Social Support Buffers the Effects of Terrorism onAdolescent Depression: Findings From Sderot, Israel

CHRISTOPHER C. HENRICH, PH.D., AND GOLAN SHAHAR, PH.D.

ABSTRACT

Objective: This prospective study of 29 Israeli middle school students experiencing terror attacks by Qassam rockets

addressed whether higher levels of baseline social support protected adolescents from adverse psychological effects of

exposure to rocket attacks. Method: Participants were assessed at two time points 5 months apart, before and after a

period of military escalation fromMay to September 2007. Adolescent self-reported depression was measured at both time

points, using the Center for Epidemiological Studies-Child Depression Scale. Social support from family, friends, and

school was measured at time 1, via a short form of the Perceived Social Support Scale. Adolescents also reported their

exposure to rocket attacks at both time points. Results: There was a significant interaction between social support and

exposure to rocket attacks predicting depression over time. As hypothesized, baseline levels of social support buffered

against the effect of exposure to rocket attacks on increased depression. Conversely, social support was associated with

increased depression for adolescents who were not exposed to rocket attacks. Conclusions: Findings highlight the

potential importance of community mental health efforts to bolster schools, families, and peer groups as protective

resources in times of traumatic stress. J. Am. Acad. Child Adolesc. Psychiatry, 2008;47(9):1073Y1076. KeyWords: social

support, early adolescence, depression, terrorism, stress buffering.

Youths exposed to acts of terrorism are at heightenedrisk of a host of internalizing problems, includingposttraumatic stress, anxiety, and depression.1 However,most research on youth exposure to terrorism is limitedto postattack data, and relatively little is known aboutprotective factors that moderate youths’ response toterrorism.1,2 This study represents the results of aunique opportunity to longitudinally study a smallsample of Israeli adolescents for 5 months during whichtime their community was subjected to a barrage of

Qassam rocket attacks launched from the neighboringGaza Strip. We investigated the effects of exposure tothe rocket attacks with a focus on the potential stress-buffering effect of social support.3Y5 The stress-bufferingmodel highlights the role of social support in promotingpsychological and physical health in the face of stress-ful life events.4 Stress can affect physical and mentalhealth adversely via numerous mechanisms, includingthe promotion of maladaptive coping strategies andthe activation of physiological pathways implicated inillness and emotional distress. According to the stress-buffering model, social support can protect individualsfrom these adverse health effects of stress exposure byhelping to either mitigate the stressor or make thestressor seem less important and by promoting effectivecoping strategies.4 Furthermore, according the stress-buffering model, the health benefits of social supportshould be found only in conditions of stress.4 Thus,social support and stress exposure should have inter-active effects on psychological and physical health.The goal of this small-scale prospective study was to

test the hypothesized stress-buffering effect of socialsupport in the context of terrorism. The sample

Accepted March 24, 2008.Dr. Henrich is with the Department of Psychology, Georgia State University;

Dr. Shahar is with the Department of Psychology, Ben-Gurion University of theNegev.

This study was funded by a grant from the Israel-U.S. Binational ScienceFoundation to Drs. Shahar (Principal Investigator) and Henrich (Co-PrincipalInvestigator).

This article is the subject of an editorial by Dr. Judith A. Cohen in this issue.Correspondence to Dr. Christopher C. Henrich, Department of Psychology,

Georgia State University, P.O. Box 5010, Atlanta, GA 30302-5010; e-mail:chenrich@gsu.edu.

0890-8567/08/4709-1073�2008 by the American Academy of Child andAdolescent Psychiatry.

DOI: 10.1097/CHI.0b013e31817eed08

WWW.JAACAP.COM 1073J . AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 47:9, SEPTEMBER 2008

comes from a larger investigation of processes of riskand resilience among adolescents from economicallydisadvantaged communities in the Negev region ofsouthern Israel. One of the communities in the largerstudy is Sderot, a town located 1 km from the GazaStrip. Although Sderot has faced ongoing threats fromQassam rockets since 2000, in May 2007 an increase inthe number of rockets fired at Sderot prompted schoolclosures and a number of families to leave town whileour data collection was under way. Once the increasedrocket attacks subsided, follow-up data were collectedon a small sample of Sderot adolescents for whombaseline data had been collected immediately before theincrease in terror violence. We tested the stress-bufferinghypothesis by investigating whether adolescents report-ing higher levels of baseline social support wereprotected from adverse psychological effects of exposureto rocket attacks on increased depression.

