THE PERSISTENCE OF RACIAL SEGREGATION

Some observers have considered trends in average Black–White segregation across all metropolitan areas. After noting the downward drift in the mean segregation values since 1970, they have concluded that segregation is declining in importance and that there is little cause for action or concern (Thernstrom and Thernstrom, 1997). This reading of the data, however, confounds two distinct trends. In the subset of U.S. metropolitan areas that contain a relatively small percentage of Black residents, levels of Black–White segregation have indeed been declining, often quite precipitously (Krivo and Kaufman, 1999; Massey and Gross, 1991). The downward trend has been especially pronounced in newer and smaller metropolitan areas of the south and west with relatively small Black populations, and in areas that contain colleges, military bases, and large stocks of post-1970 housing (Farley and Frey, 1994). In contrast, among older metropolitan areas that contain a disproportionate share of the nation's Black population, segregation levels remain stuck at very high levels. Considering the average across all metropolitan areas misrepresents the situation of African Americans because the declines are concentrated in places where few African Americans live.

Iceland et al. (2002) provide the latest data on levels and patterns of racial segregation. I draw upon their figures to consider the degree of residential segregation experienced by African Americans in the United States, dividing the population into three groups: those living in metropolitan areas characterized by low to moderate levels of segregation; those living in areas of high segregation; and those living in areas characterized by an intense form of racial isolation known hypersegregation (Massey and Denton, 1989, 1993). Iceland and colleagues present Black–White dissimilarity indices for all U.S. metropolitan areas in the year 2000. The dissimilarity index measures the unevenness of Black and White settlement across neighborhoods, and is the most widely used measure of residential segregation (see Massey and Denton, 1988).

Following convention, I define dissimilarity values of sixty or more as “high” and those below this mark as either low or moderate. I then divide highly segregated metropolitan areas into two categories: those that are hypersegregated and those that are not. Hypersegregation exists whenever Blacks display an index value of sixty or greater on four of the five dimensions of segregation defined by Massey and Denton (1988, 1989): evenness, isolation, clustering, centralization, and concentration. Some thirty metropolitan areas satisfy the criteria for hypersegregation.

Rather than considering the share of cities experiencing each level of segregation, I focus on the percentage of African Americans living under each segregation regime. At the dawn of the twenty-first century, a clear majority of African Americans (almost 60%) lived in a metropolitan area that was highly segregated, and a substantial minority (some 41%) lived under conditions of hypersegregation. Not all Blacks live in metropolitan areas, of course, so in Figure 1, I present a pie chart that includes only metropolitan residents.

Segregation experienced by URBAN African Americans in 2000. Source: Iceland et al. (2002.)

As can be seen, nearly half of all metropolitan Blacks (48%) live under conditions of hypersegregation, and another fifth (21%) live under a regime of racial segregation that is “merely” high. In other words, among African Americans who reside in U.S. metropolitan areas, a distinct minority—less than a third—currently enjoy conditions of low or moderate residential segregation.

Recent trends in Black–White segregation within the nation's hypersegregated metropolitan areas are not very encouraging. Figure 2 presents Black–White dissimilarity indices for 1980, 1990, and 2000 for the five most segregated metropolitan areas and for the aggregate of all thirty areas that satisfy the criteria for hypersegregation. Across all hypersegregated areas, the average level of Black-White segregation went from seventy-seven in 1980 to seventy-one in 2000, a drop of just 8% in twenty years. In some metropolitan areas, change was barely detectable. For example, Chicago, Detroit, Newark, and Milwaukee display indices above eighty throughout the period. No other group in the history of the United States has ever experienced such high levels of segregation, even for a brief historical moment (Lieberson 1980).

Trends in Black-White residential segregation for hypersegregated metropolitan areas 1980–2000. Source: Iceland et al. (2002).

