increased mortality.

Kawachi, Kennedy, and Glass (1998) expanded on these findings. Theyused General Social Survey (GSS) measures of civic trust, reciprocity (helpful-ness of others), and civic engagement. Using a multilevel model, they foundthat a person living in a state with low levels of social trust was 40 percentmore likely to report lower self-assessed health status than someone living inan area of higher social trust. Similar results were found for those living instates characterized as having low and medium group membership and lowand medium reciprocity (Kawachi, Kennedy, and Glass 1998).

Kawachi, Kennedy, and Wilkinson (1999) further expanded on the mean-ing of social capital by proposing that among U.S. states, inequalities in envi-ronmental and social characteristics help to predict geographic variation incrime. The results suggested that state-level income inequality (as measuredby the Robin Hood Index) was highly associated with violent crime such ashomicide. Low levels of interpersonal trust and the number of female-headedhouseholds were correlated with higher homicide rates. Health status mea-sures were not included in the study. These studies follow others showingassociations between income inequality, crime, and other unfavorable socialconditions (see Chiu and Madden 1998; Sampson, Raudenbbush, and Earls1997; Walberg 1998).

Kawachi, Kennedy, and Brainerd (1998) applied measures of social capitalto explain increased mortality in Russia after the fall of the Soviet Union. Theyused voting rates and trust in government as social capital measures and alsoincluded a number of variables related to what the authors term social cohe-sion. These include crime, divorce rates, and conflicts in the workplace. Eachof these social indicators was strongly associated with all cause age-adjustedmortality and life expectancy for men and women.




NEO-MATERIAL PATHWAYS

The second main set of explanations emphasizes neo-material and eco-nomic factors and stems from either a Marxist or rational choice orientation.These studies primarily explore the relationships among income inequality(B.1), macroeconomic factors (C.2), individual economic resources (D.2), andhealth (E.1). They also occasionally incorporate measures of the physical envi-ronment (C.1) and individual exposures (D.1).

In many ways, the neo-material approach developed out of critiques of thepsychosocial pathway literature. It is founded on the proposition that materialfactors, such as income and living conditions, are the most importantdeterminants of health. In contrast to the Durkheimian theories of psychosocialstress and social cohesion, the individual-level income hypothesis holds thataggregate-level associations between income inequality and health reflectonly the individual-level associations between absolute income and health.The neo-material explanation incorporates the individual-level healthhypothesis but goes further by claiming “health inequalities result from thedifferential accumulation of exposures and experiences that have theirsources in the material world” (Lynch, Smith, et al. 2000, 1202). The effect ofincome inequality on health is thought to be the result of negative exposures,lack of resources, and systematic underinvestment in human, physical, health,and social infrastructure. Thus, it incorporates individual- and contextual-level factors but emphasizes that accumulated exposures resulting frompoorer physical conditions and reduced quality of education and other socialservices adversely affect the life chances and health outcomes of those at thelowest end of the social spectrum. As an illustration of this concept, Kaplan etal. (1996) found that U.S. states with high income inequality also had lowerrates of medical care expenditures and higher rates of unemployment.

Muntaner and Lynch (1999) argued that the psychosocial explanation isflawed because of the following: (1) Class relations are ignored, (2) interna-tional economic relations are not analyzed, (3) politics are not included as adeterminant of population health, (4) social cohesion is not well defined, (5)the role of medical care is underemphasized, and (6) there is little evidencethat social cohesion and equality are related.

Lynch, Due, et al. (2000, 406) offered an alternative explanation for the roleof social capital and income inequality on population health. They claimedthat “absolute and relative income differences may represent the unequal dis-tribution of the material conditions that structure the likelihood of possessingand accessing health protective resources; of reducing negative health expo-sures; and of facilitating full participation in the society.” Essentially, they




argued that disturbed social relations are the effect of health and socialinequalities, not their cause.

Lynch, Smith, et al. (2000) asserted that data for the association betweenpsychosocial pathways and income inequality are ambiguous at best. Theirobjections include the following: (1) the macrolevel social cohesion hypothe-sis ignores structural causes of the inequality, (2) it oversimplifies the role ofsocial cohesion, and (3) the data do not clearly show that as income inequalityincreases, social indicators also uniformly decrease. Instead, they suggestedthat systematic differences in material conditions (at the individual and con-textual levels) better explain the observed relationship between incomeinequality and health. Lynch, Smith, et al. suggested specific public policyinterventions such as investments in schooling, health care, social welfare,and working conditions and more equitable distribution of public and privateresources as ways to address inequalities in health. A similar argument waspresented by Bobak et al. (2000).

Lynch et al. (2001) conducted an ecological study that was improved overprevious studies because it used income inequality measures from the Lux-embourg Income Survey (LIS)—widely considered to be the most accurateand consistent income data available (Atkinson, Rainwater, and Smeeding1995)—and tested a wide range of health outcomes. The authors found that ina sample of 16 OECD countries, higher income inequality was consistentlyassociated with higher infant mortality. Associations between incomeinequality and all-cause mortality decreased with age, and actually reversedin populations older than 65 years. Cause-specific mortality showed no clearrelationship. Psychosocial variables (trust, organizational membership, vol-unteering) showed weak and inconsistent relationships. They concluded thatoverall, psychosocial pathways seem to play little role in the relationshipbetween income inequality and health (Lynch et al. 2001). The study did notinclude any measures of the medical or social services system.

THE ROLE OF HEALTH SERVICESAND PRIMARY CARE

In general, the studies reviewed here did not incorporate the role of healthservices in the analysis of the relationship between income inequality andhealth. In our conceptual framework, there are two main levels at which thehealth system may influence the relationship between income inequality andhealth: through the structure and availability of health and other social ser-vices (C.4) and individual access and use of specific services (D.4).




Although the literature indicates some skepticism as to the overall contri-bution of medical care to the improvement of population health worldwide(McKeown 1976; McKeown, Record, and Turner 1975; McKinlay and McKin-lay 1977), there is evidence that access to certain types of medical care may bemore beneficial than others in reducing a country’s overall burden of disease(Starfield 1996). For example, Bunker (2001) has suggested that in some coun-tries, as much as half of the 7.5 years of increased life expectancy since 1950 isthe result of improved medical care. Another study on the determinants ofdecline in mortality rates in the Netherlands from 1875 to 1924 likewise high-lighted the historical role of medical care—particularly public health inter-ventions and health education provided through primary care settings—incontributing to long-term improvements in population health (Wolleswinkelet al. 2001).

Other authors have suggested that at least some of the historical differencesin health status between the rich and poor may be in part due to differentialaccess to basic health services (Mackenbach, Stronks, and Kunst 1989). In par-ticular, primary care, defined as “that level of a health service system that pro-vides entry into the system . . . provides person-focused care over time, pro-vides care for all but very uncommon or unusual conditions, and coordinatesor integrates care provided elsewhere or by others” (Starfield 1998, 8-9), hasbeen shown to have an important impact on health outcomes for some of themost common medical problems (Starfield 1996). Furthermore, several stud-ies have demonstrated that the extent of primary care orientation (as opposedto a system based on physician specialists) has a statistically significant effecton improving life chances, particularly among the disadvantaged (Shi 1994).

Among the studies abstracted in the appendix, there were contradictoryfindings as to whether access to health services was associated with health sta-tus and income inequalities. Flegg (1982) found that the number of physiciansand nurses per capita had a statistically significant effect on health outcomesin models including income inequalities, while Judge, Mulligan, andBenzeval (1998a, 1998b) and van Doorslaer et al. (1997) found no consistentassociation. One possible explanation for this discrepancy is that Flegg’s anal-ysis included primarily developing and middle-income countries, whileJudge et al. and van Doorslaer et al. included only wealthy OECD countries intheir analysis. This choice of countries is important because generic measuresof medical care may be stronger predictors of overall population health statusin developing countries than in more wealthy ones. Moreover, none of theseauthors differentiated between primary care and specialty care, which is aparticularly important distinction when examining the impact of health careon outcomes in wealthy OECD countries (Shi and Starfield 2000; Starfield andShi 2002).




Other health and social services indicators, such as overall social expendi-tures (Gustafsson and Johansson 1999) and investments in medical care(Kaplan et al. 1996), have been found to be associated with lower incomeinequality, but these same expenditure variables have not always been foundto be associated with mortality (Judge, Mulligan, and Benzeval 1998a; vanDoorslaer et al. 1997).

Shi et al. (1999), in an ecologic study in the United States, found that incomeinequality and primary care (measured by primary care physicians per 10,000population) exerted a strong and statistically significant influence on state-level mortality and life expectancy. Path analysis suggested that primary caremight overcome some of the adverse health impacts of income inequality onpopulation health.

In a multilevel model including individual-, community-, and state-levelvariables, Shi and Starfield (2000) present stronger evidence for the ability ofprimary care to partially attenuate the adverse health effects of incomeinequalities. In multivariate analyses, income inequality and primary carewere significantly associated with self-rated health. Primary care significantlyattenuated the effects of income inequality on self-reported health status.Adding individual-level socioeconomic status variables somewhat reducedthe magnitude of the association between income inequality, primary care,and self-reported health. The study found that while controlling for allcovariates, an increase of one primary care physician per 10,000 populationwas associated with a 2 percent increase in the odds of reporting excellent/good health.

