Stephanie A Robert & James S House. Handbook of Social Studies in Health and Medicine. Editor: Gary L Albrecht, Ray Fitzpatrick, Susan C Scrimshaw. Sage Publications, 2000.
Socioeconomic inequalities in health have been observed persistently over the course of human history. These differences are manifest across individuals, communities, and societies, and recent analyses suggest that for the most part they have increased over the past century, and even in the past few decades (Marmot et al. 1987; Pappas et al. 1993; Preston and Haines 1991). The nature and size of these inequalities make them arguably the major problem of population and public health in America and many other societies for reasons that will become clearer below. Hence, socioeconomic inequalities in health have increasingly become a focus of health policy (Department of Health and Human Services 1990) as well as health research.
We still do not well and consensually understand, however, why socioeconomic inequalities in health exist and persist, nor what policies are most likely and necessary to reduce these inequalities. In seeking such understanding, research has increasingly focused on socioeconomic differentials in health at the level of communities and societies as well as at the level of individuals. Yet, there has been little integration of our substantial knowledge of the relationship between individual-level socioeconomic position and health with our growing understanding of the relationship between community-level socioeconomic position and health. Thus, this chapter seeks to focus on: (1) what we have learned from studies of socioeconomic inequalities in health at the levels of individuals, communities, and societies, (2) whether and how the information from such multiple levels of analysis helps us to understand better the nature and explanations for socioeconomic inequalities in health at all levels, and (3) how social and health policy might address socioeconomic inequalities in health.
A comprehensive review of the literature on individual- and family-level socioeconomic inequalities in health is beyond the scope of this or any single chapter or paper of moderate length. There are, however, sources that have reviewed much of the literature, and we will refer readers to these while highlighting only the major theories, findings, and issues relevant to our discussion. We will first summarize major findings of individual- and family-level research on socioeconomic differentials in health, so that our later discussion of community-level research will indicate: (1) how findings from research on individual- and family-level socioeconomic position might inform research and theory at the level of communities or societies on socioeconomic differentials in health, and (2) how multilevel research on socioeconomic differentials in health may answer some of the questions that have previously been left unanswered in individual- and family-level research.
The Relation of Individual- and Family-Level Socioeconomic Position to Health: Major Findings and Issues
A large body of research, well-reviewed elsewhere (Adler et al. 1994; Antonovsky 1967; Feinstein 1993; Kaplan and Keil 1993; Krieger and Fee 1994; Marmot et al. 1987; Townsend and Davidson 1982; Williams 1990; Williams and Collins 1995), has documented the higher rates and risk of mortality and morbidity from most causes, as well as of functional limitations, among persons who have lower socioeconomic position – people with lower levels of education, income, occupation, material possessions and/or wealth, or people who are part of marriages or households with such characteristics. There remains a number of issues and factors that may qualify this generalization and which deserve to receive increasing attention in research.
Variations across Measures of Socioeconomic Position
A good deal has been written on how to measure socioeconomic position, which are the best indicators of socioeconomic position, and whether some indicators are more predictive of health in different populations (e.g., Berkman and Macintyre 1997; Krieger et al. 1997; Liberatos et al. 1988). To a considerable degree these questions remain unanswered, and perhaps unanswerable in a generic sense. There is often considerable variation in the extent to which different indicators of socioeconomic position can be measured in different populations and in the precision and reliability of such measures. European research makes heavy use of occupational indicators, while research in the United States relies more heavily on income and education, the last being the most widely used indicator in underdeveloped nations. Wealth or permanent income is now being used more in health research in the United States and Canada (e.g., Kington and Smith 1997; Robert and House 1996; Schoenbaum and Waidmann 1997; Wolfson et al. 1993). Material circumstances (such as car ownership or housing tenure) have been additional indicators used primarily in British research (e.g., Arber and Ginn 1993; Marmot et al. 1987).
We support the position of Krieger et al. (1997) that composite indices of socioeconomic position are generally to be avoided in favor of using a variety of separate indicators. Education and income, if measured reasonably well, have the virtues of being applicable to all individuals and being relatively continuous in nature, and the same is true of wealth and material possessions, although these are more difficult to measure well. Occupation, in contrast, works well for employed populations, but becomes increasingly difficult or even inappropriate to apply to those not or never in the labor force. Researchers should try to use multiple measures of socioeconomic position, as there is evidence that:
- Different measures have both common and independent pathways linking them to health (e.g., education affects health both through and independent of its impact on income (Reynolds and Ross 1998));
- Some measures of socioeconomic position may be particularly salient for specific populations or subgroups (e.g., wealth may increasingly rival or surpass income as a measure of the socioeconomic position of the elderly (Robert and House 1996));
- Different socioeconomic factors may affect different health outcomes in different degrees and ways (e.g., education may be more important for health outcomes and risk factors such as cardiovascular disease or smoking that have their origins earlier in life (Blane et al. 1997; Davey Smith et al. 1998a)).
Temporality and Causality
In addition, care needs to be given in several ways to conceptualizing and measuring what role time plays in measuring socioeconomic position and its relationship to health. First, rather than measuring socioeconomic position at one point in time and assessing its relationship to health and mortality, we need to understand how stability and change in socioeconomic position relate to health and mortality (Lynch et al. 1997b). For example, McDonough et al. (1997), using a longitudinal panel study of adults ages 45 and older in the United States, found that persistent low income was a particularly strong determinant of mortality, but that income instability was also an important predictor of mortality among middle-income adults.
Second, research needs to clarify the complex relationships among childhood socioeconomic position, childhood health, adult socioeconomic position, and adult health. Do socioeconomic conditions of childhood have a profound effect on health in adulthood, or are socioeconomic conditions in adulthood the primary determinants of health in adulthood? Measuring the association between adult socioeconomic position and health ignores the potential role of childhood socioeconomic position on both adult socioeconomic position and health, and may thereby overstate the role of adult socio-economic position and/or understate the role of childhood socioeconomic position. Recent research from Power and colleagues (Power and Matthews 1997; Power et al. 1996, 1998) suggests that occupational class differences in health at age 33 in Great Britain result from the accumulation of conditions and experiences throughout both childhood and early adulthood. Other studies generally conclude that childhood socioeconomic conditions are related to adult health and mortality both through and independent of adult socioeconomic conditions. However, childhood socioeconomic conditions are not fully, or even primarily, responsible for the robust association between adult socioeconomic position and health (Blane et al. 1996; Brunner et al. 1996; Elo and Preston 1992; Kaplan and Salonen 1990; Kuh and Ben-Shlomo 1997; Lynch et al. 1994, 1997a; Peck 1994; Reynolds and Ross 1998). Again, childhood socioeconomic position appears to be more consequential for health outcomes and risk factors with long-term etiologies, whereas adult socioeconomic position may be more consequential for other health outcomes and risk factors (Davey Smith et al. 1997, 1998b). Further research measuring socioeconomic inequalities in health at different points in the life course can help us understand the pathways linking socioeconomic position to health. It can also help us better understand at what point in the life course different types of interventions might be most beneficial (Bartley et al. 1997; Kuh and Ben-Shlomo 1997).