In their 1993 book, Massey and Denton referred to the regime of Black-White segregation in the United States as “American Apartheid.” Figure 3 brings this metaphor to life by comparing levels of Black–White dissimilarity in hypersegregated U.S. metropolitan areas with the degree of segregation experienced by Africans in the Union of South Africa under apartheid (taken from Christopher 1993). Whereas the de jure apartheid of South Africa produced an average dissimilarity index of ninety in South African urban areas as of 1991, the de facto apartheid in the United States yielded values that were not much lower: Eighty-six for Detroit in the year 2000, eighty-three in Milwaukee, and eighty-one in both Chicago and Newark. The average across all hypersegregated areas was seventy-two.

Degree of Black-White segregation in hypersegregated metro areas of the U.S. compared with metro areas in South Africa under apartheid. Sources: Christopher (1993); Iceland et al. (2002).

EXPLAINING RACIAL SEGREGATION

Data thus reveal that a majority of all African Americans, and the large majority of urban African Americans, continue to experience high levels of residential segregation in U.S. cities, and that about half of all urban Blacks and more than 40% of all African Americans experience hypersegregation, a degree of racial separation that is little different from that achieved in South Africa under apartheid. A variety of hypotheses have been offered to explain persistent Black segregation. The easiest hypothesis to dismiss is the hypothesis that racial segregation reflects socioeconomic differences between African Americans and Whites. This explanation proposes that because the former generally have lower incomes than the latter, more African Americans are channeled into lower-class neighborhoods, on average, than Whites.

To test this hypothesis, Figure 4 presents Black–White dissimilarity indices (taken from Massey and Fischer, 1999) that were computed within income categories of the fifty largest metropolitan areas. A line corresponding to a high level of segregation is positioned just above the Segregation Index of 60 to facilitate interpretation. As can be seen, at all income levels the degree of Black–White segregation remains “high.” Although we observe a slight decline from the poorest to the lower-middle income category, thereafter the trend is flat. At all levels of income, Blacks are highly segregated. In contrast, among Latinos and Asians, the level of segregation is moderate among the poorest families and falls even lower as income rises. Indeed, the poorest Latinos and Asians (those earning under $15,000) are more segregated than the most affluent African Americans (those earning at least $50,000).

Segregation of Blacks, Latinos, and Asians from Whites by income in 1990. Source: Massey and Fischer (1999).

Other social scientists have argued that persistent racial segregation reflects the preference of African Americans for living in segregated Black neighborhoods (Clark 1992; Patterson 1998). When Charles (2003) tabulated nationally representative survey data on housing preferences, however, she found that Blacks expressed weaker preferences for co-residence with members of their own group than did Whites, Asians, or Latinos. The bar chart in Figure 5 shows the ideal neighborhood racial composition expressed by White and Black respondents to the General Social Survey. The data come from a “show card experiment” where respondents were shown a picture of a neighborhood containing blank houses and were asked to color them in to indicate their preferred distribution of Black, White, Asian, and Latino neighbors.

Ideal neighborhood desired by Whites and Blacks in 2000. Source: Charles (2003).

This figure makes it clear that Whites very strongly prefer same-race neighbors, but that Blacks do not. Whereas the ideal neighborhood for the typical White person is 57% White (containing just a smattering of other groups), the ideal neighborhood for Blacks is only 30% Black and would, in fact, contain a larger share of Whites (42% on average). As of the year 2000, therefore, the degree of in-group preference expressed by Whites was about twice that of Blacks whereas the willingness of African Americans to tolerate out-group neighbors was 2.6 times that of Whites.

The exceptional nature of White racial intolerance is indicated forcefully by Figure 6, which shows the percentage of Whites reporting an ideal neighborhood that is all White (with no other groups present), and the percentage reporting an ideal neighborhood with no Blacks present (although allowing in other minorities). Some thirty years after the civil rights era, about a fifth of White Americans would still prefer to inhabit a neighborhood that was all White, and a quarter would prefer to live in a neighborhood that had no African Americans. In contrast, only 6.5% of African Americans wished to live in an all-Black neighborhood and just 9% preferred one with no Whites.

Preference for all in-group and no-outgroup neighborhoods in 2000. Source: Charles (2003).

