Neighborhood distress and school dropout: the variable significance of community context

Social

Social Science Research 32 (2003) 659–698

Science

RESEARCH

www.elsevier.com/locate/ssresearch

Neighborhood distress and school dropout: the variable significance of community context Kyle Crowdera,* and Scott J. Southb,1 a

Department of Sociology, Western Washington University, Bellingham, WA 98225, USA b Department of Sociology and Center for Social and Demographic Analysis, State University of New York at Albany, Albany, NY 12222, USA

Abstract Although a substantial body of recent research has examined the impact of neighborhood socioeconomic distress on youth socioeconomic attainment and urban social dislocations, few studies have determined under what conditions, and for what types of adolescents, neighborhood characteristics matter most. Drawing on theories of collective socialization, social capital, and social control, we develop hypotheses regarding the conditional nature of neighborhood effects on the risk of dropping out of high school, and we then test these hypotheses by estimating event history models based on data from the 1968–1993 waves of the Panel Study of Income Dynamics. We find that, among African Americans, the detrimental impact of neighborhood socioeconomic distress on school dropout has increased significantly over the past quarter-century, a probable repercussion of the increasing geographic concentration of urban poverty. The negative effect of neighborhood distress on high school completion is particularly pronounced among black adolescents from single-parent households and among white adolescents from low-income families, results broadly consistent with WilsonÕs claim that exposure to neighborhood poverty reinforces the damaging consequences of individual disadvantage. Supporting the social capital perspective, among both black and white adolescents the deleterious impact of neighborhood distress on school dropout is stronger for recent in-movers than for long-term residents. The impact of neighborhood disadvantage also varies significantly by gender for both racial groups and, among whites, is stronger for younger than older adolescents. We conclude with a discussion of the implications of these findings for theories of neighborhood effects. Ó 2003 Elsevier Science (USA). All rights reserved.

*

Corresponding author. Fax: 1-360-650-7295. E-mail addresses: [email protected] (K. Crowder), [email protected] (S.J. South). 1 Fax: 1-518-442-4936. 0049-089X/$ - see front matter Ó 2003 Elsevier Science (USA). All rights reserved. doi:10.1016/S0049-089X(03)00035-8

660

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

Keywords: Social capital; Social control; Collective socialization; Dropout; Neighborhood distress

1. Introduction Few events in the adolescent life course determine subsequent social and economic opportunities more than dropping out of school. Adolescents who leave school before graduating enjoy fewer prospects for employment and earnings later in life (Murnane et al., 1995; Pallas, 1987; Rumberger, 1987), are more likely to become involved in criminal activity (Griffith et al., 1989; Kasen et al., 1998) and drug use (Swaim et al., 1997), and face a variety of other potentially deleterious outcomes, including reduced productivity and increased psychological stress (Wehlage and Rutter, 1985). Given these myriad repercussions of prematurely leaving the educational system, it is not surprising that the determinants of school performance and the risk of school dropout have been the subject of intense investigation. While most research has focused on the role of individual, family, and school characteristics, a recent wave of studies has attempted to identify the impact of the broader social context on adolescent educational outcomes. Based on observed spatial variations in school performance and dropout rates, and drawing heavily on WilsonÕs (1987, 1996) seminal thesis, this research has sought to identify the extent to which neighborhood socioeconomic characteristics affect various aspects of adolescentsÕ academic performance. While the results of these investigations are far from uniform, most have revealed that neighborhood context, in many ways, complements the impact of individual- and family-level attributes. In comparison to young people from wealthier neighborhoods, those from areas with high levels of poverty and distress tend to have lower test scores and grades (Dornbusch et al., 1991; Gonzales et al., 1996; Turley, 2003), reduced cognitive abilities and higher retention rates (Entwisle et al., 1994; Halpern-Felsher et al., 1997), a higher risk of dropping out of school (Aaronson, 1997; Brooks-Gunn et al., 1993; Connell and Halpern-Felsher, 1997; Connell et al., 1995; Crane, 1991; Ensminger et al., 1996), a lower likelihood of post-secondary education (Duncan, 1994), and ultimately complete fewer years of schooling (Corcoran et al., 1992). Thus, although dissenting evidence can be found (Evans et al., 1992; Plotnick and Hoffman, 1999), in general this body of research suggests that neighborhoods play a meaningful role in determining academic outcomes. But despite this general conclusion, we currently have relatively little knowledge about the degree to which these neighborhood effects vary across time or by personal and family attributes. For example, despite strong but sometimes opposing theoretical suggestions, only a few studies have explored whether the impact of neighborhood characteristics on the risk of dropping out is conditioned by family- and individual-level factors, and none of these studies utilizes data that allow for the examination of how these effects may have changed in recent decades or vary across the

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

661

adolescent life course. In this paper we provide a more nuanced perspective on this issue by exploring the conditions under which neighborhood socioeconomic status most strongly influences adolescentsÕ risk of dropping out of school. We draw on a vast theoretical literature on the variable nature of community influences to develop hypotheses regarding temporal changes in these effects and about the extent to which these influences vary by race, gender, social class, and other micro-level characteristics. We then test these hypotheses using longitudinal data drawn from the Panel Study of Income Dynamics for the period between 1968 and 1993. Our results provide important clues about the mechanisms through which neighborhood contexts influence the risk of dropping out of school.

2. Background and theory Neighborhood characteristics have been linked to individual educational outcomes through a variety of interrelated theoretical mechanisms [see reviews by Jencks and Mayer (1990) and Gephart (1997)]. While some authors have pointed to the important role of local opportunity structures (e.g., Duncan and Hoffman, 1991; Haveman and Wolfe, 1995) or the strength of neighborhood institutions (e.g., Wilson, 1987, 1996), variants of three theoretical perspectives have received most scholarly attention. Perhaps the most popular of these theoretical models, the collective socialization perspective, assumes that the norms, values, aspirations, and ultimately the behaviors of adolescents are shaped by their interaction with non-parental adults in their neighborhood (Gephart, 1997; Jencks and Mayer, 1990). According to the socialization perspective, adult neighbors provide important models of behavior for local adolescents so that youth living in neighborhoods in which many residents experience school failure, joblessness, poverty, and family instability, will themselves be less likely to complete school, gain employment, or forego early, non-marital childbearing. According to contagion models (Crane, 1991), which represent a key variant of the collective socialization perspective, adolescents are also affected by the attitudes and behaviors of their peers in the neighborhood so that dropping out of high school and other deleterious outcomes spread in epidemic fashion in poor and otherwise distressed neighborhoods. Wilson (1987, 1996) offers perhaps the most influential statement of the collective socialization perspective, arguing that neighborhoods with high levels of poverty, joblessness, and other forms of disadvantage are less likely than affluent neighborhoods to stimulate in young people the planfulness, efficacy, and organizational skills that facilitate academic achievement and other conventional behaviors. As Wilson (1987: 56) puts it, high-status neighbors provide ‘‘mainstream role models that help keep alive the perception that education is meaningful, that employment is a viable alternative to welfare, and that family stability is the norm, not the exception.’’ Consistent with the basic assumptions of the socialization perspective, Brooks-Gunn et al. (1993; see also Crane, 1991; Halpern-Felsher et al., 1997) find that the presence of high-status neighbors is particularly important for establishing a setting in which success in school is normatively and institutionally supported. In addition, a variety

662

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

of ethnographic works suggest that many poor neighborhoods provide a normative atmosphere that undermines academic success and emphasizes instead alternative markers of adult status, including early parenthood (Anderson, 1990, 1999; Burton, 1990; MacLeod, 1987). The social capital perspective complements the socialization perspective in many ways, but places greater emphasis on social networks and the flow of information and resources across these networks. Coleman (1990: 300) defines social capital as the set of connections between individuals that are inherent in ‘‘family relations and in community organization and that are useful for cognitive and social development.’’ For adolescents, relationships with non-parental adults and integration in local institutions serve as important sources of social capital that not only allow for the transmission of norms and attitudes, but also provide information about social, educational, and employment opportunities. From the perspective of the social capital approach, academic success is largely a function of the degree to which an adolescent is able to maintain ties to local institutions and is incorporated into a cohesive social network. But this social capital is likely to be especially beneficial to those adolescents integrated into relatively advantaged neighborhoods because these youth are likely to have access to social networks that contain high-status individuals who can provide educationally beneficial resources and information. Thus, consistent with BourdieuÕs (1986) proposition on the convertability of social capital, access to these high-status social networks is often translated into educational capital (Harker et al., 1990). In contrast, the social networks of adolescents living in disadvantaged neighborhoods are likely to be smaller (Wacquant and Wilson, 1989) and local sources of social capital are apt to be weak (Sampson and Groves, 1989; Sampson and Wilson, 1995), containing few individuals with access to educational resources. Thus, as Wilson (1996) argues, adolescents residing in communities fraught with concentrated poverty and related dislocations are not only denied access to mainstream role models, but also face a deficit of interpersonal networks that could enhance their social and economic opportunities. Finally, the social control perspective posits that adolescents living in poor and distressed neighborhoods are at greater risk of school failure and other non-conventional outcomes simply because these neighborhoods provide less oversight of adolescent activities than do more affluent neighborhoods. Jencks and Mayer (1990) argue that adults in more affluent neighborhoods may be more invested in their neighborhoods than are residents of poor areas and are more likely to act as ‘‘potential enforcers’’ who keep children from engaging in risky behavior. This notion is echoed by Brooks-Gunn et al. (1993) and by Hogan and Kitagawa (1985). Similarly, studies of juvenile delinquency suggest that control over the activity of adolescents declines as poverty and social disorganization in a neighborhood increase (Bursik and Grasonick, 1993; Sampson and Groves, 1989; Sampson et al., 1997). Thus, adolescents in distressed neighborhoods may be at greater risk of dropping out not only because they face weak social institutions, ineffective sources of social capital, and a dearth of economically successful role models, but also because local adults may be less likely to intervene during times of educational crisis.

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

663

2.1. The variable significance of neighborhood context Theories of collective socialization, social capital, and social control all imply that, beyond the effects of individual- and family-level characteristics, neighborhood socioeconomic disadvantage increases adolescentsÕ risk of dropping out of school. While this central hypothesis assumes an additive effect of neighborhood characteristics, there are also reasons to expect significant differences in the effect of neighborhood distress across various individual- and family-level characteristics. According to these theoretical mechanisms, neighborhood characteristics exert their influence by shaping the quality of social interactions and the composition of social networks to which individual adolescents have access. Given these mechanisms, the extent to which individualsÕ personal contacts are concentrated within the immediate neighborhood, rather than dispersed across a broader geographic area, represents a key source of potential variability in the effects of neighborhood characteristics. Indeed, an extensive body of theory and empirical research points to important variations in the extent to which neighborhoods circumscribe personal networks and social interactions and, in so doing, suggests a number of micro-level characteristics that might moderate the impact of neighborhood conditions on school dropout. The theoretical perspectives on the effects of neighborhood distress, when viewed in combination with the literature on what Wellman (1979) refers to as ‘‘the community question,’’ point to some complementary—but also some contradictory—hypotheses regarding the conditional and contingent nature of neighborhood effects. But despite the potential implications for both theory and policy, little prior research has explored the conditions under which neighborhood characteristics most strongly influence the risk of dropping out of school. 2.1.1. Variations by race Theoretical statements on the salience of neighborhoods suggest that, largely because of residential segregation, community cohesion may vary substantially by race and ethnicity. A number of authors have suggested that cohesive communities, characterized by dense social networks, are most likely to be found in areas dominated by immigrants and racial and ethnic minority groups (Abrahamson, 1996; Fischer, 1984; Gans, 1962; Greer, 1962; Suttles, 1968; Wellman, 1977). In these neighborhoods community cohesion and interaction between neighbors are likely to be reinforced by the shared ethnic identity of residents. Because of their continuing experience with unprecedented levels of racial isolation (Massey and Denton, 1993), the effects of neighborhood conditions may be especially pronounced among African Americans whose social networks, reference groups, and primary ties are more likely to be constrained to the local neighborhood than are those of other racial and ethnic group members (Lee et al., 1991). Given these dynamics we should expect that neighborhood poverty and distress have a stronger impact on the risk of dropping out among blacks than among nonblacks. Past research provides mixed evidence of racially-differentiated neighborhood effects. Dornbusch et al. (1991) find a significantly greater impact of neighborhood

