Health & Place 16 (2010) 1007–1013

Contents lists available at ScienceDirect

Health & Place

1353-82

doi:10.1

n Corr

Cancer

Tel.: +1

E-m

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

Associations between perceived social environment and neighborhoodsafety: Health implications

Maria De Jesus a,b,n, Elaine Puleo c, Rachel C. Shelton d, Karen M. Emmons a,b

a Center for Community-Based Research, Dana-Farber Cancer Institute, 44 Binney Street, LW 703, Boston, MA, USAb Department of Society, Human Development, and Health, Harvard School of Public Health, Boston, MA, USAc Department of Public Health, Division of Biostatistics and Epidemiology, University of Massachusetts, Amherst, MA, USAd Department of Oncological Sciences, Division of Cancer Prevention & Control Mount Sinai School of Medicine, New York, NY, USA

a r t i c l e i n f o

Article history:

Received 14 September 2009

Received in revised form

9 June 2010

Accepted 9 June 2010

Keywords:

Social networks

Social support

Social cohesion

Neighborhood safety

Public housing residents

92/$ - see front matter & 2010 Elsevier Ltd. A

016/j.healthplace.2010.06.005

esponding author at: Center for Community-

Institute, 44 Binney Street, LW 703, Boston, M

617 582 7468; fax: +1 617 632 5690.

ail address: Maria_Dejesus@dfci.harvard.edu

a b s t r a c t

This study examined the associations between social networks, social support, social cohesion, and

perceived neighborhood safety among an ethnically diverse sample of 1352 residents living in 12 low-

income public housing sites in Boston, Massachusetts. For males and females, social cohesion was

associated with perceived safety. For males, a smaller social network was associated with greater

feelings of safety. Social support was not a significant predictor of perceived safety. The findings

reported here are useful in exploring a potential pathway through which social environmental factors

influence health and in untangling the complex set of variables that may influence perceived safety.

& 2010 Elsevier Ltd. All rights reserved.

1. Introduction

A growing body of public health research suggests thatperceptions of neighborhood safety are linked to health outcomes(Macintyre and Ellaway, 2000; Chandola, 2001; Ziersch et al.,2005; Baum et al., 2009). Perceiving one’s neighborhood as unsafehas been significantly associated with anxiety (Middleton, 1998),poor health outcomes (Macintyre and Ellaway, 2000), and poorself-rated health (Chandola, 2001). Furthermore, evidence sug-gests that individuals’ sense of neighborhood safety is associatedwith the extent to which they participate in and interact withtheir community (Sampson, 2003; Young et al., 2004; Zierschet al., 2007; Lochner et al., 1999; Baum et al., 2009). Empiricalstudies examining the role of neighborhood factors on healthdemonstrate that higher levels of safety are associated withhigher respondent perceptions of social cohesion and betterhealth outcomes (e.g., Wen et al., 2007; King, 2008). Scientificresearch demonstrates that ‘‘the degree to which an individual isinterconnected and embedded in a community—is vital to anindividual’s health and well-being y’’ (Berkman and Glass, 2000).This paper examines perceptions of neighborhood safety as a

ll rights reserved.

Based Research, Dana-Farber

A 02115, USA.

(M. De Jesus).

potential mechanism through which social environmental factors– social support, social networks, and social cohesion – mayinfluence health (Diagram 1).

Two commonly measured characteristics of social relation-ships are social support and social networks (Cohen and Syme,1985; Heaney and Israel, 1997). Social support has been broadlydefined as resources (e.g., emotional, instrumental, and financialsupport) provided in the context of a relationship (Cohen andSyme, 1985). Social networks refer to the web of social relation-ships that surround an individual, and provide information on theextent to which an individual is connected with others (Berkmanand Syme, 1979). Another key aspect of the social environment issocial cohesion which refers to the extent of connectedness andsolidarity among groups (Kawachi and Berkman, 2000).

Studies exploring the interrelationships among residents ofdifferent public housing programs have resulted in mixedfindings. Social network structure tends to vary by socioeconomicstatus (SES), and the social networks of those with lower incomestend to be more place-based, be homogeneously low income,contain more close relationships, and contain more overlappingrelations (e.g., Campbell and Lee, 1992). However, several studiescontradict these findings (e.g., Hurlbert et al., 2001; Kleit, 2001;Chaskin and Joseph, 2010). For example, Hurlbert et al. (2001)examined poor people’s networks in a low-income neighborhoodand found that those who lived in low-income neighborhoods,rather than being involved in more intense and strong relation-ships, instead tended to have weaker relationships than those in

Social networks

Social support

Social cohesion

Health outcomes

Mediating Mechanisms

Perceptions of neighborhood safety

Diagram 1. Conceptual pathway between perceptions of neighborhood safety and health.

M. De Jesus et al. / Health & Place 16 (2010) 1007–10131008

middle-income areas. Kleit’s (2001) study demonstrated thatscattered-site and clustered public housing residents differ verylittle in terms of the number of neighbors they know, the degreeof embeddedness of relationships within their neighborhoods,and the amount of aid received from their neighbors.

