Spatial Associations of Alcohol Outlet Density with Neighborhood Crime and Child Opportunity: A Hierarchical Bayesian Ecological Study

Abstract

Passaic county, New Jersey (NJ) has a high population density and diverse racial composition, with significant socioeconomic disparities that vary by city. Such disparities expose residents, and particularly children, to environmental conditions that may be harmful to their health and wellbeing such as high exposure to crime, violence, and high amounts of alcohol outlets. This study explores the association between alcohol outlet density (AOD) and neighborhood-level crime rates and child opportunity index (COI) measures in Passaic county, NJ. We applied a hierarchical Bayesian spatial approach to model the relationships between AOD and crime and child opportunity at the census tract level, controlling for neighborhood-level socioeconomic covariates (unemployment, access to health insurance, and living below the federal poverty level) and spatial correlation. In total there were 120 census tracts in Passaic county which had a total of 554 alcohol outlets. The median number of alcohol outlets in a single census tract was four; the maximum number of outlets in a single census tract was 26. Our results show that neighborhoods in the highest tertile of AOD experienced significantly higher rates of property crime compared to those in the lower tertiles. Furthermore, lack of access to health insurance was consistently associated with both an increase in crime and a lower child opportunity index. These findings suggest that reducing AOD and increasing access to health insurance in high-risk neighborhoods could mitigate crime and enhance opportunities for children. Policy interventions addressing these factors may be critical for improving community health and safety.

Background

Research suggests that the location of alcohol outlets within a neighborhood can contribute to health disparities which negatively impact resident health and overall quality of life, particularly for youth residents of the neighborhood (Milam et al., 2014). The geography of alcohol outlets is primarily measured through alcohol outlet density (AOD), which includes the number of functioning alcohol outlets within defined geographical boundaries and can infer the availability of alcohol access in a geographical region (Hobday & Meuleners, 2018). High AOD can typically be seen in downtown areas of urban cities which can lead to increased rates of alcohol-related issues in these areas, and further, there is a trend of higher AOD seen concentrated in lower income areas, exposing those residents to greater risks of alcohol-related issues (Morrison et al., 2015).

Economic theories have previously suggested that alcohol outlets tend to be in the downtown areas of cities due to their higher population and easier access, which increases the demand for their products (Morrison et al., 2015). In addition, these businesses may choose to locate in lower income areas to reduce operating costs compared to locating in higher income areas, thereby exposing lower income residents of these urban areas to alcohol-related issues (Morrison et al., 2015). Studies have shown higher rates of AOD and alcohol advertising in urban areas with more African American/Black and Hispanic populations along with more impoverished areas, which leaves vulnerable populations more susceptible to negative health effects from alcohol consumption (Milam et al., 2014).

Some health effects associated with residents living in areas with high AOD include liver problems and death, hospital admissions, trauma, chronic disease, suicide, pedestrian collisions, self-reported injury, and interpersonal violence (Gmel et al., 2016). High AOD has been found to be positively associated with increased alcohol consumption among adolescents and exposure to alcohol, tobacco, and other drugs (Milam et al., 2014). Children who live in urban, racial minority neighborhoods are more likely to see tobacco and alcohol advertisements on their way to school than adolescents who live in predominantly White neighborhoods (Milam et al., 2014). These factors reflect the importance of the built environment on children’s health.

The built environment is defined as the physical, man-made structures in the environment that influence a person’s ability to live and work (Gascon et al., 2016). Some factors of the built environment demonstrated to impact population health include the walkability of a neighborhood, neighborhood greenspace, access to grocery stores with nutritious foods, access to facilities for physical activity, transportation systems, traffic, and general urban infrastructure (Gascon et al., 2016; Masoumi, 2017; Villanueva et al., 2023). Children’s health in the built environment is especially important as they are more susceptible to environmental exposures harming their health due to their typically higher rates of being outside and physical activity than adults (Gascon et al., 2016). These factors can also affect children’s health by not only impacting the resources families need to raise children but influencing how children adopt health behaviors through interacting with their environment (Ortegon-Sanchez et al., 2021). Multiple studies have shown that positive interactions with the built environment can increase positive behaviors in children (e.g., playing and walking outdoors) (Gascon et al., 2016; Masoumi, 2017; Ortegon-Sanchez et al., 2021), while negative experiences with the built environment can lead to less time spent outside (Ortegon-Sanchez et al., 2021). Therefore, high AOD in neighborhoods with a large population of children can affect their health--not only through exposure to violent and/or neglectful behavior from adults consuming alcohol, but also by reducing their time spent outside due to the high levels of crime, less greenspace, and other factors associated with alcohol outlets in their neighborhoods.

