Delft University of Technology
Biological Sensitivity to Context
Cortisol Awakening Response Moderates the Effects of Neighbourhood Density on the
Development of Adolescent Externalizing Problem Behaviours
Yu, Rongqin; Nieuwenhuis, Jaap; Meeus, Wim; Hooimeijer, Pieter; Koot, Hans M.; Branje, Susan
DOI
10.1016/j.biopsycho.2016.08.004
Publication date
2016
Document Version
Final published version
Published in
Biological Psychology
Citation (APA)
Yu, R., Nieuwenhuis, J., Meeus, W., Hooimeijer, P., Koot, H. M., & Branje, S. (2016). Biological Sensitivity
to Context: Cortisol Awakening Response Moderates the Effects of Neighbourhood Density on the
Development of Adolescent Externalizing Problem Behaviours. Biological Psychology, 120, 96-107.
https://doi.org/10.1016/j.biopsycho.2016.08.004
Important note
To cite this publication, please use the final published version (if applicable).
Please check the document version above.
Copyright
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons. Takedown policy
Please contact us and provide details if you believe this document breaches copyrights. We will remove access to the work immediately and investigate your claim.
This work is downloaded from Delft University of Technology.
Contents lists available atScienceDirect
Biological
Psychology
j o u r n a l h o m e p a g e :w w w . e l s e v i e r . c o m / l o c a t e / b i o p s y c h o
Biological
sensitivity
to
context:
Cortisol
awakening
response
moderates
the
effects
of
neighbourhood
density
on
the
development
of
adolescent
externalizing
problem
behaviours
Rongqin
Yu
a,∗,
Jaap
Nieuwenhuis
b,c,
Wim
Meeus
a,d,
Pieter
Hooimeijer
c,
Hans
M.
Koot
e,
Susan
Branje
aaDepartmentofYouthandFamily,UtrechtUniversity,Utrecht,TheNetherlands
bDepartmentOTB–ResearchfortheBuiltEnvironment,DelftUniversityofTechnology,Delft,TheNetherlands cDepartmentofHumanGeographyandPlanning,UtrechtUniversity,Utrecht,TheNetherlands
dDepartmentofDevelopmentalPsychology,TilburgUniversity,Tilburg,TheNetherlands
eDepartmentofClinicalDevelopmentalPsychologyandEMGOInstituteforHealthandCareResearch,VUUniversityAmsterdam,Amsterdam,The
Netherlands
a
r
t
i
c
l
e
i
n
f
o
Articlehistory: Received30July2015
Receivedinrevisedform6July2016 Accepted13August2016 Availableonline16August2016 Keywords:
Biologicalsensitivity Externalizingbehaviours Cortisolawakeningresponse(CAR) Neighbourhooddensity
a
b
s
t
r
a
c
t
Thisfour-yearlongitudinalstudyattemptedtotestperson-environmentinteractiontheoryandbiological sensitivitytheorybyassessingwhetherindividuals’biologicalstressactivityCARAUCg(CortisolAwakening
ResponseAreaUndertheCurvewithrespecttoground)moderatestheeffectsofneighbourhooddensity onthedevelopmentofadolescentexternalizingproblembehaviours.Participantswere358Dutch adoles-centswithameanageof15yearsatthefirstmeasurement.OuranalysesshowedthatCARAUCgmoderated
theeffectsofneighbourhooddensityonthelevelofparent-reporteddelinquencyandaggressionand adolescentself-reporteddelinquency.Morespecifically,foradolescentswithhighCARAUCg,higher
neigh-bourhooddensitysignificantlypredictedhigherlevelsofparent-reportedandadolescentself-reported delinquencyandaggression,whereastheassociationwasreversedornon-significantforadolescents withlowCARAUCg.OurfindingssuggestthatadolescentswithdifferentlevelsofCARAUCgrespond
differ-entiallytothedensityoftheneighbourhoodtheylivein,supportingforperson-environmentinteraction perspectivesandbiologicalsensitivitytheory.
©2016TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/4.0/).
1. Introduction
Highneighbourhooddensityhasbeencommonlyconsidered
anenvironmentalstressor.It couldcreate variousstressessuch
aslimitedenvironmentalresources,restrictedfreedomofchoice,
andtheexperienceofacrowdedandnoisyenvironment,which
could potentially stimulate individuals’ externalizing problem
behaviours(Regoeczi,2008).Theextenttowhichhigh
neighbour-hood density contributes toadolescents’ externalizing problem
behavioursmight bedetermined bywithin-individualvariables
thataffectthewayneighbourhooddensityisprocessed.Theeffects
of neighbourhooddensity on thedevelopment of externalizing
problembehaviours mightbe dependentonadolescents’stress
sensitivityasindicatedbywell-knownbiologicalmarkerslikethe
cortisolactivity(Hellhammer,Wüst,&Kudielka,2009).Cortisolis
∗ Correspondingauthor.Currentaddress:DepartmentofPsychiatry,Universityof Oxford,WarnefordHospital,WarnefordLn,OxfordOX37JX,UnitedKingdom.
E-mailaddress:Rongqin.yu@psych.ox.ac.uk(R.Yu).
oneofthestresshormonessecretedbyamajorhuman
physiolog-icalstresssensitivitysystem,thehypothalamic-pituitary-adrenal
(HPA)axis.Cortisolactivity isanimportant biological indicator
ofself-regulationandit playsacrucial roleintheregulationof
individuals’emotionalandbehaviouralresponsetoenvironmental
stressors(Fries,Dettenborn,&Kirschbaum,2009).Thus,cortisol
activitymightmoderatetheeffectsofneighbourhooddensityon
adolescents’externalizingproblembehaviours.
Several theories support the assumption that effects of
neighbourhooddensitydependonindividualcharacteristics.An
interactionistperspectivepositsthatindividuals’developmental
outcomes depend on the interplay between individual
charac-teristicsandenvironmentalfactors(Magnusson&Stattin,2006).
Morespecifically,hormones-contextinteractiontheoryassertsthat
individuals’hormone secretioncaninteractwithenvironmental
factorstopredictantisocialbehaviours(Susman,1997).
Further-more,accordingtothenotionofbiologicalsensitivity-to-context
(Ellis,Essex,&Boyce,2005),certainbiologicalmarkersthatmake
individualsmorevulnerabletoenvironmentaladversityalsomake
them morelikely to benefitfrom positive environmental
influ-http://dx.doi.org/10.1016/j.biopsycho.2016.08.004
0301-0511/©2016TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/4. 0/).
ences.Takentogether,individualswithdifferentstressresponse
setsasindicatedbytheircortisolactivitywouldbeexpectedto
differentially respondtohighdensity neighbourhoods in terms
ofelevatedlevelsofexternalizingproblembehaviours.
Individu-alshighincortisolactivitymay,duetotheirhighersensitivityto
theenvironment,bemorelikelyinfluencedbybothpositiveand
negativeaspectsoftheneighbourhoodinwhichtheylive.The
cur-rentstudyaimedtoexaminewhetherbiologicalcharacteristics(i.e.,
morningcortisolactivity)moderatedtheeffectsof
environmen-talcontext(i.e.,neighbourhooddensity)onthedevelopmentof
adolescents’externalizingproblembehaviours.
1.1. Neighbourhooddensityandexternalizingproblem
behaviours
Environmentalstressorssuchasahighdensityneighbourhood
might have a negative impact on youthexternalizing problem
behaviours(Lepore,Evans,&Palsane,1991).Increasesin
neigh-bourhood density may be accompanied by thepresence of an
increasingnumberofpeople,whichcouldcauseacrowding
experi-ence,andbyadecreaseinopenspace,whichmayposeconstraints
suchaslimitedaccesstopublicresources.Thesepotential
nega-tiveenvironmentalstressorsmightrestrictandinterferewiththe
attainmentofone’sgoals,whichmaythusgeneratefrustrationthat
couldstimulateaggressiveanddelinquentbehaviours(Regoeczi,
2003).Inaddition,high-densityenvironmentsmightcreateamore
impersonalatmosphereandprovideagreaternumberoftargets
forexternalizingbehaviours(Cohen&Felson,1979;Stark,1987).
Theaccompanyinganonymityof highpopulationdensitymight
alsoformastructurallimittowhatcanbeachievedthroughsocial
control(Sampson&Raudenbush,1999).
Empirical studies investigating the associations between
neighbourhood density and adolescents’ externalizing problem
behaviourshaveyieldedmixedresults.First,studieswhich
mea-suredthephysical density ofa neighbourhood(e.g.,number of
residentialaddressespersquarekilometre)reportedinconsistent
findings.Weenink(2011)revealedthatcomparedtoadolescents
livinginDutchurbanareas(>1000addressespersquarekilometre),
adolescentslivinginruralareasareslightlylesslikelytoengage
indelinquentbehavioursincludingpropertyoffensesandviolence.
