Children and Youth Services Review 35 (2013) 1933–1940

Contents lists available at ScienceDirect

Children and Youth Services Review

j ourna l homepage: www.e lsev ie r .com/ locate /ch i ldyouth

The moderating effect of substance abuse service accessibility on therelationship between child maltreatment and neighborhoodalcohol availability

Cory M. Morton ⁎National Development and Research Institutes, United States

⁎ 435 E 70th Street, 31 F, New York, NY 10021, UnitedE-mail address: cory.michael.morton@gmail.com.

0190-7409/$ – see front matter © 2013 Elsevier Ltd. All rihttp://dx.doi.org/10.1016/j.childyouth.2013.09.019

a b s t r a c t

a r t i c l e i n f o

Article history:Received 3 April 2013Received in revised form 27 September 2013Accepted 28 September 2013Available online 5 October 2013

Keywords:Child maltreatmentNeighborhood effectsSubstance abusePreventionAlcohol outlets

This study investigates how the relationship between dense concentrations of alcohol retailers and high rates ofchild maltreatment may be moderated by the presence of substance abuse service facilities. Using a cross-sectional design, the study utilized data from Bergen County, New Jersey on child maltreatment reports,alcohol-selling retailers, substance abuse service facilities, and the United States Census. Findings indicate childmaltreatment rates were higher in neighborhoodswith lower socioeconomic status and those with greater alco-hol outlet density. Neighborhoods with easily accessed substance abuse service facilities had lower rates of childmaltreatment. Additionally, the relationship between child maltreatment and alcohol outlet density wasmoder-ated by the presence of substance abuse service facilities. The study findings highlight the relevance of makingprimary prevention approaches readily available and using multi-sector collaboration to reduce childmaltreatment.

© 2013 Elsevier Ltd. All rights reserved.

1. Introduction

In contrast to the typical child maltreatment prevention strategiesthat focus on changing individual risk behavior, structural or environ-mental prevention approaches attempt to locate the source of riskwith-in the physical environment of communities and make changes thatsupport pro-social behavior (Blankenship, Friedman, Dworkin, &Mantell, 2006).Most of the existing programs to prevent childmaltreat-ment rely on secondary or tertiary prevention approaches and generallyfocus on family or individual (i.e. parent) level interventions aimed atimproving parent–child interactions, educating families, increasing for-mal and informal family support, or improving family home environ-ments through home visitation (Daro & Donnelly, 2002; Stagner &Lansing, 2009). Although these approaches may be successful for indi-vidual families, there has been little empirical support for the successof these family- and individual-level interventions in reducing overallrates of child maltreatment (reviewed in Reynolds, Mathieson, &Topitzes, 2009). Efforts rooted at the community level seek to perma-nently alter the environment, rooting out the structural determinantsof behavior and altering that structure to promote well-being(Yacoubian, 2007). There has been increasing attention in the field ofchild welfare in the identification of structural risk factors for child mal-treatment with particular attention to the role of neighborhood envi-ronment. This research has investigated the linkages between high

States. Tel.: +1 917 952 5760.

ghts reserved.

rates of neighborhood level drug and alcohol availability, poverty, resi-dential instability, and child care burden to higher rates of child mal-treatment, and findings suggest changes in the structure ofneighborhoods could influence rates of child maltreatment (Coulton,Crampton, Irwin, Spilsbury, & Korbin, 2007; Freisthler, Merritt, &LaScala, 2006).

1.1. Neighborhood substance use environment and child maltreatment

Investigating the substance use environment of neighborhoods isimportant as parental substance abuse has long been recognized as aproblem inextricably linked with child maltreatment. It is estimatedthat 40–80% of all children who come to the attention of child welfareagencies are living in homes with a substance abusing parent (Banks& Boehm, 2001; Besinger, Garland, Litronwick, & Landsverk, 1999;Young, Boles, & Otero, 2007). Of those substance abusing parents, alco-hol has been identified as the primary problem (Young et al., 2007). Theenvironment of substance availability as evidenced by the density anddistributions of alcohol retailers has important consequences for ratesof child maltreatment. Freisthler and her colleagues have conducted aseries of studies on neighborhoodswithin California counties to explorehow alcohol availability, operationalized as the density of alcohol re-tailers in a neighborhood, was linked with increased rates of child mal-treatment. Rates of child maltreatment were found to be higher inneighborhoods with dense concentrations of alcohol outlets(Freisthler, Midanik & Gruenewald, 2004; Freisthler, Needell, &Gruenewald, 2005). Adding an additional bar per 1000 people in the

1934 C.M. Morton / Children and Youth Services Review 35 (2013) 1933–1940

population was found to increase rates of child maltreatment by 2.2children per 1000 (Freisthler, 2004).

1.2. Neighborhood effects and child maltreatment

The bodyof research on structural risk factors for childmaltreatmenthas shown clear linkages between socioeconomic factors and features ofa neighborhood's built environment as being related to child maltreat-ment. Higher poverty, neighborhoods with high rates of residentialturnover, an environment that places a burden on caregivers in termsof diminished social networks, and neighborhoods inundatedwith alco-hol retailers all serve to influence greater rates of child maltreatment(Coulton, Korbin, & Su, 1999; Coulton, Korbin, Su, & Chow, 1995;Ernst, 2001; Freisthler, 2004; Molnar, Buka, Brennan, Holton, & Earls,2003; Paulsen, 2003).

1.2.1. PovertyImpoverished neighborhoods have been consistently linked with

high rates of child maltreatment (Coulton et al., 1999; Drake &Pandey, 1996; Freisthler, 2004; Paulsen, 2003). Rather than use eco-nomic indicators of poverty alone, researchers have utilized a varietyof U. S. Census indicators to proxy not only the economic dimension ofpoverty but the structural and demographic as well. Using indicatorsfrom the U. S. Census, Coulton and her colleagues have utilized principalcomponents analysis to reveal the underlying dimensions of poverty inneighborhoods, finding percent single, female-headed households, per-cent living below the poverty line, percent unemployed, number of va-cant housing units, 5year population loss, and percent African Americancombined to represent an impoverished neighborhood. They found anoverall strong relationship between child maltreatment and poverty,and further analysis showed that poverty was an important predictorin both predominately black and predominately white neighborhoods(Coulton et al., 1995, 1999; Korbin, Coulton, Chard, Platt-Houston, &Su, 1998). Similarly, Freisthler, Bruce, and Needell (2007) found neigh-borhood level measures of poverty were positively related to child mal-treatment substantiation rates for African American, Hispanic, andwhite children. At a population level, poverty has consistently beenthe best predictor of a family's chances for childwelfare system involve-ment (Coulton et al., 2007; Freisthler et al., 2006).