664

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

conditions on educational outcomes for blacks than for whites, but Brooks-Gunn et al. (1993), Clark (1992), and Halpern-Felsher et al. (1997) report that neighborhood conditions exert a greater influence for whites than for blacks, perhaps reflecting the greater variability of neighborhood conditions experienced by the whites in these studies. The use of different measures of educational success and different analytical strategies makes it difficult to reconcile the conflicting results of these studies. One possibility is that the apparent ambiguity of past findings actually reflects substantial variations in the way neighborhood conditions are measured, with some studies focusing on the presence of affluent neighbors and others concentrating on measures of distress, most often the presence of poor neighbors. These inter-study differences would produce substantially different estimates of racial differences in the effects of neighborhood conditions if black and white adolescents responded differently to different aspects of neighborhood socioeconomic distress. Thus, the mixed picture provided by past research leaves open the question of racial differences in the effects of neighborhood conditions on educational outcomes when multiple aspects of community conditions are considered. In addition, none of these past studies has systematically investigated the possibility that interactions between neighborhood distress and other micro-level characteristics may themselves differ between blacks and whites. Racial differences in levels of residential isolation and neighborhood distress, as well as in a wide variety of individual- and family-level characteristics, certainly raise this possibility. 2.1.2. Variations by gender For several reasons the impact of neighborhood disadvantage on school dropout might be more pronounced for young women than for young men. Distressed neighborhood conditions may be more likely to produce for young women events that disrupt progress towards school completion. Several authors have argued that, for both sexes, neighborhood distress affects educational outcomes indirectly by increasing the occurrence of a number of other behaviors that tend to undermine educational progress (Elliott et al., 1996). For example, neighborhood disadvantage appears to significantly increase the likelihood of early sexual activity and premarital pregnancy (Baumer and South, 2001; Billy and Moore, 1992; Billy et al., 1994; Brooks-Gunn et al., 1993; Crane, 1991; South and Crowder, 1999). Because young women assume primary responsibility for raising children, teenage premarital childbearing will be more disruptive to their educational progress than to young menÕs (Klepinger et al., 1995; Ribar, 1992; Winquist Nord et al., 1992). Gender differences in the effects of neighborhood conditions may also develop because, relative to young men, young women tend to have more extensive social networks (Feiring and Lewis, 1991; Fuhrer et al., 1999), higher rates of neighboring (Campbell and Lee, 1990), and greater investment and involvement in local neighborhoods (Susser, 1982). Following the socialization perspective, this greater integration into the social environment of the local area may increase the susceptibility of young women to the inimical influence of local role models and peer group norms that undermine educational success. As a result, young women may be more detri-

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

665

mentally impacted by high levels of neighborhood poverty and distress than are young men. On the other hand, greater integration into the social networks of the neighborhood may imply that local adults are more active in overseeing the educational progress of young women. According to the social control perspective, this supervision might help to attenuate the impact of the neighborhood socioeconomic status on young womenÕs risk of dropping out. Similarly, the more extensive social networks and neighboring of young women may mean that females have greater access to some sources of social capital that enhance educational opportunities than do young men. Although neighborhood-based sources of social capital might be less beneficial in poor neighborhoods than in more affluent areas, they may still partially counteract the generally detrimental impact of neighborhood disadvantage for young women. Thus, while the collective socialization perspective points to a greater impact of neighborhood distress on the risk of dropping out of school for females than for males, both the social control and social capital perspectives imply the opposite interaction. 2.1.3. Variations by socioeconomic status While each of the theoretical arguments outlined above suggests a general effect of exposure to neighborhood poverty and distress, this effect might also be substantially moderated by the socioeconomic conditions of the family. In particular, the influence of neighborhood distress on the risk of dropping out of school is likely to vary across two components of family socioeconomic status—income and parental education. Consistent with the socialization perspective, Wilson (1996: 78) argues explicitly that exposure to neighborhood poverty is detrimental because it exacerbates the impact of family poverty, noting that individualsÕ immediate ‘‘experiences involving unstable work and low income are reinforced or strengthened by the similar feelings and views of others who share the conditions and culture of the neighborhood.’’ According to Wilson, deep feelings of fatalism and hopelessness thrive among poor residents of distressed neighborhoods because their economic struggles are echoed by those of their neighbors. In such settings, family poverty and neighborhood poverty reinforce one another, thereby reducing the willingness of adolescents to invest in their own education and increasing their risk of dropping out. In contrast, contact with more economically successful neighbors instills in adolescents from poor families the idea that conventional economic and educational success is possible, despite their own economic deprivation (Wilson, 1987, 1996). In a similar way, low parental education may bolster norms devaluing educational attainment that are thought to prevail in poor neighborhoods. The greater concentration of social networks and patterns of interaction within the local neighborhood may also render poor adolescents especially susceptible to local normative and structural conditions (Bott, 1957; Coulton et al., 1990). Alternatively, Jencks and Mayer (1990), drawing on relative deprivation theory, suggest that family and neighborhood socioeconomic status may interact in the opposite direction. They argue that deviant or oppositional behaviors are most likely to arise when disadvantaged children are exposed to more advantaged neighbors,

666

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

thereby fostering the invidious comparisons that generate a ‘‘collective reaction to relative failure’’ (1990:117). Low individual or family SES is most likely to give rise to adolescent behavioral problems when poor children are surrounded by comparatively large numbers of well-off peers. Thus, according to the relative deprivation hypothesis, low family socioeconomic status may be less detrimental in areas containing large poor populations while adolescents from higher-status families may garner more benefit from relative neighborhood advantage (Turley, 2002), suggesting a positive interaction between neighborhood disadvantage and favorable family socioeconomic characteristics on the risk of dropping out. 2.1.4. Variations by household structure Both the social control and socialization perspectives imply that the type of household structure to which an adolescent is exposed may also create variation in the impact of neighborhood distress. McLanahan and Sandefur (1994), for example, argue that children living in single-parent families are less likely to receive encouragement to excel academically. Single parents, they argue, may simply hold lower educational expectations for their children than do other parents since, on average, single parents have lower levels of educational attainment. If such differences do exist, it is reasonable to assume that this family-level socialization may complement the collective socialization process resulting from interaction with poor neighbors. Growing up in a single-parent household might also exacerbate the impact of neighborhood distress by further diminishing adolescent supervision. Although data on the existence and magnitude of the differences are currently underdeveloped, it is often assumed that single parents are able to dedicate less time to their childrenÕs school-related activities, such as assisting with homework or attending school functions, and usually lack support for such activities from another parent (Astone and McLanahan, 1991; Sandefur et al., 1992). To the extent that this is true, children from single-parent households would presumably be less likely to receive additional support during times of educational crisis. If in poor neighborhoods other adults fail to provide such oversight as well, as the social control perspective implies, then the risk of school failure and voluntary dropout is likely to be especially pronounced. In contrast, children growing up in two-parent households may, according to this perspective, receive relatively more support at home, making the context of the neighborhood and the supervision of neighbors comparatively less important. 2.1.5. Variations by length of residence According to the social capital, social control, and collective socialization perspectives, the length of residence in a community may significantly alter the impact of neighborhood context on the risk of dropping out. However, these perspectives lead to opposing predictions regarding the direction of this difference. According to the collective socialization perspective, long-term residents of a neighborhood should be especially susceptible to the effects of neighborhood context. Community attachment, neighborhood-based friendship ties, and local social activities all increase with

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

667

the length of time spent in the local area (Kasarda and Janowitz, 1974; Sampson, 1988). Extended residence in the neighborhood is likely to increase exposure to, and presumably the influence of, prevailing normative conditions and opportunity structures. Conversely, the educational expectations of more recent entries into the neighborhood are presumably shaped, at least in part, by the characteristics of their previous neighborhood. Thus, according to this perspective, long-term residence will intensify the impact of neighborhood poverty and disadvantage on school dropout. In contrast, the social capital and social control perspectives raise the possibility that the impact of neighborhood socioeconomic distress on school dropout is stronger among short-term than among long-term residents. At the individual level, residential mobility has been implicated as a key threat to the maintenance of social capital (Coleman, 1988; Hagan et al., 1996) and, in turn, the loss of social capital has been used to explain the generally negative impact of geographic mobility on childrenÕs academic achievement (Ingersoll et al., 1989; Pribesh and Downey, 1999) and overall educational attainment (Astone and McLanahan, 1994; Hagan et al., 1996; Haveman et al., 1991; McLanahan and Sandefur, 1994; Teachman et al., 1996). According to this argument, adolescent residential mobility often entails a disruption of ties to individuals and institutions that might provide support for, and access to, educational and economic opportunities. Although social networks in poor neighborhoods tend to be smaller and contain relatively fewer high-status residents, the maintenance of at least some form of social capital might be especially important in distressed neighborhoods where normative and structural conditions present additional barriers to school completion. Furthermore, the scarcity of social capital in disadvantaged neighborhoods may simply mean that only the most socially integrated adolescents are able to garner any benefit from local social networks and the tenuous links they provide to educational and economic opportunities. In contrast, in more advantaged neighborhoods, sources of beneficial social capital may be plentiful enough to benefit even recent in-movers and other less-integrated residents. In this sense, relatively new arrivals to the neighborhood are likely to be most detrimentally impacted by poor local socioeconomic conditions because they lack access to already scarce sources of social capital and the advantage of protective, neighborhood-based social networks. In a similar way, it is reasonable to assume that the lack of social integration associated with recent in-mobility reduces the chances that neighborhood adults will supervise adolescents or intervene in a time of crisis. A neighborhood is likely to better control the behavior of adolescents with a longer history in the area and who are well known to the adults and institutions in the area. Because extended residence enhances protective social capital and increases levels of social control, long-term residents of a neighborhood may be somewhat insulated against the damaging influence of neighborhood distress. 2.1.6. Variations by age Neighborhood involvement is thought to vary by individualsÕ stage in the life course (Greer, 1962; Wellman, 1977) such that children and younger adolescents, whose social worlds seldom extend beyond the immediate neighborhood, are most likely to be affected by neighborhood conditions (Greer, 1962; Sorin, 1990).

668

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

Progression towards adulthood is typically accompanied by an expansion of social networks and contacts outside of the immediate neighborhood and a general increase in the scale of an individualÕs social world. Through this process, normative and structural forces outside of the local area presumably gain increasing influence over attitudes, opportunities, and behaviors while the conditions of the neighborhood become less influential. Accordingly, the negative socialization associated with exposure to high levels of poverty, school failure, and other dislocations is likely to be most detrimental to the school progress of young adolescents who draw on the neighborhood for more of their social contacts and role models. Similarly, the weak social control and more tenuous sources of social capital that may exist in distressed neighborhoods are likely to increase the risk of dropping out more for younger adolescents than for older adolescents. 2.1.7. Variations by historical period Finally, it is reasonable to hypothesize that the impact of neighborhood conditions has declined over recent decades. A common theme in community sociology is the waning salience of neighborhoods in the daily lives of their residents (Hunter, 1974; Keller, 1968; Stein, 1960; Warren, 1963), a trend that, according to Putnam (1995), reflects a more general erosion of civic engagement in the US. According to some observers, geographic mobility and increasingly sophisticated communication and transportation technologies have allowed for social networks to be more spatially diffuse, effectively reducing individualsÕ reliance on, and integration in, local communities (Campbell, 1990; Wellman, 1977). The collective socialization perspective implies that as social networks increasingly extend beyond the local community, the normative, and structural conditions of the area become less influential in determining adolescent development. The presence or absence of conventional role models in the immediate neighborhood becomes increasingly irrelevant, and the knowledge of, and access to, opportunities outside of the neighborhood take on greater importance. Thus, based on this popular claim, the impact of local socioeconomic conditions on the risk of dropping out of school may have diminished over recent decades. On the other hand, there is also reason to believe that, at least among African Americans, the impact of neighborhood disadvantage may have become more pronounced in recent decades. Since the 1970s, the increasing concentration of poverty has not only meant that poor urban residents are more likely to share their neighborhoods with larger percentages of other poor individuals, but that these areas are increasingly surrounded by other high-poverty neighborhoods (Jargowsky, 1997). Consequently, residents of poor neighborhoods not only have limited access to high-status neighbors in the immediate area, but they are decreasingly likely to come into contact with such individuals even if their social contacts extend into adjacent areas. These ecological trends have resulted in residents of poor neighborhoods becoming increasingly isolated from conventional role models and potential sources of the most convertible social capital. According to Jargowsky (1997), this increasing concentration of poverty, and presumably the resulting social isolation experienced by residents of poor neighborhoods, have been particularly pronounced among

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

669

African Americans, raising the possibility that, at least among this group, the impact of neighborhood distress on school dropout may have strengthened over time.

