NANCY KRIEGER Our future survival is predicated our ability to within equality.
Inequality hurts. Discrimination harms embodiment in population patterns of health,
health. These seems like straightforward, even disease, and well-being. self-evident, statements. They are propositions that can test, just like any other proposition about health that we investigate. Yet epidemiologic research explicitly foe cused discrimination as a determinant of population health is in its infancy. At issue are both economic consequences of discrimination and accumulated insults arising from everyday and at times violent experiences of tying treated as a second-class citizen, at each and every economic level. In asking whether discrimination harms health, this new work builds on a century and a half of research demonstrating that racial/ethnic economic dtspanties often but not always explain- U.S. racial/ ethnic inequalities in health (DuBois 1906; Tibbitts 1937; Krieger 1987; Krieger et al. 1993; Williams and Collins 1995; LillieBlanton et al. 1996). And it extends this work to address health consequences Of other tyrrs of discrimination. based on gender, sexuality, disability. and age (Table 3-1). Testing the hypothesis that discrimination harms health requires clear concepts, measures, and methods. This chapter will accordingly review definitions and patterns Of discrimination within the United States, evaluate analytic strategies and instruments researchers have developed to study health effects of different kinds of discrimination, and conclude by delineating diverse path. ways by which discrimination can harm health, both outright and by distorting pro. duction of epidemiologic knowledge about determinants Of population health. Although the examples primarily are U.S., based and pertain chiefly to racial discrimination and physical health, the broader issues raised should be relevant to other countries, to Other types Of discrimination, to mental health, and to overall well-being. Throughout, the framework I use to conceptualizc and orrrationalizr relationships between discrimination, inequality. and health is theory (Kneger 1994). Taking literally the notion Of “embodiment,” this theory asks how we incorporate biologically — from conception to death - our social experiences and express this Bringing the metaphor of the body what drives patterns of health, ease, and well- politic to life—a body “ruled" by a “head" being so as to produce knowledge useful for and sustained by laboring —hands," a body guiding policies and actions to reduce strial that creates, consumes, excretes, reproduces. inequalities in health and promote social and evolves— this theory draws attention to well-being. why and how uxietal conditions daily produce population distributions of health. DISCRIMINATION IN THE UNITED Critical causal components conjointly STATES: DEFINITIONS AND PATTERNS include: societal arrangements of power and property and contingent patterns of Definitions Of Discrimination production and con. sumption and (2) constraints and possibilities of our biology, According to the Oxford English as shavrd by our species' evolutionary Dictionary, the word "discriminate" derives history. Our ecologic context, and individual from the Latin term discriminare, which trajectories of biological and social means "to divide, separate, distinguish" development. These factors together (1971, p. 746). From this standpoint, Structure inequalities in exposure and discrimination" simply means "a distinction susceptibility to and also options for resisting (made with the mind, or in action)." Yet, pathogenic insults and processes across the when people are involved, as both agents and life course (Krieger 1994; Kuh and Ben- objects Of discrimination, di«rimination Shlomo 1997). Ecosocial theory thus that takes on a new meaning: —to discriminate how we develop, grow, age, ail. and die against" is "to make an adverse distinction necessarily reflects a constant interplay, with regard to; to distinguish unfavorably within our bodies, of our intertwined and from Others" (p. 746). In Other words, when inseparable social and biological history. people discriminate against each Other, more Three additional assumptions, relevant to this than simple distinctions are at issue. Instead, chapter, are that we, as human beings, desire those who discriminate restrict, by judgment and are capable of living fully expressed lives and action, the lives Of those whom they replete with dignity and love. that di«rirninate against. epidemiologists are motwated to reduce The invidious meanings of discrimination human suffering, and that social jusrice is the readily apparent in the legal realm, where foundation of public health. people have created and enforce laws both to Before considering how to uphold and to challenge discrimination. conceptualize, measure, and quantify health Legally, discrimination can be of two forms. consequences of discrimination, one caveat One is de jure. meaning mandated by law; the immediately is in order: The purpose of Other is de facto, without legal basis but studying health effects of dixriminanon is not sanctioned by Custom or practice. Examples to prove that oppression is bad because it of de iure dis. crimination include Jim Crow harms health. Unjustly denying people fair laws. now overturned, that denied African treatment abrogating human rights and Americans access to facilities and services constraining possibilities for living fully used by white Americans (Jaynes and expressed, digrufied, and loving lives is, by Williams 1989. pp. 57—111) and Current definition, wrong (United Nations General laws prohibiting gay and lesbian marriage Assembly 1948; of effects on health. (Vaid 1995). By contrast, Rather, the rationale for studying health of underrepresentation of people of color and discrimination is to enable full accounting of white women in clinical trials constitutes a form of de facto discrimination (Sechzer et a socially defined group is, or are, treated al. 1994; King 1996). differently (especially unfairly) because of Whether de jure or de facto. his/her/their membership of that group" (Jary discrimination can be perpetrated by a and Jary 1995, p. 169). Extending this diverse array of actors. These include the definition, the Concise Ox. ford Dictionary state and its institutions (ranging from law ofSociology holds that discrimination courts to public schools), nonstate institutes involves not only "socially derived beliefs (e.g., private sector employers, private each group holds about the other" but also schools, religious organization), and "patterns of dominance and oppression, indviduals. From a legal or human rights viewed as expressions of a struggle for power perspective, however, it is the state that and privilege" (Marshall 1994, pp. 12.5— possesses critical agency and estabilished the 126). In other words, random acts of unfair context-whether pcmissive or prohibitive for treatment do not constitute discrimination. discriminatory act can enforce, enable, or Instead, discrimination is a serially structured condone discrimination, or, alternatively, it and sanctioned phenomenon, justified by can outlaw and seek ro redress its effects ideology and ex. pressed in interactions, (Table J-2) (Tomascvski 1993). A powerful among and between individuals and example of the latter is the new pose- institutions, intended to maintain privileges apartheid Sou th African constitution (de Vos for members of dominant groups at the cost 1997). This document mandates, in the most of deprivation for others. inclusive language of any na nona l Although sharing a common thread of conscirution in the world, that "The stare may systemic unfair treatment, discrimination not unfai rly discriminate directly or nevertheless can vary in form and type, de. indirectly against anyone on one or more pending on how it is expressed, by whom, grounds, including race, gender,sex, and against whom. As summarized in Table pregnancy, marital status, ethnic or social 3—3, forms identified by social sci. entists origin, colour, sexual orientation, age, include: legal. illegal, overt (or blatant), and disability, religion, conscience, belief, Covert (or subtle) discrimination, and also culture, language and birth"; discrimination instltutional (or organizational), structural (or by individuals on these terms is likewise systemic), and interpersonal (Or indnnduat) prohibited. discrimination (Benokratis and Feagin 1986; Despite its legal dimensions, Rothenberg 1988; Feagin 1989; Essed 1992). however, discrimination is never simply a Although usage of these terms varies, legal affair. Conceptualized more broadly, it institutional discrimination typically refers to refers to all means of expressing and discriminatory policies or practices carried institutionalizing social relationships of Out by state or nonstate institutions; dominance and oppression. At issue arc structural discnmination refers to the totality practices of dominant groups to maintain of ways in which societies foster privileges they accrue through subordinating discrimination; and interpersonal the groups they oppress and ideologies they discrimination refers to directly perceived use to justify these practices; these ideologies discnminatory interactions between revolve around notions of innate superiority individuals—whether in their institutional and inferiority, difference, or roles (e.g., employer/employee) or as public deviance. Thus, the Collins Dictionary of or private individuals (e.g., Sociology defines discrimination as the shopkeeper/shopper). In all cases, process by which a member, or numbers, Of perpetrators of discrimination act unfairly towards members Of socially de. fined this experience of multiple subordination subordinate