1.2.2. Residential instabilityResearchers have also investigated the relationship between neigh-

borhood residential instability and rates of child maltreatment. Stableneighborhoods are operationalized as those where residents have along tenure, housing units are fully/mostly occupied and there is lessmovement in and out of the area. The relationship to rates of child mal-treatment here has been weaker and less consistent for residential in-stability than that for impoverishment. Investigations have foundmixed effectswhen using residential instability to predict overall higherrates of childmaltreatment. Ernst (2001) found residential instability tobe a positive, significant predictor when investigating a county inMary-land, but Freisthler, Midanik, and Gruenewald (2004) and Coulton et al.(1999) did not find a relationship for neighborhoods in California andOhio, respectively.

1.2.3. Child care burdenChild care burden has been defined as the “amount of adult supervi-

sion and resources thatmay be available for children in the community”(Coulton et al., 1995, p. 1270).When children outnumber adults in areasand there is a lack of natural support networks (i.e. elderly residents),that child care burdenmay become stressful for parents. Child care bur-den suggests a breakdown in the structure of helping networks in aneighborhood where parents have few choices for help when it comesto caregiving as well as possible reservations about children being ableto play freely in the neighborhood. If there are no neighborhood senti-nels in the form of adult or elderly residents who can act as de facto

guardians for the children in a neighborhood, the result is an increasein stress as parents and children are either in constant contact, childrenare left alone more frequently without adequate adult supervision, orparents must travel outside their community to obtain competentchild care, incurring both financial and human costs.

Coulton and her colleagues operationalized child care burden byusing a factor score representing the indicators of percent elderly,ratio of children to adults, and the ratio of males to females. Using thisapproach, child care burdenwas found to be positively related to overallrates of child maltreatment (Coulton et al., 1999; Coulton et al., 1995).When investigating child maltreatment reports separated by race,only predominately white neighborhoods and substantiation rates forwhite children were found to have a positive relationship to child careburden (Freisthler, Bruce, et al., 2007; Korbin et al., 1998). Child careburden has been less consistent in predicting rates of child maltreat-ment, suggesting it may operate differently for different forms of childmaltreatment and among ethnic and racial groups.

This body of research has been able to definewith some consistencythe structural risk factors for child maltreatment, but there has beenlimited attention so far to the structural component of protectionagainst child maltreatment. Klein (2011) found the availability anduse of early childhood education services within neighborhoods was amitigating factor for child maltreatment among children aged0–5 years old when controlling for the neighborhood risk factors men-tioned above. This paper investigates the possibility that access to sub-stance abuse services in a community could serve as a mitigatingfactor for child maltreatment.

1.3. Substance abuse services and child welfare outcomes

Substance abuse service facilities are not often structured in a waythat ensures participation of impoverished clients due to their inacces-sibility (i.e. not convenient to public transportation) (Semidei, Radel,& Nolan, 2001). Jacobson (2004) referred to this problem as the travelburden, or the difficulty experienced when neighborhood geographyplaces an extreme distance between one's home locale and the servicefacility. If it is difficult to access services, individuals may be less likelyto engage with service providers to start treatment for substanceabuse problems or continue in treatment when the travel burden out-strips an individual's tolerance for the travel costs incurred. However,when clients involved in child welfare services are able to access sub-stance abuse services, the permanency outcomes have been shown tobe largely positive, with completion of treatment associated with ashorter tenure in foster care for children and greater odds ofreunification (Choi, Huang, & Ryan, 2012; Green, Rockhill, & Furrer,2007). Additionally, Marsh, D'Aunno, and Smith (2000) found motherswith children in foster care who were provided transportation to over-come the geographic inaccessibility of substance abuse services weremore likely to refrain from substance use than mothers who did not re-ceive transportation services. While the studies referenced above werecompleted at the individual level, several studies offer additional sup-port that accessibility to other social services could be an important fac-tor in reducing rates of child maltreatment.

Bai, Wells, and Hillemeier (2009) investigatedmental health serviceutilization among child welfare involved children with mental healthservice needs and found the number of mental health practitionersper 100,000 childrenwere positively related to the use of mental healthservices, but the number of mental health clinics per 100,000 childrenwas negatively associated with the use of these services. The negativefindings in the Bai et al. (2009) county-level study could be developedand clarified by including indicators of how the location of servicesare geographically distributed. Freisthler (2013) investigated the rela-tionship between child maltreatment referrals and foster care entriesto the spatial distribution of child welfare-related services. Findingshere indicated a greater overall density of social services predictedlower rates of child maltreatment referrals and foster care entries.

1935C.M. Morton / Children and Youth Services Review 35 (2013) 1933–1940

Interestingly, when social services were separated by type, findings in-dicated the density of substance abuse services was related to higherrates of both referrals and entries (Freisthler, 2013). These paradoxicalfindings were posited to occur due the possibility of substance abuseservice facilities locating in high-need areas and additional work isneeded to further explore this issue in other areas of the United States.

2. Theoretical framework

The social disorganization perspective suggests communities char-acterized by poverty, residential instability, racial and ethnic heteroge-neity or isolation, and disruptions in family life often suffer frombreakdowns in social cohesion and control (Bursik & Grasmick, 2002;Kornhauser, 1978; Shaw &McKay, 1942). The consequences for a com-munity lie in the inability to solve shared problemsbecause of structuralimpediments to the development of coherent, interconnected socialnetworks (Sampson & Groves, 1989). Thus the breakdown of formaland informal social networks restricts the ability of a community to reg-ularly enforce social control in terms of working together to reduce in-terpersonal violence, prohibit an inundation of unhealthy retail, orattract investment in the community. Rates of child maltreatmenthave consistently been found to be higher in neighborhoods that canbe characterized as socially disorganized (Coulton et al., 2007;Freisthler, Merritt & LaScala, 2006). Poor neighborhoods are often struc-tured in a way that concentrates disadvantage, leading to an erosion ofsocial control allowing for not only negative outcomes in terms of inter-personal violence but also the inundation of addictive retail that is notwelcome in more affluent areas. As a result, researchers have consis-tently found alcohol outlets are more likely to be located in areas char-acterized by social disorganization (Gorman, Speer, Gruenewald, &Labouvie, 2001; LaVeist & Wallace, 2000; Popova, Giesbrecht,Bekmuradov, & Patra, 2009). The current study concentrates on neigh-borhood availability of alcohol, its relationship to child maltreatment,and the ability of substance abuse service facilities to act as mitigatingfactors for child maltreatment, while controlling for the socioeconomicrisk factors representative of social disorganization.