METHOD

Permission for conducting the original larger study was securedfrom the Israeli Ministry of Education, schools, and the UniversityDepartmental Review Committee. Data collection began in May2007. On the first day of the study in Sderot, data were collectedfrom 45 students with active parental consent in one of the town’spublic religious middle schools. Under the supervision of trainedresearch assistants, students completed a 45-minute battery ofquestionnaires assessing depressive and anxious symptoms, violenceexposure and commission, personality, life events, and exposure toQassam rockets. During data collection, the research team heardwarning sirens, and we later learned that several rockets had hit thetown that day.A serious military escalation transpired the next day, leading to the

discontinuation of the study in Sderot. We secured permission fromthe Ministry of Education to reinterview participants on thetelephone once the escalation subsided. These follow-up data werecollected in September 2007. The telephone interview, conductedby trained research assistants, assessed depressive symptoms andexposure to Qassam rockets during the escalation period. Of the 45students participating at time 1, 30 also participated at time 2, and29 completed the full interview. Of the remaining 15, one refused tobe interviewed at time 2, and the rest were unreachable despiteseveral attempts to contact them. They may have permanentlyrelocated when the violence escalated.At time 1, the sample of 45 students was 64% female (n = 29).

Twenty-five students (56%) were in seventh grade, with theremainder in eighth grade. Ages ranged between 11 and 14 (mean12.63, SD 0.68). Parents of most participants were married (n = 37;82%), parents of seven participants were divorced, and oneparticipant had a single parent who was a widow. At time 1, 56%(n = 25) of the participants reported having been present at aprevious terror/rocket attack.The sample at time 2 was similar demographically. Nineteen

(66%) participants were female, and 18 (62%) were seventh graders.Ages ranged between 12 and 14 (mean 12.58, SD 0.62). Parents of23 participants (79%) were married, parents of seven participants

were divorced, and one participant had a single parent who was awidow. Fifty-nine percent (n = 17) of the sample who were retainedat time 2 had reported being present at a terror/rocket attack beforetime 1. Chi-square analyses indicated that the participants remainingin the study through time 2 did not differ statistically from thosewho dropped out across any of the demographic variables.Depression was measured by the Center for Epidemiological

Studies-Child Depression Scale (CES-CD),6 an adaptation of theadult CES-D.7 Twenty items assess symptoms of depression, with afocus on depressed mood (e.g., ‘‘I felt sad’’), on a 0 to 3 scale. Itemsare summed to form a total depression score. At time 1, the averageCES-CD depression score was 13.44 (SD 8.20), and 38% of thesample (n = 11) exceeded the clinical cutoff score of 16. The averagescore at time 2 was 16.44 (SD 8.70), and 52% (n = 15) exceeded theclinical cutoff. Participants who dropped out of the study by time 2did not differ statistically from those remaining in terms of time 1depression scores, t(42) = 0.80, p = .43.Perceived social support was measured by an abbreviated form of

the Perceived Social Support Scale,8 which assesses support providedby friends, family, and school personnel. Our abbreviated formincluded 16 items, six each for the friend and family subscales andfour for the school subscale. For each item, respondents hadto endorse a ‘‘yes,’’ ‘‘no,’’ or ‘‘don’t know’’ response. An internallyconsistent total scale score was created by summing ‘‘yes’’ responsesacross the 16 items, " = .81. The average social support score was9.62 (SD 3.48). Participants who dropped out of the study by time2 did not differ statistically from those who stayed in terms of socialsupport, t(42) = 1.15, p = .23.Previous exposure to terrorism was assessed through one item

at time 1 asking whether participants had ever been present at aterror/rocket attack. Exposure to Qassam rocket attacks wasassessed at time 2 by six items adapted from previous research inIsrael9 to assess the types of attacks participants were most likelyto have experienced during the 5-month period of militaryescalation. Items asked whether participants had been injured, hadfriends or family injured, had property damaged, had left home,had friends or family whose property was damaged, and hadfriends or family who had been affected mentally by the rocketattacks in the past 5 months. The average exposure sum score attime 2 was 2.10 (SD 1.45).