3. Data and methods To test for general and conditional effects of neighborhood disadvantage on school dropout, we use data from the Panel Study of Income Dynamics (PSID), linked with decennial census data. The PSID is a nationally representative longitudinal survey of US families and their individual members (Hill, 1992) initiated in 1968 with approximately 5000 families. The panel has been interviewed annually since then, and new families have been added to the sample as children and other members of the original families form new households. By 1993, the cumulative total of individuals participating in the PSID had grown to over 50,000, representing some 8700 families. Sample attrition has been fairly modest and, in general, has not compromised the representativeness of the sample (Fitzgerald et al., 1998). A series of retrospective educational variables makes it possible to determine the timing of educational transitions for most panel members across all years of the PSID. In addition, the PSID data contain an extensive battery of individual- and family-level characteristics that may influence the risk of school dropout.2 What makes the PSID especially well suited for examining contextual effects on educational outcomes is the availability of supplementary Geocode Files. These files allow us to attach to the individual PSID records data drawn from the 1970, 1980, and 1990 censuses describing the neighborhood conditions experienced by PSID respondents and their families at each annual interview. We use census tracts to approximate neighborhoods for the vast majority of cases. Census tracts average about 4000 people, and their boundaries are drawn to circumscribe fairly homogenous populations. Although they imperfectly capture many social dimensions of communities (Tienda, 1991), tracts are by far the most commonly used geographic representation of neighborhoods (Jargowsky, 1997; White, 1987). For respondents residing in untracted areas, we use data for the enumeration district (ED) or, in those rare cases in which both the tract and ED code are unavailable, data for the minor civil division (MCD). Enumeration districts are the basic unit for census enumerators; they are geographically larger than census tracts, but because they are most often used in non-metropolitan areas, they typically contain fewer people (average of 570, versus about 4000 for tracts). Minor Civil Divisions are typically townships, and average about 6000 people. The PSID respondentsÕ residential addresses for the period 1968–1975 are linked to 1970 census data on these neighborhood equivalents, 2 A primary drawback of the PSID data is that they do not provide information about the schools to which children are exposed. This raises questions about whether neighborhood effects may actually represent the impact of local school characteristics. While schools are likely important in shaping educational expectations and performance, there is also evidence to indicate that neighborhood characteristics have an effect of educational outcomes independent of the effects of school characteristics (Ainsworth, 2002; Jencks and Mayer, 1990).

670

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

addresses for the 1976–1985 period are linked to 1980 census data, and addresses for the 1986–1993 period are linked to 1990 census data. Our sample includes black and white PSID family members who were between the ages 14 and 19 between 1968, the first year of PSID data collection, and 1993, the latest year for which cleaned data are currently available. Because the PSID panel is based on an original sample drawn in 1968, members of other racial groups are too small in number to support a separate analysis. Our effective sample includes 3067 black and 3689 white individuals for a total of 6762. For these individuals we use retrospective and annual educational information to measure the timing of the final exit from school without completing high school.3 Just over 15% of the adolescents in our sample experienced such a dropout event before reaching age 20.4 Previous studies of neighborhood effects on educational outcomes have used a wide variety of variables to tap the socioeconomic quality of neighborhoods. For example, Crane (1991) differentiated neighborhoods based on the percentage of their residents that are employed in professional or managerial jobs, Brooks-Gunn et al. (1993) focused on the percentage of residents with high and low incomes, and Ensminger et al. (1996) tapped both the poverty level and occupational structure of residents to characterize neighborhoods as ‘‘poor’’ or ‘‘middle-class.’’ While each of these variables has demonstrated some impact on individual behavior, it is unlikely that any of these single measures can, by itself, fully characterize a neighborhoodÕs socioeconomic status. Thus, we measure the socioeconomic quality of neighborhoods using a multi-item Neighborhood Disadvantage Index (South and Crowder, 1999). The index is constructed as an additive scale comprised of standardized values of six common measures of neighborhood SES: the poverty rate, the percentage of families receiving public assistance, the male joblessness rate (the percentage of working-age men who are either unemployed or out of the labor force), the percentage of families without high incomes (less than $15,000 in 1970, less than $30,000 in 1980,

3 We use a combination of three sets of variables to detect these events. First, for those respondents who were heads of PSID households or spouses/partners of household heads sometime between 1985 and 1993 (many of whom were children in PSID households in earlier years) we use variables tapping retrospective information on the timing and type of school completion. Second, we use similar retrospective variables on month and year last in school and completed education for other members of PSID households who were age 16 or older in interview years between 1985 and 1993. Finally, for those not captured with the first two sets of variables we utilize information, collected annually between 1969 and 1984, indicating current student status and/or whether the respondent left school during the previous year and, if so, the number of years of school completed. This last set of variables was collected for all household members from 1969 to 1974, those age 25 and younger and not a household head or spouse/ partner from 1975 to 1978, and those 16 or older and not a household head or spouse/partner from 1979 to 1984. These inconsistencies in the schedule of questions on educational transitions make it impossible to measure the timing of dropout events for some PSID respondents. In addition, the reliability of information about individualsÕ educational transitions may vary depending on the knowledge of the person supplying the information. Despite these inconsistencies, the similarity of regression results and trends in dropout rates based on specific subsets of cases utilizing these different sets of variables indicate that differences in question wording and other variations do not substantially affect our results. 4 Our data show that only about 3% of young adults had neither graduated nor dropped out of school by age 20. Less than 1% of all dropout events occurred after age 19.

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

671

and less than $50,000 in 1990), the percentage of residents age 25 and over without a college education, and the percentage of workers who are not in managerial or professional occupations.5 Using this index we examine the impact of the average neighborhood disadvantage experienced by each adolescent in the three years leading up to the annual observation period (times t, t
1, and t
2). One of the many difficulties in estimating the effects of neighborhood conditions on individual outcomes is that individuals are not randomly distributed across neighborhoods. A key implication of this ‘‘endogeneity problem’’ is that the apparent impact of neighborhood context might actually reflect the impact of unmeasured individual- and family-level characteristics (Aaronson, 1997; Duncan et al., 1997; Jencks and Mayer, 1990; Tienda, 1991). Especially important in this regard are family income and other socioeconomic factors that may affect both neighborhood location and adolescentsÕ school performance. Fortunately, the PSID allows us to include in our regression models controls for a wide range of family- and individual-level characteristics, helping us to move towards isolating the net effect of neighborhood disadvantage.6 These data also allow us to test key hypotheses regarding the variation of neighborhood effects across these micro-level conditions. Because both educational attainment (Pollard and OÕHare, 1999) and residential location (Logan et al., 1996; South and Crowder, 1997) vary by race and ethnicity, we include controls for whether the youth is black (1 ¼ black and 0 ¼ white) and for 5

The scale has considerable internal reliability (CronbachÕs a ¼ .89) and a principal components analysis revealed that all six neighborhood characteristics load highly on a single factor. This was the only factor extracted with an eigenvalue greater than 1.00. The average inter-item correlation among the six variables is .599. Regression models in which neighborhood disadvantage is measured using single measures of neighborhood disadvantage, including male joblessness, a variable on which Wilson (1996) has focused attention, or aggregate educational attainment, a variable with a direct conceptual link to the risk of dropping out, produced effects that are similar but generally weaker than those using the full index. 6 Controls for these sociodemographic characteristics, of course, are unlikely to remove the influence of all uncontrolled factors that may influence both the selection of families into particular types of neighborhoods and an adolescentÕs risk of dropping out. To further investigate the potential impact of endogeneity on our results, we used a two-stage instrumental-variable approach similar to that employed by Duncan and his colleagues (1997). In this test we used the characteristics of the neighborhood occupied by the adolescentÕs parent (or household head) in the final observation period after the exit of all children from the parentÕs home to create an instrumental version of the annual NDI. We assume that the residential location of parents after their children are out of the home reflects their latent residential preferences but not parental investment, attitudes, and parenting prerogatives that remain unobserved but that might influence adolescent outcomes. We find that, despite the reliance on a smaller sub-sample for whom our instrument is available and the possible introduction of multicollinearity in our two-stage procedure (Maddala, 1983), the effects of the instrumental version of the average NDI are actually somewhat stronger than the effects of the standard measure of neighborhood disadvantage as reported in our main analysis. Thus, we have no reason to believe that the effects of neighborhood distress reported in our analysis reflect the impact of endogeneity. Alternative strategies for dealing with this endogeneity, or omitted-variables problem, have generated inconclusive results. While some studies that adopt an instrumental-variable approach continue to find evidence of neighborhood effects (Duncan et al., 1997), others do not (Evans et al., 1992) and studies that use fixed-effect sibling models to solve the problem have generated similarly inconsistent findings (Aaronson, 1997; Plotnick and Hoffman, 1999). Quasi-experimental studies such as the Gautreaux and MTO programs are somewhat more consistent in finding evidence of true neighborhood effects, although they too are not without limitations (Duncan and Raudenbush, 2001).

672

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

their Hispanic status (1 ¼ Hispanic and 0 ¼ non-Hispanic).7 A control for whether the youth is female (1 ¼ female and 0 ¼ male) is also included to capture gender differences in the risk of dropping out. Many previous studies have found that adolescentsÕ educational attainment is affected by the composition of their households. Most notably, living in a single-parent household appears to decrease educational attainment among adolescents (Astone and McLanahan, 1991; McLanahan, 1985; McLanahan and Sandefur, 1994). The variable single-parent household is scored 1 if the youth lived, at the beginning of the annual interval, in a household in which the head did not have a partner and 0 for all other household types. We also include a variable for the number of children age 18 or under in the household to control for the fact that, in general, family size is negatively associated with educational attainment (Blake, 1989). We include three measures of the socioeconomic conditions of adolescentsÕ families in order to account for the generally positive association between family resources and adolescentsÕ educational outcomes (Ainsworth-Darnell and Downey, 1998; Duncan, 1994). Family income-to-needs ratio measures the annual income of the family relative to the total annual economic need of the family (Survey Research Center, 1974). This measure is adjusted for inflation and is averaged over the three years leading up to the observation period so that we may more effectively differentiate between the impacts of recent socioeconomic conditions of the family and those of the neighborhood. Parental education reflects the number of years of school completed by the more highly educated of the adolescentÕs parents as of the beginning of the annual interview.8 The final measure of family socioeconomic status, home ownership, intended to tap the familyÕs level of wealth, is coded 1 if the youthÕs family owns the current dwelling and 0 otherwise. As noted earlier, residential mobility is negatively associated with childrenÕs academic performance and overall educational attainment. We measure residential stability with a dichotomous variable scored 1 if, at the time of each annual interview, the youthÕs family had lived in the dwelling for at least three years and 0 if they had lived in the same dwelling for fewer than three years.9 We also account for possible 7 Prior to 1990 the PSID had no mechanism for incorporating immigrants into the survey unless they married into an existing panel family. Thus, our sample is not representative of all Hispanic residents of the United States over the period 1968–1993 (Hill, 1992). Given this, ethnic differences in the probability of dropping out of school should be interpreted with caution. 8 In those few cases in which the educational attainment of both parents was not available, we substituted the mean value for parentsÕ education. We also included in all initial regression models an indicator of this mean substitution and found that it was always statistically non-significant and had no impact on the effects of other variables in the models. We therefore excluded this indicator from the final models. 9 Preferable for testing the theoretical models under examination here would be a measure of length of residence in the current neighborhood. Unfortunately, since geocodes are available only for PSID interview years, it is impossible to examine how long respondents in the early years of the survey had been present in the neighborhood. In contrast, the PSID does contain an indicator of length of residence in the current dwelling in 1968 only, allowing for a consistent measure of recent inter-dwelling mobility across all years. It should be noted that the analysis was replicated using an indicator of 3-year inter-neighborhood mobility, including only those years for which such a measure is possible (1972–1993). The results of this supplemental analysis were substantively identical to those reported here.