These racial preferences appear to be driven more by negative stereotyping toward African Americans than by attachment to other Whites. When Charles (2003) developed indicators of stereotyping, in-group attachment, and perceptions of class difference and used them to predict White avoidance of Blacks within neighborhoods, she found that avoidance was most powerfully explained by the holding of negative images about Blacks. As shown in Figure 7, the standardized effect of racial stereotyping on neighborhood preferences (0.390) was about seven times that of perceived class differences (0.056), and about four times that of in-group preferences (0.091).

Explaining White avoidance of Black neighbors. Source: Charles (2003).

Black residential segregation is not only a function of anti-Black attitudes, of course. Substantial evidence suggests that discrimination remains a powerful force in American housing (Galster 1990a, 1990b; Ross and Yinger, 2002; U.S. Department of Housing and Urban Development 2002; Yinger 1993). The limitation of housing opportunities for African Americans was clearly demonstrated in a recent analysis done by Massey and Lundy (2001), who assigned auditors to call advertised rental units in the Philadelphia metropolitan area and inquire about the availability of apartments. Male and female speakers of White, middle-class English, Black-accented English, and Black English Vernacular called selected listings and read a standardized script inquiring about the unit's cost and availability. Results showed that callers who spoke an identifiably “Black” linguistic register achieved far less access to rental housing than callers speaking White, middle-class English.

Figure 8 shows the percentage of auditors who reached a rental agent and were told that a unit was still available. Access is always greater for Whites and a significant interaction between race, class, and gender appears to exist. Middle-class White males always achieve the greatest access, followed by middle-class White females and middle-class Black males. Behind them are lower-class Black males, and in last place are lower-class Black females (assuming that Black English Vernacular indicates lower class origins). Whereas White, middle-class males gained access to rental housing on 76% of their attempts, Black lower-class females did so on only 38% of theirs. Moreover, having gained access, Black females were far more likely to have the issue of credit problems raised and to be assessed application fees. Whereas rental agents mentioned credit worthiness as a potential problem to 3% of White, middle-class males, they did so to about a quarter of lower-class Black females.

Percentage gaining access to rental units and percentage having credit raised as an issue, Philadelphia 2000. Source: Massey and Lundy (2001).

SEGREGATION AND STRATIFICATION

Persistent residential segregation undermines the social and economic well-being of African Americans in a variety of ways. First, by restricting spatial mobility it necessarily limits social mobility because of the close interconnection between the two processes (Massey et al., 1987; Massey and Fong, 1990). Second, by segmenting Black housing demand and channeling White buyers away from Black neighborhoods, it reduces the value of Black housing, making it more difficult for African Americans to accumulate wealth in the form of home equity (Conley 1999; Oliver and Shapiro, 1995; Yinger 1993). As a result, Black wealth remains a small fraction of White wealth despite improvements in employment and earnings (Keister 2000). Third, segregation contributes to the spatial mismatch between the geographic placement of jobs and the residential location of the people who need them (Kain 1968; Preston and McLafferty, 1999).

Consistent with Wilson's (1987) emphasis on concentration effects and their role in perpetuating socioeconomic disadvantage, perhaps the most important mechanism of racial stratification operates through segregation's role in promoting the spatial concentration of poverty. As already noted, high levels of racial segregation interact with shifts in the distribution of income to concentrate poverty geographically (Massey and Fischer, 2000). Under conditions of high or rising Black poverty, segregation necessarily produces neighborhoods of concentrated poverty because the disadvantage created during economic downturns is confined to a small number of racially isolated neighborhoods that are clustered together in space and concentrated in high densities at the center of the metropolitan area.

The interactive effect of rising segregation and increasing poverty is illustrated in Figure 9 (drawn from the simulation developed by Massey 1990). The bottom (solid) line shows what happens to the spatial concentration of poverty as the level of Black-White segregation increases from minimum to maximum, assuming a constant Black poverty rate of 20% and a fixed but moderate level of class segregation between poor and non-poor Black households. Under conditions of racial integration with a 20% poverty rate, the average poor African American lives in a neighborhood that is 25% poor (owing to modest class segregation). As racial segregation increases, however, the concentration of poverty nearly doubles. Under conditions of complete segregation, the average poor African American lives in a neighborhood where 40% of the families are poor.

