Chronic Exposure to Community Violence and Criminal Behavior: A Marginal Structural Modeling Approach

Abstract

Objectives

Assess the hypothesis that the effects of community violence on delinquency accumulate as exposures increase in chronicity over time. Few existing studies have examined this hypothesis, and none have accounted for the possibility of time-variant confounding.

Methods

The sample comprised seven years of prospective, longitudinal data on 986 juvenile offenders from the Pathways to Desistance study. Inverse probability of treatment-weighted marginal structural models were used to account for time-variant confounders that influence selection into victimization and witnessing violence across the first six years of the study. Outcomes (any, aggressive, and income offending) were assessed in the seventh year.

Results

Analyses illustrate that failing to account for time-variant confounding—or doing so with traditional regression methods—substantially biases effect estimates. Properly specified marginal structural models show that each additional year with at least one victimization event increased the probabilities of participants’ offending during the seventh year of the study by 4.0–5.5 percentage points, whereas witnessing violence increased these probabilities by 2.9–3.7 percentage points.

Conclusions

The study shows that chronic exposure to community violence poses a substantial threat to development, including the cessation of crime by justice-involved adolescents and emerging adults. More work is needed, however, to understand the intervening mechanisms involved. Implications for policies to reduce community violence and for analyses of other time-varying exposures are discussed.

Appendices

Appendix 1: Prevalence of Exposures and Outcomes

See Table 5.

Table 5 Percentages of analytic sample reporting exposures and outcomes

Appendix 2: Descriptions of Measures Used for Weighting and Control Variables

Baseline Measures

Age The age of participants in years. Mean = 16.01 (SD = 1.14, range = 14–19).

Gender The biological sex of participants. 15.41% Female, 84.49% male. Reference category was Male.

Race/ethnicity dummy variables White was the reference category (22.63%). Other categories were Black (37.72%), Hispanic (35.09%), and Other (4.56%).

Maternal warmth Mean of 9 items from the Quality of Parental Relationships Inventory tapping the perceived warmth of treatment by his/her mother (e.g., “How often does your mother let you know she really cares about you?”). Scale mean = 3.18 (SD = 0.65, range = 1–4, α = 0.92).

Maternal hostility Mean of 12 items from the Quality of Parental Relationships Inventory tapping maternal hostility toward the participant (e.g., “When you and your mother have spent time talking or doing things together how often did your mother shout or yell at you because she was mad at you?”). Scale mean = 1.60 (SD = 0.43, range = 1–4, α = 0.85).

Maternal substance use Indicator of whether participant’s mother has a past or current drug problem. No = 72.92%, Yes, in the past = 23.53%, Yes, currently = 3.55%.

Family criminal involvement Count of the number of family members that have been arrested or jailed. Mean = 1.87 (SD = 1.70, range = 0–6).

Socioeconomic status Index combining information on parents’ education and occupation, via Hollingshead (1957). Mean = 51.30 (SD = 12.12, range = 16.5–77.0).

Parental knowledge Mean of 5 items from the Parental Monitoring Inventory (e.g., “How much does [primary caregiver] how who you spend time with?”). Mean = 2.72 (SD = 0.81, range = 1–4).

Early behavior problems Count of the presence of five behavior problems prior to age 11 (i.e., fighting, stealing, cheating at school, disturbing class, trouble for being drunk/high). Mean = 1.48 (SD = 1.18, range = 0–5).

School fighting Single Likert-style item indicating how often participants get into fights at school. Mean = 5.86 (SD = 1.80, range = 1–7).

Psychopathy Checklist—Youth Version—Factor 1 Sum of scores from 8 items of the Psychopathy Checklist—Youth Version tapping the interpersonal and affective dimensions of psychopathy. Mean = 4.91 (SD = 3.40, range = 0–19, α = 79).

Psychopathy Checklist—Youth Version—Factor 2 Sum of scores from 9 items of the Psychopathy Checklist—Youth Version tapping the socially deviant lifestyle dimension of psychopathy. Mean = 8.26 (SD = 3.80, range = 0–21, α = 0.93).

Time-Variant Measures

Prior offending variety Proportion of the types of 22 offending behaviors participants reported involvement in during the recall period. Baseline mean = 0.15 (SD = 0.15, range = 0–0.91).

Mental health problems Mean of subscale scores from the Brief Symptom Inventory. Baseline mean = 0.56 (SD = 0.53, range = 0–3.25).

Substance use consequences Count of life consequences due to substance use (e.g., “had problems or arguments with family or friends?”). Baseline mean = 2.07 (SD = 3.21, range = 0–15).

