“some of my best friends are ”: interracial friendships ... · “some of my best friends...

28
“Some of My Best Friends Are ...”: Interracial Friendships, Class, and Segregation in America Xavier de Souza Briggs Massachusetts Institute of Technology Ties among persons of different backgrounds, when such ties act as social bridges, play a vital role in diverse societies—expanding identities, opening insular commu- nities of interest, containing intergroup conflicts, and reducing inequalities. Using a phone survey of 29 city-regions matched with census data, this study analyzes predic- tors of interracial friendships for Whites, Blacks, Asians, and Hispanics, with single- and multilevel path models. Results underscore the social significance of the work- place and of civic involvement. Those who report ties to other races tend to be “joiners,” in the broad sense: Involvement in nonreligious groups, socializing with coworkers, and having more friends are robust predictors for all racial groups. These are strongly associated with each other and with higher socioeconomic status, high- lighting a powerful class dimension to accessing intergroup ties. But macro-level op- portunity for contact (metro-level racial makeup) dominates the variation in friend- ship exposure patterns for Whites, whereas associations and other “substructures” are more predictive for minorities. Consistent with immigrant assimilation theory, among minorities, sharing neighborhoods with Whites remains an important—and apparently unique—social marker for the personal relationships of Blacks. INTRODUCTION What’s in a friendship? In an increasingly diverse and unequal society, quite a lot. On individual as well as collective levels, intergroup ties—informal personal networks, formal associations, and other connections among socially dissimilar persons or groups—are uniquely important in social life as well as social theory. This is especially true when such ties function as bridges , enabling meaningful exchanges across social divides and the social worlds they help define. Bridging ties are particularly crucial where they help bind diverse societies, expanding social and civic identities, opening up insular communities of interest, containing ethnic and other intergroup conflicts, and reducing status inequalities, for example, by widening access to valuable information and endorsements. As exceptions to the rule of homophily in social relations, ties among birds of different feathers are important in public as well as private life and in multiple dimensions—social, economic, and political. I am grateful to anonymous reviewers, Robert Putnam, Tom Sander, Barry Wellman, William Julius Wilson, Michael Woolcock, and participants in seminars at Harvard University, the Lincoln Institute for Land Policy, and the University of Cambridge for helpful comments. Meaghan Dolan, Marilyn Milliken, and Michael New gave invaluable assistance with data access, and Benjamin Keys provided superb analytic support. Correspondence should be addressed to Xavier de Souza Briggs, Department of Urban Studies and Plan- ning, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Room 9-521, Cambridge, MA 02139; [email protected]. City & Community 6:4 December 2007 C American Sociological Association, 1307 New York Avenue, NW, Washington, DC 20005-4701 263

Upload: others

Post on 10-Jun-2020

3 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

“Some of My Best Friends Are . . .”: InterracialFriendships, Class, and Segregation in America†

Xavier de Souza Briggs∗

Massachusetts Institute of Technology

Ties among persons of different backgrounds, when such ties act as social bridges,play a vital role in diverse societies—expanding identities, opening insular commu-nities of interest, containing intergroup conflicts, and reducing inequalities. Using aphone survey of 29 city-regions matched with census data, this study analyzes predic-tors of interracial friendships for Whites, Blacks, Asians, and Hispanics, with single-and multilevel path models. Results underscore the social significance of the work-place and of civic involvement. Those who report ties to other races tend to be“joiners,” in the broad sense: Involvement in nonreligious groups, socializing withcoworkers, and having more friends are robust predictors for all racial groups. Theseare strongly associated with each other and with higher socioeconomic status, high-lighting a powerful class dimension to accessing intergroup ties. But macro-level op-portunity for contact (metro-level racial makeup) dominates the variation in friend-ship exposure patterns for Whites, whereas associations and other “substructures”are more predictive for minorities. Consistent with immigrant assimilation theory,among minorities, sharing neighborhoods with Whites remains an important—andapparently unique—social marker for the personal relationships of Blacks.

INTRODUCTION

What’s in a friendship? In an increasingly diverse and unequal society, quite a lot. Onindividual as well as collective levels, intergroup ties—informal personal networks, formalassociations, and other connections among socially dissimilar persons or groups—areuniquely important in social life as well as social theory. This is especially true when suchties function as bridges, enabling meaningful exchanges across social divides and the socialworlds they help define. Bridging ties are particularly crucial where they help bind diversesocieties, expanding social and civic identities, opening up insular communities of interest,containing ethnic and other intergroup conflicts, and reducing status inequalities, forexample, by widening access to valuable information and endorsements. As exceptionsto the rule of homophily in social relations, ties among birds of different feathers areimportant in public as well as private life and in multiple dimensions—social, economic,and political.

†I am grateful to anonymous reviewers, Robert Putnam, Tom Sander, Barry Wellman, William Julius Wilson,Michael Woolcock, and participants in seminars at Harvard University, the Lincoln Institute for Land Policy, andthe University of Cambridge for helpful comments. Meaghan Dolan, Marilyn Milliken, and Michael New gaveinvaluable assistance with data access, and Benjamin Keys provided superb analytic support.∗Correspondence should be addressed to Xavier de Souza Briggs, Department of Urban Studies and Plan-ning, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Room 9-521, Cambridge, MA 02139;[email protected] & Community 6:4 December 2007C© American Sociological Association, 1307 New York Avenue, NW, Washington, DC 20005-4701

263

Page 2: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

CITY & COMMUNITY

Unfortunately, given the long-standing importance of the color line in American life,prior research has shown interracial ties to be rare in the close friendships of many Amer-icans (Marsden, 1988), that is, relative to what random choice would predict in a diversesociety. These are friendships through which important material exchanges, everyday ex-pressive support, and attitudinal influence operate (Jackson, 1977). Still, between 1985and 2004, the share of Americans reporting someone of another race among their con-fidants rose somewhat, from 9 to 15 percent, even as those core networks became morefamily-centered and less grounded in the neighborhood and associational ties that offermore diverse social contacts (McPherson, Smith-Lovin, and Brashears, 2006). And therate of other reported interracial friendships has increased sharply over the past genera-tion, at least between Whites and Blacks—the groups longest surveyed (Thernstrom andThernstrom, 1997). Popular discussions of race in America often center on such interper-sonal relations, which relate so closely to the respect, security, and feelings of mutualitywe often crave.

Despite the importance of interracial ties, to date, research has provided limited viewsof the kinds of people who have them and of how opportunities to form them arestructured. Using census data matched to a restricted-use version of the Social CapitalCommunity Benchmark Survey (SCCBS), a uniquely detailed survey of 29 city-regionsin the United States that range widely in racial makeup and cover all major regionsof the country, this study analyzes the factors associated with interracial exposure infriendships.1 The study tests hypotheses linked to macrostructural, group assimilation,and group threat theories. My focus is on majority-to-minority ties, that is, the analysisconsiders Whites’ friendship ties to members of minority out-groups (Black, Hispanic,and Asian) and those groups’ ties to Whites, not minority-to-minority ties. More specif-ically, the study analyzes racial exposure (or isolation) in friendships, in answer to thesurvey question “Do you have a personal friend who is [white, black, Hispanic, Asian]?,”not the extent of bridging in each person’s network (how many friends) or the contentof the friendships. This research complements recent studies of social trust and othermeasures of social capital in racially diverse settings (Alesina and La Ferrara, 2002;Putnam, 2000, 2003), as well as a wide array of studies focused on the implications of in-creased diversity for America’s communities, schools, workplaces, associations, and otherinstitutions.

The contributions of this study are threefold, reflecting specific limitations of earlierresearch. First, drawing primarily on the General Social Survey, most studies of diversity inAmericans’ personal relations that were conducted over the past two decades have focusedon effects of individual traits and the kinship, neighborhood, and other sources of personalties rather than traits of the local context. Second, studies that have explored effects oflocal context on intergroup ties and other network traits have thus far been limited toa single city or metro area. This multicity, multiregion study affords substantial variationacross local contexts and allows one to distinguish variation within communities fromvariation across them (using multilevel models). Third and finally, socioeconomic status,life stage, and other individual factors associated with forming and sustaining interracialties may operate through direct effects or more indirectly, such as by shaping participationin civic, religious, or other associations, habits of socializing, or other aspects of social life,and also by influencing what kinds of neighborhoods we live in. The links among some ofthese factors differ dramatically across racial groups, yet such path effects have not beendirectly analyzed before.

264

Page 3: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

INTERRACIAL FRIENDSHIPS, CLASS, AND SEGREGATION IN AMERICA

In the next section, I briefly outline scholarly and popular interest in social bridges alongwith the significance of race bridging, in particular, for changing societies. Following thatcontext are three bodies of theory from which I develop hypotheses about correlates ofinterracial ties, data and methods, results, and a discussion of the findings.

WHY DO SOCIAL BRIDGES MATTER?

Social bridges resting on intergroup ties have important consequences for individuals andfor society, for social equality as well as for democracy. Among social scientists, sociologistshave shown a particularly long interest in intergroup ties, most explicitly in such classics asLaumann’s (1973) study of social differentiation in the personal networks of city dwellersand Granovetter’s (1973) essay on the value of acquaintanceships and other “weak” ties,which are more likely than strong ties to cross group boundaries. Seminal analyses of eco-nomic and social status include Blau’s study of cross-cutting social circles (Blau, 1977; Blauand Schwartz, 1984), as well as analyses of status attainment that emphasize the benefitsof diverse social contacts for network range and function (Erickson, 1996; review in Lin1999). Further back, Durkheim’s (1893) theorizing about cross-cutting ties as buttressingthe interdependence among differentiated segments of emerging industrial cities andSimmel’s (1923) conception of social structure as organized by multiple identities, inter-sections of roles, and “webs” of group affiliation helped define modern life in complexsocieties. They also foreshadowed today’s fascination with far-flung networks (includingthe Internet), lifestyle niches, and social boundaries. Over the past generation, sociolog-ical analyses evolved from Laumann’s (1973) study of social distance among subgroupsof White men, which examined ties that crossed national ancestry and religious lines,to contemporary studies across the fuller spectrum of racial identity and gender lines inAmerica.

But it is political scientist Robert Putnam’s work on social capital that has turned ascholarly concern for social bridges into something of a public debate in recent years,encouraged attention to the multiple benefits of such ties, and stirred an activist interestin creating more bridging ties—and, through them, more “bridging social capital” (Briggs,1997; Gittell and Vidal, 1998; Putnam, 2000; Woolcock, 1998). In a comparative analysisof social capital in eight democracies, Putnam and Goss (2002, pp. 11–12) offer a pointedcase for bridging: “It is true . . . that without the natural restraints imposed by members’crosscutting alliances and diverse perspectives, tightly knit and homogeneous groups canrather easily combine for sinister ends. In other words, bonding without bridging equalsBosnia.” Bonding without bridging may also undermine collective action where broaderperspectives and coalitions are required, as when tightly knit neighborhoods are underexternal threat (Granovetter, 1973).