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

673

urban/rural differences in the likelihood of dropping out using a variable coded 1 for those adolescents living in a census-defined metropolitan area as of the annual interview and 0 for those outside of a metropolitan area. Age, which captures the duration dependence of the estimated hazard of dropping out, is a continuous variable measured in years.10 Finally, secular declines in the risk of dropping out are captured using a continuous variable for year of observation. The operationalization of all of our variables is summarized in Table 1. To estimate the additive and interactive effects of these micro-level characteristics and neighborhood conditions on the risk of dropping out, we use discrete-time event history-models (Allison, 1984). While they generally yield results similar to continuous-time event-history models (Allison, 1984), we utilize discrete-time models largely for the ease with which time-varying explanatory variables can be incorporated. For these models each adolescentsÕ experiences are segmented into a series of person–year observations defined by the period between each successive annual interview. The dependent variable is a binary variable indicating whether the youth made their final exit from school prior to graduation during the annual interval. With the exceptions of race, ethnicity, and gender, the independent variables are treated as time-varying covariates, measured at the beginning of each annual interval. Once having left school or graduated from school, the adolescent is no longer at risk of dropping out, and is no longer observed. In addition, adolescents who leave the panel prior to leaving school are censored at the time of attrition. Discrete-time hazards models of this type can be estimated using logistic regression techniques (Allison, 1984).11 To determine whether neighborhood socioeconomic disadvantage increases the risk of dropping out, we examine the effect of the Neighborhood Disadvantage Index while controlling for the other explanatory variables. To explore whether the effect of neighborhood disadvantage has declined over time or varies by race, gender, socioeconomic status, or other micro-level conditions, we include the appropriate product terms that test for these hypothesized interactions. Our analytical approach differs from those of previous studies in potentially important ways. First, our event-history models are sensitive to the timing of dropout events 10 More complex parameterizations of duration dependence, including sets of dummy variables for age, generated similar findings. The main advantage of capturing age effects with a continuous variable is that the hypothesized interaction between age and neighborhood distress can be tested with a single product term. 11 The possibility exists that the estimated standard errors of these models will be affected by the clustering of sample members within families and neighborhoods, a data structure that often necessitates the estimation of multilevel, or hierarchical, models (e.g., DiPrete and Forristal, 1994). However, these techniques would be of little value in the current situation because the level of clustering of sample members within families and neighborhoods is extremely low (see Duncan et al., 1997). In our data, the average number of observations per family per year is only 1.40 and over 68% of the person–year observations represent individuals in families with no other sample member present. Similarly, for most of our neighborhood units there is only one respondent and the average is only 1.71 per neighborhood. This low level of clustering makes it difficult to obtain stable estimates of family- or neighborhood-level variation in the effects of variables at the individual level, or to partial out that part of the error structure due to similarities between individuals sharing the same household or geographic unit (Teachman and Crowder, 2002).

Whether R left school without graduation between time t and t þ 1 (1 ¼ yes) Whether R is Hispanic (1 ¼ yes) Whether R is female (1 ¼ yes) Whether head of RÕs household is single (1 ¼ yes) Number of people age 18 or under in family at time t Years of schooling completed by parent with highest educational attainment Average ratio of family income to total needs (adjusted for inflation) for RÕs family, time t
2 to t Whether RÕs family owns their home (1 ¼ yes) Whether R has lived in the current dwelling for at least 3 years at time t (1 ¼ yes) Whether R lives in metropolitan area at time t (1 ¼ yes) RÕs age in years at time t Observation year (t) Average additive scale of neighborhood poverty rate, percent of families receiving public assistance, male joblessness, percent of families without high incomes, percent of adults with less than a college education, and the percent of adults not employed in professional or managerial occupations, time t
2 to t

Dropout

Number of person–year observations Number of persons

Metropolitan resident Age Year Neighborhood disadvantage index

Homeownership Long-term resident

Family income/need

Hispanic Female Lives with single HH head Number of children ParentÕs education

Description

Variable

3689

3067

.82 16.05 1979.18 )2.72

.82 .75

2.85

.05 .49 .15 2.36 12.93

16,329

.36 1.53 6.41 4.05

.50 .47

1.00

.11 .50 .50 1.98 2.93

13,678

.85 16.09 1979.65 2.42

.53 .68

1.35

.01 .51 .47 3.12 10.71

.03

Mean

.20

Mean .04

Whites

Blacks SD

Table 1 Weighted descriptive statistics for variables in the event history models of dropping out of school: adolescents age 14–19, 1968–1993

.38 1.48 6.48 4.18

.39 .43

2.07

.23 .50 .35 1.48 2.63

.16

SD

674 K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

675

and move us beyond efforts to model completed educational attainment or whether an individual has completed high school by a given age (e.g., Brooks-Gunn et al., 1993). Second, and perhaps more importantly, by using a time-varying measure of neighborhood disadvantage, we are able to take into consideration the characteristics of the neighborhood to which a youth was exposed immediately preceding a dropout event. In contrast, most past studies have examined the impact of neighborhood conditions measured at some arbitrary single age (e.g., age 14) on educational outcomes in all subsequent years (e.g., Duncan, 1994). Given substantial movement into and out of poor and distressed neighborhoods (Quillian, 1999; South and Crowder, 1997), this measurement of contextual conditions at a single, arbitrary point may potentially obscure the full effects of neighborhood disadvantage on adolescent behavior.

4. Results Table 1 presents descriptive statistics, by race, for all variables included in the analysis. These statistics are based on person–year observations, and thus represent the values of these variables averaged across person–years of exposure to the risk of dropping out.12 Sample weights are applied to these statistics and all regression models to account for the PSIDÕs oversampling of poor, urban families and to enhance the generalizability of our results.13 These statistics point to important similarities between the two racial groups included in the analysis. On average, about 4% of black adolescents and 3% of white adolescents who began the annual interval in school dropped out before the end of the interval. Only a small percentage of the individual adolescents in the sample are from Hispanic families—about 1% of blacks (contributing 1% of the person–year observations in the black sub-sample) and 7% of whites (contributing just over 5% of the person–year observations in the white sub-sample). Adolescents from each racial group average about 16 years of age across the person–year observations, and there is a fairly even split of young men and young women in both racial groups; 52% of the individuals in the black sample are female (contributing 51% of the black person–year observations) and 49.5% of the individuals in the white sample are female (contributing 49% of the white person–year observations). However, there are sharp racial differences in the family context experienced by these adolescents. In a typical observation year, blacks were three times more likely to live in a single-parent household than were white adolescents (47% versus 15%)

12 Although respondents experiencing a dropout event contribute fewer person–year observations, these statistics are similar to those based on person-level observations. 13 One drawback of weighting the analyses is that no sample weights are assigned to individuals who are not immediate members of original PSID panel families. Thus, for example, those marrying into a panel family are excluded from our weighted analyses because they are assigned sample weights of zero. In reality, this selection by itself affects a small number of cases since we also lack information on characteristics of most of these individuals during adolescence and the context of the family or neighborhood in which they grew up.

676

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

and were members of families with a larger total number of children (3.12 versus 2.36 for whites). Even more pronounced are racial differences in socioeconomic conditions; compared to whites, black adolescents spent their at-risk years in families with substantially lower average levels of income and parental education. Reflecting racial differences in wealth, 82% of white adolescentsÕ at-risk years, but only 53% of black adolescentsÕ at-risk years, were spent in owner-occupied homes. Also apparent are sharp racial differences in the average neighborhood context experienced by black and white adolescents. Because the Neighborhood Disadvantage Index is a standardized scale, a score of zero indicates an average level of overall neighborhood disadvantage as measured by the combination of our six neighborhood characteristics. The descriptive statistics, then, indicate that black adolescents spent their at-risk years in neighborhoods that had above-average levels of distress while, on average, white adolescents spent their at-risk years in areas with disadvantage scores that were below average by about the same magnitude. A consideration of some of the components of this index makes this racial difference clearer. For example, in the average observation year, black adolescents in the sample lived in neighborhoods in which the poverty rate was 26%, 16% of families received public assistance income, and 91% of adults had less than a college education. In contrast, white adolescents in the average observation year lived in a neighborhood in which the poverty rate was 11%, fewer than 6% of families received public assistance, and 84% of adults had less than a college education. Table 2 presents the results of logisitic regression models relating these variables to the log of the estimated hazard of dropping out of school. Records for black and white youths are pooled in these models in order to distinguish the overall effects of these variables and to test for racial differences in the impact of neighborhood distress. The first model shows that once other background characteristics are controlled, the risk of dropping out is actually significantly lower for black adolescents than for white adolescents (see also Teachman et al., 1996) and there is no significant difference by hispanicity. However, it is important to note here that immigrants were not actively incorporated into the PSID panel until 1990 so that the Hispanic members of our sample are unlikely to be representative of the US Hispanic population, a fact that warrants caution in interpreting the effects of ethnicity in Table 2. The risk of dropping out is generally lower for females than for males and is higher for adolescents from larger families. However, once family socioeconomic conditions are controlled, living in a single-parent household has no apparent effect on the probability of dropping out.14 The likelihood of dropping out is lower for adolescents from higher-income families and for those whose parents have higher levels of education. Consistent with the social capital perspective, residential stability sig-

14

Living in a single-parent household has a significant positive impact in a bivariate analysis (b ¼ :691; SE ¼ .077; p ¼ :000), but the effect is attenuated by controls for differences in income, parental education, and neighborhood context across household types.

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

677

nificantly reduces the risk of dropping out of school. Finally, the risk of school dropout increases significantly as adolescents age. The coefficient for year of observation in Table 2 indicates that, when all other variables are controlled, the rate of school dropout has increased over time. This ostensible upward trend contradicts the general temporal decline in the rate of dropout Table 2 Logistic regression analysis of the annual log-odds of dropping out of school: adolescents age 14–19, 1968–1993 1 Independent variables Black Hispanic Female Lives with single HH head Number of children ParentÕs education Family income/need Home ownership Long-term resident Metropolitan resident Age Year Neighborhood disadvantage index (NDI) Interactions NDI X black Constant Model v2 df

).378 (.103) .235 (.140) ).248 (.073) .035 (.093) .081 (.024) ).192 (.013) ).332 (.057) ).233 (.087) ).472 (.079) .104 (.097) .348 (.025) .049 (.006) —

2

3

).532 (.107) .178 (.140) ).246 (.073) .051 (.093) .087 (.024) ).183 (.013) ).251 (.058) ).175 (.088) ).497 (.079) .179 (.097) .351 (.025) .046 (.006) .054 (.011)

).652 (.129) .185 (.140) ).246 (.073) .046 (.093) .086 (.024) ).187 (.013) ).264 (.059) ).164 (.088) ).505 (.080) .158 (.099) .351 (.025) .047 (.006) .042 (.013)

—

—

)6.823 (.474) 806.822 12

)7.066 (.476) 831.688 13

.046 (.025) )7.012 (.477) 835.215 14

Notes. Numbers in parentheses are standard errors. Year has been re-scaled as a counter variable with 1968 ¼ 0. N ¼ 30; 007. * p < :05. ** p < :01. *** p < :001 (two-tailed tests).

678

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

(US Department of Education, 2000). It is important to note that the secular decline in the percentage of individuals dropping out by a given age is observed with the raw, non-person–year PSID data. In addition, the effect of observation year is statistically non-significant in a bivariate model. The second model of Table 2 adds to the equation the measure of neighborhood disadvantage. The coefficient for the Neighborhood Disadvantage Index indicates that, as the level of neighborhood distress increases, the risk of dropping out increases significantly. For example, the estimated odds of dropping out for an adolescent living in a neighborhood at the 10th percentile of neighborhood disadvantage [NDI ¼ )8.185; odds ¼ eð
8:185Þð:054Þ ¼ .642] are 36% lower than the odds for an adolescent in a neighborhood with an average level of disadvantage (NDI ¼ 0). For an adolescent at the 90th percentile of neighborhood disadvantage, the odds of dropping out are about 20% higher [NDI ¼ 3.319; odds ¼ eð3:319Þð:054Þ ¼ 1.196] than at the average. And, importantly, this effect of neighborhood socioeconomic status appears to operate independently of the effects of family resources and other micro-level characteristics.15 While the second model indicates a general impact of neighborhood disadvantage on the individual risk of dropping out, our central concern is with variations in this effect across micro-level attributes. Model 3 of Table 2 begins our exploration of these interactive effects, displaying the results of a regression model that includes a product term representing the interaction between the Neighborhood Disadvantage Index and the race of the adolescent. While it barely fails to achieve statistical significance with a two-tailed test (p ¼ :06), the positive interaction in Model 3 points to a moderate racial difference in the impact of neighborhood conditions. Specifically, this coefficient implies that the impact of neighborhood disadvantage on the logodds of dropping out is over twice as large for black adolescents (b ¼ :042þ :046 ¼ :088) than for white adolescents (b ¼ :042). The magnitude of the racial difference is further illustrated in Fig. 1, which compares predicted probabilities of dropping out of school for black and white adolescents across the range of neighborhood conditions actually experienced by the members of each group in our sample. We use the coefficients in Model 3 of Table 2 and the mean values of all other explanatory variables in the model to generate these predicted probabilities. The figure shows that, once racial differences in family sociodemographic characteristics are controlled, black adolescents are less likely than are whites to drop out of school and this differential exists across the entire portion of the range of neighborhood disadvantage shared by black and white adolescents ()16 < NDI < 12). However, as the level of neighborhood disadvantage increases, the risk of dropping out increases somewhat more strongly among black adolescents, converging toward the net risk of dropping out among white adolescents in the most disadvantaged areas to which whites are exposed. Furthermore, 15 As expected, the bivariate impact of neighborhood disadvantage on the risk of dropping out is even stronger (b ¼ :115; SE ¼ .008; p < :001) than in the multivariate model. Thus, controlling for the effects of individual- and family-level attributes diminishes, but does not completely eliminate, the impact of neighborhood disadvantage.