3. Study hypotheses

The specific study hypotheses are as follows, based on the empiricaland theoretical review above:

1. Higher densities of alcohol outlets will be relatedwith higher rates ofchild maltreatment, controlling for socioeconomic and demographicfactors.

2. Easier access to substance abuse service facilities will be associatedwith lower rates of child maltreatment.

3. A neighborhood's substance abuse service accessibilitywillmoderatethe relationship between alcohol outlet density and child maltreat-ment, controlling for socioeconomic and demographic factors. Inareas with a higher density of alcohol outlets, child maltreatmentrates will be lower if there is easy access to substance abuse services.

4. Methods

4.1. Data sources

Data for this study were drawn from five sources: (1.) 2003 NewJersey Department of Children and Families (DCF) Bergen Countychild maltreatment referral data, (2.) 2003 New Jersey Division ofAlcoholic Beverage Control (ABC) listing of alcohol-selling retailers,(3.) New Jersey Division of Mental Health and Addiction Services(DMHAS) listing of licensed substance abuse providers active in 2003,(4.) Bergen County Center for Alcohol and Drug Resources' (CADR) list-ing of substance abuse service facilities active in 2003, and (5.) the 2000United States Census. The DCF data provide information for all child

maltreatment referrals to Child Protective Services and the addressesincluded in this data reflect the residential address of the family in-volved in the referral. The ABC listing of alcohol retailers contains infor-mation about the licensee including the address where each outlet islocated. The DMHAS and CADR data combine to create a listing of all li-censed substance abuse service facilities in Bergen County, New Jerseyand contains the address where each facility is located. The addressfor each child maltreatment report, alcohol outlet and substance abuseservice facility was geocoded using ArcMap 10.0, with match rates ex-ceeding 90% for each.

Finally, the Census data were used to create demographic and socio-economic profiles of the neighborhoods in Bergen County, New Jersey.Bergen County is themost populous county in New Jersey with a popu-lation of 884,118 that includes 235,070 families in the year 2000. Resi-dents were 10.3% Hispanic (of any race), 78.4% White, 10.7% Asian,and 5.3% African American. It encompasses 247mile2 and has a popula-tion density of 3776 people per square mile.

This study uses administrative units defined by the U. S. Census Bu-reau to represent neighborhoods, specifically theCensus tractwhichhasanywhere from 1500 to 8000 residents with the average number of res-idents being 3000 (U.S. Census Bureau, 2002). Census tracts are chosenboth to follow the neighborhood effects child maltreatment researchand to provide a unit of aggregation that is relatively homogenous,while at the same time providing enough variability in the rates ofchild maltreatment for valid statistical estimates.

4.2. Measures

4.2.1. Child maltreatmentThe dependent variable for this analysis was the rate of child mal-

treatment referrals per 1000 children in Bergen County, NJ for theyear 2003. Using child maltreatment referrals, instead of substantiatedreports or foster care entries, follows previous research in order to ac-count for the number of cases that may go unreported (Deccio,Horner, & Wilson, 1994; Garbarino & Sherman, 1980). To date, onestudy has compared three operationalizations of child maltreatmentcommonly found in the neighborhood effects literature (CPS referrals,substantiated referrals, and foster care entries). The findings across thethree child maltreatment measures did not conflict with one anotherin terms of finding a positive correlation for one and negative for anoth-er, and the model fit statistics indicated referrals as marginally betterthan substantiation and foster care entry (Freisthler, Gruenewald,Remer, Lery, & Needell, 2007). This study is concernedwith the possibil-ity that increased access to substance abuse services could act as a pri-mary prevention strategy for child maltreatment, meaning therewould in effect be a reduced need for CPS referrals in those neighbor-hoods with easily accessed substance abuse service facilities.

4.2.2. Alcohol outlet densityAlcohol outlet density was measured as the number of alcohol out-

lets (this includes, bars, liquor/beer stores, and restaurants that serve al-cohol) per 10kmof roadwaywhich operationalizes availability based onhow frequently alcohol retailers are encountered in the street networkof a community (Gruenewald & Johnson, 2010). A roadway measurefor Bergen County was favored here over a population based measurebecause of its dense population. Consider the following example: ifthere is an alcohol outlet on the first floor of a large, mixed-use buildingwith 1000 residents, one assumes that all residents have equal access tothe outlet. Using a population density measure would produce a smalldensity value, describing the residents' access to alcohol as relativelylow; however, the roadwaymeasure produces a higher density statistic,indicating easier access to alcohol.

4.2.3. Substance abuse service accessibilityService accessibility wasmeasured as the distance from the centroid

of each census tract to the nearest substance abuse service facility

Table 1Descriptive statistics for rates of child maltreatment, alcohol outlet density, substanceabuse service accessibility, and socioeconomic controls for Bergen County,NJ census tracts,2003 (n=163).

Mean Standard deviation

Child maltreatment ratea 3.97 4.06Alcohol outlet densityb 2.46 2.80Substance abuse service accessc 1.39 .90Socioeconomic controls% in poverty 4.78 3.29% unemployed 3.92 1.96% vacant housing units 2.45 1.38% 5 year residential movement 36.62 8.62% single female-headed households 5.38 3.28Child to adult ratio .31 .07Male to female ratio .89 .06% population over 65 15.09 3.50% African American 4.67 10.92% Asian 10.46 8.58% Latino/Hispanic 10.01 8.56% immigrant population 24.52 11.96

a Measured as number of child maltreatment reports per 1000 children.b Distance in miles from centroid of each census tract to nearest facility.c Measured as number of outlets per 10 km of roadway.

1936 C.M. Morton / Children and Youth Services Review 35 (2013) 1933–1940

(Schonlau et al., 2008). This measure was chosen due to the low occur-rence of these facilities in census tracts in order to assign each censustract an overall measure of accessibility to substance abuse servicefacilities.