McPherson et al. (2001, p. 415), summarizing the consequences of forces that favorbonding over bridging, argue that “homophily limits people’s social worlds in a way thathas powerful implications for the information they receive, the attitudes they form, and theinteractions they experience.” But networks, schools, workplaces, and formal associations(civic clubs, faith institutions, etc.) that bond along one dimension of social identity(socioeconomic status, for example) can bridge on others (race and gender, say). Suchcross-cutting ties derive their special significance from the fact that they bond on the socialtrait shared by the linked actors while bridging their social differences. Cross-cutting ties

265

Page 4: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

CITY & COMMUNITY

are essential to the development of broader identities and communities of interest (Blauand Schwartz, 1984; Briggs, 2004; Horton, 1995; Varshney, 2002; Warren, Saegert, andThompson, 2001), as well as the reciprocity and learning crucial to democratic behavior(Putnam, 2000). These are the social foundations of power sharing, without which theformal machinery of democratic government tends to falter around the world (Lipset,1994).

Ties that meaningfully bridge social boundaries can also reduce inequality directly,by improving access by lower-status out-groups to information, vouching (recommenda-tions and other social endorsements), preparation, mentoring, and other keys to eco-nomic access and attainment (Dickens, 1999; Ferrand, Mounier, and Degenne, 1999; Lin,1999)—at least when one’s social contacts are willing to help (Smith, 2005). Analyzingthe powerful connection between social capital and cultural capital as resources for gettingahead, Erickson (1996, p. 247) argues, “Social networks are the continuing adult educa-tion of culture, and diversified networks are the liberal arts programs teaching a little ofalmost everything.” In the U.S. context, research confirms the particular importance ofbridging network ties for advancement by poor minorities in inner cities (Briggs, 1998;Crain and Wells, 1994; Dominguez and Watkins, 2003; Johnson, Bienenstock, Farrell, andGlanville, 2000).

THEORETICAL BACKGROUND: DETERMINANTS OF INTERGROUP FRIENDSHIP

Several strands of social theory provide testable claims about community context andother sources of variation in interracial friendship: structural theories of social associationand homophily studies, in particular, which respectively emphasize opportunity for contactand preferences for in-group over out-group relations; group assimilation theory, which con-siders how immigrant groups incorporate into receiving societies; and group threat theory,which emphasizes the effects of relative group size on intergroup dynamics, includingavoidance.

STRUCTURAL THEORY AND HOMOPHILY

Structural Theories of Association

Central to social-structural theories of association, including friendship choices, is thenotion that both “supply side” and “demand-side” factors matter, that is, that choices toassociate are based both on opportunities for meaningful social contact and individualpreferences and action (Blau, 1977; Marsden, 1990), which vary across the life course. Alarge empirical literature confirms the importance of population makeup, for example,for the structure of interpersonal relations, including interracial marriage, friendships,and crime (Blau and Schwartz, 1984; Blum, 1985; Marsden, 1990; South and Messner,1986). On this opportunity-for-contact dimension, most studies have analyzed effects ofmacrostructures, such as group populations in a metropolitan area or nation, not what Blau(1977) termed substructures—the family, workplace, school, neighborhood, and voluntaryassociations in which most daily life is conducted, where attention is focused on accessiblecontacts.

266

Page 5: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

INTERRACIAL FRIENDSHIPS, CLASS, AND SEGREGATION IN AMERICA

Structures, Large and Small

Marsden (1990) found that individuals were more likely to identify members of other racialgroups in these “core” (confidant) networks if the networks were less kin-based (had alower proportion of family members) and more coworker-based (though these small corenetworks give a limited view of the diversity in respondents’ social ties). No association wasfound between interracial ties and proportion neighbor or group (voluntary association)member. Marsden (1987) found that being young, having a larger network, and living inlarger cities were associated with having more racially diverse contacts. Higher-SES peopletend to have larger networks and be more organizationally active (Putnam, 2000), but itis unclear whether these “returns to status” correlate with interracial ties, or link to oneanother in the same ways, for all racial groups.

If small-scale settings, not just big local pools, matter, what is the racial composition ofthe everyday substructures in which people in America have interpersonal contacts? Mostworkplaces are somewhat racially mixed (Estlund, 2003; Fernandez and Nichols, 2002),more so than most K-12 public schools (Clotfelter, 2004; Frankenberg, Lee, and Orfield,2003) or residential neighborhoods (Logan, 2003). And secular voluntary associations aremore diverse, on average, than religious institutions (Jackson, 1977; Putnam, 2000). Butthese general patterns vary by region, and occupational and other structures can promoteracial segregation within workplaces and other substructures.

Homophily

The folk wisdom that “birds of a feather flock together”—homophily, in the label em-ployed first by Lazarsfeld and Merton (1954)—has been traced back at least as far as theancient Greeks in the Western tradition, and the social-psychological attractions of similar-ity have been confirmed in countless studies. In the United States, this finding holds mostpowerfully for race/ethnicity, followed by class status and religion (review in McPhersonet al., 2001). While homophily shapes many types of relationships, it appears to act morepowerfully on close or strong ties, including marriage and friendships, than on acquain-tanceships or other “weak” ties (Granovetter, 1973; Marsden, 1988). Furthermore, thelimited available evidence on network change over time suggests that bonding ties are notonly more likely to form than bridging ties but less likely to fade as well (McPherson et al.,2001).

Using the GSS, Marsden (1987) found that racial minorities were more likely thanWhites to have significant ties to members of out-groups. In general, White non-Hispanics(Anglos) have the most racially homogeneous networks of any Americans. African Ameri-cans’ and Hispanics’ networks show intermediate levels of homogeneity, and Asian Amer-icans and smaller ethnic groups tend to report networks dominated by the majority racialgroup (McPherson, Miller, and Smith-Lovin, 2001).

Integrating Opportunity Structure and Preferences

The findings outlined above suggest three specific keys to the formation of the ratherexceptional (nonhomophilous) ties that are interracial. The first is “baseline” oppor-tunity for contact (baseline homophily), measured by population or pool sizes, andits significance is clearest in the personal networks of majority-group members: Onereason the average White American’s social ties are mostly to other Whites is that

267

Page 6: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

CITY & COMMUNITY

most Whites live, work, study, worship, play, and mobilize politically in majority-Whitemilieus.

Hypothesis 1, Opportunity for contact (macro structure): In general, interracial exposure infriendships will vary positively with the size of the out-group pool available in a localcommunity. The range will be greater for Whites than for racial minorities, given relativepool sizes for each.

Hypothesis 2, Opportunity for contact (substructure): Interracial exposure in friendships will bemore strongly associated with the workplace and with participation in secular voluntaryassociations than with participation in religious institutions (the survey does not includedata on the makeup of schools attended).

The second key factor (inbreeding homophily) is most evident in minority-group net-works. Inbreeding helps explain the deviation of network composition from what grouppopulation sizes alone would predict if social ties were chosen at random. Even in majority-White milieus, racial minorities tend to report high proportions of coethnics in theirnonkin networks (Tatum, 1987, 1999). The powerful attraction of coethnics reflects avariety of shared traits, such as language (or code or dialect), regional or national ori-gin, tastes and normative disposition, physical appearance (physiognomy and dress), andmore (Laumann, 1973; Blau and Schwartz, 1984). Racial and ethnic minorities appearespecially likely to seek social support from coethnics, a point psychologist Beverly Tatumcaptures vividly in the title of her book, Why Are All the Black Kids Sitting Together in theCafeteria? (1999; and cf. Baerveldt et al., 2004; Menjıvar, 2000). Likewise, in a study ofneighboring in Nashville, Lee and Campbell (1999, p. 136) report, “Despite the potentialfor interracial contact in [racially] mixed areas, over 80% of the ties of black residentsare to black neighbors, while over 95% of the ties of whites in mixed areas are to whiteneighbors.”

A third key factor shaping the formation of bridging ties is others’ preferences, whichshape our own opportunities and preferences.2 Group agency, which I do not model di-rectly, can include pressure to associate within one’s own group, outward pressure fromone’s own group aimed selectively at out-groups deemed desirable (for status advance-ment, historical similarity, or other reasons), exclusionary pressure by out-groups, or allof the above. A history of intergroup hostility (animus), a strong norm of cultural andreligious preservation, perceived economic threat or opportunity, political conflicts orpolarizing episodes, and other factors may all contribute to this kind of collective agencyfor or against intergroup ties.

The racial segregation of neighborhoods, the final substructure of interest, may interactperniciously, with all three factors favoring bonding over bridging, that is, by limitingopportunity for cross-group contact, increasing in-group salience by adding territorial dif-ferences to other cross-group differences, perpetuating negative stereotypes, and more(Massey and Denton, 1993). But thanks to new technologies and other factors, increas-ingly, neighbors tend to be casual contacts rather than socially significant ties (Wellman,1996, 2001; Wellman and Leighton, 1979), and this is especially true where neighbor-hoods are heterogeneous—in race, religion, and other dimensions—and relatively tran-sient (Greenbaum and Greenbaum, 1985; Sampson, 1988; Sampson et al., 1997). Amongadults, only the ties of the poor, physically isolated, and linguistically isolated tend to

268

Page 7: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

INTERRACIAL FRIENDSHIPS, CLASS, AND SEGREGATION IN AMERICA

be highly localized (Briggs, 1998; Fischer, 1982; Smith, 2005; Wellman, 1971; Wilson,1987). Urban Blacks have somewhat more localized support ties than do Whites (Lee andCampbell, 1999; Oliver, 1988).

Very few studies have directly investigated links among neighborhood racial makeup,interracial social ties, and participation in local institutions or substructures in the UnitedStates. First, using a 1975 survey of Philadelphia, Yancey, Ericksen, and Leon (1985)found that residential segregation and friendship segregation (each measured in termsof dissimilarity) among ethnic groups were highly correlated. But the authors remindus that this is partly a matter of selection effects—like friends recommend and chooselike neighborhoods—and not strictly of residence determining social relations. In gen-eral, the researchers found friendship and associational involvement alike to be morelocalized to the neighborhood for respondents with less education who lived in stablecommunities. These were largely working-class White ethnics still clinging to Philadel-phia’s declining industrial base. Lower-status respondents were also more likely to reportprejudice and racially insular friendships. Second, using a 1992 survey of metropolitanDetroit, Welch et al. (2001) found that living in mixed-race neighborhoods, whetherin city or suburb, generally predicted more casual interracial contact, more interra-cial friendships, and less prejudice and stereotyping on the part of both Whites andBlacks.5

European researchers report mixed findings on the association between residentialsegregation and intergroup contacts (net of individual traits), depending on how contactand “ethnic concentration” in neighborhoods are specified (Bouma-Doff, forthcoming;Drever, 2004; Esser, 1986). Although minority segregation rates are much lower in Europethan in the United States, this research generally concludes that residential segregation perse has either a modest effect or no effect on the formation of ties between, say, native Dutchor Germans and immigrant ethnic minorities in their countries. The third hypothesis—a“straw man”—reflects these mixed findings, as well as the conventional wisdom held byadvocates of residential integration:

Hypothesis 3, Segregation in neighborhoods and friendships: A given racial group’s friendshipexposure to out-groups will be positively associated with residential exposure to out-groupsat the neighborhood level, net of other factors.

GROUP THREAT THEORY

There is some evidence for a nonlinear relationship between population (contact pool)diversity and intergroup relations (review in Goldsmith, 2004). These patterns may re-flect a psychological sense of threat or intergroup competition for material resources(Blalock, 1967; Goldsmith, 2004; Moody, 2001; cf. Allport, 1954; Pettigrew, 1998) or othertipping-point phenomena (a) where race becomes salient when a critical mass of out-groupmembers is reached, and (b) where population diversity is so high that every group isa minority, leading to an uneasy accommodation or other responses. Although I do notmodel attitudinal factors, only structural ones, Hypotheses 1, 2, and 3 may capture groupthreat as a source of inbreeding, especially if interracial friendship exposure does not risemonotonically with the size of the out-group pool (Hypothesis 1). I return to this in thesection on results.