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

679

Fig. 1. Predicted probability of dropping out of school by neighborhood disadvantage and race: adolescents age 14–19, 1968–1993.

as the NDI increases beyond a score of 12, characterizing levels of neighborhood disadvantage experienced by none of the white but about 1 in 100 of the black adolescents in our sample, the likelihood of dropping out continues to rise for black adolescents, reaching a maximum of 7.5% in the most disadvantaged neighborhoods. While this racial difference in the impact of neighborhood disadvantage is substantial in our sample, we remind the reader that it just fails to attain statistical significance at conventional confidence levels.16 It is worth noting that the direction of the racial interaction revealed in Table 2 and Fig. 1 is consistent with theoretical expectations but contrasts with the results of some prior research that found stronger effects of affluent neighbors on the educational progress of white adolescents than for blacks (e.g., Brooks-Gunn et al., 1993; Halpern-Felsher et al., 1997). Our analysis indicates that when a time-varying indicator that encompasses multiple dimensions of neighborhood disadvantage is incorporated into an event-history approach, the effects of neighborhood context as measured here are greater among black adolescents than among white adolescents. The moderately strong racial interaction in Table 2 also raises the possibility that there are significant racial differences in the characteristics that moderate the impact of neighborhood socioeconomic conditions. In fact, given racial differences in family context, socioeconomic resources, and levels of neighborhood disadvantage, racial differences in these cross-level interactions seem quite likely. Thus, following prior explorations of neighborhood effects on educational outcomes (see e.g., Crane, 16

The apparent curvilinearity in this and subsequent figures reflects the transformation of coefficients from log-odds to probabilities rather than non-linear effects of neighborhood disadvantage. Guided by the epidemic/contagion hypothesis proposed by Crane (1991) and others we did investigate the possibility of non-linear effects by estimating models that included a coefficient for neighborhood disadvantage squared. However, this polynomial was statistically non-significant in all models, thereby contradicting the idea of non-linear effects. In addition, the inclusion of the polynomial had no impact on the substantive conclusions about any cross-level interactions.

680

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

1991; Halpern-Felsher et al., 1997), subsequent analyses are performed separately for black and white adolescents. Table 3 presents the results of logistic regression models for black adolescents only. Model 1 shows that among blacks the effects of individual- and family-level characteristics on the hazard of dropping out are also generally consistent with past observations. There are, however, some noteworthy differences between the effects of these variables in the model for blacks and the effects in the pooled sample. All else being equal, black female and male adolescents appear to be equally likely to leave school. However, black adolescents living with a single parent are significantly more likely to drop out than are those from other types of families. Consistent with the interactive model (Model 3) of Table 2, the socioeconomic quality of the neighborhood has a relatively large and statistically significantly effect on the risk of dropping out of school for black adolescents, independent of individual- and family-level characteristics. Most important for our purpose are the coefficients for the interaction terms included in the remaining models of Table 3. The significant interaction term in Model 2 indicates that, among black adolescents, the impact of neighborhood distress varies significantly by gender, and this interaction between gender and neighborhood disadvantage remains strong and statistically significant even when all other interactions are included in the model (Model 9).17 The nature of this interaction is depicted in Fig. 2A which compares predicted probabilities of dropping out of school for black male and female adolescents across the full range of neighborhood conditions experienced by blacks in the sample. Once again, we use the coefficients in Model 2 of Table 3 and the race-specific mean values of the other explanatory variables to generate these simulated probabilities. The figure indicates that the risk of dropping out is similar for black female and male adolescents in neighborhoods with low to moderate levels of distress. However, for black males the risk of dropping out increases at a more dramatic pace as the level of neighborhood distress increases and at high levels of neighborhood disadvantage (NDI > 5; above the 76th percentile for blacks) the risk of dropping out for young black women and men diverge markedly. In the most highly disadvantaged neighborhoods, young black men are almost twice as likely to drop out than are young black women. These results are generally inconsistent with the central assumptions of the collective socialization perspective that imply that, because of their relatively limited social integration and weaker contacts with neighbors, young black men should be somewhat less susceptible to the negative social influences and the dearth of positive role models said to prevail in disadvantaged areas. In fact, the gender difference in the impacts of neighborhood disadvantage revealed here might be used to challenge this prevailing assumption about gender differences in levels of social integration, or at least raise questions about the uniformity of these gender differences. Especially in

17

The difference between the model v2 in Model 9 and v2 in Model 1 is statistically significant (445:55
413:67 ¼ 31:88; df ¼ 19
12 ¼ 7; p < :01) indicating that the inclusion of these interactions contributes significantly to the predictive power of the model.

NDI X parentÕs education

Interactions NDI X female

Neighborhood disadvantage index (NDI)

Year

Age

Metropolitan resident

Long-term resident

Home ownership

Family income/need

ParentÕs education

Number of children

Lives with single HH head

Female

Independent variables Hispanic

).515 (.489) .036 (.090) .274 (.108) .071 (.025) ).075 (.016) ).165 (.074) ).191 (.105) ).438 (.096) ).041 (.132) .424 (.031) .040 (.008) .082 (.012)

1

).057 (.023)

).484 (.489) .245 (.124) .280 (.108) .072 (.025) ).075 (.016) ).159 (.074) ).196 (.105) ).425 (.096) ).035 (.132) .425 (.031) .041 (.008) .113 (.018)

2

).003 (.004)

).518 (.489) .036 (.090) .276 (.108) .071 (.025) ).066 (.019) ).169 (.074) ).187 (.105) ).433 (.096) ).041 (.133) .424 (.031) .040 (.008) .113 (.042)

3 ).514 (.489) .036 (.090) .274 (.108) .071 (.025) ).075 (.016) ).162 (.081) ).190 (.105) ).438 (.096) ).041 (.132) .424 (.031) .040 (.008) .083 (.019)

4 ).399 (.490) .023 (.090) ).045 (.138) .071 (.025) ).076 (.016) ).199 (.075) ).202 (.106) ).439 (.096) ).036 (.132) .427 (.031) .039 (.008) .019 (.021)

5 ).503 (.489) .048 (.090) .266 (.108) .073 (.025) ).073 (.016) ).163 (.074) ).223 (.106) ).242 (.131) ).032 (.132) .426 (.031) .041 (.008) .114 (.019)

6

Table 3 Logistic regression analysis of the annual log-odds of dropping out of school: black adolescents age 14–19, 1968–1993

).521 (.489) .034 (.090) .275 (.108) .070 (.025) ).075 (.016) ).166 (.074) ).189 (.105) ).439 (.096) ).042 (.132) .388 (.042) .040 (.008) ).084 (.132)

7 ).525 (.489) .020 (.090) .260 (.108) .069 (.025) ).075 (.016) ).156 (.075) ).214 (.106) ).449 (.096) ).051 (.132) .425 (.031) .026 (.010) .024 (.029)

8

).065 (.024) ).006 (.005)

).387 (.490) .248 (.124) ).076 (.143) .073 (.025) ).058 (.021) ).262 (.090) ).251 (.107) ).291 (.132) ).030 (.132) .389 (.041) .031 (.010) ).140 (.140)

9

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698 681

)9.94 (.596) 413.67 12

1

)10.12 (.602) 419.71 13

2

)10.03 (.607) 414.28 13

3

4

)9.94 (.598) 413.68 13

).001 (.013)

)9.74 (.598) 427.53 13

.093 (.025)

5

)10.12 (.603) 418.65 13

).054 (.024)

6

Notes. Numbers in parentheses are standard errors. Year has been re-scaled as a counter variable with 1968 ¼ 0. N ¼ 13; 678. * p < :05. ** p < :01. *** p < :001 (two-tailed tests).

Model v2 df

Constant

NDI X year

NDI X age

NDI X long-term resident

NDI X single HH head

NDI X family income/need

Table 3 (continued)

)9.33 (.762) 415.26 13

.010 (.008)

7

.004 (.002) )9.73 (.604) 418.55 13

8

9 .029 (.016) .100 (.028) ).037 (.025) .011 (.008) .004 (.002) )9.27 (.772) 445.55 19

682 K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

A

B

C

D

683

Fig. 2. Predicted probability of dropping out of school by neighborhood disadvantage and selected sociodemographic characteristics: black adolescents age 14–19, 1968–1993. (A) Interaction by gender. (B) Interaction by household composition. (C) Interaction by residential mobility. (D) Interaction by year.

the most disadvantaged neighborhoods, where the gender difference in the risk of dropping out is most apparent, the prevalence of gangs and the lack of alternative activities and general social cohesion may actually increase the social integration (in this case, into social networks supporting unconventional behavior) of young black men relative to young black women, thereby increasing their relative risk of dropping out. More directly, based on the assumption that young women tend to be more socially integrated into conventional social networks than are young men, these results may be viewed as supportive of hypotheses derived from the social control and social capital perspectives. Specifically, it appears that the relatively greater social integration and more extensive social networks of young black women may provide them with access to a measure of protective social capital or more intensive supervision needed to shield them from the generally harmful impact of neighborhood disadvantage. The results presented in Models 3 and 4 of Table 3 indicate that, among black adolescents, the generally positive impact of neighborhood disadvantage on school dropout does not vary significantly by parental education or family income. Model 5, however, indicates that the impact of neighborhood socioeconomic conditions does vary significantly by family living arrangements, an interaction that again remains significant in the fully interactive model (Model 9). As predicted by both the collective socialization and social control perspectives, among black adolescents exposure to distressed neighborhood contexts appears to exacerbate the effect of

684

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

living in a household headed by a single parent. This result also provides support for WilsonÕs (1987) claim that that exposure to neighborhood poverty reinforces the damaging consequences of individual disadvantage. Once again, the nature of this interaction is depicted in Fig. 2B indicates that the risk of dropping out is similar for black adolescents from single-parent and two-parent households when the level of neighborhood disadvantage is relatively low. However, once the level of disadvantage exceeds the average, the risk of dropping out diverges sharply for these two groups. At the highest level of neighborhood disadvantage, adolescents from single-parent households are over four times more likely to drop out than are children from two-parent households (predicted probabilities of .199 and .046, respectively). This interaction is consistent with the hypothesis, derived from collective socialization theory, that adolescents from two-parent households are exposed to more educationally supportive home environments, and that these advantages help to counter the otherwise detrimental effects of living in a distressed neighborhood. This finding is also consistent with the hypothesis, drawn from the social control perspective, that adolescents from two-parent households benefit from increased supervision that provides protection against the harmful impact of neighborhood poverty and helps to compensate for the weak oversight offered by the poor community. Model 6 of Table 3 provides a test of the socialization and social capital perspectivesÕ competing hypotheses about potential interactive effects of length of residence and neighborhood distress on school dropout. The negative interaction in this model implies that the impact of residential stability is especially pronounced among those adolescents entering into more disadvantaged neighborhoods. In fact, the predicted probabilities in Fig. 2C indicate that in the most distressed neighborhoods the risk of dropping out is about two times higher for recent in-movers than it is for longer-term residents (predicted probabilities of .180 and .089, respectively). This difference refutes the hypothesis, drawn from the collective socialization perspective, that the impact of neighborhood distress is likely to be strongest among long-term residents because of their sustained exposure to community norms and attitudes that fail to encourage educational attainment. This effect is, however, consistent with the hypothesis derived from the social capital and social control perspectives. Specifically, short-term residents appear to be most vulnerable to the impacts of neighborhood distress possibly because their recent mobility undermines the cohesion of their social networks and their ties to local institutions that might otherwise counterbalance the effects of neighborhood disadvantage and prevent an early exit from school. Stated another way, recent in-movers are least likely to gain access to local sources of social control and social capital, and if these resources are already limited (as they likely are in disadvantaged neighborhoods) then this mobility-related disadvantage is magnified, producing a negative interaction between neighborhood disadvantage and residential stability. While the impact of neighborhood disadvantage on the risk of dropping out does not appear to vary by the age of black adolescents (Model 7), Model 8 of Table 3 indicates that this effect has changed significantly over recent decades. Contrary to the presumed secular decline in the significance of neighborhoods, the interaction between the Neighborhood Disadvantage Index and year of observation (Models