Third, and related to the prior discussion, we need to better understand the causal relationship between socioeconomic position and health. Some suggest that it is the effects of poor health on restricting or reducing socioeconomic position that drives the overall socioeconomic inequalities in health. Such claims rely primarily on research in economics showing that extreme levels of ill health constrain the ability of individuals or whole populations to engage in productive work roles (Fogel 1991; Fuchs 1983; Shaar et al. 1994; Smith and Kington 1997). In contrast, most sociologists and social epidemiologists, though recognizing that health must play some role in affecting socioeconomic position, view the causal direction as running primarily from socioeconomic position to health. Such conclusions are usually based on research that (a) shows a prospective effect of socioeconomic position on health and mortality while adjusting for health at baseline (e.g., House et al. 1994; Marmot et al. 1997; Mirowsky and Hu 1996), (b) probes in various ways the ability of selection effects to explain the association between socio-economic position and health and finds that selection effects of health on socioeconomic position cannot be the major explanation (Blane et al. 1993; Fox et al. 1985; Lichtenstein et al. 1993; Lynch et al. 1997b; Power et al. 1996; Wolfson et al. 1993), and (c) indicates that actual patterns of downward and upward mobility work to constrain rather than cause overall patterns of socioeconomic inequalities in health (Bartley and Plewis 1997). Research is still needed to estimate more precisely the relative effects of socioeconomic position on health and vice versa.
An intriguing finding of some research on socio-economic inequalities in health is that it is not just those who are in the lowest socioeconomic groups that have poorer health than those in the higher socioeconomic groups. Rather, a relationship between socioeconomic position and health has been observed across the socioeconomic hierarchy, with even those in relatively high socioeconomic groups having better health than those just below them in the socioeconomic hierarchy (Adler et al. 1994; Marmot et al. 1991). Perhaps the most important implication of this finding is that it is not just the material, psychological, and social conditions associated with severe deprivation or poverty (such as lack of access to safe housing, healthy food, and adequate medical care) that explain socio-economic inequalities in health among those already at relatively high levels of socioeconomic position.
Despite evidence for gradient effects of socio-economic position on health, it is also important to note that many studies indicate that the relationship of socioeconomic position to health is monotonic but not a linear gradient, particularly when socioeconomic position is indexed by a measure of income. Although increasingly higher levels of socioeconomic position may be associated with increasingly higher levels of health, there are also substantially diminishing returns of higher socioeconomic position on health. For example, studies have found diminishing and even nonexistent relationships between income and mortality (Backlund et al. 1996; Chapman and Hariharan 1996; McDonough et al. 1997; Wolfson et al. 1993) and morbidity (House et al. 1990, 1994; Mirowsky and Hu 1996) at higher levels of income (e.g., above the median). This trend partially reflects a health ‘ceiling effect’ caused by the fact that people in the upper socioeconomic strata maintain overall good health until quite late in life, leaving little opportunity for improvements in health among these groups throughout much of adulthood (House et al. 1994). Thus, it is most important to understand what accounts for socioeconomic inequalities in health across the broad lower range (e.g., the lower 40-60 per cent) of socioeconomic position, rather than focusing mainly or only on factors that might explain this relationship across the gradient or at higher levels.
In the United States, race and socioeconomic position are intertwined in complex ways, making it crucial that research on race differentials in health consider the role of socioeconomic position, and that research on socioeconomic inequalities in health consider the role of race. Regarding the former, a sizable and growing number of studies find that much, but not all, of race differences in health in the United States are explained by socioeconomic factors (Clark and Maddox 1992; Kington and Smith 1997; Krieger and Fee 1994; Krieger et al. 1993; Lillie-Blanton and LaVeist 1996; Mendes de Leon et al. 1997; Mutchler and Burr 1991; Rogers et al. 1996; Schoenbaum and Waidmann 1997; Williams and Collins 1995). However, these studies on race differences in health have not included a full range of socio-economic measures – most notably excluding community-level socioeconomic measures. Many have argued that simply controlling for individual- and family-level socioeconomic position when looking at race differences in health overlooks the significant race differences in the types of neighborhoods that whites and non-whites live in, even at similar levels of individual-and family-level socioeconomic position (Anderson and Armstead 1995; Lillie-Blanton and LaVeist 1996). For example, in metropolitan areas in 1990, only 6.3 per cent of poor white people lived in high poverty areas, compared with 33.5 per cent and 22.1 per cent of poor black and poor Hispanic people, respectively (Jargowsky 1997). Thus, the socioeconomic characteristics of the community may further explain observed race differences in health, a point we return to in discussing community-level socioeconomic effects on health.
Often research focusing on socioeconomic inequalities in health does not investigate whether this relationship differs by race, and what little research there is has found inconsistent results. Krieger et al. (1993) summarize work showing that education does not have the same economic return (e.g., actual salary, nonwage benefits, or occupational status) for blacks as it does for whites, raising the question of whether there might also be differential socio-economic returns to health by race. Some research does find that education has less of an effect on measures of self-rated health (Mutchler and Burr 1991; Reynolds and Ross 1998), coronary heart disease (Diez-Roux et al. 1995), and infant mortality (Din-Dzietham and Hertz-Picciotto 1998; Schoendorf et al. 1992) among blacks compared with whites, whereas other research finds virtually no race differences in the effects of income (Diez-Roux et al. 1995; Hahn et al. 1996; McDonough et al. 1997; Mutchler and Burr 1991) and occupation (Gregorio et al. 1997) on health and mortality. Gillum et al. (1998) found that both education and income predict coronary heart disease incidence in white men, white women, and black men, but neither predict coronary heart disease incidence in black women.
In sum, race and socioeconomic position are inextricably related to each other and to health, and hence one cannot be considered without the other. Socioeconomic position is a major explanation of race differences, but not the full one. Other experiences associated with race in our society, such as discrimination (Hummer 1996; Krieger and Sidney 1996; Krieger et al. 1993; Williams 1997; Williams and Collins 1995) and residential segregation (Jargowsky 1997; Massey and Dentón 1993), may also account for some race effects on health. Finally, the relation of different indicators of socioeconomic position to health may vary across racial/ethnic populations due to the differential importance or sensitivity of different socioeconomic measures across these populations.
Despite the fact that women are more likely than men to have lower socioeconomic position and higher morbidity, socioeconomic inequalities in health have often been found to be stronger in men than in women. This finding has resulted in much debate about whether standard measures of socioeconomic position are equally appropriate for men and women, particularly whether married women should be classified according to their own socioeconomic position, that of their husbands, or both. Although some research finds that measuring socioeconomic position at the individual or at the family level makes little difference in patterns of socioeconomic inequalities in health for women (Arber and Ginn 1993), other research suggests that measuring socioeconomic position at both the individual and family level may be important to understanding the full association between socioeconomic position and health – for both women and men (Krieger et al. 1993; Pugh and Moser 1990). For example, Krieger et al. (1993) suggest that individual-level socioeconomic position may be most directly related to working conditions, whereas family-level socioeconomic position may be most directly related to one’s overall standard of living. Community-level socioeconomic conditions might be considered additional measures of a family’s overall standard of living, and one that may be particularly salient for women who do not work and who spend a substantial amount of time in their community environment.