4.2.4. Socioeconomic controlsFinally, measures of community structure tested in other studies on

neighborhood effects on child maltreatment and population character-istics unique to Bergen County, NJ were used as control variables(Coulton et al., 1995; Freisthler, 2004). These were taken from the2000 U.S. Census and include: poverty rate, unemployment rate, per-cent vacant housing units, percent of people who moved from1995–2000, percent single female-headed households, child to adultratio, male to female ratio, percent of population over 65 years of age,percent African American, percent Latino/Hispanic, percent Asian, andpercent immigrant population.

4.3. Analytic strategy

The analytic strategy relied on Ordinary Least Squares (OLS) or spa-tial regression to assess the relationship between predictor and criterionvariables. This technique of choosing between two regression tech-niques is most appropriate when considering data that is spatially or-dered as one must address spatial autocorrelation. For this analysis, itmay be that not only are independent observations highly correlated(multicollinearity), but also observations can be influenced by theirproximity in a spatial plane violating the assumption in regressionthat observations are independent. So, measures from neighborhoodsthat share a boundary may be highly correlated, introducing measure-ment error into the statistical test employed as relationships vary as aresult of space (Cahill & Mulligan, 2007; Graif & Sampson, 2009).

To assess the presence of significant spatial autocorrelation thatwould bias statistical estimates, a two-step process was conducted.First, the dependent variable was investigated to determine the leveland statistical significance of spatial autocorrelation using a rookweights correlation matrix. This process produced the Moran's Index(Moran's I) statistic which is interpreted like a correlation coefficient:values range from −1 to 1 indicating neighboring areas are either per-fectly dissimilar (negative spatial autocorrelation) or perfectly similar(positive spatial autocorrelation). Second, it is possible that the arrayof independent variables can explain away the spatial autocorrelationof the dependent variable, if present (Charlton & Fotheringham, 2009;Freisthler, Bruce, et al., 2007; Griffith, 1988). Here, the residuals fromthe OLS regression model were analyzed and, again, Moran's I wasassessed for significance. If the Moran value was significant in both ac-counts, spatial autocorrelation was controlled for in subsequent analy-ses using spatial regression model (Fotheringham & Rogerson, 2009;Freisthler, 2004; Freisthler, Bruce, et al., 2007; Freisthler et al., 2005).

Another concern is the small area analysis problem(hetereoskedasticity) where areas with exceedingly small populationsare given the same weight in the regression equation as areas withlarge populations. To control for this, each unit of observation wasweighted by the square root of the child population for that area(Freisthler, 2004; Freisthler et al., 2005; Klein, 2011). This study usedArcGIS 10 and SPSS 19.0 to test the relationship between predictorand criterion variables, controlling for spatial autocorrelation, wherepresent, and hetereoskedasticity.

The moderating effect of substance abuse service accessibility wastested following Baron and Kenney's (1986) conceptualization. Themoderator model was tested by entering an interaction term in theregression equation, here the product of the alcohol outlet densitymeasures and the substance abuse service accessibilitymeasure. Hierar-chical regressionwas employed to lend support to themoderatormodelby examining the significance of the R2 and F change in themodel as theinteraction between alcohol outlet density and substance abuse serviceaccessibility was stepped into the model. To aid in the visualization of

the moderator effect, post hoc analysis of covariance (ANCOVA) wasperformed to illustrate the relationship between alcohol outlet densityand childmaltreatment at different levels of the substance abuse serviceaccessibility, controlling for the sociodemographic profile of neighbor-hoods. This follows other public health research investigating modera-tor effects (Lachman & Weaver, 1998; Peterson, Lowe, Peterson, &Janz, 2006).

4.3.1. Criterion variable transformationThe criterion variable of child maltreatment rate had an extreme

positive skew with a skewness statistic of 2.92. In order to correct forthis skew, a Box Cox transformation was applied to rates of child mal-treatment. This transformation is a power transformation, where eachobservation is raised to some power in order to achieve normality(Osborne, 2010). Box Cox transformations are different from otherpower-family transformations (e.g., log, square root transformations)in that, rather than picking an arbitrary value to use as the power toraise all values by, one is presented with a range of power transforma-tions in order to choose the value that best normalizes the data(Allison, Gorman, & Kucera, 1995). These values are known as lambda(λ) and the value of λ that best normalized the data was chosen asthe transformation power which eliminated skew. Using the Box Coxtransformation, the rate of childmaltreatmentwas normalized, improv-ing skew to .16where λ=.10. The transformed variable was used in theregression equation.

4.3.2. Principal Components AnalysisPrincipal Components Analysis (PCA) with varimax rotation was

conducted on the set of 12 socioeconomic variables as a data reductionstrategy to protect against the issue of multicollinearity. This followsprevious research in the area of child maltreatment neighborhood ef-fects research and allows for the identification of the underlying struc-ture of the data (Coulton et al., 1995; Ernst, 2001; Korbin et al., 1998).Results from the PCA are presented in Table 2.

PCA on the 163 census tracts revealed three componentswith eigen-values greater than 1, explaining 68% of the variance in the set of socio-economic variables. Seven variables loaded onto the first component,PC1: poverty rate, immigrant population, 5-year residential movement,percent Latino, vacant housing units, and child to adult ratio. Here, dis-advantage seems concentrated in certain areas with the economic indi-cator of poverty status combining with residential movement andvacant housing to suggest capital disinvestment. Additionally, ethnicminorities and immigrant populations loaded onto this component.

Table 2Principal Components Analysis of socioeconomic controls.

Variable PC1 PC2 PC3

% immigrant population .909 .032 .122% poverty .763 .380 .136% 5-year residential movement .787 .246 .214Child to adult ratio −.705 −.131 .420% Latino/Hispanic .596 .384 .301% vacant housing units .576 .202 −.030% African American .087 .765 −.049% single female headed households .492 .701 .192% unemployment .425 .678 .146% Asian .599 −.566 .056% over 65 years of age .059 −.234 −.846Male to female ratio .183 −.066 .742

Note: N=163. These 3 factors explained 68% of the variance in the set of variables.* FactorLoadings above .500.

