The Political Context of the Percent Black-Neighborhood Violence Link: A Multilevel Analysis

Transcription

1 Social Problems, 2015, 62, doi: /socpro/spu005 Article The Political Context of the Percent Black-Neighborhood Violence Link: A Multilevel Analysis María B. Vélez, Christopher J. Lyons, and Wayne A. Santoro University of New Mexico ABSTRACT A century of urban research has established that percentage black associates positively with violence at the neighborhood level. We extend traditional structural explanations for this association by drawing attention to the political contexts of cities that may influence the race-violence link. Drawing on insights from social movement and racial politics literatures, we contend that the relationship between percentage black and neighborhood violence will be attenuated in cities with greater black political opportunities and black mobilization. We examine this thesis using multilevel data from the National Neighborhood Crime Study that provide sociodemographic and violence data for census tracts nested within 87 large cities. We pair these data with city-level measures of black political opportunities and mobilization. Multilevel analyses reveal that the relationship between percentage black and violence varies substantially across cities, and that the average positive relationship often is attenuated, and reduced to statistical insignificance, in cities with favorable political contexts. We propose that the substantive and symbolic benefits set in motion by favorable political contexts lay the foundation for neighborhood organization against violence. KEYWORDS: neighborhoods; violence; race; political opportunity structures; social movements. A century of urban research documents a strong correlation between percentage black and neighborhood violence (Curtis 1975; Du Bois [1889] 1973; Peterson and Krivo 2005, 2010a; Pratt and Cullen 2005; Sampson 2012; Sampson and Wilson 1995; Shaw and McKay 1942). This robust association has served as a springboard for much investigation and theorizing into the connections between racial composition and neighborhood well-being. Dominant approaches call attention to the acute socioeconomic disadvantages that isolate black residents and their communities from important resources and opportunities (McNulty and Holloway 2000; Peterson and Krivo 2010a; The authors are grateful to Ruth Peterson and Laurie Krivo for support and data, Alma Hernandez for data assistance, Blake Boursaw for statistical advice, the editor and reviewers for helpful comments on prior drafts, and Arlene Santoro. The authors presented earlier versions of this article at the 2012 Annual Meetings of the American Sociological Association and the 2011 Annual Meetings of the American Society of Criminology. Direct correspondence to: María B. Vélez or Christopher J. Lyons, Department of Sociology, University of New Mexico, MSC , 1 University of New Mexico, Albuquerque, NM mvelez@unm.edu or clyons@unm.edu. VC The Author Published by Oxford University Press on behalf of the Society for the Study of Social Problems. For Permissions, please journals.permissions@oup.com 93

2 94 Vélez/Lyons/Santoro Sampson 2012; Sampson and Wilson 1995; Wilson 1987). These prevailing structural explanations trace racial inequality in the concentration of disadvantage to conditions that undermine organization against crime and posit that adjusting for socioeconomic inequality should largely explain the association between percentage black and neighborhood violence. The common persistent and positive association of percentage black and violence, even after accounting for socioeconomic inequalities (e.g., Hipp 2007; Krivo and Peterson 1996; McNulty 2001), however, has pushed scholars to look beyond the internal characteristics of neighborhoods to conceptualize the divergent social conditions of black and white areas (Peterson and Krivo 2010a). We extend traditional structural approaches by situating the neighborhood race-violence link within political dimensions of cities. In so doing, we elaborate on an embedded ecological approach, which takes seriously that neighborhoods are nested within cities that vary in their willingness to support minority constituents (Lyons, Vélez, and Santoro 2013). In line with public social control and political economy traditions, our embedded ecological approach also recognizes that city political environments shape outcomes for minority neighborhoods in important ways (Bursik and Grasmick 1993; Logan and Molotch 1987; Squires and Kubrin 2006; Vélez 2001). Building on our prior work that unpacks the relationship between immigrant concentration and violence (Lyons et al. 2013), we investigate whether political contexts influence the long-established connection between percentage black and neighborhood violence. 1 We contend that the association between percentage black and neighborhood violence depends on the degree to which cities present favorable political contexts for blacks. We conceptualize favorable political contexts along two dimensions: black political opportunities and political mobilization. Political opportunities capture the extent that the city is receptive or vulnerable to black political demands, as indicated by cities with significant shares of black elected officials. Political mobilization reflects the ability of blacks and their neighborhoods to organize and make public demands for change. We argue that favorable black political contexts lessen the neighborhood-level association between percentage black and violence and potentially render the link trivial net of common controls. Drawing on insights from social movement and racial politics literatures and our prior work on immigration and violence (Lyons et al. 2013), we suggest that the substantive and symbolic benefits that arise from black political opportunities and mobilization lay the foundation for neighborhood organization against violence. Substantively, favorable political opportunities and political mobilization can lead to policies and resources that advance the social, economic, and political standing of black communities (Browning, Marshall, and Tabb 1984; Karnig and Welch 1980; Fainstein and Fainstein 1996; Saltzstein 1989). Symbolically, favorable city political contexts can engender trust in the political system and other local institutions and guard against legal cynicism and hostility towards authorities that undermine civic engagement (Bobo and Gilliam 1990; Gay 2002; Rahn and Rudolph 2005; Williams 1998). We argue these substantive and symbolic benefits encourage black residents to engage in a host of actions that facilitate neighborhood organization against violence. We employ unprecedented data from the National Neighborhood Crime Study (NNCS), pioneered by Ruth Peterson and Lauren Krivo (2010b). The NNCS supplies multilevel demographic and violent crime data for 8,931 census tracts embedded within a representative sample of 87 large cities circa We pair these data with measures of city characteristics that reflect political dimensions relevant to black residents. Our study makes two contributions. First, our focus on the political realm of cities takes seriously scholarship from urban sociology, political economy, and public social control traditions that have long highlighted the forces beyond neighborhood borders that shape the 1 Although we adopt the same multilevel modeling strategy and core theoretical premise as Lyons and colleagues (2013), we push the conceptualization of the political context further to appreciate resources that facilitate black mobilization. Rather than asking solely how actors and structures in the larger political environment (political opportunities) affect minorities and their neighborhoods, we also examine how black advocacy organizations and protest affect neighborhood processes. This approach allows us to avoid taking an overly structural view by better appreciating how black agency the proactive actions of black residents shapes in part the fate of their neighborhoods.