269

Page 8: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

CITY & COMMUNITY

GROUP ASSIMILATION THEORY

Assimilation theory considers alternative experiences for immigrant groups in America,as well as contrasts between immigrant outcomes over time and those of native-bornBlacks (involuntary immigrants). Based on the experiences of European immigrants inthe nineteenth and twentieth centuries, traditional assimilation theory predicts that thesocial and economic outcomes of Asians and Hispanics will gradually converge, acrossgenerations, toward an American mainstream (Alba and Nee, 2003; Lee and Bean, 2004),as immigrants and their children acquire language skills, human capital, and social capitalon the road to economic success and social acceptance. But newer theories and evidencesuggest a segmented assimilation (Portes and Rumbaut, 2001; Portes and Zhou, 1993),in which immigrant group segregation may persist over generations or immigrants mayselectively acculturate—assimilating much more in the job market, for example, than inpersonal friendship or other culturally organized spheres of life.

In a study of interracial friendships among adolescents in schools, Quillian and Camp-bell (2003) compared patterns for first-, second-, and third-generation Asians and His-panics, as well as Blacks and Whites. The researchers found weak support for traditionalassimilation theory: there was some convergence over generations, but immigrant youth’sfriendships showed a high degree of in-group homophily; racial similarity was a much bet-ter predictor of friendships, for example, than having parents of the same socioeconomicstatus. While my data do not allow me to test patterns across immigrant generations, thedata include many immigrants as well as direct measures of language isolation and foreigncitizenship among Asians and Hispanics (the number of immigrant Blacks and Whites inthe sample is trivial), suggesting a final hypothesis:

Hypothesis 4, Group assimilation: Asians and Hispanics will not differ significantly from Blacksin the determinants of friendship exposure to Whites.

DATA AND METHOD

I model the structural correlates of interracial exposure in the friendships of Whites,Blacks, Hispanics, and Asians—that is, the odds of having at least one “personal friend”of another race—in 29 U.S. cities and metro areas. The data are from a restricted-use,geocoded version of the Social Capital Community Benchmark Survey 2000 (SCCBS)matched to the 1990 and 2000 censuses.4 The survey included a national survey of adults(N = 3,003) that oversampled for Blacks and Hispanics, as well as 41 “community sam-ples” in selected metropolitan regions and states (N = 26,200). The community samplesranged from 500 to 1,500 respondents each and employed proportionate sampling. Thisrandom-digit-dial phone survey, averaging 26 minutes in length, was conducted betweenJuly and November 2000.5 Overall, the SCCBS achieved an adjusted cooperation rate of42.3 percent for the national sample and 41.6 percent for the combined community sam-ples (Roper Center, 2001); these rates compare favorably with other major phone surveysin recent years, despite the fact that the average SCCBS interview was longer.6

Because the large number of local surveys was made possible by actively engaged localphilanthropic foundations, these funders helped determine the final sample geography.But since some of their “communities” in fact comprise entire states or, conversely, only

270

Page 9: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

INTERRACIAL FRIENDSHIPS, CLASS, AND SEGREGATION IN AMERICA

central cities, I selected 29 of the 41 community samples that matched or closely approx-imated the Metropolitan Statistical Area (MSA or SMSA) boundaries designated by thefederal government as census geographies; this enables more consistent single- and mul-tilevel modeling. Three city-only samples were part of the SCCBS, and for these I use thecentral-city data on population and segregation. The 29 community samples, with N =23,028, are the focus of this study. As local labor markets, metro areas may be treated asmeaningful live–work areas and thus as catchments for local social relations (Wellman,1996), though this localism is probably not constant across locales: more populous areascan meet a wider array of social needs, not just offer a greater diversity of social contacts.

The combined community sample provides statistically adequate sample sizes forAfrican Americans, Asian Americans, and Hispanics on many variables. The latter twogroups include large proportions of foreign-born noncitizens in the sample (34.8 and31.3 percent, respectively) and, in the case of Hispanics, respondents who preferred toconduct the interview in Spanish (36.1 percent). These adequate samples are not avail-able for the minority groups in each of the 29 locales, yet excluding community sampleswith fewer than 25 or fewer than 50 respondents of a given minority group did not affectthe models shown, and I do not report community-specific multivariate results (whereinreliability would be an issue given small ns). In addition, while I am able to estimaterobust, multivariate two-level models for White respondents, only one-level path modelswere estimated for the smaller samples of racial minorities.

All of the models are recursive and use observed, not latent, variables.7 I employed Mplus3.0 to compute multilevel path models (structural equations), also to enable exploratoryfactor analysis and latent variable tests that served as background for the results presented.Mplus generates a maximum likelihood (probit) estimator, with robust standard errors(Muthen and Muthen, 2004). The two-level models directly estimate between-communityversus within-community variation, while the one-level models cluster standard errors toreflect the sample design. All models employ sample weights.

Locales

The 29 surveyed city-regions represent almost one-tenth of the nation’s (331) metro areas.While not nationally representative in the formal sense, the 29 encompass all major censusregions, small and large population centers, and diverse demographic makeup and trendsas well as segregation rates. In preliminary analyses, I employed the two leading segregationmeasures: dissimilarity (D) and exposure (P∗) (Massey, White, and Phua, 1996), while thereported results are limited to exposure (neighborhood racial makeup). I tested a varietyof specifications for out-group presence (pool size): number, log number, percent, andpercent squared. I present the percent measures, as explained in Results.

Friendship Measures

Although the SCCBS survey included an extensive battery of questions on associationalparticipation, the survey includes data on traits of respondents’ social ties (alters) only inthe case of “personal friends.” These data therefore understate the full range of intergroupties that respondents may have. When the interviewer asked about friendship ties, therespondent reported first on the number of close friends (specifying “people you feel atease with, can talk to about private matters, or call on for help”), then the number of

271

Page 10: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

CITY & COMMUNITY

people with whom “you can share confidences,” and then traits of “personal friends”(race, sexual preference, religious affiliation, and economic and social status) consideredas a set, not tie by tie through the more time-intensive name generator approach usedin the General Social Survey and other surveys (Marsden, 1987). We know, for example,whether the respondent has any personal friends who are White, Black/African American,Hispanic, or Asian American (or who own their own business) but not how many suchfriends or what the content of those friendships may be.8

Researchers have found that “friend” is an ambiguous descriptor unless some relationalcontent is specified (Fischer, 1982; Marsden, 1987). This leads to considerable variationand some error when researchers attempt to measure rates of interracial friendship andto compare results over different surveys and different years. Smith (1999) notes thatone-step measures of interracial friendship, such as the measures employed by the SCCBSanalyzed here, which did not define “personal friendship” for respondents, are likely tooverstate actual interracial contact and closeness, for example, because of favorable per-ception bias: the desire to avoid the perception of racial prejudice by reporting “some”friend of the racial out-group or even to associate having a personal friend with being friendlytoward that group. Asking about “personal friends” after specifying relational content for“close friends” should mitigate this bias. Nevertheless, the SCCBS rates should be readas a generous estimate of friendships maintained by Whites as well as other groups, andmeasurement error should not bias analyses of structural determinants of friendship ex-posure significantly. This is the focus of my study, not precise estimates of friendship ratesfor particular communities, nor of the strength or contents of respondents’ interracialfriendships. Finally, about 4 percent of both the national and combined community sam-ples reported that they had no “close” friends, but some of this small group neverthelessreported having “personal” friends. I dropped cases reporting no close friends, since theseoffered no measure of the size of friendship network.9

Covariates

The SCCBS offers rich data on associational life, political participation, and informal so-cializing in the subject communities. Three key indicators are: frequency of socializingwith coworkers (“How many times in the past twelve months have you socialized withcoworkers outside of work?”); a multiitem factor score indicating nonreligious social partic-ipation combining membership with frequency and breadth of involvement in activities(see Table 1 notes); and a similarly constructed index of faith-based participation (text forall items is available online, cf. Roper Center, 2001).10

Achievements vary along the life course, along with choices about residence, social re-lationships, and associational involvement. Though network analyses typically considerlinks among these traits in single-equation multivariate models, the links may be betterconceived and analyzed as path relationships. For example, among the predictor vari-ables available in the SCCBS data, ascriptive traits (race, gender, age) and achieved traits(education, income, labor participation, marital status, parenthood) may exert directand indirect effects on friendship choices, and the pathways may be distinct across racialgroups. I use education, income, and at-home Internet access as indicators of respondent’ssocioeconomic status (Cronbach’s alpha = 0.64); respondent’s occupation and parent’seducation are not available.

272

Page 11: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

INTERRACIAL FRIENDSHIPS, CLASS, AND SEGREGATION IN AMERICA

TABLE 1. MEANS AND STANDARD DEVIATIONS OF COVARIATES, BY RACIAL/ETHNIC GROUP: Personal, household,associational, friendship, and contextual measures (29 locales)

All Groups White Black Hispanic Asian

Mean SD Mean SD Mean SD Mean SD Mean SDIndividual-level traitsAge group (4-level) 2.18 1.06 2.32 1.08 2.06 1.02 1.72 .88 1.64 .82Language of interview

(1 = Spanish).04 .20 .01 .09 .01 .00 .37 .40 .00 .00

U.S. citizen .94 .24 .99 .11 .98 .14 .65 .48 .69 .46Household-level traitsSocioeconomic Status (scale) −.01 .77 .09 .74 −.26 .75 −.51 .74 .45 .67Social participation and network sizeReligious participation (index) −.08 .75 −.07 .76 .10 .72 −.27 .65 −.29 .71Non-religious organizational and

civic participation (factorscore)

.06 1.03 .09 1.02 .11 1.11 −.30 .88 −.17 .87

How often socializes withco-workers (5-level)

1.22 1.28 1.28 1.28 1.09 1.22 .95 1.22 1.22 1.26

Number of close friends (5-level) 3.32 1.09 3.44 1.04 2.91 1.09 3.01 1.20 3.22 1.06Has friend who owns a business

(1 = yes).63 .48 .66 .47 .59 .49 .43 .50 .51 .50

Ecological traitsNeighborhood (1990)Mean percent white, non-Hispanic 77.6 26.4 86.1 17.6 47.7 34.4 57.8 29.0 64.7 24.6Mean percent black, non-Hispanic 11.2 21.6 5.8 12.0 43.8 35.4 11.4 18.8 6.2 11.8Mean percent Hispanic (all races) 6.9 13.1 4.6 9.5 5.1 10.0 23.9 23.3 12.6 14.0Metro area (2000)% white, non-Hispanic 72.3 15.3 74.9 14.6 69.9 11.9 61.8 16.2 58.3 14.5% black, non-Hispanic 11.5 8.6 10.8 8.5 17.4 8.6 10.4 7.0 8.3 5.6% Hispanic 9.5 10.5 8.3 9.5 7.8 9.0 18.1 13.3 16.0 10.2% Asian 5.4 7.1 4.7 6.4 3.8 4.7 8.3 8.3 16.0 9.9Total (N) (20,025) (13,602) (2,409) (1,917) (885)

Coding of indicator variables: Age group in years (18–34, 35–49, 50–64, 65 up); Socioeconomic Status, (meanzero scale), measured by education, income, access to internet at home, alpha = .64; Religious participationfactor score (continuous) combines membership in religious institution, worship attendance, participation inother activities, volunteering; Non-religious organizational activism factor index (Low, Medium, High) combinesmembership in associations, serving as an officer, frequency of recent participation in group activities,participation in community civic events (public meeting, etc.); How often respondent socialized with co-workers in past12 months (Never, Once, 2–4 times, 5–9 times, 10 or more times); Number of close friends (none, 1–2, 3–5, 6–10,more than 10). Some respondents chose not to provide racial/ethnic identity to the interviewer. See onlinecodebook, Social Capital Benchmark Survey 2000, Roper Center for Public Opinion Research, University ofConnecticut. All neighborhood (tract-level) composition measures are for the 1990 census because the Roperdataset includes 1990 geocodes; metro-level measures are 2000 census.

