ContentslistsavailableatScienceDirect
EarlyChildhoodResearchQuarterly
Expandingnotionsofsocialreproduction:Grandparents’educationalattainmentandgrandchildren’scognitiveskills
JasonL.Fergusona,1,2,DouglasD.Readyb,∗,1
ab
UniversityofCalifornia,Berkeley,410BarrowsHall,Berkeley,CA94720,UnitedStates
TeachersCollege,ColumbiaUniversity,525W.120thSt.,Box67,NewYork,NY10027,UnitedStates
articleinfoabstract
InheritedprivilegeandstatusremainpowerfulfactorsinthedistributionofopportunityinAmericanlife.Thesetransfersofsocioeconomicresourcesacrossgenerationsarefacilitatedbythelinksbetweenadulteducationalattainmentandchildren’scognitiveskills.Ourcurrentstudyexpandsthenotionofsocialreproductionbeyondthisnarrowtwo-generationapproachtoinvestigatethelinksbetweengrandparents’educationalattainmentandtheirgrandchildren’sacademicabilities.Usinganationallyrepresentativesampleofover13,000childrenwhoparticipatedintheEarlyChildhoodLongitudinalStudy—KindergartenCohort(ECLS-K),wefindthatfamilialadvantagesinhumancapitalpersistovertimeandthattheseadvantagesareassociatedwithimprovedcognitiveoutcomesamonglatergenerations.Evenaftercontrollingforawidearrayofsocioeconomicanddemographiccharacteristics,youngchildrenwithcollege-educatedgrandparentspossessstrongerliteracyandmathematicsskillsatthestartoffor-malschooling.Propensityscoreapproaches,whichaddressconcernsregardingtheendogeneityinherentinthetopic,yieldsimilarresults,suggestingtherobustnessofourfindings.
©2010ElsevierInc.Allrightsreserved.
Articlehistory:
Received17March2010Receivedinrevisedform17September2010
Accepted2October2010Keywords:GrandparentsSocialclassInequalityAchievement
Large-scaledemographicshiftssincetheendofWorldWarIIhavechangedthelandscapeofgrandparenthoodintheUnitedStates.Post-warfamiliesgenerallyhavetighter,moresustainedlinkagesbetweengrandparentsandgrandchildren,andincreasedlifeexpectancyhasallowedmoregrandparentstobecomeactiveparticipantsinthelivesoftheirgrandchildren(Barranti,1985).Moreover,decreasingfertilityrateshaveenhancedtheper-childinvestmentgrandparentscanmake,whileincreasesinparentaldivorcehavefurtherheightenedgrandparents’roles(Uhlenberg&Kirby,1998).Inrecentdecades,grandparentshaveexpandedtheirresponsibilitiesfurthertoencompassco-residencyandpri-marycaregiverstatus,withroughly6.6millionchildrenlivingwiththeirgrandparents.Nearly23%ofthesechildrenhavenoparentsinresidence—a66%increasefrom1970(Simmons&Dye,2003).Despitethesetransformationstofamilystructures,theextantlit-eratureontheassociationsbetweensocioeconomicstatusandchildoutcomesgenerallymaintainsafocusalmostexclusivelyontwogenerations—parentsandtheirchildren.Thisiscertainlyaconceptuallimitation,asgrandparentsrepresentanincreasinglyimportantlinkinthechainofsocialreproduction.
Ourcurrentstudyrecognizesthesubstantialtimeandresourcesgrandparentsofteninvestintheirgrandchildren’slives.Indeed,familiesrarelyraisetheirchildrenincompleteisolation.Theseintergenerationalfamilialnetworksoftenserveasimportantsourcesoflogistic,financial,andpsychologicalsupport.Butaswithparents,grandparentsdifferwidelyintheresourcestheybringtobearinchildrearing.Assuch,grandparents’influenceislikelyconditionedbysocio-demographicfactorssuchastheirhealth,economicsecurity,maritalstatus,proximity,andthefocusofthisstudy—educationalattainment.Specifically,theseanalysesinvesti-gatetheextenttowhichgrandparents’educationisassociatedwiththeiryounggrandchildren’scognitiveskillsevenafteraccountingforthesocioeconomicbenefitsthesegrandparentspassedontotheirownchildren.
