Racial Inequality, Ethnic Inequality, and the System Involvement of At-Risk Youth: Implications for the Racial Invariance and Latino Paradox Theses

Abstract

Drawing from the inequality and crime, racial invariance, and Latino paradox literatures, the effects of inequality on youth reoffending are examined. Specifically, hierarchical logistic regression models are estimated to determine: (1) whether racial and ethnic inequality have similar contextual effects on the continued delinquent behavior of at-risk youth and (2) whether these effects are specific to black or Latino/a youth residing in Maricopa County, Arizona (N = 13,138). Findings suggest that racial inequality increases reoffending while ethnic inequality decreases reoffending. Additionally, Latino/a youth are less likely to reoffend in areas characterized by high income and racial inequality. Structural theories of crime should continue to account for the importance of culture and the resilient responses employed by Latinos/as living in criminogenic environments.

Keywords:

• racial inequality
• ethnic inequality
• youth reoffending
• racial invariance
• Latino paradox

Notes

¹ These works typically focus on absolute poverty (e.g. concentrated disadvantage); yet inequality is simply placing the living conditions of individuals in a relative sense. Both classic (Blau & Blau, 1982) and current (Hipp, 2007; Hipp & Boessen, 2013) empirical works have confirmed the importance of relative poverty for the prediction of crime. These works suggest that both the racial invariance and Latino paradox literatures may be incomplete when focusing solely on absolute poverty, and the inclusion of racial/ethnic inequality within these literatures answers the call by scholars (e.g. Steffensmeier et al., 2010) to incorporate additional structural conditions into tests of the racial invariance thesis.

² It is instructive to note that some studies at the macro-level do examine the racial invariance of inequality by examining whether “gaps” in offending between whites, blacks, and Latinos can be explained by disparities in structural disadvantage (Velez, Krivo, & Peterson, 2003). Importantly, this research finds that gaps in disadvantage do not fully explain racial and ethnic gaps in violence (Ulmer, Harris, & Steffensmeier, 2012) and that exposure to disadvantage may be more important for explaining white–Latino gaps in violence as compared to white–black gaps in violence (cf. McNulty & Bellair, 2003; Phillips, 2002). Part of the reason that inequality has not been incorporated more often into the racial invariance literature is due to the trend toward race-specific indicators of crime (Peterson & Krivo, 2005, p. 336). By definition, inequality cannot be race-specific, unless using a racially-bound measure of inequality. Accordingly, studies find intraracial inequality to be more predictive of white crime than black crime (in disagreement with the racial invariance thesis) (Harer & Steffensmeier, 1992). Two studies (Harer & Steffensmeier, 1992; Krivo & Peterson, 2000) that did examine the invariance of interracial inequality found it to be unrelated to either black or white crime—yet both studies were limited by a focus on crime rates between blacks and whites only.

³ Some studies refer more broadly to an “Immigrant Paradox”—though they tend to focus on foreign-born individuals who would identify as Latino.

⁴ The population of referrals/arrests during this time period included 50,398 cases and 23,500 youth. Among the 23,500 unique youth cases, only 65% (n = 15,275) with valid address information could be followed for 24 months. Consistent with our theoretical focus, we excluded from the sample those youth of racial/ethnic backgrounds other than white, black, or Latino/a (n = 963). Results from supplemental analyses that included Asian Pacific Islander and American Indian youth were consistent with the findings reported below. To prevent jeopardizing the reliability of our estimates, we also removed from the sample all individuals residing in zip code areas containing fewer than 10 white, black, or Latino/a youth (n = 1,174). The removal of these cases did not alter our substantive findings.

⁵ Only youth who could be tracked by juvenile court for a 24-month period were included in the sample. Thus, by design, no youth in the sample is older than 16 years of age since age of juvenile jurisdiction ends at 18. A review of demographic characteristics among youth in the sample and those who were unable to be followed for 24 months yielded no statistically significant differences other than age.