Although the authors do not propose any specific mechanisms, there maybe several possible explanations for the observed relationship betweenincome inequality, primary care, and health. One mechanism for such aninfluence is that access to a regular source of primary care may improve healthpromotion and early detection of diseases such as complications of hyperten-sion (Shea et al. 1992). Primary care is conceived of as that aspect of the healthcare system where the majority of care is provided through its role of first con-tact with the health system and its attributes of providing comprehensive,coordinated, and longitudinal care (World Organization of National Colleges,Academies and Academic Associations of General Practitioners/FamilyPhysicians [WONCA] 1991).

Second, there is evidence that strong primary care can lead to appropriateand more efficient secondary and tertiary care (Casanova, Colomer, andStarfield 1996; Casanova and Starfield 1995).

Third, because primary care implies person-oriented, longitudinal, andfirst-contact care, it is expected that in the best of circumstances, individuals




living in areas with high levels of primary care providers may develop animportant social tie with their primary care provider.

Finally, primary care may function as an important measure of social trans-fer that could compensate individuals for their relative income deprivationwith an essential social service.

CRITIQUES OF THE RELATIVEINCOME HYPOTHESIS

Although much of the early literature on income inequality and health con-firmed the findings that such a relationship existed, there is now a growingbody of literature that questions the validity of the relative income hypothesisand its explanatory theories. The studies reviewed here consist of critiques ofthe relative income hypothesis, negative findings, and alternative explana-tions for the observed relationship. These studies are organized into severalcategories to aid analysis and describe trends in the intellectual developmentof this literature. These categories include discussions of data quality, method-ological discussions, and alternative explanations for the observed relation-ship. Taken as a whole, these studies cast doubt on several features of the rela-tive income hypothesis and argue that the role of income inequality on healthmay be modest and perhaps limited to only certain societies and certain healthoutcomes.

MEASURING INCOME AND INCOME INEQUALITY

Income and income inequality are notoriously difficult to measure. Table 2provides a summary of the major measures of income inequality and theiradvantages, disadvantages, and most common data sources. In addition todifferences in the construction and interpretation of the income inequalitymeasures, the household income data on which inequality indices are derivedcan differ significantly from study to study. Two key concepts are disposable(net) versus total (earned) income, and income that has been equivalized forhousehold size. Disposable income refers to household income that is net oftaxes and other income transfers. The idea is that disposable income is a morevalid and comparable income measure because different households willhave different purchasing power, depending on the taxes and social transfers.Such differences in real purchasing power would not be captured by usingtotal (gross) family income. This distinction can be important. For example, ina country with a highly progressive tax system, using gross household incometo assess income distribution may indicate high levels of inequality, whileusing a posttax and transfer measure of income to calculate the income




TABLE 2 Commonly Used Measures of Income Inequality

Example

United UnitedStates, Kingdom,

Measure Definition Advantage Disadvantage Data Source 1994 1995

Gini coeffi-cient

Ranges from 0 (perfectequality) to 1 (perfectinequality); ratio ofarea between Lorenzcurve and a line ofperfect incomeequality

Most commonlyused; simpleinterpretation

Comparabilityproblems; notalways constructedidentically; lack ofgood data; notavailable for manycountries/years

LIS; World Bank(Deininger andSquire 1996);WIID; U.S.census

0.355 0.344

Incomeshares

Ratio of income ofperson at the xthpercentile (often the90th) to a person atthe yth percentile(often the 10th)

Easily interpreted;can examine arange of extremedistributions

Lack of good data;not available formany countries/years

LIS; World Bank(Deininger andSquire 1996);WIID; U.S.census

90:10,5.85;80:20,3.11

90:10,4.87;80:20,2.84

AtkinsonIndex

Ranges from 0 to 1;relative position ofthe poorest is weightedby parameter e thatmeasures society’saversion to inequality

Weighs the relativeposition of thepoorest; allowsfor a range of eweights

Not commonly used;no consensus onbest value for e; lackof good data; notavailable for manycountries/years

LIS; World Bank(Deininger andSquire 1996);U.S. census

e = 1,0.214;

e = 0.5,0.105

e = 1,0.204;

e = 0.5,0.100

423

(continued)

at WE

LCH

ME

DIC

AL LIB

RA

RY

on Septem

ber 7, 2010m

cr.sagepub.comD

ownloaded from


Theil Index Ranges from 0 (completeequality) to infinity; amember of the entropyclass of inequality mea-sures

Weighs relative po-sition of poorest;reliable data setsavailable

Not commonly used;not easily inter-preted; based onwages, not income

University of TexasInequality Project(UTIP)

.0280 .0189

Robin HoodIndex

Ranges from 0 to 1; per-centage of incomeneeded to transfer fromthe richest 50 percent tothe poorest 50 percentto obtain equality

Intuitively appeal-ing; less sensitiveto highly skeweddistributions

Ignores the distribu-tion of incomewithin each 50 per-cent share; not avail-able for many coun-tries/years

U.S. census NA NA

Note: LIS = Luxembourg Income Survey; WIID = World Income Inequality Database.

TABLE 2 (continued)

Example

United UnitedStates, Kingdom,

Measure Definition Advantage Disadvantage Data Source 1994 1995

424

at WE

LCH

ME

DIC

AL LIB

RA

RY

on Septem

ber 7, 2010m

cr.sagepub.comD

ownloaded from


distribution in the same population could reveal a more egalitarian society.The World Income Inequality Database (UNU/WIDER 2000) suggests thatGini coefficients calculated from gross income are 5 to 10 points higher thanthose calculated from net income.

Another distinction in terms of household income data is whether suchdata have been “equivalized” for household size. Dividing household incomeby family size (often the square root of family size) is considered good practicebecause a family of four does not necessarily need twice the income of a familyof two to live equally well. Failure to make this adjustment may renderinterhousehold comparability significantly less reliable, because the unit ofanalysis (the household) is in reality composed of a number of different units(the family) that vary in size and composition. Atkinson, Rainwater, andSmeeding (1995), Judge and Paterson (2002), and Coulter (1989) provide moredetailed discussions of income inequality measurement issues.

Judge (1995) presented a critique of Wilkinson’s early work that addressessome of the problems inherent in many income inequality measures. Judgepointed out that Wilkinson and others used measures of income inequality(the Gini coefficient) that were derived from household surveys conductedindependently in each country. One of the key difficulties in comparinghousehold income studies is that they often differ in how they solicit totalincome from the households they interview. It is therefore possible thatreported differences in income distribution between countries could be biasedbecause each country did not measure income the same way (Galbraith,Conceição, and Kum 2000). Second, some income inequality databases (suchas the World Bank’s “Deininger and Squire” or the United Nations’ “WIDER”databases) include a mix of Gini coefficients calculated from household datathat differ not only in how income was measured but also in whether the finalGini coefficients are adjusted for taxes or other income transfers (Deiningerand Squire 1996). Although within such data sets one can select incomeinequality data that have more or less consistent definitions, this is not alwaysdone in practice because data are rarely available for every country and yearof interest. Third, not all income inequality measures were calculated by usingequivalized family income.

Judge (1995) used the LIS as a source of internationally comparable, stan-dardized measures of income inequality. The LIS income inequality measuresare based on a similar household income data collection methodology, arebased on net income, and are equivalized for family size. Using the LIS incomeinequality measures, Judge did not find a relationship between life expec-tancy and the distribution of income. He concluded that the previouslyreported relationship was in large part due to the weaknesses of the incomedata used to calculate the income inequality measures.




METHODOLOGICAL CRITIQUES

Judge, Mulligan, and Benzeval (1998a) provided a critique of the methodol-ogy used in much of the income inequality and health literature. Their objec-tions included the observations that (1) authors generally used small samplesizes so their conclusions are prone to error, (2) some studies used data ondeveloping and developed countries but then generalize the results only tothe developed countries, (3) many studies used bivariate analyses only andthus fail to control for other health determinants, and (4) income inequalitydata were generally unreliable, are not available for a wide range of years, andwere not consistently adjusted for taxation and family size.

Judge, Mulligan, and Benzeval’s (1998a) critique, although pertinent tomuch of Wilkinson’s work, is not a generalizable critique; multilevel and time-series studies in the United States (see Blakely et al. 2000; Kennedy et al. 1998;Lynch et al. 1998; Wolfson et al. 1999), studies using a variety of outcome vari-ables (Soobader and LeClere 1999), studies conducted in Canada (Cairney andWade 1998) and Australia (Clarke and Smith, 2000), and studies using equiva-lent disposable income to construct inequality data (Humphries and vanDoorslaer 2000; Shi and Starfield 2000) provide much stronger evidence forthe relative income hypothesis and address each of Judge, Mulligan, andBenzeval’s objections.

However, more sophisticated research designs have also resulted in nega-tive findings. For example, Fiscella and Franks (1997) used a U.S.-based longi-tudinal cohort study (National Health Examination Survey and its follow-upcomponents) and a multilevel model to test the relationships between individ-ual income, neighborhood-level income inequality, and an individual’s risk ofdying. Their results showed that although there seems to be a bivariate rela-tionship between survival rates and community income inequality, afteradjustment for household income, no significant relationship was foundbetween neighborhood income inequality and mortality. They suggested thatprevious ecological-level studies using larger geographical units may have suf-fered from the confounding influence of the interrelations between contextual-level income inequality measures and individual-level family income. Theyconcluded that poverty, rather than income inequality, determines mortalityat the community level.