There are avenues to mitigate the variety of issues stemming from high AOD in neighborhoods. Researchers have focused on the negative health effects resulting from AOD in neighborhoods (e.g., violent crimes, assault) (Milam et al., 2014). However, previous studies have not investigated the potential of improving outcomes for children specifically if the number of alcohol outlets were to decrease in their neighborhood. The purpose of this analysis was to assess the associations between modifiable aspects of the built environment (i.e., location and number of alcohol outlet retailers) and neighborhood-level crime and child opportunity in Passaic county, New Jersey (NJ), United States (US).

Methods

The work is part of a larger study funded by the National Institutes of Health (DP5OD029636; PI: Opara) which focuses on neighborhood level factors that impact substance use and mental health in the city of Paterson, a racially diverse and urban city in Northeastern NJ. The present analysis focused on Passaic county, NJ, in which Paterson is a part of.

Community context

Passaic county is one of the most populous, dense, and socioeconomically diverse counties in NJ. Passaic county has a diverse population (37.4% Hispanic; 62.6% White, 12.8% Black/African American, 0.67% Native American, 5.0% Asian, 0.03% Pacific Islander, 15.1% other races, and 3.71% two or more races). Racial and ethnic distributions of individual cities within Passaic county demonstrate that the county remains racially segregated (Flaxman, 2013). In addition, there is a large socioeconomic gap in Passaic county with some cities experiencing alarmingly high numbers of residents living below the poverty line or having poverty rates up to double the county and state rates (Census, 2021b; Jersey, 2015).

Measures

Outcomes

Child opportunity index

The Child Opportunity Index (COI) was used to measure social disadvantage for children. The COI is a validated composite index comprised of 29 indicators of social determinants of health (SDOH) linked to the US Census (Noelke et al., 2020). COI contains three overall domains: health opportunity (comprised of measures of airborne microparticles, ozone concentration, industrial pollutants in air, water, or soil, hazardous waste dump sites, fast food restaurant density, healthy food retailer density, extreme heat exposure, NatureScore [green space, tree canopies, parks, air, noise, and light pollution], walkability, community safety-related non-profits, vacant housing, health-related non-profits, health insurance coverage); socioeconomic opportunity (comprised of measures of employment rate, high-skill employment rate, full-time year-round earnings, median household income, poverty rate, public assistance rate, adults with advanced degrees, very high-income households, adults without high school degrees, very low-income households, broadband internet access, crowded housing, mobility-enhancing friendship networks, single-parent families, non-profit organizations, homeownership rate, aggregate home values, aggregate capital income, aggregate real estate taxes); and educational opportunity (comprised of measures of private pre-K enrollment, public pre-K enrollment, reading and math test scores, reading and math test score growth, poverty-adjusted reading and math test scores, advanced placement course enrollment, college enrollment in nearby institutions, high school graduation rate, adult educational attainment, child enrichment-related non-profits, teacher experience, and school poverty). COI indicators are assigned individual z-scores and summed to an overall z-score using indicator-specific weights that reflect how strongly each indicator predicts children’s long-term health and economic outcomes. COI data in this analysis are provided at the census tract level. Each of the individual measures for a selected census tract (health, socioeconomic, and education) were modeled separately as three outcomes of interest in the first multivariate spatial regression model.

Crime

Crime data were obtained from ArcGIS Applied Geographic Solutions, which supplies a collection of databases that aggregates and analyzes 2015-2021 data from the FBI Uniform Crime Reporting Program (UCRP) (Crime/Law Enforcement Stats (Uniform Crime Reporting Program), 2024; CrimeRisk Methodology, 2023). Personal crime (murder, rape, robbery, and assault) and property crime (burglary, larceny, and motor vehicle theft) were the focus of the present analysis. Crime index values at the national level are 100, representing the average crime in the US. Values more than 100 indicate a higher crime estimate than the national average, whereas values less than 100 represent lower crime than the national average (CrimeRisk Methodology, 2023). To improve model stability, these values were divided by 100 (i.e., the national standard) to reduce the magnitude and variance of the crime variables. The individual measures for a selected census tract (personal crime, property crime) were modeled jointly in the second multivariate spatial regression model.