To assert causality, however, solving a system of equations would require a host of in-strumental variables not available in these cross-sectional survey data, and in the analysisof segregation’s effects, such instrumenting has only been accomplished at the aggre-gate (metro) level (Cutler and Glaeser, 1997; Galster, 1987), with a much larger sampleof metro areas than the SCCBS affords. Furthermore, path analysis including discretedata presents unique conceptual and statistical challenges (Winship and Mare, 1983).This further warrants against strong causal claims based on the measures of statisticalassociation that follow, though the models are well identified using accepted criteria(Bollen, 1989).

273

Page 12: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

CITY & COMMUNITY

RESULTS

DESCRIPTIVE RESULTS: INTERRACIAL FRIENDSHIP EXPOSURE

The frequency of interracial friendships varies widely across the 29 communities in thesample and reflects important asymmetries between Whites and minorities, as well assome tendency for immigrant minority group members to be generally isolated fromfriends—even coethnics. Table 2 describes friendship exposure across lines of race interms of the probability that a respondent of race A reports having a personal friendof race A (coethnics); of race B, C, or D (out-groups); or any out-group or all out-groups. Some 79 percent of respondents in the 29-community sample report having atleast one personal friend of some other racial-ethnic group, and more than one-fifth (21percent) report having friends of all three out-groups. Consistent with the opportunity-for-contact principle, relative group sizes matter: members of each racial minority groupare much more likely to report having a personal friend who is White than Whites areto report having a friend of that other race. The degree of this asymmetry is inverselyrelated to the size of the racial minority group. For Blacks, for example, this asymme-try is 74 percent versus 61 percent, but for Asians, it is much sharper: 74 percent ver-sus 38 percent. Results for the national sample (available from the author) are highlycomparable.

Yet one-quarter (25 percent) of Whites reported no interracial friendships at all, andthis measure of racial isolation ranges from a low of 8 percent in Los Angeles to a high of55 percent in Bismarck, North Dakota (data not shown). In addition, Asians and Hispan-ics report more varied interracial ties. These groups are more likely than either Blacks orWhites to report having personal friends of every other major racial-ethnic group; almost40 percent of Asians in the combined community sample report this, for example. Hispan-ics and Asians in these locales are somewhat more likely to have a personal friend of someother race/ethnicity than of their own group: Some 11 percent of Asians and 16 percentof Hispanics reported no personal friends who are coethnics. As one reviewer noted, thisunderscores the point that not all interracial friendships (with particular individuals) can

TABLE 2. FRIENDSHIPS, BY RACE OF RESPONDENT AND ALTER: Within and Across-Group Friendships (Frequencies),29 Communities

P(Aj)/ P(white) P(black) P(Hispanic) P(Asian) P(any other) P(all other)R’s race/ethnicity (N) (N) (N) (N) (N) (N)

White, non-Hispanic .98 .62 .45 .40 .75 .26(13,922) (13,908) (13,884) (13,888) (13,857) (13,857)

Black, non-Hispanic .73 .94 .44 .27 .95 .20(2,394) (2,395) (2,388) (2,393) (2,387) (2,387)

Hispanic (all races) .69 .51 .84 .34 .88 .29(1,863) (1,863) (1,867) (1,863) (1,860) (1,860)

Asian .77 .51 .47 .89 .95 .40(579) (579) (577) (578) (575) (575)

All groups .91 .65 .50 .40 .79 .23(19,914) (19,902 (19,871) (19,875) (18,679) (20,021)

P(Aj) = Probability(respondent of race A has a “personal friend” of race j). Pairwise Ns shown in cells. Theboldface diagonal is co-ethnic ties.

274

Page 13: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

INTERRACIAL FRIENDSHIPS, CLASS, AND SEGREGATION IN AMERICA

function as true bridges to their groups. Moreover, the isolation from coethnics appearsto be a function, at least in part, of an overall isolation from friendship ties. Consistentwith the General Social Survey and other studies, minorities were more likely than Whitesto report having no close friends at all (7.7 percent vs. 2.6 percent, p < 0.001, data notshown); the same is true for foreign-born noncitizens when compared to citizens on thisquestion (9.7 percent vs. 3.7 percent, p < 0.001). In a related vein, minorities are muchmore likely than Whites to report that the membership of their “most important group”is racially mixed or composed mostly of persons of other races (54.4 percent vs. 30.4 per-cent; p < 0.001). This figure is highest for Asians (72.8 percent), followed by Hispanics(56.1 percent) and Blacks (47.4 percent, p < 0.001).

Multiple Forms of Friendship Isolation (or Exposure) Consolidate

I estimated logistic regression equations to examine whether interracial friendship expo-sure makes it more likely that racial minorities in the sample have friends with compara-tively high social status and influence, such as those who own a business, own a vacationhome, or are considered community leaders by respondents, holding respondent’s ownSES constant. The survey data do not allow us to test for bridging race and status in thesame social ties but, more simply, to test whether low-income, less-educated minorities,say, who report having a White friend also tend to have a friend with the abovementionedstatus traits. Table 3 shows odds ratios for minority respondents, including the relativelysmall sample of Asians with less than a high school diploma and less than $20,000 peryear income in 1999. Net of respondent’s own low SES, having an interracial tie is stronglyassociated with having a friendship tie to someone with these status traits (and also tosomeone of a different religious background, data not shown). For example, Blacks whohave White friends are about three times more likely than fellow Blacks at comparableeducation and income levels to have a friend who owns a business. For Hispanics, theanalogous rate is almost five times more likely.11

TABLE 3. CONSOLIDATION OF RACE AND STATUS IN FRIENDSHIPS OF LOWER-SES MINORITY RESPONDENTS: BinomialLogistic Regression Coefficients: Odds of Having a High-Status Friend on Odds of Having a White Friend, byRespondent’s Race

Black Respondents Hispanic Respondents Asian Respondents

Dependent Variables Exp (B) and 95% Confidence Intervals

Has a friend who owns a business 3.00 4.68 model not significant(2.21–4.07) (3.32–6.58)

Has a friend who owns a vacation home 3.34 3.80 9.98(2.19–5.07) (2.50–5.78) (1.15–86.6)

Has a friend who is a community leader 2.65 3.08 6.52(1.96–3.59) (2.12–4.49) (1.71–24.8)

Model chi-squares <.001 <.001 <.01Mean (N) 859 829 90

Notes: all models include controls for income level and educational attainment (income below $20,000/yearand less than high school completion).

275

Page 14: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

CITY & COMMUNITY

The Significance of Community Racial Composition

How does interracial friendship exposure vary by the racial makeup of a person’s com-munity (metro area)? Consider a random-choice model of friendship selection: A Whiterespondent making, say, three random friendship choices in a local area that is 60 percentBlack would have a 1 – (1 – p)N = 94 percent probability of choosing at least one Blackfriend among the three, where p is the out-group proportion in the local population and Nthe number of choices. In this model, the relationship between friendship exposure andracial makeup is curvilinear, with diminishing “returns to diversity” as the out-group shareof population rises. I assume, for purposes of this descriptive model, that the localism offriendships does not vary significantly across locales, that respondents in Baton Rouge,say, are no more and no less likely to have local friends than respondents in Los Angeles,although the latter have more choices overall. In addition, I take the local catchmentarea to be metropolitan, but community diversity might affect respondents’ propensityto “shop” outside their jurisdiction for friends. While racial and ethnic diversity is on therise in most metro areas, using metro-level racial composition even for central-city surveyrespondents is conservative (metro Detroit was 70 percent White in 2000, for example,while the city was just 12 percent White).

Figure 1 plots rates of friendship with Blacks that were reported by White respondentsagainst the percentage of local population that is Black12. The plot employs a locallyweighted scatterplot smoother (loess) to directly detect nonlinear patterns in the data(Cleveland, 1993). The pattern matches the functional form predicted by the random-choice model. The odds of a White respondent reporting some friendship exposure (at

FIG. 1. White friendship ties to blacks: Observed Rates by Community Diversity Across-Group Friendships(Frequencies), 29 Communities

276

Page 15: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

INTERRACIAL FRIENDSHIPS, CLASS, AND SEGREGATION IN AMERICA

FIG. 2. Black friendship ties to Whites: Observed Rates by Community Diversity Across-Group Friendships(Frequencies), 29 Communities

least one friend) across lines of race rise quickly as minority out-group share rises from0 to 10 percent of the local population. The curve then flattens quickly. The same formobtains for White friendships with Hispanics and Asians (plots not shown). Functionalform aside, however, actual rates of interracial friendship exposure generally fall shortof predicted levels except where the minority out-group share is below 10 percent (notshown).13 These results directly suggest why local pool sizes (of racial out-groups) are somuch more predictive for White friendships with other races than vice-versa, consistentwith hypotheses derived from macrostructural theory. This is the probabilistic aspect of ties toother races (for Whites) in a nation where most metropolitan areas are still majority White (note the“stacking” of data points on the left side of Figure 1).

Conversely, Figure 2 plots rates of friendships with Whites that were reported by Blackrespondents, who live, for the most part, in majority-White metro areas. This plot em-phasizes a quite different pattern. For Blacks, increased White population share does notsignificantly reduce racial isolation. The relationship is largely flat, with the Lowess plotdownweighting the extreme outliers of Lewiston and Yakima, which have only a handfulof Black respondents reporting. Results for Asian and Hispanic respondents (not shown)are comparable, except that there is much less variance in White share of metro popu-lation for those two groups as compared to Blacks, leaving only the flat-line portion ofthe relationship evident in Figure 2. These results suggest isolation from Whites (evenone friend) that may owe to preferences (inbreeding), racially segregated substructuresof potential contact, or other factors.

In preliminary analyses of the same survey data, Putnam (2003) has obtained resultscomparable to those for Figures 1 and 2 above, using a Herfindahl index of diversity rather

277

Page 16: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

CITY & COMMUNITY

than race-group–specific measures to measure actual versus predicted interracial friend-ship rates. Yes, White residents of Los Angeles are far more likely than their counterpartsin Maine to have a friend of another race but not nearly as much as the racial compositiondifferences between the two places would predict. What further differences in communitycontext, then, or in the substructures of social participation and neighborhood segrega-tion, might explain this?