Although social support, social networks, and social cohesionfactors have been found in the literature to have importantimplications for health (Macintyre and Ellaway, 2000; Cattell,2001; Ziersch et al., 2005; Almedom, 2005; De Silva et al., 2005;Ziersch et al., 2007), the mechanisms through which they workare not well-understood, particularly among low-income popula-tions in urban settings. In this paper, we were interested inexamining whether public housing residents’ perceptions ofspecific characteristics of the social environment—social support,social networks, and social cohesion—influence their perceptionsof neighborhood safety as a potential pathway through whichsocial environmental factors influence health. We hypothesizedthat residents with larger social networks, higher levels of socialsupport, and higher levels of social cohesion in their neighbor-hood would be more likely to perceive their neighborhood as safercompared to their counterparts. Additionally, we examinedwhether each of these associations differed by age, ethnicity/race,health status, financial status, and employment status.

2. Methods

2.1. Study design and sample

This study analyzed baseline data collected in 2004–2005 fromOpen Doors to Health (ODH), a cluster randomized control trial ofcolorectal cancer (CRC) prevention targeting CRC screening,physical activity, and multi-vitamin intake. Twelve urban publichousing communities in low income Boston neighborhoods werethe primary sampling units and individuals within housing sitesserved as secondary sampling units. Ninety-two percent ofhousing sites approached (12 out of 13) agreed to participate.The physical properties of the sites varied, from high-risebuildings clustered around common space, to distributed sites,where buildings with subsidized units were interspersed betweenbuildings that did not belong to that site. Unequal probabilitysampling across housing sites was employed due to the varyingsize of housing sites. In half the sites where the population wasless than 300, all adult residents were sampled. In the other half ofthe sites where the population size was greater than 300, arandom sample was drawn to obtain a 35% sample with aminimum of 250 participants per site.

An initial sample of 3688 participants was drawn. Of them, 747(20%) were deemed ineligible for various reasons including had a

cognitive or physical disability, had active cancer, were too ill toparticipate, or were no longer at that address, leaving an eligiblesample population of 2941 individuals. Of these, 828 (28%)refused participation and 559 (19%) were never reached. Enroll-ment and baseline surveys were obtained on 1554 participants.This yielded an overall 53% response rate, which ranged from alow of 34% to a high of 92% across the housing sites. This responserate is comparable to that of other community-based studies(Ellison-Loschmann et al., 2007; Tan et al., 1997). Participantsprovided verbal informed consent and completed an interviewer-administered survey in either English or Spanish. They receivedUS $25 compensation for their participation. The study protocolwas approved by the human subjects committee at the HarvardSchool of Public Health.

On the basis of the cluster design, data for all analyses wereweighted up to the population size within each housing site(weighted sample size of 1980). The data reported here representbaseline data, among participants who answered questions onperceived neighborhood safety. Those missing data for thisoutcome variable (n¼202) were dropped (n¼66 due to missingdata; n¼136 due to completion of a short form of the baselinesurvey because of time and other constraints). The short form didnot include perceived safety questions.

2.2. Study recruitment

Study recruitment has been described in full detail elsewhere(Bennett et al., 2007). Briefly, participant recruitment began withhousing site representatives sending letters announcing the studyto their eligible residents. A total of 14–16 contact attempts weremade (10 phone attempts and a minimum of 4 door knocks, 6 ifno phone). Eligibility criteria for the study survey included: (1)residence in one of the participating housing communities, (2) ageof at least 18 years, and (3) fluency in English or Spanish.

2.3. Measures

Sociodemographic characteristics: Race/ethnicity (categorizedas Hispanic, Black, White, and Other), gender, and age wereassessed using standard demographic questions during the inter-view-administered survey. Health status was captured by askingparticipants whether current health problems make it difficult forthem to exercise (yes/no). Household financial status wasassessed on four dimensions and dichotomized as comfortablewith some extras/enough but no extras versus have to cut back/can’t make ends meet. Poverty status was assessed by combiningyearly household income and the number of people supported bythis income. Poverty status was dichotomized as being above orbelow the poverty level based on the 2005 Federal Poverty

M. De Jesus et al. / Health & Place 16 (2010) 1007–1013 1009

Guidelines on income and household size (HHS, 2005). Inaddition, employment status was assessed and organized intofour categories and dichotomized into work full-time/part-timeversus disabled/not working. Participants’ first or native language,level of education, and immigrant status were also assessed.

2.4. Perceived social environmental factors

Social networks were measured using the Berkman–SymeSocial Networks Index (BSSNI), a composite measure of four typesof social connection: marital status, sociability, church groupmembership, and membership in other community organizations(Berkman and Syme, 1979). Items included: (1) Are you marriedor living with a partner now? (Responses: No, Yes); (2) How manyrelatives, like your spouse, children, aunts, uncles, brothers, orsisters do you feel close to? (Responses: None; 1 or 2; 3–5; 6–9;10, or more); (3) How many close friends do you have?(Responses: None; 1 or 2; 3–5; 6–9; 10 or more); and (4) Areyou an active member of any of these groups or clubs?(Responses: (a) Church, temple, mosque, or other religious group;(b) recreation or sport league; (c) civic, political, service, housingsite, or other community organization; (d) professional trade orlabor organization; (e) any other organization (write-in)). Thesummary scores ranged from 0 to 4 reflecting the number ofsocial network ties (Berkman and Syme, 1979). A median split oftwo categories was then created (o2 ties; 3–4 ties) for theseanalyses, based on the distribution of responses.

