Structure and Culture in African American Adolescent Violence: A Partial Test of the “Code of the Street” Thesis

Abstract

Researchers studying the race–violence relationship have tended to focus on either structural or cultural explanations. Although both explanations are important, they tend to be incomplete. We draw on Anderson’s “code of the street” thesis, which combines structural and cultural explanations to explain the high rates of violence among African American adolescents. Anderson argues that the street code, which supports the use of violence, is a cultural adaptation to negative neighborhood structural conditions, as well as family characteristics and racial discrimination. Using two waves of data from 720 African American adolescents from 259 neighborhoods, we investigated whether neighborhood context, family type, and discrimination influenced adoption of the street code. We also assessed whether the street code mediated the effects of neighborhood context, family characteristics, and racial discrimination on violent delinquency. Consistent with Anderson’s hypotheses, neighborhood structural characteristics, living in a street family, and discrimination significantly predicted adopting the street code. Moreover, the street code mediated about one fifth of neighborhood effects on violent delinquency, about one fifth of the effect of racial discrimination, and about 4 percent of the effect of family characteristics on violent delinquency. Overall, the results suggest that neighborhood context, family characteristics, and racial discrimination directly influence adopting the street code, and partially influence violence indirectly through the street code.

Keywords:

• race
• violence
• neighborhood context
• code of street

Acknowledgments

This research was supported by the National Institute of Mental Health (MH48165, MH62669) and the Center for Disease Control (029136‐02) and the National Institute of Justice (2005‐IJ‐CX‐0035). Additional funding for this project was provided by the National Institute on Drug Abuse, the National Institute on Alcohol Abuse and Alcoholism, and the Iowa Agriculture and Home Economics Experiment Station (Project #3320). The authors are grateful to Robert Adelman, Elijah Anderson, Eric Baumer, Robert Bursik Jr., Vincent Roscigno, Richard Rosenfeld, Eric Silver, and Richard Wright for their helpful and detailed comments on earlier drafts of this manuscript. We also wish to thank Chester Britt and the anonymous reviewers for their comments. All errors and omissions are those of the authors.

Notes

1. It is important to note that the families in the neighborhoods that Anderson studied used the terms “decent” and “street.” The families in our study also used these labels. In no way are we imposing a moral judgment by suggesting that “decent” families are good and “street” families are bad. We are simply following Anderson’s description distinguishing the two (for a lively debate on this topic, see the exchange between Wacquant, 2002 and Anderson, 2002).

2. Brezina et al. (2004) addressed how individual‐level covariates influence street code beliefs and violent behavior. However, they were unable to account for neighborhood structural disadvantage, family characteristics, and racial discrimination in relation to violence. Moreover, Baumer et al. (2003) focused on how neighborhood conditions shaped the nature of violence. They were unable to test how the street code developed and whether it influenced violence above and beyond neighborhood context, family characteristics, and racial discrimination.

3. We thank Eli Anderson and Bob Bursik for this clarification.

4. Since the Kuder–Richardson (1937) reliability coefficient is a special case of the Cronbach alpha, it is interpreted in the same manner as the Cronbach alpha reliability coefficient. The Kuder and Richardson coefficient (KR₂₀) was introduced to estimate the reliability of scales composed of dichotomously scored items. The following formula is used to assess the Kuder–Richardson reliability coefficient:

where N is the number of dichotomous items; p_i is the proportion responding “positively” to the ith item; q_i is equal to 1 − p_i ; and is equal to the variance of the total composite.

5. To assess the validity of our construct of neighborhood violence, we correlated our measure of neighborhood violence with police records of neighborhood violence. The correlation between the police reports and our construct of neighborhood violence was .58. This suggests that there is a moderately high level of agreement between the two constructs.

6. We estimated a measurement model with maximum likelihood estimation procedures. We used AMOS (Arbuckle & Wothke, 1999) to create single‐factor, latent variables for the decent and street family constructs. Each of the observed indicators of the latent variable was specified as a linear combination of a latent factor plus random measurement error. In each case, assessment of the overall measurement models demonstrated that the items selected to measure the theoretical constructs “decent” and “street” were statistically significant.