Exposure

Alcohol Outlet Density

Information on establishments with an active liquor license in Passaic County were county was obtained from the NJ Division of Alcoholic Beverage Control (Division of Alcoholic Beverage Control). Information including license number, type, location of establishment, and the date of active licensing was obtained. At the time of analysis, there were a total of 554 active alcohol distributers in Passaic county across 120 census tracts. Tertiles of the distribution of outlets across all tracts were computed and each tract was assigned to one of the three groups based on their observed counts. This categorical predictor was used in the subsequent statistical modeling to capture any nonlinear behavior in the association. The first tertile (reference category) as was for census tracts with zero to one alcohol outlet, the second was two to five, and the third was six or more.

Covariates

At the census tract level, socioeconomic demographic data including percentage with no health insurance, percent unemployment, and percent living 100% below the federal poverty level were obtained from the American Community Survey, a survey administered by the US Census; details of which can be found elsewhere (Census, 2022). Percentages were used rather than raw counts to account for varying population size in each tract. These variables measuring percentages were standardized to their respective interquartile ranges (IQR) to improve model stability and fit. The regression coefficients for these variables represent the expected change in the outcome associated with an increase equal to the interquartile ranges (i.e., difference between the 25th and 75th percentiles).

Statistical approach

Data were aggregated into ESRI ArcGIS Pro to generate maps. The Passaic County shapefile (obtained from the US Census) was used to create a spatial adjacency matrix using R to account for spatial clustering in the regression model (i.e., to define which tracts were bordering each other) (2020 TIGER/Line^® Shapefiles: Census Tracts, 2020). Multivariate spatial regression models with the three components of child opportunity included as the primary outcome and AOD and other tract-level covariates as explanatory variables, were fitted using R statistical software. In a separate analysis, we included the previously described multivariate crime vector as the outcome with the predictors unchanged. The unit of analysis was the census tract; number of alcohol outlets in each census tract was calculated in ArcGIS. Census tracts provide a more appropriate unit of measurement due to their relatively consistent size (ranging from 2,500 to 8,000 residents) and demographic composition. This ecological unit of analysis offers a more precise means of accounting for population heterogeneity compared to the broader, more variable ZIP code areas. The multivariate spatial model for a vector of continuous outcomes is given as

$$Y_{kj}={\mathbf{x}}_k^{\mathrm{T}}{\beta }_j+{\varphi }_{kj}+{ϵ}_{kj},{ϵ}_{kj}\sim \mathrm{N}(0,{\sigma }_{ϵ}^2)$$

where $Y_{kj}$ refers to the COI or crime measure for domain $j$ at census tract $k$; ${\mathbf{x}}_k$ is the vector of covariates specific to tract $k$ with ${\beta }_j$ is the corresponding vector of regression parameters specific to COI or crime domain $j$; and ${\varphi }_{kj}$ is the location/domain specific random effect that accounts for spatial correlation across census tracts and cross-correlation between COI or crime outcomes from the same tract. The spatially correlated random effects are modeled using the multivariate version of the Leroux conditional autoregressive model (Halloran, 1999) such that

$${\varphi }_k\mid {\varphi }_{-k},\rho ,\Sigma \sim \text{MVN}\left(\frac{\rho {\sum }_{i=1}^n{\mathrm{w}}_{ki}{\varphi }_i}{\rho {\sum }_{i=1}^n{\mathrm{w}}_{ki}+1-\rho },\frac{\Sigma }{\rho {\sum }_{i=1}^n{\mathrm{w}}_{ki}+1-\rho }\right)$$

where ${\varphi }_k$ is a vector of three random effects specific to tract $k$; ${\varphi }_{-k}$ is the collection of all effects except those from tract $k$; $\rho \in [0,1)$ describes the strength of spatial correlation in data with $\rho =0$ indicating spatial independence; $\Sigma$ is a three-by-three covariance matrix (two-by-two for the crime model) that describes the correlation/variation among outcomes from same tract; and${\mathrm{w}}_{ki}$ is a binary variable indicating whether tract $k$ and $i$ share a common border with $w_{ii}\equiv 0$ by definition (i.e., a tract not cannot be a neighbor of itself). We specified weakly informative prior distributions for the model parameters with full details presented in the supplementary material. The model was fitted using the CARBayes package in R statistical software using Markov chain Monte Carlo sampling techniques (Lee, 2013; Team, 2013). Specifics regarding posterior sampling and convergence checking are also given in the supplement. Posterior means and 95% quantile-based credible intervals are presented to represent point estimates and measures of uncertainty, respectively.