Wagstaff and van Doorslaer (2000, 543) conducted a literature review ofindividual-level studies on the impact of income inequality on health. In theirreview of six major studies, they found that the literature “reveals strong sup-port for the absolute income hypothesis and little or no support for the incomeinequality hypothesis.” The authors concluded that many of the individual-level studies conducted in the United States suffer from methodological flaws




in the income inequality measures used, do not account for fixed state andyear effects, and are not generally consistent with several hypothetical path-ways through which income inequality is posited to influence individualhealth.

Several individual-level international studies further called into questionthe universality of the relative income hypothesis. Research conducted inJapan (Shibuya, Hashimoto, and Yano 2002) and Denmark (Osler et al. 2002)showed that individual-level health (measured by self-rated health status andmortality risk, respectively) was not associated with income inequality mea-sured at the level of prefectures (states) in Japan, or smaller areas of aggrega-tion (parishes) in Denmark. Both studies suggested that individual income is amore important predictor of health outcomes at the individual level thanincome inequality. The authors of both studies also suggested that features ofthe Japanese and Danish welfare state may protect people from the adverseeffects of income inequality in ways that social protections in countries such asthe United States, Taiwan, and the United Kingdom do not. Finally, anotherstudy reported that income inequality is not significantly related to mortalityin Canada at either the regional or metropolitan level, although there was arelationship between income inequality and health in the United States (Rosset al. 2000).

Sturm and Gresenz (2001) present a further methodological critique of theliterature. The authors argue that associations between self-rated health andincome inequality became nonsignificant when adjusted for other individual-level covariates (income, age, gender, race/ethnicity, family size). Further-more, they found a strong gradient in the relationship between better healthand improved levels of education and absolute (not relative) family income.

Perhaps the most comprehensive study to date is that by Mellor and Milyo(2001). Their study examined the country-level relationship between incomeinequality (measured by the Gini coefficient and income ratios of the bottom20th percentile to the top 20th percentile) and aggregate health outcomes (lifeexpectancy at birth, infant mortality) across 30 countries over a four-decadespan. It also examined 48 U.S. states over five decades using the Gini coeffi-cient to measure income inequality and all-cause mortality, infant mortalityrates, low-weight births, homicides, suicides, and six different specific causesof death as dependent variables. At the international and state levels, theauthors found that contrary to previous literature, there was no consistentrelationship between income inequality and health outcomes. The analysiscontrolled for demographic variables such as median income, educationallevels, and year-specific effects. The state-level analyses also controlled forpercentage of population that is urban, black, and college educated. In the 54regression equations reported, income inequality was significantly associated




with poorer aggregate health outcomes in only 11 cases but was significantlyassociated with better health outcomes in 15 cases.

ALTERNATIVE HYPOTHESES

Gravelle (1999) presented an alternative explanation for the observed rela-tionship between income inequality and health outcomes. He claimed that therelationship between individual income and mortality is curvilinear—beyond a certain point, increases in income have little effect on mortality.Because of this relationship, he argued that a statistical artifact might explainthe observed relationship between income inequality and health. His logicwas that if one population has a more equal distribution of income thananother, then there are more individuals at the middle income levels andfewer outliers at the high and low ends. This means, in general, better healthstatus for those at the lowest income groups (especially since more people atthe higher end of the income distribution will not be expected to influencehealth outcomes because of the curvilinear relationship between income andhealth). The average poor person can thus appear to be better off in a moreegalitarian society because he or she is likely to have a higher individualincome, which will reduce the risk of mortality. Therefore, the relative incomehypothesis is simply an artifact of absolute income.

Gravelle, Wildman, and Sutton (2001) then reexamined Rodgers’s (1979)study using more consistent income inequality data derived from the LIS. Inmultivariate regression analyses using data from 56 countries at two timeperiods (1980-1982 and 1988-1990) and controlling for GDP and several trans-formations of average national income, the relationship between incomeinequality and life expectancy was not significant. Box-Cox andnonparametric estimations also revealed no significant association betweenlife expectancy and income inequality. They concluded that the relationshipbetween income inequality and health was indeed a statistical artifact.

Other alternate explanations for the observed relationship between incomeinequality and health include the hypothesis that racial and ethnic minoritypopulations within countries (which often have a poorer health profile andrepresent a larger share of the lower income quintiles) may explain the highermortality among lower income deciles. Wilkinson (1992) showed that minor-ity groups have a small impact on health in England (one of the more ethni-cally diverse countries in Europe), implying that the scale of income inequal-ity was too large to be accounted for by ethnic minorities alone (Wilkinson1992). However, studies in the United States showed a different pattern.

Deaton and Lubotsky (2002) examined the impact of race on the relation-ship between income inequality and mortality among U.S. states and MSAs in




1980 and 1990. They found that “once the fraction [of the state or MSA that is]black is controlled for, income inequality has no effect” (p. 11). Their findingsare significant even after stratifying by age groups and gender and after con-trolling for income, education, and state fixed effects. The authors argue thatthe association between income inequality and health in the United Statesmight be the result of confounding with race, since those areas with a higherproportion of black population also have higher income inequality and highermortality—for whites as well as for the entire population. They proposed thatthe social cohesion argument made to explain the impact of income inequalityon health might be applied to segregation based on race. Muller (2002) foundsimilar results. While Shi and Starfield (2001) found that in an ecological studyof U.S. MSAs in 1990, income inequality was a significant predictor of mortal-ity for both black and white populations; even while conditioning on percapita income, education, unemployment, urban residence, and primary carein both combined and race-stratified analyses.

Other authors suggested that the health selection hypothesis might be atwork. This hypothesis holds that ill health precedes poor educational or workperformance. That is, sicker people are more likely to be poorer because of theeconomic consequences of their illness (West 1991).

Finally, Judge and Paterson (2002) discussed the possibility that incomeinequality might also have a negative impact on the health of those at the high-est income levels. This is supported by Weich, Lewis, and Jenkins (2001), whofound an association between regional income inequality and self-reportedmental health in Britain. This relationship was statistically significant onlyamong individuals in the wealthiest income quintile, and held when control-ling for a range of individual-level health determinants, including social class,employment status, age, gender, income, and physical health status.

COMPARISON OF EMPIRICAL STUDIES

The appendix presents an analysis of the main peer-reviewed studies onthe relationship between income inequality and health. Forty-five empiricalstudies and two literature reviews are included in the table. Criteria for com-parison include the type of income inequality measure; main outcome vari-able(s); whether the study controlled for at least the following covariates—GDP per capita or mean/median income for ecological studies and income,age, and gender for individual-level studies; unit of analysis (ecological, indi-vidual, mixed); study population and time period; which pathway (if any)was tested; and whether income inequality was found to be associated withhealth outcomes.




Table 3 presents a summary of the findings and main characteristics of thestudies reviewed. The table presents the study results according to three over-all classifications—by health outcome, income inequality measure, and levelof analysis. This table is not meant to present results of a formal meta-analyticmethod. Instead, it is intended to facilitate analysis of the results of the manystudies contained in the appendix. It is also important to note that publicationbias is likely to have limited the number of studies published that showed norelationship between income inequality and health. Therefore, we cautionagainst a simple comparison of number of studies that found or failed to findan association between income inequality and health, particularly whenexamination of the authorship of the articles points to a group of less than 10researchers being responsible for the vast majority of the studies reviewed.Instead, we constructed the table to see if there might be patterns in the studydesign, health outcomes tested, or country analyzed that were more likely tofind positive or negative results. The results of this analysis follow.

Several conclusions can be drawn from an examination of Table 3. First,there are contradictory findings based on the health outcomes measure used.Overall, 33 analyses showed a statistically significant relationship betweenincome inequality and health outcomes, while 12 studies generally showed norelationship. For each health outcome (life expectancy, infant mortality, all-cause adult mortality, and self-rated health), there is evidence both for andagainst an effect of income inequality.

Studies also differed by the measure of income inequality used. The mostprevalent measure of income inequality was the Gini coefficient, followed bydifferent combinations of income shares. For each type of income inequalitymeasure, there is evidence both for and against its association with health out-comes. There was little consistency even among studies using the same mea-sure of income inequality. For example, among those studies using the Ginicoefficient, some calculated it by using household as opposed to familyincome, while others used gross (pretax) instead of net (posttax) income. Byand large, international studies using the LIS Gini measures (adjusted fortaxes and household size) did not find an association between incomeinequality and health, while studies using other Gini measures did find anassociation. But it is important to note that not all of these studies used exactlythe same countries, the same years, the same control variables, or even thesame outcomes measures, rendering comparison among even seeminglysimilar studies inconclusive.

In terms of level of analysis, evidence is also mixed. Ecological studies werethe most prevalent, and the results, often conducted on the same unit of analy-sis (U.S. states or OECD countries), varied. Individual-level and multilevelstudies all revealed positive and negative findings. This was the case even




when the same outcomes, data sets, and units of analysis were used. For exam-ple, Lochner et al. (2001) found a significant relationship between incomeinequality and mortality using a multilevel design with the U.S. NationalHealth Interview Survey, while Fiscella and Franks (1997), using a similardesign and the same data set, did not find a relationship.