Other research suggests that the issue of gender differences in the relationship between socioeconomic position and health goes beyond determining how to classify the socioeconomic position of married women and homemakers. Gender differences in labor force participation and in the structure and quality of occupations themselves may play a role in explaining gender differences in the relationship between socio-economic position and health (Arber 1991; Arber and Lahelma 1993; Stronks et al. 1995). For example, Arber and Lahelma (1993) compared Finland and Britain and found that socioeconomic inequalities in health are strong for both women and men in Finland, but only for men in Britain. However, housewives in Britain were found to have particularly poor health. The researchers suggest that in countries with a high degree of female labor force participation, socioeconomic position may be strongly related to health for both men and women, whereas in countries with less female labor force participation, women’s family roles and housing characteristics may play more of a role than socioeconomic position in affecting women’s health. Other research in The Netherlands suggests that the more pronounced relationship between socio-economic position and health in men compared with women may partially reflect the poor working conditions of men with low socioeconomic position (Stronks et al. 1995).
In sum, research generally demonstrates a stronger relationship between socioeconomic position and health for men compared with women, which challenges us to consider: (1) whether community-level socioeconomic conditions may play an additional role in affecting health, particularly for women who do not work; (2) how gender differences in labor force participation and in family roles both directly affect and interact with socioeconomic position to ultimately affect health; (3) what role gender differences in working conditions may play in explaining gender differences in the relationship between socioeconomic position and health.
Despite the strong overall relationship between socioeconomic position and health, this relationship varies by age. Socioeconomic differences in prenatal, neonatal, and infant health and mortality are large (Aber et al. 1997; Singh and Yu 1996), but there are strikingly diminished socio-economic differences by adolescence (Ford et al. 1994; West 1997; West et al. 1990). With few exceptions (Ross and Wu 1996), research suggests that socioeconomic inequalities in adult health and mortality are generally small in early adulthood, increasingly larger through middle and early old age, and then smaller again in later old age (Elo and Preston 1996; Haan et al. 1987; House et al. 1990, 1994; Kaplan et al. 1987; Kitagawa and Hauser 1973; McDonough et al. 1997; Sorlie et al. 1995; Wilkins et al. 1989). This age variation in the relationship between socioeconomic position and health challenges researchers to discover why such age variation exists. Robert and House (1994) have described some of the potential explanations for this diminished relationship between socioeconomic position and health at older ages. (1) Health and social policies targeted to older people (such as Medicare and social security benefits) might help equalize access to care and resources at older ages. (2) Only the hardiest and healthiest people with low socioeconomic position may survive infancy and into older ages, making their health increasingly similar with age to that of people with higher socioeconomic position. (3) There may be age variation in how socioeconomic position affects exposure to and impact on mediating psychosocial, behavioral, and environmental factors that are known to help explain socioeconomic inequalities in health. (4) Standard measures of socioeconomic position may be less applicable to older adults, thereby showing a diminished relationship between socioeconomic position and health at older ages that reflects poor measurement rather than a true relationship. (5) The biological robustness of late adolescence/early adulthood and the frailty of later old age may somewhat limit the ability of socio-economic position to affect health at these ages. To date, there is some evidence for each of these explanations, although we are still far from understanding the relative importance of these and other explanations. Yet, if we can better understand why age variations in the relationship between socioeconomic position and health exist, we will certainly be much closer to having a more comprehensive understanding of the overall relationship between socioeconomic position and health.
Explanations for Socioeconomic Inequalities in Health
The Central Mediating Role of Psychosocial Risk Factors
As other reviews have pointed out (Feinstein 1993; Krieger et al. 1993), the literature has been more successful at documenting the existence and patterns of socioeconomic inequalities in health than in explaining why these inequalities persist. Recent work, however, suggests an emerging consensus. Research indicates that people in lower socioeconomic strata tend to be disadvantaged in a broad array of biomedical, environmental, behavioral, and psychosocial risk factors for health, which mediates the relationship between socioeconomic position and health (see Chapter 1.9 in this volume for more extensive discussion of mechanisms by which this occurs). Central among these are health behaviors (Berkman and Breslow 1983), chronic and acute stress in life and work (Karasek and Theorell 1990; Theorell 1982), hostility and depression (Scheier and Bridges 1995), lack of social relationships and supports (House et al. 1988), and lack of control, efficacy, or mastery (Rodin 1986). While any single or small subset of these risk factors (e.g., medical care, health behaviors such as smoking, drinking, eating, and exercise, or biomedical risk factors such as blood pressure and cholesterol) can account for only a small fraction (10-20 per cent) of the association between socioeconomic position and health (e.g., Feldman et al. 1989; Lantz et al. 1998; Marmot et al. 1984), a broad array (12-25 or so) of such risk factors can explain 50-100 per cent of the relationship between socioeconomic position and various measures of health, functional status, and mortality (House et al. 1992, 1994; Lundberg 1991; Lynch et al. 1996; Marmot et al. 1991, 1997; Power et al. 1998; Ross and Wu 1995). For example, Lynch et al. (1996) found that an array of 23 biological, behavioral, psychological, and social risk factors accounted for much of the income differentials in mortality. For men in the lowest income quantile, adjusting for the set of risk factors reduced the excess relative risk of all-cause mortality by 85 per cent, the risk of cardiovascular mortality by 118 per cent, and the risk of acute myocardial infarction by 45 per cent. Although no study has been able to include a full range of environmental, psychosocial, and behavioral factors, it is likely that including a full range of these factors might consistently account for the relationship between socioeconomic position and health at one point in time. However, determining which factors mediate the relationship between socioeconomic position and health at one point in time is just that – a snapshot of current relationships. Since the mechanisms linking socioeconomic position to health may evolve and change over time (House et al. 1990; Link and Phelan 1995; Williams 1990), research must continue examining which factors become more or less crucial in perpetuating the relationship between socioeconomic position and health.
In addition, some studies suggest that it is not only differential exposure to these mediating factors that lead to poorer health among people with lower socioeconomic position, but that differential vulnerability to those exposures may also help explain social inequalities in health (Krieger et al. 1993). The strongest evidence for this comes from studies of racial, gender, and socioeconomic differences in mental health, with more suggestive evidence for physical health (House et al. 1992; McLeod and Kessler 1990).
