Youth Violence 10.1177/1541204004267780 Brezina et al. / THE and CODE JuvenileOF Justice THE STREET

THE CODE OF THE STREET A Quantitative Assessment of Elijah Anderson’s Subculture of Violence Thesis and Its Contribution to Youth Violence Research Timothy Brezina Tulane University

Robert Agnew

Emory University

Francis T. Cullen John Paul Wright

University of Cincinnati Based on extensive field research, Elijah Anderson argues that the behavior of many youths is influenced by a street culture or “code” that prescribes violent reactions to interpersonal attacks and shows of disrespect. Although Anderson’s account has been well received by the criminological community, questions remain about the validity and generality of his findings. To address these issues, the authors review other (mostly quantitative) studies of youth violence and consider whether the findings of these studies are consistent with Anderson’s account. The authors also conduct analyses to determine whether Anderson’s observations can help to extend the existing body of quantitative research and, hence, further our understanding of youth violence. Overall, the results of this assessment suggest that Anderson’s observations provide valuable insight into the general problem of youth violence and that future research in this area would benefit from greater attention to the social processes he has served to highlight. Keywords: youth violence; strain theory; ethnographic research; victimization

Few writings in contemporary criminological theory match the eloquence and, arguably, the insight of Elijah Anderson’s (1999) Code of the Street. Drawing on extensive field research, Anderson observed that the behavior of many low-status youths is influenced by a street culture or “code” that prescribes violent reactions to interpersonal attacks and shows of disrespect. At the heart of this code is the belief that it is necessary to Authors’ Note: An earlier version of this article was presented at the 2001 meetings of the American Sociological Association, Anaheim. The authors would like to thank Elijah Anderson, Scott N. Brooks, and members of the Tulane Sociology Research Group for the helpful comments they provided on earlier drafts. Address all correspondence to Dr. Timothy Brezina, Department of Sociology, Tulane University, 220 Newcomb Hall, New Orleans, LA 70118; e-mail: [email protected]. Youth Violence and Juvenile Justice, Vol. 2 No. 4, October 2004 303-328 DOI: 10.1177/1541204004267780 © 2004 Sage Publications

303

304

Youth Violence and Juvenile Justice

show no tolerance for interpersonal transgressions; otherwise, one displays weakness to others and will likely become the target of further transgressions (also see Anderson, 1990, 1994, 1998). An important argument in Anderson’s (1999) account is that adherence to the street code (hereafter, the code) is not simply a reflection of corrupt values or deviant socialization. Rather, it represents in part an adaptation to status insecurity and to the persistent threat of violence that is present in some urban communities. In such communities, many young men believe that enactment of the code is a necessary aspect of street survival whether or not they are personally inclined to physical aggression and whether or not they are completely committed to elements of the code (Anderson, 1999). Anderson’s “code of the street” thesis is of special interest to criminologists because it represents a particularly rich and detailed example of a subcultural account of violence (also see Bernard, 1990; Luckenbill & Doyle, 1989; Wolfgang & Ferracuti, 1967). As some criminologists have observed, “Anderson’s account is perhaps the best description of a subculture of violence now available, and it is important because it makes us aware of the complexity of this subculture” (Cullen & Agnew, 1999, p. 107; also see Kasinitz, 1992; Short, 1991). The ethnographic nature of Anderson’s research is partly responsible for its level of detail and complexity and is arguably one of the strengths of his work. The ethnographic approach, however, is not without certain limitations. As Anderson (1999) has recognized, the goal of objectivity is not easy to achieve in the context of ethnographic research because it requires researchers to try to set aside their own values and assumptions about what is and is not morally acceptable—in other words, to jettison the prism through which they typically view a given situation. By definition, one’s own assumptions are so basic to one’s perceptions that seeing their influence may be difficult, if not impossible.1 (p. 11)

Although Anderson is clearly sensitive to the subjectivity inherent in ethnographic research, this sensitivity has not shielded his work from allegations of “observer bias,” as seen in a recent debate between Anderson (2002) and Wacquant (2002) over the pitfalls of urban ethnography. Moreover, Anderson’s observations are limited to a specific geographical location, encompassing two neighborhoods in Philadelphia. Thus, although Anderson (1999, p. 9) hopes that his research “may offer insight into the problem of youth violence more generally,” questions have also been raised about the generality of his findings (see Short, 1991). One possible way to answer such questions is to supplement Anderson’s qualitative observations with quantitative data. If quantitative data are consistent with Anderson’s observations, this should increase confidence in the plausibility and generality of his findings, especially if such data are collected by a variety of researchers and are drawn from other areas. As we describe below, relevant data are already available in the form of existing quantitative studies on youth violence. Although not designed to provide a direct test of Anderson’s account, the results of several previous studies have implications for the validity and generality of Anderson’s findings and conclusions. Thus, to facilitate an assessment of Anderson’s subculture of violence thesis, we review other (mainly quantitative) studies of youth violence and consider whether the findings of these studies are consistent with his account. We also conduct our own analyses to build on the existing body of research and to help assess the generality of Anderson’s

Brezina et al. / THE CODE OF THE STREET

305

findings. In particular, we analyze longitudinal data from a national survey of adolescents to examine key social processes anticipated by Anderson involving the demographic and social correlates of code-related beliefs and the impact of such beliefs on subsequent violent behavior. Drawing on this general sample of youth, we also examine the extent to which code-related beliefs mediate the effects of other variables on subsequent violence. Before turning to our assessment, we provide a brief review of Anderson’s subcultural account and derive a number of testable hypotheses. The Code of the Street Nature of the Code According to Anderson (1999), people in mainstream society generally do not “feel required to retaliate physically after an attack” (p. 76). With the exception of self-defense during a physical attack, most individuals generally prefer to avoid physical confrontation. In fact, mainstream people may feel that true “nerve” is displayed by walking away from a fight. This attitude partly reflects the relatively secure status of mainstream people. Mainstream persons can afford to walk away from public shows of disrespect because, although aware that they have been humiliated, they also know that they can leave with their self-esteem intact. After all, they typically have other areas of life where their self-worth is validated and accorded legitimacy, such as their occupation or profession (also see Rubin, 1976). Unlike their mainstream counterparts, members of low-status groups possess fewer legitimate means of demonstrating their worth and competence. This is said to be especially true of young men belonging to the “urban underclass” due to the absence of steady work in their communities and to the traditional emphasis on work as a means of constructing a “respectable” masculine image (also see Messerschmidt, 1993; Wilson, 1997). Consequently, these persons have less freedom to walk away from a fight because their status is relatively insecure. Moreover, to the extent that such persons see little hope of future social advancement, they may believe that their sense of self-worth is dependent on the ability to command respect in public. In their experience, this may be one of the few avenues through which respect can be obtained (also see Cloward & Ohlin, 1960; Stoll, 1974; Wilkinson, 2001; Zinn, 1982). Along with the need to maintain status and respect, individuals in disadvantaged communities may also face a need to cultivate a “tough” reputation for the purpose of selfprotection. Transgressions against the self—even the appearance of transgression—must be avenged. One must demonstrate a taste for violent counterattack; otherwise, one appears weak and runs the risk of being “rolled on” or physically assaulted “by any number of others” (Anderson, 1999, p. 73). The importance of such a reputation is highlighted by the fact that, in certain communities, “there are always people around looking for a fight”— people who actively campaign for respect on the streets by taking it away from others (Anderson, 1999, p. 73). It is also highlighted by the fact that, in such communities, people may have little confidence in the ability of the police to provide adequate protection (also see Kubrin & Weitzer, 2003). These facts of life in the urban ghetto are ultimately reflected in a street code—or “code of conduct”—that embodies a set of informal rules and expectations governing

306

Youth Violence and Juvenile Justice

public interaction. Among other things, these rules and expectations prescribe violent retaliation in response to real or perceived attacks against the self.2 Based on our interpretation of Anderson’s account, it appears that this code is understood and justified by individuals in terms of three interrelated beliefs or propositions: The first proposition embodies basic street wisdom and states that, to retain respect and deter future transgressions, it is necessary to respond to public assaults against the self with physical retaliation. The second proposition states that, given the first proposition, public attacks or shows of disrespect can be expected to elicit a violent response from streetwise persons. The third and final proposition flows from the assumption that the first two propositions are self-evident, and that everyone “has the opportunity” to know and learn them (Anderson, 1999, pp. 33, 91-92). It states that violent counterattacks against transgressors are morally justified because anyone who has engaged in public attack (or has given the appearance of attack) should have anticipated the possibility of physical retaliation and, hence, assumes responsibility for the consequences. Before proceeding, it should be noted that an important distinction exists between Anderson’s account and classic formulations of the subculture of violence thesis (e.g., Wolfgang & Ferracuti, 1967). Earlier theorists understood code-related values and beliefs to be relatively stable elements that had been internalized by most, if not all, members of high-crime communities. The importance attached to such values and beliefs, it was believed, transcended situational demands and served to define the overall culture of the entire community. According to Anderson (1998), however, “the code of the street applies predominantly to public behavior and is normative only for the most alienated and socially isolated segment of the community” (p. 69, note 1). Moreover, Anderson noted that, unless they are completely immersed in street behavior, young people tend to engage in “codeswitching”: that is, conforming to the expectations of the code in public encounters—where it may be necessary or justified—while conforming to more conventional norms in private settings. Learning the Code In the urban neighborhoods described by Anderson (1999), familiarity with the code typically begins early in life. From parents and other streetwise adults, many children are exposed to the following rules and advice: “Watch your back.” “Protect yourself.” “Don’t punk out.” “Respect yourself.” “If somebody messes with you, you got to pay them back.” “Don’t you come in here crying that someone beat you up; you better get back out there and whoop his ass.” (Anderson, 1999, pp. 70-71)