Social support was assessed in three of the four domains of theInventory of Socially Supportive Behaviors (emotional, instru-mental, and financial support) (Barrera et al., 1981). Residentswere asked: (1) Can you count on a friend or relative to give youemotional support; (2) When you need some extra help, canyou count on a friend or relative to help you with daily tasks?;(3) When you need some extra money for basic needs, can youcount on a friend or relative to help you?; (4) If you needed helpgetting to the doctor, is there someone who would help you?; and(5) When you are sick and need extra help, can you count on afriend or relative to help you? Responses included: Yes, always;Yes, sometimes; No, there isn’t anyone like that; I don’t need help;Yes, but I wouldn’t accept help. A summary score for socialsupport ranging from 0 to 5 was created by adding the number ofresponses to the five questions. Due to limited variability,social support was dichotomized into high level of social support(score of 5) or lower level of social support (score of o5) for thepurposes of analyses.

Social cohesion was assessed on dimensions suggested bySampson and colleagues, examining perceptions of trust andshared values in one’s neighborhood (Sampson et al., 1997).Residents were asked to report their agreement (Strongly agree;Somewhat agree; Somewhat disagree; Strongly disagree) withfive statements: (1) People around here are willing to help theirneighbors; (2) This is a close-knit neighborhood; (3) People in thisneighborhood can be trusted; (4) People in this neighborhoodgenerally do not get along with each other; and (5) People in thisneighborhood do not share the same values (Cronbach’salpha¼ .7801). A summary score ranged from 1 to 4 with a higherscore indicating higher social cohesion.

2.5. Perceptions of neighborhood safety

The outcome variable ‘‘perceived neighborhood safety,’’ wasmeasured using previously tested questions (with slight mod-ifications to accommodate literacy concerns) (Pebley and Nara-yan, 2003; Rountree and Land, 1996). For both ‘‘daytime’’ and‘‘night-time,’’ participants were asked: ‘‘y how safe do you feel

walking alone in your neighborhood?’’ These questions aredesigned to capture global perceptions of neighborhood safetyand as such, may reflect views on a variety of factors (e.g., crime,traffic, green space, etc.). Response options included: unsafe,safe, and a little unsafe. For analysis purposes, we computed a3-category outcome variable: (1) ‘‘unsafe’’ if perceived neighbor-hood as ‘‘not at all safe’’ either at night or during the day; (2)‘‘safe’’ if perceived neighborhood as ‘‘safe’’ both at night andduring the day; and (3) ‘‘a little unsafe’’ if any combination of dayor night was reported as ‘‘a little unsafe.’’

2.6. Data analysis

Frequency distributions and estimates of means and standarderrors were calculated for all exposures, outcome, and covariatesof interest to assess distributional assumptions and outliers.Crosstabs of demographic characteristics, organized by percep-tions of neighborhood safety for males and females separatelyrevealed gender differences, and thus gender-stratified bivariateanalyses were also conducted. Based on the bivariate associationsand consideration of confounders, proportional odds polytomouslogistic regression models accounting for the cluster design of thedependent variable were created (McCullagh, 1980; Stokes et al.,2000). In these models, odds ratios represent incremental oddsbetween the 3 categories of the outcome variable (feeling safe vs.a little unsafe and unsafe, and, feeling a little unsafe vs. unsafe).Planned interaction analyses examining potential modifiers (e.g.,age, race/ethnicity, and health status) in the association betweensocial environmental factors (social networks, social support, andsocial cohesion) and perceived safety were also conductedseparately for each social factor. Only those variables that weresignificantly associated with neighborhood safety at the pr .05level or were found to be effect modifiers were retained in thefinal models. Appropriate analyses were conducted usingSUDAAN 9.01 and SAS 9.1 statistical software for clustered data.

3. Results

3.1. Sociodemographic analyses

Table 1 presents demographics for the overall subsample ofODH participants for this study and by perceptions ofneighborhood safety. The participants were largely non-white(96%), female (74%), had low levels of education (67%oHS), andwere either not currently working (40%) or were disabled (31%).Average age was 48 years old. Slightly more than half theparticipants were born in the US (55%), reported English as theirfirst language (54%), and were living below the poverty level(54%). A large percentage of the female residents (66%) and morethan half of the male residents (57%) reported feeling ‘‘a littleunsafe’’ or ‘‘unsafe’’ walking alone in their neighborhoods.

3.2. Perceptions of neighborhood safety for male residents

Bivariate analyses for male residents (Table 2) indicated thatthose who had completed high school (po .01) or at least somecollege (po .001), or who were born in the US (po .03), or whosefirst language was English (po .03) were each significantly morelikely to feel greater levels of safety compared to each of theircounterparts. In addition, not having current health problems thatmade it difficult to exercise (po .001) was associated withperceptions of neighborhood safety in bivariate analyses formales.