Research on the gradient relationship between socioeconomic position and health suggests that we pay less attention to mediating factors associated with extreme material deprivation, such as unsanitary or inadequate material living conditions, and focus primarily on the psychosocial factors that may be more directly related to relative deprivation across the entire socio-economic scale (Adler et al. 1994). However, a recent analysis by Cohen et al. (forthcoming) found that psychosocial factors were of equal or even greater importance in explaining socio-economic inequalities in self-rated health among people in the lowest socioeconomic categories, rather than being more important among people in the higher socioeconomic categories. There is also speculation that gradient effects of socio-economic position on health are seen because there may be something about simply being lower in any hierarchy that may be detrimental to health (Adler et al. 1994). Despite a number of suggestive findings from animal studies about physical reactions to social ordering or hierarchical status (Sapolsky 1992), studies on the potential direct effects of socioeconomic inequality or relative deprivation on health among humans are essentially nonexistent at this time.
Moreover, excessive focus on the mediating environmental, psychosocial, and behavioral pathways that help to ‘explain’ socioeconomic inequalities in health can lead to ignoring or downplaying the role of socioeconomic position as a fundamental cause of health (House et al. 1990, 1994; Link and Phelan 1995; Williams 1990). Focusing on the mediating factors may lead policy makers to conclude that these factors are the more important causes of ill health that should be targeted in efforts to improve health. In contrast, by recognizing the primary importance of socioeconomic position in affecting a broad range of health outcomes through multiple pathways, attention may be focused on more broad-based interventions, such as altering social and economic policies, which may more effectively improve health. This argument is particularly compelling in view of the fact that although the factors mediating the relationship between socioeconomic position and health have changed over time, the association between them has persisted (House et al. 1990; Link and Phelan 1995; Williams 1990).
Importance of Medical Care
Most research suggests that access to medical care plays a relatively minor role in explaining socioeconomic inequalities in health. Such conclusions have been made for three primary reasons. First, cross-national research indicates that socioeconomic inequalities in health have persisted or increased even in countries that have initiated national health programs that somewhat equalize access to health care (Adler et al. 1993; Roos and Mustard 1997; Townsend and Davidson 1982; Wilkins et al. 1989). Second, socioeconomic differences are seen both in diseases that are amenable to medical treatment and in diseases that are not amenable to medical treatment (Adler et al. 1993), with deaths from diseases amenable to treatment representing only a fraction of all deaths in any case (Marmot et al. 1987; Poikolainen and Eskola 1986). Third, some studies have controlled for factors related to health care (such as health insurance, number of visits to doctors) and found that these factors do not account for much of the association between socioeconomic position and health (Marmot et al. 1987; Ross and Wu 1995; Williams 1990).
However, we should not be too quick to dismiss the role of access to and quality of medical care. First, even countries that are supposed to have more equal access to health care have found that differential access to and quality of care still exist (Katz and Hofer 1994). Also, large socioeconomic position and race differences in health, mortality, and health care exist even among participants in the Medicare program in the United States (Gornick et al. 1996). Second, studies that investigate the role of mediating factors in explaining the association between socioeconomic position and health have had inadequate controls for the many aspects of access to medical care (e.g., adequate transportation, prescription coverage) and quality of medical care (e.g., continuity of care, access to preventive care) that may be distributed unevenly by socioeconomic position (Feinstein 1993). The downplaying of the role of medical care is consistent with findings that medical care plays only a minor role in the overall health of populations in more developed countries (Bunker et al. 1994; McKeown 1976), but a more careful look needs to be taken at whether particular aspects of medical care may still explain some of the impact of socioeconomic position on health.
In sum, although it was once hoped and believed that socioeconomic differentials in health would wither away with increasing economic development and improvements in the technology, practice, and availability of medical care, several decades of research in the most developed countries (and increasingly in the developing countries as well) indicate that this has not come to pass in the United States and most other developing countries. This has led to a veritable explosion of research (Kaplan and Lynch 1997), largely at the level of individuals and families, which has not only documented the persistence, pervasiveness, and even perhaps increase of socioeconomic inequalities in health, but has also substantially advanced our understanding of the processes producing them.
We now know that these inequalities exist across a wide range of dimensions of socioeconomic position, though we still need to understand better how these interrelate in affecting health. We have substantial evidence that temporal and causal priority flows from socioeconomic position to health in a process that may cumulate over the life course, but which is driven most strongly by the current and recent socio-economic positions of individuals in families, though again more research is needed in this area. We know that the relationship between income, and probably many or most other socio-economic dimensions, and health is monotonic but nonlinear (following a path of diminishing impact, especially at levels above the mean or median), though again further research is needed to refine our understanding, and we know that the effects of socioeconomic position on health may vary by race, age, and gender in ways not yet well understood. There is a great need for more research on how socioeconomic inequalities combine with inequalities by race, gender, and age in affecting health.
We have evidence that the impact of socio-economic position on health is mediated by our exposure to a very broad range of psychosocial and behavioral, as well as physical-chemical-biological risk factors to health. However, we also increasingly understand that these factors can never fully explain or eliminate socio-economic differences in health because, as new risk factors emerge, exposure to them ultimately comes to be differentiated by socioeconomic position. In this regard, the role of medical care, especially in the forms of both prevention and advanced treatment, needs to be reexam-ined, even as we know that medical care is at best a minor part of the explanation of existing socioeconomic inequalities in health. Finally, the role of socioeconomic position as a fundamental cause of health suggests a need to look not only at how these socioeconomic effects may be understood or explained on the individual and family levels or more microscopic ones (e.g., psychophysiology), but also how broader community, state, and national contexts may affect and interact with individual- and family-level socio-economic position in affecting individual and population health.
The Health Impact of Community- and Societal-Level Socioeconomic Conditions
An emerging focus of the study of socioeconomic inequalities in health is the potential health impact of the socioeconomic characteristics of communities and other large social aggregates (e.g., states, regions, and countries). Two types of studies involving community socioeconomic conditions offer new perspectives and information about socioeconomic inequalities in health. One investigates the health impact of the socioeconomic level of communities, whereas the other investigates the health impact of socio-economic inequality within and between communities, counties, states, and countries. These new lines of research, however, must both inform and be informed by research on individual-and family-level socioeconomic position.
Socioeconomic Level of Communities
Research consistently shows that communities with lower average levels of income, education, etc., have higher rates of morbidity and mortality than communities with higher socioeconomic levels (Crombie et al. 1989; Guest et al. 1998). However, because these findings derive from ecological data, it is unclear to what extent communities with worse socioeconomic conditions have worse overall health. This is because (1) people with lower socioeconomic position in those communities have poor health, or (2) living in communities with worse socioeconomic conditions is detrimental to the health of all residents, in addition to (or interacting with) their individual- or family-level socioeconomic position. That is, is the aggregate relation at the community level simply reflecting the relationship at the individual or family level discussed above, or is there an effect of community socio-economic level on individual health that is over and above the effect of individual- or family-level socioeconomic position? Few studies have actually tested this question, primarily because most existing data sets do not contain adequate information about the socioeconomic characteristics of respondents, their families, and their communities, as well individual health information.