Lessons of this nature are especially likely to be handed down to children from socalled “street” families, who “grow up hard,” and whose home lives are marked by drugs, alcohol, abuse, and neglect.3 In the most desperate cases, parents fail to provide adequate supervision and leave children to fend for themselves. Such children “may well have learned the first lesson of the street: survival itself, let alone respect, cannot be taken for granted; you have to fight for your place in the world” (Anderson, 1999, p. 49). A commitment to nonviolence has a chance to prevail among children from homes that reflect mainstream values. Nevertheless, parents who worry about the safety of their children often recognize that “nonviolence prevails at the cost of relinquishing a claim to

Brezina et al. / THE CODE OF THE STREET

307

being bad and tough and therefore sets a young person up as . . . quite possibly a target of derision or even violence” (Anderson, 1994, p. 94). Thus, even parents who are strongly oriented to mainstream values (i.e., most parents) may stress the importance of observing the code in public encounters. Moreover, the demands of the street may ultimately undermine mainstream orientations to violence. Whether or not violent behavior is discouraged at home, young people learn the importance of the code through their personal experiences with violence, including rough and tumble play and fights with peers. They also witness the consequences of violence and nonviolence by observing the behavior of adults in their community. Through such experiences, many young people learn that, on the streets, respect is something to be negotiated; the code provides the framework for negotiating respect; toughness is both a virtue and a necessity; and a reluctance to adopt the code can carry a heavy price tag in terms of status and personal safety. Specifying Key Hypotheses If the above account is valid and if Anderson’s observations can be generalized beyond the communities he examined, we would expect to observe a number of patterns or associations. First, we should find that adherence to the code (or to code-related beliefs) is associated with a range of demographic factors, including race, low socioeconomic status (SES), and residence in urban neighborhoods—especially neighborhoods characterized by high levels of crime, violence, and other signs of social disorganization, where it may be difficult to place confidence in the police or the law. Second, we would also expect to find significant associations between adherence to the code and a number of social processes that are believed to encourage or facilitate the learning of code-related beliefs among young people, including lack of adequate parental supervision, abusive parenting practices, personal experiences with violence and victimization, exposure to violent or aggressive peers, and the belief that the personal attainment of status or respect through legitimate pursuits is not a realistic option. Third, we should find that adherence to the code promotes subsequent violent behavior and, moreover, mediates the effects of other variables on violent conduct. Previous Research on Youth Violence Although not designed to provide a direct or complete evaluation of Anderson’s subcultural account, a number of previous studies have examined processes relevant to this account, including the development of code-related beliefs and the contribution of such beliefs to youth violence. Thus, the results of these studies have implications for the validity and/or generality of Anderson’s findings and, for this reason, we highlight key findings below. Analyzing cross-sectional data on African American children in small towns and rural areas, Stewart, Simons, and Conger (2002) find a positive relationship between coderelated beliefs and aggressive behavior, controlling for other relevant variables. Children in the study who believed it was important and/or necessary to use violence to prevent intimidation and loss of respect tended to exhibit relatively high levels of aggressive behavior, such as physical assault, threatening and bullying behavior, and use of a weapon.

308

Youth Violence and Juvenile Justice

Analyzing data from a national survey of youth, Agnew (1994) observed that only a very small percentage of young people express general approval of violent behavior (e.g., hitting others for no reason). However, a sizable percentage of young people believe that violence is appropriate or necessary in response to insults or various other provocations. Moreover, Agnew found that such beliefs or justifications—often referred to as “definitions favorable to violence”—exhibit significant effects on subsequent violent behavior, controlling for prior violent behavior and other relevant variables. This finding suggested that code-related beliefs or definitions play a causal role in the development of future violence and are not simply rationalizations that develop “after the fact.” A study by Markowitz and Felson (1998) employed alternative, but relatively detailed, measures of code-related beliefs and attitudes, including disputatiousness (sensitivity to provocation), retribution (importance placed on retaliation), and courage (importance placed on the display of courage during a fight or argument). Although the study is based on adults and is limited to cross-sectional data, the findings are suggestive and indicate a significant relationship between code-related beliefs and self-reported violent behavior. In particular, the authors find that individuals who emphasize retribution and courage tend to be especially sensitive to slights and therefore earn relatively high scores on the measure of disputatiousness. High scores on disputatiousness, in turn, are associated with relatively high levels of violent behavior. Moreover, Markowitz and Felson (1998) found that the code-related beliefs and attitudes measured in their study tended to mediate much of the relationship between violence and certain sociodemographic characteristics, including age, gender, and SES. Code-related beliefs and attitudes (especially courage and/or retribution) tend to be more prevalent among young adults, males, and those of low SES, and this fact may help to explain why low-status male youths engage in relatively high levels of violent behavior. After controlling for SES, however, the authors did not observe evidence of a racial effect. Non-White respondents in the study did not express relatively high levels of support for code-related beliefs, nor did they report relatively high levels of violent behavior—a finding that appears to run counter to expectations but that is consistent with many other self-report studies (see Bridges & Weis, 1989). As noted by Markowitz and Felson (1998), their study does not help to explain why certain sociodemographic factors, such as SES, are associated with code-related beliefs and attitudes. Karen Heimer has presented what may be the most complete model of youth violence in the quantitative research literature and her empirical findings help to shed light on this issue. Heimer (1997) outlined a developmental model of youth violence that integrates structural (SES) and cultural (parental and peer influence) factors. In this model, low SES is associated with certain parenting factors that promote the development of violent behavior among young people, such as power-assertive (punitive) parenting styles and lack of adequate parental supervision. These parenting factors are believed to foster the learning of rationalizations and justifications for violent behavior either directly—by teaching children that force and coercion can be used to solve problems—or indirectly by increasing the likelihood that children will associate with aggressive peers.4 Heimer (1997) reported empirical findings that are consistent with her overall model, based on analyses of data from a national sample of male adolescents. In general, the results of her analyses indicate that definitions favorable to violence (in this case, the belief that violence is acceptable or necessary in response to various slights) mediated the effects of SES, parental influence, and peer influence on subsequent violence. In comparison to their higher status counterparts, youths of lower SES are more likely to be exposed to power-

Brezina et al. / THE CODE OF THE STREET

309

assertive parenting; they are subject to lower levels of parental supervision; and they tend to have more extensive histories of aggressive behavior. Consequently, low-status youths are more likely to associate with aggressive peers, who tend to reinforce violent definitions (or code-related beliefs). The learning of violent definitions, in turn, represents the best predictor of future violent behavior after taking into account prior violence and other variables (also see Heimer & De Coster, 1999). Interestingly, Black racial status did not exhibit a total direct or indirect effect on violent behavior after adjusting for SES, neighborhood crime, urban residence, and other factors. Urban residence and the presence of substantial neighborhood crime were both positively associated with prior aggressive behavior, which in turn affected adherence to code-related beliefs and thus future violence. A relatively recent study by Baron, Kennedy, and Forde (2001) may also help to shed light on the origins of code-related beliefs. The authors of this study interviewed a local sample of male street youths and collected measures of power-assertive parenting, victimization (having been robbed and/or assaulted), anger, peer aggression, prior violence, and code-related beliefs (in this case, beliefs that justify violent retaliation in response to various slights). They also presented various hypothetical scenarios, which depicted realistic situations, to the youths in their study and asked each respondent to imagine himself as the object of an insult or other provocation. Following the presentation of each scenario, the respondent was asked to state the likelihood that he would respond to such situations with physical aggression. Five sets of findings from this study are noteworthy. First, consistent with prior research, the results indicated that power-assertive parenting, peer aggression, and prior violence are positively associated with adherence to code-related beliefs. Second, the results indicated that anger is also positively associated with code-related beliefs. Third, the authors of this study observed that victimization is negatively associated with code-related beliefs. This particular finding appears to run counter to expectations derived from Anderson’s work, as Anderson (1999) suggested that personal experiences with violence and victimization tend to increase youths’ appreciation for the “real consequences” of winning or losing in a fight and, hence, the importance they attach to the code (p. 70). It is possible, however, that a reciprocal relationship exists between victimization and code-related beliefs: Although past victimization may encourage youths to adhere to the code, enactment of the code may then help to send a message that one is “no longer to be messed with” and thereby reduce the chances of future victimization. Indeed, Baron et al. (2001) interpreted the negative association between victimization and coderelated beliefs to mean that adherence to the street code may serve a protective function and allow young people to avoid victimization. Yet, as the authors recognized, the crosssectional nature of their study precludes firm conclusions regarding the causal direction of observed relationships. Fourth, the authors found that code-related beliefs mediate the effects of powerassertive parenting, victimization, anger, peer aggression, and prior violence on scenariobased intentions to engage in violence. Fifth, the findings of this study suggested that young males would be most likely to use force during disputes that take place on the street (e.g., while passing by a convenience store on an afternoon walk) as opposed to disputes that occur in more structured settings such as schoolyards or sporting events. In sum, the results of past research on youth violence lend support to many of the hypotheses we derived from Anderson’s subcultural account. In particular, quantitative studies have highlighted the role of “definitions favorable to violence”—definitions that appear to resemble the type of beliefs that are said to underlie the street code. For example,