groups to reinforce relations Of cannot simply reduced to the "sum" Of each dominance and sulx»rdination, thereby type. Recent U.S. scholarship on gendered bolstering privileges conferred to them as Of racism. for example, has begun to examine a dominant group. how, in a context of overall neganve stereotypical portrayals of black Americans Patterns Of Discrimination as lazy and unintelligent (Schuman et al. A full accounting of discrimination in 1985; Kinder and Mendelberg 1995), black the United States today is beyond the Kope women as black women are stereotyped, as Of this chapter. Instead, to provide a Patricia Collins has observed, as "mammies, reminder of its ubiquity as well as matriarchs, welfare recipients and hot background to considering how it can harm mammas" (1990, p. 67), while black men as health, I next review, briefly. five notable black men are stereotype as criminals and ways that discriminanon can permeate rapists (Rothenlk•rg 1988; Collins 1990; peoples lives. Essed 1992). Understanding discrimination First, as summanzed in Table 3 —l. experienced by black women and men thus many groups experience discrimination in requires considering the salience Of both the United States at present. Dominant types their race/ethnicity and gender. Third, singly of discrimination are based on race or or combined, different types Of ethnicity", gender. sexuality (including discrimmation can occur in just about every sexual orientation and identity), disability, facet of public and private life (Table 3—3). age, and, although not always recognized as The full gamut extends from the grinding such. social class (Rothenberg 1988; daily realities of what Philomena Essed has Jackman 1994; Essed 1996; vaid 1995; Gill termed "everyday" discrimination ( 1992) to 1996; Minkler and Estes 1991; Sennett and the less common yet terrifying and life- Cobb 1972). Other types, more pronounced transforming events, as being victim of a hate in the past. include discrimination based On crime (Pierce 1995). religion and nationality (U.S. Equal Employ. In a typical day exrxriences with ment and Opportumty Commission 1992). di«rim• ination accordingly can start These latter types are still highly relevant for depending on type—in the morning, at home, American Indians and other indigenous continue with public encounters en route to people in the United States. for whom many or while at or work or when shopping or governmental policies (e.g„ restrictions on eating at a restaurant or attending a public religious expression, abrogation of treaty event, and extend through the evening, rights, removal of children to non-Indian whether in the news or entertainment or families) have Iren gencxidal in effect, if not while engaging with family members intent (norton 1987). (Rothenberg 1988; Jaynes and Wilhams Second, as explicitly recognized by 1989; Feagin 1989; Feagin and Sikes 1994; the South African constitution, people often Esgd 1992; Sennett and Cobb 1972; Jackman can experience multiple forms of 1994; Gardner 1995; Vaid 1995; Minkler and discrimination. Whereas white women may Estes 1991; Gill 1996). Other Common but be sublect, as women, to gender not typically daily scenarios for experiencing discrunination, women of color—whether McKtnlay 1996; Greiger 1996; Gill 1996). or black, Latina, Asian or Pacific Islander. or interacting with the public agencies the American Indian—may be subject to both legal System (Rothenberg 1988; Jayne and gender and racial discrimination. Moreover. wilams 1989; Feagin 1989). Fourth, while Of discrimination may interpersonal and negative attitudes. Strikingly, then, data from obvious, they are also likely to institutional the 1990 General Social Survey reveal that and invisible. To know, for example, that you fully 75 % of white Americans agree that — have been discriminated against in your black and Hispanic people are more likely salary, or that you have been denied a than whites to prefer living on welfare and a mortgage, or an apartment, or been steered majority concur that —black and Hispanic away from certain when you are looking for people are more likely than whites to lazy, a home, requires knowing how the employer violence-prone, less intelligent, and less bank, landlord, or real estate agent treats patriotic" (Associated Press 1991; Kinder indivduals (Fix and Stryck 1993; Feagin and and Mendelberg 1995). These are ugly social Stryck 1993; Feagin and Sikes 1994; Essed). facts, with profound implications for not only Typically, it is only when people file our body politic but also the very bodies in charges of discrimination in court that which we live, love, rejoice, suffer, and die. evidence of such patterns of inequality can be obtained. Other clues can obtained by MEASURING DISCRIMINATION examining social pattermng of economic in. TO ESTIMATE ITS EFFECTS equality, since acts of discrimination whether ON POPULATION HEALTH institutional or interpersonal, blatant or covert—usually harm economic as well as How, then, can epidemiologists study social well-being. Table 3-4 illustrates this discrimination as a determinant of population point for racial/ethnic discrimination, health? Figure 3—1 summarizes three depicting marked racial/ethnic inequalities in approaches to quantify health effects of income, wealth, education, and discrimination: (l) indirectly. by inference, at unemployment. the individual level; (2) directly, using Fifth and finally, attesting to some of measures of self-reported discrimination. at the animosity that feeds and justifies the individual level; and (3) in relation to discrimination are, to give but one example. institutional discrimination, at the population numerous surveys of US. racial attitudes level. All three approaches are informative. (Schuman et al. 1985; Jaynes and Williams complementary, and necessary. I review and 1989, pp. Kinder and Mendel. berg 1995). provide examples for each method, below. Despite declines in racial preludice over time, reported levels remain high, even Indirectly Measuring Health Effects Of taking into account that ( people underreport Discrimination, Among Individuals negative social attitudes (Schuman et al. One of the more common approaches 1985); (2) dominant groups typically deny to studying health consequences of discrimination exists, especially, as Essed discrimination is indirect. Recognizmg that has noted (1996), if it is no longer legal (see, discrimination may difficult to measure, for example, Herrnstem and Murray 1994; investigators instead compare health Thernstrom and Thernstrom 1997), and (3' as outcomes of subordinate and dominant Jackman has argued (Jackman 1994), groups (Fig. 3— la). If distributions of these paternalism combined with (a) friendly outcomes differ, then researchers determine feelings toward individual memlxrs of if observed disparities can be explained by subordinate groups and (b) denial of any “known risk factors". If so, investigators responsibility for institutional discrimination interpret their findings in light of how is as much a hallmark of contemporary discrimination may shape distribution of the discrimination as is outright conflict and relevant "risk factors." If, however,a residual difference persists, even after controlling for whose incidence increases with poverry, these other risk factors, then additional with incidence rates identical among aspects of discrimination may be inferred as African Americans and white America ns at a possible explanation for the remaining each income level. Under these disparities (assuming no unmeasured circumstances, if African Americans below confounders). the poverry l ine were much poorer than wh Exemplifying this indirect method are ite Americans below the poverty line, then U.S. studies examining whether analyses adjusting for being "above" vs. socicrcononlic factors "explain" black/white "below" poverty would fail ro explain excess inequalities in health Status (Kneger et. al rates of disease among African American - 1993; Williams and Collins 1995; even though black/white income disparities LillieBlanton et al. 1996; Lillie-Blanton and in fact fully explained black/white LaVeist 1996; Navarro 1990; US. differences in disease incidence. A second Department of Health and Human Services hypothesis, discussed in the next section, is 1991), exposure to occupational and that the remaining difference reflects health environmental health hazards (Robinson consequences of unmeasured noneconomic 1987; Brown 1995; Northridge and Shepard aspects of racial discrimination, e.g., chronic 1997), or receipt of medical services psychologic stress (Krieger et al. 1993; (Council on Ethical and Judicial Affairs Wdliams 1997a). A third explanation, 1990;Gornick et al. 1996;Geiger 1996; King unrelated to discrimination, posits that 1996; Peterson et al. 1997). In their earliest unexplained differences reflect unmeasured form, starring in the mid-1800s, these kinds factors that are asscxiated with race/ ethnicity of investigations compared health of and the specified outcome but are not related enslaved and free blacks and also poorer and to either discrimination or socioeconomic wealthier whites, there by exposing bow position, e.g., culturally shaped patterns of slavery and poverty, and not “race" per se, food consumption. Finally, a fourth largely explained the poorer health of “the explanation—often invoked but rarely tested Negro" (Krieger 1987;Smith 1859; eyburn (Cooper and David 1986; Williams et al. 1866). The basic strategy, then and now, is to 1994) speculates that innate genetic determine whether "adjusting" for differences are responsible. Whether and socioeconomic position (along with relevant how investigators address these alternative confounders) eliminates observed explanations, when