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

685

8 and 9) indicates that, between 1968 and 1993, neighborhood disadvantage became increasingly more important in determining black adolescentsÕ chances of staying in school. These historical changes in the effect of neighborhood distress on school dropout are illustrated in Fig. 2D. As shown here, neighborhood disadvantage had a relatively modest impact on the likelihood of leaving school in 1970. But by 1990, there was an almost 26% point difference in the probability of dropping out between black adolescents from the least distressed neighborhoods and those from the most distressed neighborhoods (probabilities of .006 and .265, respectively). The historical rise in the impact of neighborhood distress on school dropout is consistent with the claim that, among African Americans, the increasing geographic concentration of poverty has exacerbated the impact of living in a distressed neighborhood. As poor, disadvantaged neighborhoods have become increasingly clustered in growing areas of concentrated poverty, residents of these communities become increasingly isolated from sources of social capital and social norms that support educational persistence and attainment. Conceivably, it is this growing social and spatial isolation of poor neighborhoods that accounts for the increasingly negative impact of neighborhood distress on school completion. Table 4 presents the results of a parallel analysis for white adolescents. The first model assumes only additive effects of the explanatory variables on the risk of dropping out of school. The effects of most variables are again consistent with earlier analyses. However, there are a few notable differences between white and black adolescents in the effects of some of the explanatory variables. As in the pooled model, once other variables are controlled, living in a single-parent household does not significantly influence the risk of leaving school early among white adolescents. Contrasting the absence of a significant net gender difference among blacks, white males are significantly more likely than white females to drop out of school. And, consistent with the analyses in Table 2, among whites the effect of the Neighborhood Disadvantage Index on the hazard of dropping out is somewhat weaker than the analogous effect among blacks (Model 1, Table 3), but it is positive and statistically significant. More central to our objectives are the equations in Models 2–9 of Table 4, which add to this baseline model the product terms representing the interactions between the Neighborhood Disadvantage Index and several other of the independent variables. As the coefficients for these interaction terms indicate, the impact of neighborhood disadvantage varies significantly across four micro-level characteristics of white adolescents. First, as in the analysis for black adolescents, the coefficient for the interaction term in Model 2 points to a significant gender difference in the impact of neighborhood conditions on school dropout among white adolescents. It is interesting to note, however, that the direction of this interaction differs by race; among whites, neighborhood disadvantage has a significantly stronger influence on the dropout rates of females than males. Fig. 3A depicts this gender difference. This figure shows that the net positive impact of neighborhood socioeconomic disadvantage on the risk of dropping out is quite modest for white males but more pronounced for white females. Moreover, while at most levels of neighborhood disadvantage white males are slightly more likely than white females to drop out of school, the risk for females actually surpasses the risk for males in the most extremely distressed neighborhoods.

Neighborhood disadvantage index (NDI)

Year

Age

Metropolitan resident

Long-term resident

Home ownership

Family income/need

ParentÕs education

Number of children

Lives with single HH head

Female

Independent variables Hispanic

.099 (.184) ).303 (.104) ).025 (.137) .105 (.036) ).231 (.019) ).113 (.040) ).144 (.127) ).541 (.114) .197 (.137) .337 (.036) .048 (.009) .040 (.016)

1 .081 (.184) ).248 (.107) ).039 (.137) .106 (.036) ).231 (.019) ).111 (.040) ).144 (.126) ).548 (.113) .199 (.137) .338 (.036) .048 (.009) .016 (.019)

2 .098 (.184) ).301 (.104) ).025 (.137) .105 (.036) ).231 (.019) ).115 (.040) ).140 (.127) ).541 (.114) .196 (.137) .338 (.036) .049 (.009) .069 (.050)

3 .086 (.184) ).300 (.104) ).041 (.138) .097 (.036) ).225 (.019) ).172 (.048) ).117 (.127) ).544 (.114) .208 (.138) .339 (.036) .047 (.009) .072 (.018)

4 .099 (.184) ).308 (.104) ).007 (.139) .104 (.036) ).230 (.019) ).116 (.040) ).136 (.127) ).544 (.114) .187 (.138) .339 (.036) .048 (.009) .029 (.018)

5 .061 (.185) ).294 (.104) ).032 (.137) .106 (.036) ).231 (.019) ).116 (.040) ).137 (.128) ).611 (.117) .210 (.138) .340 (.036) .049 (.009) .106 (.027)

6

Table 4 Logistic regression analysis of the annual log-odds of dropping out of school: white adolescents age 14–19, 1968–1993

.091 (.184) ).301 (.104) ).030 (.137) .105 (.036) ).230 (.019) ).114 (.040) ).135 (.127) .540 (.114) .198 ()137) .321 (.037) .049 (.009) .376 (.160)

7

.096 (.184) ).299 (.104) ).023 (.137) .105 (.036) ).229 (.019) ).115 (.040) ).139 (.127) ).538 (.113) .208 (.138) .337 (.036) .046 (.009) .066 (.032)

8

.031 (.185) ).245 (.107) ).049 (.139) .100 (.037) ).224 (.020) ).165 (.049) ).104 (.128) ).604 (.117) .222 (.139) .329 (.037) .046 (.009) .323 (.160)

9

686 K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

)6.44 (.678) 449.34 12 )6.47 (.678) 454.02 13

.063 (.029)

)6.47 (.680) 449.72 13

).003 (.004)

)6.43 (.679) 455.64 13

).014 (.004)

)6.48 (.680) 451.27 13

.050 (.036)

)6.49 (.682) 460.08 13

).099 (.031)

Notes. Numbers in parentheses are standard errors. Year has been re-scaled as a counter variable with 1968 ¼ 0. N ¼ 16; 329. * p < :05. ** p < :01. *** p < :001 (two-tailed tests).

Model v2 df

Constant

NDI X year

NDI X age

NDI X long-term resident

NDI X single HH head

NDI X family income/need

NDI X parentÕs education

Interactions NDI X female

)6.21 (.687) 453.85 13

).020 (.009) ).002 (.002) )6.46 (.678) 450.23 13

.055 (.028) .004 (.005) ).013 (.005) .029 (.036) ).080 (.031) ).015 (.009) ).001 (.002) )6.30 (.693) 472.49 19

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698 687

688

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

A

B

C

D

Fig. 3. Predicted probability of dropping out of school by neighborhood disadvantage and selected sociodemographic characteristics: white adolescents age 14–19, 1968–1993. (A) Interaction by gender. (B) Interaction by family income-to-need ratio. (C) Interaction by residential mobility. (D) Interaction by year.

The direction of this gender interaction disconfirms the hypothesis derived from the social control and social capital perspectives; apparently, neither the greater supervision of femalesÕ activities nor their more extensive local social networks renders young white women less susceptible than young white men to the harmful impact of neighborhood distress. Indeed, young womenÕs social integration and neighborhood involvement appear to exacerbate the harmful impact of the social norms that undermine educational progress in the most disadvantaged neighborhoods, as would be expected according to the hypothesis derived from the collective socialization perspective. Relatedly, this gender difference may also reflect the greater vulnerability of young white women in poor neighborhoods to events, such as premarital childbearing, that disrupt or curtail their educational trajectories. Indeed, past research has demonstrated that the impacts of neighborhood distress on premarital childbearing and early marriage are more pronounced among white women than among black women (cf., South and Crowder, 1999). Thus, the sharp racial difference in the interaction between gender and neighborhood disadvantage points to possible racial differences in the mechanisms through which neighborhood disadvantage influences the risk of dropping out, but is also likely rooted in racial differences in the impacts of neighborhood distress on more proximate determinants of school completion, impacts that are especially prominent for young white women.

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

689

The analyses in Table 4 also demonstrate that the socioeconomic resources of white adolescentsÕ families significantly moderate the impact of neighborhood disadvantage. Specifically, the negative interaction in Model 4 of Table 4 indicates that higher levels of family income provide a measure of protection against the detrimental impact of neighborhood distress on school completion. In fact, as the predicted probabilities in B of Fig. 3 show, neighborhood socioeconomic conditions appear to have little or no effect on white adolescents from the wealthiest families (income-to-needs ratio at the 90th percentile). In contrast, there is a moderate impact of neighborhood disadvantage for those adolescents from families in the middle of the income distribution (50th percentile) and a fairly pronounced effect for those from the poorest families (10th percentile). Contradicting the relative deprivation hypothesis, when neighborhood disadvantage is low, the predicted probability of dropping out for adolescents from relatively poor families is similar to the probability for adolescents from relatively wealthy families, at least when other variables are held constant at their means. But at high levels of neighborhood disadvantage, the differences in dropout rates by family economic status are prominent; poor residents of the most distressed neighborhoods are over four times more likely to drop out in a given year than are better-off residents of similarly distressed neighborhoods (predicted probabilities of .087 and .021, respectively). This finding is also consistent with WilsonÕs (1987, 1996) contention that the impact of family poverty on adolescent socioeconomic attainment is exacerbated by poverty at the neighborhood level. According to the collective socialization perspective, this interaction occurs because the frustration and hopelessness resulting from a familyÕs experiences with poverty are amplified by neighborhood-level normative conditions that minimize the importance of educational success. As with blacks, the results in Model 6 of Table 4 indicate that the impact of neighborhood disadvantage on the risk of dropping out for white adolescents also differs for long-term residents and recent in-movers. Among whites, however, this interaction is especially pronounced. In fact, as depicted in Fig. 3C, neighborhood socioeconomic conditions appear to have almost no net impact on the risk of leaving school early for long-term residents. In contrast, for recent in-movers the risk of dropping out increases sharply with neighborhood distress. In the most distressed neighborhoods to which whites in the sample are exposed, the predicted probability of dropping out is almost six times higher for recent in-movers than for longer-term residents (.163 versus .028, respectively). Once again, this interaction contradicts the hypothesis drawn from the collective socialization perspective but lends substantial support to the social capital and social control perspectives. It appears that residential instability is generally detrimental to the prospects of completing school because it undermines social integration, disrupting ties to local sources of adult oversight and network-based opportunities. However, this disruption appears to be especially detrimental for those entering into distressed neighborhoods, where these resources are likely to be scarce for even the most integrated residents. The coefficient for the interaction term in Model 8 of Table 4 indicates that, in contrast to the case for black adolescents, the impact of neighborhood

690

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

socioeconomic conditions has not changed over the past several decades. The absence of a time trend in the effect of neighborhood disadvantage on school dropout among whites may reflect the fact that poor white neighborhoods have not witnessed the increasing spatial and social isolation that has characterized poor black neighborhoods over the past quarter century. Our results do, however, indicate that the impact of neighborhood conditions does vary by the age of white adolescents. Consistent with the collective socialization perspective, the interaction in Model 7 of Table 4 implies that neighborhood disadvantage has an especially profound impact on school completion for the youngest white adolescents whose spatially limited social networks presumably leave them most vulnerable to local normative conditions. As the predicted probabilities in D of Fig. 3 indicate, the risk of dropping out increases for adolescents of all ages as neighborhood disadvantage increases, and under all neighborhood conditions older adolescents are more likely than their younger counterparts to leave school. However, as neighborhood distress increases into moderate and high levels (those with NDI scores above 0), the risk of dropping out converges for these age groups such that in the most distressed neighborhood environments adolescents of all ages are almost equally likely to leave school.

