Home Articles Reader Opinion Editorial Book Reviews Discussion Writers Guide About TCRecord
transparent 13
Topics
Discussion
Announcements
 

Learning Apart, Living Apart: How the Racial and Ethnic Segregation of Schools and Colleges Perpetuates Residential Segregation


by Pat Rubio Goldsmith - 2010

Background: Despite a powerful civil rights movement and legislation barring discrimination in housing markets, residential neighborhoods remain racially segregated.

Purpose: This study examines the extent to which neighborhoods’ racial composition is inherited across generations and the extent to which high schools’ and colleges’ racial composition mediates this relationship. To understand the underlying social processes responsible for racial segregation, I use the spatial assimilation model, the place stratification model, and perpetuation theory.

Population: Data for this project are from the National Education Longitudinal Study (NELS), the Integrated Postsecondary Education Data System (IPEDS), and the U.S. Census.

Research Design: A longitudinal design tracks the racial composition of the schools, colleges, and neighborhoods from adolescence through age 26.

Findings: Holding constant the percent white in teenagers’ neighborhoods, socioeconomic status, and other variables, the percent white that students experience in high school and college has a lasting influence, affecting the percent white in young adult neighborhoods and explaining 31% of intergenerational continuity of neighborhood racial composition.

Conclusions: The analyses suggest that racial segregation in high schools and colleges reinforces racial segregation in neighborhoods.

In 1976, Bowles and Gintis proposed the existence of a rigid stratification system in which an individual’s position in the social class hierarchy was largely inherited from his or her parents. Upward and downward mobility across generations, they argue, was relatively rare. They also argue that schools were an integral component of this relationship. Schools did not create paths for upward mobility, but were institutions that reinforced the intergenerational transmission of class position.


In this article, I adopt a similar theoretical perspective, but instead of analyzing social class, I examine the residential racial segregation system. The work follows directly from Sharkey (2008), who argues that neighborhood contexts are stratified into more advantaged and less advantaged areas, and individuals’ positions in these strata are largely inherited from the residential backgrounds of their parents’ household. His empirical analyses demonstrated little intergenerational mobility in regard to neighborhood strata defined by income levels.


My main contribution to the study of intergenerational mobility in regard to neighborhood context, in addition to looking at race rather than class, is to examine the role of schools and colleges. I examine whether segregation in schools and colleges contributes to the intergenerational transmission of residential locations in terms of racial composition. I ask: To what extent does the racial composition of teenagers’ neighborhoods associate with the racial composition of their adult neighborhoods, and to what extent can the relationship be explained by the racial composition of students’ high schools and colleges?


Considering the role of schools and colleges in the intergenerational transmission of context makes sense for a number of reasons. The racial composition of neighborhoods is similar in youth and in adulthood (Dawkins 2005). Residential racial segregation systems can also be viewed as stratification systems because the percent white in a neighborhood is related to youths’ life chances. More racially segregated metropolitan areas have greater racial differences in educational achievement, high school completion, youth employment, and single parenthood (Card and Rothstein 2005; Cutler and Glaeser 1997). Compared with blacks who move from one inner-city neighborhood to another, blacks who move from an inner-city neighborhood to a suburban neighborhood find more prejudice and discrimination from peers and adults, but they go to schools with higher academic standards, receive more help in school, attain more education, and, when they are adults, have higher employment rates (Kaufman and Rosenbaum 1992; Rosenbaum and Popkin 1991). Predominantly black and Latino neighborhoods, in comparison with predominantly white ones, often have inferior institutions (including schools), less network density, and less mutual support for raising children (Kozol 2005; Sampson, Morenoff, and Earls 1999).


The racial segregation of schools also can be considered a form of stratification. Although students of color often are devalued and stigmatized in predominantly white schools (Foster 1997), attendance at these schools privileges students in many ways. Compared with schools with fewer whites, predominantly white schools have smaller class sizes, higher average socioeconomic status (SES), a greater curricular focus on college preparation, and climates of high expectations (Goldsmith 2003; Oakes and Guiton 1995). Politicians stigmatize and neglect predominantly minority schools (Kozol 2005), while students attending predominantly white schools attain more education and are less likely to go to prison (Guryan 2004; LaFree and Arum 2006; Wells and Crain 1994).


Understanding the extent to which school segregation mediates the intergenerational inheritance of neighborhood racial composition is both theoretically and practically important. Theoretically, it will improve our understanding of the role of schools in perpetuating residential segregation and creating different life chances for blacks, Latinos, and whites. Politically, it will inform debates about the benefits of school desegregation programs. Will school desegregation programs promote residential integration? A finding that levels of exposure to whites in schools and colleges explain part of the intergenerational inheritance of neighborhood racial compositions would suggest that they can.


This article begins with a review of the literature on neighborhood and school segregation. I use three theoretical models: the spatial assimilation model, the place stratification model, and perpetuation theory. After reviewing the relevant literature, I report findings from an empirical study of these issues. I employed research methods widely used in studies of residential segregation (e.g., Crowder and South 2005; Crowder, South, and Chavez 2006). In these studies, researchers appended information about the racial and ethnic composition of residential neighborhoods to individual-level data. By doing so, they could identify which people tended to leave which neighborhoods and where they tended to move. In my study, I appended information on percent white in residential areas during the high school years and at age 26 to see how strongly racial compositions are inherited. I also appended information about the percent white in students’ high schools and colleges, making it possible to describe the percent white that students experience as they transition from their neighborhoods of origin through high school and college, and then into their young adult neighborhoods.


THEORIES OF RACIAL SEGREGATION


The United States is a racially hierarchical society that privileges people with white identities culturally, economically, and socially (Bonilla-Silva 1997). The racial segregation of institutions monopolizes resources for whites and isolates them from other groups. Residential segregation is the bedrock upon which other forms of segregation are formed. As Feagin (2006) writes, “Residential segregation reinforces, even creates, segregated schools, religious organizations, recreational facilities, and workplaces. All such segregated organizations in turn reinforce residential segregation—and thus reinforce white isolation . . . from people of color” (247).


Residential segregation in terms of race in the United States developed over the 20th century as “[t]he real estate industry, banks, appraisers, and insurance agents translated private prejudice into public action ultimately sanctioned by the federal government in Federal Housing Administration loan policies and the federal highway program” (Denton 2001, 94). These structural and individual processes resulted in whites becoming increasingly concentrated in suburban and gentrified neighborhoods, and blacks and Latinos becoming more isolated in racially distinct neighborhoods within the core cities of large metropolises. Reinforcing and adding to this process was the widespread flight of white families with school-age children from neighborhoods with predominantly nonwhite schools to more-white areas (Clotfelter 2004).


Even though the Fair Housing Act of 1968 made many of the practices involved in the creation of residential segregation illegal, residential segregation has proved to be highly persistent. Since 1970, blacks’ segregation from whites has changed little, and Latinos’ has actually increased (Charles 2003). Average percent white in the neighborhoods of the 331 U.S. metropolitan areas of blacks and Latinos is only 37% and 51%, respectively (Charles 2003).


Recent research on residential segregation has been guided by the spatial assimilation model and the place stratification model. I use these two models, as well as perpetuation theory, to understand the connections between neighborhood and school segregation over time.


The spatial assimilation model (Alba and Logan 1993; South and Crowder 1998; South, Crowder, and Chavez 2005) contends that neighborhoods and schools become more desirable as they become relatively more white. People try to exchange their capital endowments for residence in whiter neighborhoods and attendance (for their children) at whiter schools, and therefore, people with more capital and their children are more likely to be in predominantly white contexts. Capital is conceived of broadly in this model to include economic, human, and cultural components. Economic capital includes income and wealth; human capital is considered to be useful skills and is often proxied with indicators about educational achievement or attainment. Cultural capital does not refer to elite culture (e.g., attending the opera) but to immigrants’ levels of cultural assimilation, and it is indicated by such things as English fluency and generation. Suburban residence is an additional stepping stone into predominantly white contexts (Alba and Logan 1993). According to the model, whites are in schools and neighborhoods with relatively more whites because they have relatively more capital than blacks and Latinos.