MULTIVARIATE MODELS OF INTERRACIAL FRIENDSHIP EXPOSURE

I estimated regression equations for measures of social participation for all racial groups,including as independent variables ascriptive traits (gender, age group) and attained traits(marital status, presence of children, SES, employment status), as well as census region,community percent non-White (metro-level), suburban residence, and neighborhoodmakeup (race, homeownership). Prior research has shown these to be associated witha range of social participation measures, such as membership in groups, churchgoing,and socializing (review in Putnam, 2000). Selecting statistically significant relationshipsfor the next stage of the analysis, the single-equation regression models informed the moreparsimonious but computationally intensive structural equation (path) models discussedbelow.

Friendship Exposure for Whites (to Racial-Ethnic Minorities)

Figure 3 shows standardized probit coefficients for predictors of exposure by Whites toBlacks in a two-level model. All of the coefficients are highly significant, and multiple in-dices confirm a good fit. This model estimates (a) direct effects of neighborhood makeup(percent Black), social participation measures, and SES on intergroup friendship expo-sure; (b) effects of SES that are mediated by those variables; and (c) patterns of associationamong the mediators.

First, at the within-community level of the diagram, socializing with coworkers, size offriendship network, and organizational participation have the largest and most robustdirect associations with friendship exposure (religious participation had little effect andso was dropped from the two-level model to conserve degrees of freedom). Hypothesis2 is strongly supported. Direct effects of SES on friendship exposure are modest oncethese mediating variables are accounted for. Second, the multiple forms of social partic-ipation are strongly correlated. White joiners tend to have more friends and to socializemore with coworkers, and all three factors are powerfully associated with friendship ex-posure across race lines. White joiners are also more likely than noninvolved Whites tolive in racially mixed neighborhoods. Third, SES shows quite different links to social par-ticipation versus neighborhood diversity. For Whites, higher SES is strongly predictiveof greater social participation. But higher-SES Whites are, as expected, less likely thanlower-SES Whites to share neighborhoods with Blacks. Below, the one-level models pro-vide a calculus of paths to reveal the relative magnitudes of key direct versus indirecteffects.

For the odds that a White person in the United States will report having a friendof another race, the single most important predictor is where that White person lives

278

Page 17: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

INTERRACIAL FRIENDSHIPS, CLASS, AND SEGREGATION IN AMERICA

FIG. 3. White-to-Black Friendship Exposure: Two-Level AnalysisNotes: Log-likelihood: -128,534. AIC: 257,111. BIC: 257,269. Free parameters = 21. R2 = .09 (within) and .60(between). Standardized coefficients with unstandardized t’s in parentheses.

in the country—more specifically, the whiteness of their metro area home (consistentwith the bivariate plots above). The observed differences in metro-area racial expo-sure between communities explain 60 percent of the variance in White friendship ex-posure to Blacks, 71 percent for White friendship exposure to Hispanics, and 39 per-cent for White friendship exposure to Asians (data not shown). The within-level fac-tors explain much less variance at that level: 9.3 percent, 7.6 percent, and 13.2 percent,respectively.

Table 4 presents one-level results for White friendship exposure to all three out-groups,by race of friendship contact (alter), affording a more detailed anatomy of the individualtraits associated with exposure. Here, metro-level racial makeup is a covariate, and stan-dard errors are clustered by locale. Older Whites are more racially insular, net of otherfactors, and the effect is particularly robust and large for White seniors (age 65 and over).Whites who have Black and Hispanic friends tend to be more residentially integrated withthose groups as well, and nonreligious joining remains highly significant. As for SES, thecalculus of paths shows direct effects of SES, as well as total indirect and specific indirecteffects of the mediating variables. For Whites, higher SES has no significant, direct associationwith friendship exposure to Blacks. The indirect effects via these mediators, however, arelarge and highly significant: For Whites, class effects are closely associated with joining,in the broad sense. Likewise, important life course effects are clearly mediated by socialparticipation (data not shown): Large, indirect effects of age (being 65 or older) onfriendship exposure are mediated by elders having fewer friends overall, joining less,and socializing far less with coworkers (particularly in the case of those no longerin the labor force). Taken together with the models below, these results for White-to-Black friendship exposure reflect the curvilinear functional form described for the simple

279

Page 18: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

CITY & COMMUNITY

TABLE 4. FRIENDSHIPS OF WHITE RESPONDENTS WITH MINORITY OUT-GROUPS: Probit regression coefficients, by Raceof Out-Group Alter (one-level models)

Standardized coefficients with unstandardized t-statistics in parentheses

Black Friends Hispanic Friends Asian FriendsIndependent Variable (Model 1) (Model 2) (Model 3)

Age 35 to 49 years −.033+ −.024 −.080∗∗∗(1.84) (1.33) (7.94)

Age 50 to 64 years −.028+ −.024 −.063∗∗(1.97) (1.52) (4.04)

Age 65 years or older −.117∗∗ −.104∗∗ −.118∗∗∗(5.68) (4.97) (6.44)

Religious involvement (index) .035+ .004 .014(1.81) (0.309) (1.23)

Non-religious involvement (factorscore)

.118∗∗∗ .127∗∗∗ .150∗∗∗

(8.40) (9.59) (13.3)Number of close friends .188∗∗∗ .117∗∗∗ .137∗∗∗

(14.4) (7.53) (11.5)How often socializes with

co-workers (5-level).140∗∗∗ .114∗∗∗ .075∗∗

(6.64) (5.78) (3.42)Percent out-group, neighborhood .106∗∗∗ .113∗∗ −.038

(8.22) (4.72) (0.177)Percent out-group, metro area .336∗ .583∗∗ .759∗∗

(2.39) (4.90) (4.18)Percent out-group, metro area,

squared−.219+ −.327∗∗ −.487∗∗

(1.80) (3.09) (5.73)Socioeconomic StatusDirect Effect .016 .023 .125∗∗∗

(1.39) (1.50) (8.32)Total indirect effects .081∗∗∗ .065∗∗∗ .078∗∗∗

(11.7) (13.3) (12.2)Specific indirect effects:Religious involvement .006+ .001 .003

(1.72) (.308) (1.230)Non-religious involvement .035∗∗∗ .038∗∗∗ .045∗∗∗

(7.42) (8.20) (10.3)Number of close friends .024∗∗∗ .015∗∗∗ .018∗∗∗

(7.85) (6.05) (8.04)Socializes with co-workers .022∗∗∗ .018∗∗∗ .012∗∗

(5.90) (5.79) (3.48)Percent out-group, neighborhood −.007∗∗ −.007∗∗ .000. (-3.27) (4.04) (.177)Degrees of freedom 3 12 13CFI 1.00 1.00 1.00TLI 1.00 .99 1.00RMSEA .00 .01 .00WRMR .159 .332 .085R-square .15 .22 .21(N) (13,522) (13,498) (13,704)

Notes: +P < .10; ∗P <.05; ∗∗P<.01; ∗∗∗P<.001 (two-tailed tests). CFI = Comparative fit index, TLI = Tucker-LewisIndex, WRMR = Weighted root mean square residual, RMSEA = Root mean square error of approximation.Standard errors are clustered. For AGE, excluded category = young adult (age 18 to 34 years).

280

Page 19: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

INTERRACIAL FRIENDSHIPS, CLASS, AND SEGREGATION IN AMERICA

bivariate plots above (as do results, not shown, wherein tract and metro-level out-grouppresence is specified by log number rather than percent; the former fits better only in theWhite-to-Black model, and so percentages are shown for consistency and simple interpre-tation). Net of other factors, there are rapidly diminishing “returns” to increased localdiversity.

The patterns for White-to-Hispanic (model two) and White-to-Asian friendship expo-sure (model three) are very similar to White-to-Black friendship exposure in the effectsof age, social participation, and the latter as a mediator of class effects. But Asian Amer-icans are the most geographically concentrated minority group in the sample. As pre-dicted in the discussion of scatterplot results above, metro-level percent Asian is the dom-inant factor in explaining variation in White friendship exposure to members of thisgroup.

It is the composition of SES effects in the White-to-Asian model that differs sharply fromthat in the other two models. The direct association between SES and friendship expo-sure to Asians is significant and positive, and for friendship exposure to Asian friends,the direct effect of SES is much larger than the sum of all indirect effects measured.These results should be interpreted cautiously, since “direct” here means not medi-ated by any of the mediator variables included in the models. One possibility is thatthe range of mechanisms linking White SES to friendships with Asians is not capturedas well by a model emphasizing social participation factors and neighborhood of res-idence as mediators. It is also possible that the observed effect reflects Asians’ higheraverage class standing than Blacks or Hispanics. In general, specific indirect effectsof White SES on friendship exposure are very similar for the three models, in direc-tion, significance, and magnitude. But residential segregation is a striking exception:The pattern for Hispanic alters matches that for Blacks—higher-SES Whites are lessresidentially integrated with Hispanics, and this is associated with more racially insularfriendships for Whites (supporting Hypothesis 3). The same is not true for exposure toAsians.

Taken together, the one- and two-level models of Whites’ interracial ties are highlyconsistent with Hypotheses 1, 2, and 4. Nonreligious joining, having more friends, andsocializing with coworkers all mediate the positive returns to SES and are much larger thaneffects of religious involvement. But these patterns encompass distinct subgroups of Whites:educated, higher-income, middle-aged married people are the primary joiners of groups(Putnam, 2000). On average, they have more friends and are more involved in groups bothreligious and secular. For Whites and other racial groups, those who reported socializingfrequently with coworkers are more often young (age 18 to 34), male, and single. Notably,this is also the demographic group most likely to move into mixed-race neighborhoods;young single renters are the nation’s primary White integrators (Ellen, 2000). In general,Whites who report having a Black or Hispanic friend are likely to share neighborhoodswith those groups as well. Finally, the long “civic generation” highlighted in Putnam’s(2000) research and the popular media in recent years is, in the case of White Americans,also the most racially insular.

Friendship Exposure for Blacks, Hispanics, and Asians (to Whites)

Two-level models (not shown, available from the author) match the White-to-minority,two-level models I reported above, except that within-level factors—respondent’s class

281

Page 20: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

CITY & COMMUNITY

status and social participation—explain much more of the within-level variance in friend-ship exposure to Whites than these factors explain for White friendship exposure to mi-norities: 14.6 percent of said variance for Black respondents, 32.1 percent for Hispanics,and 25.3 percent for Asians. The effect of variance in metro-level out-group size (Whitepresence) is trivial, consistent with the limited variance evident in Figure 2.

The one-level models allow a concise comparison of patterns for minority-group respon-dents, including factors of special relevance to the social worlds of immigrant minorities,such as citizenship and language use. Multiple indices indicate a strong fit in the threemodels in Table 5. Nonreligious joining and socializing with coworkers show strong posi-tive associations with friendship exposure to Whites for all three minority groups. These,then, appear to be the most universal correlates of friendship exposure, consistent withHypothesis 2. These factors are robust and large as predictors of both White-to-minorityand minority-to-White friendship exposure, and in the case of immigrant minorities, beingEnglish proficient and being a citizen are strong predictors of joining, in particular (p <

0.001, data not shown). Likewise, important SES effects on interracial friendship exposureare mediated by joining, broadly defined.

Yet direct SES effects on friendship exposure are large and positive for the Asians andHispanics and modest for Blacks, for whom the effects mediated by substructures repre-sent some 70 percent of the total SES effect (compared to 33 and 28 percent for Hispanicsand Asians, respectively). Residential integration appears to be a robust marker of friend-ship exposure to Whites only in the case of Black respondents, for whom neighborhoodintegration has a very large direct effect (comparable in order of magnitude to the ef-fects of social participation variables) and also acts as a mediator of the effects of socialclass on friendship exposure to Whites. However, in reduced-form models that excludecitizenship, Hispanic residential integration with White non-Hispanics is associated withfriendship exposure (p < 0.05).