5. Discussion and conclusions Given the profound importance of school retention for economic success in later life, the determinants of adolescent educational performance and attainment have long been the subject of intense scholarly interest. Inspired primarily by the work of Wilson (1987, 1996), the search for risk factors associated with educational failure have recently extended beyond the characteristics of individuals, families, and schools to focus on the impact of broader contextual conditions. Consistent with the results of other recent studies (e.g., Brooks-Gunn et al., 1993; Halpern-Felsher et al., 1997), our analysis shows that the socioeconomic characteristics of residential neighborhoods have a substantial impact on adolescentsÕ likelihood of dropping out of school, and suggests that these contextual effects operate largely independently of the influence of family background, socioeconomic resources, and other micro-level conditions. But in contrast to most prior studies, our analysis also demonstrates that the magnitude of this impact of neighborhood disadvantage is quite variable, differing substantially by a number of individual- and family-level characteristics. Moreover, we find that for African Americans the impact of neighborhood socioeconomic distress on school dropout has increased markedly over the past quarter century. This variation in the salience of neighborhood socioeconomic disadvantage in determining adolescent educational attainment highlights the efficacy of several theoretical perspectives and, in doing so, provides clues about the mechanisms through which adolescent behavior is influenced by the local socioeconomic context. Our results lend support to some hypotheses based on each of three theoretical perspectives. First, the collective socialization perspective posits that local peers and

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

691

adults serve as role models for adolescents, shaping their beliefs, aspirations, and behaviors. Young people living in neighborhoods in which many residents live in poverty, experience persistent joblessness, and have low levels of education are themselves likely to develop limited educational aspirations and, thus, are more likely to leave school before graduating. These contextual influences, according to the socialization perspective, might reinforce similar normative structures thought to exist in relatively disadvantaged families. This hypothesis is supported by our finding that the positive impact of neighborhood distress on the risk of school dropout is most pronounced among young blacks from single-parent households and among young whites from relatively poor families. Similarly, the collective socialization perspective implies that neighborhood distress will be most damaging to the educational experiences of adolescents who have the greatest contact with their neighbors and are most completely or exclusively integrated into the local normative environment. This hypothesis is supported by the fact that, among white adolescents, the effects of neighborhood socioeconomic conditions vary significantly by age and are stronger for young white women than for young white men. Presumably, the greater neighborhood involvement and social integration of young women and the more spatially circumscribed social networks of young adolescents increases their susceptibility to neighborhood norms that fail to encourage educational persistence. Not all of our evidence, however, supports the collective socialization perspective. For example, despite evidence that contact with neighbors tends to increase with length of residence, our results indicate that for both racial groups neighborhood disadvantage has a much more pronounced impact on the risk of dropping out for short-term residents than for long-term residents. Although seemingly contradicting the collective socialization approach, this negative interaction with length of residence lends support to the social capital perspective. According to this perspective, neighborhood disadvantage is less detrimental to long-term residents precisely because they have more extensive social networks within the neighborhood and more stable ties to local institutions. These network ties ostensibly provide access to forms of social capital, such as information about educational and economic opportunities, that can be translated into other, more tangible sources of capital. Although potential sources of social capital might be less beneficial overall in poor than in more affluent neighborhoods, the greater access to some social capital appears to provide for long-term residents a degree of insulation against the generally deleterious effects of neighborhood distress. In a similar way, black femalesÕ greater integration into local social networks might afford them access to levels of protective social capital not enjoyed by young black males, thereby ameliorating the negative impact of neighborhood distress on school completion for young black women. Also consistent with this perspective, adolescents from higher-income families may be protected from the damaging effects of neighborhood distress by having access to social networks whose members provide valuable forms of social capital, a notion supported by the significant interaction between neighborhood conditions and family income among whites. Finally, the social control perspective on neighborhood effects also receives partial support from our findings. The social control perspective essentially holds

692

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

that neighborhoods influence adolescentsÕ risk of dropping out of school by affecting the quantity and quality of supervision received by adolescents. In poor neighborhoods, this supervision is presumed lacking, leading to the observed additive effect of neighborhood distress on the risk of dropping out. However, the impact of a lack of supervision at the collective level may be counterbalanced by an increased level of supervision received at home. Thus, according to the social control perspective, the stronger effect of neighborhood distress on school dropout among black adolescents from single-parent households than from two-parent households indicates that, when both family and neighborhood level controls are absent, adolescents are especially likely to engage in non-normative behaviors. Similarly, the stronger impact of neighborhood disadvantage on school dropout among short-term residents than among long-term residents may result from particularly low levels of neighborhood supervision for recent movers to the neighborhood. And, the fact that black women appear to be somewhat insulated from the impact of neighborhood distress may reflect gender differences in levels of social integration that result in different levels of oversight for young black men and women. Clearly, then, our findings suggest that several different theoretical models are useful in interpreting the impact of neighborhood socioeconomic disadvantage on adolescentÕs risk of dropping out of school. Our results, for example, are generally consistent with the idea that neighborhood socioeconomic conditions affect adolescents by shaping the normative environment, the quality of institutions, and the opportunity structures to which they are exposed. However, not all adolescents are affected by these structural conditions to the same degree. Those with access to stable sources of social capital and are in position to receive support and oversight by area adults are likely to be somewhat protected from these contextual influences, even in the poorest of neighborhoods. Thus, a complete understanding of this micro-macro linkage is likely to be found in the confluence of extant theoretical perspectives on neighborhood effects. Although these variations in the effects of contextual conditions have valuable theoretical implications, their ramifications for public policy are equally important. By pointing to individual- and family-level conditions under which neighborhood disadvantage exerts its greatest influence, these results suggest that efforts to combat the deleterious impact of neighborhood disadvantage might be most effectively focused on certain types of adolescents. For example, programs to enhance educational opportunities for young women and to reduce the impact of neighborhood conditions on teenage childbearing may prove to be particularly fruitful in distressed neighborhoods occupied by whites, while efforts to enhance the social integration and social capital of young men in black communities are likely to produce the greatest results. In both black and white communities, programs to reduce the disruptive impacts of residential mobility and to enhance the social integration of the recently mobile are likely to be especially important for both black and white adolescents. Similarly, programs that focus on assisting the poorest students from distressed neighborhoods and those from single-parent households are likely to be most worthwhile.

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

693

Perhaps our most intriguing finding, however, concerns historical change in the effect of neighborhood distress on school dropout. In contrast to claims that transportation and communication advances have increasingly weakened the influence of neighborhood conditions on individual behavior, we find that over the past quarter-century the impact of neighborhood socioeconomic disadvantage on school discontinuation has remained consistently strong for white adolescents and has actually increased substantially among black adolescents. A possible explanation for the increasing salience of neighborhood socioeconomic distress for black adolescentsÕ school persistence is that, over time, disadvantaged black neighborhoods have themselves become more isolated—both socially and spatially—from middle-class neighborhoods and the role models, sources of support, and social capital that the latter provide. The increasing geographic concentration of poverty means that, more and more, severely disadvantaged black neighborhoods are surrounded entirely by other disadvantaged communities, and thus residents of these areas are decreasingly likely to encounter middle-class role models from contiguous or otherwise nearby areas. Future research might profit from exploring how the changing ecological position of disadvantaged neighborhoods, including their proximity to less disadvantaged communities, influences adolescent development. But a key implication of this finding is that, for the substantial segment of the population relegated to distressed neighborhoods, the pernicious repercussions of residential inequality are not likely to abate anytime in the near future.

Acknowledgments This research was supported by grants to the first author from the National Institute of Child Health and Human Development (RO3 HD39325), to the second author from NICHD (RO1 HD35560) and the National Science Foundation (SBR-9729797), and to the University at Albany Center for Social and Demographic Analysis from NICHD (P30 HD32041) and NSF (SBR-9512290). We thank the anonymous Social Science Research reviewers for their helpful comments on earlier drafts of the paper.

References Aaronson, D., 1997. Sibling estimates of neighborhood effects. In: Brooks-Gunn, J., Duncan, G.J., Aber, J.L. (Eds.), Neighborhood Poverty, Volume II: Policy Implications in Studying Neighborhoods. Russell Sage Foundation, New York, pp. 80–93. Abrahamson, M., 1996. Urban Enclaves: Identity and Place in America. St. MartinÕs Press, New York. Ainsworth-Darnell, J.W., Downey, D.B., 1998. Assessing the oppositional culture explanation for racial/ ethnic differences in school performance. American Sociological Review 63, 536–553. Ainsworth, J.W., 2002. Why does it take a village? The mediation of neighborhood effects on educational achievement. Social Forces 81, 117–152.

694

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

Allison, P.D., 1984. Event History Analysis: Regression for Longitudinal Event Data. Sage, Newbury Park, CA. Anderson, E., 1990. Streetwise: Race, Class, and Change in an Urban Community. University of Chicago Press, Chicago. Anderson, E., 1999. Code of the Street: Decency, Violence, and the Moral Life of the Inner City. W.W. Norton, New York. Astone, N.M., McLanahan, S.S., 1991. Family structure and high school completion: the role of parental practices. American Sociological Review 56, 309–320. Astone, N.M., McLanahan, S.S., 1994. Family structure, residential mobility, and school dropout: a research note. Demography 31, 575–584. Baumer, E.P., South, S.J., 2001. Community effects on youth sexual activity. Journal of Marriage and the Family 63, 540–554. Billy, J.O.G., Moore, D.E., 1992. A multilevel analysis of marital and nonmarital fertility in the United States. Social Forces 70, 977–1011. Billy, J.O.G., Brewster, K.L., Grady, W.R., 1994. Contextual effects on the sexual behavior of adolescent women. Journal of Marriage and the Family 56, 387–404. Blake, J., 1989. Family Size and Achievement. University of California Press, Berkeley, CA. Bott, E., 1957. Family and Social Network: Roles, Norms, and Relationships in Ordinary Urban Families. Havistock, London. Bourdieu, P., 1986. The forms of capital. In: Richardson, J.G. (Ed.), Handbook of Theory and Research for the Sociology of Education. Greenwood Press, New York, pp. 241–258. Brooks-Gunn, J., Duncan, G.J., Klebanov, P.K., Sealand, N., 1993. Do neighborhoods influence child and adolescent development? American Journal of Sociology 99, 353–395. Bursik, R.J., Grasmick, H.G., 1993. Neighborhoods and Crime: The Dimensions of Social Control. Lexington, New York. Burton, L.M., 1990. Teenage childbearing as an alternative life course strategy in multigenerational African American families. Human Nature 1, 123–143. Campbell, K.E., 1990. Networks past: a 1939 Bloomington neighborhood. Social Forces 69, 139– 155. Campbell, K.E., Lee, B.A., 1990. Gender differences in Urban neighboring. Sociological Quarterly 31, 495–512. Clark, R.L., 1992. Neighborhood Effects of Dropping Out of School Among Teenage Boys. Urban Institute. Mimeographed, Washington, DC. Coleman, J.S., 1990. Foundations of Social Theory. Harvard University Press, Cambridge, MA. Coleman, J.S., 1988. Social capital in the creation of human capital. American Journal of Sociology 94 (supplement), S95–S120. Connell, J.P., Halpern-Felsher, B.L., 1997. How neighborhoods affect educational outcomes in middle childhood and adolescence: conceptual issues and an empirical example. In: Brooks-Gunn, J., Duncan, G.J., Aber, J.L. (Eds.), Neighborhood Poverty, Volume I: Context and Consequences for Children. Russell Sage Foundation, New York, pp. 174–199. Connell, J.P., Halpern-Felsher, B., Clifford, E., Crichlow, W., Usinger, P., 1995. Hanging in there: behavioral, psychological, and contextual factors affecting whether African American adolescents stay in high school. Journal of Adolescent Research 10, 41–63. Corcoran, M., Gordon, R., Laren, D., Solon, G., 1992. The association between menÕs economic status and their family and community origins. Journal of Human Resources 27, 575–601. Coulton, C.J., Pandey, S., Chow, J., 1990. Concentration of poverty and the changing ecology of lowincome, Urban neighborhoods: an analysis of the cleveland area. Social Work Research and Abstracts 26, 5–16. Crane, J., 1991. The epidemic theory of ghettos and neighborhood effects on dropping out and teenage childbearing. American Journal of Sociology 96, 1226–1259. DiPrete, T.A., Forristal, J.D., 1994. Multilevel models: methods and substance. Annual Review of Sociology 20, 331–357.