The place stratification model views racial segregation as the outcome of whites’ attempts to maintain spatial distance from subordinated groups (South and Crowder 1998). Massey and Denton (1993) applied this framework in their history of the structural causes of segregation. The theory contends that two dominant processes create or reproduce residential segregation. The first is racial differences in preferences. In general, whites have less tolerance than blacks and Latinos for living in predominantly nonwhite contexts and even integrated contexts (Charles 2003). Blacks’ and Latinos’ attitudes show more tolerance for nonwhite contexts or integrated contexts (Charles 2003). Racial differences in preferences are thought to be responsible for white flight from predominantly nonwhite neighborhoods. The other process focuses on the multiple forms of discrimination that blacks and Latinos experience when attempting to enter predominantly white contexts (Charles 2003). This discrimination includes unequal treatment from real estate agents, lenders, city planners, homeowners, renters, and neighbors (Charles 2003).


In addition to these two models, research on individual moves has uncovered an additional pattern responsible for residential segregation not anticipated by either model: the tendency for people to move to neighborhoods with a similar racial composition to the neighborhood they left. Crowder, South, and Chavez (2006) claimed that this relation is “by far the most important [empirical] explanation” for “black-white gaps in exposure to whites in residential neighborhoods” (86). In other words, the main reason that whites move to neighborhoods with relatively more whites than blacks do is that whites move from neighborhoods with relatively more whites. If whites and blacks moved from neighborhoods with similar percentages of whites, they would move into neighborhoods with much more similar percentages of whites. Because the spatial assimilation and place stratification models do not consider the racial composition of origin neighborhoods as a determinant of neighborhood destinations, neither model can anticipate this outcome.


The similarity in the racial compositions of origin and destination neighborhoods is one instance of a pattern noted by those who developed perpetuation theory. Perpetuation theory (Braddock 1980; Braddock and McPartland 1989; Wells and Crain 1994) holds that individuals perpetually experience the same racial compositions in their neighborhoods, schools, and other institutions over time. This theoretical camp emerged from research on school desegregation. Researchers from this camp saw school desegregation as a program that took blacks from segregated environments and put them into integrated ones, and they wanted to know whether these experiences led blacks toward more integrated lives after they left school.


The theory acknowledges that whites actively construct exclusionary barriers limiting black and Latino access to white contexts but argues that blacks and Latinos who have been in integrated institutions are better able than their counterparts from segregated institutions to navigate these barriers. Researchers argue that blacks and Latinos who attend schools with whites become more comfortable interacting with whites and being in white contexts and that they become more skilled at interacting with whites (Braddock 1980; Braddock and McPartland 1989; Eaton 2001). In addition, blacks and Latinos in places with relatively more whites develop more social ties with whites. Even if these ties are not close, they provide blacks and Latinos with access to information in white networks about opportunities or processes for entering other predominantly white institutions (Dawkins and Braddock 1994; Eaton 2001; Wells and Crain 1994). This could include information about applying to predominantly white colleges or housing opportunities in predominantly white neighborhoods.


Recent research on whiteness is helpful for understanding the effects of segregation on whites. Eduardo Bonilla-Silva and David Embrick (Bonilla-Silva and Embrick 2007; Bonilla-Silva, Goar, and Embrick 2006) argue that whites who grow up in all-white or nearly all-white settings develop a white habitus. A white habitus is a mental lens that leads whites to avoid and exclude blacks from their social networks and institutions. Whites who grow up in isolation from people of color, moreover, come to see their isolation and exclusionary practices as natural and hence unproblematic, rather than as a highly discriminatory process that they actively perform. The obfuscation of their racial motives enables them to continue maintaining their isolation after they become adults.


According to perpetuation theory, experiences in relatively less white contexts will ease whites’ resistance to integrated environments. In particular, theorists working in this tradition argue that whites from desegregated schools become more comfortable in integrated settings than whites from segregated-white schools. I consider this a weakening of resistance to desegregation because white flight from integrated contexts has been one of the primary ways that whites have resisted integration (Wells et al. 2004).


FROM NEIGHBORHOOD TO HIGH SCHOOL


The series of school desegregation programs resulting from the Supreme Court’s decision in Brown v. Board of Education (1954) led to a dramatic increase in interracial contact in schools, especially in the South, which went from the most to the least segregated region in the United States between 1968 and 1972 (Clotfelter 2004). However, schools remained highly segregated, even relative to neighborhoods. Most of the reasons why schools are so segregated are consistent with the place stratification model. In residential areas where whites lived near people of color, whites persuaded local government agents, especially school officials, to gerrymander school attendance zones so they corresponded to racially defined neighborhood boundaries (Clotfelter 2004). Whites also tended to send their children to private schools and other schools of choice in larger numbers than did nonwhite families, especially as the percent of nonwhites in their school zone, neighborhood school, or neighborhood rose (Bankston and Caldas 2002; Clotfelter 2001; Ledwith and Clark 2007; Renzulli and Evans 2005; Saporito 2003).


Saporito and Lareau (1999) argue that when white families consider schooling options, they begin by eliminating all predominantly minority schools and then choose among the remaining ones. They also argue that nonwhites put less weight on the racial composition of schools when making choices. To determine whether whites’ greater inclinations to opt out of neighborhood schools resulted in greater school segregation, Saporito and Sohoni (2006) compare the percent white in school attendance zones with that in neighborhood schools. They found that percent white was, on average, greater in attendance zones than in neighborhood schools, suggesting that whites were using school choice options to isolate their children from nonwhite students. Adding further support to their argument, they found that the gap between the percent white in attendance zones and schools was largest in areas that had more school choice options. The gap was smaller in places with school desegregation programs.


Research has also shown, consistent with the spatial assimilation model, that immigrants are less likely than natives to send their children to schools of choice, although results vary by ethnic group (Betts and Fairlie 2001). Independent of their neighborhood context, children from high-SES backgrounds are also more likely to attend schools of choice (Lauen 2007).


The outlined considerations about neighborhoods and high schools help explain the kinds of racial compositions that teenagers have in these two institutions. The prevalence of neighborhood schools results in many teenagers experiencing similar racial compositions in their high schools and neighborhoods. Net of the percentage white in neighborhoods, whites are likely to have greater percentages of whites in their schools than blacks and Latinos, who, along with immigrants and those from low-SES backgrounds, experience downward mobility into schools with relatively fewer whites in the transition from neighborhood to school (Betts and Fairlie 2001; Clotfelter 2004; Lauen 2007; Saporito and Sohoni 2006).


FROM NEIGHBORHOOD AND HIGH SCHOOL TO COLLEGE


Colleges and universities within the United States constitute a highly stratified system of schools that range from public two-year colleges that have largely open admissions that generally serve local constituents, to highly selective private institutions that recruit nationally. As in K–12 education, interracial contact in higher education increased during the 20th century. As late as 1954, 83% of black collegians were in Historically Black Colleges and Universities (HBCUs). By 1998, this number dropped to 20%, and large numbers of blacks were attending historically white colleges (Clotfelter 2004).


Like K–12 education, colleges and universities are also racially segregated. Blacks and Latinos tend to be concentrated in two-year colleges and nonselective four-year colleges, and whites are concentrated in selective four-year colleges (Deil-Amen and Lopez Turley 2007). However, researchers have not examined how the racial compositions of a student’s neighborhood and high school relate to the racial composition of colleges and universities to the same extent that they have examined other forms of segregation.


Most researchers have seen colleges and universities as heavily stratified along the dimensions of SES and standardized test scores (Davies and Guppy 1997; Karen 2002). As per the spatial assimilation model, the racial segregation of colleges is partially attributable to differences among whites, blacks, and Latinos in SES and standardized test scores (Alon and Tienda 2007). In fact, blacks are more likely than whites, and Latinos are equally likely as whites, to attend selective colleges and universities net of differences in SES and test scores (Alon and Tienda 2007). Karen (2002) attributes blacks’ and Latinos’ high rates of attendance at these institutions to social pressure from these groups. In addition, researchers working from the perspective of perpetuation theory have shown that in comparison with blacks from predominantly black schools, those from integrated schools attended colleges and universities with relatively more whites (Braddock 1980; Braddock and McPartland 1982; Crain and Weisman 1972 [as cited in Dawkins and Braddock 1994]).