Effect of racial segregation on concentration of Black poverty: simulation results for city of 128,000 inhabitants that is 25% Black and has class segregation. Source: Massey (1990).

The top (dashed) line shows what happens to the concentration of poverty when the rate of Black poverty is increased to 30%. Under conditions of racial integration, this shift in the distribution of income raises the concentration of poverty somewhat: the share of poor in the neighborhood of the typical poor black person goes from 25% to 30%. Under conditions of total segregation, in contrast, an already disadvantaged neighborhood environment becomes markedly worse, with the concentration of poverty rising from 40% to 60%. The difference between a neighborhood where the poverty rate is 25% and one where it is 60% is a slightly higher rate of poverty, and a much higher level of segregation.

SEGREGATION, STRESS, AND STRATIFICATION

As segregation concentrates poverty, it also concentrates anything that is correlated with poverty to create a uniquely disadvantaged social environment characterized by high rates of joblessness, welfare dependency, substance abuse, and single parenthood. Because crime is also associated with poverty, segregation likewise ends up concentrating social disorder and violence, yielding an unusually hostile and threatening environment to which poor African Americans must adapt (Anderson 1999; Massey 1995).

Figure 10 shows how segregation increases exposure to major crimes within neighborhoods because of the observed correlation between the poverty rate and crimes such as murder, rape, assault, robbery, burglary, larceny, and auto theft (taken from Massey 2001b). As can be seen, given complete integration and a Black poverty rate of 20%, the average poor African American is predicted to reside in a neighborhood with a crime rate of around 56 per 1000 inhabitants. In contrast, given total racial segregation and a Black poverty rate of 30%, the typical poor African American is expected to live in a neighborhood where the crime rate is 84 per 1000, and in some cases, 50% or higher.

How segregation increases exposure to major crime (murder, rape, assault, robbery, burglary, larceny, auto theft). Source: Massey (2001a, 2001b).

Because African Americans experience elevated rates of poverty and high levels of segregation, they are fated to live in environments characterized by much higher rates of crime and violence compared with other groups. This fact is clearly illustrated by recent data from the National Longitudinal Survey of Freshmen, which interviewed African Americans and Latinos entering twenty-eight selective colleges and universities in the Fall of 1999 (see Massey et al., 2003). Respondents were asked to estimate the racial composition of the schools and neighborhoods they inhabited at ages 6, 13, and 18, and to report the frequency with which they witnessed various examples of violence (shootings, stabbings, beatings, etc.) and social disorder (graffiti, prostitution, drunkenness, etc.).

Using these data, Massey and Fischer (2002) constructed severity-weighted indices of exposure to violence and disorder. Figure 11 presents the resulting indices for African Americans and Latinos classified by the average level of segregation they experienced while growing up, and then compares these measures to indices computed for Asians and Whites. The figure clearly reveals that African Americans and Latinos who grew up in segregated schools and neighborhoods (>70% minority, on average) experienced far greater exposures to dangerous and threatening events than those who grew up in integrated circumstances (<30% minority). Those who came of age in racially mixed schools and neighborhoods (30%–70% minority) generally fell in-between.

Exposure of freshmen at selective schools to social disorder and violence in schools and neighborhoods while growing up. Source: Massey and Fischer (2002).

Consider the index of exposure to social disorder. African Americans and Latinos who grew up in integrated circumstances were exposed to about the same level of social disorder as Whites and Asians. Whereas the index of exposure to social disorder was 18.5 for Whites and 18.3 for Asians, it was only slightly higher at 19.5 for those African Americans and Latinos who grew up in integrated schools and neighborhoods. Among those coming of age in racially mixed settings, in contrast, the index was 25.4; and among those growing up under conditions of segregation, the index was 31.7. Thus, moving from integration to segregation increased a student's exposure to social disorder by around 63%.

The effect of segregation on exposure to violence is even more pronounced. Whereas the severity-weighted index of exposure to violence stood at 11.8 for African Americans and Latinos from integrated backgrounds—only slightly more than the values of 10.9 and 10.3 observed for Asians and Whites—it was 18.1 for minorities from racially mixed backgrounds and 26.0 for those from segregated backgrounds. In other words, segregation was responsible for increasing a student's prior exposure to violence by a factor of around 2.5 compared with Whites, Asians, and minorities who grew up within integrated schools and neighborhoods. Recall that these particular African Americans and Latinos had already been admitted into the most elite segment of American higher education, suggesting that the differential in exposure to disorder and violence by level of segregation would probably be even greater among African Americans and Latinos generally.

