FINDINGS
Overview and “Vote Count”
The appendix lists the studies used for each measure of community characteristics. If community factors are differentially associated with violent crime, one pattern of supportive findings would indicate that the association between community factors and violence is more consistent than the association between community factors and measures of nonviolent offending. Table 10.1 summarizes the comparisonlevel analysis across the studies. Turning first to the structural factors that are theoretically predicted to increase crime, we see that ethnic heterogeneity is significantly and positively related to violence in 21 out of 38 models (55%), and negatively related in 1 out of 38 models (2.6%). The association was not statistically significant in the other models. In comparison, ethnic heterogeneity is also positively related to nonviolent crime in most models (67%). At the study level, 40% of studies (4 out of 10) reported a preponderance of comparisons (PoC^{3}) that was positive and statistically significant for violence (see Table 10.2). That percentage was slightly higher for nonviolence (50%). We take the similarity of findings across violent and nonviolent outcomes as evidence that ethnic heterogeneity is not a good differential predictor of violence.
Another core construct of social disorganization theory, residential instability, was positively related to violence in 76 out of 122 models (62%), and negatively related in only 6 out of 122 models (4.9%). Residential instability is similarly, positively related to nonviolent crime in 61 out of 93 models (66%). At the study level,
Table 10.1 ComparisonLevel Analysis of Community Factors and Violent
and Nonviolent Crime, by Category
Independent 
Total 
X 
% 
О 
% 

Variable 
Comparisons 
Comparisons 
Comparisons 

Ethnic 
Violent 
38 
21 
55.3 
1 
2.6 
Heterogeneity 
Nonviolent 
45 
30 
66.7 
2 
4.4 
Residential 
Violent 
122 
76 
62.3 
6 
4.9 
Instability 
Nonviolent 
93 
61 
65.6 
4 
4.3 
Family 
Violent 
46 
28 
60.1 
0 
0 
Disruption 
Nonviolent 
38 
22 
57.9 
1 
2.6 
Subculture of 
Violent 
4 
4 
100 
0 
0 
Violence 
Nonviolent 
4 
4 
100 
0 
0 
Disorder and 
Violent 
22 
16 
72.7 
0 
0 
Incivilities 
Nonviolent 
17 
7 
41.2 
0 
0 
Residential 
Violent 
54 
18 
33.3 
4 
7.4 
Stability 
Nonviolent 
41 
17 
41.5 
5 
12.2 
Social 
Violent 
40 
32 
80.0 
1 
2.5 
Networks/ 
Nonviolent 
42 
33 
78.6 
3 
7.1 
Collective 

Efficacy 
X Coefficient in the predicted direction (greater neighborhood instability, disorder, more crime or greater stability, collective efficacy, less crime)
О Coefficient is in the opposite direction of that predicted
Table 10.2 StudyLevel Analysis of Community Factors and Violent and Nonviolent Crime, by Category
Independent Variable 
Total Studies (k) 
Violent PoC X (k) 
% X 
Nonviolent PoC X (k) 
% X 
Ethnic Heterogeneity 
10 
4 
40% 
5 
50% 
Residential Instability 
17 
11 
64.7% 
11 
64.7% 
Family Disruption 
12 
7 
58.3% 
7 
58.3% 
Subculture of Violence 
1 
1 
100% 
1 
100% 
Disorder and Incivilities 
5 
4 
80% 
1 
20% 
Residential Stability 
11 
3 
27.3% 
3 
27.3% 
Social Networks/Collective 
8 
4 
50% 
5 
62.5% 
Efficacy 
PoC X The preponderance of comparisons (PoC) reported in the study are in favor of the predicted hypothesis
the PoC was roughly equal between violent and nonviolent offending as well. This evidence suggests that residential instability is also not a differential predictor of violence.
A frequentlytested structural characteristic is the prevalence of disrupted families in a community. The prevalence of disrupted families is positively related to violence in 28 out of 46 models (60%), and negatively related in none of the comparisons in our sample (see Table 10.1). Measures of disrupted families are positively related to nonviolent crime in similar proportion: 22 out of 38 models (58%), and negatively related in 1 out of 38 models (2.9%). At the study level, the proportions are identical (see Table 10.2). As before, this set of comparisons suggests that the prevalence of disrupted families is not a differential predictor of violence.
A different pattern is observed in the last community measure expected to increase violence. Disorder is positively related to violence in 16 out of 22 models (73%), and negatively related in none of the sampled studies. Disorder is positively related to nonviolent crime in a smaller proportion of models (7 out of 17, or 41%), and negatively related in none of the models. Consistent with the comparisonlevel evaluation, the studylevel evaluation shows that in 4 out of 5 (80%) of studies reporting on disorder/incivilities and violent crime, the PoC is in the expected direction and statistically significant, compared to only 1 out of 5 in the studies of nonviolent offending (20%) (see Table 10.2). Thus far, disorder and incivilities are the only measures with support as a differential predictor of violence.
Now we turn to studies analyzing community factors expected to reduce violence. Residential stability is unsurprisingly expected to decrease crime because it is the opposite of residential instability. We separate our summary of the two complementary sets of variables, in part, to avoid confusion due to the different expected direction of effects and, in part, because the findings have been much less consistent when indicators of stability are used compared to indicators of instability (see also Chapter 3). Across studies, measures of residential stability are negatively related to violence in 18 out of 54 models (33.3%), and positively related in 4 out of 54 models (7.4%). Residential stability is negatively related to nonviolent crime in a somewhat higher proportion of models, 17 out of 41 (42%). At the study level, the proportion of studies reporting a PoC that is statistically significant is exactly equal when we look at violent crime and nonviolent crime (3 out of11 studies, 27.3%). Thus, no evidence for the differential etiology hypothesis emerges from the dualdependent variable analysis, though researchers may take heed of the weaker results when residential stability measures are used.
The next construct is neighborhood ties, which is operationalized as “neighboring activities,” local organizational participation, and collective efficacy. Across studies, neighborhood integration measures are negatively related to violence in 32 out of 40 models (80%), and positively related in just 1 out of 40 models (2.5%). Measures of neighborhood ties are also consistently negatively related to nonviolent crime (33 out of 42 coefficients, 78.6%). At the study level, measures of neighborhood ties are slightly more likely to have a PoC that is negative and significant when the dependent variable is nonviolent (k = 5) compared to violent (k = 4).
This suggests that while indicators of neighborhood social ties are consistently related to both violent and nonviolent crime, there is no evidence here suggesting that they differentially predict of violence.
Only one study presented relationships between an indicator of a subculture of violence and both violent and nonviolent offending. Felson et al. (1994) found that both violent and nonviolent crime were significantly associated with their measure of subculture in all models.