In addition, the prevalence of high-SES students with high standardized test scores in predominantly white neighborhoods and high schools (Goldsmith 2003) may contribute to segregation in college, as these students often attend more selective colleges and universities than their counterparts from predominantly nonwhite contexts. Nevertheless, the transition from neighborhoods and high schools to college should allow for much more mobility than the transition from neighborhoods to high schools because colleges enroll students from geographic areas much larger than neighborhoods.


FROM NEIGHBORHOOD, HIGH SCHOOL, AND COLLEGE TO NEIGHBORHOOD


The spatial assimilation model, the place stratification model, and perpetuation theory are all important for studying residential segregation. Empirical research on the spatial assimilation model shows that racial differences in destination neighborhoods between whites and blacks, and to a lesser extent between whites and Latinos, is weakly related to differences in economic and human capital. People with more education, income, and wealth tend to live in whiter neighborhoods, though the associations are weak. Many educated middle-class whites live in racially mixed neighborhoods through processes such as gentrification, and many middle-class blacks and Latinos live in predominantly nonwhite neighborhoods (Crowder and South 2005; Crowder, South, and Chavez 2006).


The spatial assimilation model proved more useful for explaining the movement patterns of immigrants (Charles 2003; South, Crowder, and Chavez 2005). Non-English speakers and first-generation immigrants are less integrated into white contexts than English speakers and second- and later-generation immigrants.


Consistent with the place stratification model, empirical research has shown that whites move into neighborhoods with relatively more whites as compared with the neighborhoods that blacks and Latinos move into (Crowder, South, and Chavez 2006; South, Crowder, and Chavez 2005). In addition, whites’ movement patterns have shown a strong disinclination toward living with nonwhites. As the percent white in the neighborhood of origin declines, whites become much more likely than blacks and Latinos to leave the neighborhood—a process usually referred to as “white flight” (Crowder 2000; Quillian 2001). Because these differences between whites, blacks, and Latinos are observed even with controls for differences in human, economic, and cultural capital, researchers have interpreted their effects as evidence of some combination of racial differences in preferences or in discriminatory or exclusionary barriers that limit nonwhites’ entrance into white contexts. However, the research tracking which individuals leave or enter which neighborhoods has not included measures of these underlying mechanisms (Sharkey 2008).


Perpetuation theorists have shown desegregated schooling to foster greater integration among blacks. In comparison with blacks from black-segregated schools, those from integrated schools, colleges, and universities move to neighborhoods with relatively more whites (Crain et al. 1992; Crain and Weisman 1972 [as cited in Dawkins and Braddock 1994]). In other words, black adults tend to live in neighborhoods with a racial context that mirrors that of the schools they attended. For this reason, I expect to observe an association between the racial compositions of schools and adult neighborhoods for all groups. Despite these impressive findings, researchers working in this theoretical tradition have not usually studied how or why these outcomes occur (Wells et al. 2004).


The only research on a large quantitative data set that has spoken to mechanisms suggested by this theory is Dawkins (2005). Dawkins found that for blacks, having a household head with more integrated friendships is associated with living in a more integrated neighborhood as an adult. Research using in-depth interviews has shown evidence for the theory’s mechanisms. Wells and others (2005) have shown that adults who attended integrated schools describe themselves as being more comfortable in integrated settings than other people they know. Eaton (2001) also found through interviews that blacks from integrated schools tapped into white networks and gained information about applying to predominantly white colleges and finding job opportunities.


In addition, research linking the racial composition of schools to the racial composition of neighborhoods has not controlled for the racial composition of the neighborhoods in which students were raised. Without controlling for the racial composition of neighborhoods, it is not possible to determine if school racial compositions have an independent affect on later neighborhood compositions. This follows because neighborhood and school racial compositions are very similar for young people, and it is easy to mistake the effect of one for the other unless they are examined together. In addition, people often develop very strong psychological ties to places and people, and these ties may root individuals within geographic areas (Altman and Low 1992). Thus, many young adults may establish independent residences in the same neighborhoods where they lived as teenagers. A finding that high schools’ and colleges’ racial composition influences where young adults live, controlling for the racial composition of prior neighborhoods, will provide strong evidence that the racial segregation of schools and colleges reinforces neighborhood segregation.


METHODS


Data for this project are from the restricted version of the National Education Longitudinal Study 1988-2000 (NELS), the Integrated Postsecondary Education Data System (IPEDS), and Summary File 3B of the 1990 and the 2000 Censuses of Population and Housing. The NELS provides longitudinal data that follow a base-year sample of respondents who were eighth graders in the United States in 1988. Follow-ups occurred in 1990, 1992, 1994, and 2000. The respondents were about 26 years of age in the final wave. I used the sample of 10,827 respondents who participated in all panels from 1988 to 2000. Information from the NELS about respondents’ families (from parents or the respondent) and schools (from principals) is attached to respondents’ records.


Information about NELS respondents’ postsecondary institutions is included by appending IPEDS data on the first postsecondary institution attended (hereafter referred to as colleges). I averaged the IPEDS data over the years 1992–96 because most students began college during that period. Information about NELS respondents’ residential neighborhoods was created by appending census data to individual cases via their five-digit residential zip code numbers, which were recorded separately in 1990, 1992, and 2000. I attached 1990 census data to the 1990 and 1992 waves and 2000 census data to the 2000 wave.


Zip code areas (ZCAs), which are called zip code tabulation areas in the 2000 Census, are second to census tracts as the most-often-used level of aggregation to study “neighborhood” effects (e.g., Ainsworth 2002; South, Baumer, and Lutz 2003). I used ZCAs rather than tracts because NELS data cannot be linked to tracts. The results of using one level of aggregation rather than other are unlikely to vary. In their recent review of a large body of research on neighborhood effects, Sampson, Morenoff, and Gannon-Rowley (2002) argue that places are stratified by race and class at many different levels and that “empirical results have not varied much with the operational unit of analyses” (446). Regardless of whether tracts or zip codes are used, the type of error introduced is likely to be similar when measuring exposure to whites. If there is segregation within ZCAs, then ZCAs’ percent white will, on average, underestimate the percent white that whites directly encounter and overestimate the percent white that blacks and Latinos directly encounter.


In all, these data contain information on individuals’ exposure to whites in their teenage ZCAs, high schools, colleges, and young adult ZCAs. The age of these individuals is apt for studying the causes of residential segregation because their mobility rates are very high. During this period, many individuals leave home as they take on adult roles (Long 1988).


LOGIC OF ANALYSES


The analyses are designed to estimate the extent to which ZCAs’ percent white is similar in youth and adulthood and how much of this relationship is mediated by percent white in schools and colleges. The analyses used data measured at four time points. At Time 1, which is 1988, or eighth grade, I measured all control variables and race/ethnicity. The control variables include the various forms of capital and regions. These variables are described in more detail next. At Time 2, the teenage years (1990–92), I measured high schools’ and ZCAs’ percent white. Colleges’ percent white is measured at Time 3, which for most students occurs between 1992 and 1996. The final outcome, ZCAs’ percent white in young adulthood, was measured at Time 4, which is age 26, or year 2000.


The control variables were measured during eighth grade to reduce overcontrol bias (Sampson, Morenoff, and Gannon-Rowley 2002). Overcontrol bias occurs when one or more of the control variables are endogenous to the contextual effects. The inclusion of such variables downwardly biases the estimates of neighborhood effects and school effects. Because the control variables were measured during eighth grade—before the contextual effects—they could not be endogenous to them.


To account for problems from missing data, I followed the advice of Allison (2002) and used SAS’s PROC MI to calculate multiple imputations. This method imputes missing values from an algorithm that predicts these values as a linear combination of other variables in the analyses. A random variance component was added to the imputed values so that their variances equaled the variance of the real scores. I recoded imputed values to be within the range of the nonimputed data and so that discrete variables would have discrete imputed scores. This process was followed five times, creating five different data sets (and hence, multiple imputations). In multiple imputations, the different data sets are used to mitigate the chances that the results obtained from the analyses are influenced by the imputed scores rather than the real ones. This is done by using the results from the five data sets to calculate (in SAS’s PROC MIANALYZE) the mean coefficients and standard errors that account for the variation of the coefficients across data sets.


Models that used colleges’ percent white as a dependent variable omitted students who did not attend college by assigning their weight a value of 0 (Lee and Forthofer 2006). When colleges’ percent white was used as an independent variable, missing values were assigned the mean. A dummy variable is used to flag where this occurred. Readers also should keep in mind that the samples of Asians and American Indians were small, and therefore, any inferences regarding these populations are speculative.