Some authors have suggested that there is more evidence of a relationshipbetween income inequality and health within the United States than within(or between) other countries (Judge and Paterson 2002). This review only par-tially supports such a conclusion. Some individual-level within-country


TABLE 3 Summary of Literature Review Findings

Association between IncomeInequality and Health

Study Characteristic Significant Not Significant

OutcomesLife expectancy 3 4Infant mortality rate (IMR) 5 0 (2)a

Mortality 9 5Self-rated health 9 3Other 7 0Total 33 12

Income inequality measureGini 15 7Robin Hood 3 0Income decile ratios 10 4Other 5 1Total 33 12

Level of analysisIndividual 10 5Ecologic 17 6Multilevel 6 1Total 33 12

Countries studiedUnited States only 17 3Non–United States (single country) 5 3International (multicountry) 11 6Total 33 12

a. Some studies examined multiple outcomes, including IMR. Although only the study’s mainfindings are presented in the table, at least two studies additionally present evidence of no statisti-cal association between income inequality and IMR.



studies conducted at the individual or state level provide limited evidencethat income inequality may affect health in Australia, New Zealand, Taiwan,and the United States, and there is evidence both for and against an associa-tion within Canada. There are also mixed findings regarding multicountrystudies. Even though recent literature (see Deaton 2002; Mellor and Milyo2001) has called into question the validity of conclusions drawn fromWilkinson’s (1992) and Rodgers’s (1979) international (multicountry) ecologi-cal studies, several more sophisticated analyses (see Lynch et al. 2001) did findan association across countries between income inequality and some, but notall, categories of health outcomes.

DISCUSSION

The literature review has described the wide range of studies exploring therelationship between income inequality and health. We have attempted toprovide structure to the literature by describing its historical development,outlining its major theoretical and empirical basis, and categorizing the manycriticisms, alternative explanations, and negative findings presented in theliterature.

This review suggests that there are inconsistent findings on the relationshipbetween income inequality and health. The most frequently reported associa-tions seem to be in studies based in the United States (primarily using self-rated health and measuring income inequality at the state level) and interna-tional studies that use infant mortality as an outcome. But even within thesestudies there is variation. For example, in studies conducted within the UnitedStates, many that find a relationship between income inequality and healthcreate categories of income inequality and then rank the unit of analysis (usu-ally U.S. states) by categories of high, medium, or low income inequality,rather than using the actual value of the income inequality measure (such asthe Gini coefficient) in the analysis. This is in contrast to many of the studiesthat do not find a relationship: they usually use the value of the incomeinequality measure itself in multivariate analyses.

In light of these findings, the main conclusion from this review is that thelack of consistency in the study design, data used, measures constructed, con-ceptual models employed, and unit of analysis makes it difficult to draw adefinitive conclusion about the relationship between income inequality andhealth. For nearly each empirical study reviewed, another mostly similarstudy produces a negative finding. Then, a new publication appears to chal-lenge the negative finding and reassert the original study’s conclusions. Nor-mally, such disputes could be resolved by analyzing of the strength of differ-ent study designs, but even prospective, multilevel designs controlling for a




variety of relevant covariates revealed contradictory results often becausethey used similar (but not identical) income inequality measures. The lack ofsystematic replication of results has certainly increased the number of articlespublished on this topic, but it has done little to address the central question ofwhether (and if it is, to what extent and how) income inequality is related tohealth outcomes.

To answer these questions, we make several suggestions that we believewould facilitate future studies and ultimately help shape future policies andprograms designed to address the impact of income and other social inequali-ties on health.

First, although the majority of studies control for covariates thought to beassociated with health, the actual model of health production specified is dif-ferent in nearly every study. Our conceptual framework shows that althoughmultiple pathways have been mentioned in the literature, few studies testcompeting pathways. Almost none have placed analyses within a largersocioeconomic framework that considers the causes of income inequality aswell as its potential impacts on health and other population outcomes. As partof the conceptual framework of health determinants, future studies shouldinclude, at the very least, a macro-level measure such as GDP per capita andhealth systems and services, as well as micro (or aggregate) measures of indi-vidual income, socioeconomic status (SES) measures of education or occupa-tional prestige (or social class), and racial/ethnic composition.

Second, the link between the theoretical literature and the empirical modelused to test hypotheses is often weak. At the very least, the range of covariatesand the specific theoretical framework for each study should be better justi-fied in terms of the health outcome being evaluated. In addition, issues ofproximal determinants of income inequality and health outcomes are rarely(if ever) addressed. Here, the economics and sociology literatures presentnumerous studies of how (and why) income and other social disparitiesdevelop within nations (see Alderson and Nielson 1999; Chase-Dunn 1975;Gradstein and Milanovic 2000; Gustafsson and Johansson 1999; Lipset 1959).Few studies in the literature reviewed have attempted to analyze the role ofmacropolitical effects through techniques such as instrumental variables orstructural equation modeling. Doing so could enrich the health literatureenormously by building on the rich theoretical and empirical traditions infields that have long studied the distribution and consequences of inequalitieson populations.

Third, there is a need to improve the framework and analyses by sorting outthe various levels of health determinants, that is, ecological and individual,and the need to assess at which ecological levels (nation, state, MSA, county,city, neighborhood, etc.) that income inequality makes most sense in




influencing health. We have attempted to provide a basic framework forordering these analyses in our conceptual framework. The literature reviewedhere suggests that different results can be obtained by examining the samepopulations at different levels of analysis. Further work will be necessary tosatisfactorily explain such discrepancies and justify the level of analysis usedin future studies.

Fourth, there are great differences in the measure of income inequality, thedata used to calculate income inequality measures, and the years for whichdata are derived. These inconsistencies greatly impede study comparability.At the very least, income inequality measures should contain some weight oridentify the position of the poorest in the income distribution. This is whyincome share ratios, concentration curves, and Atkinson and Theil indicesshould be preferred to the Gini or the Robin Hood Index. Moreover, house-hold inequality data must be consistent. The economics literature suggeststhat household income used for calculating income inequality measuresshould be net of taxes and other transfers and should be equivalized for familysize. Future studies will need to be explicit about the income data used toderive the inequality index and discuss the implications of the limitations ofthe income data they employ.

Fifth, very few of the studies reviewed included health system compo-nents. Those studies that do incorporate health system variables, particularlymeasures of primary care, found that they are important determinants ofhealth outcomes and in some cases reduce the impact of income inequalityand other socioeconomic measures on health outcomes. The finding thataspects of the health system can interact in some way with social or economicdeterminants of health offers what Starfield and Shi (2002) term a “palliativestrategy” to partially reduce the negative effects in inequalities on health inthe absence of more comprehensive efforts to tackle the root causes of suchinequalities. Greater efforts should be taken to ensure that the health system isincluded not only in further research but also in the design of policies and pro-grams meant to address social disparities.

Sixth, because of the complexity of the analyses required, methodologicalproblems continue to plague most published studies. Time lags (Blakely et al.2000), data quality (Judge 1995; Judge, Mulligan, and Benzeval 1998a; Wag-staff and van Doorslaer 2000), controlling for median versus mean income(Blakely and Kawachi 2001)—just to name a few issues—have all been foundto influence results. Moreover, issues such as multicolinearity among eco-nomic variables are rarely dealt with systematically. To date, no single studyhas comprehensively addressed each of these potential limitations. For this




reason, it is difficult to compare different studies since each contains its own,often unique, flaws. Although it will be a challenging task, if we are to satisfac-torily investigate the impact (if any) of income inequalities on health,resources should be directed at designing a thorough, rigorous, and replicablestudy that addresses each of these methodological issues.

CONCLUSIONS

Some authors have recently asserted that “evidence favoring a negativecorrelation between income inequality and life expectancy has disappeared”(Mackenbach 2002, 1). Based on the analysis of the literature reviewed here,this conclusion appears premature. It is difficult to draw a firm conclusionabout the true nature of the relationship between income inequality andhealth because a combination of limited data of questionable quality,noncomparable study designs, radically different conceptual frameworks,and different analytical methodologies has complicated rather than clarifiedour understanding of this relationship.

Moreover, none of the studies or editorials discussing the relationshipbetween income inequality and health has claimed that income inequalityitself has a direct influence on health. Instead, income inequality is viewed as aproxy for something else—some other psychological, social, economic, politi-cal, or environmental determinant of health. Based on the literature reviewedhere, it appears that income inequality performs rather poorly as a proximatecause of health.

Finally, regardless of the relationship between income inequality andhealth detected, nearly every study has confirmed the importance of individ-ual income on health outcomes—even within countries with universal insur-ance and relatively generous social welfare policies. This suggests that onebenefit of research on income inequality is that it has highlighted the role ofeconomic and social resources and their impact on health inequalities(Mackenbach 2002). We also believe another important contribution has beena reexamination of the role of the health system, particularly primary care, asan important determinant of population health in advanced industrializednations, an area that deserves further thought, analysis, and action.




APPENDIXLiterature Review

Income Main Hypothesis/ IncomeInequality Outcome Controlled Unit of Study Design, Pathway Inequality Related

Author Measure Variable for Covariates Analysis Population Explored to Health?