Results

Figure 1 displays the map of the three domains of child opportunity (socioeconomic, health, and education) and alcohol outlets in Passaic county. Figure 2 displays the map of personal crime index and property crime index and alcohol outlets in Passaic county. Figure 3 displays the map of socioeconomic characteristics of the census tracts (unemployment, living 100% below the federal poverty level, and no health insurance) and alcohol outlets in Passaic county. Maps showed clustering of low levels of child opportunity, high rates of crime, and lower socioeconomic characteristics in areas with AOD, with hot spots (i.e., very high clustering) in the densely populated urban cities of Paterson and Passaic City located in the southeastern region of Passaic county.

Fig. 1.

Map of distribution of the three domains of Child Opportunity Index (COI) (socioeconomic status [SES], health, and education) z-scores and alcohol outlets at the census tract level in Passaic county, NJ.

Fig. 2.

Map of distribution of Personal Crime Index and Property Crime Index and alcohol outlets at the census tract level in Passaic county, NJ.

Fig. 3.

Map of distribution of percentage unemployment, living 100% below federal poverty level, and no health insurance and alcohol outlets at the census tract level in Passaic county, NJ.

Table 1 displays the descriptive characteristics of the census tract sample. In total, there were 120 census tracts in Passaic county, with a total of 554 alcohol outlets. The median number of alcohol outlets in a single census tract was four, ranging from zero to 26. Approximately 18% (n=21) of census tracts contained 0 or 1 alcohol outlet, 54% (n=64) had two to five, and 29% (n=35) had six or more outlets. Median child educational opportunity z-score was −0.45, median child health opportunity z-score was −0.22, and median child socioeconomic opportunity z-score was 0.22. The personal crime index in Passaic county was 56 with scores ranging from 3 to 406 (standardized values were median=0.56, range: 0.03-4.06), and property crime index was 67.5 with scores ranging from 13 to 176 (standardized median=0.68, range: 0.13-1.76). Approximately 8% of Passaic county was unemployed, 12% had no health insurance, 20% were living 100% below the federal poverty level.

Table 1.

Descriptive characteristics of Passaic county at the census tract level (n=120).

Alcohol outlets in a single census tract (median (min. value, max. value))	4 (0, 26)
Alcohol outlet tertiles (n (%))
0-1	21 (17.5)
2-5	64 (53.3)
>6	35 (29.2)
Child Educational Opportunity Z-score (median (min. value, max. value))	−0.45 (−1.28, 1.60)
Child Health Opportunity Z-score (median (min. value, max. value))	−0.22 (−1.9, 1.45)
Child Socioeconomic Opportunity Z-score (median (min. value, max. value))	0.22 (−1.27, 1.40)
Personal Crime Index (median (min. value, max. value))	0.56 (0.03, 4.06)
Property Crime Index (median (range))	0.68 (0.13, 1.76)
Percent unemployment* (mean (SD)) [IQR]	0.18 (0.81) [−0.36, 0.64]
Percent with no health insurance* (mean (SD)) [IQR]	0.16 (0.68) [−0.43, 0.57]
Percent living 100% below the federal poverty level* (mean (SD)) [IQR]	0.29 (0.77) [−0.32, 0.68]

^*Percentage variables standardized to interquartile range.

Abbreviations: SD, standard deviation, IQR, interquartile range.

Table 2 displays the results of the multivariate regression model for the three domains of child opportunity by AOD and demographic covariates. A one-IQR increase in the percent of the population who were uninsured was associated with significantly lower educational opportunity z-scores at the census tract level (β=−0.38; 95% Credible Interval [CI]: −0.53, −0.23). A one-IQR increase in the percent unemployed within a census tract was associated with significantly lower child health opportunity scores (β=−0.10; 95% CI: −0.18, −0.02), and a one-IQR increase in the percent of residents living below the federal poverty level had significantly lower child health opportunity scores (β=−0.09; 95% CI: −0.17, −0.01).