310

Youth Violence and Juvenile Justice

violent definitions are typically indexed by the belief that insults and other provocations justify—if not require—physical retaliation (Agnew, 1994; Baron et al., 2001; Heimer, 1997; Heimer & De Coster, 1999; Markowitz & Felson, 1998; Stewart et al., 2002). Prior studies have also established links between violent definitions and many of the socioenvironmental factors highlighted by Anderson, including parenting practices, peer relations, neighborhood context, and various demographic factors such as SES and urban residence. Moreover, the findings of several quantitative studies indicated that code-related beliefs tend to mediate the effects of socioenvironmental factors on future violent behavior. Although these studies were not specifically designed to test Anderson’s subculture of violence thesis, the findings increase confidence in the validity and generality of various ethnographic findings that Anderson has served to highlight. One possible exception involves the lack of a clear racial effect in previous quantitative studies of youth violence. Anderson appeared to suggest that the stigma of race and the resulting alienation from the wider society play a unique role in fostering acceptance of the street code. To the extent that it symbolizes opposition to White society and culture, socially isolated Black youths may “pride themselves on knowing and being able to enact the code of the street” (Anderson, 1998, p. 94). However, after taking into account SES, neighborhood characteristics, and other variables, existing quantitative studies typically fail to observe an effect of racial status on code-related beliefs. One possible reason for this unexpected finding is that previous studies have simply failed to include a sufficient number of inner-city Black youths, including the type of alienated youths that Anderson described. Yet this may not be the whole story. Prior research on economically disadvantaged communities suggested an alternative explanation, including the possibility that neighborhood context is more critical to the development and acceptance of code-related beliefs than race alone (also see Baumer, Horney, Felson, & Lauritsen, 2003). For instance, a large study of Chicago residents found that, once neighborhood economic disadvantage is taken into account, racial differences in attitudes toward violence and the law (e.g., the attitude that “fighting between friends or families is nobody else’s business”) tend to disappear (Sampson & Bartusch, 1998). We examine the role of race in our own analyses, reported below, and eventually return to a discussion of this issue. Drawing on Anderson’s Insights to Extend Youth Violence Research Although the results of previous quantitative studies are largely consistent with Anderson’s subculture of violence thesis (with the possible exception of race effects at the person level), several of the hypotheses we derived from Anderson’s account require further examination or have yet to be evaluated. These include the hypothesized roles of victimization and perceived opportunity in the development of code-related beliefs. We believe that an examination of these hypotheses is important because they appear to reflect some of the more distinctive features of Anderson’s subcultural account. As described earlier, Anderson argued that adherence to the code partly represents an adaptation to the social environment faced by some young people. In particular, youths who run the risk of victimization come to learn the importance and necessity of the code for the purpose of self-protection, even if they are not personally inclined toward aggression. This is said to be especially true of youths who have experienced the real consequences of

Brezina et al. / THE CODE OF THE STREET

311

violence, who understand the threat of being “rolled on” if identified as an easy target, and who have found that they cannot rely on the police or other adults for protection. Although it is possible to imagine a complex, reciprocal relationship between victimization and coderelated beliefs (see above), these arguments suggest that, at the very least, prior victimization is likely to play a significant role in the formation or development of coderelated beliefs. An evaluation of this hypothesis requires longitudinal data to establish proper temporal ordering of key variables. Adherence to the code is also said to be relatively strong among youths who see little hope of attaining status and respect through legitimate means. According to Anderson, the code is partly a function of the restricted opportunities available to disadvantaged youths for obtaining respect in mainstream culture—especially poor Black males. This fact helps to explain why communities racked by poverty, unemployment, and racism tend to possess a relatively large share of young men who are invested in the code. Among such men, the aggressive campaign for status and respect on the streets is perceived to be “the only game in town” (Anderson, 1994, p. 94). At its core, then, the code is described as a cultural adaptation to the hopelessness and alienation generated by blocked opportunities. These arguments echo the statements of early strain theorists, who interpreted violent subcultures as functional adaptations to environmental pressures or demands. For example, Albert Cohen (1955) interpreted the formation of juvenile gangs as a problem-solving response to the status frustration experienced by lower class males. As Cohen observed, boys from lower class backgrounds often lack the skills necessary to achieve success or status as it is defined by middle-class society. In comparison to their middle-class counterparts, lower class boys are more likely to experience failure in school, for example. Consequently, some lower class males engage in collective rebellion against the middleclass “measuring rod” by defining their own standards for success—standards they can more readily achieve—such as toughness and fighting ability. Cloward and Ohlin (1960) also interpreted the violent subculture as an adaptation to the frustrations experienced by disadvantaged young people. They emphasized that violent adaptations are most likely to develop in certain types of neighborhoods, especially disorganized neighborhoods that have been isolated from both conventional and illegitimate opportunity structures. In response to such isolation, frustrated and alienated youths “seize upon the manipulation of violence” as an alternative route to status (Cloward & Ohlin, 1960, p. 175). By allowing young people to demonstrate “guts,” violence provides limited but readily available opportunities for cultivating respect and self-worth (also see Canada, 1995; Majors & Billson, 1992). Surprisingly, these strain arguments have received little attention in quantitative studies of youth violence, perhaps because such studies have been dominated in recent years by social control and social learning interpretations (e.g., Felson, Liska, South, & McNulty, 1994; Heimer, 1997; Heimer & De Coster, 1999). Although certain previous studies have observed a significant relationship between perceived opportunity and delinquency (for reviews, see Burton & Cullen, 1992; Vowell & May, 2000), we know little about the role of perceived opportunity in the development of code-related beliefs in particular and the extent to which code-related beliefs mediate the relationship between perceived opportunity and violent behavior. Nor do we have an adequate understanding of the influence of victimization, the threat of victimization, and other factors related to environmental pressures or demands. Thus, in our own set of analyses, we examine the roles of perceived opportunity and victimization in conjunction with social control and social learning variables.

312

Youth Violence and Juvenile Justice

Time 1 Variables: Perceived Lack of Opportunity Victimization Power-Assertive Discipline Parental Supervision

SES

Time 2 Peer Aggression

Race Urban Time 2 Code–Related Beliefs

Neighborhood Crime Family Structure Age Time 1 Violent Behavior

Figure 1.

Time 2 Violent Behavior

Time 3 Violent Behavior

Diagram of the Proposed Causal Model Linking Social Position, Perceived Opportunity, Victimization, and Other Variables to CodeRelated Beliefs and Violent Behavior

NOTE: SES = socioeconomic status. Dashed lines represent unanalyzed correlations.

Figure 1 depicts the causal model we seek to test. This model is based on Heimer’s (1997) earlier research on the development of youth violence. However, to build on this research, we have added perceived opportunity and victimization to the causal process. In the analyses described below, we give special attention to the roles of perceived opportunity and victimization because, as stated above, these factors represent distinct elements in Anderson’s subcultural account and have not been a major focus of prior research on coderelated beliefs. Data and Methods Data To conduct our examination, we draw data from the first three waves of the National Youth Survey (NYS), a panel survey of self-reported delinquent behavior conducted by the Behavioral Research Institute (Elliott, Huizinga, & Ageton, 1985). The NYS is based on a national probability sample of youths aged 11 to 17 years and is available to researchers through the Inter-university Consortium for Political and Social Research (ICPSR), University of Michigan. The study sampled households and, in each household, selected all youths who were capable of being interviewed.

