ilable at ScienceDirect

Social Science & Medicine 67 (2008) 23–30

Contents lists ava

Social Science & Medicine

journal homepage: www.elsevier .com/locate/socscimed

Pathways to depression: The impact of neighborhood violentcrime on inner-city residents in Baltimore, Maryland, USAq

Aaron Curry b, Carl Latkin a, Melissa Davey-Rothwell a,*

a Johns Hopkins Bloomberg School of Public Health, Health, Behaviour & Society, 1629 East Baltimore Street, Baltimore, MD 21201, USAb Surveillance Data Incorporated, Philadelphia, PA, USA

a r t i c l e i n f o

Article history:Available online 8 April 2008

Keywords:USANeighborhoodViolenceDepressionCrimeUrbanInner-cityDrug users

q This work was funded by the National InstituteR01DA016555.

* Corresponding author. Tel.: þ1 410 502 5368; faE-mail addresses: aaron_curry@yahoo.com (

jhsph.edu (C. Latkin), mdavey@jhsph.edu (M. Davey

0277-9536/$ – see front matter � 2008 Elsevier Ltddoi:10.1016/j.socscimed.2008.03.007

a b s t r a c t

Crime and neighborhood disorder may negatively impact the health of urban residents.Neighborhoods with high levels of violent crime may also increase residents’ risk ofexperiencing violence. Most studies supporting the assertion that neighborhood disorderimpacts mental health have used residents’ own ratings of their neighborhoods. The pres-ent study examines the relationships among block-group level crime, perceived neighbor-hood disorder, violence experienced in the neighborhood, and depression. The samplecomprising the current and former drug users (n¼ 786) nested in 270 block groups withinBaltimore, Maryland, USA. Using path analysis, we tested the hypothesis that neighbor-hood violent crime has a direct impact on experiences of violence. Also, we hypothesizedthat neighborhood violence had a direct and indirect impact on depressive symptoms. Re-sults support a model in which violence is associated with psychological distress throughperceptions of neighborhood disorder, and through experiences of violence. We concludethat community and structural level interventions are needed to decrease neighborhoodcrime and improve residents’ perception of their neighborhood.

� 2008 Elsevier Ltd. All rights reserved.

Introduction

Depression is a major health problem in itself and hasbeen prospectively linked to cardiovascular disease andother serious morbidities (Glassman & Shapiro, 1998; Jiang,Krishnan, & O’Connor, 2002; MacMahon & Lip, 2002;Maddock & Pariante, 2001). In addition, depression isstrongly patterned by socioeconomic status (Inaba et al.,2005; Kessler, 1979; McLeod & Kessler, 1990). Althoughthe majority of studies focusing on correlates, conse-quences, and causes of depression have largely focusedon individual, family, and social network factors, some

on Drug Abuse, grant

x: þ1 410 502 5385.A. Curry), clatkin@-Rothwell).

. All rights reserved.

researchers have evaluated the role of structural and con-textual factors such as neighborhood.

Studies conducted in Canada, the United Kingdom, andthe US have examined how neighborhood problems, suchas place of residence and exposure to violent crime and vacanthousing, might contribute to mental health (Bogat et al.,2005; Dupere & Perkins, 2007; Fauth, Leventhal, &Brooks-Gunn, 2004; Galea et al., 2007; Goldsmith, Holzer,& Manderscheid, 1998; Propper et al., 2005; Ross, 2000).Neighborhood conditions in these studies are usuallyviewed as chronic stressors, producing psychological dis-tress (Avison & Turner, 1988; Matheson et al., 2006; Steptoe& Feldman, 2001). However, it is still unclear how thesechronic stressors operate to produce psychological distress.Do these conditions impact individual perceptions of envi-ronment, leading to feelings of fear, anxiety, or hopeless-ness? Or, do violent neighborhoods impact psychologicaldistress by increasing individual exposure to actualviolence?

A. Curry et al. / Social Science & Medicine 67 (2008) 23–3024

After screening 8562 studies, Truong and Ma (2006)conducted a systematic review of 29 studies that examinedthe relationship between neighborhood factors and mentalhealth. The authors included studies that assessed neigh-borhood factors through objective measures as well as sub-jective measures. Objective measures refer to empiricaldata collected at the group level, rather than from an indi-vidual, such as number of dwellings, unemployment rates,crime rates, and income levels. Subjective measures refer toself-reported data collected from individuals which assessperceptions, attitudes, and personal experiences withinthe neighborhood. The authors concluded that there wasa consistent positive relationship between neighborhoodcharacteristics and mental health. However, this associa-tion tended to be modest, especially after adjusting for in-dividual-level factors. This review indicates that bothsubjective and objectives measures of neighborhood fac-tors are associated with mental health.

Researchers have demonstrated a consistent link be-tween self-reported perceptions of one’s neighborhood,a subjective measure, and physical and mental health. Ina sample of Caucasians and African Americans living inBaltimore, MD, Gary, Stark, and LaVeist (2007) found thatindividuals who perceived that their neighborhood hadmore severe problems (e.g. physical, social, and criminalproblems) were more likely to experience higher levels ofstress and depression. Similarly, Latkin and Curry (2003)found that baseline perceptions of one’s neighborhoodproblems predicted higher rates of depression at a follow-up assessment in Baltimore, MD.