interpreting unexplained racial/ethnic disparities in the specified differences in health Status subordinate and ourcome. If so, economic consequences of dominant groups, varies considerably across racial discrimination are inferred to underline studies. the observed (unadjused) disparities; in other Illustrating both the importance and words, both racism and class matter (Krieger ambiguity of research using indirect method. er al. 1993; Williams and Collins 1995; Ologies to study health effects of discrimi• Lillie-Blanton and LaVeist 1996; Nava rro nation is research on a well-known public 1990). health problem: black/white disparities in If, however, racial/ethnic differences risk Of low birth weight (Institute of persist, four alternative explanations can be Medicine 1985; Rowley et al. 1993). offered. One is that inadequate measurement Numerous investigations have demonstrated of socioeconomic position produces residual that poverty is associated with elevated risk confounding (Kneger et al. 1993; Kaufman et Of low birth weight among both African al. 19971. Consider, for example, a disease Americans and white Americans and also that "adjusting" for poverty substantially measures of socioeconomic position for reduces—but does not eliminate—excess woman for example, will rarely adequate for risk among African Americans (Institute Of properly detecting gradients in wornen’s Medicine 1985; Rowley et al. 1993). Even health (Krieger et al. 1999; Arber 1990). So, not only is risk of low birth weight IS to Moreover, as illustrated by a study which two times higher among African American found that childhood but adult rneasures of compared to white and Hispanic infants born socioeconomic account for adult racial/ethnic to poor or less educated parents (Rowley et disparities in infection by Helicobacter pylori al. 1993; National Center for Health Statistics (Malaty and Graham 1994) presumably 1997, p. 90), but it is also two times higher because most infection occurs in childhood – comparing black to white infants born of socioeconomic position should be rneasured college-educated parents (Schoendorf et al. at relevant across the life span, in relation to 1992). even after controlling for numerous both acute exposures and cumulative covariates. Although additional disadvantage (Krieger et al. 1997; Kuh and noneconomic and economic dimensions of Ben-Sholomo 1997). For guidance on racial discrimination could account for these measuring socioeconomic position in findings, so could other unmeasured epidemiologic studies, overall and with determinants or confounders. Absent data on respect to time penod and level of these unmeasured factors. discrimination can measurement, as well race/ethnicity, gender. at inferred, not demonstrated. as a and sexual orientation, readers are determinant of health outcomes. These same encouraged to consult the cited references caveats apply to the Other mayor strand of (above) as well as chapter 2 of this book. research indirectly assessing effects of Lastly, one further indirect approach discrimination and health, which focuses on to measuring health effects of discrimination differentials in diagnosis and treatment of on individuals - albeit relevant only to racial women and men with the same symptoms or discrimination - addresses associations diseases (Council on Ethical and Judicial between skin color and health status. This Affairs 1991; McKinlay 1996). approach has been employed in 17 U.S. The importatance Of discrimination epidemiologic studies focusing on health of in restricting economic resources, coupled African Americans (Boyle et al. 1967; Boyle with evidence Of the profound impact of 1970; Harburget al. 1973, 1978; Keil et al. economic well-being health (Townsend et al. 1977, 1981, 1992; Coreshetal. 1991; NelMin 1990; Krieger et al. 1993; Evans et al. 1994; et al. 1988, 1993; Garry et al. 1989; Klag et Williams and Collins 1995; Amick et al. al. 1991; Dressier 1991a; Knapp et al. 1995; 1995; also Chapters 2, 4, and 6). accordingly Gleiberman et al. 199.5; Schwam etal. 1995; that one Strategy for reducing ambiguity and Churchill et al. 1996). Although most of improving epidemiologic research is these studies actually were attempting to use employing appropnate measures of skin color as a biological marker for genetic s.xi«rconornic position (Krieger al. 1997; admixture, several also conceptualized skin Lilberatos et al. 1988). Failing to take into color as a marker for discrimination. account such issues as level Of nEaSurement underlying presumption is that darker skin (e.g.. individual, household, neighborhood color increases risk of discrimination above region) and time period (e.g., childhood, and beyond a powerful "color line" markedly adult) can introduce bias and produce distinguishing people of color from white considerable residual confounding. Using Americans. individual level instead of household level Notably. among these 17 direct experiences of discrimination. Nor can epidemiologic Studies. 12 reported they investigate effects related to intensity, assciations (all modest) between skin color duration, or time period of exposure to and the specified outcomes (ranging from discrimination. What such studies can blood pressure to all-cause and cause - address, however, are health effects of types specific mortality) (Boyle et al. 1967; Boyle of discrimination not readily perceived by 1970; Harburg et al. 1973, 1978; Keiletal. individuals (e.g., treatment decisions of 1977, 1981; Cor. eshet al. 1991; Klag, et. al individuals' physicians), and (2) whether 1991; Dlesser et al. 1991; Knapp et al. 1995; economic disparities or other factors Gleiberman et al. 1995; Churchill etal. 1996). presurned to be related to discrimination Of these 12, the ten collecting soci€rconomic account for observed differences in health data (all but Boyle 1970 and Coresh et al. between dominant and subordinate groups. 1991) all found that socioeconomic position For these reasons. studies using indirect either typically explained or else approaches to measuring health effects of substantially the observed association. discrimination can and do provide essential, Additionally, the single published US. study powerful, and important evidence that examining associations between skin color. discrimination shapes societal distributions socioeconomic position, and self-reported of health and disease. To ask and answer the ecperiences of racial discrimination among question of how directly perceived African Americans dccumented that while discrimination affects health accordingly darker skin color was moderately associated requires a different set of questions and a with socioeconomic deprivation (among men different research strategy. only), skin color and self-reported experiences of racial discrimination were Measuring Self-Reported Experiences Of largely unrelated (Krieger et al. 1998). Other Direct Discrimination and Health Effects sociologic research similarly has shown that Among Individuals while moderate associations exist between To meet the challenge of explicitly skin color and income among both African measuring direct experiences of Americans and Mexican American (chiefly discrimination and relating this to their health among men), income disparities are far statuss a new generation of public health greater comparing African Americans or researchers is devising new methods and Mexican Americans with light skin to white approaches. Indicating the novelty of this Euro-Americans than when companng work, at the time of preparing this chapter I African Americans or Mexican Americans could identify only 20 studies in the pub. lic with dark vs. light skin (Telles and Murguia health literature employing instruments to 1990; Keith and Herring 1991). The net measure self-reported experiences of implication is that while skin color may serve discrimination (Table 3—5). Of these, 15 as a modest indirect marker for aspects of focused on racial discrimination (13 on racial discrimination, it is not a direct marker African Americans, two on Hispanics and for self-reported experiences of racial Mexican Americans), two of which discrimination. additionally addressed gender Taken together, then, existing discrimination; another solely examined research relying upon indirect strategies to gender discrtmination; three investigated measure health effects of discrimination discrimination based on sexual orientation; provides precisely this: indirect evidence. and one concerned discrimination based on They do not and cannot explicitly measure disability. I could find no published empirical studies on health effects of self-reported outcomes, within the same study population, experiences of discrimination based on age. has yet to conducted. Without such validation In Table 3-5, I summarize measures research, choice of appropriate arasures is of discrimination employed in. along with likely to remain problematic. findings of, these 20 investigations. The most Also contributing to eclectic of questions common outcome (ten studies) was mental about self-reported experiences Of health, e.g., depression, psychological discrimination is an overall dearth of distress; the second most common (five empirical studies on this topic, nor just in studies) was hypertension or pressure. public health but in research more broadly. Overall, studies consistently reported higher Often. when epidenuologtsts decide to levels of self-reported experiences of measure social phenomena to their impact on discrimination were associated with poorer health, we look to social sciences for mental health; associations with somatic guidance. Yet, neither the sociologic nor health, as discussed below, were more psycologic literature currently offers complex. wellcharacterized, “ready-to-use," validated