Table 1Participant characteristics overall and by perceptions of neighborhood safety (weighted N¼1980**).

Overall weighted (%) Males weighted (%) Females weighted (%)

n (%) Safe n (%) A little unsafe n (%) Unsafe n (%) Safe n (%) A little unsafe n (%) Unsafe n (%)

TOTAL 1980 (100%) 222 (44%) 176 (35%) 111 (22%) 499 (34%) 484 (33%) 488 (33%)

Race/EthnicityBlack 957 (49%) 109 (50%) 76 (35%) 34 (16%) 270 (37%) 253 (34%) 216 (29%)

White 89 (4%) 15 (44%) 14 (43%) 5 (13%) 23 (41%) 21 (39%) 11 (20%)

Hispanic 863 (44%) 89 (38%) 73 (32%) 70 (30%) 186 (29%) 196 (31%) 250 (40%)

Other 64 (3%) 9 (40%) 12 (51%) 2 (9%) 19 (46%) 14 (34%) 8 (20%)

Ageo35 469 (24%) 54 (49%) 49 (45%) 7 (6%) 117 (33%) 147 (41%) 96 (27%)

35–49 518 (26%) 46 (42%) 37 (34%) 26 (24%) 171 (42%) 148 (36%) 89 (22%)

50–64 628 (32%) 65 (42%) 55 (35%) 36 (23%) 147 (31%) 129 (27%) 196 (41%)

65+ 365 (18%) 57 (43%) 35 (26%) 42 (31%) 64 (28%) 61 (26%) 107 (46%)

Employment statusWork full time 459 (23%) 60 (52%) 38 (33%) 17 (14%) 139 (40%) 131 (38%) 74 (22%)

Work part time 290 (15%) 30 (47%) 24 (38%) 9 (15%) 80 (35%) 92 (40%) 55 (24%)

Disabled 433 (22%) 37 (32%) 42 (38%) 34 (30%) 99 (31%) 78 (25%) 142 (44%)

Not working 799 (40%) 95 (44%) 71 (33%) 51 (24%) 181 (31%) 183 (32%) 216 (37%)

Poverty levelAbove poverty level 820 (46%) 96 (42%) 82 (36%) 50 (22%) 198 (33%) 221 (37%) 172 (29%)

Below poverty level 965 (54%) 101 (44%) 77 (33%) 54 (23%) 245 (34%) 211 (29%) 275 (38%)

EducationLess than HS 784 (40%) 92 (38%) 73 (30%) 78 (32%) 149 (28%) 146 (27%) 245 (45%)

Completed highSchool/vocational 543 (27%) 59 (48%) 46 (37%) 19 (15%) 158 (38%) 142 (34%) 119 (28%)

At least some college 650 (33%) 71 (50%) 56 (40%) 14 (10%) 191 (38%) 195 (38%) 123 (24%)

ImmigrantBorn in US 1078 (54%) 120 (48%) 96 (39%) 32 (13%) 294 (35%) 306 (37%) 229 (28%)

Born in Puerto Rico 500 (25%) 56 (38%) 39 (26%) 53 (36%) 113 (32%) 89 (25%) 152 (43%)

Not born in US or PR 401 (20%) 46 (41%) 41 (37%) 26 (23%) 92 (32%) 89 (31%) 106 (37%)

English 1st languageNo 904 (46%) 95 (38%) 80 (32%) 76 (30%) 199 (30%) 194 (30%) 260 (40%)

Yes 1076 (54%) 127 (49%) 96 (37%) 35 (14%) 300 (37%) 290 (35%) 227 (28%)

M. De Jesus et al. / Health & Place 16 (2010) 1007–10131010

Table 3 displays the polytomous multivariable model ofperceived neighborhood safety. As hypothesized, the overallassociation between social cohesion and perceptions ofneighborhood safety remained significant (po .03), controllingfor education. We did not find a main effect of social support onperceptions of neighborhood safety as we had hypothesized.However, social networks (po .05) was significantly associatedwith perceived safety for male residents in the multivariableanalyses, but in the opposite direction to our hypothesis. Menwho reported having smaller social networks were more likely toperceive their neighborhoods as safer than those with largersocial networks. Educational level was also a significant predictorof perceptions of neighborhood safety for male residents withhigher levels of perceived safety among those with higher levelsof education.

We also tested specific interaction effects for our hypothesizedmoderators (age, ethnicity/race, health status, financial status,and employment status). None of these variables were found tomoderate the association between perceptions of neighborhoodsafety and social cohesion or social networks for male residents.

3.3. Perceptions of neighborhood safety for female residents

Bivariate analyses on primary exposures and outcomes ofinterest for female residents (Table 2) indicated that those whohad completed high school (po .01) or at least some college(po .001), or who were born in the US (po .03), or whose firstlanguage was English (po .001) were significantly more likely tohave higher levels of perceived safety than their counterparts.Female residents who were Black (po .001), between 35 and 49years old (po .001), or working were significantly more likely to

have higher perceived neighborhood safety compared to theircounterparts. In addition, not having current health problems thatmade it difficult to exercise (po .01), having higher levels of socialsupport (po .03), and higher perceived social cohesion (po .001)were statistically significant predictors of perceptions of neigh-borhood safety among women. Having larger social networks,however, was not associated with perceptions of greater neigh-borhood safety for female residents.