⁶ Our measure of family structure does not differentiate between married and cohabitating parents or guardians. Individual-level economic indicators (e.g. parent or guardian income) were not included as they were unavailable in the present data.

⁷ Several additional structural control variables were included in the analyses such as percent renting homes, tenure, percent foreign-born, and percent female-headed households. These were weak and insignificant correlates of youth reoffending alongside robust individual-level predictors and had no bearing on the present findings. An indicator of urban or rural was not included since the zip codes in the sample were almost exclusively urban (range = 90.4–100.0% urban, mean = 99.9% urban).

⁸ An alternative approach to using cross-level interactions is to estimate race-specific models and compare the inequality coefficients across samples. The findings using this approach appeared consistent with those presented here, although methods typically used to compare coefficients across maximum likelihood regression models (e.g. Brame, Paternoster, Mazerolle, & Piquero, 1998) cannot be applied to hierarchical logistic regression.

⁹ Correlations between independent variables did not exceed .60, and variance inflation factors among independent variables were below 2.5 (well below the standard “conservative” cutoff of 4.0; Fox, 1991). Furthermore, none of the condition index values for equations presented in Tables 2 and 3 exceeded 20, which is under the critical threshold of 30 specified by Belsley, Kuh, and Welsch (1980). According to this evidence, observed correlations between the independent variables should not result in biased estimates or inefficient SE due to multicollinearity.

¹⁰ The ICC represents the proportion of the variance in the outcome that is between the level 2 units. For hierarchical logistic regression models, the interclass correlation is calculated by the following formula: (Raudenbush & Bryk, 2002). The small ICC found here is expected due to the fixed nature of the individual-level variance and because of the dichotomous nature of the dependent variable.

¹¹ As seen in Table 2, economic deprivation did not emerge as a significant predictor of reoffending. In supplemental analyses that assessed only structural-level (level 2) variables on youth reoffending, the effect of economic deprivation was positive and significant (p < .05). When individual-level variables (e.g. age, race/ethnicity, and risk score) are included alongside economic deprivation, however, its effects diminish to non-significance. Although the null effect of economic deprivation on criminal behavior may seem inconsistent with macro-level criminological research, this is not anomalous to the study of individual-level outcomes (e.g. Haynie, Silver, & Teasdale, 2006; Wright & Rodriguez, 2014). Moreover, to ensure that the effects of Latino and black disadvantage on youth reoffending were not being masked by the general measure of economic deprivation, supplemental analyses were conducted using race- and ethnic-specific indicators. The pattern of findings remained unchanged.

¹² The Gini index was calculated using the prln04.exe program provided by Francois Nielsen at the following website: http://www.unc.edu/~nielsen/data/data.htm.

¹³ Based on reference group theory, within-group inequality recognizes that group members look internally for standards of comparison. People assess how well (or how badly) they are faring by comparing themselves not with the population as a whole, but with particular reference groups with whom they share similar attributes (Harer & Steffensmeier, 1992). Moore (1991), for example, demonstrated that high-risk Latino youth tended to acknowledge others close by as references for feelings about themselves, with few comparisons to those in white or black communities.

Additional information

Notes on contributors

Kevin A. Wright

Kevin A. Wright is an assistant professor in the School of Criminology and Criminal Justice at Arizona State University. His research interests include criminological theory and correctional policy. His recent work has appeared in Criminology, Criminology and Public Policy, and Justice Quarterly.

Jillian J. Turanovic

Jillian J. Turanovic is an assistant professor in the College of Criminology and Criminal Justice at Florida State University. Her research interests include criminological theory, victimization and offending, and the collateral consequences of incarceration. Her recent work is published in journals including Criminology and the Journal of Quantitative Criminology.

Nancy Rodriguez

Nancy Rodriguez is a professor in the School of Criminology and Criminal Justice at Arizona State University. Her research interests include juvenile justice, criminal justice policy, and the intersection of race, ethnicity, gender, and crime.  Her recent work has appeared in Criminology, Journal of Research in Crime and Delinquency, and Justice Quarterly.