However,anotherDutchstudyreportedthaturbanization(i.e.,high
residentialaddressesdensity)hadnoeffectsonyouthparent-and
self-reportedexternalizingproblembehaviours(Reijneveldetal.,
2010).Otherstudieswhich used social density (i.e.,population
density:numberofinhabitantsinacertainarea)toassess
neigh-bourhooddensityalsorevealedcontradictoryresults.Somestudies
foundpopulationdensitytobelinkedtoadolescentdelinquentand
criminalbehaviours(Schmitt, 1957,1966),whereas other
stud-iessuggestthatpopulationdensityhaslittleeffectondelinquency
orcrime(Gillis&Hagan,1981;Gillis,1974).Furthermore,Harden
etal.(2009)reportedinconsistentresultsregardingthe
neighbour-hooddensity-externalizingproblembehavioursassociationacross
informants.Inparticular,theyfoundthatpopulationdensitywas
positively associated withyouth self-reportedbut not
mother-reporteddelinquentbehaviours.Hence,existingempiricalstudies
didnotalwaysfindasignificantassociationbetween
neighbour-hooddensityandadolescentexternalizingproblembehaviours.
1.2. Cortisolawakeningresponseandexternalizingproblem
behaviours
TheCortisolAwakeningResponse(CAR)isconsideredanindex
oftheeverydayhumanadrenalcortisolsecretoryactivityandis
definedasthechangeincortisolconcentrationthatoccurs
dur-ingthefirsthourafterawakingfromsleep(Clow,Thorn,Evans,&
Hucklebridge,2004).Itistypicallyfeaturedbyabriskincreaseof
cortisollevelswithin20-30minafterawakeninginthemorning
andadeclinethereafter(Clowetal.,2004).Theoverallvolumeof
cortisolreleasedoverthewakingperiod(AreaUndertheCurvewith
respecttoground[AUCg];furtherreferredtoascortisolactivity),
whichwasthemainindicatorofCARinthecurrentstudy,presentsa
usefulandreliableindexofadrenocorticalactivity(Pruessneretal.,
1997;Schmidt-Reinwaldetal.,1999).
Different theories have hypothesized different directions of
the link between cortisol activity and externalizing problem
behaviours. Two major theories have postulated an
associa-tionbetweenexternalizingproblembehavioursandphysiological
hypo-arousal.Thefearlessnesstheorysuggeststhatalowtendency
tobecomearousedinreactiontofearfulstimuliwouldresultina
higherlikelihoodtobecomedisruptiveandantisocial(Raine,1993).
Inaddition,thesensation-seekingtheory(Raine,1993;Zuckerman
&Neeb,1979)hypothesizesthatlowarousalisanunpleasant
phys-iologicalstateandinordertogetridofthisstate,individualswith
lowarousallevelswouldseekstimulationbyinitiatingantisocial
behavioursthat increasephysicaltension. Followingthis lineof
reasoning,lowercortisollevelsshouldberelatedtohigherlevelsof
externalizingproblembehaviours(Brennan&Raine,1997).Onthe
otherhand,however,researchersrecentlyhavehypothesizedthat
hyper-arousalisariskfactorinstressfulcontexts.Hyper-arousal
maybeevidentasheightenedphysiologicalresponsestostressor
threats,suchasirritability,hypervigilance,frustration,andan
exag-geratedstartleresponse,whichcouldinturntriggerexternalizing
problembehaviourssuchasaggressiveanddelinquentbehaviours
(Kerig,Vanderzee,Becker,&Ward,2012).Hence,followingthis
lat-terlineofreasoning,highercortisollevelsindicatinghyper-arousal
shouldbeassociatedwithhigherlevelsofexternalizingproblem
behaviours.
Evidenceregardingtheassociationsbetweenmorningcortisol
levelandadolescentexternalizingproblembehavioursismixed.
Several studies examined cross-sectional associations between
activityoftheHPA-axisanddisruptivebehavioursinadolescents.
Some researchers found an association between low morning
basalcortisollevelsandadolescentdisruptivebehaviours(Pajer,
Gardner,Rubin,Perel,&Neal,2001),whereasothersdidnotfound
suchassociation(Ruttleetal.,2011;Scerbo&Kolko,1994).
Longi-tudinalstudieslookingatwhetherlowmorningcortisolactivity
is a risk factor for future externalizing problembehaviour also
revealedinconsistentresults(Shirtcliff,Granger,Booth,&Johnson,
2005; Shoal, Giancola, & Kirillova, 2003; Van Bokhoven et al.,
2005).Specifically,lowmorningcortisollevelspredicted
adoles-centaggressivebehaviours5yearslater(Shoaletal.,2003),risky
behaviours oneyearlater(Shirtcliffetal.,2005), andpersistent
aggressivebehaviouroverthreeyears(Platjeetal.,2013).Onthe
contrary, Van Bokhoven et al. (2005) found that higher
morn-ingcortisollevelsatage13predictedhigherconductdisorderin
boysatage14–16.Furthermore,studiesalsoreportedalackof
associationsbetweenmorningbasalcortisolleveland
externaliz-ingproblembehavioursinadolescents(Dabbs,Jurkovic,&Frady,
1991;Klimes-Dougan,Hastings,Granger,Usher,&Zahn-Waxler,
2001;Sondeijkeretal.,2008).Therefore,evidenceforthe
associa-tionbetweenmorningcortisolactivityandexternalizingproblem
behavioursinadolescenceislargelyinconsistent,asnegative,
pos-itive,aswellasnon-significantrelationshavebeenreported.
1.3. Moderationofneighbourhooddensityeffectson
externalizingproblembehavioursbyCARAUCg
According to person-environment interaction perspectives,
externalizing problem behaviours are likely the result of both
environmentalinfluencesandindividuals’biological
individualswithhighbiologicalsensitivitymightespeciallybenefit
fromapositiveenvironmentbutmightbevulnerabletothe
neg-ativeinfluenceofanadverseenvironment(Ellis&Boyce,2011).
Thepresenceofinteractioneffectsbetweenneighbourhood
den-sityandstresssensitivityasindicatedbymorningcortisolactivity
mightprovidesomeexplanationsfortheinconsistentfindingson
theassociationbetweenneighbourhooddensityandexternalizing
problems.Responses toa high-densityneighbourhood
environ-mentmaydifferforadolescentswhovaryintheirmorningcortisol
levels(Hellhammeretal.,2009).Forinstance,althoughsome
ado-lescentsmayviewahighdensityneighbourhoodanopportunity
forsocialinteractionsandnetworks,otheradolescents,whohave
difficultyinsocialandemotionalregulation,mayseeitascrowded
andnoisyorfeelthreatedintheiraccessibilitytoneighbourhood
recourses.
Apattern of hypo-arousalhasbeentermed a low biological
sensitivitytocontext (Elliset al.,2005).A lowbiological
sensi-tivitytocontextis definedasindividuals’lowabilitytoprocess
environmentalinfluences. Low biological sensitivity individuals
mightdisplayalackofbiologicalsensitivitytoenvironmental
chal-lenges,theymightbebufferedagainststressors(Ellisetal.,2005).
Ontheotherhand,however,individualswithhyper-arousalhave
a higher biological sensitivity to context, which means greater
responsivenesstobothpositiveandnegativeenvironmental
influ-ences.Therefore,adolescentswithhighCARAUCgmightbemore
likelytoshowincreasesinexternalizing problembehaviours in
responsetoahigh-densityneighbourhoodandalsobemorelikely
toshowdecreasesinexternalizingproblembehavioursinresponse
tolow-densityneighbourhood.Incontrast,adolescentswithlow
CARAUCgmaybenotaffectedbytheirenvironmentandtheirlevels
ofexternalizingbehavioursmightnotvarywiththedensityoftheir
neighbourhood.Empiricalresearchontheinterplaybetweenbasal
cortisollevelandneighbourhoodadversityhasbeenlimited.This
studyattemptedtofilltheknowledgegapbytestingthe
moderat-ingroleofCARAUCgontheeffectsofneighbourhooddensityonthe
developmentofadolescents’externalizingproblembehaviours.
1.4. Thecurrentstudy
The aims of the current study were to examine
person-environmentinteractioneffectsinpredictingadolescents’
devel-opmentofexternalizingproblembehaviours.Wefirstreplicated
previousresearchbyestimatingthemaineffectsofneighbourhood
density and CARAUCg on adolescents’ development of
adoles-centexternalizingproblembehavioursincludingaggressionand
delinquencyacross a period of four years frommiddle to late
adolescence.Asresultsfrompreviousstudieswereinconsistent,
noconcrete hypothesescouldbe formulated.Subsequently, we
investigatedinteractioneffectsbetweenCARAUCgand
neighbour-hooddensityonthedevelopmentof adolescentaggression and
delinquency. We expectedthat adolescents withhighCARAUCg,
comparedtothosewithlowCARAUCg,wouldscorehigherin
exter-nalizingproblembehavioursinahigh-densityneighbourhoodbut
scorelowerinexternalizingproblembehaviourinalow-density
neighbourhood.