Table 2.

Results from the multivariate spatial linear regression model examining the association between child opportunity at the census tract level and number of alcohol outlets in Passaic county.

	Beta estimate	95% Credible Interval
Childhood educational opportunity
Number of alcohol outlets
0-1	Ref.	Ref.
2-5	0.07	−0.09, 0.24
≥6	−0.004	−0.18, 0.18
Percent without health insurance*	−0.41	−0.56, −0.25
Percent unemployment*	−0.07	−0.16, 0.02
Percent living 100% below federal poverty level*	−0.02	−0.10, 0.07
Childhood socioeconomic opportunity
Number of alcohol outlets
0-1	Ref.	Ref.
2-5	−0.03	−0.28, 0.21
≥6	−0.09	−0.35, 0.16
Childhood health opportunity
Number of alcohol outlets
0-1	Ref.	Ref.
2-5	−0.09	−0.24, 0.07
≥6	0.02	−0.19, 0.15
Percent unemployment*	−0.10	−0.18, −0.02
Percent living 100% below federal poverty level*	−0.09	−0.17, −0.01

^*Percentage variables standardized to interquartile range (IQR); Beta estimates interpreted with every IQR increase.

Three multivariate spatial models; educational opportunity and health opportunity adjusted for (at the census tract level) percent of the population with no health insurance, percent of the population that are unemployed, and percent of the population living 100% below the federal poverty level. Socioeconomic opportunity did not include these covariates to prevent perfect prediction. Health opportunity index model did not include ‘Percent without health insurance’ as a covariate as health insurance coverage was included in the original index.

Notes: bold text indicates parameters whose 95% credible intervals excluded zero.

Table 3 displays the results of the multivariate regression model for personal crime and property crime by AOD and demographic covariates. Results showed that AOD was not statistically associated with the three types of child opportunity. Results showed that the third and highest tertile of AOD (six or more alcohol outlets within one census tract) was statistically associated with increased property crime (β=0.18, 95% Credible Interval [CI]: 0.04, 0.33) at the census tract level. Additionally, a one-IQR increase in the percent uninsured was significantly associated with higher rates of personal crime (β=0.31; 95% CI: 0.05, 0.57).

Table 3.

Results from the multivariate spatial linear regression model examining the association between crime at the census tract level and number of alcohol outlets in Passaic county.

	Beta estimate	95% Credible Interval
Personal crime index
Number of alcohol outlets
0-1	Ref.	Ref.
2-5	−0.01	−0.28, 0.27
≥6	0.08	−0.22, 0.37
Percent without health insurance *	0.31	0.05, 0.57
Percent unemployment*	0.04	−0.10, 0l19
Percent living 100% below federal poverty level*	0.10	−0.03, 0.24
Property crime index
Number of alcohol outlets
0-1	Ref.	Ref.
2-5	0.02	−0.11, 0.16
≥6	0.18	0.04, 0.33
Percent without health insurance*	0.06	−0.06, 0.17
Percent unemployment*	−0.04	−0.11, 0.03
Percent living 100% below federal poverty level*	0.04	−0.03, 0.10

^*Percentage variables standardized to interquartile range (IQR); Beta estimates interpreted with every IQR increase.

One multivariate spatial model adjusted for (at the census tract level) percent of the population with no health insurance, percent of the population that are unemployed, and percent of the population living 100% below the federal poverty level.

Notes: bold text indicates parameters whose 95% credible intervals excluded zero.