Brezina et al. / THE CODE OF THE STREET

313

The first wave of data collection (Time 1) was conducted in 1977. A total of 1,725 respondents agreed to participate, representing a response rate of 73%. In addition, one parent per youth was interviewed at Time 1. Subsequent interviews were conducted with the youths on an annual basis. The second wave of data (Time 2) contains responses from 1,655 youths (96% of the Time 1 respondents). The third wave of data (Time 3) contains responses from 1,626 youths (94% of the Time 1 respondents). Elliott et al. (1985, pp. 9293) found that the representativeness of the survey, with respect to key variables, was not seriously affected by the loss of respondents over the three waves. One may question the use of national survey data to test hypotheses that were derived from observations of disadvantaged youth. However, the NYS contains a substantial number of youths from lower class backgrounds, including many hardcore delinquents (Dunford & Elliott, 1984; Elliott & Ageton, 1980). Moreover, as stated above, Anderson (1999) hoped that his research “may offer insight into the problem of youth violence more generally” (p. 9). An examination of national data will help us to gauge the generality of his findings. One might also question the use of adolescent data to examine developmental processes that are believed to begin in early childhood. As Anderson (1999) argued, the basic elements of the code are typically learned in childhood: “By the time they are teenagers, most youths have either internalized the code of the streets or at least learned the need to comport themselves in accordance with its rules” (p. 72). Anderson also argued, however, that the code becomes particularly meaningful during adolescence and, for this reason, the use of adolescent data may be especially appropriate. During the course of adolescence, the importance of the code is likely to be impressed on youths as never before. Young people’s understanding of the code is likely to become refined and their beliefs about violence are likely to crystallize as they attempt to establish their identities, “as they learn how the code defines their relationship to their peers” (Anderson, 1998, p. 83), and “as they come to appreciate the real consequences” of winning or losing in a fight (Anderson, 1999, p. 70). Following Heimer (1997), our main analysis is based on the male subsample of the NYS, which included a total of 918 boys during the first wave of data collection.5 By restricting our examination to male respondents, we facilitate direct comparisons between our findings and those of prior research. Also, by studying young males, we focus on the population most likely to engage in violent behavior. Analyses To test the causal model in Figure 1, we estimated a structural equation model using the maximum-likelihood estimator available in LISREL 8 (Joreskog & Sorbom, 1993). We also added a measurement model to the structural equation (substantive) model to correct for error in the measurement of the theoretical (latent) constructs depicted in Figure 1. This methodological approach offers several advantages. First, this approach allows us to avoid bias in the estimation of substantive parameters that would otherwise result from unreliable survey items or indicators. Second, this approach allows us to estimate the substantive parameters simultaneously in the context of a full-information model—a model that corresponds to the causal ordering of both the theoretical arguments and the longitudinal data contained in the NYS.

314

Youth Violence and Juvenile Justice Measures

In addition to perceived opportunity and victimization, we include all the variables that were found by Heimer (1997) to have a significant direct or indirect effect on male violence. SES is treated as an exogenous, latent construct, and is measured by three indicators: annual family income, principle wage earner’s occupation, and mean level of parents’ education. Neighborhood crime is also treated as a latent construct and is measured by three items that were asked of the respondents’ parents at Time 1. These items index the extent to which muggings, theft, and vandalism are perceived to be a “big problem” in the neighborhood. The age of each youth respondent is recorded in years by a single item. Other exogenous variables in the model include the following single-item dichotomous variables: race (1 = Black), urban residence (1 = residence in a Standard Metropolitan Statistical Area), and family structure (1 = nonintact home). Perceived opportunity, victimization, and parenting factors are treated as endogenous variables in the model and are all measured at Time 1. Respondents who score high on a perceived “lack of opportunity” measure estimate a low probability of obtaining the kind of job they would like to have after finishing school. Responses to this item range from 1 (chances are good) to 3 (chances are poor). Respondents who score high on a three-item measure of victimization report that, during the past year, they have been “beaten up,” have had things taken directly from them by force, and have had their personal items intentionally damaged or destroyed by others. Responses (recoded) to each item range from 0 (never) to 3 (three or more times). A two-item measure of power-assertive discipline is based on parental reports of the disciplinary practices that are used when the child respondent “does something wrong.” High scores on each item correspond to power-assertive parenting strategies, such as the use of threats and physical punishment.6 A measure of parental supervision is also based on parental reports and is composed of three items that ask about parents’ level of familiarity with their children’s friends, familiarity with the parents of these friends, and the number of times they have asked their children’s friends to join in family activities. Peer aggression is measured at Time 2 by a single item that indexes the proportion of the respondent’s close friends who, according to the respondent, have hit others during the previous year. Responses to this item range from 1 (none of them) to 5 (all of them). Coderelated beliefs are also measured at Time 2 and are based on three items that index the extent to which the respondent believes physical retaliation is an appropriate response. In particular, respondents who score high on this measure agree that “it is sometimes necessary to get into a fight to uphold your honor or to put someone in his or her place,” it is “all right to beat up another person if he or she started the fight,” and it is “all right to beat up another person if he or she called you a dirty name.” Taken together, these items tap the extent to which individuals believe physical retaliation is justified and/or necessary in response to various provocations and the extent to which they believe their honor and/or status is tied to the display of aggression—both important aspects of the code (see Anderson, 1999, p. 76). Although this measure of code-related beliefs is similar to those used in previous research, it should be noted that the items that comprise this measure do not specify the context in which it would be appropriate to use force. According to Anderson (1999), the code is mainly applicable to disputes that erupt on the street or in other public locations. Unless they are completely immersed in street life, young people are less likely to enact the code in other, more civil, settings. The failure to specify the context of the dispute is a

Brezina et al. / THE CODE OF THE STREET

315

potential limitation of our measure of code-related beliefs. However, the most likely consequence of this limitation is that, in the analysis, it may be somewhat more difficult to observe a relationship between code-related beliefs and violent behavior. Violent behavior is measured at three time points with a 10-item scale that includes all of the offenses in Elliott and Huizinga’s (1983, p. 157) Crimes Against Persons Scale. High scorers on this scale state that, during the previous year, they frequently hit (or attempted to hit) students, parents, and teachers; strong-armed students, teachers, and others; participated in gang fights; and “attacked someone with the idea of seriously hurting or killing him or her.” Responses to each item in the scale range from 1 (never) to 9 (two or more times a day). The inclusion of Time 1 and Time 2 measures of violence in the model (along with Time 3 violence) increases confidence in proper causal ordering and should help to control for the influence of various individual traits that were not measured explicitly in the survey. Following previous research, our measurement model adjusts for unreliability (or measurement error) in both multiple-indicator and single-indicator constructs. Measurement error in single-indicator constructs cannot be estimated empirically. It is possible, however, to set the measurement error in single-indicator constructs at values deemed to be reasonable on the basis of past research. Failure to do so may bias the results in favor of single-indicator constructs, because it is unreasonable to assume that these constructs are measured with perfect reliability. Following Heimer (1997), we fixed the reliability of all single-item exogenous variables in the model (race, urban, family structure, and age) to .81. The reliabilities of all single-item endogenous variables (perceived lack of opportunity, peer aggression, and violent behavior at each time point) are fixed at .75, because it is reasonable to assume a greater level of error in the measurement of these social psychological variables in comparison to demographic factors. In a series of sensitivity analyses, the reliabilities of these single-item variables were set at different values, ranging from .70 to 1.00. The substantive parameter estimates did not vary greatly as a result, indicating that our findings are stable and robust. It is also important to adjust for certain correlations between measurement errors, especially the autoregressive errors associated with the repeated measurement of violent behavior over time. Failure to correct for these error correlations is likely to lead to the overestimation of stability effects (i.e., the effect of past violence on future violence) and the underestimation of other effects in the model (Kessler & Greenberg, 1981). Following Heimer (1997), we fixed a total of five error correlations to .15. These include the two firstorder autoregressive error correlations associated with violent behavior, along with three additional error correlations involving race, family structure, and age.7 The results of sensitivity analyses show that the substantive parameter estimates do not vary greatly when these correlations are varied between 0 and .25. Results We estimated the measurement and substantive models simultaneously as a single system. Following the recommendation of Browne and Cudeck (1993), we use the root mean square error of approximation (RMSEA) to assess the fit of the overall model to the data. This statistic takes into account the error of approximation in the population and addresses the question, How well would the model, with unknown, but optimally chosen, parameter values fit the population covariance matrix if it were available (Browne &

316

Youth Violence and Juvenile Justice TABLE 1 Parameter Estimates for the Measurement Model

Latent Variable 1. SES 2. Race 3. Urban residence 4. Neighborhood crime 5. Family structure 6. Age of respondent 7. Perceived lack of opportunity 8. Victimization