In an Australian sample, Ziersch, Baum, Macdougall, andPutland (2005) evaluated the impact of a variety of per-ceived neighborhood characteristics including perceivedsafety, perception of the physical environment, neighbor-hood trust, neighborhood connections, and social capitalon physical and mental health. These researchers foundthat perceived neighborhood safety was associated withphysical health. In addition, mental health was related toperceived safety and neighborhood connections.

There is a growing body of literature assessing theimpact of objective measures of neighborhood on healthstatus. Research conducted among 4.5 million Swedeshas shown associations between neighborhood level fac-tors, including proportion of residents with low incomeand low social capital, and mental health hospitalizationsand disorders (Lofors & Sundquist, 2007; Sundquist &Ahlen, 2006). Likewise, Silver, Mulvey, and Swanson(2002) reported that neighborhood disadvantage andmobility were associated with higher rates of depressionand substance abuse in a large community sample of U.S.residents.

Many researchers have posited that objective measuresof neighborhood influence physical and mental healththrough mediating variables such as perceptions of one’sneighborhood. Cutrona, Wallace, and Wesner (2006) pro-posed three mechanisms by which neighborhood charac-teristics affect individual depression levels. Thesepathways include (1) level of daily stress imposed by lackof resources, physical stressors, and other people; (2)heightened vulnerability to experiencing negative events;and (3) disruption of social networks.

Through multi-level analyses, Matheson et al. (2006)studied the role of neighborhood stress, operationalizedas residential mobility and material deprivation, on depres-sion among Canadian residents. They found that aftercontrolling for individual characteristics, there was a signif-icant association between neighborhood stress and depres-sion. In a study conducted in Michigan US, Kruger, Reischl,and Gee (2007) found that the relationship between phys-ical deterioration of neighborhood and depression wasmediated by social contact with neighbors, social capital,and perceptions of crime.

Wen, Hawkley, and Cacioppo (2006) evaluated the rela-tionships between objective and perceived neighborhoodcharacteristics, psychosocial variables, and self-ratedhealth in a sample of older adults in Illinois. They utilizedneighborhood census data to construct an objective mea-sure of neighborhood SES. Subjective assessment of neigh-borhood was measured as perceptions of physical, social,and service environment. Through a series of regressionmodels, their findings indicated that objective measuresof neighborhood affected self-reported health throughperceptions of neighborhood and psychosocial factors in-cluding stress and depression.

Using structural equation modeling, Stiffman, Hadley-Ives, Elze, Johnson, and Dore (1999) examined thepathways between objective neighborhood conditions,perceived neighborhood conditions, environmental sup-port, and mental health outcomes among adolescents liv-ing in inner-city areas of St. Louis. Although they foundno direct impact of neighborhood conditions on mentalhealth, they did find an indirect pathway through individ-ual perceptions of neighborhood conditions.

Our previous research tested and confirmed a direct as-sociation between perceptions of neighborhood disorderand depression using the same SHIELD dataset that isreported on here (Latkin & Curry, 2003). Although the pres-ent study also considers the role of neighborhood percep-tions on distress, we have expanded our model to assessthe impact of an objective neighborhood measure (i.e.,police crime report) and personal experiences of violence.The present study prospectively examined the pathwaysbetween an objective measure of neighborhood violenceand depressive symptoms in a population of adults livingin an inner-city environment. Unlike Stiffman et al. (1999)and Wen et al. (2006), who used census variables at a cen-sus tract level of aggregation, our measure of objectiveneighborhood conditions comprising police crime records,aggregated to a block-group level. Research on violence inBaltimore has shown that the block group is an appropriatelevel of aggregation due to clustering of violent events andstressors (Harries, 1997).

While there is often a high degree of correlation be-tween census variables such as median household income,and rates of violent crime, violent crime may be more prox-imal in the causal pathway that leads from neighborhoodfactors to mental health outcomes. Our choice of blockgroups versus census tracts was motivated by previous re-search that found that perceptions of neighborhood disor-der, aggregated to a block-group level, were a morereliable measure than perceptions of neighborhood disor-der aggregated to a census tract level (Curry, 2004). In the

A. Curry et al. / Social Science & Medicine 67 (2008) 23–30 25

present study, we employed correlation and path analysisto help answer the following questions:

- To what extent does neighborhood violent crime, at theblock-group level, correlates with perceptions of neigh-borhood problems at the individual level among currentand former drug users living in an inner-cityenvironment?

- Is neighborhood violent crime positively associated withindividual experiences of violent crime?

- Do perceptions of neighborhood problems mediate therelationship between neighborhood violent crime andlevel of depressive symptoms; or

- Is there a direct impact of neighborhood violent crimeon levels of depressive symptoms?