Large differences in lifetime exposure to disorder and violence carry important implications for the process of stratification because of the well-documented effects of chronic stress on human capacities. Frequent or prolonged exposure to disorder and violence within schools and neighborhoods because of racial segregation is quite likely to produce a chronic activation of the human stress response. To understand the manifold effects of stress on human beings, biomedical researchers have developed the concept of allostasis, which refers to the tendency of organisms to perpetuate their survival and maintain stability through bodily change in response to changes in the environment (McEwen and Lasley, 2002; Sterling and Ayer, 1988)

Whenever a person perceives an external threat, a brain organ known as the hypothalamus triggers an allostatic response, which is a complex interaction between the brain, the endocrine system, and the immune system. Upon perceiving the threat, the hypothalamus immediately signals the adrenal glands to release adrenaline (McEwen and Lasley, 2002). The flow of this hormone into the bloodstream accelerates the heartbeat, constricts blood vessels in the skin, increases blood flow to internal organs, dilates the bronchial tubes, triggers the release of fibrogen into the circulatory system (to promote clotting), releases glucose and fatty acids into the bloodstream from stored fats (to provide a ready source of energy), and signals the brain to produce endorphins (to mitigate pain).

While all this is going on, the hypothalamus simultaneously signals the pituitary gland to release an adrenocorticotropic hormone, which, in turn, causes the adrenal glands to secrete cortisol into the blood (McEwen and Lasley, 2002). Cortisol acts to replace the energy stores depleted by adrenaline, converting energy into glycogen and fat. Cortisol also promotes the conversion of muscle protein to fat, blocks insulin from taking up glucose, subtracts minerals from bones, and changes the external texture of white blood cells to make them “stickier” and more adhesive.

The allostatic response is nature's way of maximizing an organism's resources to meet an immediate, short-term threat. Long-term functions such as the building of muscle, bone, and brain cells are temporarily sacrificed to put more energy into the bloodstream for evasive or aggressive action (McEwen and Lasley, 2002). The hypothalamic-pituitary-adrenal (HPA) axis is common to all mammals and is designed for infrequent and sporadic use. Unlike most mammals, however, humans are capable of keeping the HPA axis turned on indefinitely because humans are capable of experiencing stress from ideas in addition to actual events. Human beings can anticipate threatening circumstances mentally—imagining events that might occur or recalling past traumas (Bremner 2002; McEwen and Lasley, 2002).

Repeated triggering of the allostatic response through chronic exposure to stressful events—as when someone is compelled by poverty and discrimination to live in a dangerous and violent neighborhood—yields a condition known as allostatic load. As allostatic load increases and persists over time, it has powerful negative effects on a variety of bodily systems (McEwen and Lasley, 2002).

One important set of effects is cardiovascular. Chronically elevated levels of adrenaline increase blood pressure and raise the risk of hypertension. Elevated fibrogen levels increase the likelihood of blood clots and increase the likelihood of thrombosis. The build-up of “sticky” white blood cells causes the formation of arterial plaques that contribute to atherosclerosis. Elevated cortisol levels, meanwhile, cause the production of excess glycogen and fat, raising the risk of obesity, while the suppression of insulin leads to excessive blood sugar and a greater risk of Type II diabetes (McEwen and Lasley, 2002).

Chronically elevated levels of adrenaline also disrupt the functioning of the vagal nervous system. This system is responsible for slowing down the heart rate and reducing bodily tension, acting as a “brake” for the ACH axis. Disruption of the vagal system contributes to the expression of a Type A personality, which is associated with aggressiveness, impulsiveness, frustration, and a low threshold for anger. People with Type A personalities often try to reduce tension by self-medicating with drugs, alcohol, and tobacco, and through these poor coping choices end up exacerbating allostatic load and causing secondary damage to vital organs such as the liver, lungs, and heart (McEwen and Lasley, 2002).