Estimates and standard errors also were adjusted for the survey design. The sample of respondents participating in all waves of the NELS was constructed by the National Center for Education Statistics (NCES) from the sampling frames of earlier waves and by oversampling individuals who were unlikely to respond or who were difficult to locate. Because of the unequal probabilities of selection, I used the weight (F4PNLWT) constructed by the NCES to make the sample representative of the eighth-grade class of 1988 (Curtin et al. 2002). Using the weight reduces the bias that might result from calculating estimates from a nonrepresentative sample, and it protects estimates from model misspecification (Lee and Forthofer 2006).


Survey design was accounted for because NELS data are stratified, and multiple students were selected within base-year schools, which are the primary sampling unit (PSU). Both sampling properties can bias standard errors, but the latter is far more important because the homogeneity of students within PSUs will downwardly bias standard errors and make hypothesis tests too liberal. I used the NCES-created variables STRATA and PSU to adjust standard errors using the Taylor Series method (Lee and Forthofer 2006), which increases standard errors by a separate amount for each variable based on the size of the intraclass correlation coefficient for the variable within PSUs. Linear regression models adjusting for survey design were estimated using the SAS procedure SURVEYREG.


MEASUREMENT


Table 1 reports the weighted means for the entire sample for all variables. Exposure to whites in ZCAs during the teenage years is the average percent non-Hispanic white in respondents’ 1990 and 1992 ZCAs. Exposure to whites in schools during the teen years is the average percent non-Hispanic white in students’ 1990 and 1992 schools, as reported by principals in the NELS data. I averaged the 1990 and 1992 reports to reduce measurement error from point-in-time estimates (Solon 1992).


Race and ethnicity were measured with dummy variables for Latinos (of any race), Asians, African Americans, and American Indians. Non-Hispanic whites are the reference category. Effects of race and ethnicity, which exist net of all other factors, are typically interpreted in the research on residential segregation as evidence of social processes proposed by place stratification models, such as differences in preferences or discrimination (Charles 2003; Sharkey 2008). Unfortunately, the data I possess, like other data sets that examine the movement of individuals into and out of different contexts, do not contain measures of the underlying mechanisms (Sharkey 2008).


I also included a dummy variable indicating which students lived with their parents at age 26. I did this because the percent white in these students’ teenage and adult neighborhoods are highly related because many of them have (presumably) not moved.


Table 1. Descriptions of Variables, Weighted Means, and Standard Errors Adjusted for Survey Design.

Variable                          

Description

M

SE

Percent white in

   

High school (1990–92)

 

70.65

0.01

College (1992–96)

 

71.19

0.01

ZCA (1990–92)

 

76.02

0.01

ZCA (2000)

 

69.24

0.01

Control variables

   

Non-Hispanic Black

(1 = yes) Reference: non-Hispanic white

  0.12

0.01

Latina/o

(1 = yes) Reference: non- Hispanic white

  0.11

0.01

Non-Hispanic Asian

(1 = yes) Reference: non- Hispanic white

  0.03

0.00

Non-Hispanic Amer. Indian

(1 = yes) Reference: non- Hispanic white

  0.01

0.01

SESa

Index of family income, education, occupational prestige, and possessions

-0.09

0.02

Family sizea

Number of people in the family

  4.62

0.02

One-parent family

(1 = yes) Reference: both biological parents

  0.17

0.01

Other nontraditional family

(1 = yes) Reference: both biological parents

  0.16

0.01

Parent expects BA

(1 = yes; 0 = no)

  0.60

0.01

Has an immigrant parent

(1 = one or more; 0 = none)

  0.13

0.01

Language minoritya

(1 = yes; 0 = no)

  0.12

0.01

Test score compositea

Summary of math and reading test scores

50.98

0.20

Grade point averagea

Self-report, covers middle school years

  2.91

0.02

Expects a BA or more

(1 = yes; 0 = no)

  0.66

0.01

Female

(1 = yes; 0 = no)

  0.50

0.01

Inner-city residence

(1 = yes) Central city of metropolitan area; reference: suburban area

  0.25

0.01

Rural residence

(1 = yes) Outside of metropolitan area; reference: suburban area

  0.32

0.01

North Central

(1 = yes) Reference category: South

  0.26

0.01

Northeast

(1 = yes) Reference category: South

  0.19

0.01

West

(1 = yes) Reference category: South

  0.20

0.01

a Indicates a variable created by the NCES that is available in the NELS.


The remaining variables were measured to correspond to 1988, the base year of the NELS. As per the spatial assimilation model, I controlled for variables related to family background, individual achievement, and acculturation. I used the NCES-created variable SES to control for variation in economic and human capital among students’ families. SES is a variable created by the NCES using information on family income, parents’ occupations and education, and household possessions. I also included an indicator of family size to adjust the SES measure for the number of people in the family. I used two dummy variables to control for variation in family structure. The reference category was families with both biological parents present, and the two dummy variables flag students in families with one parent and with some other family structure. The models also include a dummy variable that indicates parents who expect their child to earn at least a bachelor’s degree.


I controlled for the students’ human capital with self-reported high school grade point averages (GPAs) and standardized test scores. Since attitudes also influence educational and occupational mobility later in life, I controlled for educational expectations with a dummy variable indicating those who expected to earn at least a bachelor’s degree. To capture variation in acculturation, I used dummy variables to identify students who had (a) a minority language background and (b) at least one immigrant parent. I included dummy variables for residency in nonmetropolitan areas and inner-city areas, using suburban residency as the reference category. Finally, I controlled for regions, as other researchers have found them to be related to the extent of school segregation and the racial mix of destination neighborhoods (Clotfelter 2004; South and Crowder 1998). I measured regions with dummy variables for the West, North Central, and Northeast. The South is the reference category. Next I examine the linear regression models.


RESULTS


NEIGHBORHOOD TO HIGH SCHOOL


Table 2 shows models predicting high schools’ percent white. Model 1 includes the effects of ZCAs’ percent white. Because many students attend neighborhood schools, this relationship should be very strong. However, there should be some mobility in the transition from ZCAs to high schools because some students will attend schools of choice or be in desegregation programs. Mobility rates can be measured with an elasticity coefficient, which equals the square root of the coefficient of determination (Solon 1992). An elasticity coefficient of one (1.0) signifies no mobility. Model 1, which estimates an elasticity coefficient of [square root (0.74) =] 0.86, shows very little mobility in the transition from ZCAs to high schools. As a point of reference, elasticity coefficients for the intergenerational inheritance of income are believed to be about 0.6 (Solon 1992); for the intergenerational inheritance of neighborhood economic context, it is 0.64 (Sharkey 2008).


Table 2. Regression of Selected Variables onto High Schools’ Percent White.

Variable

Model 1

 

Model 2

Model

3

 

Model 4

 

Intercept

5.32

**

83.32

***

16.28

***

19.78

***

ZCA % white (1990–92)

0.54

***

  

0.45

***

0.42

***

ZCA % white squared (1990–92)

0.004

***

  

0.004

***

0.003

***

Black

  

-48.20

***

-9.80

***

-9.31

***

Latina/o

  

-48.52

***

-10.97

***

-7.09

***

Asian

  

-26.14

***

-7.53

***

-4.71

**

American Indian

  

-48.83

***

-15.39

***

-13.01

**

Family SES

      

0.83

 

Family size

      

0.13

 

One-parent family

      

-1.31

 

Other nontraditional family

      

0.08

 

Parent expects college

      

0.75

 

Immigrant parent

      

0.62

 

Grade point average

      

-0.57

 

Standardized test score

      

0.07

 

Language minority

      

-5.84

***

Expects BA or more

      

1.16

 

Female

      

0.09

 

Core city residence

      

-5.13

***

Rural residence

      

1.34

 

Northeast

      

1.19

 

West

      

-1.41

 

North Central

      

1.38

 

Adjusted R2

0.74

 

0.44

 

0.75

 

0.76

 

*p < .05. ** p < .01. ***p < .001.