We hypothesized that higher levels of neighborhood vi-olence would be directly associated with higher levels ofdepressive symptoms. Also, we posited that higher levelsof neighborhood violence would be indirectly associatedwith higher levels of depressive symptoms, with percep-tions of neighborhood crime and experiences of violenceacting as intermediate variables.

Methods

Survey assessment and data collection

The survey data used in this analysis were collected asa part of a social network-based HIV prevention interven-tion. Targeted outreach was used to recruit participants.Areas of high drug activity were assessed using focusgroups, geocoding of drug-related arrests over a 3-yearperiod, and ethnographic observations. A description ofthe study and a telephone number to call was providedto potential participants. Those that contacted ourresearch staff were given a brief screening to assess eligi-bility. Individuals who were eligible and agreed to partic-ipate were administered consent information and formsfollowed by a face-to-face interview on their socioeco-nomic background, HIV related behaviors, and their socialnetworks.

Study inclusion criteria consisted of: (1) being at least 18yearsold, (2) having daily or weekly contact with drug usersthrough either drug-related or non-drug-related interac-tions, (3) willingness to conduct AIDS outreach education,(4) willingness to recruit social network members intothe study, and (5) not being enrolled in other HIV preven-tion or network studies. Prior to data collection, the JohnsHopkins School of Public Health Committee on Human Re-search reviewed and approved all study protocols.

The present study is a longitudinal study of former andcurrent drug users. In this study, we analyze data collectedfrom participants at two time points. Baseline data werecollected from June 1997 through February 1999 and arereferred to as Time 1 data. Time 2 data refer to data col-lected during the follow-up assessment which occurred3 years (April 2000–June 2002) after baseline assessment,on average. In addition to the data collected from partici-pants, geographic data, specifically police crime reports,were analyzed. These data were collected 1 year prior to

baseline assessment (i.e. 1996) and are referred to asTime 0 data.

Measures

This study includes both individual-level measures andgroup-level measures (i.e. referred to as objective datathroughout the remainder of this manuscript).

Block-group neighborhood violenceOur objective measure of neighborhood disorder was

crime data obtained from the Baltimore City Police Depart-ment. We used crime data for 1996 (referred to as Time 0),since most of the interviews were conducted in 1997–1998.Only Part One crimes were available, meaning that misde-meanor and drug-related crimes such as drug possessionor selling could not be included in the analysis. To calculaterates of violent crime in each block group, we geocoded thelocation of each crime using the address provided by theBaltimore City Police Department. We then aggregated byblock group, and computed a rate per 1000 residents peryear for each category of crime. Since the distributions ofthe rates of crime were highly skewed, we used a naturallog transformation. We were interested only in violent orperson-to-person crimes, and thus used 4 of the 8 cate-gories of crimes available: assaults, murders, rapes, androbberies. Arson, larceny, burglary, and stolen vehiclecrimes were not used in the analysis. To create an overallmeasure of violent crime, we summed the rates of thefour violent crime categories. The Cronbach’s alpha forthis scale was 0.79 (N¼ 385 block groups, in the full Balti-more dataset).

Perceived neighborhood disorderPerceptions of neighborhood disorder were collected at

baseline (Time 1) through an adapted survey based on Per-kins and Taylor’s Block Environmental Inventory (Perkins,Meeks, & Taylor, 1992). The Perceptions of NeighborhoodDisorder Scale (NPDS) consists of seven items measuredon a three-point scale. Participants were asked about thefollowing events on their block: vandalism, litter or trashin the streets, vacant housing, groups of teenagers hangingout on the street, burglary, people selling drugs, and peoplegetting robbed on the street. Response options included‘‘not a problem,’’ ‘‘somewhat of a problem,’’ or a ‘‘big prob-lem.’’ The Cronbach’s alpha for this scale was 0.88. Amongparticipants, the NPDS scores ranged from 0 to 14 witha mean score of 7.2 (SD¼ 4.3).

Experiencing crime and violenceParticipants’ experience with crime and violence in the

past year was collected at follow-up assessment (Time 2).To measure this construct, we utilized a scale that wasadapted from the My Exposure to Violence Scale, originallydesigned for children and adolescents, to capture exposureto violent events that occur in an inner-city environment(Selner-O’Hagan, Kindlon, Buka, Raudenbush, & Earls,1998). Participants were asked to recall violent eventsthat occurred in their neighborhoods. First, participantswere asked if they had ever experienced nine events: (1)chased when you thought that you could really get hurt;

A. Curry et al. / Social Science & Medicine 67 (2008) 23–3026

(2) hit, slapped, punched, or beaten up; (3) attacked witha weapon, like a knife or bat; (4) shot; (5) shot at, but notactually wounded; (6) been in a serious accident whereyou or someone was hurt very badly, or someone elsedied; (7) sexually assaulted, molested, or raped; (8) threat-ened to seriously hurt you; (9) found a dead body. For eachevent that a participant had experienced, they were askedhow many times they had experienced it, the last time theyhad experienced it, and if the event had occurred in theirneighborhood.