With few exceptions (Ecob 1996; Reijneveld 1998; Sloggett and Joshi 1994), extant studies do find that community socioeconomic conditions are associated with various measures of health status (Diez-Roux et al. 1997; Hochstim et al. 1968; Jones and Duncan 1995; Kaplan et al. forthcoming; Krieger 1992; Morgan and Chinn 1983; O’Campo et al. 1997; Reijneveld 1998; Robert 1998; Sloggett and Joshi 1998) and mortality (Anderson et al. 1997; Davey Smith et al. 1998c, 1997; Haan et al. 1987; LeClere et al. 1997, 1998; Waitzman and Smith 1998a), over and above the impact of individual-and family-level socioeconomic position.
For example, early work by Haan et al. (1987) showed that, in a population of adults age 35 and older living in Oakland, California, in 1965, the effects of residence in a poverty area on 9-year mortality persisted after controlling separately for individual-level measures of socio-economic position, age, sex, and race, and even after adjusting for mediating behavioral factors. A more recent study by Waitzman and Smith (1998) found that residence in a poverty area predicted mortality among individuals aged 25-54 years across thirty-three metropolitan areas, after adjusting for age, race, sex, marital status, household income, formal education, poverty status, baseline health, and multiple behavioral and biological risk factors. Similarly, focusing on health status rather than mortality in a national sample of adults in the United States in 1986, Robert (1998) found that the percentage of families earning $30000 or more and the percentage of adult unemployment in respondents’ census tracts each had an independent association with the number of chronic conditions, over and above the effects of respondents’ individual- and family-level education, income, and assets, and their age, race, and sex. Similarly, the percentage of households receiving public assistance had independent associations with self-rated health, after controlling for the same individual- and family-level variables.
Although such multilevel studies indicate an independent role of community socioeconomic conditions in predicting health and mortality, most of the community-level effects have been relatively small in size, tempering any grand conclusions that can be made about the importance of community-level socioeconomic effects. For example, Robert (1998) found that not all measures of community socioeconomic conditions were associated with all measures of health, after controlling for individual- and family-level socioeconomic position and demographic variables. In addition, all the observed community socioeconomic effects were substantially smaller in size than individual- and family-level socioeconomic effects. Contextual or multilevel analyses in other areas of research have similarly found relatively weak independent community-level effects (Brooks-Gunn et al. 1997; Elliott et al. 1996; Jencks and Mayer 1990).
Many of the multilevel studies have also had limitations that further temper conclusions about the existence or strength of community-level socioeconomic effects on health. For example, some studies (Anderson et al. 1997; Haan etal. 1987; Hochstim et al. 1968; O’Campo et al. 1997) have had few or inadequate individual-and family-level socioeconomic controls, making it difficult to conclude that observed community-level socioeconomic effects were not just picking up unmeasured individual- or family-level socio-economic effects. Alternatively, community socioeconomic characteristics might just be proxies for a person’s ‘usual’ or ‘permanent’ socioeconomic position – balancing out the short-term variability of reported income, assets, or occupation that is inherent in a one-time cross-sectional survey (Davey Smith et al. 1996; Krieger et al. 1997). Some other methodological and substantive criticisms of multilevel analyses include (Blalock 1984; Diez-Roux 1998; Duncan et al. 1997; Hauser 1970, 1974; Krieger et al. 1997): not considering the existence and impact of stability and change in community characteristics; not considering length of time respondents lived in their communities; not fully accounting for factors affecting selection into and out of certain types of communities; not having measured ‘community’ in any meaningful way since the focus is primarily on census areas rather than more subjective community boundaries. Other criticisms include not using appropriate statistical methods that account for the multilevel nature of the data, and not taking into account potential differences in measurement error at the individual vs. aggregate level. Every existing multilevel analysis of the effects of socioeconomic position on health has been characterized by a number of these and other problems that could bias the effects of community variables either upward or downward, both absolutely and relative to individual- or household-level effects. As a result, making generalizations about findings in this area is still tentative.
Comparisons with and Contributions to Studies on Individual- and Family-Level Socioeconomic Position
Nevertheless, multilevel research on socioeconomic inequalities in health has the potential to amplify and clarify the complex pathways through which socioeconomic position affects health, if this multilevel research simultaneously informs and is informed by individual- and family-level socioeconomic research. Just as studies on individual- and family-level socioeconomic position have noted gradient effects on health, some of the community-level studies suggest that not only does living in a poverty area have an independent impact on health and mortality, but there also seem to be gradient effects of community socioeconomic conditions on health as well, such that those living in the highest socioeconomic communities have better health even than those living in communities just below them on the socioeconomic scale (e.g., Jones and Duncan 1995; LeClere et al. 1997; Robert 1998). It may be that the impact of community socioeconomic conditions on health helps to explain the moderate gradient effects seen at higher levels of socioeconomic position in some studies at the individual and family level. That is, these individual-level gradient effects may reflect the fact that those with similar high income levels may nevertheless reside in different types of communities that vary by socioeconomic profile. If so, the gradient association between individual- or family-level socioeconomic position and health may weaken or disappear once community socioeconomic conditions are controlled for.
Research on community-level socioeconomic effects on health may also help us better understand how socioeconomic inequalities in health vary by race, gender, and age. Regarding race, as discussed earlier, research on race differences in health often controls for individual- or family-level socioeconomic position to see how much of the race effects are explained by individual- or family-level socioeconomic position. However, community socioeconomic conditions may further explain the relationship between race and health, since nonwhites are more likely to live in lower socioeconomic communities than whites of the same individual socioeconomic position (Jargowsky 1997). In a national sample of adults in the United States in 1986, race differences (black vs. nonblack) in number of chronic conditions persisted after controlling for age, sex, income, education, and assets, but disappeared after controlling further for community socioeconomic characteristics (Robert 1998). Similarly, Haan et al. (1987) found that the association between race and mortality (controlling for individual-level socioeconomic position) was virtually eliminated after controlling for residence in a poverty area in Alameda County, California. In contrast, LeClere et al. (1997) found that in a sample of adults in the United States, the effect of race on mortality (controlling for individual-level socioeconomic position) was not entirely eliminated by adding community median income, but was eliminated after including community-level measure of racial minority concentration. Although generally consistent with theories about the role of residence in explaining racial disadvantage (Jargowsky 1997; Massey and Dentón 1993; Wilson 1987), such ideas have gone relatively untested, particularly as they relate to health outcomes.
Further, the relationship between community socioeconomic conditions and health may vary by race. Diez-Roux et al. (1997) found that for African-American men living in poor neighborhoods in Jackson, Mississippi, in the late 1980s, the prevalence of coronary heart disease actually decreased as neighborhood characteristics worsened. However, Collins et al. (1997) found that both African-American and white infants born in Chicago were less likely to be very-low-birth-weight infants if their parents had lived in communities with a higher income level than their own. Anderson et al. (1997), similarly found no race differences in the association between community-level median income and all-cause mortality after controlling for family income in a national sample of adults. Kaplan et al. (forthcoming) found no race differences in the impact of living in a poverty area on 9-year incidence of disability in a sample of adults in Oakland, California.
