Rethinking Community Organization and Robbery: Considering Illicit Market Dynamics

Abstract

Research indicates that socially integrated communities experience lower rates of violent crime. However, we have a limited understanding of the specific neighborhood‐level processes accounting for this pattern. In particular, we know little about the convergence of informal control mechanisms and other contextual processes such as drug market activity. This paper addresses this issue by assessing the mediating role of drug markets and informal social control in the relationship between levels of violence and social structural characteristics of neighborhoods. Three research questions are examined: Do drug markets account for the relationship between social structural factors and robbery? Does informal social control mediate the link between social structural factors and drug market activity? Is informal social control directly related to robbery or indirectly through drug market activity? Results indicate that drug markets mediate the relationship between structural factors and robbery. After including drug markets in our models, the relationship between informal social control and robbery was no longer significant, suggesting that prosocial regulatory mechanisms reduce drug market activity, which in turn is associated with less violence.

Keywords:

• communities
• robbery
• social disorganization

Notes

1. These are the two largest cities in Kentucky with each urban center reported to have a population of more than a quarter‐million residents (US Census, 2000).

2. The level of drug use in a neighborhood was determined previously in a government funded study that examined injection drug users (see Warner, Leukefeld, & Kramer 2002). Specifically, the study investigators recorded the addresses of not‐in‐treatment injection drug users who were arrested in Louisville and Lexington (n = 1393). Addresses were complied so that investigators could then calculate the number of arrestees (i.e., non‐treated injection drug users) residing in the census block groups of each city. A block group was designated as a “high drug use” unit if it had at least ten known non‐treated injection drug users residing within its borders (see Warner et al. 2002).

3. Blocks adjacent to those designated as high drug use were purposely oversampled (Warner & Coomer, 2003)

4. Several studies in the literature specify census areas (e.g., block groups) as proxies for neighborhood social units. They are considered generally homogenous areas in terms of their socioeconomic, cultural and demographic makeup (see Sampson et al., 2002).

5. The chosen measure of drug activity, however, does not offer any insight as to which type of drug is specifically exchanged in the transactions that respondents report witnessing. To explore what specific type of drug may be changing hands in the incidents respondents observed, analysis was conducted comparing their observations to official drug arrest data. Results of correlation equations indicate that the measure derived from the survey responses relates the strongest with the cocaine arrest rate (r = .70), followed by arrests for marijuana (r = .41). These correlations provide rudimentary evidence suggesting that cocaine, and or its derivatives (i.e., crack) is the type of illicit drug that residents most likely view being sold in their neighborhood.

6. Initial estimates from model 1 indicate that it is just identified, or saturated (Kline, 1998). In other words, there as many parameters specified as there are variables, thus there are no available degrees of freedom to approximate model fit indices. The chi‐square value associated with the model is consequently zero. Though inferences can still be made regarding the effects of the predictors on the outcome variable in the absence of these indices, in order to draw meaningful conclusions about the results it is imperative to have knowledge of overall model fit. To obtain these measures, while achieving at the same time a more parsimonious model consistent with the theory, model re‐specification procedures are conducted (Arbuckle, 2005; Hox & Bechger, 1998; Kline, 1998). According to the literature, in order to resolve this issue each regression estimate (paths linking a given exogenous factor to robbery) should be constrained to zero and the remaining paths should be specified as free (Arbuckle, 2005; Kline, 1998). An inspection of model modification indices in AMOS will convey whether certain paths exert a meaningful decrease in the chi‐square if they are relaxed, thus allowed to associate with the outcome measure (Arbuckle, 2005). In applying this technique to the present model, three separate versions of model 1 are computed. In each equation only one exogenous factor is allowed to effect robbery, while the others are not (Arbuckle, 2005; Kline, 1998). According to the results reported by the model modification indeces, the paths connecting population structure to robbery can be fixed to zero, however the paths from the other two structural factors must be linked to robbery to achieve a sound causal model (Arbuckle, 2005; Kline, 1998). By constraining the path from population structure to robbery the model is left with one available degree of freedom, which allows for the calculation of model fit indices.

7. To evaluate model fit this study utilizes the goodness of fit index (GFI), the Root Mean Square Error of Approximation (RMSEA) and the chi‐square. These three indices are commonly employed as complementary indicators of model fit in covariance structure analysis (Hu & Bentler, 1999; Kline, 1998). Research recommends that acceptable levels of the GFI exceed .90 and that the RMSEA should be less than .05. Also, the literature indicates that a good fitting model should have an insignificant chi‐square discrepancy/df value that does not exceed 3 (Kline, 1998).

8. In each of the path models analyzed in this study the exogenous variables are allowed to covary. However, in order to conserve space these paths are not visually displayed in the model figures.

9. An alpha of .05 is used to determine statistical significance in all reported analyses.

10. Reported in panel B of Appendix C are the model fit indices, which taken together suggest that the model fit is adequate.

11. As with path model 1, in the current model the association between population structure and robbery is constrained to zero. A model specification procedure was again carried out in model 2 and the modification indices reveal that the addition of a free path between this variable and the robbery outcome is not necessary. However, due to theoretical reasons and practical considerations population structure is allowed to be associated with the neighborhood drug market measure.

12. This assumption is consistent with prior research. Indeed, previous studies concerned with the nature of informal social control have not specified population structure as an exogenous predictor (e.g., Bellair, 2000; Sampson et al., 1997; see especially Silver & Miller, 2004).

13. The model fit indices outlined in panel B of Appendix C suggests that the fit of model 4 meets conventional standards.

14. Since our measure of robbery fuses commercial as well as street robberies, it is possible that the measure is capturing the opportunity presented by commercial activity or “land use” within neighborhoods (Kurtz, Koons, & Taylor, 1998; Stark, 1987). We address this potential source of bias by implementing models based on the estimation of non‐commercial robberies. To do so we incorporated the non‐commercial robbery measure as an outcome, replacing it with the joint robbery measure used here. The results were remarkably similar to those reported in the present paper. What this may suggest is that our findings regarding the distribution of robbery are not a product of the prevalence of commercial activity, and instead, they are likely influenced by both systemic control properties and drug markets embedded within neighborhoods.

15. We are grateful to an anonymous reviewer for pointing out this limitation.

16. The model and results are available upon request from the authors. The model fit indices generated in the supplementary model suggest the fit is good (RMSEA 0.00; GFI: 0.990; chi‐square: 0.698; p value: 0.581).

17. Our measure of public control does not reflect the actual response of formal authorities to resident's calls for assistance. This measure is consistent with our construct of informal control, in that it captures perceptions of action. However, because the public control measure does not capture responsiveness, admittedly this may explain our failure to find a significant relationship between it and drug market activity.

18. In the conclusion section of their research in which they used the data employed in the current study, Warner et al. (2002) discuss the existence of feedback loops in neighborhoods much like those proposed here. For instance, they suggest that “weak collective efficacy leads to the emergence of street behaviors, which then increase the perception of street values” (p. 92). And in their view, widely held perceptions of street values may perhaps consequently weaken collective efficacy.