To create a single measure of exposure to violence, weadded the number of violent events experienced in thepast year within the participant’s neighborhood (blockgroup). To maintain validity, we did not include eventsthat occurred outside of the participant’s neighborhoodor events that occurred more than 1 year prior to thetime of the survey. In this sample, experiencing violencescores ranged from 0 to 5 with a mean score of 0.2(SD¼ 0.6).

Depressive symptomsData on depressive symptoms were collected at follow-

up (Time 2). Depression was assessed using the Centers forEpidemiological Studies Depression Scale (CES-D) (Radloff,1977). The CES-D, which has high validity and reliability, isa 20-item, four-point scale developed for use in the generalpopulation (Radloff & Locke, 1986). Also, this scale has beenshown to have a high sensitivity for DSM-IV major depres-sion, and to have adequate specificity as a screening instru-ment for depression (Zimmerman & Coryell, 1988). In thissample, the CES-D scores ranged from 0 to 58 witha mean score of 15.8 (SD¼ 11.9).

Additional covariatesIn addition to the aforementioned constructs, several

other variables that have been linked to depression, vio-lence, and neighborhood characteristics were collectedthrough the survey and used in the analyses. Sociodemo-graphic variables include gender, education (<12 grade ver-sus high school diploma or higher), employment (at leastemployed part-time versus unemployed), age, and havinga main partner status (yes or no). Also, behaviors such asinjected any drugs in the past 6 months (yes or no), numberof hours per day spent on the street, and number of roles inthe drug economy (i.e. selling drugs, preparing drugs, ex-changing sex for money or drugs, etc.) were assessed sincethey may influence perceptions of one’s neighborhood aswell as experience with violence. These data were collectedduring baseline assessments (Time 1).

Analyses

GeocodingTo calculate the rates of violent crime in each block

group, the location of each crime was geocoded using theaddress provided by the Baltimore City Police Department.These data were then aggregated by block group, and com-puted a rate per 1000 residents per year for each categoryof crime. Since the distributions of the rates of crimewere highly skewed, a natural log transformation wasapplied.

Our sample (n¼ 786) resided in 270 of the 385 Balti-more City block groups. Of the 270 represented blockgroups, 175 block groups had 1 or 2 participants, and 95block groups had 3 or more participants.

Correlation analysisTo examine the bivariate relationships among block-

group level violent crime, participants’ experience with vi-olent crime, perceptions of neighborhood disorder, andlevel of depressive symptoms, we used Spearman Correla-tion Coefficients since the violent crime variable was highlyleft skewed. For comparison, however, we present bothPearson and Spearman coefficients.

Path modelTo test the hypothesis that neighborhood rates of crime

impacted CES-D scores directly and indirectly though per-ceptions of neighborhood problems, we estimated a pathmodel using Mplus (Muthen & Muthen, Los Angeles, CA,2002). We chose this strategy because it allowed us to si-multaneously assess perceptions of neighborhood disorderand experiences of violence as outcomes of neighborhoodlevel violence, and as predictors of depressive symptoms.A series of ordinary regression models would not allowsimultaneous estimation of coefficients. The path analysismethod was also preferred over a structural equationapproach, since our primary interest was in the pathwaysbetween neighborhood violence and level of depressivesymptoms, and not in the measurement models. We pres-ent standardized path coefficients (standardized to 0 mean,unit variance) to facilitate comparisons among thecoefficients.

We tested a model in which objective neighborhood vi-olence impacted depression through perceptions of neigh-borhood disorder. In the initial path model (see Fig. 1), thelevel of neighborhood block-group crime impacted bothperceptions of neighborhood level of disorder and level ofdepressive symptoms. The level of neighborhood crimealso impacted experiences of violence in the neighborhood,which in turn impacted level of depressive symptoms.Other factors (gender, age, injection drug use status, mainpartner status, and number of hours per day spent on thestreet) were hypothesized to impact level of depressivesymptoms, while the amount of time spent on the streetwas hypothesized to impact both experiencing violenceand perceptions of neighborhood disorder.

We used a Weighted Least Squares Minimum Varianceestimation method to compute path coefficients, since ourmodel included several categorical variables. All coefficientsare standardized to allow direct comparisons. For model fitevaluation, we used a chi-square test of Model Fit and eval-uated the Root Mean Square Error of Approximation. Asrules of thumb, the RMSEA should be below 0.05 and thechi-square statistic should be non-significant ( p> 0.05).

Results

Sample characteristics

Data were collected from 786 participants who werenested in 270 residential block groups. The sample was

CES-DTime 2

ExperiencedViolence Scale

Time 2

Block groupviolence, 1996 Time 0

Perceptions ofNeighborhood Disorder Scale

Time 1

Hours spenton StreetTime 1

MaleGender

+

+

+

+

+

-

Age

+

+

+

DrugUse

+

Has MainPartner

-

*All exogenous variables are specified as correlated, curved arrows not shown.

Fig. 1. Initial path model of effects of block-group level violence on psychological distress among sample of current and former drug users in Baltimore,Maryland.