1.Background
Wesituatethisstudywithinbothmicro-andmacro-levelconceptualizationsofchilddevelopment.Atthemicrolevel,ouranalysesareinformedbydevelopmentalecologicaltheory,whichrecognizesthatcognitivedevelopmentflowsfrommultiplesourcesrelatedtochildren’senvironments(seeBronfenbrenner,1979;Bronfenbrenner&Morris,1998).Thesocial,economic,andphysi-calcontextsinwhichyoungchildren’slivesareenmeshedformanecologyofinterconnectedfactorsthatsupporttheirsocialandintel-lectualgrowth(Brooks-Gunn,1995;Eccles,2005).Thesemultiple
∗Correspondingauthor.Tel.:+12126783850.
E-mailaddress:ddr2111@columbia.edu(D.D.Ready).1
Theauthorsarelistedalphabeticallyandeachcontributedequallytothiswork.2
JasonFerguson’seffortsweresupportedinpartbyaGraduateResearchFellow-shipfromtheNationalScienceFoundation.
0885-2006/$–seefrontmatter©2010ElsevierInc.Allrightsreserved.doi:10.1016/j.ecresq.2010.10.001
J.L.Ferguson,D.D.Ready/EarlyChildhoodResearchQuarterly26(2011)216–226217
spheresofinfluence,reflectedintheattributesandcapacitiesoffamilies,caregivers,peers,andneighborhoods,mustbeinterpretedwithanunderstandingofthetightlinksbetweensocioeconomiccircumstancesandchilddevelopment(seeMcLoyd,1998).Indeed,aconsiderablebodyofresearchinthedevelopmentalecologicaltraditionfocusesontheassociationsbetweenchildren’smate-rialconditionsandtheiroverallwell-being(seeBrooks-Gunn&Duncan,1997;Smith,Brooks-Gunn,&Klebanov,1997).
Centraltotheresearchpresentedhere,however,istherecog-nitionthatchildren’ssocialecologiesarenotbornanewwitheachgeneration.Rather,theyarepartlytransmittedwithinfamiliesfromonegenerationtothenextintheformofresources,attributes,andskills.Thisnecessarilyrequiresustoextendbeyondtraditionalcon-ceptualizationsofdevelopmentalecologicaltheory.Inresponse,atthemacrolevelweturntosocialreproductiontheory,whichexam-ineshowsocioeconomicrelationsandstructuresarereproducedfromonegenerationtothenext(seeBourdieu,1977,1984,1985;Bowles&Gintis,1976,2001;Weber,1946;Willis,1977).Through-outthisstudy,wedistinguishbetweensocialinequality—theextenttowhichstatusandresourcesareunequallydistributedatagivenpointintime,andsocialmobility—thedegreetowhichsocioeco-nomicoriginspredicttheeducational,occupational,andeconomicoutcomesofindividualsandthoseofsubsequentgenerations.Thisfocusisimportant,asU.S.domesticsocialpolicyhashistoricallyconcerneditselfmorewithsocialmobilityandopportunitythanwithequalizingthedistributionofsocialandeconomicresources(Corcoran,2001).Assuch,effortsaimedatreducingthenegativeeffectsofsocioeconomicdisadvantageonyoungchildrenmustcon-siderhowsuchdisadvantageisreproducedovertime.
2.Socialreproduction
Inheritedprivilegeandstatusremainpowerfulfactorsinthedis-tributionofopportunityinAmericanlife.Indeed,oneoftheleastdisputedfindingswithinthesocialscienceliteratureisthatparentspassonaportionoftheirsocioeconomicadvantage(ordisadvan-tage)totheirchildren(seeBrooks-Gunn&Duncan,1997;Jencksetal.,1972;Mayer,1997).Thetopichasoccupiedsocialscientistsfornearlyacentury.Twooftheearlieststudiesinthisarena,LyndandLynd’s(1929)MiddletownandHollingshead’s(1949[1975])Elmtown’sYouth,bothportrayedtheintricateandenduringclassstructuresintworelativelysmallandraciallyhomogenousU.S.Midwesterntowns,anddetailedtheimportanceofparentalhuman,social,cultural,andeconomiccapitaltochildoutcomesandtotheperpetuationofsocioeconomichierarchiesacrossgenerations.