As indicated by the diversity of instruments appropriate for large-scale questions listed in Table 3-5, public health empirical studies. Instead, most emptrwal research presently lacks a standardized sociologic studies on discrimination either methodology to measure self-reported have focused chiefly on racial attitucks of experiences of direct discrimination. Of people who discriminate. rather than particular note is variability in assessing: (1) experiences of those who have endured time period of exposure (ever vs. recently), discrimination (Schuman et al. 1985; (2) domain of such exposures (globally or in Jackman 1994), or else, as is also the case in specific situations), (3) intensity and psychological research, they have employed frequency of exposure (major events or in depth interviews and qualitative everyday of and (4) targets of discrimination approaches not readily transferable to (respondents only or also memebers of their epidemiologic research (Essed 1992; Feagin family or their group overall). Only eight and Sikes 1994; Mays 1995; Bobo et al. studies included additional questions asking 1995; Parker et al. 1995). The net effect is an respondents how much they were upwt by uncanny silence on em. pirical estimates of and how they responded to experiences of the prevalence (let alone the effects) of self- discrimination. Less than half the studies reported experiences of discrimination. even reported psychometric measures regarding as this experience is widely recognized in validity or reliability of their instruments. many other avenues of discourse. e.g.. law. At least two factors underlie political science, history, literature, film, proliferation of different measures of self- other art forms, and the media to name a few. reported experiences of and responses to Fortunately, epidemiologic principles discrimination in epidemiologic research. can nevertheless provide useful guidance in One is the recent emergence of public health rneasunng and analyzing self-reported research on this topic. Thus, investigators are experiences of discrimination and its effects only now starting to develop, employ, and on health. At issue, as many epidemiologic validate instruments appropriate for large- study, are (l) measurement of exposure, in scale epidemiologic investigations. relation to intensity, frequency, duration, and Methodologic research comparing relevant etiologic period, i.e., time between associations of diverse measure of self- exposure, onset of pathogenic processes. and reported discrimination with selected health occurrence of disease, (2) measurement of susceptibility, and (3) effect modification of acceptance, verbal confrontation. physical associations between exposures and confrontation. or legal action¯ (Feagin and outcomes by relevant covariates. In the of Sikes 1994. p. 274; see also: Lalonde and studies of discrimination and health, issues of Cameron 1994; and Taylor 1995). susceptibility notably in. elude responses to Studies listed in Table sup-Bitt the and ways of resisting discrimination. while recommendation to use specific, rather than those involving effect modification require global, questions about experiences of considering how self-reported experiences of discrimination, rather than ask about discrimination and ways of responding to experiencing, say, racial discrimination such experiences may have different overall, it is likely to be more informative to meaning or impact depending on a inquire about experiencing a specific type of respondent's position. as related to multiple discrimination in several different situations, subordination. degree of social and matenal e.g., at school, at work, on the street. Even depnvation, and historical cohort. better would asking separately about ha ving First. regarding measurement of experienced racial discrimination in work exposure, extant research suggests questions assignments, promotions, pay, layoffs, should be direct and address multiple facets interactions with co-workers, and of discrimination for each type of interactions with supervisors (Bobo et al. discrimination being studied. Conversely, 199.5; Feagin and Sikes 1994). The studies should avoid global questions about importance of considering muultiple types of experiences or awareness of discrimination - discrimination, moreover, is illustrated by whether for all types combined or even iust one study of antigay discrimination which for one type of discrimination - since global found that while white gay men reported questions are likely to underestimate chiefly antigay discrimination, white lesbians exposure and are of little use for guiding reported both antigay and gender interventions and policies to reduce discrimination, and black gay and lesbians exposure. Recognizing the importance of additionally reported racial discrimination assessing multiple domains of (Kneger and Sidney 1997); another study discrimination, the few larg-scale social notably found that lesbian and gay African science surveys investigating self-reported Americans reported higher rates of experiences of discrimination - whether depressive distress than would be predicted racial discrimination (Campbell and 1968; based on summing risk for their Kerner race/ethnicity, gender, and sexual orientation Commission 1968; Sigelman and Welch (Cochran and Mays 1994). 1991; Jackman 1994; Taylor et al. 1994), In addition to specifying domains in gender discrimination (Womerfi Bureau which different types Of discrimination 1994; Jackman 1994), or antigay occur, questions should also address extent of discrimination (Herek have asked exposure in relation to the presumed etiologic respondents questions about experiencing period. Depending on the health outcomes distinct of discrimination or unfair treatment under study, both chronic and acute in a variety of policy-relevant situations. exposures may matter, as will intensity. Multiple options for questions about duration, and frequency of exposure. Thus, in responses to discrimination and unfair the case of asthma attacks or other outcomes treatment are likewise advisable, since with sudden onsets that can be triggered by studies show reactions can span from adverse events, acute as well as cumulative "careful assessment to wlthdrawal, resigned exposure to discrimination may be relevant. By contrast, in the case of hypertension or about their lifetime experiences and fears not other conditions with gradual onset only for themselves but for their family cumulative exposure, rather than recent or members and their appraisal of risk for their acute most likely will have greatest etiologic social group more generally. These estimates relevance (Krieger and Sidney 1996). of individual and group exposure, moreover, Furthermore, just as "daily hassles” and may be influenced by period and cohort “major life episodes" often differentially effects due to historical changes in legal affect health (Cohen et al. 1995), daily wear- statuS, intensity, and domains of and-tear of everyday discrimination may discrimination. e.g., coming of age before, pose health hazards distinct from those during, or after the heyday of the Civil Rights resulting from major episodes of Movement in the 1960s. discrimination (such as losing a job) Even assuming questions adequately (Williams et al. 1997b). Designing questions address the breadth of individuals' about exposure to discrimination accordingly experiences, awareness, and fears of requires careful development of a priority discrimination, however, data on self- hypotheses about timing and intensity of reported experiences of discrimination exposure in relation to the outcome(s) under necessarily - and importantly - are inherently study. subjective. Issues of validity are thus the Additionally, adequate measurement of same as those with any epidemiologic data on exposure requires considering whether it is self-reported exposures, particularly those suffcient to ask individuals about only their about personal social experiences (Cohen et own experiences Of discrimination. Also of al. 1995). concern may be fears of experiencing In the case of discrimination, at least discrimination and their awareness of or fears four factors may contribute to individuals about discrimination directed against Other reporting different experiences of Of their family or their social group. Notably, discrimination even when subjected to the recent research on what has been termed same "exposure" (e.g., a specific act). The “personal/group discrimination disrepancy” first involves what has been termed documents that people typically report "internalized oppression”, whereby members perceiving greater discrimination directed of subordinated groups - especially those toward their group than toward themselves experiencing greater social and material (Crosby 1984; Taylor et al. 1990, 1994; and deprivation - internalize negative views of Tap lor 1995; Mays and Cochran 1997). the dominant culture and accept their possible explanations of this phenomenon subordinate status and related unfair range from overestimation of group treatment as "deserved" and hence expenences of discrimination to recognition nondiscriminatory (Fanon 1965; Krieger of patterns of discrimination not readily 1990; Kneger and Sidney 1996; Sigelman discerned by personal experience (e.g., and Welch 1991; Essed 1992; Crosby 1984; discriminatory hiring practices, as discussed Taylor et al. 1994; Feagin and Sikes 1994; earlier) to denial of personal expenences of Meyer 1995). The second concerns ways discrimination, positive coping, optimism, members of subordinate groups relate to and even illusions of invulnerability (Crosby "positive" traits— if any—attributed to them 1984; Sigelman and Welch 1991; Taylor et by dominant groups, e.g., some