Impulse control Mean of 8 items from the Weinberger Adjustment Inventory (e.g., “I’m the kind of person who will try anything once, even if it’s not that safe”). Baseline mean = 2.93 (SD = 0.94, range = 1–5, α = 0.76).

Suppression of aggression Mean of 7 items from the Weinberger Adjustment Inventory (e.g., “People who make me angry better watch out.”). Baseline mean = 2.79 (SD = 0.99, range = 1–5, α = 0.78).

Future orientation Mean of 8 items from the Future Outlook Inventory (e.g., “I will keep working at difficult, boring tasks if I know they will help me get ahead later”). Baseline mean = 2.33 (SD = 0.54, range = 1–4, α = 0.68).

Aspirations Mean of 7 items capturing future aspirations for work, family, and law abidance (e.g., How importance is it to graduate from college?”). Baseline mean = 4.45 (SD = 0.54, range = 1–5, α = 0.67).

Expectations Mean of 6 items capturing future expectations for work, family, and law abidance (e.g., “What do you think your chances are to earn a good living?”). Baseline mean = 3.43 (SD = 0.82, range = 1–5, α = 0.81).

Moral disengagement Mean of 32 items from the Mechanisms of Moral Disengagement scale (e.g., It is alright to beat someone who bad mouths your family”). Baseline mean = 1.62 (SD = 0.36, range = 1–3, α = 0.88).

Crime rewards—stealing Mean of 5 items tapping the social rewards of stealing (e.g., If I take things, my friends will like me more for the clothes, music, drugs, or other things that I can get now”). Baseline mean = 1.91 (SD = 0.50, range = 1–4, α = 0.76).

Crime rewards—fighting Mean of 5 items tapping the social rewards of fighting (e.g., “If I beat someone up, other people my age will respect me more”). Baseline mean = 2.19 (SD = 0.53, range = 1–4, α = 0.75).

Crime rewards—robbery Mean of 5 items tapping the social rewards of robbery (e.g., “If I rob someone, I'll get more respect from adults in my neighborhood”). Baseline mean = 1.86 (SD = 0.50, range = 1–4, α = 0.82).

Crime rewards—personal Mean of 7 items tapping the personal rewards of crime (e.g., “How much of a ‘thrill’ or ‘rush’ is it to stab someone?”). Baseline mean = 2.42 (SD = 2.44, range = 0–10, α = 0.88).

Psychosocial maturity Mean of 30 items from the Psychosocial Maturity Inventory tapping the domains of self-reliance, identity, and work orientation (e.g., “I change the way I feel and act so often that I sometimes wonder who the 'real' me is”). Baseline mean = 3.02 (SD = 0.46, range = 1–4, α = 0.89).

Gang involvement Binary indicator of whether a participant admitted being part of a gang in the past 6 months (19.2% Yes, 80.8% No at baseline).

Antisocial peers Mean of 12 items asking participants how many friends had engaged in 12 forms of antisocial behavior. Baseline mean = 2.32 (SD = 0.92, range = 1–5, α = 0.92).

Routine activities: Mean of 4 items tapping participants’ involvement in unsupervised activities (e.g., “How often do you go to parties or other social gatherings?”). Baseline mean = 3.83 (SD = 0.85, range = 1–5, α = 0.62).

Neighborhood conditions Mean of 21 items tapping participants perceptions of physical (e.g., cigarettes on ground, graffiti, abandoned cars) and social disorder (e.g., adults loitering, people drunk or passed out, presence of sex workers) in their neighborhoods. Baseline mean = 2.31 (SD = 0.73, range = 1–4, α = 0.94).

Opportunities for success Mean of 8 items tapping whether participants perceive opportunities to success in their neighborhoods (e.g., “In my neighborhood, it is hard to make money without doing something illegal”). Baseline mean = 3.27 (SD = 0.65, range = 1–5, α = 0.66).

Neighborhood gun access Response to the question: “If a young person in this neighborhood wants to buy a guy, he/she can”). Baseline mean = 2.75 (SD = 1.32, range = 1–5).

Street time The proportion of the recall period that participants spent outside of institutional settings that do not have community access. First measured at the year one data collection interview. Mean = 0.58 (SD = 0.40).

Witnessed violence Binary indicator of whether participants witnessed any form of violence during the recall period. At baseline, 93.00% of participants reported ever witnessing violence.

Victimization Binary indicator of whether participants experienced any form of direct victimization during the recall period. At baseline, 67.13% of participants reported ever being directly victimized.

Appendix 3: Pre- and Post-Weighting Covariate Balance for Year 1 Exposures

See Table 6, 7.

Table 6 Covariate balance for year 1 victimization

Table 7 Covariate balance for year 1 witnessing violence