Model 1 also shows that as percent white in ZCAs increases, so does percent white in schools, although the latter increases slightly more rapidly. This is seen in the model by the positive and significant linear and squared term for ZCAs’ percent white. Model 2 removes the ZCA effects and adds the race dummy variables. This estimates the aggregate amount of segregation in high schools because the coefficients show that on average, blacks’ and Latinos’ high schools have 48.2 and 48.5 percentage points fewer white, respectively, than the high schools of whites, which are, on average, 83.3 percent white (as indicated by the intercept). Adding the ZCA effects (Model 3) shows that the primary reason that schools are so segregated is due to differences in the racial composition of residential neighborhoods (that is, because of neighborhood segregation). Holding ZCAs’ percent white constant reduces the black-white and Latino-white gaps to just 9.8 and 11.0 percentage points, respectively, suggesting that ZCAs’ percent white explains [(48.2 – 9.8) / 48.2 * 100 =] 80 and 77% of the school segregation of these two groups, respectively.


Figure 1. Effects of ZCAs’ Percent White (1990–92) on High Schools’ Percent White, by Race ____________________________________________________________________________

[39_15690.htm_g/00002.jpg]
click to enlarge

_____________________________________________________________________________

Note. The figure was created from Model 3 in Table 3.


Figure 1 graphs the estimated effect of ZCAs’ percent white from Model 3 on high schools’ percent white for whites, blacks, and Latinos.1 The figure helps us see who is upwardly and downwardly mobile in the transition from neighborhoods to high schools. Upward mobility occurs for individuals above the identity line; downward mobility occurs for individuals below the identity line. As the graph shows, the only teenagers who, on average, experience upward mobility are whites from ZCAs with very few whites. This finding is likely to result from white flight: white teenagers in very nonwhite ZCAs find their way into schools with relatively more whites than in their neighborhoods. Other whites, in contrast, have nearly identical percentages of whites in their ZCAs and high schools.


The lines for blacks and Latinos show that almost regardless of the percent white in their ZCA, they experience downward mobility. The only blacks and Latinos who do not are those who live in ZCAs with very few whites. In essence, blacks and Latinos from predominantly minority neighborhoods are spared downward mobility in the transition to high school because they have no place lower to fall.


Model 4 adds the covariates (which are grand-mean centered in all models to make the intercept easier to interpret). These variables improve the adjusted coefficient of determination only slightly (to 0.76). However, the covariates and the race effects explain part of the ZCA effects. The linear term and quadratic term are reduced in magnitude by [(0.54 – 0.42) / 0.54 =] 22% and 12%, respectively. Presumably, the covariates explain part of the ZCA effects because characteristics of parents and their children are related to movement patterns into neighborhoods and schools. Nevertheless, about 80% of the effect of ZCAs’ percent white is independent of the covariates.


The model shows that even net of any differences in capital, blacks and Latinos experience downward mobility relative to whites. While the causes of the gaps cannot be further ascertained, racial and ethnic gaps that are estimated net of capital are frequently interpreted as evidence that is consistent with place stratification models. In this case, racial differences in mobility may result from whites’ tendency to avoid blacks and Latinos by enrolling in schools of choice (Clotfelter 2004; Saporito and Sohoni 2006).


Two other groups experience downward mobility. Schools’ percentage white is 5.8 and 5.1 points lower for students with a language minority background and for students in inner cities (rather than suburbs), respectively. Both of these effects are consistent with the spatial assimilation model, but most of the effects predicted by this model, including those about SES and achievement, do not reach statistical significance.


Table 3. Regression of Selected Variables onto Colleges’ Percent White.

Variable

Model 1

 

Model 2

 

Model 3

 

Model 4

 

Model 5

 

Intercept

29.43

***

29.75

***

79.07

***

39.59

***

42.43

***

ZCA % white (1990–92)

0.55

***

0.33

***

  

0.26

***

0.24

***

High school % white

  

0.23

***

  

0.20

***

0.17

***

Black

    

-28.22

***

-7.23

***

-9.29

***

Latina/o

    

-31.20

***

-10.79

***

-7.50

***

Asian

    

-21.15

***

-10.66

***

-4.71

**

American Indian

    

-25.22

***

-2.94

 

-3.67

 

Family SES

        

-1.51

*

Family size

        

0.01

 

One-parent family

        

-1.07

 

Other nontraditional family

        

-1.21

 

Parent expects college

        

0.20

 

Immigrant parent

        

-3.96

*

Grade point average

        

0.41

 

Standardized test score

        

0.07

 

Language minority

        

-2.55

 

Expects BA or more

        

1.95

 

Female

        

0.10

 

Core city residence

        

2.91

**

Rural residence

        

5.23

***

Northeast

        

1.53

 

West

        

-4.55

***

North Central

        

3.03

**

Adjusted R2

0.43

 

0.45

 

0.29

 

0.45

 

0.50

 

*p < .05. **p < .01. ***p < .001


FROM NEIGHBORHOODS AND HIGH SCHOOLS TO COLLEGE


Table 3 uses colleges’ percent white as the dependent variable. Only respondents with postsecondary enrollment are included. Model 1 of Table 3 examines the amount of mobility between ZCAs and college by regressing ZCAs’ percent white onto colleges’ percent white. The elasticity coefficient of [square root (0.43) =] 0.66 indicates more mobility in this transition than in the one from ZCAs to high schools, but the amount of mobility is still small. The coefficient for ZCAs’ percent white indicates that a 10-percentage-point increase in percent white in the ZCA associates with a 5.5-percentage-point increase in percent white in the college.


Model 2 adds the effect of high schools’ percent white to see whether this variable acts as a mediator. Adding it only slightly improves the model fit, as seen by the slightly higher adjusted r2 (0.45). However, it explains [(.55 – 0.33) / 0.55 * 100 =] 40% of the ZCA effect, indicating an important mediating effect. The coefficient for high schools’ percent white indicates that controlling for ZCAs’ percent white, a 10-percentage-point increase in high schools’ percent white translates into a gain of 2.3 percentage points in college.


Model 3 estimates the aggregate amount of college segregation by including the race dummy variables, but not the effects of percent white in prior contexts. As seen by the intercept, whites’ colleges are, on average, 79.1% white, and the net black-white and Latino-white gaps are 28.2 and 31.2 percentage points, respectively. Model 4, which includes the contextual variables and the race variables, shows that net of differences in ZCAs’ and high schools’ percent white, the black-white and Latino-white gaps are 7.3 and 10.8 percentage points, respectively. The data therefore suggest that the racial segregation of neighborhoods and high schools is one of the main reasons that there is racial segregation in colleges.


Model 5 adds the covariates. Once again, adding these variables only improves the model fit slightly, as seen by the adjusted coefficient of determination, which rises to 0.50. The covariates reduce the ZCA coefficient by 27% and the high school coefficient by 26%. Thus, ZCA and high school context are less predictive of college context when other variables are held constant, but about three-fourths of the effects of prior contexts are independent of these covariates. Increases of 10 percentage points in ZCAs’ and high schools’ percent white are associated with increases of 2.4 and 1.7 percentage points in colleges, respectively.


Once again, the coefficients for blacks and Latinos are significant. Consistent with place stratification models, blacks and Latinos experience downward mobility relative to whites as they transition into college net of differences in capital. The only covariate consistent with the spatial assimilation model is the one for having an immigrant parent. Students with an immigrant parent attend colleges with 4 percentage points fewer whites than students without an immigrant parent. The effects of the other significant covariates are somewhat surprising. Students with high-SES backgrounds experience slight downward mobility. A one-standard-deviation increase in SES results in 1.5 percentage points fewer whites in college. In addition, students from inner cities and from rural areas have 2.9 and 5.2 percentage points more whites, respectively, in their colleges than students from suburban backgrounds.


Determining why these counterintuitive findings occur is not within the bounds of this analysis, but one possibility may be that high-SES students from suburban backgrounds tend to enroll in very selective institutions that use affirmative action policies (Alon and Tienda 2007); these policies may help provide more contact with nonwhites than would be expected based on the percent white in their neighborhoods and high schools. Regardless, the main point is that percent white in ZCAs and high schools is strongly determinative of percent white in college and that, on average, students of color experience downward mobility more often than white students in the transition to college.


Table 4. Regression of Selected Variables onto ZCAs’ Percent White in Young Adulthood.