For Hispanics, citizenship is a marker of intergroup friendship and attainment: higherSES, joining, socializing, and friendship networks account for almost 40 percent of theassociation between citizenship and Hispanics’ friendship exposure to Whites (data notshown, and comparable results obtain when language use—asking for a Spanish-languageinterview—replaces citizenship as the predictor variable). There is no such pattern forAsians, though a sizeable share of Asian respondents (35 percent), like Hispanic ones(31 percent), were noncitizens.

Consistent with the scatterplots, macrostructural opportunity for contact (Hypothe-sis 1) is much less dominant in explaining minority friendship exposure to Whites in thesample—a reflection of the fact that most of the communities are majority White (again,at the metro level). This is not to say that the White share of population is unimportantfor racial minorities’ friendship exposure to Whites, only that such population variationis limited in the sample, as in the nation as a whole. Finally, Hypothesis 4, which pre-dicted parallel patterns for Asians and Hispanics in comparison to Blacks, was partiallysupported and Hypothesis 2 strongly so. Social participation factors are robust predic-tors of interracial friendship exposure for all groups, so Asians and Hispanics, many ofwhom are immigrants or the children of immigrants, are not only like Blacks but also likeWhites in this respect. Differences between Blacks and other minority groups are limitedto residential integration (Hypothesis 3 is supported only for Blacks), language use, andcitizenship.

282

Page 21: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

INTERRACIAL FRIENDSHIPS, CLASS, AND SEGREGATION IN AMERICA

TABLE 5. FRIENDSHIPS OF MINORITY RESPONDENTS WITH WHITES: Probit regression coefficients, by Respondent’sRace (one-level models)

Standardized coefficients with unstandardized t-statistics in parentheses

Black Respondents Hispanic Respondents Asian RespondentsIndependent Variable Model 4 Model 5 Model 6

Age 35 to 49 years −.023 −.039 −.129+(.536) (1.77) (2.42)

Age 50 to 64 (50 up for Hispanicsand Asians)

.056 −.030 −.076

(1.71) (1.33) (1.07)Age 65 or older .119∗ .101 .001

(2.62) (1.89) (.032)Respondent is a citizen – .213∗∗∗ .091

(6.85) (1.73)Religious involvement (index) .000 .072 .107

(.002) (1.55) (1.69)Non-religious involvement (factor

score).201∗∗∗ .115∗∗ .204∗∗

(6.24) (4.28) (4.36)Number of close friends .177∗∗∗ .172∗∗∗ .151∗

(5.35) (6.69) (3.19)How often socializes with

co-workers (5-level).173∗∗∗ .213∗∗ .194∗

(7.31) (4.21) (3.13)Percent white non-Hispanic,

neighborhood.167∗∗ .022 .018

(3.94) (.595) (.204)Percent white, metro area −.320∗ −.081 −.395

(2.55) (.425) (1.05)Percent out-group, metro area,

squared.342∗ .163 .573

(2.81) (.761) (1.57)Socioeconomic StatusDirect Effect .082∗∗ .237∗∗∗ 0.259∗∗∗

(3.14) (7.46) (6.41)Total indirect effects .170∗∗∗ .146∗∗∗ .126∗∗

(8.76) (6.99) (5.28)Specific indirect effects:Religious involvement −.000 .010 .003

(.002) (1.28) (.809)Non-religious involvement .058∗∗∗ .039∗∗ .025∗∗

(4.36) (4.13) (4.25)Number of close friends .028∗∗∗ .035∗∗ .038∗

(5.77) (4.15) (3.34)Socializes with co-workers .044∗∗∗ .057∗∗ .057∗

(7.22) (4.58) (3.37)Percent white, neighborhood .039∗ .005 .003

(3.09) (.577) (.213)Degrees of freedom 11 10 6CFI 1.00 .99 .99RMSEA .018 .024 .028WRMR .244 .230 .236R-square .19 .41 .34(N) (2,310) (1,723) (567)

Notes: +P < .10; ∗P < .05; ∗∗P < .01; ∗∗∗P < .001 (two-tailed tests). And see note below Table 4.

283

Page 22: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

CITY & COMMUNITY

DISCUSSION

Prior research has indicated why interpersonal bridges across major social divides areso important as America becomes more ethnically diverse and at the same time moreeconomically unequal. By various measures, Whites’ friendships with members of otherracial groups have increased sharply over the past generation, and the fact that youngerWhites are more likely to report interracial ties bodes well for the future—despite the factthat those ties are more likely than same-race ties to fade over the life course. For all racialgroups, “joining” in the broad sense—involvement in nonreligious groups, socializing withcoworkers, and having more friends overall—is the structural factor robustly associatedwith interracial exposure in friendships. And these factors channel much of the friendshipadvantages of higher class status. For Whites, all of the positive association between higherstatus and interracial friendships is mediated by joining.

But the results also emphasize the continuing importance and range, for White Ameri-cans, of opportunity for contact based on racial settlement patterns. For Whites, the oddsof having an out-group friend increase rapidly as the metro population goes from all Whiteto “just” mostly White (10 percent other) and then flatten quickly. Across communities,White isolation from friendships with minorities—in a nation where most metro areas arestill non-Hispanic White by a large majority and only large metros are becoming major-ity minority—is shaped significantly by this macrostructural factor. Conversely, for racialminorities, access to “substructures” of contact—participation in secular civic groups andwork-based socializing, in particular—is most important as a predictor of having a Whitefriend.

Net of class status, joining, and other traits, residential segregation and racial isolation infriendships, show some association—mainly in ways that support immigrant assimilationtheory. Among Whites, sharing neighborhoods with a given racial out-group is stronglyassociated with friendship exposure to that group in the cases of exposure to Blacks andHispanics but not Asians; this is another indicator of Asians’ comparatively advantagedposition vis-a-vis interpersonal relations with White Americans (Lee and Bean, 2004).Among minorities, living in whiter neighborhoods appears to be a unique marker for thesocial relationships of Blacks. It is also a controversial marker, inside and outside the Blackcommunity, as discussions of “integration fatigue” and social engineering challenge theintegrationist ideal (Charles, 2005).

Higher-status minorities of all groups are more likely to report White friends, but thislink is much weaker for Blacks than for immigrant groups and more likely to be mediatedby joining, which may suggest less opportunity and/or less effort by Blacks to createand sustain friendships with Whites outside of organized social groups. Higher-statusBlacks have more opportunity for interracial contact than lower-status Blacks but may alsodiffer in the conscious choice to cultivate more diverse contacts, that is, in preferences forsocial integration. Other researchers have emphasized the power of conscious choicesto integrate—or not—with Whites, including patterns of partial and “strategic” socialassimilation by middle-class Blacks (Lacy, 2004) and by immigrant minorities (Portes andZhou, 1993).

The positive paths linking class status to having friends of another race signal doubletrouble in the sense that people with less education and lower income are doubly dis-advantaged in the market for bridging ties and the social capital those ties can channel,with both less opportunity for contact and fewer resources with which to manage active

284

Page 23: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

INTERRACIAL FRIENDSHIPS, CLASS, AND SEGREGATION IN AMERICA

friendships that cross social borders. As for the former, lower-status people are less organi-zationally active and have smaller friendship networks. If low-skilled, they are also less likelyto work and thus to have racially diverse coworkers with whom to socialize. Finally, lower-status minorities are more likely to live in racially segregated neighborhoods and, if theyare employed, to work in racially segregated workplaces. Yet those who do report havingWhite friends are those whose social worlds also include people of comparatively higherstatus and influence, for example, business owners and community leaders. Less educated,lower-income Blacks, Hispanics, and Asians who “bridge” on one important dimension(race) often bridge on others (economic status and community influence) as well.

These findings underline the importance of strengthening civic life—for purposes ofbridging, not just bonding—in what Arthur Schlesinger (1944) famously termed “a nationof joiners” but where contemporary research has charted marked decline in associationallife and other traditional indicators of connectedness or social capital (Paxton, 1999;Putnam, 2000). The effects of socializing with coworkers highlight the importance of theworkplace as a social institution for crossing racial and other divides (Estlund, 2003).Some workplaces are quite segregated by race/ethnicity, of course, and this appears tovary by locale according to industrial structure. Workplaces can also be conflict-ridden.But like organized associations, workplaces show multiple features, missing or fragilein neighborhoods and other loosely bounded substructures, to encourage relationshipsacross social borders: tangible forms of interdependence and collective identity buttressedby material incentives, shared norms, and authority structures for learning and enforcingthose norms.

For these reasons and more, more research is needed on the causal mechanisms thatunderlie the structural patterns uncovered in this study and on the links among networks,associational involvement, trust, political and social attitudes, lifestyle, and experiencesof discrimination in varied contexts. Using the same survey data, for example, Putnamhas found that social trust and socializing are lower—for all racial/ethnic groups—inracially diverse local communities (Putnam, 2007; Saguaro Seminar, 2001). Majority- andminority-group persons living in such communities, when compared to counterparts liv-ing in more homogeneous settings, even appear to be less likely to trust members of theirown ethnic group. On one hand, this is consistent with decades of empirical research andtheoretical work on cities as diverse environments (Ross, Mirowsky, and Pribesh, 2002)and with the general finding, for workgroups and institutions, that diversity strains soli-darity, making heterogeneous workgroups less effective, on average, than homogeneousones (Estlund, 2003). People who live in cities are generally less trusting than counter-parts living in more socially homogeneous suburbs and rural towns, and however rationalthat may be, this trust deficit holds enormous implications for how society functions(or dysfunctions).

But diverse environments also make it more possible to create social bridges in waysthat are important for each of us and for society as a whole. The evidence is that fornow, we have not marshaled the bridging institutions—voluntary associations, workplaces,schools, and more—equal to that urgent project of connection. If twentieth-century socialscience, and sociology most of all, grappled with the implications of a massive populationshift from countryside to city in rapidly industrializing societies, then a crucial puzzlefor twenty-first-century social research centers on the trade-offs entailed—the good, thebad, and the unexpected—as cities dominate settlement patterns worldwide and becomeethnically diverse to an unprecedented degree.

