Immigrant Culture and Neighborhood Perceived Violent Crime and Violent Victimization: A Multi-Level Test of Enculturation to México, Acculturation to the US, and Support for the Code of the Street

Abstract

Research on neighborhood immigration and crime rarely tests cultural explanations for the null or inverse associations typically found. We examine if support for the code of the street and measures of immigrant culture (acculturation and enculturation) mediate this relationship. We employ multilevel path modeling and survey data from neighborhoods in El Paso County, Texas, merged with census data. Supporting predictions, enculturation is inversely associated with support for the code of the street and enculturation mediates an indirect relationship between neighborhood immigration and neighborhood violent victimization. However, this relationship does not appear when perceived neighborhood violent crime is used as the dependent variable or the acculturation measure. While the code of the street is positively associated with both perceived neighborhood violent crime and violent victimization, it is not associated with neighborhood immigration and does not serve as a mediator. Implications for cultural explanations of the neighborhood immigration-crime association are discussed.

Keywords:

• Acculturation
• code of the street
• enculturation
• immigration
• neighborhoods
• perceived violent crime
• violent victimization

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Notes

1 The distribution was divided into thirds with 32 clusters classified as low immigration (4.0-22.5%), 32 clusters as medium (22.6-31.0%), and 31 clusters as high immigration (31.5-54.1%).

2 A total of 78 census tracks are distributed among 46 neighborhood clusters. To test if there is meaningful population variation across tracts within clusters, we conducted two-sample proportion tests using 59 census tracks within 27 clusters (19 clusters contain only one census track). We compared the indicators across census tracts in the clusters that have more than one and found no significant differences between the census tracks within clusters.

3 While this approach to sample size within clusters is common in such research (see Schmidt et al., 2014), others (i.e., Clarke & Wheaton, 2007) argue that sample size within clusters has not been adequately studied and too often relies on unsubstantiated rules of thumb in the range of 15-30; using Monte Carlo simulations they suggest that the numbers of clusters is more important than the size of samples within clusters.

4 We excluded observations using listwise deletion, which was 3% of the total sample.

5 Approximately 81% of respondents are Hispanic, 12.8% are non-Hispanic White, 2.6% are African American, and 2.8% are Asian, American Indians, or another race/ethnicity.

6 In results not reported but available upon request we re-analyzed our models using official measures of aggravated assault, robbery, and homicide. Findings show few significant associations and no support for hypothesized relationships and that there were no significant association between the official measures and the survey measures we use. These results could be, in part, due to the very low levels of violent crime in El Paso. They could also be due to low and/or inconsistent levels of reporting violent crime to law enforcement across these neighborhoods which, importantly for our research setting, may stem from the tendency of immigrants to seek a “low profile” and avoid calling the police, even when they or a family member are a crime victim (see Guerra & Curry, 2023).

7 We excluded neighborhood percent Black in this study. Although it is proposed by Sampson et al. (1997) as another factor in the concentrated disadvantage scale, percent Black did not load the same as other items within concentrated disadvantage construct which is not surprising considering that the Black population averages about 3% across the clusters.

8 We conducted multilevel logistic path modeling for the paths that have binary outcomes (i.e., perceived neighborhood violent crime and violent victimization) and multilevel linear path modeling for the paths that have continuous outcomes (i.e., code of the street, acculturation to the US, and enculturation to the México).

9 Although bootstrapping is a popular approach for reflecting symmetry in the sampling distribution of the indirect effect, it can be problematic when data are clustered because the manner of data resampling is not necessarily straightforward (Leeden et al., 2008; Preacher & Selig, 2012; Wang et al., 2006). This difficulty is also reflected in Mplus’s lack of support for bootstrapping with multilevel data (Muthén & Muthén, 2017).

10 Because enculturation to México and acculturation to the US are key research variables that are highly correlated (r=-.52), we present two separate sets of models – one including enculturation to México but excluding acculturation to the US, and the other using acculturation but not enculturation.

11 This relationship is statistically significant at .10 level. Reporting significant results using a p-value that is less than .10 is debatable; however, we report this result because valuable information may be lost by ignoring “insignificant” results (Grabowski, 2016; see also Wasserstein et al., 2019).