Variable

Model 1

 

Model 2

Model 3

  

Model 4

 

Model 5

 

Intercept

17.96

***

12.37

***

79.52

***

18.86

***

28.18

***

ZCA % white (1990–92)

0.68

***

0.47

***

  

0.43

***

0.42

***

High school % white

  

0.13

***

  

0.11

***

0.10

***

College % white

  

0.18

***

  

0.16

***

0.14

***

Black

    

-36.87

***

-5.83

***

-7.37

***

Latina/o

    

-37.20

***

-5.87

***

-4.74

***

Asian

    

-19.40

***

-3.56

**

-0.33

 

American Indian

    

-30.62

***

-2.74

 

-3.15

 

Family SES

        

-0.34

 

Family size

        

-0.02

 

One-parent family

        

-2.04

*

Other nontraditional family

        

-1.34

 

Parent expects college

        

-2.26

***

Immigrant parent

        

-1.79

 

Grade point average

        

0.82

 

Standardized test score

        

-0.12

**

Language minority

        

-2.88

*

Expects BA or more

        

-0.76

 

Female

        

1.01

 

Core city residence

        

-0.05

 

Rural residence

        

1.06

 

Northeast

        

1.37

 

West

        

-2.43

**

North Central

        

2.28

***

Lives at home

-18.31

***

-17.41

***

-4.26

***

-17.31

***

-16.96

***

Lives home * ZCA % white

0.23

***

0.23

***

  

0.23

***

0.23

***

No college

  

1.99

   

2.09

 

0.18

 

Adjusted R2

0.56

 

0.60

 

0.35

 

0.60

 

0.61

 

*p < .05. **p < .01. ***p < .001.


FROM NEIGHBORHOODS, HIGH SCHOOLS, AND COLLEGES TO NEIGHBORHOODS


Table 4 uses adult neighborhoods’ percent white as the dependent variable. As seen in Model 1, ZCAs’ percent white in the teenage years has a coefficient of 0.68, and the model has an adjusted r2 of 0.56 and an elasticity coefficient of 0.75, suggesting a very strong intergenerational inheritance of racial composition in neighborhoods.2


Model 2 adds the effects of high schools’ and colleges’ percent white to see whether they mediate the ZCA effect.3 As seen in the model, the effects of percent white in high schools (0.13) and colleges (0.18) are very robust. Together, they explain [(0.68 – 0.47) / 0.68 =] 31% of the effect of ZCAs’ percent white. Thus, they are very strong mediators. The coefficients indicate that 10% increases in high schools’ and colleges’ percent white are associated with having 1.3% and 1.8% more whites, respectively, in the adult neighborhood, holding constant percent white in the origin neighborhood. In addition to mediating the neighborhood effect, they also slightly improve the model fit, as seen in the increase in the adjusted r2 from 0.56 to 0.60.


Model 3 estimates the aggregate amount of residential segregation among young adults by including the race variables but not the contextual variables. It shows that on average, whites’ neighborhoods are 79.5% white, and the black-white and Latino-white gaps are 36.9 and 37.2 percentage points, respectively. Model 4 suggests that most of these gaps can be attributed to the percent white in the residential ZCA during the teenage years, high school, and college. Holding these constant reduces the black-white and the Latino-white gaps to just 5.8 and 5.9 percentage points, respectively. These are both 84% smaller, suggesting that prior experiences in neighborhoods, high schools, and colleges are by far the most important empirical explanation for racial segregation among young adults.


Adding the covariates (Model 5) improves the model fit only slightly, from 0.60 (model 4) to 0.61. The covariates explain little of the contextual effects. The estimated effects of the percent white of ZCAs, high schools, and colleges are just 2%, 9%, and 13% smaller, respectively, when the covariates are included. Thus, the contribution of the covariates to the model is far less than that of the contextual effects.


The coefficients for high schools’ and colleges’ percent white in Model 5 are 0.10 and 0.14, respectively. While these coefficients are modest, their effects are important because the gaps in high schools’ and colleges’ percent white are so large. In high schools, the black-white and Latino-white gaps are about 48 percentage points; in colleges, they are 28 and 31 percentage points for blacks and Latinos, respectively. Because the gaps are larger in high schools, high schools’ percent white actually explains more of the black-white and Latino-white gaps in ZCAs’ percent white in young adulthood. This can be seen by multiplying the gaps by the coefficients. Closing blacks’ and Latinos’ gaps in high schools would reduce the ZCA gap by (0.1 * 48 =) 4.5 percentage points. Closing the college gap would reduce blacks’ and Latinos’ gaps by 3.9 and 4.3 percentage points, respectively.


However, the effects of schools’ and colleges’ percent white are still smaller than teenage ZCAs’ percent white. The gap in teenage ZCAs’ percent white is 45 percentage points for blacks and 44 percentage points for Latinos. If the teenage ZCA gap were closed, blacks’ and Latinos’ gaps in adult ZCAs would be smaller by 18.9 and 18.5 percentage points, respectively.


Model 5 also shows support for the place stratification model. Other things being equal, the black-white gap is 7.4 percentage points, and the Latino-white gap is 4.7 percentage points. Consistent with the spatial assimilation model, language-minority and single-parent backgrounds associate with downward mobility, but their effects are modest (2.9 and 2.0 percentage points, respectively). The other two significant covariates are surprising. Students whose parents expect them to earn at least a bachelor’s degree and students with higher standardized test scores in eighth grade have relatively fewer whites in their adult neighborhoods. Interpreting these counterintuitive coefficients is complicated because these models hold constant the percent white in their high schools and colleges. The lack of significance for the SES coefficient is consistent with the research showing that such factors as income, wealth, and education play trivial roles in explaining residential racial segregation (e.g., Crowder, South, and Chavez 2006).


To illustrate the importance of prior context, I calculated the expected percent white in adult ZCAs for blacks, whites, and Latinos from teenage ZCAs that are 40% and 80% white. I selected these percentages because few whites live in neighborhoods less than 40% white, and few blacks and Latinos live in neighborhoods that are more than 80% white. I used the slopes from Model 5 of Table 4, so these effects are net of all the covariates. For the levels of high schools’ and colleges’ percent white, I used the values calculated from Model 4 of Table 2 and Model 5 of Table 3. The values for the covariates are their grand mean.


These calculations show that blacks who begin in ZCAs that are 40% white and 80% white are expected to live in ZCAs as young adults that are 48% and 69% white, respectively. This is a gap of 21%. In contrast, the effect of being black compared with white is 7.4%. A 40% difference in percentage white in the initial ZCA is thus nearly three times as influential as identifying as black rather than white. Latinos who begin in ZCAs that are 40% white and 80% white are expected to live in ZCAs as adults that are 51% and 72% white, respectively; for whites, the expected ZCAs are 59% and 80% white, respectively.


DISCUSSION


This article examines the extent to which the racial compositions experienced in schools and colleges mediate this relationship. The data showed that when teenagers grew up and left their parents’ home, they frequently moved into a ZCA that had a very similar percentage of whites to the ZCA they lived in as teenagers. In between these two time points, they also attended high schools and colleges that had similar percentages of whites as in their teenage and young adult ZCAs. This pattern reflects what perpetuation theorists have long argued: that individuals tend to experience the same racial context across institutions and over time.


However, not all individuals in these data experienced the same percentage of whites in all these institutions. It was the small fluctuations in these percentages—the small amounts of upward and downward mobility in institutional transitions—that the analyses exploited to see the role of schools and colleges in perpetuating neighborhood segregation across generations. Individuals who experienced upward mobility in their transition from teenage ZCA to high school and college benefited in the long run by having a higher percentage of whites in their young adult ZCA. Similarly, individuals who experienced downward mobility in these transitions lost in the long run by having a lower percentage of whites in their young adult ZCA.


These upwardly and downwardly mobile individuals tell us what might happen were there more mobility across institutions. If all students were in colleges and high schools that had the same percentage of whites (that is, if percent white in these institutions were constant), the intergenerational transmission of ZCAs percent white would be 31% weaker. The data therefore suggest that school desegregation programs could slowly but steadily integrate neighborhoods.


The absence of desegregation programs results in two problems. First, there is little mobility in the transitions from ZCAs to high schools and to colleges. Second, teenage neighborhoods, high schools, and colleges are very segregated. Because of these two underlying problems, the experiences that people have in school and college actually perpetuate the racial segregation of neighborhoods. To paraphrase the quote from Feagin (2006) cited earlier, residential segregation creates segregated schools and colleges, and segregation in these institutions reinforces residential segregation.