Ben-Schlomo,White, andMarmot(1996)

TownsendIndex (IQrange)

All-causemortalityyoungerthan age 65

Yes (usingTownsendIndex)

Ecological(wards)

Cross-sectional;England only, 1981and 1985 data

Material Yes

Blakely et al.(2000)

Gini (state-level)

Self-ratedhealth

Yes (gender, age,race, medianhouseholdincome, stateincome)

Individuals Cross-sectional withtime lag; U.S. CurrentPopulation Survey(1979-1981, 1983-1985,1987-1989, 1991-1993,1995-1997)

Time lag Yes

Blakely,Lochner,andKawachi(2002)

Gini (high,medium, lowcategories)

Self-ratedhealth

Yes (individualandmetropolitanarea)

Multilevel Cross-sectional; U.S.Current PopulationSurvey data, 1996-1998; census data1990 for income

Individual healthand metropolitanarea (MA) incomeinequality

Yes (but not aftercontrolling forhouseholdincome and notat county level)

Bobak et al.(2000)

Gini (country-level)

Self-ratedhealth

Yes (materialdeprivation =food, clothing,heating; control,education,income)

Individuals Cross-sectional study;representativesamples of adults in1996/1998 in Russia,Estonia, Czech,Lithuania, Poland,Latvia, Hungary

Material andpsychosocial

No (not aftercontrolling formaterialdeprivationscore)

436

at WE

LCH

ME

DIC

AL LIB

RA

RY

on Septem

ber 7, 2010m

cr.sagepub.comD

ownloaded from


Chiang(1999)

Incomereceived bypoorest 50percent ofpopulation

All causemortality,younger thanage 5mortality

Yes (medianhouseholdincome, shareof household)

Individuals Ordinary least squares(OLS) regressionmodel comparingthree time periods:1976, 1985, 1995

Income inequality,mortality, eco-nomicdevelopment

Yes

Clarke andSmith(2000)

HealthConcentra-tion Index

Self-reportedHealthConcentra-tion Index

Yes Individuals Cross-sectional study;Australian Nationalhealth survey, 1990/1995

Magnitude ofhealth inequalitydue to income

Yes

Diez-Roux,Link, andNorthridge(2000)

Robin HoodIndex (bystate)

Smoking,Body MassIndex (BMI),hyperten-sion,sedentarism

Yes Multilevel U.S. state- andindividual-levelindicators fromBehavioral RiskFactor SurveillanceSystem and census,1990

Income inequalitiesand risk factors

Yes (BMI,hypertension,sedentarismstrongest effectamong poorest)

Duleep(1995)

Averageincomereceived bylowestincomedecile

Adult malemortality

Yes (percentage ofindustrializationand averageincome)

Ecological(countries)

Cross-sectionalecologic study on 38industrializedcountries; InternalRevenue Service,International LaborOrganization, andother data sources(mid-1970s comparedto late 1980s)

Mortality andincomeinequality

Yes

Fiscella andFranks(1997)

Lower 50percentincomeshare incommunity

Mortality Yes (householdincome,demographics,meancommunityincome)

Multilevel Longitudinal cohortstudy; sample ofadults included in theNational Health andNutrition Examina-tion Survey, 1971-1975 to 1987

Mortality relatedto individual orcommunityincome inequality

No

(continued)

437

at WE

LCH

ME

DIC

AL LIB

RA

RY

on Septem

ber 7, 2010m

cr.sagepub.comD

ownloaded from


Flegg (1982) Gini, Theil,coefficient ofvariation

Infantmortalityrate (IMR)

Yes (femaleilliteracy andlabor force,physician/nurseper 1,000,GDPpc, incometransfers)

Ecological(country)

Cross-sectional study;multivariate andbivariate OLSregressions for 46underdevelopedcountries, data from1970s

IMR and differentmeasures ofincomeinequality

Yes (with Gini)

Gravelle(1999)

Relative andabsoluteincome levelsby incomeshares

Mortality No Theory Reexamination of therelationship betweenincome and mortal-ity; no data used.

Is the relationshipa statisticalartifact

No

Gravelle,Wildman,and Sutton(2001)

Gini Lifeexpectancy(gender-specific)

Yes (GDP,income, andpermutations)


Cross-sectional 75countries, 1981 and1989

Reanalyze Rodgers(1979) data

No

Hales et al.(2000)

Gini—notequivalized(fromDeiningerand Squire1996)

IMR Yes (GNP percapita)

Ecological(country)

Cross-sectional; 38countries, 1970; 26countries, 1990

Infant mortality,GNP, and incomeinequalities

Yes (morepronounced inindustrializedcountries)

Humphriesand vanDoorslaer(2000)

Health Con-centrationIndex

Self-assessedhealth status

Yes (income,Health UtilityIndex)

Individual Cross-sectional; 1994Canadian NationalPopulation HealthSurvey

Absolute versusrelative income

Yes (but does notexplain allincome-relatedhealthinequalities)

APPENDIX (continued)



438

at WE

LCH

ME

DIC

AL LIB

RA

RY

on Septem

ber 7, 2010m

cr.sagepub.comD

ownloaded from


Judge (1995) Gini, percent-age of popu-lation withless than 50percentincome

Lifeexpectancy

No Ecological(countries)

Cross-sectional study;13 Organization forEconomic Coopera-tion and Develop-ment (OECD) coun-tries; LuxembourgIncome Survey (LIS)data, 1994

Is the relativeincome hypothe-sis valid?

No

Judge, Mulli-gan, andBenzeval(1998a)

Share ofincome goingto differentincome dec-iles, Gini,90:10 ratio

Life expec-tancy, infantmortality

Yes (social secu-rity, geographicregion, percent-age femaleworkforce,healthexpenditures)

Ecologic(country)

Literature review andcross-sectional study;OLS regression; 16OECD countries; LISincome inequalitydata

Further critique ofthe relativeincomehypothesis

No

Kahn et al.(1998)

Percentagetotal incomereceived byrichest 50percent ineach state

Abdominalobesity

Yes Individual Cohort from 21 states;follow-up of Ameri-can Cancer NutritionSurvey

Psychosocial stress Yes (men only)

Kahn et al.(2000)

Gini (statesranked in cat-egories basedon Gini)

Depressivesymptomsand self-rated health

Yes (householdincome, age,marital status,education, race/ethnicity, house-hold size)

Individual Cross-sectional;national sample ofwomen who gavebirth in 1998(National Maternaland Infant HealthSurvey)

Income inequalityand women’shealth

Yes (most pro-nounced amonglowest incomegroups)

(continued)

439

at WE

LCH

ME

DIC

AL LIB

RA

RY

on Septem

ber 7, 2010m

cr.sagepub.comD

ownloaded from


Kaplan et al.(1996)

Incomereceived bylowest 50percent instate

Age-adjustedall-causemortality;low birthweight

No (bivariate rela-tionships only)

Ecological(states)

Cross-sectional study;U.S. states, 1980, 1990,1989-1991; data fromthe National Centerfor Health Statistics(NCHS) and 1980 and1990 census

Income inequalityand mortality

Yes (also withhomicide, crime,insurance,imprisonment)

Kawachi etal. (1997)

Robin HoodIndex

Social capitalvariables

No (only con-trolled for statepoverty)

Ecological(states)

U.S. cross-sectionalecological analysis; 39U.S. states

Components ofsocial capital

Yes

Kawachi,Kennedy,and Glass(1999)

State-levelsocial capital(categories oftrust, groupmembershipreciprocity)

Self-ratedhealth

Yes (income edu-cation, smoking,age, gender)

Multilevel Multilevel analysis;Behavioral Risk Fac-tor Surveillance Sys-tem (U.S. adults)

Social capital Yes

Kennedy etal. (1998)

Gini for statesclassified byfour levels ofinequality

Self-ratedhealth status

Yes (risk factors,demographic,and socioeco-nomic status[SES] factors)

Multilevel Cross-sectional; U.S.states in 1993 and1994, data frombehavioral risk factorsurvey

Is there an effect ofincome inequalityin individual-level health?

Yes

Kennedy,Kawachi,andProthrow-Stith (1996)

Robin HoodIndex, Ginicoefficient

Mortality,IMR,coronarydisease,cancer,homicide

No (poverty andsmoking only)

Ecological(states)

Cross-sectional analy-sis; U.S. states, 1990census data

Income inequalityand mortality

Yes




440

at WE

LCH

ME

DIC

AL LIB

RA

RY

on Septem

ber 7, 2010m

cr.sagepub.comD

ownloaded from


LeGrand(1987)

Share ofincome forbottom 20percent

Age at death Yes (GDP, welfare,health care)


Cross-sectional analy-sis of 17 developedcountries; 1980s data

Income inequalityand health

Yes

Lochner et al.(2001)

Gini (state-level in fivecategories)

Individual riskof mortality

Yes (age, income,race, gender,marital status)

Multilevel Prospective design;U.S. National HealthInterview Surveylinked to NationalDeath Index, 1991

State-level incomeinequality andrisk of death

Yes (most pro-nounced fornear-poorwhites)

Lynch et al.(1998)

Gini, Atkinson,Theil, below50, 50:10 and90:10 incomeshares

All-cause mor-tality by agegroup

Yes (per capitaincome, house-hold size, 200percent povertyrate)

Ecological(metropoli-tan statisticalarea [MSA])

Cross-sectional design;U.S. population using1990 census data

Income inequalityand mortality inMSAs in theUnited States

Yes (for infantmortality andaged 15-64group)

Lynch et al.(2001)

Gini coeffi-cient, LISdata

Mortality cate-gories, IMR,life expec-tancy, dis-trust, organi-zationalmembership,control,union,women ingovernment

Yes (GDP andpopulation size)