3 Political Context of the Percent Black-Neighborhood Violence Link 95 fate of local areas (Bursik and Grasmick 1993; Logan and Molotch 1987; Squires and Kubrin 2006; Vélez 2001). Despite the apparent relevance of extraneighborhood political contexts for understanding neighborhood processes, sociologists have been slow to incorporate these insights into their models of neighborhood violence. Second, embedding the neighborhood race-violence link across multiple cities with varying political contexts allows us to shed light on divergent findings in previous research. Whereas a few studies that examine neighborhoods within single cities are able to explain away the percent black and violent crime relationship by adjusting for internal socioeconomic characteristics (e.g., Messner and Tardiff 1986; Shihadeh and Shrum 2004), many others still find positive associations net of controls (e.g., Hipp 2007; Krivo and Peterson 1996; McNulty 2001). We suggest that heterogeneity across cities especially in black political opportunities and mobilization partly explain these divergent findings. By attributing variation to both neighborhood and city levels with rare multilevel data, we are in a unique position to examine the contingency of the neighborhood relationship between percentage black and violence across political contexts. PERCENT BLACK AND NEIGHBORHOOD VIOLENCE Social scientists have long documented positive associations between violent crime and black racial composition at the neighborhood level (Du Bois [1889] 1973; Hipp 2007; Peterson and Krivo 2005, 2010a; Pratt and Cullen 2005; Sampson 2012; Sampson and Wilson 1995; Shaw and McKay 1942). An influential body of criminological work has in turn developed to explain this well-established relationship. Predominant racial invariance explanations focus on inequality in criminogenic conditions within neighborhoods (Shaw and McKay 1942, 1949). Neighborhood racial composition correlates with a variety of deleterious conditions, including poverty, male joblessness, and population loss that concentrate to produce distinct structural milieus conducive to crime. These conditions are prevalent especially in neighborhoods with relatively high percentages of blacks (Krivo and Peterson 1996; McNulty 1999; McNulty and Holloway 2000; Peterson and Krivo 2010a; Sampson and Wilson 1995; Shaw and McKay 1942; Shihadeh and Shrum 2004). William Julius Wilson (1987, 1996) contends that these disadvantaged conditions have increased the social isolation and marginalization of inner-city black neighborhoods, whereby residents routinely are deprived of resources, conventional role models and institutions, and cultural learning from mainstream social networks that facilitate social and economic advancement (Sampson and Wilson 1995). Socially isolated and marginalized contexts hamper social organization including mechanisms of social control, such as collective efficacy, that discourage criminal involvement (Bellair 1997; Peterson and Krivo 2010a; Sampson, Raudenbush, and Earls 1997; Wilson 1996). Such contexts also encourage violence by legitimating codes of legal cynicism and violence in light of limited economic opportunities (Anderson 1999; Sampson and Bartusch 1998). Using data on neighborhoods within a single or small number of cities, many studies test the above theoretical logic by examining how socioeconomic disadvantage within neighborhoods explains the race-violence link. Generally, studies find that accounting for internal indicators of disadvantage explains a portion of the relationship between percentage black and neighborhood violence. In a few cases, researchers are able to account for the positive association between percentage black and violence with controls for neighborhoods within a single city (e.g., Messner and Tardiff 1986; Shihadeh and Shrum 2004). More typically, however, studies report a residual association even after accounting for key neighborhood variation in the concentration of socioeconomic disadvantages (e.g., Hipp 2007; Krivo and Peterson 1996; McNulty 2001). Scholars often attribute this persistent and positive association of percentage black and violence to the imperfect measurement of structural disadvantage and social isolation. More recently, researchers have begun to look beyond the internal characteristics of neighborhoods to define the divergent social conditions of black and white neighborhoods. Peterson and Krivo (2010a), for instance, point to spatial inequalities whereby black neighborhoods are disproportionately embedded in areas with high levels of disadvantage and other crime producing

4 96 Vélez/Lyons/Santoro conditions. Likewise, Thomas McNulty and Steven Holloway (2000) examine proximity of public housing as extralocal factors that contribute to the race-crime link. We subscribe to the call to move beyond the internal characteristics of neighborhoods and propose that city contexts are a strategic site for elucidating the relationship between percentage black and neighborhood violence. As previous research demonstrates (Jargowsky 1997; Krivo, Peterson, and Kuhl 2009; Massey 1996; Massey and Eggers 1990; Peterson and Krivo 2010a; Wilson 1987, 1996), cities vary greatly in the conditions that set the stage for neighborhood crime. We extend this research by emphasizing the political contexts of cities that underlie the marginalization versus empowerment of blacks and their neighborhoods. BEYOND NEIGHBORHOOD BOUNDARIES: CITY POLITICAL CONTEXT While also acknowledging the importance of internal neighborhood processes, proponents