Nevertheless, why segregation is so perpetual across these institutions is only partially clear. The data revealed that the lack of mobility in these transitions is largely unrelated to variation in capital. Only cultural capital was consistently related to mobility. Less acculturated individuals tended to experience downward mobility in each institutional transition. In the transition to college, some evidence suggested that affirmative action policies may be increasing interracial contact for high-achieving, affluent whites from suburban areas. Future research might explore the affects of capital in more depth by looking for interactions by race.


In addition, blacks and Latinos consistently experience downward mobility relative to whites. Downward mobility by blacks and Latinos in the transition from neighborhoods to high schools is consistent with research showing gerrymandered school zones and whites’ inclinations to avoid interracial contact by using school choice policies (Clotfelter 2004; Saporito and Sohoni 2006). Downward mobility by blacks and Latinos in the transition to adult ZCAs is consistent with the research showing racial differences in preferences regarding the racial composition of neighborhoods and discrimination in housing markets (Charles 2003).


The racial differences in mobility are large, but they explain little of the contextual effects related to racial composition. They are also not as influential as prior neighborhood context in determining the racial context of neighborhoods. For example, a 40-percentage-point difference in percent white in teenagers’ ZCAs is about three times more influential than the effect of being black versus white, and more than four times as influential as being Latino versus white.


Perpetuation theory anticipates the strong connections between percent white in teenage ZCAs, high schools, colleges, and adult ZCAs. It holds that people become accustomed to the racial compositions they have already experienced, that people develop skills for navigating these racial compositions, that people develop social ties within their institutions, and that all these factors combine to orient them toward other institutions with similar racial compositions (Wells and Crain 1994). Research on whites in segregated-white neighborhoods (Bonilla-Silva, Goar, and Embrick 2006) and on a more diverse sample of people who attended integrated high schools (Wells et al. 2005) has shown with in-depth interviews the social-psychological effects of racial compositions posited by the theory. In addition, Dawkins’s (2005) research has suggested that levels of interracial contact while growing up explain who lives in more or less integrated neighborhoods as an adult.


However, the data analyzed here do not contain measures of the mechanisms posited by perpetuation theory. In the absence of these measures, the analyses are unable to rule out alternative explanations for the findings. In this case, at least two categories of alternative explanations may be particularly important for future research to consider. First, individuals may form attachments to the particular people and institutions they inhabit, and these may be more or less independent of the racial context. In particular, individuals form strong attachments to their neighborhoods and the people who live there (Altman and Low 1992). Second, there may be organizational links between institutions. These links are best exemplified by neighborhood schools. These educational institutions are linked concretely to neighborhoods and almost surely explain why there is so little mobility from neighborhoods to high school.4


The possibility that these alternative explanations—particular attachments and institutional links—may account for the lack of mobility across institutions means that the data must be interpreted cautiously in regard to perpetuation theory. Nevertheless, some findings lend themselves less to alternative explanations than others. For example, the effects of high schools’ and colleges’ percent white on the percent white in adult ZCAs do not lend themselves to other explanations so easily. The effects from both of these institutions exist net of the percent white in the teenage ZCA, so they are unlikely to indicate particular attachments to a place. In addition, there is no obvious process by which high schools and colleges link their students to adult neighborhoods. High schools and colleges have no formal procedures guiding where students live (years) after they leave school.


The effects of high schools’ and colleges’ percent white on percent white in adults’ neighborhoods imply that experiences in these educational institutions affect individuals in long-lasting ways. Here, the data indicate that blacks and Latinos who went to school or college with relatively more whites moved, years later, into neighborhoods with relatively more whites. And whites who went to school or college with relatively fewer whites moved, years later, to neighborhoods with relatively fewer whites. These findings are consistent with the views of perpetuation theorists, who argue that experiences in racial contexts shape young people’s skills, dispositions, and social networks in ways that direct them toward institutions with racial contexts like those they experienced in youth. Therefore, programs that increase interracial contact in youth, like school desegregation and affirmative action programs, may lead in the long run to more integrated neighborhoods over time.


Acknowledgements


This research was supported by a grant from the American Educational Research Association, which receives funds for its AERA Grants Program from the U.S. Department of Education’s National Center for Education Statistics of the Institute of Education Sciences, and the National Science Foundation under NSF Grant No. RED-0310268. Opinions reflect those of the author(s) and do not necessarily reflect those of the granting agencies.


Notes


1. I do not use separate slopes for each racial group because interactions between the racial and ethnic dummy variables and ZCAs' percent white are not significant.

2. All the models using ZCAs’ percent white as a dependent variable include a control variable for people who still live with their parents, and an interaction between this term and the effect of ZCAs’ percent white during the teenage years. Including these terms makes the main effect of the ZCAs’ percent white its effect on those young adults who do not live with their parents. The effect of ZCAs’ percent white for those who live with their parents is the sum of the main effect (0.68) plus the interaction (0.23), or 0.91. This effect is so strong because many of these people live in the same zip code, and probably the same residence, at the two time points.

3. Students who did not attend college are assigned the mean of colleges’ percent white. A dummy variable flagging where this occurred is also added to the model. The dummy variable captures the effect of colleges’ percent white for people who did not attend college, and the effect of colleges’ percent white captures its effect for people who did attend college. Models that exclude people who did not attend college yielded similar results.

4. If links across institutions, like neighborhood schools, help explain why individuals perpetually experience the same racial context across institutions and over time, then perpetuation theory should be expanded. That is, institutional links should be theoretically incorporated into the model.


References


Ainsworth, James W. 2002. Why does it take a village? The mediation of neighborhood effects on educational achievement. Social Forces 81:117–52.


Alba, Richard D., and John R. Logan. 1993. Minority proximity to whites in suburbs: An individual-level analysis of segregation. American Journal of Sociology 98:1388–427.


Allison, Paul D. 2002. Missing data. Sage University Paper Series on Quantitative Applications in the Social Sciences, 07-136. Thousand Oaks, CA: Sage.


Alon, Sigal, and Marta Tienda. 2007. Diversity, opportunity, and the shifting meritocracy in higher education. American Sociological Review 72:487–511.


Altman, Irwin, and Setha M. Low, eds. 1992. Place attachment, human behavior, and environment: Advances in theory and research, Vol. 12. New York: Plenum.


Bankston, Carl L., III, and Stephen J. Caldas. 2002. A troubled dream: The promise and failure of school desegregation in Louisiana. Nashville, TN: Vanderbilt Univ. Press.


Betts, Julian R., and Robert W. Fairlie. 2001. Explaining ethnic, racial, and immigrant differences in private school attendance. Journal of Urban Economics 50:26–51.


Bonilla-Silva, Eduardo. 1997. Rethinking racism: Toward a structural interpretation. American Sociological Review 62:465–80.


Bonilla-Silva, Eduardo, and David G. Embrick. 2007. “Every place has a ghetto…”: The significance of whites’ social and residential segregation. Symbolic Interaction 30:323–45.


Bonilla-Silva, Eduardo, Carla Goar, and David G. Embrick. 2006. When whites flock together: White habitus and the social psychology of whites’ social and residential segregation from blacks. Critical Sociology 32 (2-3): 229–54.


Bowles, Samuel, and Herbert Gintis. 1976. Schooling in capitalist America. New York: Basic Books.


Braddock, Jomills Henry, II. 1980. The perpetuation of segregation across levels of education. Sociology of Education 53:178–86.


Braddock, Jomills Henry, II, and James M. McPartland. 1982. Assessing school desegregation effects: New directions in research. In Research in sociology of education and socialization, Vol. 3, ed. R. Corwin, 259–82. Greenwich, CT: A1 Press.


Braddock, Jomills Henry, II, and James M. McPartland. 1989. Social-psychological processes that perpetuate racial segregation: The relationship between school and employment desegregation. Journal of Black Studies 19:267–89.


Brown v. Board of Education, 347 U.S. 483, 1954.


Card, David, and Jesse Rothstein. 2005. Racial segregation and the black-white test score gap. Working paper 500, Princeton University, Industrial Relations Section.


Charles, Camille Zubrinsky. 2003. The dynamics of racial residential segregation. Annual Review of Sociology 29:167–207.