1937C.M. Morton / Children and Youth Services Review 35 (2013) 1933–1940

The second component, PC2, was characterized by high percentages ofAfrican American residents, single female-headed households, unem-ployment, and lowpercentages of Asian residents. PC2 hints at econom-ic disadvantage as well, but is differentiated from the first by beingpredominately African American with high percentages of singlefemale-headed households. Lastly, two variables loaded onto the thirdcomponent, PC3, where census tracts were characterized by low per-centages of residents over 65 years of age and a high male to femaleratio. The factor scores from the PCA were used in the regressionequation.

5. Results

5.1. Descriptive statistics

Descriptive statistics for each variable included in the analysis arepresented in Table 1. The mean overall child maltreatment rate for Ber-gen County, NJ was 3.97/1000 which is a bit higher than the child mal-treatment rate of 3.9/1000 for New Jersey in 2003 (U.S. Department ofHealth and Human Services, Administration on Children, Youth andFamilies, 2009). Mean alcohol outlet density was 2.46 outlets per10kmof roadway and themean community substance abuse service fa-cility distance was 1.39 miles to the nearest facility from each censustract's centroid.

Table 3OLS regression model of child maltreatment rate, alcohol outlet density, substance abuse servi

Variables

Spatial autocorrelation (Moran's I)Socioeconomics

Component 1: Social disorganizationComponent 2: Predominately African AmericanComponent 3: Young male

Alcohol densityOutlet density per 10km

Substance abuse servicesDistance to nearest facility

InteractionAlcohol outlet density × distance to nearestfacility

Model 1 R2= .41, Model 2 R2= .42.F change between Models 1 and 2 was 79.92⁎⁎⁎.⁎ p b .05.⁎⁎ p b .01.⁎⁎⁎ p b .001.

5.2. Effect of alcohol outlet density and substance abuse service accessibilityon child maltreatment rates

Table 3 presents the OLS results where the three components ofneighborhood structure, the number of alcohol outlets per 10 km ofroadway, and substance abuse service accessibility were all significantlypositively related to higher rates of child maltreatment. The results of ahierarchical regression to support themoderator model for the interac-tion of alcohol outlet density and substance abuse service accessibilityindicate significant R2 and F change statistics when the interactiontermwas stepped into the regression equation.While first order spatialautocorrelationwas positive and significant (MI=.220, pb .001), the setof independent variables explained away the significant autocorrelationin the regression model where the Moran's I for the residuals from theOLS were negative and non-significant (MI=-.057, p= .279).

Post hoc tests of the moderator effect using ANCOVA wereperformed to further explore the interaction effect for illustration pur-poses only. For the post hoc analysis, alcohol outlet density was recodedinto tertile groups: low, medium, and high density. Similarly, substanceabuse service accessibility was recoded into short, medium, and longdistance categories. As in the regression analysis, the factor scoresfrom the principal components analysis were used as controls/covari-ates. Fig. 1 illustrates the post hoc analysis. In areas with low accessibil-ity (longer distance to nearest substance abuse service facility), asignificant difference in mean child maltreatment rates was found be-tween tracts with low and high alcohol outlet density. In areas withthe highest alcohol outlet density, child maltreatment rates weregreatest in areas with long distance to substance abuse services.

6. Discussion

It was hypothesized that neighborhoods with greater alcohol outletdensities would have correspondingly high rates of child maltreatment.Controlling for socioeconomic and demographic factors, this was foundto be true. Those areas inundated with alcohol outlets had higher ratesof child maltreatment. A greater distance to the nearest substanceabuse service facility was related to higher rates of child maltreatmentas a main effect. Additionally, it was hypothesized that access to sub-stance abuse service facilitieswouldmoderate the relationship betweenalcohol outlet density/accessibility and rates of childmaltreatment. Spe-cifically, child maltreatment rates would be higher in neighborhoods

ce facility access, and socioeconomic controls.

Dependent variable: child maltreatment rate per 1000 children

Model 1Socioeconomics, alcohol,and prevention access

Model 2+Interaction effect

−.059 −.057⁎⁎

B se B se.288⁎⁎⁎ .010 .294⁎⁎⁎ .010.383⁎⁎⁎ .009 .376⁎⁎⁎ .009.070⁎⁎⁎ .009 .067⁎⁎⁎ .009

.022⁎⁎⁎ .004 .022⁎⁎⁎ .007

.020⁎ .010 .093⁎⁎⁎ .013

−.026⁎⁎⁎ .003

Fig. 1. Interaction of alcohol outlet density and substance abuse service facility accessibilityon rates of child maltreatment.

1938 C.M. Morton / Children and Youth Services Review 35 (2013) 1933–1940

with greater alcohol outlet density and less accessible substance abuseservice facilities. The interaction between alcohol outlet density andsubstance abuse service access was significant. Neighborhoods withthe greatest distance to substance abuse service facilities and thehighest alcohol outlet density had the highest rates of child maltreat-ment. Interestingly, neighborhoods with what was defined as mediumaccess to substance abuse services had the lowest rates of child mal-treatment. This finding and may be indicative of desirable neighbor-hoods being those that are not too close to commercial activity butnot too far from it either to be cut off from adjacent communityresources.

Overall, the strongest predictors of increased rates of child maltreat-ment were the measures of community structure found in previouswork in this area (Coulton et al., 2007; Freisthler et al., 2006). Resultsshow predominately African American census tracts with high unem-ployment and single female-headedhouseholdswere the strongest pre-dictor of increased rates of child maltreatment, which mirrors much ofthe literature on the racial composition of the child welfare system(Drake et al., 2011; Fluke, Yuan, Hedderson, & Curtis, 2003; Morton,Ocasio, & Simmel, 2011). Census tracts characterized by poverty andresidential instability were the second strongest predictor of increasedrates of childmaltreatment. Again, this finding is confirmed by prior re-search; living in poverty and its attendant consequences is a consistentpredictor of child maltreatment (Coulton et al., 2007; Freisthler et al.,2006). Lastly, neighborhoods characterized by a great number of maleand younger residents were connected with increased rates of childmaltreatment. Thismay be a novel finding in terms of the neighborhoodeffects literature, but it does find support elsewhere. Some studies havefound perpetrators of severe physical abuse to be predominately male(Naidoo, 2000; Ricci, Giantris, Merriam, Hodge, & Doyle, 2003). Thiscould be an important finding in terms of needs assessment and pro-gram development to reduce the risk of child maltreatment.