9. Power-assertive discipline 10. Parental supervision 11. Peer aggression 12. Code-related beliefs

13. Violent behavior

Observed Indicator Annual family income Principle wage earner’s occupation Mean level of parents’ education Dichotomous variable (1 = Black) Respondent resides in SMSA Vandalism is a big problem in neighborhood Burglaries are a big problem in neighborhood Muggings are a big problem in neighborhood Respondent resides in nonintact home Youth’s age in years Perceived low probability of getting desirable job after finishing school Number of times “beaten up” during past year Number of times had things taken by force Number of times had personal items damaged/destroyed Mother’s Father’s Parents invite friends to home Parents know friends’ parents Parents know youth’s friends Proportion of close friends who have hit others during past year It’s sometimes necessary to get into a fight to uphold your honor or put someone in his or her place It’s all right to beat up others who call you names It’s all right to beat up others if they started the fight 10-item violent delinquency scale (constructed at 3 time points)

Metric Slope

Validity Coefficient

1.00 0.86 0.68 1.00a a 1.00 a 1.00 1.05 0.32 1.00a a 1.00

a

.65 .73 .78 .90a a .90 .77 .82 .48 .90a a .90

1.00a a 1.00 0.87

.87a .65 .57

0.68 1.00a 1.11 1.00a 1.07 0.86

.47 .67 .62 .67 .73 .79

1.00a

.87a

1.00a 1.45 1.49

.45 .64 .71

1.00a

.87

a

NOTE: SES = socioeconomic status; SMSA = Standard Metropolitan Statistical Area. a. Fixed parameter.

Cudeck, 1993, pp. 137-138)? An RMSEA of .05 or lower is said to be indicative of a close fit. Table 1 presents the results of the measurement portion of the model. The results show that our measures serve as reasonably valid and reliable indicators of the latent constructs. At the same time, however, enough response error exists to warrant the inclusion of a measurement model to avoid bias in the estimation of substantive parameters. Table 2 presents the results of the substantive portion of the model. These results are discussed below. The chi-square statistic for the overall model is 458.41, with 212 degrees of freedom. The RMSEA is .039, suggesting that the model fits the data reasonably well. Other goodness-of-fit statistics also indicate an acceptable model fit8 (AGFI = .93, CFI = .93). In general, the results in Table 2 show that we have successfully replicated the findings of Heimer’s (1997) research. Like Heimer, we find that the development of violent behavior among young males is influenced by their social and geographic location, as

317

–.025** (.011) –.14 –.042 (.047) –.05 .062* (.033) .10 .033 (.028) .06 .067** (.033) .10 .007 (.008) .05 .— .— .— .—

Race (1 = Black)

Urban residence (1 = urban)

Neighborhood crime

Family structure (1 = nonintact home)

Age

Time 1 Violence

Perceived lack of opportunity

Time 1 Violence

SES

Independent Variables

.—

.073 (.081) .04

.001 (.013) .00

–.069 (.055) –.06

.019 (.047) .02

.110** (.054) .11

–.083 (.077) –.06

–.029 (.019) –.10

Lack of Opportunity

.—

1.370** (.140) .55

–.041** (.020) –.11

.100 (.087) .06

–.064 (.075) –.05

.066 (.086) .04

–.062 (.120) –.03

.043 (.029) .10

Victimization

.—

.110 (.081) .07

–.026** (.013) –.11

–.049 (.055) –.05

.026 (.047) .03

.044 (.054) .05

.210** (.078) .17

–.056** (.019) –.21

Power-Assertive Discipline

.—

–.054 (.150) –.02

–.036 (.023) –.07

–.210** (.100) –.10

.024 (.088) .01

–.190* (.100) –.09

–.160 (.140) –.06

.079** (.035) .14

Parental Supervision

Dependent Variables

.061** (.029) .09

.160** (.070) .15

.013* (.008) .08

.012 (.034) .02

.069** (.029) .11

.009 (.033) .01

–.059 (.048) –.07

–.021* (.012) –.11

Time 2 Violence

.160* (.091) .08

.360* (.200) .12

–.039 (.024) –.08

.100 (.100) .05

.009 (.088) .00

–.027 (.100) –.01

–.170 (.150) –.07

.010 (.036) .02

Peer Aggression

TABLE 2 Unstandardized and Standardized Parameter Estimates for the Structural Model

–.046 (.048) –.05

.—

.019 (.013) .08

–.090 (.055) –.09

–.029 (.046) –.03

.005 (.053) .01

–.024 (.076) –.02

–.025 (.019) –.08

Code-Related Beliefs

(continued)

.037 (.025) .07

.—

–.002 (.006) –.01

.025 (.028) .04

.034 (.024) .06

–.025 (.027) –.04

–.026 (.039) –.03

–.008 (.010) –.05

Time 3 Violence

318 .— .— .— .— .— .— .— .— .— .— .— .— .— .— .— .048

Power-assertive discipline

Parental supervision

Time 2 violence

Peer aggression

Code-related beliefs

R2 .019

.— .— .—

.— .— .—

.— .— .—

.— .— .—

.— .— .—

.— .— .—

Lack of Opportunity

.320

.— .— .—

.— .— .—

.— .— .—

.— .— .—

.— .— .—

.— .— .—

Victimization

.130

.— .— .—

.— .— .—

.— .— .—

.— .— .—

.— .— .—

.— .— .—

Power-Assertive Discipline

.064

.— .— .—

.— .— .—

.— .— .—

.— .— .—

.— .— .—

.— .— .—

Parental Supervision

.290

.— .— .—

.— .— .—

.— .— .—

–.027* (.015) –.08

.050 (.039) .07

.160** (.032) .37

Time 2 Violence

.120

— — —

— — —

— — —

–.120** (.048) –.12

.240* (.130) .11

.220** (.089) .17

Peer Aggression

.330

.— .— .—

.120** (.029) .23

.670** (.110) .46

.034 (.025) .07

.062 (.065) .06

.026 (.042) .04

Code-Related Beliefs

.300

.087** (.037) .15

.010 (.014) .03

.210** (.055) .24

–.029** (.013) –.10

.008 (.033) .01

.086** (.023) .23

Time 3 Violence

NOTE: N = 752. SES = socioeconomic status. Unstandardized effects shown, with standard errors in parentheses. Standardized coefficients are presented in italics below the standard errors. *p < .05 (one-tailed test). **p < .05 (two-tailed test).

.— .— .—

Time 1 Violence

Victimization

Independent Variables

Dependent Variables

TABLE 2 (continued)

Brezina et al. / THE CODE OF THE STREET

319

indicated by SES, urban residence, residence in a crime-prone neighborhood, and family structure. These factors exhibit direct and/or indirect effects on violence. For instance, SES exhibits a significant direct effect on Time 1 violent behavior, along with urban residence and family structure (p < .05). However, SES also exhibits indirect effects on future violent behavior through its association with other consequential variables, such as power-assertive discipline and parental supervision. In particular, lower status youths are more likely than their upper status counterparts to be exposed to powerassertive disciplinary practices. They are also more likely to be free of close parental supervision. Power-assertive discipline and parental supervision, in turn, affect subsequent violence either directly or indirectly through a significant pathway that involves association with aggressive peers, a strong investment in code-related beliefs, and, ultimately, violent behavior at Time 3. Like Heimer (1997), we also find that adherence to code-related beliefs (or violent definitions) tends to mediate—in whole or in part—the effects of other variables on Time 3 violence, such as association with aggressive peers and prior history of violent behavior. Consistent with previous research (e.g., Heimer, 1997; Markowitz & Felson, 1998), the total direct and indirect effects of Black racial status on code-related beliefs, and on future violence, do not achieve statistical significance. Nor are these effects in the expected direction. Although Black racial status exhibits a positive effect on power-assertive parenting, it appears that this effect is counterbalanced by weak negative associations that exist between Black racial status and code-related beliefs and between Black racial status and Time 3 violence (also see Heimer, 1997). As stated earlier, quantitative studies of youth violence tend to find that the effects of race diminish or disappear once SES and neighborhood context are taken into account. Our ability to successfully replicate the results of past research is reassuring. Our attention now turns to those findings that extend beyond the results of previous quantitative research. In particular, we now discuss the effects of perceived opportunity and victimization—factors that receive special attention in Anderson’s account of youth violence. Figure 2 depicts selected significant pathways, with a special focus on those pathways that are especially relevant to Anderson’s account and that link perceived opportunity and victimization to subsequent violence. As seen in Figure 2, both perceived opportunity and victimization are consequential factors in the development of violent behavior. These factors exhibit significant indirect effects on Time 3 violence through two possible pathways. First, perceived lack of future opportunity and victimization are both associated with an increased likelihood of association with aggressive peers (i.e., “close friends” who have a history of physical aggression). Association with aggressive peers, in turn, contributes to the development of code-related beliefs and, ultimately, an increase in violent behavior at Time 3. As suggested by classic strain theorists (Cloward & Ohlin, 1960; Cohen, 1955), youths who face problems involving status, respect, and personal protection may be attracted to rebellious peer groups. Such groups may provide protection and also highlight alternative standards by which to measure one’s dignity and self-worth (also see Jankowski, 1995). As also suggested by classic strain theorists, it appears that the development of subcultural (code-related) beliefs is partly dependent on exposure to aggressive peers—perhaps because strong peer support plays an important role in the maintenance of such beliefs.

