Professional Documents
Culture Documents
Too Many Friends - Social Integration, Network Cohesion and Adolescent Depressive Symptoms - Falci & McNeely (2009)
Too Many Friends - Social Integration, Network Cohesion and Adolescent Depressive Symptoms - Falci & McNeely (2009)
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
Oxford University Press is collaborating with JSTOR to digitize, preserve and extend access to Social Forces.
http://www.jstor.org
Theauthors funding
acknowledge
gratefully fromtheWilliamT. GrantFoundation. We
wouldalsoliketoexpress to
thanks JimMoody his
sharing
forgraciously SAS for
Programs
AnalyzingNetworks toChristina
usersmanualDirectcorrespondence Falci,University
of
Nebraska-Lincoln, 711OldfatherHall,
ofSociology,
Department NE 68588-0324.
Lincoln,
E-mail:cfalci2@unl.edu.
of NorthCarolinaPress
© The University Social Forces 87(4): 203 1-62.June 2009
and DepressiveSymptoms
Under-integration
Adolescentswhoare under-integrated (i.e.,theyhaveveryfewornofriends)
are at greaterriskfordepressivesymptoms(Brendgenet al. 2000; Ueno
2005). Adolescents seek social connectionwithpeers (Baumeisterand
75OneFriend
Enough?
Some arguethata singleclose friend
can providesufficient
intimacy,support
and companionshipforan adolescent'swell-being(Baumeisterand Leary
1995). We hypothesizethatsmall networksize contributes to depressed
mood even when an adolescent has a reciprocatedclose friendship.
we do notexpectone close friend
Essentially, to meetallofan adolescent's
needs forsocial connection.A singlefrienddoes notgiveaccess to social
channels,bothofwhichhelpa studentfitinat school
statusor information
(Crosnoeand McNeelyin press; Walker,Wassermanand Wellman1993).
Thereis previousempiricalsupportforthishypothesis(Ueno 2005).
MeditationofUnder-Integration
byPerceptions and Support
ofBelonging
fromFriends
We examineperceptionsof belongingand supportfromfriendsas two
mediatingmechanismsby whichunder-integration mightlead to higher
levels of depressivesymptomsamong adolescents (House et al. 1988;
Ueno 2005). Perceivedbelongingat school is the sense of being a part
of the social fabricat school, of fitting
in. Not havingfriendsto sit with
in the lunchroom or to pass notes to in class can underminefeelings
of belonging.Seeminglyinnocuous moments,such as passing time or
choosingteams fora class project,become laden withthe potentialfor
feelingsof rejectionand isolation.Thus,adolescentswithfewfriendsare
less likely
to feelthattheybelongat school. Perceivedsupportfromfriends
is theextenttowhichadolescentsbelievethattheirfriends careaboutthem.
Adolescentswithfewfriendsmightperceiveless supportthanadolescents
withmorefriends.In lightofthese predictions, we expectthatperceived
belongingand friendsupportwillmediatethe relationship betweensmall
network size and depressivesymptomsinadolescence. Previousresearch
has demonstrated thatsupportfromfriendsand a sense of belongingare
inverselyrelatedwithdepressive symptoms(Laible,Carlo and Farraelli
2000; McNeelyand Falci2004). Furthermore, usingthe same Add Health
and DepressiveSymptoms
Over-Integration
Does theEffect
ofOver-Integration Cohesionand Gender?
VarybyNetwork
Previousresearch has focused on the independenteffectsof distinct
networkcharacteristics; however,the negativeeffectof havingtoo many
friendsmaydepend on levelsofnetworkcohesion.Networkcohesioncan
be representedas a continuum fromlowto highcohesion.Atone extreme,
an adolescent mighthave a completelyfragmentedlocal networkwhere
noneoftheadolescent's friendsare friendswithone another.Attheother
extreme,an adolescent could have a closed network,where all of the
adolescent's friendsnominateeach otheras a friend.Adolescentstend
to fallsomewhere in the middleof these two extremes,butthe former
extremeis morecommonthanthe latter(see appendixA).