6.1. Limitations

This study is limited by its reliance on secondary data as well as itscross-sectional design. The child maltreatment indicators used were re-ports of child abuse and neglect. As such, they may not be indicative ofwhat the true rates of child maltreatment are in the areas under inves-tigation, they only include those families who were reported to CPS.Measuring childmaltreatment has always beendifficult and researchersare developing innovative ways to couple different sources of adminis-trative data to compute child maltreatment rates that do not rely exclu-sively on CPS reports, retrospective surveys, or community sentinels.

Recent work has combined child maltreatment reports with child fatal-ity data to develop indices of child maltreatment (Putnam-Hornstein,2011). Future research should continue to explore ways to presentrates of child maltreatment that attempt to account for its hidden na-ture. Additionally, the data for this study were drawn from 2003; futurework could compare thesefindingswithmore recent data to account forany changes in the landscape of child welfare practice in New Jersey aswell as expand to include data for the entire state.

This study treated all alcohol retailers the same in that risk was con-sidered to be distributed equally among the set of outlets. Itmay be thatthere are alcohol outlets in a neighborhood that are considered especial-ly problematic in terms of crime or property damage occurring in andaround them, compared to outlets that have little or no collateral dam-age associated with their location. Future research could identify prob-lem alcohol outlets by connecting rates of alcohol-related crimeoccurring in close proximity and investigating whether a difference ex-ists between good and bad outlets in terms of their relationship to childmaltreatment. This study relied on the physical location of alcohol re-tailers and substance abuse services facilities to make connections topopulation-level rates of child maltreatment. The indicators of alcoholoutlet density and substance abuse service facilities do not necessarilymean residents are purchasing and abusing alcohol or accessing the ser-vices of substance abuse professionals. Future work could utilize multi-levelmodeling that couples survey data on individual drinking behaviorand service use with community indicators of child maltreatment, alco-hol access, and access to substance abuse service facilities. Research thatcan link individual behavior to increased access to either alcohol or sub-stance abuse service facilities will help close the gap in linking these at-tributes of the built environment to rates of child maltreatment. Theinability to determine causality is also a limitation of this study. It is im-possible to knowwhether drinking behavior was influenced by the easyaccess to alcohol or if those with alcohol problems are drawn to live inareas with easy access to alcohol. Accordingly, it is not known whetherincreased child maltreatment rates were caused by increased alcoholaccessibility, conducting time-series designs that track the changing al-cohol retail environment along with changes in rates of child maltreat-ment would be necessary to add weight to any causality argument.

6.2. Implications for child welfare policy and practice

The results from this study support a primary prevention approachto reducing child maltreatment through environmentally-focused in-terventions. This study adds to the existing literature confirming the im-portance of neighborhood structure in influencing rates of childmaltreatment. This is not only limited to the socioeconomic and demo-graphic profile of an area, but also the study adds to evidence of the im-portance of the built environment, both for risk and protection.Prevention strategies could investigate the possibility of limiting alcohollicenses in areas that are already saturated with alcohol retailers as away to reduce the harms associated with alcohol use. Restricting thenumber or density of alcohol retail licenses is seen as a promising areaand others have made the call to target alcohol outlet density as anenvironmentally-focused prevention goal (Campbell et al., 2007). As aresult, the implementation of policy changes aimed at zoning and landuse regulations to limit the physical availability of alcohol are seen assome of the most effective interventions to reduce alcohol-relatedharms (Sparks, Jernigan, &Mosher, 2011). These strategies have limitedthe total number of retail licenses for alcohol, limited the days and hoursalcohol is sold or increased the price of alcohol through taxation. Childwelfare agencies could lend considerable support to substance abuseprevention agencies in terms of leveraging support for this type of pol-icy change. Connecting the reduction of alcohol-related harms to theprevention of child abuse could open policywindows to limit the densi-ty of alcohol retailers in ways that substance abuse prevention agenciesworking alone could not.

1939C.M. Morton / Children and Youth Services Review 35 (2013) 1933–1940

Additionally, the significant findings for substance abuse service facil-ity access and the significant moderation effects also support a primaryprevention approach in terms of democratizing access to substanceabuse treatment. This may be an important protective factor againstchild maltreatment. Areas with easier access to substance abuse servicesnot only had lower rates of child maltreatment but also impacted the re-lationship between alcohol outlet density and overall childmaltreatmentand neglect. This type of strategywould not focus on those at risk of com-mitting child abuse or those who had already become involved with thechild welfare system. Rather, increasing access to substance abuse ser-vices targets the population as a whole to see change across systems.

For child welfare clients, reducing the logistical barriers to substanceabuse treatment has been shown to increase the likelihood of positivepermanency outcomes (Marsh et al., 2000; Rockhill, Green, & Newton-Curtis, 2008). Attention should be paid to the location of human servicefacilities as both a functional benefit and an ethical imperative.Impoverished communities are often over-looked and segregatedfrom resources such as substance abuse services (Massey, 2004), loca-tional strategies for these facilities should consider how to best diffusethe network of locations to reach communities in need. Community as-sessments could utilize GIS in order to locate areas considered treat-ment deserts and those historically segregated from quality servicesdue to economic or racial attributes and engage community leaders inplanning any extensions of service.

Primary prevention and especially environmentally-focused pre-vention are not familiar waters for the child welfare field. Its focus hastraditionally been on providing services to families and not on organiz-ing communities in a way that benefits child well-being. The ability toshift some resources from providing services to investing in communi-ties in a way that builds capacity would take considerable partnershipbetween governmental departments in terms of linking resources be-tween health and human services, child welfare, mental health, addic-tion services, and alcoholic beverage control. At the federal level, theUnited States Department of Health and Human Services created theNational Center on Substance Abuse and Child Welfare (NCSACW)with funding from the Substance Abuse and Mental Health ServicesAgency and the Administration on Children, Youth and Families(USDHHS, n.d.). This multi-sector initiative is meant to support state-level collaborations between systems of substance abuse services,child welfare, and family courts by providing technical assistance forcollaborative efforts. This type of collaboration is crucial as innovativeand environmentally-focused strategies aimed at the primary preven-tion of child maltreatment would require a critical mass of support tobuild communities that facilitate family and child well-being.

Acknowledgment

The first authorwas supported as a postdoctoral fellow in the Behav-ioral Sciences Training Program in Drug Abuse Research sponsored byPublic Health Solutions and National Development and Research Insti-tutes with funding from the National Institute on Drug Abuse [grantnumber T32 DA007233]. Points of view, opinions, and conclusions inthis paper do not necessarily represent the official position of the U.S.Government, Public Health Solutions, National Development and Re-search Institutes, or the State of New Jersey Department of Childrenand Families.