Table 3 displays the polytomous multivariable model ofperceived neighborhood safety. As hypothesized, the overallassociation between social cohesion and perceptions of neighbor-hood safety remained significant (po .001), controlling foreducation, English as a first language, age, employment status,and health problems that make it difficult to exercise. We did notfind a main effect of social network or social support onperceptions of neighborhood safety. Educational level, English asa first language, age, and employment status also remainedsignificant predictors of neighborhood safety in the multivariablemodels for female residents.

We also tested specific interaction effects to examine differencesin potential moderators (age, ethnicity/race, health status, financialstatus, and employment status) of the association betweenperceptions of neighborhood safety and social cohesion, socialsupport, and social networks, respectively. None of these variableswere found to moderate the association between perceptions ofneighborhood safety and social cohesion for female residents.

4. Discussion

Although there has been a growing interest in the relationshipbetween people’s perceptions of specific characteristics of the

Table 2Bivariate model predicting neighborhood safety (predicting greater feelings of neighborhood safetya).

Males Females

Weighted n ORa (95% CI) Weighted n ORa (95% CI)

Demographics 504 1471

EducationLess than HS (Ref) 228 REF 541 REF

Completed HS/Voc 122 1.85 (1.24, 2.76)n 420 1.69 (1.20, 2.36)n

At least some college 131 2.19 (1.37, 3.48)n 509 1.82 (1.43, 2.32)n

Immigrant statusBorn in US 830 1.89 (1.06, 3.40)n

Born in Puerto Rico (ref) 353 REF

Not born in US or PR 288 1.11 (.54, 2.30)

English 1st languageYes 817 1.99 (1.53, 2.60)n

No (Ref.) 653 REF

Race/EthnicityBlack 739 1.96 (1.48, 2.60)n

Hispanic (Ref) 631 REF

White/Other 97 1.75 (.77, 3.95)

Ageo35 109 1.52 (0.86, 2.69) 359 1.31 (.97, 1.77)

35–49 109 1.01 (0.60, 1.70) 408 1.82 (1.36, 2.44)n

50+ (ref) 290 REF 703 REF

Employment statusWork full time/part time 178 1.43 (0.87, 2.35) 571 1.59 (1.27, 2.00)n

Disabled/not working (Ref) 331 REF 900 REF

Health problems make it difficult to exerciseYes (Ref) 186 REF 675 REF

No 323 1.55 (1.21, 1.99)n 794 1.38 (1.09, 1.75)n

Social networksBelow median split 216 1.28 (0.88, 1.86) 581 1.05 (.86, 1.28)

Median split or higher (Ref) 292 REF 884 REF

Social supportHas all5 356 1.08 (0.64, 1.81) 999 1.29 (1.03, 1.62)n

Does not have all 5 148 REF 462 REF

Social cohesionContinuous 482 1.49 (0.95, 2.35) 1345 1.64 (1.32, 2.05)n

n pr .05.a Polytomous model odds ratios represent odds for the 2 relationships: feeling safe vs. a little unsafe and unsafe, and feeling a little unsafe vs. unsafe.

M. De Jesus et al. / Health & Place 16 (2010) 1007–1013 1011

social environment and their perceptions of safety, very fewempirical studies have examined how perceptions of the socialenvironment might impact perceptions of perceived neighbor-hood safety, especially among public housing residents. Our studycontributes to the scientific literature in this area by providingsome insight into the association between social environmentalfactors—social networks, social support, and social cohesion—andperceptions of neighborhood safety among low-income publichousing residents as a potential pathway through which socialenvironmental factors influence health. As hypothesized, resi-dents who reported higher levels of social cohesion perceivedtheir neighborhood as safer compared to residents who reportedlower levels of social cohesion. This association fits previousempirical evidence linking social cohesion to neighborhood safety(e.g., Kuo et al., 1998; Sampson and Raudenbush, 1999; Wen et al.,2007; Baum et al., 2009).

Given the extensive literature on gender-based differences inperceptions of safety and the fear of personal and propertyvictimization (e.g., Warr, 2000; Mesch, 2000; Pain, 2001; Scott,2003; Lane and Meeker, 2003; Schafer et al., 2006), it is notsurprising that there were divergent findings for males andfemales in terms of effect of social networks on perceptions ofsafety. Contrary to our hypothesis, male residents with smallersocial networks were more likely to feel safe than thosewith larger social networks. It is possible that the associa-tion between social networks and perceived safety does notfollow a linear dose–response curve. Another possibility is that