Data Analysis

Multiple regression analysis was used to test whether time 1 totalsocial support buffered against an adverse effect of exposure torocket attacks between time 1 and time 2 on increased depressivesymptoms at time 2. This analysis tested the stress-bufferinghypothesis by examining whether there was a statistical interactionbetween social support and exposure to rocket attacks, such thatthe effects of exposure to rocket attacks on depression weredependent on the level of social support. In the regression model,time 2 depression was regressed on time 1 social support, exposureto rocket attacks between time 1 and time 2, and a social supportby exposure to rocket attacks product interaction term, controllingfor time 1 depression and whether participants reported havingbeen present at a terror/rocket attack at time 1. All of the predictorvariables were entered simultaneously in one step. Followingguidelines for testing interaction effects in multiple regression, thesocial support and exposure variables were mean centered.10 Beforethe regression analysis was conducted, all of the continuous var-iables in the model were tested for univariate normality. None wassignificantly skewed or kurtotic.

HENRICH AND SHAHAR

1074 WWW.JAACAP.COM J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 47:9, SEPTEMBER 2008

RESULTS

The multiple regression model explained 38% ofthe sample variance in time 2 depression (F5,23 = 2.91;p = .04). The social support by exposure interactionterm uniquely explained 18% of the variance in time2 depression ( p = .02), which according to Cohen’sguidelines is a large effect size.10 Regression coefficientsare presented in Table 1.

The buffering effect of social support on the effect ofexposure to rocket attacks on depression was probed byrecentering the social support variable in the multipleregression model to estimate the effects of exposure atlow and then high levels of social support.11 Resultsindicated that at low levels of social support (j1 SD),greater exposure to rocket attacks was associated withincreased depression, B = 3.57, SE 1.56, p = .03,whereas at high levels of social support (1 SD), there wasa trend for greater exposure to rocket attacks to beassociated with decreased depression (B = j3.21, SE1.75, p = .08). This interaction is graphed in Figure 1.The figure also illustrates a crossover pattern withrespect to the association between social support anddepression: When exposure to rocket attacks was high(i.e., 1 SD), there was a trend for more social support tobe associated with decreased depression over time(B = j1.47, SE 0.83, p = .09). Conversely, whenexposure to rocket attacks was low (j1 SD), more socialsupport was associated with increased depression overtime (B = 1.40, SE 0.57, p = .02).

DISCUSSION

Results supported the hypothesized stress-bufferingeffect of social support in the context of terrorism. Tothe best of our knowledge, these are the first findingsdocumenting the protective effect of social support forterrorized adolescents based on a longitudinal study

design in which social support was assessed before theputative stress (i.e., exposure to terrorism) and theoutcome (i.e., time 2 depression). Because analysescontrolled for time 1 depression, they represent arigorous test of the interactive effects of social supportand terrorism exposure on change over time indepression. These findings add to the stress-bufferingliterature and attest to the key protective roles offamily, school, and peer relationships during adoles-cence.12,13 Because families in the sample chose to sendstudents to a religious public school, the protective effectof social support may also be attributed in part toreligious support.14

We also found some evidence for a crossoverinteraction, in which social support predicted increaseddepression over time for participants who were notexposed to terrorism. Such crossover effects are notuncommon in the stress-buffering literature, suggestingthat some protective factors (e.g., social support) arebeneficial under conditions of stress but can beassociated with elevated distress in the absence ofstress.5,15,16 Furthermore, this study’s population is notonly terrorized but also suffers from a host of poverty-related stressors. It may be that higher levels of socialsupport reflect in part a heightened attention of socialnetworks to adolescents who are becoming increasinglydepressed over time due to chronic stressors other thanexposure to terrorism. It is also possible that underconditions of chronic versus acute stress, the course ofdepression and function of social support operatedifferently. For the adolescents who were not exposedto terrorism, higher levels of social support may actuallyanticipate an evolving course of depression due topoverty-related factors, as opposed to buffering againstthe immediate effects of acute stress for adolescents

TABLE 1Multiple Regression Model Predicting Time 2 Depression (N = 29)

B SE p

Previous terror exposure, time 1 3.16 2.99 .30Depression, time 1 0.44 0.19 .03Social support, time 1 j0.03 0.46 .94Exposure, time 2 0.18 1.05 .87Social support time 1 � exposure, time 2 j0.99 0.38 .02

Note: Predictor variables entered simultaneously. Effects of p < .05interpreted as statistically significant. Exposure= exposure to rocketattacks; B = unstandardized coefficients.

Fig. 1 Time 1 social support moderates the effects of exposure to rocketattacks between time 1 and time 2 on time 2 depression.