285

Page 24: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

CITY & COMMUNITY

Notes

1 Both race and ethnicity have been employed in the social network literature (see, e.g., Fischer, 1982; Marsden,

1987), but race is more commonly employed in reference to residential segregation in the United States (see,

e.g., Lewis Mumford Center, 2001; Massey, 2001). In such usage, race denotes the five principal classifications that

combine, however imperfectly, racial and ethnic identification: white non-Hispanic (Anglo); black non-Hispanic;

ethnic Hispanic (of any race); Asian American/Pacific Islander; and Native American/American Indian.2 Citing longitudinal linguistic research, for example, Massey and Denton (1993, pp.162–165) note that Black

English has become more uniform among blacks over the past few decades—reflecting a more distinct group

boundary. It is the only spoken code or dialect in America that has shown this pattern, a testament to the uniquely

high and persistent rates of segregation faced by blacks.3 Welch et al. also found churchgoers more likely to have interracial ties, but since the analyses did not control

for other, non-religious involvement, it may be that the more organizationally active overall have more bridging

friendships and that some of these people include religious activity in their lives.4 In the models, I used Census 2000 data on racial make-up at the metro level, since these data were collected

in the SCCBS survey year (1999) and thus offer the most contemporary measures of the local “pool” of contacts

for survey respondents. But since the SCCBS data included only 1990 tract geography, the tract-level measures

are from the 1990 Census. Friendship formation lags changes in the pool of potential contacts.5 The survey was conducted by TNS Intersearch for Harvard University’s Saguaro Seminar on Civic Engage-

ment in America, led by Robert Putnam. Data are available on-line through the Roper Center for Public Opinion

Research Archive, University of Connecticut.6 These rates compare favorably with random digit dial (RDD) phone surveys conducted by major polling

organizations and leading survey research firms, although the average SCCBS interview was 50 percent longer

(in call time) than the average for those others (Holbrook, Krosnick, and Pfent, n.d.). The adjusted cooperation

rate, which varied across community samples in the SCCBS, is the share of eligible respondents reached who

agreed to participate and completed the interview. Cooperation with telephone surveys has dropped in the

past two decades even as costs of face-to-face interviewing have risen sharply (Weisberg, Krosnick, and Bowen,

1996). Phone surveying is much more cost effective, but how representative, reliable, and valid are the data?

The SCCBS samples are carefully weighted to reflect age, gender, education level, and ethnic composition of

the local and national populations. In addition, to maximize both contact and cooperation rates, the SCCBS

survey personnel made up to 10 call-backs for all but “hard” refusals (Roper Center, 2001). But failed contact

and cooperation raise the question of response bias in the specific relationships of interest in the study. The

most careful studies of response rate and data quality across multiple surveys and survey modes suggest that

nonresponse bias is typically modest in comparison to social desirability, satisficing, and other sources of bias in

survey research (Chang and Krosnick, 2003; Green and Krosnick, 2001; Holbrook, Green, and Krosnick, 2003;

Holbrook, Krosnick, and Pfent, n.d.). In addition, Robert Putnam reports, in personal communication with

the author, that “the harder interviews were only modestly different from the easy ones,” that “the people who

were harder to find at home” (limiting the contact rate) were somewhat more likely to be young, black, and

urban, while “the people who were harder to convince to take the interview” (limiting the cooperation rate)

were slightly older, whites, and more prejudiced. “But these differences were modest,” he notes. For this study, I

found no significant associations between cooperation rates, which range from 30 percent to 57 percent across

the 29 locales, and key variables or effects of interest.7 I ran tests of several theoretical constructs, including “race bridging” as a multidimensional construct com-

posed of network, trust, and group-activism elements, using Mplus’ latent variable and exploratory factor analysis

functions. These provide a background for analysis; data are available from the author.8 On a more technical note, the SCCBS network data are ego-centric; we do not know whether and how

respondents’ friends are tied to each other or whether their friendship choices are reciprocated. While the first

286

Page 25: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

INTERRACIAL FRIENDSHIPS, CLASS, AND SEGREGATION IN AMERICA

point is unimportant for a direct assessment of interracial exposure in the respondent’s own first-order ties,

earlier research indicates that homophilous ties are not only more common in respondents’ networks but more

likely to know each other (McPherson et al., 2001). This is another confirmation of the more dispersed quality of

non-homophilous ties, a quality that tends to make them social bridges (Granovetter, 1973). The reciprocation

point emphasizes advantages of studying friendships dyads, rather than individuals, as the unit of analysis (e.g.,

Baerveldt et al., 2004; Quillian and Campbell, 2003).9 While the data and prior research (Smith, 1999) indicate clearly that respondents interpreted the items for

“close friend” and “personal friend” differently, the number of close friends represents our closest approximation

of the size of a respondent’s overall friendship network (N). Because close friends are fewer in number, the

approximation is a conservative one.10 Because reporting errors frequently obtain when respondents are asked to estimate totals, data for the total

number of close friends was collected categorically (no close friends, 1–2, 3–5, 6–10, more than 10). To model

random-choice versus observed rates of bridging (see findings), I recoded the variable to select on the mean

of each category greater than zero (i.e., 1.5, 4, 8 for the middle categories) and to impute a value of 12 as a

conservative estimate of the mean for “more than 10.”11 Models for Asian Americans, for whom the sample size available on these survey items is about one-tenth

that of the other racial minority groups, were generally not significant.12 Community (MSA) abbreviations in Figures 1 and 2: Atl (Atlanta, GA), Bat (Baton Rouge, LA), Birm

(Birmingham, AL), Bis (Bismarck, ND), Bld (Boulder-Longmont, CO), Bos (Boston, MA city), Char (Charlotte-

Gastonia-Rock Hill, NC-SC), Chi (Chicago, IL), Cin (Cincinnati, OH-KY-IN), Clv (Cleveland-Lorain-Elyria, OH),

Den (Denver, CO), Det (Detroit, MI), Hou (Houston, TX), Grn (Greensboro-Winston Salem-High Point, NC),

GrR (Gran Rapids, MI city), Kal (Kalamazoo-Battle Creek, MI), Knx (Knoxville, TN), Lew (Lewiston-Auburn,

ME), Los (Los Angeles-Long Beach, CA), Man (Manchester, NH), MSt (Minneapolis-St. Paul, MN-WI), Phx

(Phoenix-Mesa, AZ), Rch (Rochester, NY), SDg (San Diego, CA), Sea (Seattle, WA city), SFo (San Francisco,

CA), Syr (Syracuse, NY), Yak (Yakima, WA), Yrk (York, PA).13 The degree of potential correction varies with percent outgroup in the metro. For example, in a metro area

that is 10 percent outgroup, an average in-group reporting of 2 “close” friends in the friendship network yields

a random-choice predicted bridging rate of 19 percent, but if the (corrected) number of “personal” friends is

actually 4, the rate is 34 percent, if 6, then 47 percent, and if 10, then 65 percent. The conservative interpretation

(in text) focuses on the direction of the deviation from baseline indicated in the scatterplots rather than the degree

of deviation at particular levels of community racial diversity.

References

Alba, R., and Nee, V. 2003. Remaking the American Mainstream. Cambridge, MA: Harvard University Press.Alesina, A., and La Ferrara, E. 2002. “Who Trusts Others?” Journal of Public Economics 85, 207–224.Allport, G. W. 1954. The Nature of Prejudice. Cambridge, MA: Addison-Wesley.Baerveldt, C., Van Duijn, M. A. J., Vermeij, L., and Van Hemert, D. A. 2004. “Ethnic Boundaries and Personal

Choice: Assessing the Influence of Individual Inclinations to Choose Intra-Ethnic Relationships on Pupil’sNetworks,” Social Networks 26, 55–74.

Blalock, H. M. 1967. Toward a Theory of Minority Group Relations. New York: Wiley.Blau, P. 1977. Inequality and Heterogeneity. New York: Free Press.Blau, P. M., and Schwartz, J. E. 1984. Cross-Cutting Social Circles: Testing a Macrostructural Theory of Intergroup

Relations. Orlando, FL: Academic Press.Blum, T. C. 1985. “Structural Constraints on Interpersonal Relations: A Test of Blau’s Macrosociological Theory,”

American Sociological Review 47, 45–61.Bollen, K. A. 1989. Structural Equations with Latent Variables. New York: Wiley.Bouma-Doff, W. V. D. L. 2007. “Confined Contact: Residential Segregation and Ethnic Bridges in the Nether-

lands,” Urban Studies 44(5–6), 997–1017.

287

Page 26: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

CITY & COMMUNITY

Briggs, X.d.S. 1997. “Social Capital and the Cities: Advice to Change Agents,” National Civic Review 86(2), 111–117.Briggs, X.d.S. 1998. “Brown Kids in White Suburbs,” Housing Policy Debate 9(1), 177–221.Briggs, X.d.S. 2004. “Civilization in Color: The Multicultural City in Three Millennia,” City & Community 3(4),

311–342.Charles, C. 2005. “Can We Live Together?: Racial Preferences and Neighborhood Outcomes,” in X. de Souza

Briggs (ed.), The Geography of Opportunity, pp. 45–80. Washington, DC: Brookings Institution Press.Chang, L., and Krosnick, J. 2003. “Comparing Oral Interviewing with Self-Administered Computerized Ques-

tionnaires,” Unpublished manuscript, Ohio State University, Columbus, Ohio.Cleveland, W. S. 1993. Visualizing Data. Summit, NJ: Hobart Press.Clotfelter, C. T. 2004. After Brown: The Rise and Retreat of School Desegregation. Princeton, NJ: Princeton University

Press.Crain, R. L., and Wells, A. S. 1994. “Perpetuation Theory and the Long-term Effects of School Desegregation,”

Review of Educational Research 64(4), 531–553.Cutler, D., and Glaeser, E. 1997. “Are Ghettos Good or Bad?” Quarterly Journal of Economics 112, 827–882.Dickens, W. T. 1999. “Rebuilding Urban Labor Markets: What Community Development Can Accomplish,” in

Ronald F. Ferguson and William T. Dickens (eds.), Urban Problems and Community Development, pp. 381–436.Washington, DC: Brookings Institution Press.

Dominguez, S., and Watkins, C. 2003. “Creating Networks for Survival and Mobility: Social Capital amongAfrican-American and Latin American Low-Income Mothers,” Social Problems 50(1), 111–135.

Drever, A. I. 2004. “Separate Spaces, Separate Outcomes?: Neighborhood Impacts on Minorities in Germany,”Urban Studies 41, 1423–1439.

Durkheim, E. 1893 [1993]. The Division of Labor in Society. New York: MacMillan.Ellen, I. G. 2000. Sharing America’s Neighborhoods: Prospects for Stable Racial Integration. Cambridge, MA: Harvard

University Press.Erickson, B. H. 1996. “Culture, Class, and Connections,” American Journal of Sociology 102(1), 217–251.Esser, H. 1986. “Social Context and Inter-Ethnic Relations: The Case of Migrant Workers in West German Areas,”

European Sociological Review 2, 30–51.Estlund, C. 2003. Working Together: How Workplace Bonds Strengthen a Diverse Democracy. New York: Oxford University

Press.Fernandez, M., and Nichols, L. 2002. “Bridging and Bonding Capital: Pluralist Ethnic Relations in the Silicon

Valley,” International Journal of Sociology and Social Policy 22(9/10), 104–122.Ferrand, A., Mounier, L., and Degenne, A. 1999. “The Diversity of Personal Networks in France: Social Strati-

fication and Relational Structures,” in B. Wellman (ed.), Networks in the Global Village, pp. 185–224. Boulder,CO: Westview Press.

Fischer, C. 1982. To Dwell among Friends: Personal Networks in Town and City. Chicago: University of Chicago Press.Frankenberg, E., Lee, C., and Orfield, G. 2003. A Multiracial Society with Segregated Schools: Are We Losing the Dream?

The Civil Rights Project, Harvard University, Cambridge, MA, January.Galster, G. C. 1987. “Residential Segregation and Interracial Economic Disparities: A Simultaneous-Equations

Approach,” Journal of Urban Economics 21, 22–44.Gittell, R., and Vidal, A. 1998. Community Organizing: Social Capital as a Development Strategy. Thousand Oaks, CA:

Sage Publications.Goldsmith, P. A. 2004. “Schools’ Role in Shaping Race Relations: Evidence on Friendliness and Conflict,” Social

Problems 51(4), 587–612.Granovetter, M. 1973. “The Strength of Weak Ties Hypothesis,” American Journal of Sociology 78(6), 1360–

1380.Green, M. C., and Krosnick, J. A. 2001. “Comparing Telephone and Face-to-Face Interviewing in Terms of Data

Quality: The 1982 National Election Studies Method Comparison Project,” in M. L. Cynamon and R. A. Kulka(eds.), Health Survey Research Methods, pp. 115–121. Hyattsville, MD: U.S. Department of Health and HumanServices.