A. Curry et al. / Social Science & Medicine 67 (2008) 23–30 27

predominantly composed of males (61.2%). The mean agewas 39.0 (SD¼ 7.3). The sample is highly disadvantaged,relative to the general population. Only 22% of the samplewas employed at least part-time at the time of the initial in-terview, and only 52% had completed at least a high schooleducation. Sixty-four percent of participants reported hav-ing a main sexual partner. Approximately 57% reportedinjecting drugs in the previous 6 months while 36%reported having at least one role in the street drug econ-omy. In addition, 14% of participants experienced at leastone violent event in their neighborhood in the past year,and the average study participant spent 6 h per day onthe street.

Block-group data

Descriptive statistics for the block groups are presentedin Table 1. The average population size of each group was

Table 1Descriptive statistics for block groups (N¼ 270) in Baltimore, MDa

Block-group characteristics Mean Minimum Maximum SD

Total population per blockgroup, 2000 census

878.4 139 2455 401.6

SHIELD participants perblock group

5.4 1 78 7.1

Rate of assault per 1000residents, 1996

28.3 0 308.1 29.5

Rate of murders per 1000residents, 1996

0.9 0 13.0 1.7

Rate of rape per 1000residents, 1996

2.1 0 46.5 3.6

Rate of robbery per 1000residents, 1996

32.7 0 875.0 59.2

Percent Completedhigh school

59.3

a The sample of 786 participants resided in 270 block groups.

878.4 (SD¼ 401.6). A mean of 2.8 (SD¼ 3.2) SHIELD partic-ipants were represented in each block group. Regardingcrime levels, the two highest rates were robbery(mean¼ 32.7 per 1000, SD¼ 59.2) and assault(mean¼ 28.3 per 1000, SD¼ 29.5).

Correlation analyses

The variable measuring neighborhood violent crimewas modestly, but significantly correlated with the Percep-tions of Neighborhood Disorder Scale (PNDS) (r¼ 0.19) andwith the experiences of violence scale (r¼ 0.09), but notwith level of depressive symptoms (see Table 2). ThePNDS was not correlated with experiences of violence inthe neighborhood. However, there was a modestly signifi-cant association with level of depressive symptoms(r¼ 0.13). Finally, the experiencing violence scale was mar-ginally correlated with level of depressive symptoms(r¼ 0.11).

Table 2Spearman and Pearson correlation coefficients among actual level of vio-lent crime, perceptions of neighborhood disorder, experiences of violence,and depressive symptoms, SHIELD Study, N¼ 786

Variable Spearman coefficients (Pearson coefficients)

Actual levelof violentcrime

Perceptions ofneighborhooddisorder

Experiencesof violence inneighborhood

Perceptions ofneighborhood disorder

0.19 (0.19)***

Experiences of violence inneighborhood

0.09 (0.07)* 0.02 (0.03)

Depressive symptoms(CES-D score)

0.03 (0.03) 0.13 (0.14)*** 0.11 (0.12)**

*p< 0.05; **p< 0.01; ***p< 0.001.

A. Curry et al. / Social Science & Medicine 67 (2008) 23–3028

Path analyses

The path model that was initially tested did not ade-quately fit the data, as indicated by both the chi-squaretest of Model Fit and the RMSEA (chi-square test significantp¼ 0.01, and RMSEA> 0.05). There were several path coef-ficients that were not statistically significant (z-score test).First, the direct path between neighborhood block-groupviolence and CES-D score was not significant, in contrastto our initial hypothesis. We removed this path from themodel. Stiffman et al. (1999) observed a similar result: theirmeasure of actual neighborhood conditions did not directlyimpact on the mental health status of the adolescents intheir study. Second, time spent on the street did not appearto impact experiences of violence in the neighborhood, andthus this path was removed from the model. Finally, age,drug use status, and main partner status were not signifi-cantly associated with level of CES-D, and were removedfrom the model. Only gender remained a significant predic-tor, with males having lower CES-D scores than females.

The final model included two indirect paths by whichlevel of violent block-group level crime was associatedwith CES-D: through perceptions of neighborhood disorderand through experiences of violence in the neighborhood.The total indirect effect of neighborhood level of violentcrime (using standardized coefficients) was 0.05. The directeffect of neighborhood violent crime was �0.02, but wasnot significant. Thus, the total effect of actual violence onCES-D was 0.03. For the NPDS, the R2 (variance explained)was 0.04; for the experiences of violence in the neighbor-hood, the R2 was 0.02, and for CES-D score, the R2 was

ExperiencedViolence Scale

Time 2

Block groupviolence, Time 0

Perceptions ofNeighborhoodDisorder Scale

Time 1

0.12

0.19

†z-test for significance p < 0.10, all other paths significantbelow p = 0.05

Chi-Square Test of Model FitValue 3.38Degrees of Freedom 6p-Value 0.76

RMSEA (Root Mean Square Error Of Approximation)Estimate

N

IAA

DA

T

0.000

Fig. 2. Final path model of indirect effects of block-group level violence on psycholMaryland.