Cross-sectional; corre-lations only; OECDcountries; WorldHealth Organization(WHO) and worldvalues survey data,from mid-1990s

Psychosocial envi-ronment, health,incomeinequalities

Yes (for IMRonly), no(psychosocialvariables showmixed results)

McIsaac andWilkinson(1997)

Decile sharesof income(LIS data)

Mortality(multiple cat-egories),IMR, poten-tial years oflife lost(PYLL)


Cross-sectional; corre-lations only; datafrom 12 wealthyOECD countries

Income inequalityand differenthealth outcomes

Yes

(continued)

441

at WE

LCH

ME

DIC

AL LIB

RA

RY

on Septem

ber 7, 2010m

cr.sagepub.comD

ownloaded from


Mellor andMilyo(2001)

Gini coeffi-cient, incomeratios

Life expec-tancy, all-cause mortal-ity, IMR,low-weightbirths, homi-cide, suicide

Yes (income, edu-cation, year,urban, black)

Ecological(countriesand U.S.states)

Cross-sectional for dif-ferent time periods;30 countries, 48 U.S.states, 1960s-1990s


No (relationshipnot consistent,income inequal-ity associatedwith both betterand pooreroutcomes)

Muller (1985) Income shareof upperquartile

Homicidefrom politicalviolence

Yes (economicgrowth, regime,oppression,years ofdemocracy)

Ecological(country)

Cross-sectional study;OLS regression andpath analysis, 57countries

Mortality frompolitical violence

Yes

Osler et al.(2002)

Medianincome shareby parish

Mortality risk Yes (householdincome, house-hold and demo-graphiccharacteristics)

Individual Pooled, representativecohort studies (morethan 25,000 peoplefollowed for 13 years)in Denmark

Income inequality,mortality, andindividualincome

No

Rodgers(1979)

Gini Life expec-tancy at birthand age 5and IMR

Yes (GNP percapita and logtransformationsof GNP)


Cross-sectional study,56 countries, 1973

Income inequalityas a determinantof mortality

Yes

Ross et al.(2000)

50 percentincome share

Mortality Yes Ecological(states andMSAs)

Cross-sectional study,OLS regressionmodel; Canadianprovinces and MSAs,U.S. states and MSAs;Census data 1990-1991


Yes for UnitedStates, no forCanada




442

at WE

LCH

ME

DIC

AL LIB

RA

RY

on Septem

ber 7, 2010m

cr.sagepub.comD

ownloaded from


Shi et al.(1999)

Gini Mortality, postand neonatalmortality, lifeexpectancy

Yes Ecological(states)

Cross-sectional study;U.S. census and oth-ers, 1990

Primary care Yes

Shi andStarfield(2000)

Gini (state) Self-ratedhealth

Yes (primary care,income, demo-graphics,employment,wages, poverty,insurance,smoking

Multilevel Cross-sectional study;U.S. CommunityTracking Survey(state, MSA/commu-nity, individual), 1996

Primary care Yes (incomeinequalityimpact isreduced in pres-ence of strongprimary care)

Shibuya,Hasimoto,and Yano(2002)

Gini Self-ratedhealth

Yes (individualincome anddemographiccharacteristics)

Individual Cross-sectional analy-sis of more than80,000 Japaneseadults in 1995

Income inequalityand self-ratedhealth

No

Soobader andLeClere(1999)

Gini Self-ratedhealth

Yes (income, pov-erty, income-to-needs ratio, edu-cation,occupation)

Individual Cross-sectional ecologi-cal study; white adultmales in the UnitedStates, NationalHealth Interview Sur-vey, 1989-1991

Income inequalityand morbidity

Yes (at countylevel, butreduced at cen-sus tract level)

Sturm andGresenz(2001)

Gini Self-ratedhealth,chronic med-ical condi-tions, andmentalhealth

Yes (familyincome, age,gender, race/ethnicity, familysize)

Individual Cross-sectional ecologi-cal study; population-based survey datafrom nationally rep-resentative commu-nity tracking study(United States only)

Income inequality,family income,and mental health

No (not after con-trolling for edu-cation and fam-ily income—strong predic-tors of health inthis study)

Turrell andMathers(2001)

Gini Propensity tonot reportincome

Yes (income level,occupationalclass, gender,education,income source)

Individual 1995 Australian HealthSurvey; analysis ofdata from survey ofAustralian adults,1995

Is incomeunderreportingsignificant?

NA (incomeunderreportingmost prevalentamong highincome groups)

(continued)

443

at WE

LCH

ME

DIC

AL LIB

RA

RY

on Septem

ber 7, 2010m

cr.sagepub.comD

ownloaded from


vanDoorslaeret al. (1997)

Health concen-tration curve

Self-reportedhealth

Yes (health andsocialexpenditures)

Individual Cross-sectional; Fin-land, Germany, Neth-erlands, Spain, Eng-land, United States,Sweden, Switzerland;data from 1982-1992

Income inequalityand health inOECD countries

Yes

Wagstaff andvanDoorslaer(2000)

NA NA NA Individual Literature review of sixindividual-levelstudies


No consistentrelationship

Waldmann(1992)

Log incomeshare of top 5percent andbottom 20percent

IMR (log-transformed)

Yes (GDP, doc-tors/1,000,urban, primaryschool, femaleliteracy, repro-ductive rate)


Cross-sectional study;57 countries; datafrom World Tables,UN statistics, WorldBank 1960/1970

Income distribu-tion and IMR

Yes

Weich, Lewis,and Jenkins(2001)

Gini (net cur-rent income)

Self-reportedmentalhealth

Yes (age, gender,social class,employment,ethnicity, physi-cal health)

Individual Cross-sectional analy-sis of adults in Eng-land, Wales, and Scot-land; data from early1990s

Income inequalityand mental health

Yes (but onlyamong thewealthiestincome quintile)

Wennemo(1993)

Gini IMR Yes (poverty, fam-ily benefits,unemploymentinsurance,GDPpc,unemployment)


Pooled cross-sectionalanalysis of 18 OECDcountries, 1990

IMR and inequality Yes




444

at WE

LCH

ME

DIC

AL LIB

RA

RY

on Septem

ber 7, 2010m

cr.sagepub.comD

ownloaded from


Wilkinson(1994)

Gini (posttax,standardizedfor house-hold size)

Lifeexpectancy

Yes (GDP percapita)


Cross-sectional study;some time-seriesanalyses; 12-22 OECDcountries, 1970-1990

Developed versusdevelopingcountries

Yes

Wilkinson(1992)

Income decileshares (fam-ily net cashincome)

Lifeexpectancy


Cross-sectional study;bivariate correlations;23 OECD countries;data from WorldTables, LIS, WHO,OECD, 1975-1985

Income distribu-tion and lifeexpectancy indevelopedcountries

Yes

Wolfson et al.(1999)

Income shares All-causemortality

Yes (state incomelevels)

Individual Multivariate statisticalsimulation model;data from 1990 U.S.census, CDC, andnational longitudinalmortality study

Relation betweenmortality andincome inequalitya statisticalartifact?

Yes

445

at WE

LCH

ME

DIC

AL LIB

RA

RY

on Septem

ber 7, 2010m

cr.sagepub.comD

ownloaded from


REFERENCES

Alderson, A., and F. Nielson. 1999. Income, inequality, development and dependence:A reconsideration. American Sociological Review 64:606-31.

Antoft, P. E., E. Gadegaard, and O. P. Lind. 1974. Social inequality and caries studies in1,719 Danish military recruits. Community Dentistry & Oral Epidemiology 2:305-15.

Antonovsky, A. 1967. Social class, life expectancy and overall mortality. Milbank Memo-rial Fund Quarterly 45:731-35.

Atkinson, A., L. Rainwater, and T. Smeeding. 1995. Income distribution in advancedeconomies: Evidence from the Luxembourg Income Study. Working Paper 120,Maxwell School of Citizenship and Public Affairs, Syracuse University, New York.

Basch, P. 1990. Textbook of international health. Oxford, UK: Oxford University Press.Ben-Schlomo, Y., I. White, and M. Marmot. 1996. Does the variation in the socioeco-

nomic characteristics of an area affect mortality? British Medical Journal 312:1013-14.Berkman, L., and I. Kawachi. Eds. 2000. Social epidemiology. New York: Oxford Univer-

sity Press.Berkman, L., and S. Syme. 1979. Social networks, host resistance, and mortality: Anine-

year follow-up study of Alameda country residents. American Journal of Epidemiol-ogy 109:186-204.

Black, D., J. N. Morris, C. Smith, and P. Townsend. 1980. The Black report. In Inequalitiesis health, edited by P. Townsend and N. Davidson, 30-213. London: Penguin.

Blackburn, C. 1994. Low income, inequality and health promotion. Nursing Times90:42-43.

Blakely, T., and I. Kawachi. 2001. What is the difference between controlling for meanversus median income in analyses of income inequality? Journal of Epidemiology andCommunity Health 55:352-53.

Blakely, T., B. Kennedy, R. Glass, and I. Kawachi. 2000. What is the lag time between in-come inequality and health status? Journal of Epidemiology and Community Health54:318-19.

Blakely, T., K. Lochner, and I. Kawachi. 2002. Metropolitan area income inequality andself-rated health—A multi-level study. Social Science & Medicine 54:65-77.