of political economy and public social control traditions emphasize how larger political contexts, particularly at the city level, facilitate or impede neighborhood viability (Logan and Molotch 1987; Smith, Caris, and Wyly 2001; Squires and Kubrin 2006). Numerous scholars argue that city political contexts influence the socioeconomic realities of minority and poor neighborhoods especially. A prime example of this is the active hand city governments played in concentrating poverty in black neighborhoods by building large-scale public housing in already impoverished areas (Farley, Danziger, and Holzer 2002; Hirsch 1998; Massey and Denton 1993; Sampson and Wilson 1995). Actions like these contributed to greater levels of violence and related social problems in minority neighborhoods (Bursik 1989; McNulty and Holloway 2000). Increasingly scholarship highlights the political forces that undergird social organization against crime (Bursik and Grasmick 1993; Carr 2005; Silver and Miller 2004; Vélez 2001; Vélez and Richardson 2012). We argue that cities with favorable black political contexts will attenuate and possibly nullify the positive neighborhood-level association between percentage black and violence. In contrast, politically unfavorable cities should exacerbate the percentage black-neighborhood violence nexus. This means that percentage black may heighten neighborhood violence only in cities where blacks experience a hostile political context. We now elaborate on two key dimensions of favorable black political contexts. Black Political Opportunities Political process theory, the dominant approach among contemporary social movement scholars, argues that the political opportunity structure profoundly shapes the fate of movement constituencies (Meyer 2004). The political opportunity structure refers to the receptivity or vulnerability of the political system to organized protest by a given challenging group (McAdam 1982). Politically open regimes are receptive or vulnerable to constituency issues, evident in part by regime willingness to listen to their concerns or grant concessions to their demands. In contrast, closed regimes are unresponsive to challengers and may in fact repress them. Common conceptualizations of open political opportunities include the receptivity of politicians to constituent demands (Costain 1992), bureaucratic incorporation of constituents into civil service positions (Eisinger 1973), preexisting policies allowing for constituent input into political decisions (Kitschelt 1986), and the amenability of the audience to constituent issues (Santoro 2008). Grounded in this literature, we offer the concept black political opportunities to capture the extent that cities are receptive or vulnerable to the grievances of blacks and their neighborhoods (cf. Lyons et al. 2013). We conceptualize black political opportunities in four ways. First, in line with Peter Eisinger s (1973) original articulation, open regimes have significant shares of black elected representation because such municipalities often enact or facilitate substantive improvements for black communities. For instance, black politicians have increased black representation on city commissions, led to greater black public-sector employment, allowed citizens more control over the police department, and more

5 Political Context of the Percent Black-Neighborhood Violence Link 97 equitably distributed city contracts to black businesses (Betancur and Gills 2004; Browning et al. 1984; Mladenka 1989; Saltzstein 1989). Research also demonstrates that cities with black politicians have fewer police killings of blacks and lower rates of interracial violence (Jacobs and Carmichael 2002; Jacobs and Wood 1999; Kent 2010). Second, the bureaucratic incorporation of blacks into civil service positions, such as in the police department, reflects black political opportunities as bureaucratic actors can institutionalize minority friendly procedures and lead to amicable interactions with black residents (Marrow 2009; Theobald and Haider-Markel 2008). Third, regimes with preexisting policies favorable to the input of residents indicate a degree of regime vulnerability to black concerns. We focus on the presence of civilian police review boards whose purpose is to provide mechanisms for residents to investigate police misconduct and push for police reforms. Such review boards have performed well in some cities but poorly in others where they can be underfunded and reluctant to hold police accountable for misconduct (Harris 2005). Nonetheless, even underfunded review boards indicate that the regime is open (if only symbolically) to citizen input. In fact, Ronnie Dunn (2010) finds that black residents in Cleveland are willing to use the complaint system despite knowing that it is ineffective because it provides a forum to voice their grievances. Finally, cities with liberal voters should be more likely to elect politicians willing to respond favorably to black demands (Browning et al. 1984). Black Political Mobilization Regardless of the predispositions or vulnerabilities of local regimes, marginalized communities often must politically mobilize to have their voices heard. Black political mobilization reflects the ability of blacks and their neighborhoods to organize and make public demands for change. Existing data allow us to examine two forms of political mobilization. First, we look to citywide levels of minority advocacy organizations. Social movement scholars highlight the ability of indigenous organizations to facilitate