larger social networks may be characterized by more negativesocial norms, influences, and interpersonal interactions, whichcontribute to increased perceptions of being unsafe. Indeed, thereis evidence that demonstrates that social networks may notalways serve positive functions (Berkman and Glass, 2000;Heaney and Israel, 2002). For female residents, our findingssuggest that social networks did not contribute to perceivedneighborhood safety, and that sociodemographic factors (e.g.,employment, education, and language) were more important.Female residents who had lower levels of education, wereborn outside the US, and who reported a language other thanEnglish as their first language were more likely to feel lower levelsof safety compared to each of their counterparts. Our studyfindings suggest that public policy and health promotionefforts need to first address female residents who are less well-educated, not born in the US, and whose first language is notEnglish. Although extant research has focused on explainingdifferences between, rather than within the genders, there issome evidence to suggest that minority women, and womenwith lower educational levels and lower incomes report greaterlevels of fear due to residential patterns and routine activities aswell as actual or perceived social and physical vulnerabilities(McGarrell et al., 1997; Goodey, 1997; Austin et al., 2002).Interestingly, health status was not a significant factor foreither male or female residents. However, this finding may bereflective of the fact that we had a very limited measure of healthstatus.

Table 3Multivariable Model Predicting Neighborhood Safety (predicting greater feelings of neighborhood safetya).

Males Females

Weighted n ORa (95% CI) Weighted n ORa (95% CI)

Demographics 504 1471

EducationLess than HS (Ref) 228 REF 483 REF

Completed HS/Voc 122 1.85 (1.24, 2.76)n 386 1.40 (.95, 2.04)

At least some college 131 2.19 (1.37, 3.48)n 476 1.46 (1.08, 1.97)n

Immigrant statusBorn in US

Born in Puerto Rico (ref)

Not born in US or PR

English 1st languageYes 760 1.69 (1.17, 2.45)n

No (Ref) 585 REF

Race/EthnicityBlack

Hispanic (Ref)

White/Other

Age

o35 331 1.25 (.88, 1.77)

35–49 387 1.85 (1.42, 2.41)n

50+ (ref) 627 REF

Employment statusWork full time/part time 516 1.28 (1.10, 1.48)n

Disabled/not working (Ref) 829 REF

Health problems make it difficult to exerciseYes (Ref)

No

Social networks

Below median split 202 1.61 (1.06, 2.53)n

Median split or higher (Ref) 279 REF

Social supportHas all 5

Does not have all 5

Social cohesionContinuous 481 1.64 (1.06, 2.53)n 1345 1.71 (1.39, 2.10)n

n pr .05.a Polytomous model odds ratios represent odds for the 2 relationships: feeling safe vs. a little unsafe and unsafe, and feeling a little unsafe vs. unsafe.

M. De Jesus et al. / Health & Place 16 (2010) 1007–10131012

Social support in this study was not significantly associatedwith perceived neighborhood safety. Most participants in thestudy actually reported fairly high levels of social support andthus, the lack of variability could, in part, explain some of thesenon-significant results. Our measure assessed overall socialsupport available through all aspects of one’s life, and not justthat available in one’s neighborhood. Further research examiningthe role of neighborhood-specific characteristics of social supporton perceptions of neighborhood safety is warranted. For example,does being able to count on neighbors for emotional, instru-mental, and financial support when one needs it contribute toperceptions of safety for residents?

4.1. Study strengths and limitations

To our knowledge, this study was the first to examine the role ofsocial networks, social support, and social cohesion in influencingpublic housing residents’ perceptions of neighborhood safety. Thisresearch was conducted among a large, random sample of diverseracial and ethnic, low-income male and female residents. Giventhat this study is cross-sectional however, causal interpretationscannot be made, and future studies should evaluate theseassociations longitudinally. Study findings are only generalizableto low income public housing residents living in the Bostonmetropolitan area. Another limitation was that social networkswere measured solely in terms of quantity of social ties

(e.g., number of close friends and relatives) and not quality ofsocial ties. Because we were specifically interested in examiningindividual perceptions of safety we did not use objective measuresof neighborhood safety (e.g., crime statistics). Further researchmeasuring both the perceived and objective environment is neededto assess the relative contribution of each construct to specifichealth behaviors and health outcomes. Finally, the response ratefor this study was lower than hoped, but should be considered inthe context of the challenges associated with conducting researchin community settings. Given the 53% response rate, the samplemay be potentially skewed in terms of representativeness of safety.However, other studies confirm that our sample is generallyrepresentative of public housing populations (i.e., low-income,predominately racial/ethnic minority women) (Lewis et al., 1993;Brunson et al., 2001; Peters et al., 2007; Manzo et al., 2008).

The findings reported here are also useful in exploring apotential pathway through which social environmental factorsinfluence health and suggest that an examination of residents’perceptions of social networks, social cohesion, and social supportis a helpful first step in untangling the complex set of variablesthat underpin perceptions of neighborhood safety. This papersuggests that social cohesion, in particular, may contribute to agreater sense of safety for residents, and potentially thereforehealth. Interventions therefore designed to improve health andwell-being for low-income residents in public housing couldusefully consider ways to promote a stronger sense of cohesionamongst residents.

M. De Jesus et al. / Health & Place 16 (2010) 1007–1013 1013

Funding

This research was supported by grants 5R01CA098864-02,1K22CA126992-01, and K05 CA124415 from the National CancerInstitute, and support to the Dana-Farber Cancer Institute byLiberty Mutual, National Grid, the Patterson Fellowship, and theYerby Fellowship Program.