Allostatic load also compromises the human immune system. Long term exposure to elevated cortisol usually lowers the immune response to increase susceptibility to illness and infection (Schulz et al., 1998). In some circumstances, however, cortisol appears to overstimulate the immune system to mistakenly goad it into attacking targets within the body that don't normally pose a threat, leading to the expression of inflammatory diseases such as asthma and autoimmune diseases such as multiple sclerosis, arthritis, and Type I diabetes (McEwen and Lasley, 2002).

Finally, allostatic load has serious consequences for a variety of brain systems, and hence, influences cognitive functioning. The organ of the brain that is primarily responsible for the consolidation and storage of memory is the hippocampus (Carter 1999). Because stressful events are important to remember, the hippocampus is rich in cortisol receptors and people are indeed more likely to remember things that are associated with strong emotions (McEwen and Lasley, 2002). Our ancestors who recalled where and under what circumstances danger occurred were more likely to survive and pass on their genes. Chronically elevated cortisol, however, causes the receptors to become permanently saturated, leading to atrophy of the hippocampus and an impairment of memory, both short-term and long-term (Bremner 2002).

Excessive cortisol also appears to interfere with the normal operation of neurotransmitters such as glutamate, which is a critical ingredient in the formation of synaptic connections. By disrupting the production and operation of glutamate at the synapse, allostatic load inhibits long-term potentiation—the formation of a relatively permanent neural connection—which is the fundamental chemical event in human learning. In this way, chronic exposure to disorder and violence may compromise the very process of learning itself (McEwen and Lasley, 2002).

Finally, the hippocampus plays an important role in shutting down the HPA axis by reducing cortisol production. As a result, damage to it is doubly detrimental. Through its effect on the hippocampus, chronic stress creates a viscous cycle whereby excessive cortisol causes shrinkage of the hippocampus, which causes less inhibition of cortisol production, which also causes more hippocampal shrinkage (McEwen and Lasley, 2002). Over the long run, this cycle leads to dendritic remodeling, wherein neurons become shorter and sprout fewer branches, as well as to the suppression of neurogenesis, or the creation of new brain cells (Gould et al., 1998). Simply put, people who are exposed to high levels of stress over a prolonged period of time are at risk of having their brains re-wired in a way that leaves them with fewer cognitive resources (Bremner 2002; McEwen and Lasley, 2002).

A BIOSOCIAL APPROACH TO STRATIFICATION

The foregoing review suggests a biosocial model of stratification that connects elements of social structure (racial segregation and income inequality interacting to produce concentrated poverty and its correlate, spatially concentrated violence) to distinctively high allostatic loads among African Americans (through their involuntary confinement in areas of concentrated poverty and violence) to an elevated risk of coronary heart disease (hypertension, thrombosis, atherosclerosis, diabetes, and obesity), a greater likelihood of inflammatory disorders (asthma, multiple sclerosis, arthritis), and impaired cognition (atrophy of memory, inhibition of synaptic learning, dendritic remodeling, and suppression of neurogenesis).

This hypothesized biosocial model is summarized in Figure 12 with the various links in the causal chain labeled A through E. Pathway A, the interaction of segregation and inequality to produce the concentration of poverty and its correlates is very well established in the research (Massey 1990, 2001b; Massey and Denton, 1993; Massey and Fischer, 2000). Likewise, pathways C, D, and E, which connect allostatic load to compromised health and cognitive outcomes, have been confirmed in a growing number of clinical and laboratory studies (reviewed in Bremner 2002; McEwen and Lasley, 2002). At the same time, Black-White differentials in mortality, and morbidity from a variety of causes are well-documented (Collins and Hawkes, 1997; Hayward and Heron, 1999; Hummer 1996; Manton et al., 1987; Stockwell and Goza, 1996), and persistent gaps in measured cognitive skills are similarly well known (Jencks and Phillips, 1998).

Biosocial model of racial stratification.