political mobilization by developing local leadership, marshaling community resources, engendering solidarity among participants, and coordinating strategies for social change (Ganz 2000; Hirsch 1990; Morris 1984). Indeed, a major theoretical advancement in the study of social movements occurred when scholars recognized the centrality of organizations to explain the ability of marginalized constituencies to mobilize and press for gains (Gamson 1975; McCarthy and Zald 1977; Tilly 1978). Black organizations in the sixties played a key role in generating protest that pushed local and federal officials to pass legislation beneficial to the black community (Andrews 1997; Button 1989; McAdam 1982; Morris 1993; Santoro 2002, 2008). More recently, minority rights organizations have negotiated with local governments to bring about beneficial municipal policies. Wayne Santoro (1995) finds that cities with advocacy organizations like local chapters of the NAACP were more likely to adopt affirmative action policies related to public-sector employment. Sandra Bass (2000) documents that minority advocacy organizations in Seattle and Oakland helped secure improvements in police accountability by lobbying city council members, conducting studies on how police handled complaints, and educating citizens on effective interactions with police officers. Dennis Keating, Norman Krumholz, and Philip Star (1996) discuss examples in Boston, Chicago, Minneapolis, and Los Angeles where politically active neighborhood organizations worked with receptive political regimes to promote city service delivery, upgrade housing stock, and halt physical decline. Second, we look to prior levels of black insurgency in the form of riots and nonviolent protests. Black communities vary considerably in the degree to which they have been able to generate public demonstrations, especially in the post-civil rights movement era. We contend that cities with significant levels of black insurgency provide incentives for elites to take seriously the contemporary political demands of black constituents. Policymakers tend to be more willing to listen to the political demands of constituencies who have demonstrated their ability to generate public protest (Gamson 1975; Lipsky 1968; Morris 1993; Santoro 2002). Moreover, prior insurgency may have led to changes

6 98 Vélez/Lyons/Santoro in the permeability of city governance that continue to facilitate resident engagement with and trust of public institutions. Take, for instance, the connection between riots and policing. L. W. Sherman (1983) and Jack Greene (2000) note that the riots of the sixties and seventies were instrumental in compelling police departments to take more community-centered approaches to improve relationships with black communities. Others link the 1992 Los Angeles riots to reforms in the local police department, including community policing, increased foot patrol, and neighborhood watch programs (Chung 2001; Friedman 1992). THE MODERATING ROLE OF POLITICAL CONTEXT While testing specific causal processes for the wide range of neighborhoods and cities is beyond our data repository, we turn now to a discussion of why favorable political contexts should attenuate the neighborhood-level association between percentage black and violence. Here, we expand on Lyons and colleagues (2013) who find that the inverse or protective association between immigration and neighborhood violence is stronger in cities supportive of immigrant concerns. We propose that favorable political contexts create conditions at the neighborhood level that bolster social organization. Neighborhood social organization refers to the ability of local areas to realize common goals such as safety (Bursik 1988). Models of social organization highlight the complex network of social relationships, ranging from friendship and kinship ties to more wide ranging associations with external actors and institutions, which set the stage for crime control. Robert Sampson (2012; see also Sampson et al. 1997) conceptualizes collective efficacy, defined as the linkage of cohesion and mutual trust among residents with shared expectations for intervening in support of neighborhood social control, as a key component of social organization. To develop our discussion of mechanisms, we emphasize the substantive and symbolic benefits that emanate from favorable political contexts, which we argue fortify and encourage community social organization against violence. Substantive Benefits Political regimes favorable to blacks face structural constraints that can limit their ability to redress the grievances of black residents (Eisinger 1980; Mladenka 1989; Patillo 2007; Peterson 1981; Reed 1988). These barriers include budget shortfalls and the need to maintain coalitions with whites. Even so, an impressive array of racial politics scholarship indicates that cities with favorable political opportunities and/or high levels of black political mobilization can yield a variety of substantive improvements for blacks and their communities (Betancur and Gills 2004; Browning et al. 1984; Jacobs and Carmichael 2002; Karnig and Welch 1980; Kent 2010; Lewis and Ramikrishan 2007; Saltzstein 1989; Santoro 1995; Sass and Mehay 2003). Local regimes favorable to black demands can stimulate neighborhood organization against crime (a process criminologists term public social control; Vélez 2001) by reallocating municipal funds in ways that benefit minority communities, initiating policies that improve the day-to-day interactions between black residents and