2. Method
2.1. Sample
Participantswere358adolescents(205[57.3%]boys)who
par-ticipatedinalongitudinalstudyonadolescentdevelopmentand
whotookpartincortisolawakingmeasurementsatWave3,with
ameanageof15.03years(SD=0.45).TheywerepartoftheRADAR
study(N=497,ResearchonAdolescentDevelopmentAnd
Relation-ships;Meeusetal.,2010).RADARisanongoinglongitudinalstudy
focusingonvariousadolescentdevelopmentaloutcomesincluding
externalizingproblembehaviours.Thecurrentstudywasbasedon
datafromthethirdtothesixthwaveofRADAR.Allparticipants
inthecurrentstudyidentifiedthemselvesasDutch.Inthis
sam-ple,10.5%ofadolescentswerefromlowSESfamilyinwhichfather
and mother were unemployedor held an elementary job (e.g.,
constructionworker,janitor,truckdriver;Statistics-Netherlands,
1993).FamilySESwasmediumorhighfortheother89.5%ofthe
adolescents,implyingthat atleastoneof theparents’ jobswas
classifiedasmediumlevel (e.g.,policeofficer, physician’s
assis-tant) or high level (e.g., doctor, scientist,high school teacher).
BetweenthetotalRADARsampleof497adolescentsandthe358
adolescentswhoparticipatedinthecortisolmeasurementatWave
3,theonlysignificantdifferencewasinself-reportedaggression
(F[1,373]=5.24, p=0.02), withadolescentswho participated in
thecortisolmeasurementreportingahigherlevelofaggression.
Therewerenosignificantdifferencesbetweenthesetwogroups
inparent-reporteddelinquency(F[N=1,384]=0.25,p=0.66)and
aggression(F[N=1,384]=0.19,p=0.66),self-reporteddelinquency
(F[N=1, 387]=2.52, p=0.11), gender distribution(2 [N=497,
1]=0.05, p=0.82),age(F[3,494]=0.05, p=0.77),and
neighbour-hooddensity(F[N=469,1]=1.92,p=0.17).
ThenumberofparticipantsacrossWave3toWave6fluctuated
peryear,with358participantsatWave3,347atWave4,338at
Wave5,and323atWave6.Attritionwas9.8%overtheperiodof
threeyears.Toestimatethepatternofmissingdata,Little’sMissing
AtRandom(MCAR)testwasconducted(Little,1988)onallvariables
usedinthisstudy.Little’sMCARtestrevealedanormed2(2/df)
of1.31which,accordingtoguidelinebyBollen(1989),indicates
thatthepatternofthemissingdatawasnotmeaningfully
differ-entfromamissingcompletelyatrandompattern.Therefore,we
appliedFullInformationMaximumLikelihood(FIML)inMplusfor
themodelestimations(Schafer&Graham,2002).
2.2. Procedure
Participants were recruited from various Dutch elementary
schools.Aninvitationletterandadescriptionofthestudywere
senttoadolescents’homeaddresses.Bothparentsandadolescents
providedinformedconsenttoparticipate.Adolescentsandparents
filledoutvarious questionnairesduringtheannual homevisits,
supervisedbytrainedresearchassistants.Inadditionto
administra-tionofthebehaviouralmeasurements,trainedresearchassistants
gavedetailedverbalandwritteninstructionforcortisol
measure-mentsduringthehomevisit.Confidentialitywasassuredexplicitly
beforeparticipationinthestudy.ParticipantsreceivedD15asa
rewardfortheirparticipationineachwave.TheRADARstudyhas
beenapprovedbytheresponsiblemedicalethicscommittee,and
wasconductedinaccordancewiththeDeclarationofHelsinki.
2.3. Measure
2.3.1. Adolescentexternalizingproblembehaviours
Externalizingproblembehavioursweremeasuredthrough
par-ents’reportoftheirchild’sbehavioursand throughadolescents’
self-report.ParentsfilledouttheDutchversionoftheAchenbach
ChildBehaviourChecklist(CBCL;Achenbach,1991),reportingtheir
child’s delinquent and aggressive behaviours. Delinquency was
measuredusingtheDelinquent Behaviourscale (13items; e.g.,
“Setsfires”)andaggressionwasmeasuredusingtheAggressive
Behaviourscale(20items;e.g.,“Fightsalot”).Parentsresponded
tothequestionsona3-pointscale(i.e.,0=never;1=sometimes;
2=often). The validity and reliability of this measure hasbeen
showntobeadequate(Achenbach,1991).Inthecurrentsample,
to0.80fortheDelinquentBehaviourscale,andfrom0.89to0.90
fortheAggressiveBehaviourscale.
AdolescentsalsofilledouttheYouthSelfReport(YSR;Verhulst,
VanderEnde,&Koot,1997),whichistheself-reportversionofthe
CBCL.Specifically,11itemswereemployedtoassessdelinquency
(e.g.,“Isetfires”)and19itemswereusedtoindicateaggressive
behaviours(e.g.,“Ifightalot”).Adolescentsratedtheirfrequency
ofdelinquencyandaggressiononathree-pointLikertscale(i.e.,
0=never;1=sometimes;2=often). Acrossfourwaves,Cronbach’s
alphasrangedfrom0.70to0.75fordelinquency,andfrom0.86to
0.88foraggression.
2.3.2. Neighbourhooddensity
Neighbourhooddensitywasassessedbytheaveragenumberof
addressespersquarekilometre.Welinkedsix-digitpostcodedata
inoursampletoStatisticsNetherlands(2011)whichprovidesdata
ontheaddressdensityofthesurroundingareaforeachpost-code
in2010(i.e.,aroundthefifthwaveofdatacollection).As
neigh-bourhoodswithahighdensitymightbepoorerthanthosewith
lowdensity,wecontrolledfortheeffectofneighbourhoodwealth
onneighbourhooddensityinthecurrentstudy.Weobtainedthe
averagepropertyvalueofdwellingsinneighbourhoodsmeasured
in 2004 from Statistics Netherlands (2006). The average
prop-ertyvaluecaptured thequalityof dwellingsand thesocial and
physicalattributesoftheneighbourhoodandthereforewas
con-sideredagoodproxyforneighbourhoodwealth(Visser,VanDam,
&Hooimeijer,2008).
2.3.3. CARAUCg
CARAUCgwasmeasuredinsalivawhichwascollectedby
pas-sivedrooling, immediately afterawakening (Cort0),30minutes
(Cort30),and60minutes(Cort60)later.Thesalivasamplingwas
scheduledonatypicalweekday,followingdetailedverbaland
writ-teninstruction.Participantswereinstructedtorinsetheirmouths
withwaterbeforesampling,andnottoeat,drink,smokeorbrush
theirteethbeforecompletingCort60.Theywererequestedto
col-lecttheirsalivathroughasmallstrawintoapolypropylenetube,
andlabelthesetubeswiththetimeanddateofsampling.After
collection,participants were askedto storethe samplesin the
refrigeratorandsendthembymailtotheresearchcenterthesame
day.
At the research center, the cortisol collections were stored
uncentrifugedat-20◦Cuntilanalysis.Salivarycortisollevelswere
analyzed using electrochemiluminescence immunoassay ECLIA
(E170 Roche, Switzerland). The lower detection limit was 0.5
nmol/l,andmeanintra-assayandinter-assaycoefficientsof
vari-ationwere,respectively,3%and12%.Allsampleswerechecked
forcorrectnessofsampling.Caseswereexcludedfromanalysesif
thecortisoldatawereofincorrectsamplingtime,unclearhowit
wassampled(i.e.,notregistered),contaminated(e.g.,bysmoking
or brushingteeth), orof extreme values (i.e.,>3 SDfrom
aver-age).Inthecurrentstudy,358participantsprovidedqualifieddata.
CARAUCgisconsideredasummaryparameteroftherepeated
mea-surementsofCAR(i.e.,0,30,and60minutesafterawakening).Thus,
itisanestimationoftotaladrenalcortisolsecretionduringthefirst
hourafterawakening.WecalculatedtheCARAUCgwiththeformula
providedbyPruessner,Kirschbaum,MeinlschmidandHellhammer
(2003).
2.3.4. Controlvariables
Asgender,physicaldevelopment,substanceuse,andstressful
lifeexperiencesmightaffectCARAUCg(Clowetal.,2004;Platjeetal.,
2013),wetestedwhetherthesevariableshadasignificanteffect
onCARAUCgandcontrolledforthesignificantpredictorsinourfinal
statisticalmodels.Thesevariableswereassessedwhenadolescents
were15yearsold,thesameagewhentheirCARAUCgwasmeasured.
Physicaldevelopmentwasindicatedwithadolescents’bodymass
index(BMI)whichwascalculatedasweightinkg/(lengthinm)2.
Substanceusewasindicatedasnicotineandalcoholuse.Nicotine
usewasassessedbyanine-optionquestionrangingfrom“Ihave
neversmoked”to“Ismokeeveryday”.Thedataweredichotomized:
responsesfrom“Ihaveneversmoked”to“Ismokelessthanoncea
month”weredefinedas“notusenicotine”andthosefrom“Idonot
smokeweekly,butatleastonceamonth”to“Ismokeeveryday”were
definedas“usingnicotine”.Alcoholuseoverthelastfourweekswas
assessedwithaquestionwithsixresponseoptions,rangingfrom
“none”to“daily”.Thesedataweredichotomizedtoo:“none”was
definedas“notusingalcohol”andresponsesfrom“1-3daysinthe
lastfourweeks”to“everyday”weredefinedas“usingalcohol”.