With respect to age, research has found smaller or nonexistent associations between community socioeconomic conditions and health for people 65 years and older (Anderson et al. 1997; Haan et al. 1987; Waitzman and Smith 1998a). In contrast, Robert (1996) found that community socioeconomic characteristics were better predictors of adult health at middle and older ages compared with younger ages, and that community socioeconomic characteristics were sometimes better predictors of health than individual- or family-level socioeconomic measures at these middle and older ages.
As discussed earlier, community socioeconomic characteristics might be particularly salient to the lives and health of women compared with men, particularly for women who do not work outside the home. Diez-Roux et al. (1997) found increased odds for coronary heart disease among white women living in more disadvantaged communities, even after controlling for individual-level socioeconomic measures, with much weaker effects for white men. In contrast, LeClere et al. (1997) found that community socioeconomic indicators were better predictors of mortality for men than for women (after controlling for measures of individual-level socioeconomic position). Anderson et al. (1997) found no gender differences in the association between community-level median income and all-cause mortality (after controlling for family income).
At this point, evidence for race, age, and gender differences in the relation of community socioeconomic characteristics to health is contradictory, but the potentially complex interactions between race, age, gender, individual-level socio-economic position, community-level socioeconomic characteristics, and residential segregation need to be explored in more detail to help us understand how they work to affect health.
Studies on the multilevel effects of socioeconomic position on health also allow us to test for possible health effects of interactions between individual- and community-level socio-economic position, and thus to test theories about double jeopardy and relative deprivation. The double-jeopardy hypothesis suggests that living in a lower socioeconomic community would be particularly detrimental to individuals with low socioeconomic position themselves. Alternately, living in a higher socioeconomic community could be worse for the health of people with lower socioeconomic position than living in a lower socioeconomic community because people with lower socioeconomic position would experience greater relative deprivation in higher socioeconomic communities. Diez-Roux et al. (1997) found no interactions between community-level and individual-level socioeconomic position when predicting coronary heart disease among white adults. However, other results from Britain (Jones and Duncan 1995) and the United States (Kaplan et al. forthcoming; O’Campo et al. 1997) also suggest that interactions between individual- and community-level socioeconomic position do exist, although they are complex and difficult to interpret.
Problems in determining explanations for the independent association between community socioeconomic characteristics and health are similar to problems determining explanations for the association between individual- and family-level socioeconomic position and health. Explanations for the association that are caused by methodological limitations (such as reverse causality or selection problems) need to be excluded before substantive or theoretical explanations for the association can be verified. Further, the mechanisms proposed to mediate this relationship are so varied (Macintyre et al. 1993; Robert 1998) that one study is unlikely to be able to explore the mediating effects of all explanations at once, necessitating a cumulative approach to building our knowledge base in this area, and we are only beginning to assemble and analyze the complex, multilevel data sets needed for such studies. Finally, even if we could fully understand which characteristics of the physical, social, and service environments of communities account for the relationship between community socioeconomic conditions and health, this would not preclude debate about how to use this information to improve health. As with individual-and family-level research, the question arises as to whether one can improve health best by addressing the characteristics of the physical, social, and service environments of communities, or by more directly improving the socio-economic profiles of communities.
Socioeconomic Inequality within and between Societies and Communities
Another line of research suggests that it is not just the absolute level of income or deprivation of communities that is associated with the health and mortality of residents. Income inequality within communities or societies is also associated with health and mortality, and may even be more important than level of income for developed countries. Cross-national comparisons indicate that country-level measures of average socioeconomic levels (e.g., per capita income) are associated with population health (e.g., life expectancy), although this relationship is nonlinear (Preston 1975; Wilkinson 1996). Per capita income is strongly and linearly associated with health among less developed countries, but as per capita income rises, the relationship weakens and becomes almost nonexistent among more developed countries. For example, among OECD countries in 1990, there was virtually no association between gross domestic product per capita and life expectancy (Wilkinson 1996).
However, although some remain to be convinced (e.g., Judge 1995), a growing body of research over two decades indicates that among developed countries, and to a lesser degree developing countries as well, the degree of income inequality within a country is strongly and linearly associated with differences in life expectancy between countries. This is the case even after controlling for average level of income within each country (Rodgers 1979; van Doorslaer et al. 1997; Wilkinson 1992, 1996). Simply put, among developed countries, the bigger the gap in income between the rich and poor, the poorer the health of the population. In addition, comparing areas within single countries rather than between countries, studies in England (Ben-Shlomo et al. 1996) and the United States (Kaplan et al. 1996; Kennedy et al. 1996; Lynch et al. 1998) suggest that differences in income inequality across local authorities in England and across states and metropolitan areas in the United States are strongly related to mortality rates (see Lynch and Kaplan 1997 for an excellent review and appraisal of this body of research and critiques of it).
Reactions to this new line of research range from extreme excitement to extreme caution. The idea that income inequality in itself may be bad for health is congruent with other arguments that the large and growing income inequality in the United States and other countries should be reversed. For those who argue that socioeconomic position be considered a ‘fundamental cause’ of health (Link and Phelan 1995), the idea that income inequality at the country and community levels may be important to health helps draw attention to socioeconomic position as an important force in itself at the macrolevel. Discussions about the importance of macrolevel income inequality also resonate with those who have been grappling with the idea that there are gradient effects on health at the more microlevel. Perhaps there is something similar between the processes that create gradient effects of socioeconomic position on health at the microlevel and processes linking income inequality to health at the macrolevel. Finally, this research suggests some new ways of thinking about inequality and how it might work to affect health. In particular, Wilkinson (1996) and others (Kawachi et al. 1997) suggested that income inequality may work to affect health through mechanisms of social cohesion and social trust. Wilkinson (1996) points out that those countries with the least income inequality are the most socially cohesive countries, and suggests that it is this social cohesion that may affect health. The melding of two resurgent research traditions – that on socioeconomic inequalities in health with that on social cohesion and social capital – promises intellectual challenges as well as potentially useful and novel policy implications.
Yet caution must also be used in interpreting these new findings on the aggregate relation of income inequality to health between countries and between subunits (i.e., states, counties, and wards) within countries. There are many potential explanations for this empirical finding, with both competing and complementary implications for how to improve health. Some recent research has explored the potential role of aggregate measures of social capital, such as social cohesion and trust, as mediators in the relationship between socioeconomic inequality and health between countries and between subunits (i.e., states, counties, and wards) within countries (Kawachi and Kennedy 1997; Kawachi et al. 1997; Wilkinson 1996). We will argue that this type of argument is neither logically necessary to make sense of the relationship between socioeconomic inequality and health at the aggregate level, nor logically or empirically consistent with and related to known empirical relationships at the individual level. We will first briefly describe some of the potentials and pitfalls (both theoretical and methodological) of current research investigating the role of social cohesion or capital in explaining the relationship between aggregate measures of socioeconomic inequality and health. Then we will suggest an alternative theoretical argument that might also explain the relationship between aggregate measures of socioeconomic inequality and health by integrating what we know at both the individual and aggregate levels about the relationship between income (and perhaps other dimensions of socioeconomic position) and health.