city officials, and implementing democratizing policies that enable greater voice for black communities. Harold Washington, Chicago s first black mayor, illustrates how favorable regimes can lead to substantive benefits (Betancur and Gills 2004). Washington mandated that private firms hire blacks if they were to receive governmental assistance, set aside 25 percent of public contracts and purchases for black and other minority firms, and financed neighborhood-based projects. Washington opened up the political process to include blacks as partners in law enforcement and school policy making, and pushed to redraw city districts that eventually produced four new minority-controlled wards. Moreover, to improve relationships between the police and black youth and their communities, he implemented a community policing program, juvenile restoration boards, and a people s court. Actions like these led to greater resources for neighborhoods and better neighborhood relations with city service providers. Increased resources contributed to lower levels of marginalization and isolation in neighborhoods with relatively large numbers of blacks. Insofar as favorable regimes lead to substantive benefits for blacks, they lay a

7 Political Context of the Percent Black-Neighborhood Violence Link 99 foundation for the willingness and ability of communities to organize against crime (Shaw and McKay 1942; Silver and Miller 2004). Symbolic Benefits Favorable political contexts also can have symbolic benefits for black residents. We emphasize stimulation of trust in local government as a key social-psychological benefit of favorable political contexts. Numerous studies link political opportunities to greater trust among black residents in the civic process. This research documents that trust leads to greater legitimacy accorded to local government, encourages blacks to be more knowledgeable of and satisfied with government actions, and heightens perceptions of personal efficacy in political and social affairs (Bobo and Gilliam 1990; Gay 2002; Marschall and Ruhil 2007; Mansbridge 1999; Rahn and Rudolph 2005; Tate 1993; Theobald and Haider-Markel 2008; Williams 1998). Melissa Marschall and Anirudh Ruhil (2007), for example, find that blacks who live in cities with greater black political representation on the city council, school board, and police department are more satisfied with their neighborhood, their schools, and the police. Similarly, bureaucratic incorporation into the police force may guard against hostility between the police and black residents, as blacks are more likely to view interactions with minority police officers as legitimate and justified (Theobald and Haider-Markel 2008). Lawrence Bobo and Franklin Gilliam (1990) show that black elected representation creates more trust among blacks toward local politics, increases knowledge of political issues, and enhances the perception that individual actions can influence political and social causes (see also Rahn and Rudolph 2005; Tate 1993). Likewise, Melissa Williams (1998) and Claudine Gay (2002) argue that black political representation generates a spiral of trust that improves communication between officials and blacks and generates greater system-level trust in government on the part of black constituents. 2 Consistent with criminological research on the importance of trust for collective efficacy against crime (Sampson et al. 1997), these social-psychological factors should facilitate subsequent levels of black participation in political and civic affairs, attachment to and ownership of their neighborhoods, and further mobilization on behalf of neighborhood concerns (Bobo and Gilliam 1990; Tate 1993; Williams 1998). Bobo and Gilliam (1990), for instance, demonstrate that the social-psychological benefits of black empowerment lead to greater social and political participation among blacks, including voting, campaigning, and membership in groups and community organizations. By enhancing neighborhood social organization, cities with favorable political contexts may guard against isolation and marginalization, and consequent legal cynicism and political disaffection, faced by neighborhoods with relatively large black populations, and instead improve the ability of these neighborhoods to organize residents and local institutions to control crime. DATA AND METHODS To test our multilevel and conditional thesis, we utilize data from the NNCS (Peterson and Krivo 2010b), the first national data set to contain violence and sociodemographic information for tracts embedded within a representative sample of U.S. cities. The complete NNCS provides information for 91 cities (with populations of more than 100,000 in the year 2000) and 9,593 census tracts. Missing data on key variables (including spatial lags of violence) reduces our sample to 8,931 tracts nested in 87 cities. The sample of NNCS cities generalizes to most large urban places regarding crime levels, racial/ethnic composition, segregation, and economic disadvantage (Krivo et al. 2009; Peterson and Krivo 2010a). We merge these data with city-level indicators of black political opportunities and mobilization collected from a variety of secondary sources. 2 Whereas this research focuses mostly on how black political opportunities translate into symbolic benefits for blacks, we suggest similar benefits may derive from black political mobilization as well, though the association with improved trust may operate indirectly through the substantive benefits secured by the mobilization of black constituents.