Discussion

This study aimed to explore the associations between modifiable aspects of the built environment (i.e., number of alcohol outlet retailers) and neighborhood-level crime and child opportunity in Passaic county, NJ. Findings indicate that, while AOD was not statistically associated with child opportunity in Passaic county, the third tertile of AOD (six or more alcohol outlets within one census tract) was associated with an increase in property crime. The lack of association between AOD and child opportunity may be due to the geographic scales of census tracts being too large to identify observable spillovers, a spatial bias known as the modifiable areal unit problem (Openshaw, 1979). Additionally, socioeconomic variables were a consistent and strong driver of higher crime and lower child opportunity. Interestingly, an increase in percent unemployment was found to be associated with decreased property crime. Consistent with previous research, spatially, higher AOD was concentrated in areas with higher rates of property crime (Lardier et al., 2020). Clusters of high AOD, high crime indexes, low COI, and low socioeconomic characteristics were found in the densely populated cities of Paterson and Passaic City, both located in Passaic county. The city of Paterson has one of the largest proportions of people of color living in the county and within the state of NJ, with 61.4% identifying as Hispanic, 25.7% as African American or Black, and 27.2% as White (Census, 2021b). Furthermore, approximately, 29.1% of Paterson residents live below the poverty line; Paterson’s child poverty rate is 41%, which is significantly higher than the state child poverty rate (16%) (Census, 2021b; Jersey, 2015). Approximately, 25.5% residents living in Passaic City have a poverty rate of 21.1% which is about 1.5 times the county rate (13%) and more than double the rate in NJ (10%) (Census, 2021a). Our maps offer a critical spatial analysis of these disparities, revealing concentrated areas of economic disadvantage that may exacerbate health inequities and further marginalize vulnerable populations within Paterson and Passaic City.

Our novel approach of modeling AOD as an independent variable provides evidence that modifying (i.e., reducing) AOD in a community may drive rates of crime down. Further, given the racial and socioeconomic distribution of Passaic county, these findings deserved to be explored in a broader context. A larger body of research has documented the associations of higher AOD in socially and economically disadvantaged neighborhoods, and predominantly communities of color (Berke et al., 2010). In fact, the increased placement of alcohol outlets in disadvantaged neighborhoods and predominantly communities of color are a manifestation of structural racism and discrimination that increase challenges in accumulating capital and economic investments in urban areas and perpetuate narratives of neighborhoods as being “risky” or less “profitable” (Scott et al., 2020). One prominent example of such was the practice of “redlining” by the Homeowner’s Loan Corporation (HOLC) which codified racially segregated housing practices and has been a highly predictive determinant on contemporary social, economic, and health-related opportunities (Fishback et al., 2022). One study by Lee and colleagues (2020) reported that HOLC’s lowest ranked neighborhoods in the Northern California region were predominantly minority residents and had significantly higher concentration of alcohol outlets compared to higher-ranked neighborhoods (Lee et al., 2020). Similarly, Sadler and colleagues (2023) reported that within the cities of Baltimore, Maryland and Flint, Michigan, AOD was significantly higher in previously redlined neighborhoods compared to neighborhoods not within redlining zones (Sadler et al., 2023). It is also necessary to consider how these historical processes have enacted ‘symbolic violence’ (i.e., the ways cultural practices produce and maintain power asymmetries) among neighborhoods with increased AOD, specifically that the placements of AOD may also lead to the perpetuation of stigmatizing narratives that isolate and degrade neighborhoods and preserving individualized responsibilities that blame residents and maintain the conditions that keep them oppressed (Scott et al., 2020). The overall findings of this study provide groundwork quantitative evidence to support this phenomenon.

The relationship between health insurance access and personal crime has been explored in previous research and can be explained through the lens of economic stability and access to healthcare services. While census tract-level associations between insurance coverage and crime provide important insights into broader socio-economic patterns, interpreting these findings at the individual level may introduce ecological fallacy, as the census-level data do not specify whether community members without health insurance are the ones directly involved in personal crimes. Nonetheless, research has found that communities with inadequate access to health insurance often experience systemic economic strain, limited healthcare infrastructure, and fewer mental health and substance use treatment resources which have been found to be associated with higher rates of criminal activity (Bondurant et al., 2018; Finkelstein et al., 2012). He & Barkowski (2018) argue that the expansion of Medicaid increased the opportunity cost of crime by improving physical and mental health and thereby lowering incentives to engage in criminal activity (He & Barkowski, 2020). Studies suggest that increased access to healthcare services may help reduce crime rates by promoting economic stability and improving overall community health (He & Barkowski, 2020).