0.07. The final model, although different from the initialmodel, provided support for the hypothesis that neighbor-hood violence was associated with depressive symptomsthrough perceptions of neighborhood and experiences ofviolence (Fig. 2).

Discussion

The findings from this current study suggest that neigh-borhood crime is associated with depressive symptoms. Al-though we did not observe a direct path betweenneighborhood level of violent crime and depressive symp-toms, we did find support for two indirect pathways bywhich neighborhood violent crime is associated with de-pressive symptoms: through perceptions of neighborhooddisorder and through experiences of violence in the neigh-borhood. Our hypothesized sequence (neighborhood vio-lence impacts perceptions of neighborhood disorder andexposure to violence, which both in turn impact level ofpsychological distress) is further supported by the use ofa prospective design. Psychological distress was measuredapproximately 2 years after measurement of neighborhoodviolent crime and perceptions of neighborhood disorder.Experiencing violence was measured concurrently withpsychological distress, at Time 2.

Our sample represented a homogenous group of indi-viduals with a low socioeconomic background living inneighborhoods with high levels of drug activity. The homo-geneity of the sample is both a strength and weakness ofthis study. If the sample had included a wider range of af-fluent neighborhoods and participants, it is likely that the

CES-DTime 2

Hours spenton Street Time 1

MaleGender

0.14

0.16 0.17

0.07†

= 786, Number of Block groups = 270

ndirect Causal Effect of Actual Violencectual Violence Perceptions CES-D = 0.03ctual Violence Experienced Violence CES-D = 0.02

irect Causal Effect of Actual Violencectual Violence CES-D = −0.02 (not significant, not shown)

otal Causal Effect of Actual Violence on CES-D = 0.03

ogical distress among sample of current and former drug users in Baltimore,

A. Curry et al. / Social Science & Medicine 67 (2008) 23–30 29

results would be strengthened due to the increase in vari-ation of both predictors and outcomes. However, differentprocesses and pathways may operate within different pop-ulations. In more affluent neighborhoods, with little crimeor disorder, there may be no observable impact of violentcrime on mental health status. Furthermore, higher levelsof available social support and social resources in more af-fluent neighborhoods may have a greater impact on mentalhealth status. The vast majority of studies that have exam-ined the mitigating impacts of social support and socialresources on the relationship between neighborhood con-ditions and negative outcomes have focused on childrenand adolescents (Drukker, Kaplan, Schneiders, Feron, &van Os, 2006; Garmezy, 1993). More research is neededthat examines this association among adults.

The significant but modest correlation between percep-tions of neighborhood disorder and neighborhood violentcrime indicates that residents’ perception of the conditionof their neighborhood may be accurate. The correlation,however, is not strong, partially due to the fact that levelof violent crime and neighborhood disorder are not identi-cal constructs. Furthermore, police data, while consideredan ‘objective’ measure of the level of violent crime in a blockgroup, may be inadequate to assess less serious and morefrequent forms of neighborhood violence. In extremely dis-advantaged areas with high levels of illicit drug trade andserious violence, police may not be able to arrest or reportmany less serious violent incidents. Additionally, violent in-cidents of a less serious nature may go unreported to thepolice, in part due to mistrust and lack of confidence inthe ability of police to address less serious violent crimes.More research is needed to examine how less violent crimemay be associated with perceived neighborhood disorderand depression.

Our findings have several alternative interpretations.First, the association between exposure to violence andpsychological distress could be accounted for by socialcapital. Using state-level crime rates, Kawachi, Kennedy,and Wilkinson (1999) found that violence is associatedwith low social capital. The authors propose that fear ofcrime and violence leads to few interactions among resi-dents and disorganization which impede social capital. Fur-ther, researchers have shown that low levels of socialcapital are associated with poor mental health outcomes(Fitzpatrick, Piko, Wright, & LaGory, 2005; Phongsavan,Chey, Bauman, Brooks, & Silove, 2006).

Another possible explanation for the violence–depres-sion association is mood-congruent bias; that is, individ-uals who are more distressed may be more likely to recallviolent events. Third, it is possible that depressed individ-uals move into areas of the city that are more violent andhave more outward signs of disorder. Lack of financialmeans could force persons to move to such neighborhoods.In this case, the pathways linking neighborhood conditionsand distress that we observed would, therefore, be spuri-ous, since neighborhood conditions did not cause psycho-logical distress. Our study found an association betweenneighborhood factors and depression. While the goodmodel fit provides support for the sequence of variableswe specify, we cannot claim that higher levels of neighbor-hood violence cause higher levels of depressive symptoms.

One limitation of this study is our modeling approach.The purpose of this study was not to account for all factorsthat may influence psychological distress. Rather, the studywas an exploratory analysis of the pathways that exist be-tween neighborhood conditions and psychological distress.Path analysis is widely used as an exploratory data analysistool. However, we recognize that path models are inher-ently sensitive to correct model specification and that ifvariables which lie in the ‘true’ causal pathways are leftout, other model parameters may be incorrectly estimated.Furthermore, it is unlikely that the two pathways we spec-ified are the only ones that lead from neighborhood condi-tions to psychological distress. However, our results aresimilar to that of Stiffman, Dore, Earls, and Cunningham(1992), which provide support for the relationship betweenneighborhood conditions and mental health outcomes. An-other limitation of the study is that due to sampling thestudy’s generalizability is limited.