women may al. 1994; Feagin and Sikes 1994). Fully Interpret men lc»king them over sexually in measuring exposure to discrimination public as evidence of their own sexual accordingly may entail asking individuals attractiveness and hence self-worth, whereas other women may perceive such staring as (CARDIA ) study, a prospective multisite public harassrnent (Jackman 1994; Gardner community-based investigation established 1995). Third, people consciously or in 1985 — 1986 that enrolled slightly over unconwiously may shape answers to be 5000 young black and white women and in "socially acceptable (Schuman et al. 1985; fairly equal proportions, who were 18 to 27 Cohen et al. 1995) and may also vary in years old at baseline. Questions on racial whether they find it helpful or distressing to discrimination ineluded in the Year 7 speak about their problems (Ross and CARDIA examination are in Table 3-5. To Mirowsky 1989). And fourth, individuals analyze data on exposure to discririmination, may exaggerate expriences of discrimination I set as referent group African Americans (system-blame) to avoid blaming themselves rerxrting moderate racial discrimination, for failure (Neighbors et al. 1996). defined as reporting racial discrimination in If operative, any of these biases could one or two of seven specified situations. I potentially affect not only estimates of based this choice on the a priori logic that directly percieved discrimination but also its moderate exposure constitutes a normal impact on health. It is important to experience for people subject to racial emphasize, however, that existence of these discrimnation, and I further hypothesized – potential biaes not render epidemiologic based on prior research - that this referent research on discrimination and health group would at lower risk of elevated blood impossible or unfalslfiable, logical inference, pressure than African Amencans reporting no for example, of a study reporting comparable or extensive discrimination, (Kreiger, 1990). health status (controlling for relevant Key findings for the African confounders) among, say, women reporting American participants were that, first, no, moderate, and high levels of reported having ever experienced racial discrimination within each and every discrimination (28% in one or two, and 52 % specified sociodemographic stratum, e.g., in three or more of seven specified class, race/ ethnicity, age. sexual orientation, situations); 20%, however, reported having would be that discrimination is not causally never experienced racial discrimination. related to the health outcomes under study. Second, systolic blood pressure (SBP) was By contrast, if asqxiations were. in some independently associated with self-reported instarwes, a lose response relationship (more experiences of racial discrimination and to discrimination associated with greater risk of unfair treatment. Third, adjusting for relevant poor health), or, in others, a J-shaped curve confounders, SBP was significantly elevated (since oppression may affect meaning of a by 2 to 4 mmHg among working class men reply), the data would offer evidence of links and women and professional women between self-reported experiences of reporting substantial compared to moderate discrimination and health. discrimination, and (2) working class men The salience of these kinds of and women reporting no compared to conceptual and rnethcxiological issues for moderate discrimination; conversely (3) studying reported exrrriences of among professional men, blood pressure was discrimination in relation to health is over 4 mmHg lower among reporting no illustrated by a recent investigation I compared to discrimination. Fourth, within conducted on racial discrimination and blood economic strata, a net difference of 7 to 10 pressure (Krieger and Sidney 1996). mmHg in average SBP existed comparing Participants were memlrrs Of the Coronary extremes of expenence involving racial Artery Risk Development in Young Adults discrimination and responses to unfair treatment. Additional novel analyses, also moderate discrimination (Krieger 1990; adjusted for relevant confounders, showed Krieger and Sidney 1997). that (l) black—white differences in SBP Resolving cotweptual and would be reduced by 33% among working methodologic questions raised by emerging class wornen and by 56% among working research on self-reported discrimination and class men if SBP of all black working class health will require conducting appropriate women and men were equal to that of those validation studies. I accordingly describe reporting only moderate discrimination four complementary research strategies that (whose SBP was the same as that Of their could potentially useful, involving smaller, white working class counterparts), and (2) no indepth studies as well as larger surveys. black—white differences in SBP occurred One approach would be to employ qualitative among professional black and men reporting, interviews to assess respondents perceptions respectively, moderate and no of discrimination and to probe manings of discrimination, as compared to their white their answers to survey questions about professional counterparts. experiences of discrimination. Along these One plausible interpretation of why a lines, one small British study found that response of no compared to racial people who initially stated on the discrimination was assosiated with elevated questionnaire that they had not experienced SBP among working class African American racial discrimination later said, in subsequent women and men but lower SBP among in-depth interviews, that they had professional black men is that, as discussed experienced such discrimination but found it above, the meaning of "no" may be related to too hard—or too frightening or too pointless social position, in this case, gender and class to discuss (Parker et al. 1995). Were this (Krieger and Sidney 1996), for people with finding to replicated, and were discrepancies relatively more mywer and resources, a may between survey responses and in depth truly mean "no." By contrast, among more answers about experiencing discrimination disenfranchised persons, especially those found to greatest among those most subject subject to multiple forms of subordination or to subordination or depnvation, it would depnvation, a "no" may reflect internalized underscore the need to (1) develop more oppression. In such cases, a disjuncture sensitive approaches to eliciting information between words and somatic evidence may be on self-reported experiences of an instance of the body revealing experiences discrimination and to (2) take into account translated into pathogenic processes that effect modification, by social of observed people cannot readily articulate with words. associations between self-reported In my view, this is the interpretation that experiences of discrimination and health makes the most which takes as real the status. patterns evinced by pressure levels in relation A second strategy could build new to self-reported experiences of racial research about physiologic responses to dicrimination. The can teach us something stimuli pertaining to the types of here, together with our words. Adding discrimination being studied. Several recent plausibility to this interpretation are results of experimental studies, for example, have two additional smaller studies, both of which shown that pressure and heart rate among found higher blood pressure among members African Americans increase more quickly of groups subjected to discrimination (black upon vtewing movie scenes or imagining women, in one; white gay men, in the other) scenarios involving racist, as compared to who said that they had experienced no vs. nonracist but angry, or neutral, encounters (Armstead et al. 1989; Jones et al. 1996). 1990; Waters and Eschbach 1995; Meyers These kinds of studies could be extended by 1995). The purpose would be to examine also querying study participants about their whether these expressions of self and social- self-reported experiences of discrimination awareness modify associations between and then analyzing associations between health status and selfreported experiences of their responses to these questions and their discrimination. Notably, each of these experimentally induced physiologic constructs is distinct from and cannot be responses to witnessing or imagining reduced to “selfesteem" and "self-effcacy." discrimination. At least among African Americans, research A third investigative technique, indicates that awareness that discrimination likewise addressing how self-reports of hinders black people from getting a good discrimination might be biased by self- education or good jobs is not associated with presentational Concerns or by impaired self-esteem and is only modestly associated ability to engage in introspection (Greenwald with selfefficacy presumably because derive and Banaii 1995), would be to use implicit their selfesteem chiefly from relations with measures designed to circumvent these family and and their sense of self-efficacy biases. One such measure, recently from how much they are able to influence developed by cognitive and social their immediate conditions. even while psychologists, is the Implicit Attitude Test understanding that societal discrimination (IAT) (Greenwald et al. 1998). This test exists (Neighbors et al. 1996). involves a computer task that assesses the degree of association between two concepts, Measuring Population-Level Experiences Of based on the assumption that people take less Discrimination and Health Effects time to categorize two concepts at the same Individual-level measures of time when they are associated with each exposures and responses to direct other than when they are not. Results indicate