One can understandthe importof the cohesiveness of network
structureintuitivelyby lookingat friendshipstructurevisually.Figure1
shows two largefriendship networksof equal size (15 actors),butwith
varyinglevels of network cohesion. The networkin Panel A has low
withVaryingLevelsofAlter-Density
Figure1. AlterNetworks
PanelA PanelB
= 15%
Alter-density = 50%
Alter-density
Large Network
Fragmented Cohesive
Large Network
MediationofOver-Integration
bySocialSupportand SocialBelonging
Previousresearchconsistentlydocuments two linearrelationships:(1.
as networksize increases,so do adolescents' perceptionsof belonging
and support (Haines, Beggs and Hurlbert2002; Walkeret al. 1993);
and (2. as perceptionsof belongingand supportincrease, depressive
symptomsdecrease (Laibleet al. 2000; McNeelyand Falci2004). Ifthis
is the case, then perceptionsof belongingand supportcannotmediate
the hypothesizedassociation between over-integration and depressive
symptoms.Perceptionsof belongingand supportcan onlymediatethe
illeffectsof over-integration
ifover-integration
leads to lower levels of
belonging and support.Although predictpositivelinearrelationships
we
betweennetworksize and perceptionsof belongingand support,we test
the competinghypothesisthatover-integration compromisesperceived
peer supportand the sense of belonging.
TheCurrentStudy
Our researchassesses forthe firsttime the possibilityof a curvilinear
relationshipbetween social integrationand depressive symptomsin
adolescence. Both under-integrated and over-integrated adolescents
are hypothesizedto reporthigherlevels of depressive symptomsthan
adolescents withaverage-sized social networks.However,the effect
of over-integration
on depressive symptomswillvaryas a functionof
both networkcohesion and gender. For girls,large networkswill not
compromisewell-beingiftheyare cohesive. Forboys,however,network
cohesion willnotprotectagainstthe negativeeffectsof over-integration.
Finally,perceptionsof friendsupportand belongingwill mediate the
association between under-integrationand depressive symptoms,but
notthe association between over-integration
and depressivesymptoms.
Higherlevelsofdepressivesymptomsamongover-integrated adolescents
probablyresultfromhigherlevelsofrolestrain,althoughitis notpossible
to testthispotentialmechanismwiththe data used inthisstudy.
Methods
Sample
AddHealthis a stratified
sampleof132 juniorand seniorhighschools inthe
UnitedSates (Udry2003). An in-schoolsurveywas administeredin 1994.
All social networkmeasures are created fromfriendshipsnominations
Measures
Analytic
Strategy
The social networkmeasureswere createdusingPROC IMLproceduresin
SAS 9.1. OLS regressionmodelstesthypothesesregarding thecurvilinear
relationshipbetween social integrationand depressive symptoms.All
analyses are run in SAS 9.1 and adjusted forAdd Health's complex
samplingdesign (Chantala2006). Specifically, all analyses are weighted
to adjust forover-samples and nonresponse,and the standarderrors
are adjusted to take intoaccount the stratified samplingplan and the
clustering of students within schools.
The jointtest of curvilinearrelationshipsand interactioneffectsfor
degree, alter-density and gender requiredtestinga four-way interaction.
Inclusionof multipleinteractionterms poses the potentialproblemof
multicollinearity. Several approaches explore potentialmulticollinearity
problems.First, network size is transformed to Z-scoresand alter-density
is mean-centered(Jaccard, Wan and Turrisi1990). The VIFs forthe
interaction termsin the four-way model ranged2-5, whichare highbut
below the acceptable thresholdof 10 (Hairet al. 2006). Second, because
of the presence of multicollinearity, the stabilityof the beta coefficients
is assessed by runningthe four-wayinteractionmodel on randomly
selected sets of halfof the analyticsample (Echambadiand Hess 2007).