Stress-fulexperiencesincludingsexualassault,physicalassault,andbeing
threatenedwithviolence,weremeasuredwiththeInternational
CrimeVictimsSurvey(ICVS;Nieuwbeerta,2002).Ourregression
model withCARAUCg as a dependentvariable andwith gender,
smoking,alcoholuse,BMI,sexualabuse,physicalassault,andbeing
threatenedwithviolenceaspredictorsindicatedasignificanteffect
ofgender(=0.14,p=0.02).Alltheothervariablesinthemodeldid
nothaveasignificanteffectonCARAUCg(srangedfrom-0.01to
0.11,andps>0.05).Therefore,wecontrolledforgendereffectson
CARAUCginourfollowingstatisticalmodels.
2.4. Statisticalanalysis
Multipleregressionmodelsincorporatinglatentgrowthmodels
wereconductedtotesttheinteractioneffectsbetweenCARAUCgand
neighbourhooddensityonthedevelopmentofexternalizing
prob-lembehaviours.Beforetestingmultipleregressionmodels,wefirst
determinedtheshapeofgrowthinadolescentexternalizing
prob-lembehaviours(i.e.,delinquencyandaggression)fromage15to18.
Tothatend,wecomparedthechi-squarevaluesofmodels
includ-ingalinearandquadraticgrowthtocapturechangesindelinquency
andaggression(Satorra&Bentler,2001).
The modelsfor parent-reporteddelinquencyand aggression
indicated that adding quadratic slopes significantly improved
modelfit(i.e.,asignificantlylowerchi-squarevalue;2[N=358,
4]=56.38,p=0.00 and 2 [N=358, 4]=21.11,p=0.01,
respec-tively). For self-reported delinquency and aggression, adding
quadratic slopes did not significantly improve model fit (2
[N=358, 4]=9.49, p=0.26 and 2 [N=358, 4]=9.54, p=0.13,
respectively). However, tofacilitate the comparability between
modelsacrossparent-and self-reportedexternalizing problem
behaviours,wechosemodelswithbothlinearandquadraticslopes.
Toavoidconvergenceproblems,thevariancesofquadraticslopes
werefixedatzero.
Afterdeterminingtheshapeofthegrowthofexternalizing
prob-lembehaviours,weaddedpredictorstothegrowthmodel.Wefirst
assessedthemaineffectsofneighbourhooddensityandCARAUCg
onthedevelopment(i.e.,interceptandlinearslope)ofadolescent
delinquencyandaggression.Next,wetestedtheinteractioneffects
byincluding theCARAUCg byneighbourhooddensity interactive
term.Inallmodels,wecontrolledforeffectsofSocial Economic
Status(SES)andgender.Furthermore,asneighbourhoodswitha
highdensitymightbepoorerthanthosewithlowdensity,
neigh-bourhooddensitywascontrolledfortheeffectsofmeanlevelsof
propertyvalues (s=-0.26, ps<0.00acrossmodels).AsCARAUCg
differbetweengirlsandboys,CARAUCgwascontrolledforgender
effects(srangedfrom0.26to0.27,psrangedfrom0.01to0.02
acrossmodels).Moreover,whentheinteractioneffectsbetween
CARAUCgandneighbourhooddensityweresignificant,weassessed
SESdifferencesintheinteractioneffectsbyconductingthree-way
interactionanalyses.Inaddition,wetestedwhetherthe
moderat-ingeffectsofCARAUCgcanbeexplainedbythepotentialoverlap
interac-Table1
MeansandStandardDeviationsoftheObservedValuesofCARAUCg,Neighbourhood
Density,Parent-andSelf-ReportedExternalizingProblemBehavioursfrom15to18 Years.
Measure M(SD)
15years 16years 17years 18years
1.CARAUCg 1030(365) – – – 2.NeighbourhoodDensity 2030(1911) – – – 3.Delinquency(CBCL) 1.85(2.15) 2.12(2.57) 2.22(2.57) 2.09(2.26) 4.Aggression(CBCL) 7.00(5.42) 6.55(5.42) 5.79(5.45) 5.05(5.04) 5.Delinquency(YSR) 3.45(2.95) 3.77(2.97) 3.83(2.87) 3.75(2.87) 6.Aggression(YSR) 7.11(5,75) 6.87(5.68) 6.43(5.49) 5.70(5.09) Note:M(SD)=Mean(StandardDeviation).CARAUCg=CortisolAwakeningResponse AreaUndertheCurvewithrespecttoground.CBCL=ChildBehaviourChecklist. YSR=YouthSelfReport.
tioneffects betweenSES and CARAUCg in predicting adolescent
externalizingproblembehaviours.Finally,wealsoexaminedifthe
moderatingeffectsofCARwerespecifictoAUCg(i.e.,theoverall
cortisolsecretionpost-awakening),oralsooccurredforthe
abso-lutechangeincortisollevelspost-awakening:AUCi(Areaunder
thecurvewithrespecttoincrease;seePruessneretal.,2003for
detailedcomputationalprocedure).Tothisend,weassessedthe
interactioneffectsbetweenCARAUCiandneighbourhooddensityon
adolescents’externalizingproblembehaviours.Allanalyseswere
conductedwithinMplus7.0(Muthén&Muthén,1998–2012).In
addition,sincethedependentvariablesdelinquencyand
aggres-sionwerenotnormallydistributed,weadoptedarobustmaximum
likelihoodestimator(MLR; Satorra&Bentler, 2001),totakethe
non-normaldistributionofthedataintoaccount.
Whentheinteractioneffectsweresignificant,weexaminedthe
shapeoftheinteractionbyprobingtheinteractioneffects.To
exam-inethis,simpleslopesforeachinteractionwerepresented(at1SD
aboveandbelowmeanofthemoderators).Moreover,weapplied
theJohnson-Neymantechnique,usingthecomputationaltoolof
Preacher,Curran,andBauer,(2006),toidentifyforwhichregions
intherangeofthemoderatorvariableeffectsofthefocalpredictor
ontheoutcomeweresignificant(p<0.05)(Bauer&Curran,2005;
Hayes&Matthes,2009).
3. Results
Table1presentsdescriptivestatistics,includingthemeansand
standarddeviationsofneighbourhooddensityandCARAUCgwhen
adolescentswereonaverage15yearsofageandparent-and
self-reporteddelinquentandaggressivebehavioursacrossfourannual
waves(i.e.,fromage15to18years).Table2presentsbivariate
cor-relationsamongCARAUCgandneighbourhooddensitymeasuredat
age15years,andparent-andself-reportedexternalizing
prob-lembehavioursassessedfromage15to18years.Tables3and4
presenttheresultsofourfinalstructuralequationmodels
examin-ingthemaineffectsofandtheinteractioneffectsbetweenCARAUCg
andneighbourhooddensityonthedevelopmentofexternalizing
problembehavioursacrossmiddletolateadolescence.
3.1. MaineffectsofneighbourhooddensityandCARAUCgonthe
developmentofparent-reportedadolescentexternalizingproblem
behaviours
Weexaminedthemaineffectsofneighbourhooddensityand
CARAUCg on the development of parent-reported externalizing
problem behaviours, controlling for effects of SES and gender.