interpersonal discrimination, however, no that white respondents more quickly matter how refined, can, by their very nature, associate typically "white" names with describe only one of several levels of positivelyvalenced words (e.g., "heaven") discrimination that affect people's lives. Also and typically "black" names with neg. potentially relevant are population-level atively-valenced words (e.g., “cancer”) a expenences of discrimination, such as result that held even among white residential segregation, and also population respondents who did not display prejudice in level expressions of empowerment, such as their explicit self-reports of racial attitudes representation in government. A small but (Greenwald et al. 1998). Such implicit growing body of research accordingly has attirude tests could be adapted to measure begun to examine whether aspects of beliefs about experiences of discrimination, discrimination that can be measured only at thereby affording a measure Of exposure less the population level themselves determine likely ro be biased by cognitive distortion population health. Thus far primarily focused than explicit self-reports (Ruggiero et al., on racial discrimination, studies employing forthcoming). this third strategy have examined A fourth approach, feasible for associations of African American morbidity largescale surveys, would be to include and mortality rates with residential questions assessing identity formation, segregation, racial/ ethnic political clout, and political consciousness, stigma, and racial attitudes (LaVeist 1992. 1993; Wallace internalized oppression (Bobo and Gilliam and Wallace 1997; Polednak 1997; Kennedy inequality. community marginalization, and et al. 1997). mortality (Wilkinson 1996; Wallace and A study on how infant postneonatal mortality Wallace 1997; see also Chapters 4 and 8). (the death rate of infants 2 to 12 months old) As in the case of studies of self- may be related to black residential reported discrimination, however, research segregation and political empower. ment on population health in relation to exemplifies this third approach to population-level measures Of discrimination quantifying health Consequences of or empower, ment is in its infancy. discrimination (LaVeist 1992). Following Potentially promising measures include prior sociological research on residential population-level indicators of social segregation (Duncan and Duncan 1955; inequality and discrimination created by the White 1986), this investigation used an index United Nations Development Programme Of dissimilarity to measure degree of (UNDP) ( 1996), none of which have residential segregation. This index ranges employed in epidemiologic studies. The from 0 to 100 and essentially measures the UNDP's gender empowerment measure, for percent of African Americans who would example, includes data pertaining to (l) have to relocate so that the ratio of blacks to "economic participation," operationalized as white in every neighborhood would bc the the percent Of women and of men in same as that for the city as a whole. Black administrative and managerial positions and political empowerment (Ikibo and Gilliam in professional and technical jobs, (2) 1990) in turn was assessed with two "political participation and decision-making measures: (1) relative black political power, power." measured as the percent of women defined as the ratio of the proportion of black and of Inen in parliamentary seats, and (3) representatives On the city council divided "power over economic resources," by the proportion of the voting age operationalized as women’s and men’s population that was black, and (2) absolute proportional share of earned income (based black political power, defined as the on the proportion of women and men in the percentage of city council who are black. economically active workforce and their This latter measures was conceptualized as average wage) (UNDP 1996, p. 108). Similar reflecting ethe level at which African- measures of economic participation and Americans are empowered to control the political empowerment could be developed political and policy-making apparatus of the for other subordinate groups, e.g„ the lesbian city" (LaVeist 1992, p. 1084). Analyses and gay or disabled populations. Also likely showed increased risk of black neonatal to informative, though not yet incorporated in mortality was independently associated with epidemiologic studies, are measures of (l) higher levels Of segregation. poverty, and economic segregation of neighborhoods lower levels of relative (but not absolute) Jargorskwy 1996); (2) occupational black political power even when controlling segregation of jobs by gender and for intracity allocation of municipal race/ethnicity (Jaynes and Williams 1989; resources (e.g„ per capita spending, by Rothenberg 1988); (3) voter registration and neighborhood, on health, police, fires, streets, voting rates of subordinate and dominant and sewers). One implication is that groups; and (4) sociodemographic community organization, in addition to other composition of additional branches of community con. ditions, may affect government, e.g„ the judiciary. population health, a finding likewise A related strategy—also not yet suggested by recent research on income employed in epidemiologicc research— would be to examine population health in Any studies investigating relation to government ratification and associations between population-level enforcement Of diverse human rights nrasures Of determinants and outcomes. instruments, ineluding existence and however, must address two concerns, enforcement Of national laws prohibiting regarding: (1) etiologic period and (2) discrimination (e.g., in the United States, the ecologic fallacy. In the case of etiologic Civil Rights Act and the Americans with period, at issue - as in the case of studies Disability Act) Table3— 2). For example, the using individual-level measures Of United States has ratified the International distinctions between acute and cumulative Convenant on Civil and Political Rights ( exposures and between with short and longer 1966) and the International Convention on latency periods. Thus, from a temporal stand the Elimination of All Forms of Racial point, an association of higher levels Of Discrimination ( 1965). but not the Universal residential segregation or negative racial Declaration of Human Rights (United attitudes say, concurrent infant mortality Nations General Assembly 1948), the rates or childhood morbidity rates or International Convenant on Economic, homicide rates would provide more Social and Cultural Rights (1966), the compelling evidence Of health effects of Convention on the Rights of the Child segregation or racial attitudes than would its (1989). nor the Convention on the association with mortality among adults. Elimination of All Forms of Discrimination given the much longer latency periOd for against Women (1979HUNDP 1996, p. 216). most caugs Of death (e.g„ cardiovawular Any or all of these human rights instruments disease, cancer). If, however current levels of could provide important benchmarks for segregation reflected past levels and little assessing how discrimination relates to bias were introduced by residential mobility. violation of these internationally stipulated then inferences about links between rights affects population health. From a segregation and adult mortality rates could policy perspective, this could be particularly warranted. Comparable caveats about useful. since popular movements and temporal plausibility have been raised fror professional organizations can hold studies examining current levels of income governments, and sometimes even nonstate inequality in relation to adult mortality rates: actors, accountable for stipulations in these These associations make only if current human rights instruments (Tomasevski income inequality is a marker for systematic 1993). Epidemiologic research, for example, underinvestment in human resources over could analyze rates of domestic violence time (Davey Smith 1996). against wornen in relation to state funding for The concern regarding ecologic police training on domestic violence (a type fallacy centers on whether causal inferences of spending called for by the Convention on at the population level are valid at the the Elimination of All Forms of individual level. As discussed also in Discrimination Against Women) or Chapters 14 and I5 of this book, ecologic racial/ethnic disparities in infant mortality in fallacy chiefly resuits from confounding relation to public expenditures to improve Introduced through the grouping variable race relations (a type of spending called for (e.g., census tract, city. state, nation) to by the lnternational Convention on define the grouplevel dependent and Elimination Of all Forms of Racial independent variables (Robinson 1950; Discrimination). Alker 1969). In the classic case, reported by W.S. Robinson in 1950 (Robinson 1950), although state-level data showed strong 1994). Using such methods, U.S. associations between high illiteracy rates and epidemiologic studies have begun to show the proportion Of states' population that was that health profiles of, say, poor people who black (Pearson correlation coefficient = live in neighboroods generally are worse than 0.946), within these states the relationship those of equally poor people who live in more between illiteracy and race/ ethinicity was affluent neighborhoods (Haan et al. 1987; much weaker (Pearson correlation coefficient Diez Roux al. 1997; see also 2, 14, and 15). = 0,203). Residential or community political A subsequent critique of Robinson’s empowerment could likewise conceivably analyses demonstrated that grouping by state modify experiences, perceptions, and effects added