Our analysis also found that lack of health insurance is significantly associated with lower child opportunity. Child opportunity has been directly linked with access to health insurance and healthcare access (Noelke et al., 2020). In addition, communities with low child opportunity have been found to be associated with high absenteeism and low academic achievement, thus making it difficult for youth to transition successfully into adulthood with such structural barriers in place. (Opara et al., 2022). Lack of health insurance can limit a child's social opportunity by restricting access to critical health services necessary for physical, emotional, and cognitive development (Currie & Chorniy, 2021). Children in families without access to health insurance are less likely to receive preventive care, routine checkups, or early interventions for developmental issues, which can negatively impact their school performance and social mobility (Noelke et al., 2020). At the neighborhood level, similar to the patterns observed with crime, communities with a high prevalence of uninsured individuals may also experience a lack of infrastructure to support key environmental health features, such as adequate green spaces, safe recreational areas, clean air and water, and access to healthy food options (Hunter et al., 2019). This has major implications for policy to expand access to public health insurance, integrate health and social services, focus on community development initiatives, and make data-driven policy decisions.

Community context and implications

There was a census tract in northeastern Passaic County which presented an unusual combination of low poverty, low crime, high insurance coverage, low educational COI, high health and social COI, and high unemployment. With guidance from our community partners, as part of our larger community-based participatory research study, we learned that this area’s geographic isolation and suburban–rural character contribute to limited access to services and employment, despite in-migration driven by more affordable housing. These structural factors may help explain the coexistence of socioeconomic advantage and disadvantage. Our findings highlight the importance of examining rural–urban variation when interpreting community-level indices like COI in areas with a mix of urban, suburban, and rural communities. Partnering with local community stakeholders and members is critical to accurately contextualize and address disparities in under-resourced or geographically isolated areas.

Strengths & limitations

This was one of the first analyses to spatially explore the association between AOD and crime and child opportunity together in Passaic county, NJ. Additionally, this analysis was novel in that it accounted for spatial dependency and variability, correlation among domains from a multivariate outcome, and the unit of analysis was at the census tract level, a smaller geographical unit that accounts for population size and demographics. Another strength of this analysis was modeling AOD as an independent and modifiable variable in relation to estimating change in crime and child opportunity. Regarding the crime data, despite the comprehensive compilation the FBI UCRP, the reporting of crime varied and were inconsistently documented due to differences in jurisdictional levels. That is, each city has its own police department and own procedures of reporting crime data, which may result in potential inconsistencies in data and comparisons across geographies. UCRP may limit accuracy in violent crimes data due to under-reporting of crime that involved minimal to no physical contact (Lardier et al., 2020). Because this study particularly attempts to focus on geographic-related factors on social opportunities, such patterns of crime are greatly influenced by local law enforcements practices, where strategic policing patrols specific neighborhoods lead to an inflated reporting of crime that may deviate from a standard reporting of crime. Previous studies have highlighted the practice of over-policing in predominantly Black and Brown neighborhoods; place-based stigma renders communities of color more susceptible to heightened scrutiny by the police (Gaston et al., 2023). Furthermore, communities of color may also have mistrust in the police, which may also result in underreporting of crimes within certain neighborhoods (Murphy et al., 2018). Another potential limitation of this study is the “edge effect,” which arises when premises located on the borders or edges of exterior geographical units are closer to individuals residing in neighboring units than to those within the same unit (Cameron et al., 2016). This phenomenon may lead to measurement inaccuracies, particularly in assessing access to alcohol outlets. For instance, children living near the edge of their geographic area might experience greater exposure to outlets in adjacent but exterior boundaries, which our density calculations cannot not fully capture. As a result, this could underestimate or misrepresent the true spatial accessibility and potential opportunity for exposure in such border areas. Finally, while the maps visually indicated higher AOD in areas characterized by lower child opportunity, this observed spatial association did not translate into statistically significant COI outcomes. This could be due to the modifiable areal unit problem when using aggregate data in a large geographical boundary results in varying or non-significant statistical findings, despite their visible representation on spatial maps. Additionally, our models adjusting for demographic and socio-environmental covariates might attenuate the direct statistical association between AOD and COI. Despite these limitations, our novel approach provides insight into socioeconomic factors impacting crime and social opportunity at the census tract level.

Conclusion

This study aimed to quantify the association between AOD and crime, and child opportunity and AOD in Passaic county, NJ at the neighborhood-level. Results showed AOD as a predictor of property crime, lack of health insurance as a consistent driver of personal crime and lower child opportunity, and higher AOD spatially concentrated in areas with higher rates of crime and lower child opportunity. Implications of this study include development of tailored community-level prevention efforts and policy to address exposure of alcohol outlets to reduce crime and raise child opportunity in high-risk neighborhoods.