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

695

Dornbusch, S.M., Ritter, P.L., Steinberg, L., 1991. Community influences on the relation of family statuses to adolescent school performance: differences between African Americans and Non-Hispanic Whites. American Journal of Education 38, 543–567. Duncan, G.J., 1994. Families and neighbors as sources of disadvantage in the schooling decisions of White and Black adolescents. American Journal of Education 103, 20–53. Duncan, G.J., Hoffman, S.D., 1991. Teenage underclass behavior and subsequent poverty: have the rules changed. In: Jencks, C., Peterson, P. (Eds.), The Urban Underclass. Brookings Institution, Washington, DC, pp. 155–174. Duncan, G.J., Raudenbush, S.W., 2001. Neighborhoods and adolescent development: how can we determine the links. In: Booth, A., Crouter, A.C. (Eds.), Does It Take a Village? Community Effects on Children, Adolescents, and Families. Lawrence Erlbaum, Mahwah, NJ, pp. 105– 136. Duncan, G.J., Connell, J.P., Klebanov, P.K., 1997. Conceptual and methodological issues in estimating causal effects of neighborhoods and family conditions on individual development. In: Brooks-Gunn, J., Duncan, G.J., Aber, J.L. (Eds.), Neighborhood Poverty, Volume I: Context and Consequences for Children. Russell Sage Foundation, New York, pp. 219–250. Elliott, D.S., Wilson, W.J., Huizinga, D., Sampson, R.J., Elliott, A., Rankin, B., 1996. The effects of neighborhood disadvantage on adolescent development. Journal of Research in Crime and Delinquency 33, 389–426. Ensminger, M., Lamkin, R.P., Jacobson, N., 1996. School leaving: a longitudinal perspective including neighborhood effects. Child Development 67, 2400–2416. Entwisle, D.R., Alexander, K.L., Olson, L.S., 1994. The gender gap in math: its possible origins in neighborhood effects. American Sociological Review 59, 822–838. Evans, W.N., Oates, W.E., Schwab, R.M., 1992. Measuring peer group effects: a study of teenage behavior. Journal of Political Economy 100, 966–991. Feiring, C., Lewis, M., 1991. The transition from middle childhood to early adolescence: sex differences in the social network and perceived self-competence. Sex Roles 24, 489. Fischer, C.S., 1984. The Urban Experience. Harcourt, Brace, Jovanovich, New York. Fitzgerald, J., Gottschalk, P., Moffitt, R., 1998. An analysis of the impact of sample attrition on the second generation of respondents in the michigan panel of income dynamics. Journal of Human Resources 33, 300–343. Fuhrer, R., Stansfeld, S.A., Chemali, J., Shipley, M.J., 1999. Gender, social relations and mental health: prospective findings from an occupational cohort (Whitehall II study). Social Science and Medicine 48, 77–87. Gans, H.J., 1962. The Urban Villagers. Free Press, New York. Gephart, M.A., 1997. Neighborhoods and communities as contexts for development. In: Brooks-Gunn, J., Duncan, G.J., Aber, J.L. (Eds.), Neighborhood Poverty, Volume I: Context and Consequences for Children. Russell Sage Foundation, New York, pp. 1–43. Gonzales, N.A., Cauce, A.M., Friedman, R.J., Mason, C.A., 1996. Family, peer, and neighborhood influences on academic achievement among African-American adolescents: one-year prospective effects. American Journal of Community Psychology 24, 365–387. Greer, S., 1962. The Emerging City: Myth and Reality. The Free Press of Glencoe, Chicago. Griffith, J., Frase, M., Ralph, J., 1989. American Education: The Challenge of Change. Population Reference Bureau, Washington, DC. Hagan, J., MacMillan, R., Wheaton, B., 1996. New kid in town: social capital and the life course effects of family migration on children. American Sociological Review 61, 368–385. Halpern-Felsher, B.L., Connell, J.P., Spencer, M.B., Aber, J.L., Duncan, G.J., Clifford, E., Crichlow, W.E., Usinger, P.A., Cole, S.P., Allen, L., Seidman, E., 1997. Neighborhood and family factors predicting educational risk and attainment in African American and White children and adolescents. In: Brooks-Gunn, J., Duncan, G.J., Aber, J.L. (Eds.), Neighborhood Poverty, Volume I: Context and Consequences for Children. Russell Sage Foundation, New York, pp. 146–173. Harker, R., Mahar, C., Wilkes, C., 1990. An Introduction to the Work of Pierre Bourdieu: The Practice of Theory. MacMillan, London.

696

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

Haveman, R.H., Wolfe, B., 1995. The determinants of childrenÕs attainments: a review of methods and findings. The Journal of Economic Literature 33, 1829–1878. Haveman, R., Wolfe, B.L., Spaulding, J., 1991. Childhood events and circumstances influencing high school completion. Demography 28, 133–157. Hill, M.S., 1992. The Panel Study of Income Dynamics: A UserÕs Guide. Sage, Newbury Park, CA. Hogan, D.P., Kitagawa, E.M., 1985. The impact of social status, family structure, and neighborhood on the fertility of black adolescents. American Journal of Sociology 90, 825–852. Hunter, A., 1974. Symbolic Communities: The Persistence and Change of ChicagoÕs Local Communities. University of Chicago Press, Chicago. Ingersoll, G.M., Scamman, J.P., Eckerling, W.D., 1989. Geographic mobility and student achievement in an Urban setting. Educational Evaluation and Policy Analysis 11, 143–149. Jargowsky, P.A., 1997. Poverty and Place: Ghettos, Barrios, and the American City. Russell Sage Foundation, New York. Jencks, C., Mayer, S.E., 1990. The social consequences of growing up in a poor neighborhood. In: Lynn Jr., L.E., McGeary, M.G.H. (Eds.), Inner-City Poverty in the United States. National Academy Press, Washington, DC, pp. 111–186. Kasarda, J., Janowitz, M., 1974. Community attachment in mass society. American Sociological Review 39, 328–339. Kasen, S., Cohen, P., Brook, J.S., 1998. Adolescent school experiences and dropout, adolescent pregnancy, and young adult deviant behavior. Journal of Adolescent Research 13, 49–72. Keller, S., 1968. The Urban Neighborhood. Random House, New York. Klepinger, D., Lundberg, S., Plotnick, R., 1995. Adolescent fertility and the educational attainment of young women. Family Planning Perspectives 27, 23–28. Lee, B.A., Campbell, K.E., Miller, O., 1991. Racial differences in Urban neighboring. Sociological Forum 6, 525–550. Logan, J.R., Alba, R.D., McNulty, T., Fisher, B., 1996. Making a place in the metropolis: locational attainment in cities and suburbs. Demography 33, 443–453. Maddala, G.S., 1983. Limited-dependent and Qualitative Variables in Econometrics. Cambridge University Press, Cambridge. MacLeod, J., 1987. AinÕt No MakinÕ It: Aspirations and Attainment in a Low-Income Neighborhood. Westview, Boulder, CO. Massey, D.S., Denton, N.A., 1993. American Apartheid. Harvard University Press, Cambridge, MA. McLanahan, S.S., 1985. Family structure and the reproduction of poverty. American Journal of Sociology 90, 873–901. McLanahan, S., Sandefur, G., 1994. Growing Up with a Single Parent: What Hurts, What Helps. Harvard University Press, Cambride, MA. Murnane, R.J., Willett, J.B., Levy, F., 1995. The growing importance of cognitive skills in wage determination. The Review of Economics and Statistics 77, 251–267. Pallas, A., 1987. School dropouts in the United States. In: Denton, W.T. (Ed.), Hot Topics Series: Dropouts, Pushouts, and other Casualties. Phi Delta Kappa, Center on Evaluation, Development, and Research, Bloomington, IN, pp. 23–40. Plotnick, R.D., Hoffman, S.D., 1999. The effect of neighborhood characteristics on young adult outcomes: alternative estimates. Social Science Quarterly 80, 1–18. Pollard, K.M., OÕHare, W.P., 1999. AmericaÕs Racial and Ethnic Minorities. Population Bulletin 54(1). Population Reference Bureau, New York. Pribesh, S., Downey, D.B., 1999. Why are residential and school moves associated with poor school performance? Demography 36, 521–534. Putnam, R.D., 1995. Bowling alone: AmericaÕs declining social capital. Journal of Democracy 6, 65–78. Quillian, L., 1999. Migration patterns and the growth of high-poverty neighborhoods, 1970–1990. American Journal of Sociology 105, 1–37. Ribar, D., 1992. A multinomial logit analysis of teenage fertility and high school completion. Economics of Education Review 12, 153–164.

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

697

Rumberger, R.T., 1987. High school dropouts: a review of issues and evidence. Review of Educational Research 57, 101–121. Sampson, R.J., 1988. Local friendship ties and community attachment in mass society: a multilevel systemic model. American Sociological Review 53, 766–779. Sampson, R.J., Groves, W.B., 1989. Community structure and crime: testing social-disorganization theory. American Journal of Sociology 94, 774–802. Sampson, R.J., Wilson, W.J., 1995. Race, crime, and Urban inequality. In: Hagan, J., Petersen, R. (Eds.), Crime and Inequality. Stanford University Press, Stanford, CA, pp. 37–54. Sampson, R.J., Raudenbush, S.W., Earls, F., 1997. Neighborhoods and violent crime: a multilevel study of collective efficacy. Science 277, 918–924. Sandefur, G., McLanahan, S., Wojtkiewicz, R., 1992. The effects of parental marital status during adolescence on high school graduation. Social Forces 71, 103–121. Sorin, G., 1990. The Nurturing Neighborhood: The Brownsville Boys Club and Jewish Community in Urban America, 1940–1990. New York University Press, New York. South, S.J., Crowder, K.D., 1997. Escaping distressed neighborhoods: individual, community, and metropolitan influences. American Journal of Sociology 102, 1040–1084. South, S.J., Crowder, K.D., 1999. Neighborhood effects on family formation: concentrated poverty and beyond. American Sociological Review 64, 113–132. Stein, M.R., 1960. The Eclipse of Community. Harper and Row, New York. Susser, I., 1982. Norman Street: Poverty and Politics in an Urban Neighborhood. Oxford University Press, New York. Survey Research Center, 1974. A Panel Study of Income Dynamics: Procedures and Tape Codes, 1974 Interviewing Year, Wave VII, Supplement. Institute for Social Research, The University of Michigan, Ann Arbor, MI. Suttles, G.D., 1968. The Social Order of the Slum: Ethnicity and Territory in the Inner City. University of Chicago Press, Chicago. Swaim, R.C., Beauvais, F., Chavez, E.L., Oetting, E.R., 1997. The effect of school dropout rates on estimates of adolescent substance use among three racial/ethnic groups. American Journal of Public Health 87, 51–55. Teachman, J.D., Crowder, K., 2002. Hierarchical linear models in family research: some conceptual and methodological issues. Journal of Marriage and the Family 64, 280–294. Teachman, J.D., Paasch, K., Carver, K., 1996. Social capital and dropping out of school early. Journal of Marriage and the Family 58, 773–783. Tienda, M., 1991. Poor people and poor places: Deciphering neighborhood effects on poverty outcomes. In: Huber, J. (Ed.), Macro–Micro Linkages in Sociology. Sage, Newbury Park, CA, pp. 244–262. Turley, R.N.L., 2002. Is relative deprivation beneficial? The effects of richer and poorer neighbors on childrenÕs outcomes. Journal of Community Psychology 30, 671–686. Turley, R.N.L., 2003. When do neighborhoods matter? The role of race and neighborhood peers. Social Science Research 32, 61–79. US Department of Education, 2000. Dropout Rates in the United States: 1999, National Center for Education Statistics 2001–2022. US Government Printing Office, Washington, DC. Wacquant, L., Wilson, W.J., 1989. The cost of racial and class exclusion in the inner city. Annals of the American Academy of Science 501, 8–25. Warren, R.L., 1963. The Community in America. Rand McNally, New York. Wehlage, G.G., Rutter, R.A., 1985. Dropping out: how much do schools contribute to the problem? TeacherÕs College Record 87, 374–392. Wellman, B., 1977. Who needs neighborhoods. In: Warren, R.L. (Ed.), New Perspectives on the American Community. Rand McNally, New York, pp. 218–223. Wellman, B., 1979. The community question: the intimate networks of East Yorkers. American Journal of Sociology 84, 1201–1231. White, M.J., 1987. American Neighborhoods and Residential Differentiation. Russell Sage Foundation, New York.

698

K. Crowder, S.J. South / Social Science Research 32 (2003) 659–698

Wilson, W.J., 1987. The Truly Disadvantaged. University of Chicago Press, Chicago. Wilson, W.J., 1996. When Work Disappears: The World of the New Urban Poor. Vintage Books, New York. Winquist Nord, C., Moore, K.A., Morrison, D.R., Brown, B., Myers, D.E., 1992. Consequences of teenage parenting. Journal of School Health 62, 310–318.