320

Youth Violence and Juvenile Justice

Perceived Lack of Opportunity

.08 .17

.11

Time 2 Peer Aggression

Victimization

.23

Urban

.55 SES

.11

.23 .12 .37

.09

-.14

.15 .46

Time 1 Violent Behavior

Figure 2.

Time 2 Code–Related Beliefs

.15

Time 2 Violent Behavior

.24

Time 3 Violent Behavior

Selected Significant Pathways Linking Anderson-Derived Variables to Code-Related Beliefs and Violent Behavior

NOTE: SES = socioeconomic status. Standardized effects shown. To reduce clutter, the effects of parenting variables are not shown (see Table 2). Dashed lines represent unanalyzed correlations.

Second, both perceived opportunity and victimization exhibit significant direct effects on Time 2 violence, controlling for Time 1 violent behavior and other variables. Time 2 violence, in turn, directly affects violence at Time 3. Time 2 violent behavior is also associated with adherence to code-related beliefs—beliefs that make violent behavior at Time 3 a more likely possibility. The observed effect of past violence on code-related beliefs is consistent with prior research (Heimer, 1997). Interpretations of this effect vary, however. One possibility is that past involvement in violence encourages the development of rationalizations and justifications for such behavior. As a result, youths with violent histories may be receptive to beliefs that condone violent retaliation in response to various provocations. An alternative possibility—suggested by Anderson (1999, p. 70)—is that experience with violence leads to greater appreciation of the “real consequences” of physical aggression, including the importance of winning or losing in a fight, the necessity of appearing “tough,” and the significant personal costs associated with the failure to observe the street code. In addition to the effects described above, past victimization exhibits a significant direct effect on violence at Time 3. The unmediated effects of victimization on violent behavior suggest that much violence may be directly motivated by a desire for selfprotection or by anger—rather than by investment in the code. As Baron and Hartnagel (1998) contended, anger may lead young people to “seek out their victimizers to settle the score regardless of learning experiences and subcultural expectations” (p. 183). However, as stated above, once such violence occurs, it tends to foster acceptance of the code. Taking into account both direct and indirect effects, the total standardized effect of victimization on

Brezina et al. / THE CODE OF THE STREET

321

Time 3 violence is .31 (p < .05). The total standardized effect of perceived opportunity on Time 3 violence is .08 (p < .05). Figure 2 also shows how various demographic factors are related to perceptions of future opportunity and to the experience of victimization. For example, it can be seen that urban residence has a significant effect on perceived lack of opportunity. Unexpectedly, however, race and SES are not significantly associated with perceptions of future opportunity. As suggested by previous research, it is possible that, although some disadvantaged young people may believe that their future opportunities are limited, many other disadvantaged youths maintain high—perhaps inflated—aspirations despite their social backgrounds (see Agnew & Jones, 1988). Age is the only demographic variable in our model that exhibits a significant direct effect on victimization. Older respondents in our sample tend to report relatively low levels of victimization (path not shown in Figure 2). The effects of other demographic variables on victimization, such as SES and urban residence, are mediated by past involvement in violent behavior (also see Esbensen & Huizinga, 1991). Consistent with previous research on the relationship between delinquent involvement and victimization (e.g., Esbensen & Huizinga, 1991; Lauritsen, Sampson, & Laub, 1991), we find that levels of victimization are relatively high among youths who have a history of violent behavior. As suggested by Esbensen and Huizinga (1991), youths who are engaged in violent pursuits may have disproportionate contact with other offenders and thus are more likely to be assaulted themselves or to be the target of other types of victimization. Yet the experience of victimization only appears to increase the chances that such youth will maintain high levels of violent behavior in the future. As described above, victims/offenders may be motivated to seek retaliation or revenge and tend to gravitate to aggressive peer groups, where they appear to learn code-related beliefs. Summary and Conclusion A number of recent studies in the subculture of violence tradition have highlighted the role of beliefs in the etiology of youth violence. Individuals who believe that violent retaliation is an acceptable, justified, and/or necessary response to provocation tend to exhibit relatively high levels of violent behavior (e.g., Agnew, 1994; Baron et al., 2001; Felson et al., 1994; Heimer, 1997; Heimer & De Coster, 1999; Markowitz & Felson, 1998). In other words, violence is more likely among individuals who endorse beliefs that are consistent with the “code of the street” described by Anderson (1999). Data also indicate that such beliefs mediate much of the relationship between various demographic variables—such as SES and underclass status—and violent behavior (Heimer, 1997; Heimer & De Coster, 1999; Markowitz & Felson, 1998). Less is known about the origins of code-related beliefs. As highlighted by previous research (Baron et al., 2001; Heimer, 1997; Heimer & De Coster, 1999), it appears that such beliefs are transmitted in part through social learning processes that involve exposure to aggressive role models, including parents who rely on physical aggression to control their children and peers who display violent behavior patterns. It is also possible, however, that investment in code-related beliefs represents an adaptive strategy, especially in contexts where individuals face a substantial threat of victimization and especially “if those who are fearless and punitive are better able to avoid victimization” (Markowitz & Felson, 1998, p. 134). Investment in code-related beliefs may also be adaptive if a fearless and punitive

322

Youth Violence and Juvenile Justice

response enhances status and respect on the streets, particularly in areas where alternative sources of status and respect are not readily available. Indeed, these latter possibilities are emphasized in Elijah Anderson’s (1999) account of youth violence. Data from the male subsample of the NYS are largely consistent with the expectations we derived from Anderson’s ethnographic study. Consistent with previous research, we find that a number of social learning factors are associated directly or indirectly with adherence to code-related beliefs, including power-assertive parenting styles, low levels of parental supervision, and involvement with aggressive peers. Code-related beliefs, in turn, predict future violent behavior. As suggested by Anderson’s account (1999), we also find that perceived lack of opportunity and victimization are consequential in the development of code-related beliefs and, ultimately, subsequent violence. For instance, we observe that youths who perceive a lack of future opportunity and those who report relatively high levels of prior victimization are more likely than other youths to become associated with aggressive peers. Youths who become involved with aggressive peers, in turn, are more likely to view physical retaliation as an acceptable and necessary response to real or perceived slights. Consequently, they are more likely to carry concealed weapons, to engage in assault, to become involved in gang fights, to hit (or threaten to hit) other students at school, and to engage in other acts of violence. Consistent with arguments advanced by classic strain theorists (Cloward & Ohlin, 1960; Cohen, 1955), these findings suggest that problems involving opportunity and personal protection may increase attraction to delinquent peer groups and, hence, facilitate the learning of violent behavior. And consistent with some of the specific arguments advanced by Anderson (1999), these findings suggest that problems involving opportunity and personal protection contribute to the acceptance of the street code among young people. In sum, our own findings provide additional empirical support for key aspects of Anderson’s account, including the hypothesized origins of violent beliefs among young people. Our findings—along with the findings of other national studies—also suggest that his observations, which were drawn from two Philadelphia neighborhoods, offer insight into the problem of youth violence more generally. Nevertheless, our study was limited in scope, and several issues must be left for future research to address. First, available data sets do not permit a complete or comprehensive assessment of Anderson’s subculture of violence thesis. Ideally, one would examine the full range of dynamics that Anderson has served to highlight, with precise measures of codeswitching, the extent to which individuals trust the abilities of the police and other authorities to protect citizens, and perhaps details on the exact nature and context of violent encounters. A more complete examination may well provide greater support for Anderson’s thesis. Second, the role of race in the development of code-related beliefs and code-related violence remains unclear. Although Anderson (1999) highlighted the role of racial dynamics, we find that SES, neighborhood context, the experience of victimization, and peer influence are more crucial factors in the development of code-related beliefs and youth violence. Moreover, our findings are consistent with other individual-level studies in this area (e.g., Heimer, 1997; Markowitz & Felson, 1998; also see Bridges & Weis, 1989; Sampson & Bartusch, 1998). Perhaps, as Anderson seemed to imply, it is not Black racial status alone that contributes to investment in the code but the combination of Black racial status, poverty, and social isolation. If this combination is critical, one would expect to find that Black racial status contributes to code-related beliefs primarily among young people who reside in