Despite these limitations, this study has several implica-tions. Both neighborhood crime and perceptions of neigh-borhood are related to mental health of local residents.Community-wide and structural violence prevention inter-ventions, such as stricter enforcement of laws, increasedpolice presence, neighborhood watch groups, and installa-tion of surveillance cameras are needed in urban communi-ties. Such interventions may serve as a barrier forperpetuators, thus decreasing neighborhood criminallevels as well as likelihood of being victimized. Further,Cohen, Farley, and Mason (2003) suggest that stricterenforcement of zoning and housing codes and additionalefforts to keep streets clean are policies that cities can con-sider to decrease the impact of ‘‘broken windows’’ on thehealth of inner-city residents. Although this theory is theproduct of debate (Taylor, 2001), these policies mayimprove residents’ perceptions of their neighborhoodwhich may ultimately lead to lower levels of depressivesymptoms.

References

Avison, W. R., & Turner, R. J. (1988). Stressful life events and depressivesymptoms: disaggregating the effects of acute stressors and chronicstrains. Journal of Health and Social Behavior, 29, 253–264.

Bogat, G. A., Leahy, K., von Eye, A., Maxwell, C., Levendosky, A. A., &Davidson, W. S. (2005). The influence of community violence on thefunctioning of women experiencing domestic violence. AmericanJournal of Community Psychology, 36, 123–132.

Cohen, D. A., Farley, T. A., & Mason, K. (2003). Why is povertyunhealthy? Social and physical mediators. Social Science & Medi-cine, 57, 1631–1641.

Curry, A. D. (2004). Pathways to depression: the impact of neighborhoodviolent crime on inner-city residents in Baltimore, Maryland. Disserta-tion Abstracts International.

Cutrona, C. E., Wallace, G., & Wesner, K. A. (2006). Neighborhoodcharacteristics and depression: an examination of stress processes.Current Directions in Psychological Science, 15, 188–192.

Drukker, M., Kaplan, C., Schneiders, J., Feron, F. J., & van Os, J. (2006). Thewider social environment and changes in self-reported quality of lifein the transition from late childhood to early adolescence: a cohortstudy. BMC Public Health, 6, 133.

Dupere, V., & Perkins, D. D. (2007). Community types and mental health:a multilevel study of local environmental stress and coping. AmericanJournal of Community Psychology, 39, 107–119.

Fauth, R. C., Leventhal, T., & Brooks-Gunn, J. (2004). Short-term effects ofmoving from public housing in poor to middle-class neighborhoodson low-income, minority adults’ outcomes. Social Science & Medicine,59, 2271–2284.

A. Curry et al. / Social Science & Medicine 67 (2008) 23–3030

Fitzpatrick, K. M., Piko, B. F., Wright, D. R., & LaGory, M. (2005). Depressivesymptomatology, exposure to violence, and the role of social capitalamong African American adolescents. American Journal of Orthopsy-chiatry, 75, 262–274.

Galea, S., Ahern, J., Nandi, A., Tracy, M., Beard, J., & Vlahov, D. (2007).Urban neighborhood poverty and the incidence of depression ina population-based cohort study. Annals of Epidemiology, 17, 171–179.

Garmezy, N. (1993). Children in poverty: resilience despite risk.Psychiatry, 56, 127–136.

Gary, T. L., Stark, S. A., & LaVeist, T. A. (2007). Neighborhood character-istics and mental health among African Americans and whitesliving in a racially integrated urban community. Health and Place,13, 569–575.

Glassman, A. H., & Shapiro, P. A. (1998). Depression and the course ofcoronary artery disease. American Journal of Psychiatry, 155, 4–11.

Goldsmith, H. F., Holzer, C. E., III, & Manderscheid, R. W. (1998).Neighborhood characteristics and mental illness. Evaluation andProgramming Planning, 21, 211–225.

Harries, K. (1997). Social stress and trauma: synthesis and spatial analysis.Social Science & Medicine, 45, 1251–1264.

Inaba, A., Thoits, P. A., Ueno, K., Gove, W. R., Evenson, R. J., & Sloan, M.(2005). Depression in the United States and Japan: gender, maritalstatus, and SES patterns. Social Science, 61, 2280–2292.

Jiang, W., Krishnan, R. R., & O’Connor, C. M. (2002). Depression and heartdisease: evidence of a link, and its therapeutic implications. CNSDrugs, 16, 111–127.

Kawachi, I., Kennedy, B. P., & Wilkinson, R. G. (1999). Crime: socialdisorganization and relative deprivation. Social Science & Medicine,48, 719–731.

Kessler, R. C. (1979). Stress, social status, and psychological distress.Journal of Health and Social Behavior, 20, 259–272.