References

Allison, D. B., Gorman, B.S., & Kucera, E. M. (1995). Unicorn: A program for transformingdata to approximate normality. Educational and Psychological Measurement, 55,625–629.

Bai, Y., Wells, R., & Hillemeier, M. M. (2009). Coordination between child welfare agenciesand mental health service providers, children's service use and outcomes. Child Abuse& Neglect, 33(6), 372–381.

Banks, H., & Boehm, S. (2001). Substance abuse and child abuse. Children's Voice(Retrieved from http://www.cwla.org/articles/cv0109sacm.htm)

Baron, K. E., & Kenney, D. A. (1986). Themoderator–mediator variable distinction in socialpsychological research: Conceptual, strategic, and statistical considerations. Journal ofPersonality and Social Psychology, 51, 1173–1182.

Besinger, B.A., Garland, A. F., Litronwick, A. J., & Landsverk, J. A. (1999). Caregiver sub-stance abuse among maltreated children place in out-of-home care. Child Welfare,78(2), 221–239.

Blankenship, K. M., Friedman, S. R., Dworkin, S., & Mantell, J. E. (2006). Structural inter-ventions: Concepts, challenges, and opportunities for research. Journal of UrbanHealth, 83(1), 59–72.

Bursik, R. J., & Grasmick, H. G. (2002). Neighborhoods and crime: The dimensions of effectivecommunity control. New York: Lexington Books.

Cahill, M., & Mulligan, G. (2007). Using geographically weighted regression to explorelocal crime patterns. Social Science Computer Review, 25(2), 174–193.

Campbell, C. A., Hahn, R. A., Elder, R., Brewer, R., Chattopadhyay, S., Fielding, J., et al.(2007). The effectiveness of limiting alcohol outlet density as a means of reducingexcessive alcohol consumption and alcohol-related harms. American Journal ofPreventive Medicine, 37(6), 556–569.

Charlton, M., & Fotheringham, A. S. (2009). Geographically weighted regression (white paper).: National Univesity of Ireland, Maynooth: National Centre for Geocomputation.

Choi, S., Huang, H., & Ryan, J. P. (2012). Substance abuse treatment completion in childwelfare: Does substance abuse treatment completion matter in the decision toreunify families? Children and Youth Services Review, 34(9), 1639–1645.

Coulton, C. J., Crampton, D. S., Irwin, M., Spilsbury, J. C., & Korbin, J. E. (2007). How neigh-borhoods influence child maltreatment: A review of the literature and alternativepathways. Child Abuse & Neglect, 31, 1117–1142.

Coulton, C. J., Korbin, J. E., & Su, M. (1999). Neighborhoods and child maltreatment: Amulti-level study. Child Abuse & Neglect, 23(11), 1019–1040.

Coulton, C. J., Korbin, J. E., Su, M., & Chow, J. (1995). Community level factors and childmaltreatment rates. Child Development, 66, 1262–1276.

Daro, D., & Donnelly, A.C. (2002). Charting the waves of prevention: Two steps forward,one step back. Child Abuse & Neglect, 26, 731–742.

Deccio, G., Horner, W. C., & Wilson, D. (1994). High-risk neighborhoods and high-riskfamilies: Replication research related to the human ecology of child maltreatment.Journal of Social Science Research, 18(3/4), 123–137.

Drake, B., Jolley, J. M., Lanier, P., Fluke, J., Barth, R., & Jonson-Reid, M. (2011). Racial bias inchild protection? A comparison of competing explanations using national data.Pediatrics, 127(3), 471–478.

Drake, B., & Pandey, S. (1996). Understanding the relationship between neighborhoodpoverty and specific types of child maltreatment. Child Abuse and Neglect, 20(11),1003–1018.

Ernst, J. S. (2001). Community-level factors and child maltreatment in a suburban county.Social Work Research, 25(3), 133–142.

Fluke, J., Yuan, Y., Hedderson, J., & Curtis, P. A. (2003). Disproportionate representation ofrace and ethnicity in child maltreatment: Investigation and victimization. Childrenand Youth Services Review, 25(5–6), 359–373.

Fotheringham, A. S., & Rogerson, P. A. (Eds.). (2009). The SAGE handbook of spatial analysis.London: Sage Publications, Ltd.

Freisthler, B. (2004). A spatial analysis of social disorganization, alcohol access, and ratesof child maltreatment in neighborhoods. Children and Youth Services Review, 26,803–819.

Freisthler, B. (2013). Need for and access to supportive services in the Child Welfare Sys-tem. GeoJournal, 78(3), 429–441.

Freisthler, B., Bruce, E., & Needell, B. (2007). Understanding the geospatial relationship ofneighborhood characteristics and rates of maltreatment for black, Hispanic, andwhite children. Social Work, 52(1), 7–16.

Freisthler, B., Gruenewald, P. J., Remer, L. G., Lery, B., & Needell, B. (2007). Exploring thespatial dynamics of alcohol outlets and child protective service referrals, substantia-tions, and foster care entries. Child Maltreatment, 12(2), 114–124.

Freisthler, B., Merritt, D. H., & LaScala, E. A. (2006). Understanding the ecology of childmaltreatment: A review of the literature and directions for future research. ChildMaltreatment, 11(3), 263–280.

Freisthler, B., Midanik, L. T., & Gruenewald, P. J. (2004). Alcohol outlets and child physicalabuse and neglect: Applying routine activities theory to the study of child maltreat-ment. Journal of Studies on Alcohol, 65(5), 586–592.

Freisthler, B., Needell, B., & Gruenewald, P. J. (2005). Is the physical availability of alcoholand illicit drugs related to neighborhood rate of child maltreatment? Child Abuse &Neglect, 29, 1049–1060.

Garbarino, J., & Sherman, D. (1980). High-risk neighborhoods and high-risk families: Thehuman ecology of child maltreatment. Child Development, 51(1), 188–198.

Gorman, D.M., Speer, P. W., Gruenewald, P. J., & Labouvie, E. W. (2001). Spatial dynamicsof alcohol availability, neighborhood structure and violent crime. Journal of Studies onAlcohol, 62(5), 628–636.