SOCIAL SUPPORT AND TERRORISM

WWW.JAACAP.COM 1075J . AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 47:9, SEPTEMBER 2008

who were exposed to terror attacks. (For a theoreticaldiscussion of the difference between course and cause inpsychopathology, see Shahar et al.17) This possibilityshould be investigated in future research.The study’s limitations include the small sample size

and reliance on self-report data from students in onemiddle school in Sderot. Furthermore, the exposure toterrorism measure was changed at time 2 to better reflectthe specific experiences of terrorism facing Sderotadolescents during the 2007 escalation. These limitationsplace restrictions on the generalizability of our findingseven to other Israeli samples, yet at the same time theyencourage further inquiries into the protective effect ofsocial support on trauma-related depression and otherinternalizing problems not assessed in the present study,such as anxiety and posttraumatic stress. These find-ings, should they be replicated, highlight the potentialimportance of communitymental health efforts to bolsterschools, families, and peer groups as protective resourcesto help prevent depression in times of traumatic stress. Inaddition, in this study, social support was only measuredat time 1, so our findings represent the protective effectsof social support networks as they existed at the startof the escalation of violence. It is likely that the periodof heightened rocket attacks caused some degree ofdisruption and stress to these social support networks.Investigating the effects over time of terrorism onadolescents’ social resources and their protective benefitsis an important avenue for future research.

Disclosure: The authors report no conflicts of interest.

REFERENCES

1. Comer JS, Kendall PC. Terrorism: the psychological impact on youth.Clin Psychol Sci Pract. 2007;14:179Y212.

2. La Greca AM. Understanding the psychological impact of terrorism onyouth: moving beyond posttraumatic stress disorder. Clin Psychol SciPract. 2007;14:219Y223.

3. Alloway R, Bebbington P. The buffer theory of social support: a review ofthe literature. Psychol Med. 1987;17:91Y108.

4. Cohen S. Social relationships and health. Am Psychol. 2004;59:676Y684.5. Cohen S, Wills TA. Stress, social support, and the buffering hypothesis.

Psychol Bull. 1985;98:310Y357.6. Weissman MM, Orvaschel H, Padian N. Children symptom and social

functioning self-report scales: comparison of mothers and children’sreports. J Nerv Ment Dis. 1980;168:736Y740.

7. Radloff L. The CES-D scale: a self-report depression scale for research inthe general population. Appl Psychosocial Meas. 1977;1:385Y401.

8. Dubois DL, Felner RD, Sherman MD, Bull CA. Socio-environmentalexperiences, self-esteem, and emotional/behavioral problems in earlyadolescence. Am J Community Psychol. 1994;22:371Y397.

9. Kirschenbaum A. Terror, adaptation and preparedness: a trilogy forsurvival. J Homeland Security Emerg Manage [serial online]. 2006;3(1):article 3. http://www.bepress.com/jhsem/vol3/iss1/. Accessed December 18,2007.

10. Aiken LS, West SG. Multiple Regression: Testing and InterpretingInteractions. Newbury Park, CA: Sage; 1991.

11. Cohen J. Statistical Power Analysis for the Behavioral Sciences. Hillsdale,NJ: Lawrence Erlbaum; 1988.

12. Kuperminc GP, Leadbeater BJ, Blatt SJ. School social climate andindividual differences in vulnerability to psychopathology among middleschool students. J Sch Psychol. 2001;39:141Y159.

13. Resnick MD, Ireland M, Borowsky I. Youth violence: what protects?What predicts? Findings form the National Longitudinal Study ofAdolescent Health. J Adolesc Health. 2004;35:424Y434.

14. Hill PC, Pargament KI. Advances in the conceptualization andmeasurement of religion and spirituality. Am Psychol. 2003;58:64Y74.

15. Cohen S, Hoberman HM. Positive events and social supports as buffersof life change stress. J Appl Soc Psychol. 1985;13:99Y125.

16. Gonzales NA, Tein JY, Sandler IN, Friedman RJ. On the limits ofcoping: interactions between stress and coping for inner city adolescents.J Adolesc Res. 2001;16:372Y395.

17. Shahar G, Bareket L, Rudd MD, Joiner TE Jr. In severelysuicidal young adults, hopelessness, depressive symptoms, and sui-cidal ideation constitute a single syndrome. Psychol Med. 2006;36:913Y922.

HENRICH AND SHAHAR

1076 WWW.JAACAP.COM J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 47:9, SEPTEMBER 2008