Acknowledgments

We would like to profoundly thank the co-investigators of thisstudy: Gary Bennett, Sapna Syngal, Robert Mayer, and MarthaZorn. We also gratefully acknowledge the efforts of the OpenDoors to Health Research Team: Elise Dietrich, Elizabeth GonzalezSuarez, Terri Greene, Lucia Leone, Mike Massagli, VanessaMelamede, Maribel Melendez, Tamara Parent, Lina Rincon,Claudia Viega, Monifa Watson, Caitlin Gutheil, Zoe Bendixen,Roona Ray, Aidana Baldassrre, David Wilson, Ruth Lederman.Finally, we would also like to thank the resident helpers andresident service coordinators at collaborating housing sites.

References

Almedom, A., 2005. Social capital and mental health: an interdisciplinary review ofprimary evidence. Social Science and Medicine 61, 943–964.

Austin, D.M., Furr, L.A., Spine, M., 2002. The effects of neighborhood conditions onperceptions of safety. Journal of Criminal Justice 30, 417–427.

Barrera, M., Sandler, I.N., Ramsey, T.B., 1981. Preliminary development of a scale ofsocial support: studies on college students. American Journal of CommunityPsychology 58, 304–309.

Baum, F.E., Ziersch, A.M., Zhang, G., Osborne, K., 2009. Do perceived neighbourhoodcohesion and safety contribute to neighbourhood differences in health? Health& Place 15 (4) 925–934.

Bennett, G.G., McNeill, L.H., Wolin, K.Y., Duncan, D.T., Puleo, E., Emmons, K.M.,2007. Safe to walk? Neighborhood safety and physical activity among publichousing residents. PLoS Med 4 (10), 1599–1606.

Berkman, L.F., Syme, S.L., 1979. Social networks, host resistance, and mortality: anine-year follow-up study of Alameda County residents. American Journal ofEpidemiology 109 (2), 186–204.

Berkman, L., Glass, T., 2000. Social integration, social networks, social support, andhealth. In: Berkman, L.F., Kawachi, I. (Eds.), Social Epidemiology. OxfordUniversity Press, New York, pp. 137–173.

Brunson, L., Kuo, F.E., Sullivan, W.C., 2001. Resident appropriation of defensiblespace in public housing: implications for safety and community. Environmentand Behavior 33 (5), 626–652.

Campbell, K.E., Lee, B.A., 1992. Sources of personal neighbor networks: Socialintegration, need, or time? Social Forces 70 (4), 1077–1100.

Cattell, V., 2001. Poor people, poor places and poor health: the mediating role ofsocial networks and social capital. Social Science and Medicine 52, 1501–1516.

Chandola, T., 2001. The fear of crime and area differences in health. Health & Place7, 105–116.

Chaskin, R.J., Joseph, M.L., 2010. Building ‘‘community’’ in mixed-incomedevelopments: assumptions, approaches, and early experiences. Urban AffairsReview 45, 299–307.

Cohen, S., Syme, S.L., 1985. Issues in the application and study of social support. In:Cohen, S., Syme, S.L. (Eds.), Social Support and Health. Academic Press,Orlando.

De Silva, M.J., McKenzie, K., Harpham, T., Huttly, S.R.A., 2005. Social capital andmental illness: a systematic review. Journal of Epidemiology and CommunityHealth 59, 619–627.

Ellison-Loschmann, L., Sunyer, J., Plana, E., Pearce, N., Zock, J.P., et al., 2007.Socioeconomic status, asthma and chronic bronchitis in a large community-based study. European Respiratory Journal 29, 897–905.

Goodey, J., 1997. Boys don’t cry: masculinities, fear of crime and fearlessness.British Journal of Criminology 37, 401–418.

Heaney, C.A., Israel, B.A., 1997. Social networks and social support in healtheducation. In: Glanz, K., Lewis, F.M., Rimer, B.K. (Eds.), Health Behavior andHealth Education: Theory, Research and Practice. Jossey-Bass, San Francisco.

Heaney, C.A., Israel, B.A., 2002. Social networks and social support. In: Glanz, K.,Rimer, B.K., Lewis, F.M. (Eds.), Health Behavior and Health Education: Theory,Research and Practice. Jossey-Bass, San Francisco, pp. 185–209.

HHS, 2005. Annual Update of the HHS Poverty Guidelines. Federal Registry,8373-8375.

Hurlbert, J.S., Beggs, J.J., Haines, V.A., 2001. Social networks and social capitalin extreme environments. In: Lin, N., Cook, K., Burt, R.S. (Eds.), SocialCapital Theory: Theory and Research. Transaction Publishers, New Jersey,pp. 209–232.

Kawachi, I., Berkman, L., 2000. Social cohesion, social capital, and health. In:Berkman, L.F., Kawachi, I. (Eds.), Social Epidemiology. Oxford University Press,New York, pp. 174–190.