7. We thank Eli Anderson for his helpful and detailed comments in aiding with the development of our “decent” and “street” family constructs. Again, we understand the complexity that underlies the family types that Anderson identified and the challenge presented in trying to quantify these types. Nevertheless, we believe that our current analysis conceptualizes the basic characteristics of decent and street families that Anderson described.

8. To assess multicollinearity among the predictor variables, we examined the variance inflation factor (VIF). In the current study, multicollinearity does not appear to be a problem, as none of the VIFs was greater than 2.0, suggesting that the variables are theoretically and empirically distinct constructs (Fisher & Mason, 1981).

9. The models were first estimated separately for boys and girls. However, since subgroup comparisons showed a similar pattern of results, only the results obtained for the combined sample are presented.

10. We began by estimating a model in which family SES was the only predictor of adopting the street code. The effect of family SES was negative and significant, suggesting that adolescents who live in families with higher SES were less likely to adopt the street code. However, this relationship was eliminated once the effects of peer associations, school attachment, strain experiences, and prior adoption of the street code were added to the model.

11. We also included neighborhood violence² (−.015; t = −.725) and neighborhood disadvantage² (−.010; t = −.639) to assess whether higher levels of neighborhood violence and disadvantage influenced the street code.

12. Because the dependent variable, violent delinquency_T2, has a distribution that is highly skewed and overdispersed, we estimated negative binomial regressions, which can be viewed as an extension of the Poisson regression. However, the negative binomial model is more appropriate for dealing with overdispersion than the restrictive Poisson model. To account for overdispersion, the negative binominal introduces an additional parameter that estimates the extent of overdispersion in the model (Long, 1997).

13. Because street code_T2 and violent delinquency_T2 are measured at wave 2, we estimated a reciprocal effects model between the two variables, while controlling for prior effects to determine the directional relationship. The results of the SEM are presented in the Appendix (Figure A1). The standardized and significant path coefficient from violent delinquency_T2 to street code_T2 is .10, while the coefficient from street code_T2 to violent delinquency_T2 is .25 and also significant. To assess whether the paths are significantly different we used the equity of coefficient test which allows us to directly compare coefficient differences across models using the following equation (Paternoster, Brame, Mazerolle, & Piquero, 1998):

The results suggest that the path from street code to violent delinquency_T2 is significantly different and stronger than the path from violent delinquency to street code_T2 (.25 versus .10; t = 2.49; p < .05). Consistent with Anderson’s (1999) hypothesis, this finding provides some support for the notion that adopting the street code leads to delinquency rather than delinquency leading to the street code.

14. We first estimated Model 2 in Table 3 by including both street code_T1 (served as a control) and street code_T2. However, multicollinearity between the two variables was a problem. We decided to remove street code_T1, which reduced the multicollinearity to non‐problematic levels. Further, to assess whether there was a difference in the mediational effect whether we used street code_T1 or street code_T2, we estimated models by alternating the two. Regardless of which measure was used, the results were virtually identical. As such, we used street code_T2 to take advantage of the longitudinal data.

15. According to Long (1997), one way to interpret the results of the negative binominal model is by exponentiating the coefficients, subtracting 1, and multiplying the result by 100.

16. We also included neighborhood violence² (−.008; z = −.239) and neighborhood disadvantage² (−.003; z = −.610) to assess whether higher levels of neighborhood violence and disadvantage influenced violent delinquency. Neither emerged significant.

Additional information

Notes on contributors

Eric A. Stewart

Eric A. Stewart is an Assistant Professor in the Department of Criminology and Criminal Justice at the University of Missouri—St. Louis and a member of the National Consortium on Violence Research (NCOVR). He is also a W. E. B. DuBois Fellow with the National Institute of Justice. His research interests include crime over the life course; neighborhood, school, and family processes on adolescent development; and testing criminological theories.

Ronald L. Simons

Ronald L. Simons is a Professor of Sociology and a Research Fellow with the Institute for Behavioral Research at the University of Georgia. Much of his research has focused on the manner in which family processes, peer influences, and community factors combine to influence deviant behavior across the life course. He has also completed work on domestic violence and the effect of racial discrimination on adolescent development.