To date, these stubborn racial differentials with respect to health and cognition have been resistant to full explanation using the usual array of socioeconomic and demographic control variables (Geronimus et al., 1996; Hummer 1993; Navarro 1990; Phillips et al., 1998). Even after exhaustive background controls are added to statistical models, a significant racial gap generally remains, leading some observers to fall back on genetic explanations (Herrnstein and Murray, 1999; Rushton 2000).

The only link in the model that has not been established empirically is pathway B, the connection between concentrated poverty/violence and high allostatic loads. No matter how reasonable or logical this pathway might seem, researchers have not yet documented it empirically, though there is substantial evidence connecting segregation to excess Black mortality (Collins and Williams, 1999; Fang et al., 1998; Guest et al., 1998; Polednak 1997). The absence of empirical evidence for pathway B is not because investigators have tried and failed to produce such evidence. Rather, owing to a lack of appropriate data, no one has yet been in a position to document the connection.

What the field needs at this point is a dataset that contains biosocial markers indicating allostatic load gathered from a large multi-racial sample whose individual, family, and neighborhood characteristics are well-defined and measured at various points in time. Compiling such a dataset should be a top priority for stratification research. The leading candidate for such a dataset is the National Longitudinal Survey of Adolescent Health, a nationally representative survey of students enrolled in grades 7 through 12 during September 1994 through April 1995 (when they were roughly 12–18 years of age). Eligible respondents were re-interviewed during April–August of 1996 (when they were aged roughly 14–20) and again during August 2001 through April of 2002 (when they were aged 19–25). Some 15,197 respondents participated in the last wave of the survey, and each person was asked to contribute 15 cc of urine to test for the presence of sexually transmitted diseases. However, 2 ml were set aside and frozen with the date and time of the collection recorded. The existence of these urine samples provides a unique opportunity to generate data on the missing link between neighborhood conditions and health outcomes.

Although measurement of allostatic load from a single urine specimen is likely to be unreliable (Stewart and Seeman, 2000) and influenced by a variety of sources of unobserved heterogeneity (individual differences in diet, sleep, diurnal patterns of hormonal secretion, etc.), these problems are common in social science. The validity and reliability of measurement will probably be no less than for widely used social science indices of self-esteem, racial prejudice, liberalism-conservatism, and other attitudes. As with these indices, access to a large sample size compensates for the lack of reliability and provides sufficient statistical power to separate patterns from noise. To the extent that the measure is unreliable, however, error will mitigate against finding any significant relationship between neighborhood conditions and allostatic load. Thus, if statistically significant relationships are found between neighborhood conditions, cortisol, and health or cognitive outcomes, they can therefore be regarded as conservative. A better approach, of course, would be to build measurement of allostatic load into the next round of the Adolescent Health survey (currently planned for 2004–2005), using multiple biomarkers and assays to achieve greater validity and reliability.

Compiling a multi-level, longitudinal data file that links individuals with measures of allostatic load is important because the biosocial model just outlined offers plausible, objective accounts of racial differentials with respect to health, cognition, and mortality that do not require one to fall back on essentialist genetic theories, which make little sense when applied to the socially constructed category of race. In the biosocial model of stratification I have sketched, racial differentials are explained by the unique social structure to which African Americans are subjected in the United States. Among all U.S. social groups, only they simultaneously experience high rates of poverty and high levels of segregation. As a result, they experience far higher rates of neighborhood poverty than members of other groups (Massey and Eggers, 1990; Massey and Fischer, 2003), thus exposing them to higher levels of violence and disorder and driving up their allostatic loads to produce a host of negative health and cognitive outcomes that undermine their ability to compete in the socioeconomic order.

In the past, many social scientists have shunned biologically grounded explanations of racial gaps for fear of legitimizing racist theories or out of a fear of being labeled a racist; but an appreciation of the biosocial mechanisms by which racial differentials are produced turns these fears on their heads. Indeed, by understanding and modeling the interaction between social structure and allostasis, social scientists should be able to discredit explanations of racial difference in terms of pure heredity. In an era when scientific understanding is advancing rapidly through interdisciplinary efforts, social scientists in general—and sociologists in particular—must abandon their hostility to biological science and incorporate its knowledge and understandings into their work.