Clotfelter, Charles T. 2001. Are whites still fleeing? Racial patterns and enrollment shifts in urban public schools, 1987–1996. Journal of Policy Analysis and Management 20:199–221.


Clotfelter, Charles T. 2004. After Brown: The rise and retreat of school desegregation. Princeton, NJ: Princeton Univ. Press.


Crain, Robert L., Randi L. Miller, Jennifer A. Hawes, and Janet R. Peichert. 1992. Finding niches: Desegregated students sixteen years later. Final report on the educational outcomes of Project Concern, Hartford, CT. Institute for Urban and Minority Education, Teachers College, Columbia University, New York.


Crain, Robert L., and C. S. Weisman. 1972. Discrimination, personality, and achievement. New York: Seminar.


Crowder, Kyle. 2000. The racial context of white mobility: An individual-level assessment of the white flight hypothesis. Social Science Research 29:223–57.


Crowder, Kyle, and Scott J. South. 2005. Race, class, and changing patterns of migration between poor and nonpoor neighborhoods. American Journal of Sociology 110:1715–63.


Crowder, Kyle, Scott J. South, and Erick Chavez. 2006. Wealth, race, and inter-neighborhood migration. American Sociological Review 71:72–94.


Curtin, Thomas R., Steven J. Ingels, Shiying Wu, and Ruth Heuer. 2002. National Education Longitudinal Study of 1988: Base-year to fourth follow-up data file user’s manual (NCES 2002-323). Washington, DC: U.S. Department of Education, National Center for Education Statistics.


Cutler, David M,. and Edward L. Glaeser. 1997. Are ghettos good or bad? Quarterly Journal of Economics 112: 827–72.


Davies, Scott, and Neil Guppy. 1997. Fields of study, college selectivity, and student inequalities in higher education. Social Forces 75:1417–38.


Dawkins, Casey J. 2005. Evidence on the intergenerational persistence of residential segregation by race. Urban Studies 42 (3): 545–55.


Dawkins, Marvin P., and Jomills Henry Braddock II. 1994. The continuing significance of desegregation: School racial composition and African American inclusion in American society. Journal of Negro Education 63:394–405.


Deil-Amen, Regina, and Ruth Lopez Turley. 2007. A review of the transition to college literature in sociology. Teachers College Record 109:2324–66.


Denton, Nancy A. 2001. The persistence of segregation: Links between residential segregation and school segregation. In Pursuit of a dream deferred: Linking housing & educational policy, ed. John A. Powell, Gavin Kearny, and Vina Kay, 89–120. New York: Peter Lang.


Eaton, Susan F. 2001. The other Boston busing story. New Haven, CT: Yale Univ. Press.


Feagin, Joe. 2006. Systemic racism: A theory of oppression. New York: Routledge.


Foster, Michéle. 1997. Black teachers on teaching. New York: New Press.


Goldsmith, Pat António. 2003. All segregation is not equal: The impact of Latino and black school composition. Sociological Perspectives 46:83–105.


Guryan, Jonathan. 2004. Desegregation and black dropout rates. American Economic Review 94:919–43.


Karen, David. 2002. Changes in access to higher education in the United States: 1980–1992. Sociology of Education 75:191–210.


Kaufman, Julie E., and James E. Rosenbaum. 1992. The education and employment of low-income black youth in white suburbs. Educational Evaluation and Policy Analysis 14:229–40.


Kozol, Jonathan. 2005. The shame of the nation: The restoration of apartheid schooling in America. New York: Crown.


LaFree, Gary, and Richard Arum. 2006. The impact of racially inclusive schooling on adult incarceration rates among US cohorts of African Americans and whites since 1930. Criminology  44:73–103.


Lauen, Douglas Lee. 2007. Contextual explanations of school choice. Sociology of Education 80:179–210.


Ledwith, Valerie, and William A. V. Clark. 2007. The effects of the residential mosaic and “white flight” on public school composition: Evidence from Los Angeles County. Urban Geography 28:160–80.


Lee, Eun Sul, and Ronald N. Forthofer. 2006. Analyzing complex survey data. Sage University Papers Series on Quantitative Applications in the Social Sciences, 07-71. Thousand Oaks, CA: Sage.


Long, Larry. 1988. Migration and residential mobility in the United States. New York: Russell Sage Foundation.


Massey, Douglas S., and Nancy A. Denton. 1993. American apartheid: Segregation and the making of the underclass. Cambridge, MA: Harvard Univ. Press.


Oakes, Jeannie, and Gretchen Guiton. 1995. Matchmaking: The dynamics of high school tracking decisions. American Educational Research Journal 32:3–33.


Quillian, Lincoln. 2001. Why is black-white residential segregation so persistent? Evidence on three theories from migration data. Social Science Research 31:197–229.


Renzulli, Linda A., and Lorraine Evans. 2005. School choice, charter schools, and white flight. Social Problems 52:398–418.


Rosenbaum, James E., and Susan J. Popkin. 1991. Employment and earnings of low-income blacks who move to middle-class suburbs. In The urban underclass, ed. Christopher Jencks and Paul E. Peterson, 342–56. Washington, DC: Brookings Institution Press.


Sampson, Robert J., Jeffrey D. Morenoff, and Felton Earls. 1999. Beyond social capital: Spatial dynamics of collective efficacy for children. American Sociological Review 64:633–60.


Sampson, Robert J., Jeffrey D. Morenoff, and Thomas Gannon-Rowley. 2002. Assessing “neighborhood effects”: Social processes and new directions in research. Annual Review of Sociology 28:443–78.


Saporito, Salvatore. 2003. Private choices, public consequences: Magnet school choice and segregation by race and poverty. Social Problems 50:181–203.


Saporito, Salvatore, and Annette Lareau. 1999. School selection as a process: The multiple dimensions of race in framing educational choice. Social Problems 46:418–39.


Saporito, Salvatore, and Deenesh Sohoni. 2006. Coloring outside the lines: Racial segregation in public schools and their attendance boundaries. Sociology of Education 79:81–105.


Sharkey, Patrick. 2008. The intergenerational transmission of context. American Journal of Sociology 113:931–69.


Solon, Gary. 1992. Intergenerational income mobility in the United States. American Economic Review 82:393–408.


South, J. Scott, Eric P. Baumer, and Amy Lutz. 2003. Interpreting community effects on youth educational outcomes. Youth & Society 35:3–36.


South, Scott J., and Kyle L. Crowder. 1998. Leaving the ‘hood: Residential mobility between black, white, and integrated neighborhoods. American Sociological Review 63:17–26.


South, Scott J., Kyle Crowder, and Erick Chavez. 2005. Geographic mobility and spatial assimilation among U.S. Latina/o immigrants. International Migration Review 39:577–607.


Wells, Amy Stuart, and Robert L. Crain. 1994. Perpetuation theory and the long-term effects of school desegregation. Review of Educational Research 64:531–55.


Wells, Amy Stuart, Jennifer Jellison Holme, Awo Korantemaa Atanda, and Anita Tijerina Revilla. 2005. Tackling racial segregation one policy at a time: Why school desegregation only went so far. Teachers College Record 107:2141–77.


Wells, Amy Stuart, Jennifer Jellison Holme, Anita Tijerina Revilla, and Awo Korantemaa Atanda. 2004. How society failed school desegregation policy: Looking past the schools to understand them. Review of Research in Education 28:47–99.




Cite This Article as: Teachers College Record Volume 112 Number 6, 2010, p. 1602-1630
https://www.tcrecord.org ID Number: 15690, Date Accessed: 5/25/2022 11:48:32 AM

Purchase Reprint Rights for this article or review
 
Article Tools
Related Articles

Related Discussion
 
Post a Comment | Read All

About the Author
  • Pat Goldsmith
    University of Wisconsin-Milwaukee
    E-mail Author
    PAT RUBIO GOLDSMITH is associate professor in the department of sociology at the University of Wisconsin-Milwaukee. Currently, he researches the causes and consequences of racial segregation; race and sport; and the mistreatment of Latinos along the U.S.-Mexico border. He has forthcoming publications in Social Forces, Sociology of Sport Journal, and Aztlan.
 
Member Center
In Print
This Month's Issue

Submit
EMAIL

Twitter

RSS