8 100 Vélez/Lyons/Santoro Dependent Variable Following previous research on race and crime using the NNCS (Kubrin and Ishizawa 2012; Peterson and Krivo 2010a), we examine violent crime as the sum of the number of homicides and robberies between 1999 and 2001 at the tract level. 3 We use multiyear counts to minimize the impact of annual fluctuations, especially for smaller units (Peterson and Krivo 2010a). Homicide is the most accurately reported violent crime and robbery is the most reliably reported form of nonlethal violence (Arnio, Baumer, and Wolff 2012; Baumer 2002; Baumer and Lauritsen 2010). 4 Tract-Level Covariates At the tract level, we measure a number of well-established predictors of violence from the 2000 U.S. Census. Our central measure of interest is the percentage of non-latino blacks at the tract level. This operationalization is consistent with most research on neighborhood violence, which has conceptualized black racial composition as the percentage of blacks in a given neighborhood. A small number of more recent studies, however, compare mostly black areas with other neighborhood types with a series of indicator variables (e.g., Peterson and Krivo 2010a). We focus on the percentage of blacks in a neighborhood to be consistent with the lion s share of prior work. This approach also allows us to capture the experience of a more diverse range of neighborhoods where blacks live. In the NNCS, in fact, only 17 percent of tracts are predominantly black and about half of the black population lives in nonmajority black tracts. Nonetheless, we discuss the robustness of our findings to alternative operationalizations of racial composition (see additional analyses section). We control for other types of racial/ethnic composition with the percentage of residents who are Latino, foreign born, and Asian. We include a variety of additional predictors of neighborhood violence. Concentrated disadvantage is an index (average z-scores) of the poverty rate, extent of joblessness (percentage of persons aged 16 to 64 who are unemployed or out of the labor force), low-wage jobs (percentage of workers in the six occupations with the lowest average incomes), professional jobs (percentage of employed civilian population age 16 years and older in management, professional, and related occupations [reverse coded]), high school graduates (percentage of adults age 25 years and older with at least a high school degree), and the percentage of households that are single-mother families (a ¼.91). We create an index of residential instability by standardizing and averaging the percentage of renter-occupied units and the percentage of residents aged five or older who lived in a different dwelling in 1995 (a ¼.69). We capture the crimeprone population with the percentage of the population that is male and between ages 15 and 34. Finally, we calculate a spatial lag for our dependent variable of violent crime to adjust for spatial autocorrelation in violence. The spatial lag stands for the average of the violent crime count for census tracts that are adjacent geographically to the immediate tract. They are constructed by multiplying tract characteristic values by a row-standardized, first-order spatial contiguity matrix that utilizes a queen criterion (see Peterson and Krivo 2009, 2010a). City-Level Covariates We measure black political opportunities in five ways. We gauge black elected representation with 2000 data from the National Roster of Black Elected Officials published by the Joint Center for 3 Some cities report homicide and robbery for only one or two of the years between 1999 and In these instances, the threeyear count is an estimate compiled from available data. According to Peterson and Krivo (2010b), (w)hen two years of crime counts were provided, the estimate was calculated by multiplying the two year count by 1.5. When only a single year s crime count was available, the estimate was calculated by multiplying the single year count by 3 (p. 8). 4 We find similar findings to those reported below with respect to key cross-level interactions when we examine a broader measure of violence that includes rape and assault along with robbery and homicide, though missing data for rape and assault reduce the number of level-two units considerably. We also find similar patterns if we limit our analysis to homicide, although interactions are less robust (in terms of statistical significance) for homicide than for other forms of violence. We employ a combined homicide and robbery measure of violence to be consistent with the majority of research on percent black and neighborhood violence and to maximize the number of cities in our analysis.

9 Political Context of the Percent Black-Neighborhood Violence Link 101 Political and Economic Studies (2011). We first measure the presence of a black mayor, arguably indicating the greatest degree of local black political incorporation. Second, we tally the total number of blacks elected to other municipal offices excluding the mayor, such as members of the city council. We divide this total count by the total number of blacks in the city, and then multiply by a constant to represent the rate of elected officials per 1,000 blacks (Stults and Baumer 2007). Third, we capture bureaucratic incorporation with black representation in the police force in We construct a ratio of the percentage of the police force (sworn officers) that is black and the percentage of the city population that is black, where values below one indicate underrepresentation of the police force relative to the city population. Fourth, we measure whether cities have a civilian police review board to reflect the possibility for citizen oversight of local law enforcement. Data on police representation and civilian review boards come from the Law Enforcement Management and Administrative Statistics in Fifth, we measure audience receptivity to black issues with liberal voting records by calculating the proportion of votes cast for the Democratic presidential candidate (Gore) in Data for this measure come from the 2000 County and City Data Book published by the U.S. Census Bureau (2001). 5 Three variables measure black political mobilization. Using data on all registered nonprofits at the county level provided by the National Center for Charitable Statistics (2000) at the Urban Institute, we count the number of minority advocacy organizations that registered with the Internal Revenue Service (in the May 2000 file). Drawing on the National Taxonomy of Exempt Entities, we enumerate organizations listed under the categories civil rights, minority rights, intergroup and race relations, and civil rights, social action, and advocacy. 