Greenbaum, S., and Greenbaum, P. 1985. “The Ecology of Social Networks in Four Urban Neighborhoods,”Social Networks 7, 47–76.

Holbrook, A. L., Green, M. C., and Krosnick, J. A. 2003. “Telephone Versus Face-to-Face Interviewing of NationalProbability Samples with Long Questionnaires,” Public Opinion Quarterly 67, 79–125.

Holbrook, A. L., Krosnick, J. A., and Pfent, A. n.d. “Response Rates in Surveys by the News Media and GovernmentContract Research Firms,” Unpublished presentation, Methods of Analysis Program in the Social Sciences,Stanford University, Stanford, CA.

288

Page 27: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

INTERRACIAL FRIENDSHIPS, CLASS, AND SEGREGATION IN AMERICA

Horton, J. 1995. The Politics of Diversity: Immigration, Resistance, and Change in Monterey Park, California. Philadelphia:Temple.

Jackson, R. M. 1977. “Social Structure and Process in Friendship Choice,” in C. Fisher (ed.), Networks and Places:Social Relations in the Urban Setting , pp. 59–78. New York: Free Press.

Johnson, J. H., Jr., Bienenstock, E. J., Farrell, W. C., Jr., and Glanville, J. L. 2000. “Bridging Social Networks andFemale Labor Force Participation in a Multi-ethnic Metropolis,” in L. D. Bobo, M. L. Oliver, J. H. Johnson,Jr., and A. Valenzuela (eds.), Prismatic Metropolis: Analyzing Inequality in Los Angeles, pp. 383–416. New York:Russell Sage Foundation.

Lacy, K. 2004. “Black Spaces, Black Places: Strategic Assimilation and Identity Construction in Middle ClassSuburbia,” Ethnic and Racial Studies 27(6), 908–930.

Laumann, E. O. 1973. Bonds of Pluralism: The Form and Substance of Urban Social Networks. New York: Wiley.Lazarsfeld, P. F., and Merton, R. K. 1954. “Friendship as a Social Process: A Substantive and Method-

ological Analysis,” in M. Berger (ed.), Freedom and Control in Modern Society, pp. 18–66. New York: VanNostrand.

Lee, B. A., and Campbell, K. E. 1999. “Neighbor Networks of Black and White Americans,” in B. Wellman (ed.),Networks in the Global Village, pp. 119–146. Boulder, CO: Westview Press.

Lee, J., and Bean, F. D. 2004. “America’s Changing Color Lines: Immigration, Race/ethnicity, and MultiracialIdentification,” Annual Review of Sociology 30, 221–242.

Lewis Mumford Center for Comparative Urban and Regional Research. 2001. “Ethnic Diversity Grows, Neigh-borhood Integration Is at a Standstill,” University at Albany, April.

Lin, N. 1999. “Social Networks and Status Attainment,” Annual Review of Sociology 25, 467–487.Lipset, S. M. 1994. “The Social Requisites of Democracy Revisited,” American Sociological Review 59, 1–22.Logan, J. 2003. “Ethnic Diversity Grows, Neighborhood Integration Lags,” in B. Katz and R. E. Lang (eds.), Re-

defining Urban and Suburban America, Evidence from Census 2000, vol. I, pp. 235–256. Washington, DC: BrookingsInstitution Press.

Marsden, P. V. 1987. “Core Discussion Networks of Americans,” American Sociological Review 52, 122–313.Marsden, P. V. 1988. “Homogeneity in Confiding Relations,” Social Networks 10, 57–76.Marsden, P. V. 1990. “Network Diversity, Substructures and Opportunities for Contact,” in C. Calhoun, M. Meyer,

and R. S. Scott (eds.), Structures of Power and Constraint: Papers in Honor of Peter Blau, pp. 397–410. New York:Cambridge University Press.

Massey, D. S., and Denton, N. A. 1993. American Apartheid: Segregation and the Making of the Underclass. Cambridge,MA: Harvard University Press.

Massey, D. S., White, M. J., and Phua, V. C. 1996. “The Dimensions of Segregation Revisited,” Sociological Methods& Research 25(2), 172–206.

McPherson, M., Smith-Lovin, L., and Brashears, M. E. 2006. “Social Isolation in America: Changes in CoreDiscussion Networks Over Two Decades,” American Sociological Review 71, 353–375.

McPherson, M., Smith-Lovin, L., and Cook, J. M. 2001. “Birds of a Feather: Homophily in Social Networks,”Annual Review of Sociology 27, 415–444.

Menjıvar, C. 2000. Fragmented Ties: Salvadoran Immigrant Networks in America. Berkeley, CA: University of CaliforniaPress.

Moody, J. 2001. “Race, School Integration, and Friendship Segregation in America,” American Journal of Sociology107(3), 679–716.

Muthen, L. K., and Muthen, B. O. 2004. Mplus User’s Guide: Statistical Analysis with Latent Variables (3rd ed.). LosAngeles: Muthen and Muthen.

Oliver M. L. 1988. “The Urban Black Community as Network,” Sociological Quarterly 29(4), 623–645.Paxton, P. 1999. “Is Social Capital Declining in the United States?: A Multiple Indicator Assessment,” American

Journal of Sociology 105(1), 88–127.Pettigrew, T. F. 1998. “Intergroup Contact Theory,” Annual Review of Psychology 49, 65–85.Portes, A., and Rumbaut, R. C. 2001. Legacies: The Story of the Immigrant Second Generation. Berkeley, CA: University

of California Press.Portes, A., and Zhou, M. 1993. “The New Second Generation: Segmented Assimilation and Its Variants among

Post-1965 Immigrant Youth.” Annals of the American Academy of Political and Social Science 530, 74–96.Putnam, R. D. 2000. Bowling Alone: The Collapse and Revival of American Community. New York: Simon & Schuster.Putnam, R. D. 2003. “Social Capital in a Diverse Society: Who Bridges? Who Bonds?” Unpublished tables and

graphs presented at the Conference, “The Opportunity and Challenge of Diversity: A Role for Social Capital?”Organization for Economic Cooperation and Development, Montreal, Canada, November 25.

289

Page 28: “Some of My Best Friends Are ”: Interracial Friendships ... · “Some of My Best Friends Are...”: Interracial Friendships, Class, and Segregation in America† Xavier de Souza

CITY & COMMUNITY

Putnam, R. D. 2007. “E Pluribus Unum: Diversity and Community in the Twenty-First Century,” ScandinavianPolitical Studies 30(2), 137–174.

Putnam, R. D., and Goss, K. A. 2002. “Introduction,” in R. Putnam (ed.), Democracies in Flux: The Evolution ofSocial Capital in Contemporary Society, pp. 1–19. New York: Oxford University.

Quillian, L., and Campbell, M. E. 2003. “Beyond Black and White: The Present and Future of MultiracialFriendship Segregation,” American Sociological Review 68(4), 540–566.

Roper Center for Public Opinion Research. 2001. Social Capital Community Benchmark Survey: Methodology andDocumentation. Storrs: University of Connecticut.

Ross, C. E., Mirowsky, J., and Pribesh, S. 2002. “Disadvantage, Disorder, and Urban Mistrust,” City & Community1(1), 59–82.

Saguaro Seminar on Civic Engagement in America. 2001. Social Capital Community Benchmark Survey: Ex-ecutive Summary. John F. Kennedy School of Government, Harvard University, Cambridge, MA, March.http://www.cfsu.org/communitysurvey/results.htm (accessed September 19, 2007).

Sampson, R. J. 1988. “Local Friendship Ties and Community Attachment in Mass Society: A Multi-Level SystemicModel,” American Sociological Review 53, 766–779.

Sampson, R. J., Raudenbush, S., and Earls, F. 1997. “Neighborhoods and Violent Crime: A Multi-level Study ofCollective Efficacy,” Science 277(5328), 918–924.

Schlesinger, A. 1944. “Biography of a Nation of Joiners,” American Historical Review 50(1), 1–25.Simmel, G. 1955 [1923]. Conflict and the Web of Group Affiliations, K. H. Worlff and R. Bendix (trans.). Glencoe:

Free Press.Smith, S. S. 2005. “‘Don’t Put My Name on It’: Social Capital Activation and Job-finding Assistance among the

Black Urban Poor,” American Journal of Sociology 111(1), 1–57.Smith, T. W. 1999. “Measuring Interracial Friendships: Experimental Comparisons,” General Social Survey

Methodological Report #91, National Opinion Research Center, University of Chicago.South, S. J., and Messner, S. F. 1986. “Structural Determinants of Intergroup Association: Interracial Marriage

and Crime,” American Journal of Sociology 91, 1409–1430.Tatum, B. D. 1987. Assimilation Blues: Black Families in a White Community. New York: Basic Books.Tatum, B. D. 1999. Why Are All the Black Kids Sitting Together in the Cafeteria? New York: Basic Books.Thernstrom, S., and Thernstrom, A. 1997. America in Black and White: One Nation Indivisible. New York: Simon &

Schuster.Varshney, A. 2002. Ethnic Conflict and Civic Life: Hindus and Muslims in India. New Haven: Yale University Press.Warren, M. R., Thompson, J. P., and Saegert, S. 2001. “The Role of Social Capital in Combating Poverty,” in S.

Saegert, J. P. Thompson, and M. Warren (eds.), Social Capital and Poor Communities, pp. 1–28. New York: RussellSage Foundation.

Weisberg, H. F., Krosnick, J. A., and Bowen, B. D. 1996. An Introduction to Survey Research, Polling, and Data Analysis(3rd ed.). Thousand Oaks, CA: Sage.

Welch, S., Sigelman, L., Beldsoe, T., and Combs, M. 2001. Race and Place. New York: Cambridge University Press.Wellman, B. 1971. “Crossing Social Boundaries: Cosmopolitanism among Black and White Adolescents,” Social

Science Quarterly 52(3) (December), 602–624.Wellman, B. 1996. “Are Personal Communities Local? A Dumptarian Reconsideration,” Social Networks 18, 347–

354.Wellman, B. 2001. “Physical Place and Cyberplace: The Rise of Personalized Networking,” International Journal

of Urban and Regional Research 25(2), 227–252.Wellman, B., and Leighton, B. 1979. “Networks, Neighborhoods and Communities,” Urban Affairs Quarterly 14,

363–390.Wilson, W. J. 1987. The Truly Disadvantaged. Chicago: University of Chicago Press.Winship, C., and Mare, R. D. 1983. “Structural Equations and Path Analysis for Discrete Data,” American Journal

of Sociology 89(1), 55–110.Woolcock, M. 1998. “Social Capital and Economic Development,” Theory and Society 27(2), 151–208.Yancey, W., Ericksen, E. P., and Leon, G. H. 1985. “The Structure of Pluralism: ‘We’re All Italian Around Here,

Aren’t We Mrs. O’Brien?’” Ethnic and Racial Studies 8, 94–116.

290