In the two models predicting development of parent-reported
delinquencyand aggression,the onlysignificanteffect wasthe
effect of SES (i.e., low vs. high) on the intercept of aggression
(N=358,B=2.47,=0.15,p=0.04).LowerSESwaslinkedtohigher Table
2 Bivariate Intercorrelations among CAR AUCg , Neighbourhood Density, Parent-and Self-Reported Externalizing Problem Behaviours from 15 to 18 Years. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 1 CAR AUCg T3 2 NB Density T3 0.05 3 Delinquency CBCL T3 0.05 0.02 4 Delinquency CBCL T4 0.04 0.03 0.74** 5 Delinquency CBCL T5 0.03 0.02 0.67** 0.77** 6 Delinquency CBCL T6 -0.04 0.05 0.60** 0.66** 0.79** 7 Aggression CBCL T3 0.00 -0.03 0.70** 0.60** 0.58** 0.55** 8 Aggression CBCL T4 0.02 -0.00 0.59** 0.72** 0.69** 0.60** 0.81** 9 Aggression CBCL T5 0.06 -0.02 0.52** 0.54** 0.73** 0.61** 0.75** 0.85** 10 Aggression CBCL T6 0.04 0.02 0.43** 0.42** 0.57** 0.66** 0.68** 0.76** 0 .83 ∗ ∗ 11 Delinquency YSR T3 0.11 0.05 0.43** 0.43** 0.41** 0.40** 0.35** 0.33** 0 .31 ∗ ∗ 0 .26 ∗ ∗ 12 Delinquency YSR T4 -0.01 0.08 0.31** 0.37** 0.37** 0.40** 0.29** 0.31** 0 .28 ∗ ∗ 0 .25 ∗ ∗ 0 .70 ∗ ∗ 13 Delinquency YSR T5 0.08 0.07 0.29** 0.33** 0.44** 0.42** 0.25** 0.29** 0 .32 ∗ ∗ 0 .26 ∗ ∗ 0 .59 ∗ ∗ 0 .66 ∗ ∗ 14 Delinquency YSR T6 -0.05 0.12* 0.17** 0.22** 0.27** 0.32** 0.17** 0.16** 0 .18 ∗ ∗ 0 .14 ∗ 0 .46 ∗ ∗ 0 .58 ∗ ∗ 0 .68 ∗ ∗ 15 Aggression YSR T3 0.12* -0.02 0.28** 0.25** 0.29** 0.25** 0.38** 0.34** 0 .35 ∗ ∗ 0 .29 ∗ ∗ 0 .68 ∗ ∗ 0 .54 ∗ ∗ 0 .50 ∗ ∗ 0 .40 ∗ ∗ 16 Aggression YSR T4 0.08 -0.01 0.27** 0.28** 0.31** 0.26** 0.39** 0.42** 0 .40 ∗ ∗ 0 .31 ∗ ∗ 0 .60 ∗ ∗ 0 .68 ∗ ∗ 0 .59 ∗ ∗ 0 .49 ∗ ∗ 0 .76 ∗ ∗ 17 Aggression YSR T5 0.14* -0.03 0.24** 0.24** 0.34** 0.24** 0.33** 0.35** 0 .38 ∗ ∗ 0 .29 ∗ ∗ 0 .47 ∗ ∗ 0 .51 ∗ ∗ 0 .68 ∗ ∗ 0 .52 ∗ ∗ 0 .68 ∗ ∗ 0 .73 ∗ ∗ 18 Aggression YSR T6 0.00 0.03 0.12* 0.10 0.17** 0.13* 0.21** 0.22** 0 .29 ∗ ∗ 0 .25 ∗ ∗ 0 .41 ∗ ∗ 0 .45 ∗ ∗ 0 .52 ∗ ∗ 0 .64 ∗ ∗ 0 .60 ∗ ∗ 0 .65 ∗ ∗ 0 .73 ∗ ∗ Note. CAR AUCg = Cortisol Awakening Response Area Under the Curve with respect to ground. NB = Neighbourhood. CBCL = Child Behaviour Checklist. YSR = Youth Self Report. *p < 0.05; ** p < 0.01.
aggression.Neighbourhooddensitydidnotpredictthe
develop-ment (i.e.,intercept and slope)of parent-reported delinquency
(N=358, B=0.04, =0.02, p=0.64 andN=358, B=0.02,=0.03,
p=0.67,respectively)andaggression(N=358,B=-0.17,=-0.03,
p=0.49andN=358,B=0.06,=0.06,p=0.46,respectively).
More-over,CARAUCgdidnothavemaineffectsontheinterceptandslope
ofparent-reporteddelinquency(N=358,B=0.17,=0.08,p=0.19
andN=358,B=-0.06,=-0.10,p=0.16,respectively),neitherdid
ithavemaineffectsontheinterceptandslopeofparent-reported
aggression(N=358,B=0.13,=0.03,p=0.69andN=358,B=0.05,
=0.05,p=0.57,respectively).
3.2. MaineffectsofneighbourhooddensityandCARAUCgonthe
developmentofadolescentself-reportedexternalizingproblem
behaviours
Weexaminedthemaineffectsofneighbourhooddensityand
CARAUCgonthedevelopmentofself-reportedexternalizing
prob-lembehaviours,controllingforeffectsofSESandgender.Inthetwo
modelspredictingdevelopmentofself-reporteddelinquencyand
aggression,significanteffectsofSESandCARAUCgwereobserved.
Specially,lowerSESwasrelatedtolowerslope(i.e.,changerate)
ofself-reporteddelinquency(N=358,B=-0.46,=-0.18,p=0.01).
In addition,lowerSES wasrelated tohigher intercept(N=358,
B=3.48,=0.20,p=0.00)andlowerslope(N=358,B=-0.85,=
-0.22,p=0.00)ofself-reportedaggression.Furthermore,CARAUCg
waspositivelyassociatedwiththeinterceptofadolescent
aggres-sion(N=358,B=0.84,=0.16,p=0.02).
No other main effects of CARAUCg on the intercepts and
slopesofadolescentexternalizingproblembehaviourswerefound.
CARAUCgdidnotpredicttheintercept(N=358, B=0.33,=0.13,
p=0.06) nor the slope of self-reported delinquency (N=358,
B=-0.12, =-0.15, p=0.09).Moreover,CARAUCg wasnot
signifi-cantly related totheslope of self-reportedaggression (N=358,
B=-0.20, =-0.16, p=0.08). Inaddition, neighbourhooddensity
did not predict the intercept and slope of self-reported
delin-quency(N=358, B=0.13, =0.05, p=0.35 and N=358, B=0.04,
=0.05, p=0.35, respectively)and aggression(N=358,B=-0.16,
=-0.03, p=0.48 andN=358, B=0.05,=0.04, p=0.50,
respec-tively).
3.3. InteractioneffectsbetweenCARAUCgandneighbourhood
densityonthedevelopmentofparent-reportedexternalizing
problembehaviours
Table3presentstheresultsoftheinteractioneffectsbetween
CARAUCgand neighbourhooddensityonparent-reported
adoles-centdelinquencyandaggression,whilecontrollingfortheeffects
ofSESandgender.Theresultsfromtheregressionanalysesshowed
significantinteraction effects between CARAUCg and
neighbour-hood density on the intercept of parent-reported delinquency
(N=358,B=0.31,=0.15,p=0.01)andparent-reportedaggression
(N=358, B=0.98, =0.19, p=0.01).These significantinteraction
effects indicated that the associations between neighbourhood
densityandparent-reporteddelinquentandaggressivebehaviours
differedforadolescentswithvaryinglevelsofCARAUCg.Regionof
significancetestsrevealedthatforadolescentswithCARAUCg
lev-els≥0.62SDand≥1.90SDabovethemean,higherneighbourhood
densitysignificantly predictedhigherlevelsof delinquencyand
aggression,respectively.Remarkably,thereversewastrueforthe
relationbetweenneighbourhooddensityandexternalizing
prob-lembehavioursforadolescentswithCARAUCglowerthancertain
levels.Specifically,foradolescentswithCARAUCglevels≤-0.97SD
and≤-0.09SDbelowthemean,lowerneighbourhooddensity
sig-nificantlypredictedhigherlevelsofdelinquencyandaggression.
Todeterminewhethertheinteractionwasinsupportofthe
biolog-icalsensitivitytheory,wealsoprobedthisinteractionforvarying
levelsofneighbourhooddensity.Wefoundthatforadolescents
liv-inginaneighbourhoodwithdensitylevels≥0.27SDand≥0.85SD
abovethemean,higherCARAUCgsignificantlypredictedhigher
lev-elsofdelinquencyandaggression,respectively.Thereversewas
truefortherelationshipbetweenCARAUCgandexternalizing
prob-lembehaviours, foradolescentsliving ina neighbourhoodwith
density levels≤-2.57SDand≤-0.96 SDbelow themean,lower
CARAUCgsignificantlypredictedhigherlevelsofdelinquencyand
aggression,respectively.Theinteractiveeffectsarevisualizedby
showingsimple slopes for adolescentshigh(at1 SDabove the
mean)andlow (at1SDbelowthemean)inCARAUCginFigs.1a
and2a,andforadolescentswholiveinhigh(at1SDabovethe
mean)andlow(at1SDbelowthemean)densityneighbourhood
inFigs.1band2b.Furtherthree-wayinteractionanalysisdidnot
indicatethatthesesignificantinteractioneffectsweredifferentby
SESlevels.
3.4. InteractioneffectsbetweenCARAUCgandneighbourhood
densityonthedevelopmentofself-reportedexternalizingproblem
behaviours
Table4presentstheresultsoftheinteractioneffectsbetween
CARAUCgandneighbourhooddensityonself-reportedadolescents’
delinquencyand aggression,whilecontrolling fortheeffects of
SESandgender.Thereweresignificantinteractioneffectsbetween
CARAUCg and neighbourhooddensity inpredictingtheintercept
(N=358, B=0.30, =0.11, p=0.04) but not the slope (N=358,
B=0.02,=0.03,p=0.68)ofself-reporteddelinquency.Therewere
nosignificantinteractioneffectsbetweenCARAUCgand
neighbour-hooddensityinpredictingtheintercept(N=358,B=0.21,=0.04,
p=0.46)neithertheslope (N=358, B=0.05,=0.04,p=0.63)of
adolescents’self-reportedaggression.Probing of theinteraction
fordifferentlevelsofCARAUCgrevealedthatforadolescentswith
CARAUCglevels≥0.02SDabovethemean,higherneighbourhood
density significantly predicted higher levels of delinquency. To
determinewhethertheinteractionwasinsupportofthebiological
sensitivitytheory,wealsoprobedthisinteractionforvarying
lev-elsofneighbourhooddensity.Wefoundthatforadolescentsliving
inaneighbourhoodwithdensitylevels≥1.07SDabovethemean,
higherCARAUCglevelspredictedhigherlevelsofself-reported
delin-quentbehaviours.Theinteractiveeffectsarevisualizedbyshowing
simpleslopesforadolescentshigh(at1SDabovethemean)andlow
(at1SDbelowthemean)inCARAUCginFig.3a,andforadolescents
wholiveinhigh(at1SDabovethemean)andlow(at1SDbelowthe
mean)densityneighbourhoodinFig.3b.Furtherthree-way
inter-actionanalysesdidnotindicateSESdifferencesinthesesignificant
interactioneffects.