Socioeconomic Inequality, Social Capital, and Health
The basic argument of Wilkinson (1996) and others (Kawachi and Kennedy 1997; Kawachi et al., 1997) is that income inequality somehow affects population health via a variable at the societal or aggregate level – social cohesion and trust. Wilkinson, however, has never directly measured this variable or assessed its empirical relationship to health. Kawachi et al. (1997) measured trust at the level of the United States via the mean levels of trust reported by residents of those states represented in the General Social Survey. They show that adjusting for this variable weakens or eliminates the relationship across these same states between income inequality and population life expectancy, consistent with both their theories and those of Wilkinson.
There are ambiguities, however, about the interpretation of Kawachi et al. (1997) and others that growing income inequality leads to a lack of trust in people, which then affects population mortality. Lacking longitudinal data, they concede that lack of trust in people may also lead to income inequalities that then affect population mortality, or even that unmeasured societal attitudes or characteristics affect both lack of trust in people and tolerance of income inequality (Kawachi et al. 1997). However, even beyond these problems in testing causation, these interpretations suffer from more important problems of failing to theoretically or empirically link aggregate properties of communities to the experiences of individuals. How does inequality at the aggregate level actually affect attitudes of trust at the individual and aggregate levels, and how do these attitudes actually impact the health of individuals? The complex multilevel approach necessary to answer these questions has been missing from both the theoretical and methodological analyses.
In fact, without controlling for individual-level socioeconomic position in a multilevel analysis, it is not clear that socioeconomic inequality at the country or community levels actually has an independent effect on the health of individuals. Fiscella and Franks (1997) found that although community income inequality (at approximately the county level) relates to individual-level mortality in the United States (just as it does to aggregate life expectancy), once family income is controlled, the relation between community income inequality and individual mortality becomes minimal and nonsignificant. Because Fiscella and Franks derive their aggregate inequality measures from the data on their survey respondents, rather than an independent (e.g., census) source, their results have been criticized as overly conservative (e.g., Soobader and LeClere, forthcoming; Waitzman and Smith 1998b). These critics and others (e.g., Daley et al. 1998) are increasingly demonstrating effects of socioeconomic inequality at the level of counties or metropolitan areas on morbidity and mortality in multilevel analyses with appropriate adjustment for individual or household income. However, even these studies indicate that the substantial majority of the impact of aggregate income equality on individual morbidity and mortality, and hence population life expectancy, operates through individual income, not via some independent effect of aggregate-level inequality or derivatives/correlates of it such as social cohesion or trust.
Integrating Individual- and Community-Level Research on Income Inequality
Arguments linking country or community socio-economic inequality to health through mechanisms of social cohesion or trust virtually ignore how country or community socioeconomic inequality may relate to individual- or family-level socioeconomic position to produce health outcomes. Yet, it is the very link between socio-economic inequality at these aggregate and individual levels that helps to explain how and why socioeconomic position relates to health at both levels.
A number of authors, beginning with Preston (1975) and especially Rodgers (1979) and most recently Gravelle (1998), demonstrate that the relationship between country or community income inequality and health is necessarily implied by the curvilinear relationship between income and health seen at the individual level. Countries or communities with higher aggregate income inequality will always have worse aggregate health than communities or countries with lower aggregate income inequality, even if they have the same average aggregate levels of income. This effect occurs because an increase in community income inequality will always disproportionately hurt the health of the poor more than it will benefit the health of the rich, which is because there is a greater impact of income on health at lower levels of individual- or family-level income. Gravelle (1998) and others (Fiscella and Franks, 1997) argue that it is the relationship between individual- or family-level income and health that determines the relationship between country or community income inequality and health. In essence, the relationship between country or community income inequality and health is simply an artifact of individual-level processes.
We find this statistical argument compelling, and believe that it and available data suggest that the relationship between country or community income inequality and health is primarily due to the curvilinear relationship between socioeconomic position and health at the individual level, rather than to effects of aggregate measures of social capital, which do not operate through individual socioeconomic position. However, we do not agree with Gravelle that this means the impact of country or community income inequality should be seen as simply a statistical artifact. Nor do we conclude that the relationship between country or community income inequality and health is necessarily totally explained by relationships at the individual level.
First, rather than seeing the relationship between country (or community) income inequality and health simply as an artifact of relationships at the individual level, as Gravelle implies, aggregate income inequality may instead be seen as the major macroeconomic force driving the levels and distribution of individual income, which then more directly affect health. Given the curvilinear relationship between income and health, any reduction in community-level income inequality that raises income levels of the poor will improve the health of both the poor and the total population.
Second, existing theory and data both suggest that characteristics of communities or societies, including both their average income and level of income inequality, have an effect on individual and population health net of individual or household socioeconomic position, although the bulk of the effects of these community or societal income levels or inequalities must and do operate via individual and household socio-economic positions. However, the other mechanisms through which community- or societal-level socioeconomic characteristics affect health remain to be elucidated, both theoretically and empirically.
We agree with Lynch and Kaplan (1997) that there are, in fact, two rather different variants of the ‘social capital’ hypothesis linking income inequality to health. Wilkinson (1996) and Kawachi and Kennedy (1997) espouse one based on the psychological perceptions and feelings of individuals in response to collective levels of ‘social cohesion’ or ‘trust.’ Alternatively, Kaplan and Lynch and their colleagues (Kaplan et al. 1996, Lynch et al. 1998) and others (Davey Smith 1996) suggest that income inequality is associated with and shapes levels of public investment in education, health care, housing, transportation, public safety, environmental quality, and other human and social capital. These more tangible forms of social capital then impact the health of individuals, independent of their socioeconomic position, although probably most importantly among persons of lower socioeconomic position. Kaplan et al. (1996) show that income inequality correlates across states with many such indicators of tangible social capital. Such tangible social capital seems to us a more plausible and likely explanation of the effects of income inequality (not mediated via individual or household socioeconomic position) than the somewhat ‘miasmalike’ constructs of social cohesion and trust. Such cohesion and trust may be necessary conditions for public actions to moderate or reduce income inequality and to invest in tangible social capital.
Comparisons with and Contributions to Studies on Individual- and Family-Level Socioeconomic Position
Research on country and community income inequality and health can increasingly be integrated with research at the individual level, but it also shares many of the same problems. Issues of causal direction arise just as they do with individual-level research. Rather than looking at how aggregate socioeconomic characteristics impact health, economists have emphasized the effects of health as a form of human capital on macroeconomic growth and performance (Fogel 1991; Fuchs 1983). Issues of causation among community-level measures of socioeconomic position, income inequality, and health need to be explored, as well as causal pathways linking community- and individual-level processes. In addition, research on community socioeconomic inequality has focused to date on income inequality with little attention to whether aggregate inequality in education, assets, etc, impact health beyond their average levels within and between countries and communities.