King, D., 2008. Neighborhood and individual factors in activity in older adults:results from the neighborhood and senior health study. Journal of Aging andPhysical Activity 16, 144–170.

Kleit, R.G., 2001. Neighborhood relations in suburban scattered-site and clusteredpublic housing. Journal of Urban Affairs 23 (3-4), 409–430.

Kuo, F.E., Sullivan, W.C., Coley, R.L., Brunson, L., 1998. Fertile ground forcommunity: inner-city neighborhood common spaces. American Journal ofCommunity Psychology 26 (6), 823–851.

Lane, J., Meeker, J.W., 2003. Women’s and men’s fear of gang crimes: sexual andnonsexual assault as perceptually contemporaneous offenses. Justice Quarterly20, 337–371.

Lewis, C.E., Raczynski, J.M., Heath, G.W., Levinson, R., Cutter, G.R., 1993. Physicalactivity of public housing residents in Birmingham, Alabama. American Journalof Public Health 83 (7), 1016–1020.

Lochner, K., Kawachi, I., Kennedy, B.P., 1999. Social capital: a guide to itsmeasurement. Health and Place 5 (4), 259–270.

Macintyre, S., Ellaway, A., 2000. Ecological approaches: rediscovering the role ofthe physical and social environment. In: Berkman, L.F., Kawachi, I. (Eds.), SocialEpidemiology. Oxford University Press, New York, pp. 332–348.

Manzo, L.C., Kleit, R.G., Couch, D., 2008. ‘‘Moving three times is like having yourhouse on fire once’’: the experience of place and impending displacementamong public housing residents. Urban Studies 45 (9), 1855–1878.

McCullagh, P., 1980. Regression models for ordinal data. Journal of ResearchStatistics Society 42, 109–142.

McGarrell, E.F., Giacomazzi, A.L., Thurman, Q.C., 1997. Neighborhood disorder,integration, and the fear of crime. Justice Quarterly 14, 479–497.

Mesch, G.S., 2000. Perceptions of risk, lifestyle activities, and fear of crime. DeviantBehavior 21, 47–62.

Middleton, J., 1998. Crime is a public health problem. Medicine, Conflict andSurvival 14, 24–28.

Pain, R., 2001. Gender, race, age, and fear of the city. Urban Studies 38, 899–913.Pebley, A., Narayan, S., 2003. The Los Angeles Family and Neighborhood Survey:

Household Questionnaires, RAND Working Paper.Peters, J.L., Levy, J.I., Rogers, C.A., Burge, H.A., Spengler, J.D., 2007. Determinants of

allergen concentrations in apartments of asthmatic children living in publichousing. Journal of Urban Health 84 (2), 185–197.

Rountree, P., Land, K., 1996. Perceived risk versus fear of crime: empirical evidenceof conceptually distinct reactions in survey data. Social Forces 74, 1353–1376.

Sampson, R.J., Raudenbush, S.W., Earls, F., 1997. Neighborhoods and violent crime:a multilevel study of collective efficacy. Science 277, 918–924.

Sampson, R., Raudenbush, S.W., 1999. Systematic social observation of publicspaces: a new look at disorder in urban neighborhoods. American Journal ofSociology 105, 603–651.

Sampson, R.J., 2003. Neighborhood-level context and health: lessons fromsociology. In: Kawachi, I., Berkman, L.F. (Eds.), Neighborhoods and Health.Oxford University Press, New York, pp. 132–146.

Schafer, J.A., Huebner, B.M., Bynum, T.S., 2006. Fear of crime and criminalvictimization: gender-based contrasts. Journal of Criminal Justice 34 (3),285–301.

Scott, H., 2003. Stranger danger: explaining women’s fear of crime. WesternCriminology Review 4, 203–214.

Stokes, M.E., Davis, C.S., Koch, G.G., 2000. Categorical data analysis using the SASsystem. SAS Institute Inc., Cary.

Tan, H., Choo, W., Archibald, C., Esuvaranathan, K., 1997. A community-based study of prostatic symptoms in Singapore. Journal of Urology 157 (3),890–893.

Warr, M., 2000. Fear of crime in the United States: avenues for research and policy.In: Duffee, D. (Ed.), Criminal Justice: Measurement and Analysis of Crime andJustice. National Institute of Justice, Washington, DC, pp. 451–490.

Wen, M., Kandula, N.R., Lauderdale, D.S., 2007. Walking for transportation orleisure: What difference does the neighborhood make? Journal of GeneralInternal Medicine 22 (12) 1674–1680.

Young, A.F., Russell, A., Powers, J.R., 2004. The sense of belonging to aneighbourhood: can it be measured and is it related to health and well beingin older women? Social Science and Medicine 12, 2627–2637.

Ziersch, A.M., Baum, F.E., MacDougall, C., Putland, C., 2005. Neighborhood life andsocial capital: the implications for health. Social Science & Medicine 60 (1),71–86.

Ziersch, A., Putland, C., Palmer, C., MacDougall, C., Baum, F., 2007. Neighborhoodlife, social capital and perceptions of safety in the Western suburbs ofAdelaide. Australian Journal of Social Issues 42 (4), 550–562.