an important confounding variable: of—as well as responses to—individually state level of spending on public education reported experiences of discrimination. The (Langbein and Lichtman 1978). Because study design of contextual analysis, however, southern states—the ones with relatively has yet to used in epidemiologic research on high proportions of black a low tax base and health effects of relatively less public education, illiteracy in Discrimination. these states was also high among their white residents. Had Robinson taken into account HOW COULD DISCRIMINATION HARM state per capita spending on education, a HEALTH? phenomenon that can only measured at the Prompting development of the kinds group level, not only would the computed of research strategies I have been describing ecologic correlations have less affected by is the persistent question: Why does health bias but the study also would have how gate status differ among subordinant and funding for education determines literacy dominant groups? More than methodology, rates. In other words, had Robinson used however, is required to conduct valid and relevant population-level data, his study informative analysis of health consequences would have avoided what has been termed of discrimination. Equally vital is systematic the individualistic fallacy” : erroneous and explicit consideration of ways that inferences about explanations of patterns discrimination can harm health. Theory individual level because they only upon matters. At issue is comprehending not only individual-level data (Alker 1969; Krieger et direct health consequences of discrimination al. 1993). that we embody but also how discrimination In addition to highlighting the can harm our very ability to understand—and importance of population-level determinants provide knowledge for effectively of outcomes measured among individuals, intervening public's health. the critique of Robinson’s study implies that population-level measures of discrimination Pathways to Embodying Discrimination could perhaps be meaningfully combined From an standpoint, one useful with individual-level rneasures to yield even concept for understanding links between more informative analyses of health discrimination and health is "biological consequences of discrimination (Krieger et express.ons of discrimination.- to extend a al. 1993; Williams 1997a). terminology I developed with Sally Zierler to Methodologically, this approach entails use discuss connections between gender and of contextual or multilevel analyses, a health. We defined biological expressions of technique first developed in the social gender (including gender discrimination) to sciences (Blam 1984; DiPriete and Forristal mean “incorporation of social experiences of gender into the body and expressed race/ethnicity, examples of racialized biologically, in ways that may or may as components of our biology include skin sociated with biological sex" (Krieger and color, hair type, and facial features, and also Zierler 1995). One example would how girls' such disorders as sickle cell anemia, cystic and women's body build and exercise patterns are affected by underfunding of girls' fibrosis, and Tay-Sachs syndrome. Rather athletic programs. By the same logic, than being conceptualized as particular biological expressions of racial aspects of human diversity, with varying discrimination (or race relations, more distribunons among populations - broadly) refer to how people literally embody distributions notably shaped by geography, and biologically express of racial oppression conquest, and laws about who can have and resistance, from conception to death, children with whom - these traits instead thereby producing racial/ethnic disparities in morbidity and mortality across a wide typically are construed, tautologically, as spectrum of outcomes (Krieger 1998). evidence of "racial types" (Krieger 1998). Similar terminology could be used to discuss Particular biologic characteristics biological expressions of other types of accordingly tycome imbued with of “race," discrimination, whether on sexual identity or conjuring up notions of fundamental orientation, age, disability, social class, or difference on a whole host of other other characteristics. For each type of characteristics, even though within-group discrimination, a key a priori assumption is that disparate social and economic conditions differences far exceed those between groups of subordinate and dominant groups will (King 1981 ; Lewontin 1982; and David produce differences in their physiologic 1986; Krieger al. 1993; Williamset al. 1994; profiles and health status. Cavalli-Sforza et al. 1996). From an Conversely. constructs as “gendered ecosocial vantage, specific pathways expressions of biology" (Krieger and Zierler potentially leading to embodiment of 1995) or “racialized expressions of biology experiences of discrimination - whether (Krieger 1998) are useful for denoting how perpetrated by institutions or individuals. tn social relations of dominance and public or private domains - are legion, as are subordination affect expression of health plausible health outcomes. This is because outcomes linked to biological procesess and discrimination creates and structures traits invoked to define rnembership in exposures to noxious physical, chemical. subordinate and dominant groups. In the case biological, and psychosocial insults, all of of biological sex and gender, for example, which can affect biological integrity at women’s ability to become pregnant has been numerous integrated and interacting levels, used to define women's roles and to restrict simultancously comprised of genes, cells, women's employment in certain male and tissues, organs, and organ systems. The net relatively well-paid occupations, even effect, as discussed in a growing literature on though other less well-paid and typically causal pathways leading to inequallties in female occupations may be equally health across the life course, is to create, hazardous-with these gendered roles in turn using Eric Brunner’s term, a "biology of shaping distributions of pregnancy outcomes inequality" (Brunner 1997; Breilh 1979; (Krieger and Zierler 1995). Or, in the case of Townsend et al. 1990; Krieger et al. 1993; Anderson et al. 1989; Essed 1992; Krieger et Doyal 1995; Evans et al. 1994; Amick et al. al. 1993; Feagin and Sikes 1994). 1995; Williams and Collins 1995; Wilkinson From a theoretical standpoint, the utility of an 1996; Kuh and Ben-Shlomo 1997; re also ecosocial framework is that it encourages Chapter 13). development of specific testable hypotheses Conceptually, however. the myriad by systematically tracing pathways between socially structured trajectories — operative social experimences and their biologic throughout the life course by which expression. Applying these five pathways to discrimination can affect health can be case of racial discrimination and distributions coalesced into five clusters. As delineated in of pressure among black and white Table 3-3, these pathways involve exposure, Arnericans, an eco. social framework thus susceptibillty, and responses (both social and guides to explore the following kinds of biologic) to: hypotheses.
l. Economic and social deprivation: at Pathway #1
work, at home, in the neighborhood, and Residential segregation and occupational other relevant socioeconomic regions segregation lead to greater economic 2. Toxic and hazardous conditions deprivation among African Americans and (pertaining to physical, chemical, and increased likelihood of living in biological agents): at work, at home and neighborhoods without good supermarkets, in the neighborhood. thereby reducing access to affordable, nutritious diets; risk of hypertension is 3. Socially inflicted trauma (mental, elevated by nutritional pathways involving physical. or sexual, ranging from verbal high fat, high salt, and low vegetable diets to violent): at work, at home, in the (Anderson et al. 1989; Troutt 1993; Khaw neighborhood, in society at large 1993). 4. Targeted marketing of legal and illegal psychoactive substances (alcohol, Pathway #2 smoking, other drugs) and other Residential segregation increases risk of commexiities (e.g., junk food) exposure to lead among Afncan Americans 5. Inadequate health care, by health care via contaminated soil (related to proximity of facilities and by specific providers neighborhood to freeways) and lead paint (including access to care, diagnosis, and (related to decreased resources for removing treatment) and replacing lead paint); lead elevates risk of hypertension by damaging renal Also relevant are health cosequences of physiology (Sorel et al. 1991; Lanphear et al. people’s varied responses to discrimination. 