Brezina et al. / THE CODE OF THE STREET

323

neighborhoods of concentrated poverty. Interactions of this nature could be explored in future research, ideally with data sets that contain an oversample of such young people.9 Alternatively, it may be the larger social context of the inner-city ghetto, where race and economic disadvantage are closely intertwined, that is more deeply implicated in the development of code-related beliefs—rather than race at the person level (see Sampson & Bartusch, 1998). Third, only a limited amount of research exists on the role of gender in the development of code-related beliefs and youth violence. Prior research has tended to focus on male youth violence and to permit comparisons between our findings and those of previous studies, we followed suit. However, certain data indicate that “a nontrivial proportion of girls are engaging in some violent delinquency” (Heimer & De Coster, 1999, p. 294). Moreover, Anderson (1994) observed that an increasing number of inner-city girls are finding it necessary to enact the code and are “doing their own fighting” (p. 92). In what may be the most complete empirical study of gender and violence to date, Heimer and De Coster (1999) found that code-related beliefs (definitions favorable to violence) are important predictors of both male and female violent behavior. Nonetheless, although code-related beliefs appear to play a significant role in female violence, their research also indicated that females are less likely than males to adopt the code, and this fact helps to explain why girls engage in relatively low levels of violent behavior. 10 Yet several gender-related issues remain to be explored. For instance, it is possible that the code is mainly expressed in indirect terms among females. As Anderson (1994) observed, “Often if a girl is attacked or feels slighted, she will get a brother, uncle, or cousin to do her fighting for her” (p. 92). To capture such dynamics in empirical studies, it may be necessary to design gender-specific measures of both code-related beliefs and participation in violent behavior. If indeed females tend to enact the code indirectly, we may find that they play a nontrivial role in the instigation of male violence and, in general, a more substantial role in violence than has been suggested by previous quantitative studies. Fourth, in terms of future research, it should be noted that our study is limited to an examination of individual-level factors. It will also be important to examine the role of contextual (neighborhood-level) factors in the development of code-related beliefs and violent behavior. Although contextual effects on individual developmental outcomes tend to be relatively small in size, this does not mean that they are inconsequential (see Elliott et al., 1996). It is possible, for instance, that individual-level investment in the code is influenced by the extent to which the code is observed in the larger community. In communities where commitment to the code is widespread, young people may face strong pressure to observe the code regardless of their personal orientations, they may have disproportionate contact with individuals who are heavily invested in the code and who campaign for status and respect by “testing” or provoking others, and they may have disproportionate contact with others who place a premium on the display of toughness, “guts,” or “nerve” (also see Agnew, 1999; Baumer et al., 2003; Bernard, 1990; Brezina, Piquero, & Mazerolle, 2001; Elliott et al., 1996; Felson et al., 1994; Stewart et al., 2002). A recent study of homicides in St. Louis is suggestive. The authors of this study (Kubrin & Weitzer, 2003) observed a disproportionate rate of “retaliatory murders” in disadvantaged neighborhoods and, in many such cases, they found evidence that the assailants were encouraged by others in their community (family members and friends) and by awareness of community norms and expectations to enact the street code and pursue physical retaliation against their victims. Thus, in addition to individual-level beliefs and orientations, a number of factors at the

324

Youth Violence and Juvenile Justice

neighborhood level of analysis may contribute to the development of violent behavior and these should be the focus of future, multilevel research. Given existing empirical support for Anderson’s subculture of violence thesis, along with the additional findings we report here, we believe that further research on the above issues is likely to be fruitful and should be encouraged. Although the problems associated with subjectivity in ethnographic research have been well documented, and although the charge of “observer bias” has recently been leveled at Anderson’s work (see Wacquant, 2002), we do not believe that such concerns leave the validity of ethnographic accounts forever in doubt. Studies of the same social phenomenon conducted by different researchers, using different methodologies, can increase confidence in “objectivity,” especially if a consistency in major findings is observed across studies. At the moment, it appears that a degree of consistency of findings has emerged with respect to Anderson’s subculture of violence thesis. This pattern of findings lends plausibility to the thesis and, as Anderson had hoped, suggests that its relevance is not limited to the violence of inner-city youth or to the local areas he observed.

NOTES 1. The difficulty of conducting unbiased field research may be compounded by personal involvement with subjects. As emphasized by Monti (1992), “Anderson lived with the persons he studied for 14 years” and he “cared a great deal about the problems they faced” (p. 330). 2. It should be noted that, according to Anderson (1999), the code not only prescribes a proper way to respond to provocation but also a proper comportment or way of carrying oneself on the street (e.g., intimidating facial expressions or gait). 3. Anderson refers to “street” and “decent” families, which are terms that the local residents themselves used to distinguish the “criminal element” from families who were oriented toward civility and mainstream values. 4. A related body of theory and research suggests that abuse or neglect in the home may contribute to “coercive ideation” or to a “defensive world view,” which encourages cynical interpretations of social interaction and helps to justify the use of force or intimidation (see Colvin, 2000; Fleisher, 1995). 5. Actual analyses are based on a pairwise covariance matrix. The median sample size of the pairwise covariances is 752. 6. During the parent interviews, one parent for each child respondent was asked, “In general how do you react when [name of the respondent] does something wrong?” The parent was then presented with three different sets of disciplinary options and, for each set, was asked to select the option that he or she would choose first when dealing with the child. For instance, one set of options listed the following possible actions: “hit or threaten to hit him,” “send him to his room,” “explain that he should accept responsibility for his behavior and request that he make up for it,” or “discuss his behavior with him as well as my reasons for being upset with it.” If the parent selected “hit or threaten to hit him” or “send him to his room,” their response was interpreted as an indication of a powerassertive parenting style. Based on responses to all three sets of options, a one-item index of powerassertive parenting was constructed and was scored in the following manner: 0 = power-assertive actions never selected, 2 = power-assertive action selected one time, and 3 = power-assertive actions selected two or more times. For each set of disciplinary options, the interviewed parent was also asked to gauge the likely response of his or her spouse, providing the basis for a second, one-item index of power-assertive parenting.

Brezina et al. / THE CODE OF THE STREET

325

7. Although it would have been possible to increase model fit by including additional, “freely estimated” error correlations, we decided not to pursue this approach for two reasons. First, resulting improvements in model fit would not have been dramatic. Second, there is concern that post hoc, datadriven modifications may simply capitalize on idiosyncratic characteristics of the sample and, hence, on chance alone (MacCallum, Roznowski, & Necowitz, 1992). 8. We conducted a series of sensitivity analyses to gauge the robustness of our findings in the face of two potential problems: (a) departures from assumptions of normality and (b) bias in estimates due to the failure to specify reciprocal relationships. First, the distributions of certain variables are skewed (e.g., violent behavior and victimization). Technically, these distributions violate the assumption of normality required by the maximum-likelihood estimation procedure in LISREL 8. To address this issue, we estimated a model using logarithmically transformed measures of violent behavior and victimization to reduce skew. In this case, we observed an identical pattern of results. Second, we estimated a series of models that specified reciprocal (nonrecursive) relationships between peer aggression and violence at Time 2 and between victimization and violence at Time 1. To specify these relationships at Time 2, we used previous measures of peer aggression and violence as instruments. To specify these relationships at Time 1, we used questions from the parents’ interview as instruments for victimization and violence (note that previous measures of victimization and Time 1 violence are not available). The inclusion of these nonrecursive effects in the model did not change the overall pattern of results. Although the nonrecursive models included in our sensitivity analyses are based on strong identifying assumptions, we believe that the results of these analyses increase confidence in our original findings. The results suggest that our main model is stable and robust and that our findings are not seriously biased by the omission of nonrecursive effects (also see Heimer, 1997). 9. With the aid of multiple regression analyses, we tested for possible interactions between Black racial status, socioeconomic status, and urban residence. We failed to observe significant effects. It is possible, however, that the National Youth Survey (NYS) may not contain a sufficient number of disadvantaged Black male youths to permit an adequate test or that more specific indicators of economic disadvantage are required (e.g., residence in an area of concentrated poverty). 10. To examine the effects of victimization and perceived opportunity on female violence, we conducted separate analyses using data from the female subsample of the NYS. Consistent with the findings of Heimer and De Coster (1999), the results of our analyses indicate that code-related beliefs contribute to female violence. We also find that victimization exhibits significant direct and indirect effects on female violence, as it did in the case of males. However, perceived opportunity did not exhibit statistically significant effects, suggesting that, in terms of violent behavior, blocked opportunity is less consequential for girls than boys.