Graif, C., & Sampson, R. J. (2009). Spatial heterogeneity in the effects of immigration anddiversity on neighborhood homicide rates. Homicide Studies, 13(3), 242–260.

Green, B.L., Rockhill, A., & Furrer, C. (2007). Does substance abuse treatment make a dif-ference for child welfare case outcomes? A statewide longitudinal analysis. Childrenand Youth Services Review, 29, 460–473.

Griffith, D. A. (1988). Advanced spatial statistics. Norwell, MA: Kluwer AcademicPublishers.

Gruenewald, P. J., & Johnson, F. W. (2010). Drinking, driving, and crashing: A traffic-flowmodel of alcohol-related motor vehicle accidents. Journal of Studies on Alcohol andDrugs, 237–248.

Jacobson, J. O. (2004). Place and attrition from substance abuse treatment. Journal of DrugIssues, 34(1), 23–50.

Klein, S. (2011). The availability of neighborhood early care and education resources andthe maltreatment of young children. Child Maltreatment, 16(4), 300–311.

1940 C.M. Morton / Children and Youth Services Review 35 (2013) 1933–1940

Korbin, J. E., Coulton, C. J., Chard, S., Platt-Houston, C., & Su, M. (1998). Impoverishmentand child maltreatment in African American and European American neighborhoods.Development and Psychopathology, 10(2), 215–233.

Kornhauser, R. R. (1978). Social source of delinquency. Chicago: University of ChicagoPress.

Lachman, M. E., & Weaver, S. L. (1998). The sense of control as a moderator of social classdifferences in health and well-being. Journal of Personality and Social Psychology,74(3), 763–773.

LaVeist, T. A., & Wallace, J. M. (2000). Health risk and unequal distribution of liquor storesin African American neighborhoods. Social Science and Medicine, 51(4), 613–617.

Marsh, J. C., D'Aunno, T. A., & Smith, B.D. (2000). Increasing access and providing socialservices to improve drug abuse treatment for women with children. Addiction, 95,1237–1247.

Massey, D. S. (2004). Segregation and stratification: A biosocial perspective. DuBoisReview, 1(1), 7–25.

Molnar, B. E., Buka, S. L., Brennan, R. T., Holton, J. K., & Earls, F. (2003). A multilevel studyof neighborhoods and parent-to-child aggression: Results from the project on humandevelopment in Chicago neighborhoods. Child Maltreatment, 8(2), 84–97.

Morton, C. M., Ocasio, K., & Simmel, C. (2011). Critiquing methods used to describedisproportionality. Children and Youth Services Review, 33, 1538–1542.

Naidoo, S. (2000). A profile of the oro-facial injuries in child physical abuse at a localhospital. Child Abuse & Neglect, 24(4), 521–534.

Osborne, J. W. (2010). Improving your data transformations: Applying the Box-Cox trans-formation. Practical Assessment, Research & Evaluation, 15, 1–9.

Paulsen, D. J. (2003). No safe place: Assessing spatial patterns of child maltreatmentvictimization. Journal of Aggression, Maltreatment & Trauma, 8(1/2), 63–85.

Peterson, J. J., Lowe, J. B., Peterson, N. A., & Janz, K. F. (2006). The relationship betweenactive living and health-rated quality of life: Income as a moderator. HealthEducation Research, 21(1), 146–156.

Popova, S., Giesbrecht, N., Bekmuradov, D., & Patra, J. (2009). Hours and days of sale ofalcohol outlets: Impacts on consumption and damage: A systematic review. Alcohol& Alcoholism, 44(5), 500–516.

Putnam-Hornstein, E. (2011). Preventable injury deaths: A population-based proxy ofchild maltreatment risk in California. Public Health Report, 127(2), 163–172.

Reynolds, A. J., Mathieson, L. C., & Topitzes, J. W. (2009). Do early childhood interventionsprevent child maltreatment? A review of research. Child Maltreatment, 14, 182–206.

Ricci, L., Giantris, A., Merriam, P., Hodge, S., & Doyle, T. (2003). Abusive head trauma inMaine infants: medical, child protective, and law enforcement analysis. Child Abuse& Neglect, 27(3), 283–287.

Rockhill, A., Green, B.L., & Newton-Curtis, L. (2008). Accessing substance abuse treatment:Issues for parents involved with child welfare services. Child Welfare, 87, 63–93.

Sampson, R. J., & Groves,W. B. (1989). Community structure and crime: Testing social dis-organization theory. American Journal of Sociology, 94, 774–802.

Schonlau, M., Scribner, R., Farley, T. A., Theall, K. P., Bluthenthal, R. N., Scott, M., et al.(2008). Alcohol outlet density and alcohol consumption in Los Angeles county andsouthern Louisiana. Geospatial Health, 3(1), 91–101.

Semidei, J., Radel, L. F., & Nolan, C. (2001). Substance abuse and child welfare: Clear link-ages and promising responses. Child Welfare, 80(2), 109–128.

Shaw, C., & McKay, H. D. (1942). Juvenile delinquency and urban areas. Chicago: Universityof Chicago Press.

Sparks, M., Jernigan, D. H., & Mosher, J. F. (2011). Regulating alcohol outlet density: Anaction guide. : Community Anti-Drug Coalitions of America (Retrieved from: http://www.cadca.org/resources/detail/strategizer-55%E2%80%94regulating-alcohol-outlet-density-action-guide)

Stagner, M. W., & Lansing, J. (2009). Progress towards a prevention perspective. TheFuture of Children, 19(2), 19–38.

U.S. Census Bureau (2002). Census 2000 Basics.Washington DC: U.S. Government PrintingOffice.

U.S. Department of Health and Human Services, & Administration on Children, Youth andFamilies (2009). Child maltreatment 2007.Washington, DC: U.S. Government PrintingOffice.

U.S. Department of Health and Human Services (n.d). National Center on SubstanceAbuse and Child Welfare: About us. Retrieved from: www.ncsacw.samhsa.gov/aboutus/default.aspx

Yacoubian, G. C. (2007). Assessing environmental prevention strategies for reducing the prev-alence and overall harm of methamphetamine use. Journal of Drug Education, 37, 31–53.

Young, N. K., Boles, S. M., & Otero, C. (2007). Parental substance use disorders and childmaltreatment: Overlap, gaps, and opportunities. Child Maltreatment, 12(2), 137–149.