Kruger, D. J., Reischl, T. M., & Gee, G. C. (2007). Neighborhood socialconditions mediate the association between physical deteriorationand mental health. American Journal of Community Psychology.

Latkin, C. A., & Curry, A. D. (2003). Stressful neighborhoods anddepression: a prospective study of the impact of neighborhood disor-der. Journal of Health and Social Behavior, 44, 34–44.

Lofors, J., & Sundquist, K. (2007). Low-linking social capital as a predictorof mental disorders: a cohort study of 4.5 million Swedes. Social Sci-ence & Medicine, 64, 21–34.

McLeod, J. D., & Kessler, R. C. (1990). Socioeconomic status differences invulnerability to undesirable life events. Journal of Health and SocialBehavior, 31, 162–172.

MacMahon, K. M., & Lip, G. Y. (2002). Psychological factors in heart fail-ure: a review of the literature. Archives of Internal Medicine, 162,509–516.

Maddock, C., & Pariante, C. M. (2001). How does stress affect you? Anoverview of stress, immunity, depression and disease. Epidemiologiae psichiatria sociale, 10, 153–162.

Matheson, F. I., Moineddin, R., Dunn, J. R., Creatore, M. I., Gozdyra, P., &Glazier, R. H. (2006). Urban neighborhoods, chronic stress, genderand depression. Social Science & Medicine, 63, 2604–2616.

Perkins, D. D., Meeks, J. W., & Taylor, R. B. (1992). The physical environ-ment of street blocks and resident perceptions of crime and disorder:

implications for theory and measurement. Journal of EnvironmentalPsychology, 12, 21–34.

Phongsavan, P., Chey, T., Bauman, A., Brooks, R., & Silove, D. (2006). Socialcapital, socio-economic status and psychological distress amongAustralian adults. Social Science & Medicine, 63, 2546–2561.

Propper, C., Jones, K., Bolster, A., Burgess, S., Johnston, R., & Sarker, R.(2005). Local neighbourhood and mental health: evidence from theUK. Social Science & Medicine, 61, 2065–2083.

Radloff, L. S. (1977). The CES-D scale: a self-report depression scale forresearch in the general population. Applied Psychological Measure-ment, 1, 385–401.

Radloff, L. S., & Locke, B. Z. (1986). The community mental healthassessment survey and the CES-D scale. In M. M. Weissman, J. K.Myers, & C. E. Ross (Eds.), Community surveys of psychiatric disorder(pp. 177–189). New Brunswick, NJ: Rutgers University Press.

Ross, C. E. (2000). Neighborhood disadvantage and adult depression.Journal of Health and Social Behavior, 41, 177–187.

Selner-O’Hagan, M. B., Kindlon, D. J., Buka, S. L., Raudenbush, S. W., &Earls, F. J. (1998). Assessing exposure to violence in urban youth. Jour-nal of Child Psychology and Psychiatry and Allied Disciplines, 39,215–224.

Silver, E., Mulvey, E. P., & Swanson, J. W. (2002). Neighborhood structuralcharacteristics and mental disorder: Faris and Dunham revisited. So-cial Science & Medicine, 55, 1457–1470.

Steptoe, A., & Feldman, P. J. (2001). Neighborhood problems as sources ofchronic stress: development of a measure of neighborhood problems,and associations with socioeconomic status and health. Annals ofBehavioral Medicine, 23, 177–185.

Stiffman, A. R., Dore, P., Earls, F., & Cunningham, R. (1992). The influence ofmental health problems on AIDS-related risk behaviors in youngadults. Journal of Nervous Mental Disorders, 180, 314–320.

Stiffman, A. R., Hadley-Ives, E., Elze, D., Johnson, S., & Dore, P. (1999).Impact of environment on adolescent mental health and behavior:structural equation modeling. American Journal of Orthopsychiatry,69, 73–86.

Sundquist, K., & Ahlen, H. (2006). Neighbourhood income and mentalhealth: a multi-level follow-up study of psychiatric hospital admissionsamong 4.5 million women and men. Health & Place, 12, 594–602.

Taylor, R. B. (2001). Breaking away from broken windows: Baltimoreneighborhoods and the nationwide fight against crime, grime, fear, anddecline, (18th ed.). Boulder, CO: Westview Press.

Truong, K. D., & Ma, S. (2006). A systematic review of relations betweenneighborhoods and mental health. The Journal of Mental Health Policyand Economics, 9, 137–154.

Wen, M., Hawkley, L. C., & Cacioppo, J. T. (2006). Objective and perceivedneighborhood environment, individual SES and psychosocial factors,and self-rated health: an analysis of older adults in Cook County, Illi-nois. Social Science & Medicine, 63, 2575–2590.

Ziersch, A. M., Baum, F. E., Macdougall, C., & Putland, C. (2005).Neighbourhood life and social capital: the implications for health.Social Science & Medicine, 60, 71–86.

Zimmerman, M., & Coryell, W. (1988). The validity of a self-reportquestionnaire for diagnosing major depressive disorder. Archives ofGeneral Psychiatry, 45, 738–740.