Blaxter, M. 1987. Evidence on inequality in health from a national survey. Lancet 2:30-33.

Bobak, M., H. Pikhart, R. Rose, C. Hertzman, and M. Marmot. 2000. Socioeconomicfactors, material inequalities, and perceived control in self-rated health: Cross-sectional data from seven post-communist countries. Social Science & Medicine51:1343-50.

Bunker, J. 2001. The role of medical care in contributing to health improvements withinsocieties. International Journal of Epidemiology 30:1260-63.

Bunker, S. G. 1983. Dependency, inequality, and development policy: A case fromBugisu, Uganda. British Journal of Sociology 34:182-207.

Cairney, J., and T. J. Wade. 1998. Reducing economic disparity to achieve better health:Modeling the effect of adjustments to income adequacy on self-reported morbidityamong the elderly in Canada. Canadian Journal of Public Health 89:424-28.




Casanova, C., C. Colomer, and B. Starfield. 1996. Pediatric hospitalization due to ambu-latory care-sensitive conditions in Valencia (Spain). International Journal for Qualityin Health Care 8:51-59.

Casanova, C., and B. Starfield. 1995. Hospitalizations of children and access to primarycare: Across-national comparison. International Journal of Health Services 25:283-94.

Chandola, T. 1998. Social inequality in coronary heart disease: Acomparison of occupa-tional classifications. Social Science & Medicine 47:525-33.

Chase-Dunn, C. 1975. The effects of international economic dependence on develop-ment and inequality: Across-national study. American Sociological Review 40:720-38.

Chiang, T. 1999. Economic transition and changing relation between income inequalityand mortality in Taiwan: Regression analysis. British Medical Journal 319:1162-65.

Chiu, W., and P. Madden. 1998. Burglary and income inequality. Journal of Public Eco-nomics 69:123-41.

Christenson, B. A., and N. E. Johnson. 1995. Educational inequality in adult mortality:An assessment with death certificate data from Michigan. Demography 32:215-29.

Clarke, P., and L. Smith. 2000. More or less equal? Comparing Australian income-related inequality in self-reported health with other industrialised countries. Aus-tralia and New Zealand Journal of Public Health 24 (4): 370-73.

Coulter, P. 1989. Measuring inequality: A methodological handbook. Boulder, CO:Westview.

Deaton, A. S. 2002. Commentary: The convoluted story of international studies of in-equality and health. International Journal of Epidemiology 31:546-49.

Deaton, A. S., and D. Lubotsky. 2002. Mortality, inequality and race in American citiesand states. Social Science & Medicine 56 (6): 1139-53.

Deininger, K., and L. Squire. 1996. Measuring income inequality: Anew database. Washing-ton, DC: World Bank.

Diez-Roux, A. V., B. G. Link, and M. E. Northridge. 2000. A multilevel analysis of in-come inequality and cardiovascular disease risk factors. Social Science & Medicine50:673-87.

Dougherty, G., I. B. Pless, and R. Wilkins. 1990. Social class and the occurrence of trafficinjuries and deaths in urban children. Canadian Journal of Public Health 81:204-9.

Duleep, H. O. 1995. Mortality and income inequality among economically developedcountries. Social Security Bulletin 58:34-50.

Durkheim, E. 1951. Suicide. New York: Free Press.Fink, A. 1998. Conducting literature reviews: From paper to the Internet. Thousand Oaks,

CA: Sage.Fiscella, K., and P. Franks. 1997. Poverty or income inequality as predictor of mortality:

Longitudinal cohort study. British Medical Journal 314:1724-27.Flegg, A. T. 1982. Inequality of income, illiteracy, and medical care as determinants of

infant mortality in developing countries. Population Studies 36:441-58.Fuchs, V. 1974. Some economic aspects of mortality in developing countries. In The eco-

nomics of health and medical care, edited by M. Perlaman, 174-93. New York: JohnWiley.




Galbraith, J., P. Conceição, and H. Kum. 2000. Inequality and growth reconsidered:Some new evidence from old data. Working Paper 17, UTIP.

Gradstein, M., and B. Milanovic. 2000. Does liberte = egalite? Asurvey of the evidence on thelinks between political democracy and income inequality. Washington, DC: World Bank.

Grant, J. 1977. Poverty and health. In Papers of the Conference on the Interaction of Healthand Development. Washington, DC: National Council for International Health.

Gravelle, H. 1999. How much of the relation between population mortality and un-equal distribution of income is statistical artifact? In The society and population healthreader, edited by I. Kawachi, B. Kennedy, and R. G. Wilkinson, 99-104. New York:New Press.

Gravelle, H., J. Wildman, and M. Sutton. 2001. Income, income inequality and health:What can we learn from aggregate data? Social Science & Medicine 54:577-89.

Gregory, J. W., and V. Piche. 1983. Inequality and mortality: Demographic hypothesesregarding advanced and peripheral capitalism. International Journal of Health Ser-vices 13:89-106.

Gustafsson, B., and M. Johansson. 1999. In search of smoking guns: What makes in-come inequality vary over time in different countries? American Sociological Review64:585-605.

Hales, S., P. Howden-Chapman, et al. (2000). National infant mortality rates in relationto gross national product and distribution of income. Lancet 354:2047.

Hollingsworth, J. R. 1981. Inequality in levels of health in England and Wales, 1891-1971. Journal of Health and Social Behavior 22:268-83.

Humphries, K. H., and E. van Doorslaer. 2000. Income-related health inequality in Can-ada. Social Science & Medicine 50:663-71.

Judge, K. 1995. Income distribution and life expectancy: A critical appraisal. BritishMedical Journal 311:1282-87.

Judge, K., J. Mulligan, and M. Benzeval. 1998a. Income inequality and populationhealth [see comments]. Social Science & Medicine 46:567-79.

. 1998b. The relationship between income inequality and population health. So-cial Science & Medicine 47:983-85.

Judge, K., and I. Paterson. 2002. Poverty, income inequality and health. Auckland: NewZealand Treasury.

Kahn, H. S., L. M. Tatham, E. R. Pamuk, and C. W. Heath, Jr. 1998. Are geographic re-gions with high income inequality associated with risk of abdominal weight gain?Social Science & Medicine 47:1-6.

Kahn, H. S., P. Wise, B. P. Kennedy, and I. Kawachi. 2000. State income inequality,household income, and maternal mental and physical health: Cross-sectional na-tional survey. British Medical Journal 321:1311-15.

Kaplan, G., E. Pamuk, J. Lynch, R. Cohen, and J. Balfour. 1996. Inequality in income andmortality in the US: Analysis of mortality and potential pathways. British MedicalJournal 312:1004-7.

Kawachi, I., B. Kennedy, and E. Brainerd. 1998. The role of social capital in the Rus-sian mortality crisis. In The society and population health reader, edited by I. Kawachi,B. Kennedy, and R. G. Wilkinson, 261-77. New York: New Press.




Kawachi, I., B. Kennedy, and R. Glass. 1998. Social capital and self-rated health: A con-textual analysis. American Journal of Public Health 89:1187-93.

Kawachi, I., B. P. Kennedy, K. Lochner, and D. Prothrow-Stith. 1997. Social capital, in-come inequality, and mortality. American Journal of Public Health 87:1491-98.

Kawachi, I., B. Kennedy, and R. Wilkinson. 1999. Crime: Social disorganization and rel-ative deprivation. Social Science and Medicine 48:719-31.

Kennedy, B., I. Kawachi, R. Glass, and D. Prothrow-Stith. 1998. Income distribution, so-cioeconomic status, and self rated health in the United States: Multilevel analysis.British Medical Journal 317:917-21.

Kennedy, B., I. Kawachi, and D. Prothrow-Stith. 1996. Income distribution and mortal-ity: Cross sectional ecological study of the Robin Hood Index in the United States.British Medical Journal 312:1004-7.

LeGrand, J. 1987. Inequalities in health: Some international comparisons. European Eco-nomic Review 31:182-91.

Lipset, S. M. 1959. Some social requisites of democracy: Economic development andpolitical development. American Political Science Review 53:69-105.

Lochner, K., E. Pamuk, D. Makuc, B. Kennedy, and I. Kawachi. 2001. State-level incomeinequality and individual mortality risk: A prospective, multilevel study. AmericanJournal of Public Health 91:385-91.

Lynch, J., P. Due, C. Muntaner, and G. Davey Smith. 2000. Social capital: Is it a good in-vestment strategy for public health? Journal of Epidemiology and Community Health54:404-8.

Lynch, J. , G. Kaplan, E. Pamuk, R. Cohen, K. Heck, J. Balfour, and I. Yen. 1998. Incomeinequality and mortality in metropolitan areas of the United States. American Journalof Public Health 88:1074-80.

Lynch, J., G. D. Smith, M. Hillemeier, M. Shaw, T. Raghunathan, and G. Kaplan. 2001.Income inequality, the psychosocial environment, and health: Comparisons ofwealthy countries. Lancet 358:194-200.

Lynch, J., G. Smith, G. Kaplan, and J. House. 2000. Income inequality and mortality: Im-portance to health of individual income, psychosocial environment, or materialconditions. British Medical Journal 320:1200-4.

Macinko, J., and B. Starfield. 2001. The utility of social capital in studies on health deter-minants. Milbank Quarterly 79:387-428.

Mackenbach, J. 2002. Income inequality and population health. British Medical Journal324:1-2.