3.5. InteractioneffectsbetweenSESandneighbourhooddensity
onthedevelopmentofexternalizingproblembehaviours
To test whether the moderating effects of CARAUCg can be
explainedbythepotentialoverlapbetweenneighbourhooddensity
andSES,weexaminedinteractioneffectsbetweenSESandCARAUCg
inpredictingadolescentexternalizingproblembehaviours.Infour
statisticalmodelspredictingeightoutcomes(i.e.,fourintercepts
andfourslopesofexternalizingproblembehaviours),theonly
sig-nificantCARAUCgby SESinteractioneffectwasin predictingthe
slopeofself-reporteddelinquency(i.e.,B[SE]=0.34[0.16],=0.14,
p=0.03).ThisinteractionshowedthatforadolescentswithCARAUCg
levels≤0.20SDbelowthemean,lowerSESsignificantlypredicted
higherslope(i.e.,fastergrowth)ofdelinquentbehaviours.Noother
Table3
InteractionEffectsbetweenAdolescentCARAUCgandNeighbourhoodDensityonParent-ReportedExternalizingProblemBehavioursfrom15to18Years.
(N=358) Delinquency(CBCL) Aggression(CBCL)
Intercept Slope Intercept Slope
B(SE)  B(SE)  B(SE)  B(SE) 
Model1
SES(Lvs.H&M) 0.73(0.40) 0.11 0.25(0.15) 0.14 2.47(1.20) 0.15* 0.43(0.31) 0.14 Gender(Girlvs.Boy) -0.28(0.23) -0.07 -0.12(0.07) -0.12 -0.41(0.58) -0.04 0.24(0.15) 0.13
CARAUCg 0.17(0.13) 0.08 -0.06(0.04) -0.11 0.13(0.33) 0.03 0.05(0.09) 0.05
NBDensity 0.04(0.04) 0.02 0.02(0.04) 0.03 -0.17(0.24) -0.03 0.06(0.08) 0.06
Modelfit CFI=0.94,RMSEA=0.07,SRMR=0.05 CFI=0.98,RMSEA=0.06,SRMR=0.05 Model2
SES 0.77(0.40) 0.12 0.25(0.15) 0.14 2.61(1.17) 0.16* 0.42(0.31) 0.14
Gender -0.26(0.22) -0.06 -0.13(0.07) -0.12 -0.33(0.57) -0.03 0.23(0.15) 0.12
CARAUCg 0.18(0.12) 0.09 -0.06(0.04) -0.11 0.15(0.31) 0.03 0.05(0.09) 0.05
NBDensity -0.02(0.08) -0.01 0.02(0.04) 0.04 -0.35(0.22) -0.07 0.08(0.07) 0.08
CARAUCg*NBDensity 0.31(0.12) 0.15** -0.02(0.04) -0.03 0.98(0.37) 0.19** -0.10(0.10) -0.11 ModelFit CFI=0.94,RMSEA=0.06,SRMR=0.05 CFI=0.97,RMSEA=0.05,SRMR=0.05
Note:Scoreofdelinquencyandaggressionwasbasedonthemeanscoreofmotherandfatherreports.CBCL=ChildBehaviourChecklist.SES=SocialEconomicStatus.Lvs. H&M=Lowvs.High&Medium.CARAUCg=CortisolAwakeningResponseAreaUndertheCurvewithrespecttoground.NB=Neighbourhood.B=unstandardizedregression coefficient.=standardizedregressioncoefficient.*p<0.05;**p<0.01.
0 1 2 3 4
Low NB Density High NB Density
De lin q ue n cy C B C L Low CARaucg High CARaucg 0 1 2 3 4
Low CARaucg High CARaucg
De lin q ue n cy C B C L Low NB Density High NB Density
a
b
Fig.1.InteractionEffectsbetweenAdolescentCARAUCgandNeighbourhoodDensityonDevelopmentofParent-ReportedDelinquency.
3 4 5 6 7 8 9 10 11 12
Low NB Density High NB Density
Aggression CBCL Low CARaucg High CARaucg 3 4 5 6 7 8 9 10 11 12
Low CARaucg High CARaucg
Aggression CBCL
Low NB Density High NB Density
a
b
Fig.2. InteractionEffectsbetweenAdolescentCARAUCgandNeighbourhoodDensityonDevelopmentofParent-ReportedAggression.
3.6. InteractioneffectsbetweenCARAUCiandneighbourhood
densityonthedevelopmentofexternalizingproblembehaviours
WeadditionallytestedwhetherthemoderatingeffectsofCAR
werespecifictocortisolactivity(i.e.,AUCg),oralsooccurredfor
cortisol reactivity (i.e., AUCi). Our analyses showed significant
CARAUCi byneighbourhooddensityinteractioneffectsin
predict-ingadolescentself-reporteddelinquency(i.e.,B[SE]=-0.21[0.11],
=-0.08, p=0.06inpredictingtheintercept;B[SE]=0.08[0.04],
=0.11, p=0.03 in predicting the slope) and adolescent
self-reported aggression (i.e.,B [SE]=0.41 [0.19], =- 0.08, p=0.03
Table4
InteractionEffectsbetweenAdolescentCARAUCgandNeighbourhoodDensityonSelf-ReportedExternalizingProblemBehavioursfrom15to18Years.
(N=358) Delinquency(YSR) Aggression(YSR)
Intercept Slope Intercept Slope
B(SE)  B(SE)  B(SE)  B(SE) 
Model1
SES(Lvs.H&M) 1.07(0.60) 0.12 -0.46(0.18) -0.18** 3.48(1.12) 0.20** -0.85(0.29) -0.22** Gender(Girlvs.Boy) -0.52(0.32) -0.10 -0.04(0.11) -0.02 -0.42(0.60) -0.04 0.09(0.17) 0.04
CARAUCg 0.33(0.18) 0.13 -0.12(0.07) -0.15 0.84(0.35) 0.16* -0.20(0.11) -0.16
NBDensity 0.13(0.14) 0.05 0.04(0.05) 0.05 -0.16(0.24) -0.03 0.05(0.07) 0.04
Modelfit CFI=0.97,RMSEA=0.05,SRMR=0.04 CFI=0.98,RMSEA=0.04,SRMR=0.03 Model2
SES 1.11(0.59) 0.13 -0.46(0.18) -0.18* 3.50(1.12) 0.20** -0.85(0.29) -0.22**
Gender -0.50(0.32) -0.09 -0.04(0.11) -0.02 -0.41(0.60) -0.04 0.10(0.18) 0.04
CARAUCg 0.34(0.18) 0.13 -0.12(0.07) -0.15 0.84(0.35) 0.16* -0.19(0.11) -0.16
NBDensity 0.08(0.13) 0.03 0.04(0.04) 0.05 -0.20(0.25) -0.04 0.04(0.07) 0.03
CARAUCg*NBDensity 0.30(0.15) 0.11* 0.02(0.05) 0.03 0.21(0.29) 0.04 0.05(0.10) 0.04 ModelFit CFI=0.97,RMSEA=0.04,SRMR=0.04 CFI=0.97,RMSEA=0.04,SRMR=0.04
Note.YSR=YouthSelfReport.SES=SocialEconomicStatus.Lvs.H&M=Lowvs.Medium&High.CARAUCg=CortisolAwakeningResponseAreaUndertheCurvewithrespect toground.NB=Neighbourhood.B=unstandardizedregressioncoefficient.=standardizedregressioncoefficient.*p<0.05;**p<0.01.
1 2 3 4 5 6 7
Low NB Density High NB Density
Delinquency YSR Low CARaucg High CARaucg 1 2 3 4 5 6 7
Low CARaucg High CARaucg
Delinquency YSR
Low NB Density High NB Density
a
b
Fig.3. InteractionEffectsbetweenAdolescentCARAUCgandNeighbourhoodDensityonDevelopmentofSelf-ReportedDelinquency.
in predicting the slope).The interaction pattern indicated that
foradolescentswithahighCARAUCi,delinquencyandaggression
were overall higher for those from a high density
neighbour-hood.TherewerenosignificantCARAUCibyneighbourhooddensity
interactioneffectsonparent-reportedexternalizingproblem
beha-viours.