These subset analyses replicatedourresults.Third,onlyinteraction terms
in an
thatexplainedadditionalvariance the model,using F-test, kept are
(Kromrey and Foster-Johnson 1998). Finally,additionalanalyses stratified
bygender and levels ofalter-density verifiedthe resultsoftheinteractions.
Applying ordinary least squares regression a skewed dependentvari-
to
able raisestheconcernofpossiblespuriousinteraction terms(Osgood,Fink-
en and McMorris 2002; Haynieand Osgood 2005). Forthisreason,theCESD
is transformed usingIRTmethods(thegradedresponsemodel; Samejima
1969) in Mplus.Then,all analyseswere duplicatedusingthe transformed
CESD (i.e., factorscores fromthe Mplus confirmatory factoranalysis)in
a Tobitregressionmodelwiththe IVE-wareSAS moduleto adjustforthe
complexsurveydesign(Raghunathan, Solenbergerand Van Hoewyk2002).
in
The resultsdid notdiffer significance or effectsize. Forsimplicity and in-
the resultsfromthe OLS regressionare presented.Finally,
terpretability, to
investigate the possibility thatunexplained variation inindividualoutcomes
might be due to unspecified differencesbetween schools random effects
models were estimatedin Stata 9 using the xtreg command. Again,the
resultsdid notdiffer in anysubstantiveway fromthe resultsobtainedus-
ingtheSAS surveyreg procedures.To do an additionalcheckofschoolsize
(range = 26 to more than 3,000 students),analyseswere runon a sample
witha minimum schoolsize of500 students,and theresultsdidnotchange.
Results
Network Statistics
Descriptive
Table 1 reportsweighteddescriptivestatistics.On average,adolescents
nominatedor receivednominationsfromalmost eightfriendsin school
NetworkStructure
and DepressiveSymptoms
Table 1: WeightedDescriptiveStatistics
St.d. Min Max
Mean/Proportion
Depressive Symptoms 9.61 6.10 1 43
Friendship Network Structure
Networksize3 8.93 4.42 1 34
Alter-density .21 0 1
Reciprocatedschool closefriend .45 0 1
Unreciprocatedschool close
friend .18 0 1
Non-school closefriend .24 0 1
Noclosefriend .12 0 1
Network Control Variables
#Non-school friends 2.14 2.21 0 10
#Missingschool friendnomination .25 .57 0 5
SocialPerception
Perceivedbelonging 9.49 2.55 1 13
Perceivedfriendsupport 4.28 .77 1 5
Demographic Characteristics
Female .52 0 1
Grade 9.35 1.62 6 12
White .75 0 1
Black .17 0 1
Latino .07 0 1
Household income (in$1,000s) 48.12 49.03 1 1000
Missingonhousehold income .20 0 1
Schoolsize 878.14 780.15 26 3334
#ofyears atcurrentschool 2.84 1.61 1 6
#ofyears atcurrent residence 6J54 5.71 0 19
Notes:Standarddeviationsareonlyreportedfornon-dummy variables.
aThecountfornetwork sizeincludesthefocaladolescent.