Mackenbach, J. P., A. E. Kunst, A. E. Cavelaars, F. Groenhof, and J. J. Geurts. 1997. Socio-economic inequalities in morbidity and mortality in western Europe: The EU Work-ing Group on Socioeconomic Inequalities in Health. Lancet 349:1655-59.

Mackenbach, J., K. Stronks, and A. E. Kunst. 1989. The contribution of medical care toinequalities in health. Social Science & Medicine 29:369-76.

Marmot, M., G. Smith, S. Stansfeld, C. Patel, F. North, J. Head, I. White, E. Brunner, andA. Feeney. 1991. Health inequalities among British civil servants: The Whitehall IIstudy. Lancet 337:1387-93.

Mathers, J. 1974. Letter: Inequality and the health service. Lancet 1:1342-43.




McIsaac, S., and R. Wilkinson. 1997. Income distribution and cause-specific mortality.European Journal of Public Health 7:45-53.

McKeown, T. 1976. The role of medicine: Dream, mirage or nemesis. London: Nuffield Pro-vincial Hospitals Trust.

McKeown, T., R. G. Record, and R. D. Turner. 1975. An interpretation of the decline inmortality in England and Wales during the twentieth century. Population Studies29:391-422.

McKinlay, J., and S. McKinlay. 1977. The questionable contribution of medical mea-sures to the decline of mortality in the United States in the twentieth century.Milbank Memorial Fund Quarterly Health and Society 55 (3): 405-28.

Mellor, J. M., and J. Milyo. 2001. Reexamining the evidence of an ecological associationbetween income inequality and health. Journal of Health Politics, Policy and Law26:487-522.

Muller, A. 1985. Income inequality, regime repressiveness, and political violence.American-Sociological-Review 50 (1): 47-61.

. 2002. Education, income inequality, and mortality: A multiple regression anal -ysis. British Medical Journal 324:1-4.

Muntaner, C., and J. Lynch. 1999. Income inequality, social cohesion, and class rela-tions: A critique of Wilkinson’s neo-Durkheimian research program. InternationalJournal of Health Services 29:59-81.

Omran, A. 1971. The epidemiological transition: Atheory of the epidemiology of popu-lation change. Milbank Memorial Fund Quarterly Health and Society 49:509-38.

Osler, M., E. Prescott, M. Gronbaek, U. Christensen, P. Due, and G. Engholm. 2002. In-come inequality, individual income, and mortality in Danish adults: Analysis ofpooled data from two cohort studies. British Medical Journal 324 (7328): 13-16.

Preston, S. H. 1976. The changing relation between mortality and the overall level ofeconomic development. Population Studies 29:231-48.

Robinson, J. C. 1984. Racial inequality and the probability of occupation-related injuryor illness. Milbank Memorial Fund Quarterly Health and Society 62:567-90.

Rodgers, G. B. 1979. Income and inequality as determinants of mortality: An interna-tional cross-section analysis. In The society and population health reader, edited by I.Kawachi, B. Kennedy, and R. G. Wilkinson, 5-13. New York: New Press.

Rodriguez-Garcia, R., and A. Goldman. Eds. 1992. The health-development link. Wash-ington, DC: Pan American Health Organization.

Ross, N. A., M. C. Wolfson, J. R. Dunn, J. M. Berthelot, G. A. Kaplan, and J. W. Lynch.2000. Relation between income inequality and mortality in Canada and in theUnited States: Cross sectional assessment using census data and vital statistics. Brit-ish Medical Journal 320:898-902.

Salkever, D. S. 1975. Economic class and differential access to care: Comparisonsamong health care systems. International Journal of Health Services 5:373-95.

Sampson, R. J., S. W. Raudenbbush, and F. Earls. 1997. Neighborhoods and violentcrime: A multilevel study of collective efficacy. Science 277:918-24.

Sen, A. 1999. Development as freedom. New York: Alfred A. Knopf.




Shea, S., D. Misra, M. Ehrlich, L. Field, and C. Frances. 1992. Predisposing factors for se-vere, uncontrolled hypertension in an inner-city minority population. New EnglandJournal of Medicine 327:776-81.

Shi, L. 1994. Primary care, specialty care, and life chances. International Journal of HealthServices 24:431-58.

. 1996. Health services research methods. Albany, NY: Delmar.Shi, L., and B. Starfield. 2000. Primary care, income inequalities, and self-rated health in

the United States: A mixed-level analysis. International Journal of Health Services30:541-55.

. 2001. The effect of primary care physician supply and income inequality onmortality among blacks and whites in US metropolitan areas. American Journal ofPublic Health 91:1246-50.

Shi, L., B. Starfield, B. Kennedy, and I. Kawachi. 1999. Income inequality, primary care,and health indicators. Journal of Family Practice 48:275-84.

Shibuya, K., H. Hashimoto, and E. Yano. 2002. Individual income, income distribution,and self-rated health in Japan: Cross-sectional analysis of a nationally representa-tive sample. British Medical Journal 321:16-19.

Smith, D. B., and A. D. Kaluzny. 1974. Inequality in health care programs: A note onsome structural factors affecting health care behavior. Medical Care 12:860-70.

Soderfeldt, B., B. Danermark, and S. Larsson. 1989. Class inequality in health: A meth-odological study of two measures of social class in relation to sickness insurance di-agnoses. Scandinavian Journal of Social Medicine 17:207-15.

Soobader, M. J., and F. B. LeClere. 1999. Aggregation and the measurement of incomeinequality: Effects on morbidity. Social Science & Medicine 48:733-44.

Starfield, B. 1996. Is strong primary care good for health outcomes? In The future of pri-mary care: Papers for a symposium held on 13th September 1995, edited by J. Griffin, 18-29. London: Office of Health Economics.

. 1998. Primary care: Balancing health needs, services and technology. New York: Ox-ford University Press.

Starfield, B., and L. Shi. 1999. Determinants of health: Testing of a conceptual model.Annals of the New York Academy of Sciences 896:344-46.

. 2002. Policy relevant determinants of health: An international perspective.Health Policy 60 (3): 201-18.

Sturm, R., and R. Gresenz. 2001. Relations of income inequality and family income tochronic medical conditions and mental health disorders: National survey. BritishMedical Journal 324:20-23.

Turrell, G., and C. Mathers. 2001. Socioeconomic inequalities in all-cause and specific-cause mortality in Australia: 1985-1987 and 1995-1997. International Journal of Epide-miology 30 (2): 231-39.

UNU/WIDER. 2000. World Income Inequality Database v1.0: User guide and data sources.September 12. Helsinki, Finland: United Nations Development Program.

van Doorslaer, E., A. Wagstaff, H. Bleichrodt, S. Calonge, U. G. Gerdtham, M. Gerfin, J.Geurts, L. Gross, U. Hakkinen, R. E. Leu, O. O’Donnell, C. Propper, F. Puffer, M. Ro-




driguez, G. Sundberg, and O. Winkelhake. 1997. Income-related inequalities inhealth: Some international comparisons. Journal of Health Economics 16:93-112.

Wagstaff, A., and E. van Doorslaer. 2000. Income inequality and health: what does theliterature tell us? Annual Review of Public Health 21:543-67.

Walberg, P. 1998. Economic change, crime, and mortality crisis in Russia: Regionalanalysis. British Medical Journal 317:312-18.

Waldman, R. 1992. Income distribution and infant mortality. In The society and popula-tion health reader, edited by I. Kawachi, B. Kennedy, and R. G. Wilkinson, 14-27. NewYork: New Press.

Weich, S., G. Lewis, and S. Jenkins. 2001. Income inequality and the prevalence of com-mon mental health disorders in Britain. British Journal of Psychiatry 178:222-27.

Wennemo, I. 1993. Infant mortality, public policy, and inequality: A comparison of 18industrialized countries 1950-85. Sociology of Health and Illness 15:429-46.

West, P. 1991. Rethinking the health selection explanation for health inequalities. SocialScience & Medicine 32:373-84.

Wilkins, R., O. Adams, and A. Brancker. 1989. Changes in mortality by income in urbanCanada from 1971 to 1986. Health Reports/Statistics Canada, Canadian Centre for HealthInformation 1:137-74.

Wilkinson, R. 1992. National mortality rates: The impact of inequality? American Jour-nal of Public Health 82:1082-84.

. 1994. The epidemiological transition: From material society to social disadvan-tage? Daedelus 123:61-77.

. 1996. Unhealthy societies: The afflictions of inequality. London: Routledge.. 1998. Income inequality and population health. Social Science & Medicine

47:411-12.Wolfson, M., G. Kaplan, J. Lynch, N. Ross, and E. Backlund. 1999. Relation between in-

come inequality and mortality: Empirical demonstration. British Medical Journal319:953-55.

Wolleswinkel, J., F. van Poppel, C. Looman, and J. Mackenbach. 2001. The role of cul-tural and economic determinants in mortality decline in the Netherlands, 1875/1879–1920/1924: A regional analysis. Social Science & Medicine 53:1439-53.

World Bank. 1993. World Development Report 1993: Investing in health. New York: OxfordUniversity Press.

World Organization of National Colleges, Academies and Academic Associations ofGeneral Practitioners/Family Physicians (WONCA). 1991. The role of the generalpractitioner/family physician in health care systems. Victoria: Author.




income inequality and health: acritical review of the ... · the theoretical literature was used to...

Documents