REFERENCES Agnew, R. (1994). The techniques of neutralization and violence. Criminology, 32(4), 555-580. Agnew, R. (1999). A general strain theory of community differences in crime rates. Journal of Research in Crime and Delinquency, 36(2), 123-155. Agnew, R., & Jones, D. (1988). Adapting to deprivation: An examination of inflated educational expectations. Sociological Quarterly, 29(2), 315-337. Anderson, E. (1990). Streetwise: Race, class, and change in an urban community. Chicago: University of Chicago Press. Anderson, E. (1994, May). The code of the streets. Atlantic Monthly, 273, 81-94. Anderson, E. (1998). The social ecology of youth violence. Crime and Justice, 24, 65-104.

326

Youth Violence and Juvenile Justice

Anderson, E. (1999). Code of the street: Decency, violence, and the moral life of the inner city. New York: Norton. Anderson, E. (2002). The ideologically driven critique. American Journal of Sociology, 107(6), 15331550. Baron, S. W., & Hartnagel, T. F. (1998). Street youth and criminal violence. Journal of Research in Crime and Delinquency, 35(2), 166-192. Baron, S. W., Kennedy, L. W., & Forde, D. R. (2001). Male street youths’ conflict: The role of background, subcultural, and situational factors. Justice Quarterly, 18(4), 759-789. Baumer, E., Horney, J., Felson, R., & Lauritsen, J. L. (2003). Neighborhood disadvantage and the nature of violence. Criminology, 41(1), 39-71. Bernard, T. J. (1990). Angry aggression among the truly disadvantaged. Criminology, 28(1), 73-96. Brezina, T., Piquero, A. R., & Mazerolle, P. (2001). Student anger and aggressive behavior in school: An initial test of Agnew’s macro-level strain theory. Journal of Research in Crime and Delinquency, 38(4), 362-386. Bridges, G. S., & Weis, J. G. (1989). Measuring violent behavior: Effects of study design on the reported correlates of violence. In N. A. Weiner & M. E. Wolfgang (Eds.), Violent crime, violent criminals (pp. 14-34). Newbury Park, CA: Sage. Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A. Bollen & J. S. Long (Eds.), Testing structural equation models (pp. 136-162). Newbury Park, CA: Sage. Burton, V. S., & Cullen, F. T. (1992). The empirical status of strain theory. Journal of Crime and Justice, 15(2), 1-30. Canada, G. (1995). Fist, stick, knife, gun: A personal history of violence in America. Boston: Beacon. Cloward, R. A., & Ohlin, L. E. (1960). Delinquency and opportunity: A theory of delinquent gangs. Glencoe, IL: Free Press. Cohen, A. (1955). Delinquent boys: The culture of the gang. Glencoe, IL: Free Press. Colvin, M. (2000). Crime and coercion: An integrated theory of chronic criminality. New York: St. Martin’s. Cullen, F. T., & Agnew, R. (Eds.). (1999). Criminological theory: Past to present. Los Angeles: Roxbury. Dunford, F., & Elliott, D. S. (1984). Identifying career offenders using self-reported data. Journal of Research in Crime and Delinquency, 21(1), 57-86. Elliott, D. S., & Ageton, S. S. (1980). Reconciling race and class differences in self-reported and official estimates of delinquency. American Sociological Review, 45(1), 95-110. Elliott, D. S., & Huizinga, D. (1983). Social class and delinquent behavior in a national youth panel: 1976-1980. Criminology, 21(2), 149-177. Elliott, D. S., Huizinga, D., & Ageton, S. S. (1985). Explaining delinquency and drug use. Beverly Hills, CA: Sage. 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(4), 389-426. Esbensen, F., & Huizinga, D. (1991). Juvenile victimization and delinquency. Youth & Society, 23(2), 202-228. Felson, R. B., Liska, A. E., South, S. J., & McNulty, T. L. (1994). The subculture of violence and delinquency: Individual vs. school context effects. Social Forces, 73(1), 155-173. Fleisher, M. S. (1995). Beggars and thieves: Lives of urban street criminals. Madison: University of Wisconsin Press. Heimer, K. (1997). Socioeconomic status, subcultural definitions, and violent delinquency. Social Forces, 75(3), 799-833.

Brezina et al. / THE CODE OF THE STREET

327

Heimer, K., & De Coster, S. (1999). The gendering of violent delinquency. Criminology, 37(2), 277-312. Jankowski, M. S. (1995). Ethnography, inequality, and crime in the low-income community. In J. Hagan & R. Peterson (Eds.), Crime and inequality (pp. 80-94). Palo Alto, CA: Stanford University Press. Joreskog, K. G., & Sorbom, D. (1993). LISREL 8. Chicago: Scientific Software International. Kasinitz, P. (1992). Bringing the neighborhood back in: The new urban ethnography. Sociological Forum, 7(2), 355-363. Kessler, R. C., & Greenberg, D. F. (1981). Linear panel analysis: Models of quantitative change. New York: Academic Press. Kubrin, C. E., & Weitzer, R. (2003). Retaliatory homicide: Concentrated disadvantage and neighborhood culture. Social Problems, 50(2), 157-180. Lauritsen, J. L., Sampson, R. J., & Laub, J. H. (1991). The link between offending and victimization among adolescents. Criminology, 29(2), 265-292. Luckenbill, D. F., & Doyle, D. P. (1989). Structural position and violence: Developing a cultural explanation. Criminology, 27(3), 419-453. MacCallum, R. C., Roznowski, M., & Necowitz, L. B. (1992). Model modifications in covariance structure analysis: The problem of capitalization on chance. Psychological Bulletin, 111(3), 490-504. Majors, R., & Billson, J. M. (1992). Cool pose: The dilemmas of Black manhood in America. New York: Lexington Books. Markowitz, F. E., & Felson, R. B. (1998). Socio-demographic differences in attitudes and violence. Criminology, 36(1), 117-138. Messerschmidt, J. W. (1993). Masculinities and crime. Lanham, MD: Rowman and Littlefield. Monti, D. J. (1992). On the risks and rewards of “going native.” Qualitative Sociology, 15, 325-332. Rubin, L. B. (1976). Worlds of pain: Life in the working-class family. New York: Basic Books. Sampson, R. J., & Bartusch, D. J. (1998). Legal cynicism and (subcultural?) tolerance of deviance: The neighborhood context of racial differences. Law and Society Review, 32(4), 777-804. Short, J. F., Jr. (1991). Poverty, ethnicity, and crime: Change and continuity in U.S. cities. Journal of Research in Crime and Delinquency, 28(4), 501-518. Stewart, E. A., Simons, R. L., & Conger, R. D. (2002). Assessing neighborhood and social psychological influences on childhood violence in an African-American sample. Criminology, 40(4), 801-829. Stoll, C. S. (1974). Female and male: Socialization, social roles, and social structure. Dubuque, IA: William C. Brown. Vowell, P. R., & May, D. C. (2000). Another look at classic strain theory: Poverty status, perceived blocked opportunity, and gang membership as predictors of adolescent violent behavior. Sociological Inquiry, 70(1), 42-60. Wacquant, L. (2002). Scrutinizing the street: Poverty, morality, and the pitfalls of urban ethnography. American Journal of Sociology, 107(6), 1468-1532. Wilkinson, D. L. (2001). Violent events and social identity: Specifying the relationship between respect and masculinity in inner-city youth violence. Sociological Studies of Children and Youth, 8, 235-269. Wilson, W. J. (1997). When work disappears: The world of the new urban poor. New York: Random House. Wolfgang, M. E., & Ferracuti, F. (1967). The subculture of violence. London: Tavistock. Zinn, M. B. (1982). Chicano men and masculinity. Journal of Ethnic Studies, 10(1), 29-44.

328

Youth Violence and Juvenile Justice Timothy Brezina is associate professor of sociology at Tulane University. His current research examines the personal consequences of criminal and delinquent involvement, including the real and symbolic rewards of deviant behavior. Recent publications have appeared in the journals Criminology, Justice Quarterly, and Social Psychology Quarterly. Robert Agnew is professor of sociology and violence studies at Emory University in Atlanta. His work focuses on the causes of delinquency, and forthcoming publications include Juvenile Delinquency: Causes and Control (2nd ed., Roxbury) and Why Do Criminals Offend? A General Theory of Crime and Delinquency (Roxbury). Francis T. Cullen is Distinguished Research Professor of Criminal Justice and Sociology at the University of Cincinnati. His most recent works include Combating Corporate Crime: Local Prosecutors at Work, Criminological Theory: Context and Consequences, and Criminological Theory: Past to Present—Essential Readings. He is president of the American Society of Criminology and a past president of the Academy of Criminal Justice Sciences. John Paul Wright is associate professor of criminal justice at the University of Cincinnati. He has published articles on the effects of adolescent employment on delinquency, the effects of parenting on offspring misbehavior, and the role of money in youthful misconduct. His current research focuses on the genetic heritability of traits related to crime, including self-control, stability in antisocial behavior over time, and the biosocial development of serious violence.