4. Discussion
Thislongitudinalstudyexaminedtheeffectsofneighbourhood
densityandadolescents’morningcortisolactivityonthe
devel-opmentofexternalizingproblembehavioursfrommiddletolate
adolescence. Our results indicated that neither neighbourhood
densitynoradolescents’cortisolactivityingeneraldidhavemain
effects ontheinitiallevelsand developmental changesof
ado-lescent externalizing problembehaviours.However, therewere
significantinteraction effects between CARAUCg and
neighbour-hooddensityinpredictingadolescents’aggressionanddelinquency
over a period of four years. Level of CARAUCg moderated the
effectsofneighbourhooddensityonthedevelopmentof
adoles-cents’externalizing problembehaviours. Thesefindingsprovide
supportforperson-environmentperspectivesandalsobiological
sensitivitytheoryonthedevelopmentofadolescent
externaliz-ingproblembehaviours(Ellis&Boyce,2011;Magnusson&Stattin,
2006;Susman&Ponirakis,1997).
4.1. Maineffectsofneighbourhooddensityonadolescent
externalizingproblembehaviours
Thelackofmaineffectsofneighbourhooddensityonaggressive
and delinquent behaviours is consistentwith someprior
stud-ies (Gillis, 1974; Reijneveld et al.,2010; Wichstrom, Skogen, &
Osia,1996).However,findingswerecontradictorytoresultsfrom
otherstudiesreportingpositiveassociations(Hardenetal.,2009;
Schmitt,1957),negativerelations(Bao,Haas,&Pi,2004),and
non-linearlinks(Browningetal.,2010).Theinconsistentfindingscould
bedue toneighbourhooddifferencesacrossstudies.Specifically,
thesestudiesarefromdifferentcountriesordifferentareaswithin
acountry,anditmightbethatindifferentcountriesorareas
liv-inginahighdensityenvironmentimpliesdifferencesintermsof
distributionsofeducationalresources(Boyle,Georgiades,Racine,&
Mustard,2007),incomeequality(Kawachi,2000),orethnic
compo-sition(Semyonov,Gorodzeisky,&Glikman,2012)whicharerelated
toadolescentexternalizingproblembehaviours(MacDonald,Hipp,
&Gill,2013;Schindleretal.,2015;Wilkinson&Pickett,2009).
Theinconsistencycouldalsoreflectotherdifferencesin
meth-ods amongstudies, suchasthecontrolling variables and study
design.Forinstance,someresearcherscontrolledfor
unemploy-ment,meanincome,andeducationallevel(Reijneveldetal.,2010),
others controlledfor parental divorce(Weenink,2011).
Differ-encein these measuredcovariates as statistical controlsmight
stud-ies adopted a cross-sectional design which does not provide
informationonthedirectionofeffects.Amoreidealapproachis
toadoptalongitudinaldesignthattakesintoaccountthemobility
ofadolescents(i.e.,movefromareasofhightolowdensity
neigh-bourhoodenvironmentand viceversa).It hasbeenshown that
somesignificant relations between neighbourhooddensity and
externalizingproblembehavioursdemonstratedincross-sectional
analysesdisappearin longitudinalinvestigations(Hardenet al.,
2009).Furthermore,individualcharacteristicsmightmoderatethe
associationbetweenneighbourhoodenvironmentsand
external-izing problembehaviours (Jaffee et al., 2015; Obradovi ´c, Bush,
Stamperdahl, Adler, & Boyce, 2010). In sum, more research is
neededonfactorsthatqualifytheeffectofneighbourhooddensity
onadolescentexternalizingproblembehaviours.
4.2. MaineffectsofCARAUCgonadolescentexternalizingproblem
behaviours
Thecurrentstudyindicatesthatingeneraltherewereno
sig-nificanteffectsofCARAUCgonthedevelopment(i.e.,initiallevels
anddevelopmentalchanges)ofexternalizingproblembehaviours
acrossmiddletolate adolescence.The onlyexception wasthat
CARAUCgwaspositivelylinkedtotheinterceptofadolescent
self-reportedaggression.Thenon-significantlinksbetweenCARAUCg
andexternalizingproblembehavioursfoundinthecurrent
com-munitysampleare inaccordance withfindingsfrom studiesof
earlytomiddlenormallydevelopingadolescents(Klimes-Dougan
et al., 2001; Sondeijker et al., 2008). In addition, the lack of
CARAUCg-externalizingproblembehavioursassociationsarein
gen-eralconsistentwitha priormeta-analyticreviewconcludingno
associationsbetween(morning)basalcortisollevelsand
external-izingbehavioursinadolescentsampleswithagesrangedfrom12
to19years(Alinketal.,2008).
Findings from the current study are, however, inconsistent
withstudiesshowingnegativeassociationsbetweenmorningbasal
cortisollevelsandadolescents’externalizingproblembehaviours
(Pajeretal.,2001;Shirtcliffetal.,2005;Shoaletal.,2003).The
divergentfindingsontherelationbetweenCARAUCgand
external-izingbehaviourproblemscouldbeduetostudiesnotaccounting
for issues of sample characteristics such as severity of
exter-nalizing behaviours and comorbidity of traits or internalizing
problembehaviours (Marsmanet al., 2008; Ruttleet al., 2011;
Shirtcliff&Essex,2008).Somestudieshavesuggestedthatthe
neg-ativecortisol-externalizingproblembehavioursassociationmay
becharacteristic for a particularsevere subgrouponly, suchas
adolescents withconduct disorders, oppositional defiant
disor-der,oradolescentswithpersistentaggressivebehaviours(Platje
et al.,2013; Shoal et al.,2003). Thisassociation might alsobe
dependentonthecomorbidityof individualtraits(e.g.,
callous-unemotional traits; Hayes & Matthes, 2009) and internalizing
problembehaviours(Shirtcliff&Essex,2008).Hence,further
stud-iesarenecessarytounderstandtherelationshipbetweenCARAUCg
andexternalizingproblembehaviours.
4.3. ModeratingeffectsofCARAUCgontheassociationbetween
neighbourhooddensityanddevelopmentofadolescent
externalizingproblembehaviours
We found significant interaction effects between CARAUCg
and neighbourhood density in predicting adolescent
parent-reported aggression and delinquency and self-reported
delin-quency.CARAUCgappearedtobeamoderatorof theassociation
betweenneighbourhood density and adolescents’ externalizing
problem behaviours. In predicting parent-reported aggression
and delinquency, the interaction effects showed a contrastive
pattern—meaningthat the valence of theassociations between
neighbourhood density and adolescent externalizing problem
behaviours runs in the opposite direction at relatively high
versuslowlevelsofthemoderatorCARAUCg(Belsky,
Bakermans-Kranenburg, & Van IJzendoorn, 2007). The contrastive pattern
suggeststhathighCARAUCgcouldbeaprotectivefactorinalow
den-sityenvironmentbutariskfactorinahighdensityenvironment.In
contrast,lowCARAUCgcouldbeaprotectivefactorinahighdensity
environmentbutavulnerablefactorinalowdensityenvironment.
ThesefindingssuggestthatneitherhighnorlowCARAUCgisa
pro-tectiveorvulnerablefactorassuch.Instead,itismoreofthematch
andcombinationofthelevelofanindividual’sCARAUCgandthelevel
ofdensityofaneighbourhoodenvironmentthatpredictsthe
devel-opmentofparent-reportedexternalizingproblembehaviours.We
alsoobservedinteractiveeffectsbetweenCARAUCgand
neighbour-hood density in predicting adolescent self-reported delinquent
behaviours.ForadolescentswithlowCARAUCg,therewasno
signif-icantassociationbetweenneighbourhooddensityanddelinquent
behaviours,whereasforthosewithhighCARAUCg,therewere
posi-tiveassociationsbetweenneighbourhooddensityanddelinquency.
Thesefindingstogetherprovidesupportforgeneralinteractive
per-spectivesandhormones-contextinteractionstheory,claimingthat
individuals’developmentaloutcomesaredependenton
interac-tiveprocessesbetweenindividuals’characteristicsandcontextual
factors(Magnusson&Stattin,2006;Susman&Ponirakis,1997).
In general, the interactivepatterns between neighbourhood
densityandCARAUCginpredictingparentalreporteddelinquency
andaggression andself-reporteddelinquencysuggestthat high
CARAUCgisasensitivityfactorwhichincreasesthenegativeeffects
of adverseenvironments (i.e.,highdensity neighbourhood)but
enhancesthebeneficialeffectsofpositiveenvironments(i.e.,low
density neighbourhood). Hence, these results support the
bio-logicalsensitivitytheory(Ellis&Boyce,2011).Morespecifically,
thesefindingsindicatethathyper-arousedadolescentsmighthave
highersensitivitytowardsnotonlythestressesthatcouldbe
gen-eratedbyahighdensityenvironmentsuchasovercrowding,noise,
impersonalspace,orlimitedaccessibilitytoresourcesbutalsothe
benefitsthatmightbeprovidedinalowdensityneighbourhood
suchasphysicalspace,quietness,oravailabilityofresources.For
instance,ithasbeensuspectedthataseriousconsequenceof
high-densitylivingmaybearealorperceivedinabilitytocontrolthe
environmentandtoregulateone’ssocialinteractions(Rodin,1976).