N = 9097
2J
2U
2 io
5fe§SSSSSSS5oS
i"
• *
i" i" r i* i i i i
I
ja
"a>
e ^-^^f^^^CM^-vOO ^^t^- ^^CM ^-.'«- ^-.CD ^^O) ^^LO ^-^CO ^-^OO ^-
^T-cNio^cN^^^c\jir>co^coirjr^ocz>cocDiqo>cocq
§ ©OO C5OC5OO
£
2
SE .q t- ^co in" oo f^c3^ o"o Pco ^cp r^
c ^> o o "
o o cz>
CO J2 i" i" i" i"
~^
CO "55
&- •Co** -»-
Q
O
&
2
I OJ
t5
^
o
i"
o
i"
o
i"
1o .* .* *
o
U CO CVj CO CO CO CO
I iS S
% o
CD
o
1 t^i>^
« u cn oco
"^ -r-; t-|
in
O
o
"3
.3
I
■S -2 o
t
CO "O CO >
3 * *c
"-it
S .3 .3 © 8
I
3 3 3
« I SI
£s co too"ov. co</>co
I ii'llii
to
3 ^
Q. N
'CO
^-^-v^-vf"^
N
'CO
N
'CO
N
'CO
^S
16
O
"CO
5
Q.-?i
-^
g
.^
g
<D
&
g
<D
&
M
<D
&
M
<D
&
M
CD
= § i § § s-s v'Sii'Si
1
O^^-^^O t_ C m m o CD <D
-C<D <D <D CD CD C O ±^ ±^ ±^ ±^ ±f
U-ZZZZD^IDZ<<<<<
CO
2
g .■
8 g 5 i
5= • I u
* -•- t* -3
o
S 00s ?P5oo)co?oS)£ g
^ CO OOx-f-OT-pCNJLOO ^
5 g '
5 52 * "§
1* 1 Ctf
t- CO ▼- UO r^"1 +-•
O O• O t- g ej
i" 1' a <u
I £3
cn f^^i o^ cFo5 r^ o to g oj
-«- o <+-.
<+h
^
o co in
O CD O t- *t3 O
8a
5 8 5 ¥? f g
i ^ 38
^coSc^ScoSf^^fe <^S
OOCMt~Ot-OOt-COO ^£ (D
ON <L> O
o >%73
* ?SSv
I IIIs
1 ic m*
s| {I
CO -g & C iiil ^ ai *
I U i i §11°
1 51 i 1
I
Kl IIK
f 11 I 1 18251^
acrossValuesof
Figure2. PredictedValueofDepressiveSymptoms
Alter-Densityand Network Size byGender
12.5
115__ +.-
-♦■•♦
1q. ..::-.;•. =L-^___
1Q5
3) **• „ s
^^ ^ *
* *
m% 9.5 ^ m^^T
0> ^^^^
^^^^^^^"^^ ^ b ww||f
jgj , . , , , , , , , , ,
1 3 5 7 9 11 13 15 17 19 21
Network
Size
" ♦ Females: networks- - Females: networks
cohesive
fragmented
networks - Males:cohesive
m^mMales:fragmented networks
withlargernetworksreportmoredepressivesymptoms.As expected,per-
ceivedbelongingand supportmediatethe illeffectsofsmallfriendship net-
worksbutnotlargeones, inwhich depressivesymptoms remain elevated.
The higherlevels of depressive symptomsamong adolescents with
manyfriendscannot be explainedby the extentto whichtheyperceive
belongingat school or perceivesupportfromtheirfriends.Furthermore,
in Model 4 of Table 4, the four-wayinteractionstillexplains additional
inthe model,overand above theeffectofbelongingand support
variation
ws ^ oincoincocococ^CNj
I
C0 *gC\|f- OC5OC5OOOOO
" " "
'* '" •" ■" •" ■" «"
^. _Q
5o icm.^m{
I
^-^ cn .-^o ^^t- ^^o ^-^o ^-^cn ^-^oo ^-^^- ^^cm ^~
Sffl(pqr-r; CI>^^OC\JC\JCNT-;^rCNJ'^C\JTtCOph-:Cpa
I -
I
S*O ur>^-vco^^r^-^-^ oo ^-^Xn ^-.o ^^i-^ ^^^- --
Q
l*-0^?"^ "§. ^ R^ ^s ^« ^fe g
(A
I £ ^> * *
O O CO ^O) ^-^ O /-sN ^-^ CO ,->O) ^-^h^ ^- ^O .- »t- ^~
If
I! 1 I !
i!