As already indicated, the curvilinear (vs. linear gradient) nature of the relationship between income and health is important in making sense of socioeconomic inequalities in health at the aggregate as well as individual levels, and relations among them. In addition, much more analysis is needed on the impact of age, race, and gender on socioeconomic inequalities in health at aggregate as well as individual levels, and on the relations between them (see discussion of sex, race, and gender variation in the effects of community characteristics on health). In particular, we need to consider how racial inequality, in general, and racial residential segregation, in particular, affects and is affected by the relationship between community income inequality and health. In cross-national studies, those countries with a greater proportion of racial minorities are the same countries with more income inequality. In the United States, those states with a higher proportion of racial minorities are the same states with more income inequality (Kaplan et al. 1996). Conceptualizing and testing the complex relationships between race, racial residential segregation, community-levels and distribution of socioeconomic position, individual-level socioeconomic position, and health will certainly be important, particularly in the United States where it is clear that race and socioeconomic position are related in complex ways (Williams and Collins 1995).
Summary and Implications for Research and Policy
The last two decades have seen a virtual explosion of research on socioeconomic differences in health (Kaplan and Lynch 1997), with all results demonstrating that socioeconomic inequalities in health are large, pervasive, and persisting even in the face of major improvements in overall levels of population health and attendant improvements in the quality and availability of modern medical care. At this point, socioeconomic and related racial and gender differences in health are arguably the major public health problem of many developed societies (and developing ones as well). The upper socioeconomic strata of these societies are increasingly experiencing levels of life expectancy and health over the life course that approximate Fries’ (1980, 1984) Utopian scenario of the compression of morbidity, and approach the biological limits of human longevity and health. Such limits may be gradually extendable, but the greatest opportunity for improving population health in these societies lies in bringing the life expectancy and life course trajectory of health of the lower socioeconomic strata closer to that of the upper socioeconomic strata, trends that seem to be occurring in those societies with the highest levels of population health (e.g., Sweden and Japan).
If we are to reduce socioeconomic inequalities in health and hence improve overall population health, we need to better understand the forces that generate and explain the existence and persistence of these inequalities. Many suggest that this implies better understanding of the mechanisms or pathways through which the socioeconomic position of individuals comes to affect their health. This is an important scientific objective, but one on which we believe substantial progress has been made, although more remains to be done.
This chapter suggests that research also needs to move in another more upstream direction: a direction of understanding how health is affected not only by a person’s own socioeconomic position, but also by the level and distribution of socioeconomic variables in the communities, states/provinces, and nations within which individuals and families live and work. The research must also address how these aggregate-level indicators of socioeconomic position combine with individual- and family-level socioeconomic indicators to affect health. At this point, however, what we have are often parallel analyses at the individual and aggregate levels rather than the multilevel research that is necessary for understanding how and why the socioeconomic characteristics of individuals, families, communities, states/provinces, and nations are so profoundly related to their health.
Such complex multilevel analyses require not only more theoretical development and clarity in proposing models and hypotheses, but also more methodological sophistication as well. First, it will be necessary to find new ways of combining individual-level information about socioeconomic position and health with community-level information about socioeconomic level, socioeconomic inequality, and other community-level characteristics. Geocoding large data sets to combine with census data is one way of accomplishing this, although we will also ultimately want more detailed information about community characteristics that cannot be obtained from the census (e.g., availability of transportation, physical environment quality, etc.). We will also want to reconsider how we conceptualize ‘communities,’ both in terms of how we measure them (e.g., census tract vs. self-reported community boundaries) and of how we expect different processes to occur at different levels (e.g., individual, family, group, community, county, state, or national levels). We will want to find or collect multilevel longitudinal data that can track an individual’s movement in and out of different communities, changes in community profiles over time, and changes in individual-level socioeconomic position, health risk factors, and health status. Furthermore, when analyzing multilevel data, we need to use appropriate statistical techniques and software that take into account the multilevel nature of the data (e.g., Hierarchical Linear Models, Bryk and Raudenbush, 1992).
Implications for Policy
Asking questions about the potential impact of community socioeconomic conditions on the health of individuals does not necessarily mean that resulting policy implications will or should focus on community-level interventions. Rather, studying the potential impact of these aggregate socioeconomic conditions should force us to consider more closely which levels of intervention – at the individual, family, community, county, state, or national levels – might best achieve our goals of improving and maintaining health.
For example, although research has found an independent association between community socioeconomic conditions and health and mortality over and above the effects of individual-and family-level socioeconomic position, this research nevertheless suggests that individual and family-level socioeconomic position is still more strongly linked to health and mortality than are community-level socioeconomic characteristics (Robert 1998). Therefore, directing interventions to lower socioeconomic communities would ignore the many people with lower socioeconomic position who live in higher socio-economic communities (Berk et al. 1991). On the other hand, directing interventions to lower socioeconomic communities might be both an efficient way of reaching many people with low socioeconomic position, and necessary to alleviate the particular detrimental health effects of living in a lower socioeconomic community. Studying the multilevel effects of socioeconomic position on health should encourage us to think about complementary intervention strategies at different levels.
The evidence of the deleterious impact of country and state income inequality on population health indicates that socioeconomic forces at those levels drive the levels and distribution of income and other socioeconomic resources at the level of families and individuals. There is increasing consensus that improving population health requires reducing socioeconomic inequality. It is important to recognize that it is not inequality per se that is the primary culprit here, but rather the greater absolute and relative deprivation of lower socioeconomic strata in more unequal societies (e.g., the United Kingdom and the United States) versus societies that are less unequal (e.g., Sweden or Japan). Direct comparison of socioeconomic differences in infant mortality and adult health for the United Kingdom and Sweden show reduced socioeconomic inequalities in infant mortality in Sweden, produced primarily by the better health of the lower socioeconomic strata in those societies (Vágeró and Lundberg 1989). Although socioeconomic inequality matters to health, reducing income inequality solely by reducing the income levels of the richest members of society is not likely to improve individual or population health. However, reducing income inequality by increasing the income levels of the poorest members of society is likely to improve individual and population health.
Thus, reducing the absolute and relative deprivation of the bottom 25-50 per cent of the socioeconomic hierarchy in societies such as the United States, rather than reducing inequality per se, is the policy goal. We know of policy mechanisms for doing this (e.g., earned income tax credits, adequate minimum wage levels, full employment policies, and adequate support systems for those, especially women and children, not employable), if we have the will to apply them (Ellwood, 1988). These are likely to require reductions in income inequality, although the extent of such reductions depends on the overall level of economic and income growth in a society or region. Both government (e.g., welfare reform) and market forces are constantly producing policy changes affecting income inequality and the absolute and relative socioeconomic position of the less ‘well off.’ We must do more to evaluate the effects of these policy changes on health as well as other outcomes.
There will probably always be a residual socioeconomic gradient in health in all societies, but the magnitude of it can and should be moderated if the United States and other societies that have mediocre and worsening levels of population health relative to other developed societies are to achieve levels of population health commensurate with their overall economic level. Health policy alone cannot now, nor could it ever, solve our problems of population health. Socioeconomic policy is equally or more important, and should be evaluated in terms of its consequences for health as well as other desirable goals.