1996; Northridge and Shepard 1997). These can range from internalized oppression and use of psychoactive substances to Pathway #3 reflective coping, active resistance, and Perceiving or anticipating racial community organizing to end discrimination discrimination provokes fear and anger; the and promote social justice et al. 1981; physiology of fear (“flight-or-fight" ability to guide healthy public policy crimped response) mobilizes lipids and glucose to — unless they take into account interweaving increase energy supplies and sensory of social and biological determinants of well- vigilance and also prroduces transient being. elevations in blood pressure; chronic trigering of physiologic pathways leads to Effects Of Discrimination Upon sustained hypertension (Krieger 1990; Epidemiologic Knowledge Krieger and Sidney 1996; Harburg et al. 1973; Anderson et al. 1989; Armstead et al, Discussion of how theory directs generation 1989; James et al. 1984; Dressier 1991 a; Of hypotheses in turn points to one important Jones et al. 1996; Williams 1997a; Williams additional way discrimination can affect et 1997b). population health: its impact on epidenuologic knowledge and public health Pathway #4 practice. At issue are the kinds of questions epidemiologists do and do not ask, the studies Targeted marketing of high-alcohol content we conduct, and ways we analyze and trverages to African-American communities interpret our data and consider their likely increase likelihood of harmful use of alcohol flaws. to reduce feelings of distress; excess alcohol consumption elevates risk of high blood That scientists' ideas are shaped, in part, by pressure (Anderson et al. 1989; Moore et al. dominant social beliefs of their times is well 1996). dcxumented by historians of public health, medicine. and wience (Haller 1971; Rose and Pathway #5 Rose 1979; Fee 1987; Haraway 1989; Rosenberg and Golden 1992). Relevant to Poorer detection and clinical management of epidemiology, during the last 20 years a hypertension among African Americans substantial body of literature has begun to increases risk of uncontrolled hypertension divument how scientific knowledge and, due to insufficient or inappropriate medical more importantly, real people, have harmed Care (Anderson et al. 1989; Ahluwalia et al. by scientific racism, sexism, and other related 1997). ideologies, including eugenics, that justify discrimination in relation to class, age, sexual By specifying these discrete pathways— orientation, and disability (Kneger et al. however entangled in people's real lives— 1993; Kneger 1987, 1992; Haller 1971; Jones ecosocial theory thus provides a coherent 1981; Navarro 1986; Gamble 1989. Hubbard way to integrate social and biologic 1990; Leslie 1990; Minkler and Estes 1991; reasoning about discrimination as a Stevens 1992; Williams et al. 1994', Fee and determinant of population health. Instead of Kneger 1994; Gill 1996; Muntaner et al. cataloguing an eclectic list of risk factors - or 1996). presuming genetic explanations as sufficient or fundamental (e.g„ Wilson and Grim 1991; At Issue are both acts of onussion and for refutation, see Curtin 1992), ecosocial commission. These range from the virtual theory proposes that explanations of invisibility of lesbians and gay men in major population health are incomplete — and their public health databases (Stevens 1992; council on Affairs 1996), to distortions of imaginable and then use these data both to and therapeutic knowledge due to explain observed racial/ethnic disparities in underrepresentation of people of color and health and to prove the "black race" was women in epidemiologic studies, clinical innately inferior to the "white race" and “fit" trials, and even medical textbooks (Sechzer only for slavery (Krieger 1987; Gamble et al. 1994; Mendelsohn et al. 1994; King 1989; Haller 1971; Cartwnght 1850; Nott 1996; Ruiz and Verbruge 1997), to the and Gliddon 1857). conduct of premised on the view that innate During the mid, however, the first differences underlie poorer health of generation of U.S. black physicians — along subordinate groups, absent consideration of with abolitionists —challenged the very how subordination might affect health. category of “race”. Arguing that people Vividly illustrating detrimental effects of were more alike than different, they instead discrimination upon generation and conducted studies showing diversity of application of scientific knowledge, to health outcomes among free and enslaved choose but one example, is the pernicious blacks and similarity of health outcomes and longstanding legacy of “race" among blacks and poor whites (Krieger epidemiology; comparable accounts exist 1987; Smith 1859; Reyburn 1866). Based on for eugenic constructions of classbased these studies, they accordingly argued that differences in health (Sy&nstricker 1933; slavery and economic duress — not innate Kevles 1986), for sexist analyses of constitution — were the principal reasons women's health (Hubbard 1990; Fee and that black Americans had or different health Krieger 1994; Doyal 1995), and, to a lesser than white Americans. This alternative extent, for heterosexist research on lesbian viewpoint flourished briefly during and after and gay health (Stevens 1992; Erwin 1993; the Civil War. After the destruction of Council on Affairs 1996). Reconstruction, however, leading medical Historically, “race" first attained and researchers again conducted studies and prominence in U.S. medical research in the proffered explanations based on the premise early 1700s (Stanton 1960; Krieger 1987; that “race” — not racial subordination — Gamble 1989). Aprrarance of “race" as was the root of racial inequalities in health category relevant to health followed (Krieger 1987; Gamble 1989; Haller 1971). institutionalization of the “one drop rule" in The next serious challenge to biologic various slave codes established in the mid- definitions of “race" in biomedical literature to-late 1600s (Stanton 1960; Davis 1991). emerged in the aftermath of World War II, in This rule specified that if someone had only part in reaction to Nazi racial science, “one drop" of African “blood” she or he was espcially its fusion of eugenics and anti- deemed “black”. Embedded in this allegedly Semitism to justify both "Aryan" supremacy biologic and innate definition of "race" was and the Holocaust (Kevles 1985; Proctor the notion of intrinsic “racial" superiority 1988). In 1951, UNESCO released its first and inferiority. Based on this belief, leading Statement on race, rebutting its validity as a scientists and physicians conducted studies biologic category; subsequent revisions, to document – and occasionally fabricate amplifying this point, were issued in 1964, Jarvis 1844; Deutsch 1944 '—racial/ethnic 1969, and, most recently, 1997 (Kupper differences in every physical feature 1975; Katz 1998). All editions emphasize 1998; President's Cancer Panel 1998). that although distributions of specific genetic Tellingly, whereas the keyword identifies traits may vary across geographic regions, no 33,921 articles indexed in Medline since ensemble of linked characteristics exists that 1966, only about 2,600 (7%) are additionally delineates distinct "races." Empirical indexed by the keyword “socioeconomic" evidence supporting this view is now so well and only the 16 studies (0.0005%) listed in established that the dominant view among Table have attempted to study self-reported contemporary population geneticists, other experiences of racial discrimination in biologtsts, anthropologists, and scientists is relation to health. Correcting this imbalance that racial categories reflect social and requires explicit attention to theories guiding ideological conventions, not meaningful research to explain population patterns of natural distinctions (King 1981 ; Lewontin health, disease, and well-being. 1982. Jary and Jary 1995; Cavalli-Sforza et al. 1996; Williams 1997a; Katz 1998). Or, as INTIMATE CONNECTIONS: stated in the 1997 revision of the UNESCO EPIDEMIOLOGY AND THE TRUTHS statement: "Pure races. in the sense of genetically homogenous populations, do not OF OUR BODY AND BODY POLITIC exist in the human species today, nor is there any evidence that they have ever existed in In summary, epidemiologists can draw on a the past" (Katz 1998). variety of study designs (Fig. 3-1) and concepts (Table 3—3) to develop and test Yet, despite this consensus. the 1995 third epidemiologic hypotheses about health edition of the Dictionary of Epidemiology consequences of discrimination. Arguably (sponsored by the International the most fruitful approaches will Epidemiological Association) continues to systematically address discrimination in defines "race" as "persons who are relatively relation to (1) its varied aspects (type; form; homogeneous with respect to biological agency; expression; domain; level); (2) inheritance" (Last 199.5, p. 139). Worse, cumulative exposure (timing; intensity; flouting contemporary scientific knowledge, frequency; duration); (3) likely pathways of it baldly asserts that "In a time of political embodiment; (4) likely forms of responses correctness, classifying by race is done and resistance and their health consequences; cautiously," as if only ideology, and not and (5) effects upon scientific knowledge. scientific evidence, were at issue. The net effect of such views has been an Stated simply, the epidemiology of health overemphasis in epidemiologic research on consequences of discrimination is, at heart, allegedly genetic explanations of the investigation of intimate connections racial/ethnic inequalities in health and a between our social and biological existence. disregard for how racism, rather than "race," It is about how truths of our body and body drives these disparities (Krieger and Bassett politic engage and enmesh, thereby 1986; Cooper and David 1986; Jones et al. producing population patterns of health, 1991; Krieger et al. 1993; Ahmad 1993; disease, and well-being. Williams et al. 1994; Williams and Collins To research how discrimination harms 1995; Muntaner et al. 1996; Lillie-Blanton health, we accordingly must draw on not only and LaVeist 1996; Williams 1997a; Freeman a nuanced understanding of the likely biological pathways of embodying discrimination, from conception to death, but also a finely tuned historical, social, and political sensibility, situating both the people we study and ourselves in the larger context of our times. Out of the epidemiologic come mitrnent to reduce human suffering, we can extend our discipline's to elucidate how oppression, exploitation, and degradation of human dignity harm health—and, simultaneously, further knowledge and inspire action illuminating how social justice is the foundation of public health. Embodying equality should be our goal for all.