W ^+^
I* **-^ T-I ■
I
- '*"
N) O
H 3
« o
"2 ~
£ 111
, « *=
3
"S S 1 5 -8 i §
cf | | 11 | i 1I
3 2
2 «
1
|l I is § § § i 1 | & h
2
"*1
a«
O l_ C_ Q> ^ -^ ^ -^ O 2_ CI (D <D
OQ> O'iZCD CD d> CD CD C o 4=« ■»=«
COO- Q- U_Z Z Z Z 0£ 3 Z < <
H o
Discussionand Conclusion
r- t- "^- CD O CO "*- OO
• i-
.-.•-.• §> o
OOOO O O O r« *+-•
* '55 ^
8 S 5 ^ ^ L
N
I
N
l\ h
<U U ^
v
•c75 a5 ^ 2 O,
-ill - i^ .aJ
§
1 1 1
CDCDCDCDOCDCD-s:
5 >"S "S i gosgo *-• t^ »- •
I I I I "I! | •§ -i ^ § « -S« 1 o
i <i <i <£ Z*II
<
I% i* iT1 llllijv
^Qi^^S^^ (^4 Cd ^ « ■»-
Notes
1. Ueno (2005) did graph mean differencesin depressive symptomsacross
networksize and founda lineartrend.Althoughour studyuses the same
data as Ueno (2005), our measureof networksize differsand we engage in
illeffectsofover-integration.
a morerigorousempiricaltest ofthe potentially
References
Echambadi, Raj, and James D. Hess. 2007. "Mean-Centering Does Not Alleviate
CollinearityProblems in Moderated Multiple Regression Models." Marketing
Science 26(3):438-45.
Eder, Donna, and Janet L. Enke. 1991. "The Structureof Gossip: Opportunities
and Constraints on Collective Expression Among Adolescents." American
Sociological Review 56(4):494-508.
Eder, Donna, Catherine C. Evans and Stephen Parker.1995. School Talk: Gender
and Adolescent Culture. Rutgers UniversityPress.
Frydenberg,Erica, and Lewis Ramon. 1993. "Boys Play Sport and GirlsTurnto
Others: Age, Gender and EthnicityAs Determinants of Coping." Journal of
Adolescence 16(3):253-66.
Haynie, Dana L, and D.W. Osgood. 2005. "Reconsidering Peers and Delinquency:
How Do Peers Matter?" Social Forces 84(2): 1 109-30.
House, James S., Debra Umberson and Karl R. Landis. 1988. "Structures and
Processes of Social Support." Annual Review of Sociology 14:293-318.
Jaccard, James, Choi K. Wan and Robert Turrisi. 1990. "The Detection and
of InteractionEffectsBetween Continuous Variables in Multiple
Interpretation
Regression." MultivariateBehavioral Research 25(4):467-78.
Kandel, Denise, Victoria Raveis and Mark Davies. 1991. "Suicidal Ideation in
Adolescence: Depression and, Substance Use and Other Risk Factors."
Journalof Youthand Adolescence 20(2):289-309.
Segregationin
and Friendship
Moody,James. 2001a. "Race, School Integration,
America."AmericanJournalofSociology107(3):679-716.
Pescosolido, Bernice A., and JudithA. Levy.2002. "The Role of Social Networks
in Health , Illness, Disease and Healing." Pp. 3-25. Social Networksand Health,
Volume 8. JudithLevy and Bernice Pescosolido, editors. Elsevier Science.
Raghunathan, T.E., Peter W. Solenberger and John Van Hoewyk. 2002. IVEware:
Imputation and Variance Estimation Software User Guide. University of
Michigan: Survey Research Center, InstituteforSocial Research.
Rosenfield, Sarah, Mary C. Lennon and Helene R. White. 2005. "The Self
and Mental Health: Self-Salience and the Emergence of Internalizingand
ExternalizingProblems." Journal of Health and Social Behavior 46(4):323-40.
Thoits, Peggy A. 1991. "On Merging IdentityTheory and Stress Research." Social
Psychology Quarterly54(2): 101-12.