6 We divide this number by the black population for each city and multiply by 1,000. Our measure thus captures all organizations that support the passage and enforcement of policy outputs that promote the rights and interests of racial/ethnic minorities. We examine post-sixties levels of two forms of black insurgency with data collected by Susan Olzak and Elizabeth West (1995) from the archives of The New York Times. We sum the total number of black (nonviolent) protest events, such as boycotts and mass marches, from 1970 to 1992 and the total number of violent black protest-events, or riots, from 1970 to We transform these counts into the rate of protests or riots per 1,000 blacks in the city in We begin counting protests and riots in 1970 to postdate the civil rights movement, a period where distinctive and often national-level events affected city-level mobilization processes. We find substantively similar results to those reported below if we restrict riots and protests to events occurring post Although ideally our data would extend to 2000, Olzak and West (1995) did not collect data beyond We strongly suspect, however, that levels of black insurgency across the 23 years we observe would be highly related to insurgency for the remaining eight years. Using 2000 Census data (U.S. Census Bureau 2001), we control for several city-level characteristics associated with differential levels of neighborhood violence: the city-level disadvantage index, measured in a parallel fashion to the neighborhood indicator; manufacturing jobs (the percentage of employed civilian population aged 16 and over working in a manufacturing industry); residential mobility (percentage of the population age five and over who lived in a different residence in 1995); percentage non-latino black; percentage non-latino Asian; percentage Latino; the percentage of the total city population that is foreign born; and the percentage of young males ages 15 to 34. We also control for South and West regions, with cities in other regions as the referent. Finally, we take into account white-black residential segregation with an index of dissimilarity across census tracts within a city (Krivo et al. 2009). Table 1 lists means, standard deviations, and minimum and maximum values for all city- and tract-level variables. 5 Most cities in the NNCS are part of a single county, and so integrating county and city-level data is unproblematic. However, four cities in our sample are located in counties that encompass multiple cities. In these instances, we adjusted the percentage of Democratic votes by the portion of the city population that is located in each county. 6 This classification system captures an array of minority rights organizations, but does not allow us to isolate organizations that serve only blacks.

10 102 Vélez/Lyons/Santoro Table 1. Descriptive Statistics of Dependent and Independent Variables Mean SD Min Max Tract level (N ¼ 8,931) Violent crimes Spatial lag of violence Disadvantage Residential instability Black (%) Asian (%) Hispanic (%) Foreign born (%) Young males (%) City level (N ¼ 87) Disadvantage Manufacturing jobs Residential mobility Black (%) Asian (%) Hispanic (%) Foreign born (%) Young males (%) South West White/black segregation Black mayor Black elected officials (per 1,000 blacks) Black police representation Civilian police review board Proportion votes Democrat Minority advocacy organizations (per 1,000 blacks) Black protests (per 1,000 blacks) Black riots (per 1,000 blacks) ANALYTIC STRATEGY We estimate a series of hierarchical generalized linear models that account for the nonindependence of observations, with 8,931 tracts (level-one units) clustered within 87 cities (level-two units). We grand-mean center all continuous variables. As we analyze relatively rare events within small units, we fit Poisson models with violence counts as the outcome. The variance in our measure of violence is considerably larger than its mean, indicating significant overdispersion. We account for overdispersion in the level-one variance. In multilevel models, a Poisson model with overdispersion is analogous to a negative binomial model. We specify that these counts have variable exposure by tract population and thereby transform the outcome to violent crimes per capita rates (Osgood 2000). By estimating random intercepts for the dependent variable, multilevel models also help address measurement error in homicides and robberies (Snijders and Bosker 1999). We investigate whether the association between percentage black and violence varies significantly across cities by testing for random variation in the slope of percentage black. Variance components within a multilevel framework allow us to gauge the statistical and substantive significance of random

11 Political Context of the Percent Black-Neighborhood Violence Link 103 slope variation for percentage black, net of common controls. Insofar as random slope variation exists, we assess whether the relationship between percentage black and violence is always positive across cities. We then account for any random variation in the estimate of percentage black on violent crime with cross-level interactions that test the moderating potential of city-level political characteristics. 7 RESULTS Table 2 presents baseline multilevel models of the association between percentage black and violent crime. Model 1 estimates a reduced-form model without controls for tract disadvantage and the spatial lag of violence, key variables proposed to mediate some of the relationship between percentage black and violence (Peterson and Krivo 2010a). In this reduced-form model, the coefficient for percentage black is substantially positive and significant, consistent with expectations of previous research. 