Itislikelythathyper-arousedadolescentssensehigherstressof
inabilitytocontroloverwhatgoesoninaneighbourhoodandover
interpersonalchange,thanless-arousedadolescents,thustheyact
moredelinquentandaggressiveinhigh-densityenvironment.In
sum,findingscorroborateaconceptualizationofstressactivityas
biologicalsensitivitytocontextbyshowingthathighCARAUCgcan
bothhinderandpromoteadaptivefunctioning.
In predicting parent-reported externalizing problem
behaviours, we observed detrimental effects of a
combina-tionofahypo-arousedbiologicalresponse,namelyalowcortisol
awakingresponse,andalowdensityneighbourhoodinpredicting
adolescents’ externalizing problem behaviours. This finding is
inconsistentwiththetheoryofbiologicalsensitivitytocontextin
whichadolescentswithalowCARAUCgwouldbeexpectedtobe
notinfluencedbytheirenvironment.Thisfindingmighthowever
providesomesupportforthesensationseekingtheoryofantisocial
behaviours(Raine,1993;Zuckerman&Neeb,1979).Alowdensity
neighbourhood environment is probably not only quieter and
calmer but also provides fewer opportunities for involvement
in interpersonal contacts or social events than a high-density
neighbourhoodenvironment.Hence,suchanenvironmentmight
be boring for hypo-aroused adolescents as they are less likely
tobearousedby theirenvironment. Aslow arousal represents
anaversivephysiologicalcondition,toattainahigherandmore
stimulation through delinquent and aggressivebehaviours in a
low-densityneighbourhood.
There was no significant association between CARAUCg and
neighbourhooddensityinthisstudy,thusitisratherunlikelythat
therewouldbeabidirectionalprocesswherebycertainadolescents
develophigherCARAUCgasaresultoflivinginhighdensity
neigh-bourhoods,orviceversa.Hence,theinteractioneffects between
CARAUCg andneighbourhood densitywerenot likely tobe
con-foundedbyanybidirectionaleffectsbetweentheindependentand
moderatingvariables.
Ourstudy showedsignificantmain effectsof SESon
aggres-sivebehaviours,withadolescentsfromlowSESfamiliesreporting
higherlevelsofparent-andself-reportedaggression.Theseresults
areconsistentwithfindingsfromalargebodyofprevious
stud-ies(seeBradley&Corwyn,2002forareview).However,despite
thesemaineffects,theonlysignificantSESbyCARAUCginteraction
effectwasontheslopeofself-reporteddelinquency.These
find-ingssuggest thatthemoderatingeffects ofCARAUCgonthelink
betweenneighbourhooddensityandexternalizingproblems
can-notbeexplainedbythepotentialoverlapbetweenneighbourhood
densityandSES.Theseresultsfurtherhighlighttheimportanceof
investigatingtheinteractioneffectsbetweenneighbourhood
den-sityandCARAUCg.
4.4. ModeratingroleofCARAUCiontheassociationbetween
neighbourhooddensityanddevelopmentofadolescent
externalizingproblembehaviours
Wefoundsignificantinteractioneffectsbetweencortisol
reac-tivityCARAUCi neighbourhooddensity inpredictingadolescents’
self-reporteddelinquentandaggressivebehaviours.The
interac-tionpatternsuggeststhatadolescentswithahigherCARAUCi(i.e.,a
biggerchangeincortisollevelspost-awakening)mightbemore
likely tobenegatively influencedby a highdensity
neighbour-hood and therefore they develop higher externalizing problem
behaviours.Thesefindingsingeneralseemtobeconsistentwiththe
interactionpatternsbetweenneighbourhooddensityandcortisol
activityCARAUCginpredictingexternalizingproblembehaviours,
providingfurthersupportforthebiologicalsensitivitytheory(Ellis
&Boyce,2011).
4.5. Limitationsandstrengths
Severallimitationsofthecurrentstudyhavetobenoted.First,
findingsmightbelimited toyouthsfrommiddletolate
adoles-cence,theageperiodcoveredbythis study.Theeffectsofbasal
cortisollevelsonexternalizing problembehaviourswere found
tobeage dependent(Alink etal., 2008).It ispossible thatthe
interactioneffectsbetweenCARAUCgandneighbourhooddensityin
predictingexternalizingproblembehavioursarealsodependenton
individuals’developmentalstages.Second,thisstudyfocusesona
generalpopulationsample.Thus,resultsreflectnormativelevelsof
aggressiveanddelinquentbehaviours.Therefore,theresultscannot
begeneralizedtoclinicallyreferredyouthswithconductdisorders
oroppositionaldisruptivedisorders.Third,inourstudyCARwas
measuredonasingledayinsteadofovermultipledays.Althougha
recentreviewhasshownthatthecortisolawakingresponsehas
relatively highbetween-visit reliabilities (r=0.33–0.67;Golden,
Wand,Malhotra,Kamel,&Horton,2011),factorssuchassubstance
use,timeofawakening,andstressfulexperiencescouldpotentially
affectCAR(e.g.,Williams,Magid,&Steptoe,2005).Inourstudy,
wepartiallyaddressthelimitationofonesinglemeasurementby
testingtheeffectsofsomepotentialcovariatesandincludingthe
significantpredictorsinouranalyses.However,tofurtherincrease
thereliabilitiesofCARmeasurement,futurestudiesareencouraged
touserepeatedmeasuresovermultipledays(Hellhammer,Fries,
Schweisthal,Schlotz,Stone,&Hagemann,2007).
This studyhas alsoseveral strengths. It is the first attempt
investigatinginteractioneffectsbetweenneighbourhooddensity
andindividuals’CARonthedevelopmentofadolescent
external-izingproblembehaviours.Themultipleinformantapproach(i.e.,
both self- and parent-report) adopted in this study provided a
morecomprehensivepictureoftheinteractionprocessesbetween
neighbourhooddensityandCAR.Moreover,thelongitudinaldesign
allowedustoinvestigatewhethertheinteractionbetween
neigh-bourhood density and CAR predicts changes in externalizing
problembehaviourswithinindividuals.Furthermore,thepresent
studycontrolledforthepropertyvalueoftheneighbourhoodas
wellasfamily SES.Asitispossiblethatneighbourhoods witha
highdensityareonaveragepoorerthanthosewithlowdensity,
controllingforbothfamilySESandneighbourhoodwealthvalue
couldprecludethisselectionbiaswhichcouldobscurethe
indepen-denceoftheeffectsofneighbourhooddensity(Hipp,Butts,Acton,
Nagle,&Boessen,2013).Finally,thefocusonneighbourhood
den-sityratherthangeneralneighbourhooddeprivationprovidesmore
specificinformationonthecontributionofneighbourhood
charac-teristicsonadolescents’externalizingproblembehaviours,which
iscalledforinarecentreview(VanVuuren,Reijneveld,Vander
Wal,&Verhoeff,2014).
5. Conclusion
The present study revealed significant interaction effects
betweenneighbourhooddensityandCARinpredictingadolescent
aggressiveanddelinquentbehaviours.Ingeneralitindicatesthat
individuals’biologicalsensitivitytoenvironmentalcontextplaysan
importantroleinmoderatingtheeffectsofneighbourhooddensity
onadolescents’adjustment. Findingshighlightthecomplexities
thatcontributetoadolescents’externalizingproblembehaviours
andsupportthetheoreticalideaofinteractiveprocessesbetween
environmentalcontextandindividualscharacteristicscontributing
toadolescents’developmentaloutcomes.
Acknowledgements
Data of the RADAR (Research on Adolescent Development
AndRelationships)studywereused.RADARhasbeenfinancially
supported by main grants from the Netherlands Organisation
forScientific Research(GB-MAGW480-03-005,GB-MAGW
480-08-006, CID; 024.001.003) and various other grants from the
NetherlandsOrganisationforScientificResearch,theVU
Univer-sityAmsterdam,andUtrechtUniversity.Thisresearchisfundedby
theSeedMoneyDOY(DynamicsofYouth),UtrechtUniversity,The
NetherlandsandbytheEuropeanResearchCouncilunderthe
Euro-peanUnion’sSeventhFrameworkProgramme(FP/2007-2013)/ERC
GrantAgreementn.615159(ERCConsolidatorGrant
DEPRIVED-HOODS, Socio-spatial inequality,deprived neighbourhoods, and
neighbourhoodeffects).Thefundersofthisresearchhadnorole
inthestudydesign,analysis,andinterpretationofdata,writingthe
report,orinthedecisiontosubmitthepaperforpublication.
References
Achenbach,T.M.(1991).Manualforthechildbehaviorchecklist/4-18and1991 profile.BurlingtonVT:UniversityofVermont,DepartmentofPsychiatry. Alink,L.R.,VanIJzendoorn,M.H.,Bakermans-Kranenburg,M.J.,Mesman,J.,Juffer,
F.,&Koot,H.M.(2008).Cortisolandexternalizingbehaviorinchildrenand adolescents:Mixedmeta-analyticevidencefortheinverserelationofbasal cortisolandcortisolreactivitywithexternalizingbehavior.Developmental Psychobiology,50,427–450.http://dx.doi.org/10.1002/dev.20300 Bao,W.N.,Haas,A.,&Pi,Y.(2004).Lifestrain,negativeemotions,and