8 This represents the average relationship between percentage black and violence for 8,931 tracts across 87 large cities. However, we find that the slope for percentage black varies substantially across cities, as evidenced by the significant variance components (bottom panel of Table 2). In Model 2, we introduce controls for disadvantage and the spatial lag of violence. As expected by prevailing structural frameworks, the introduction of these additional controls reduces the estimate for tract-level percentage black substantially (from.025 in Model 1 to.005 in Model 2). Yet the average estimate remains positive and statistically significant. Furthermore, even in this more fully specified model, the relatively large standard deviation of the random slope for percentage black (.016) indicates substantial variation in the association between percentage black and violence across cities. Specifically, the estimated association between percentage black and violence ranges above and below zero for 95 percent of cities in our sample (calculated by.005 þ / [2*.016]; see Snidjers and Bosker 1999). To our knowledge, these results provide the first evidence across a large sample of cities that the relationship between percentage black and violence is geographically variant net of controls. Table 3 investigates the extent to which black political opportunities and mobilization moderate the percentage black-neighborhood violence relationship. To do so, we test for cross-level interactions between tract-level percentage black and each of the eight city-level measures. Models 1 and 2 indicate support for our thesis that cities with greater black elected representation not only attenuate the race-violence link, but also nullify it. Model 1 shows a negative product term between black mayor and percentage black, and Model 2 reveals a similar significant interaction with the number of other black municipal officials. We display these interactions graphically with predicted values. Figure 1 illustrates that in cities without a black mayor, the predicted relationship between percentage black and neighborhood violence is positive. In contrast, cities with black mayors (17 percent of our sample) reveal a slightly inverse but nonsignificant association. Net of controls, this indicates that percentage black plays no role in predicting neighborhood violence in cities with black mayors. Figure 2 shows similar patterns via the interaction with black incorporation in other municipal offices. The figure separates cities into those with no black elected officials (38 percent of the sample), average representation, and high representation (1.5 standard deviations above the mean). In cities with significant black political incorporation, the predicted values reveal a negative slope for percentage black. Additional tests indicate that the negative slope for cities with above average 7 Following Peterson and Krivo (Krivo et al. 2009; Peterson and Krivo 2010a), we also let the slope of concentrated disadvantage vary randomly in all models. With minor exceptions, results are similar to those reported below if we do not allow tract disadvantage to vary randomly across cities. The exceptions are that, without a random slope for disadvantage, the cross-level interaction between percentage black and the rate of black elected officials is no longer significant, whereas cross-level interactions with proportion votes Democrat and the presence of a civilian review board become more efficient (p <.05). 8 In supplementary models, we find some evidence of nonlinearity in the association between percentage black and violence with a significant and negative coefficient for percentage black squared. The estimate for the quadratic term, however, is substantively small and at no point in the distribution is the slope for percentage black less than zero. Cross-level interactions between citylevel contexts and linear terms for percentage black mirror those reported below with the addition of the quadratic term.

12 104 Vélez/Lyons/Santoro Table 2. Multilevel Poisson Models (with variable exposure) of Neighborhood Violence, National Neighborhood Crime Study 2000 Model 1 Model 2 b SE b SE Tract level (N ¼ 8,931) Disadvantage.693***.037 Residential instability.335*** ***.011 Black (%).025*** **.002 Asian (%).005** Hispanic (%).013*** Foreign born (%) ***.002 Young males (%) ***.002 Spatial lag of violence.005***.001 City level (N ¼ 87) Disadvantage.263** Manufacturing jobs.017* **.008 Residential mobility Black (%) **.004 Asian (%) Hispanic (%) *.004 Foreign born (%).018** Young males (%) South *.087 West White/black segregation.020*** ***.003 Intercept 4.766*** ***.075 Variance components (SD) Intercept Black (%) Tract disadvantage.397 Notes: SE ¼ robust standard errors; tract population as variable exposure. * p <.1; ** p <.01; *** p <.001 (two-tailed tests) representation is not significantly different from zero; that is, percentage black is unrelated to violent crime in these cities. In contrast, the positive slopes for cities with no or average black representation are significant at traditional levels. In terms of black bureaucratic incorporation into the police force, although the product term in Model 3 is negative it is not significant statistically. In contrast, Model 4 indicates that in cities that have a civilian police review board, the relationship between percentage black and violence crime is null; only in cities without this policy is the slope positive and significant. Similarly in Model 5, the product term for the interaction between the percentage of votes cast for Gore in 2000 and tract-level percentage black indicates that cities with liberal voting patterns undercut the positive relationship between percentage black and violence. Cities with high levels of Democratic Party voting (1.5 standard deviations above the mean) render the association between percent black and violent crime not significant. In contrast, in cities with limited Democratic Party voting (1.5 standard deviations below the mean), the relationship is positive and significant. In sum, these findings suggest that percentage

13 Political Context of the Percent Black-Neighborhood Violence Link 105 Table 3. Multilevel Poisson Models (with variable exposure) of Neighborhood Violence with Cross-Level Interactions, National Neighborhood Crime Study 2000 Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 b SE b SE b SE b SE b SE b SE b SE b SE Tract level (N ¼ 8,931) Disadvantage.677*** *** *** *** *** *** *** ***.037 Residential instability.241*** *** *** *** *** *** *** ***.011 Black (%).006** ** ** ** ** ** **.002 x Black mayor.007**.003 x Black elected officials.055**.028 x Black police representation x Civilian police review board.007*.004 x Prop. votes Democrat.027*.015 x Minority advocacy organizations.004***.001 x Black protests x Black riots.036**.015 Asian (%) Hispanic (%) Foreign born (%).010*** *** *** *** *** *** *** ***.002 Young males (%).008*** *** *** *** *** *** *** ***.002 Spatial lag of violence.005*** *** *** *** *** *** *** ***.001 City level (N ¼ 87) Disadvantage Manufacturing jobs.022** ** *** ** ** ** ** **.007 Residential mobility Black (%).011** ** ** ** ** ** ** **.004 Asian (%) Hispanic (%) * * * * (continued)