Sincetheseearlystudies,anexpansivebodyofresearchhasexaminedtheextenttowhichsocioeconomicstatusis“shuf-fled”fromonegenerationtothenext(Becker,1991;Becker&Tomes,1986;Blau&Duncan,1967;Ganzeboom,Treiman,&Ultee,1991;Mazumder,2005;Musick&Mare,2006).AlthoughtheU.S.experiencesconsiderableupwardanddownwardsocialmobility,children’ssocioeconomicoriginsremainstronglyassociatedwiththeiradultoutcomes.Authorshavereportedcorrelationsbetweenmulti-yearindicatorsoffatherandson’searningsofbetween0.4and0.6(Corcoran,2001;Mazumder,2005;Solon,1999).Putanotherway,over30%ofchildrenbornintofamilieswithincomesinthebottomdecilewillfindthemselvessimilarlydisadvantagedasadults(Hertz,2005).However,theextentofincomemobilityacrossgenerationsvariesdramaticallybyracialbackground.Lessthan17%ofWhitechildrenbornintothebottomincomedecilewillfindthemselvesthereasadults,comparedtoover40%ofBlackchildren(Corcoran,2001).Similarpatternsemergewitheducationalattain-ment;correlationsbetweenfatherandson’scompletedschoolingarereportedtobeintherangeof.35(Mulligan,1999).Theseasso-ciationsarereflectedinthefactthatstudentswhoseparentswere
collegegraduatesarealmosttwiceaslikelytoattendcollegecom-paredtothosewhoseparentshadobtainedahighschooldiplomaorless(U.S.DepartmentofEducation,2001).Similarly,66%ofstu-dentswithfamilyincomeinthetopquartilebeginpost-secondaryschoolingcomparedtoonly28%ofstudentsinthebottomincomequartile(Ellwood&Kane,2000).
Whiletheabovestudiesfocusedonadultoutcomes,extantresearchhasfoundevidenceofsocialreproductionduringearlierstagesofthelifespan.Becausechildrenhavenotyetattainedtheirownsocialclass,researchershaveinsteadinvestigatedthelinksbetweenparentalsocialclassandchildren’sacademicoutcomes,usuallystandardizedtestscores.Evenpriortothestartofformalschooling,higher-SESchildrenpossessconsiderablystrongercog-nitiveskillsthantheirlower-SESpeers(Brooks-Gunn&Duncan,1997;Entwisle,Alexander,&Olson,1997;Lee&Burkam,2002;Smithetal.,1997).Myriadexplanationshavebeenofferedforthisinequality,includingdisparitiesinfamilyandneighborhoodresources,thepersistentlinksbetweensocioeconomicstatusandrace/ethnicity,andthenegativesocial,environmental,andphysicaleffectsassociatedwithpoverty(forareview,seeRothstein,2004).Inaddition,higher-SESparentsareoftenabletodevotemorehourstoparentingandaremorelikelytohavecognitivelystimulatingverbalexchangeswiththeirchildreninearlychildhood(Hart&Risley,1992;Heath,1983;Lareau,2003).
Asthediscussionabovesuggests,therelationshipbetweenparentalcharacteristicsandchildoutcomeshasbeenwellestab-lished.However,muchofthisliteraturehasfocusedexclusivelyontwogenerations—parentsandchildren.Thissuggestsanimpor-tantlimitation,asgrandparentscanalsoplayacentralroleintheintergenerationaltransmissionofsocietaladvantage.Ouraimhereistodiscernwhetherthelinksbetweengrandparents’educationalattainmentandtheirgrandchildren’scognitiveperformanceoper-atesolelythroughparents,astheabovestudiessuggest,orwhethertheassociationremainsafterholdingconstantparentaleducationandahostofotherparentalsocioeconomiccharacteristics.
3.Threegenerations:consideringtheroleofgrandparentsSystematicresearchongrandparenthoodbeganduringthelat-terhalfofthetwentiethcentury;previously,studieswerebasedonanecdotalevidencethatportrayedgrandparentsasadistur-bancetothenormalfunctioningofthenuclearfamily(Tomlin,1998).Asrecentlyasthe1980s,studiesessentiallyignoredtheinfluenceofgrandparentsongrandchildren,arguingthattherealforcesofsocializationoperatedonlythroughparents(Szinovacz,1998).Adevelopingliterature,however,isexaminingtheassocia-tionsbetweengrandparentsocialclassandgrandchildoutcomes.Aswiththetwo-generationstudiesdescribedabove,researchadoptingathree-generationapproachcanbedividedintooneoftwodistinctgroups.Thefirstgroupexamineshowgrandpar-entsocialclassinfluencesthesocialclassultimatelyattainedbytheiradultgrandchildren.Thesecondareaofinquiryfocusesontheconnectionsbetweengrandparentcharacteristicsandtheaca-demicskillspossessedbytheiryounggrandchildren.Intermsofadultgrandchildren,studieshavereportedgrandparents’edu-cation,occupation,andincometobenon-significantfactorsinoccupationalstatusandeducationalattainmentamongadultgrandchildren,holdingparentalcharacteristicsconstant(Warren&Hauser,1997).OtherstudiesusingFinnishdata,however,havereportedpositivelinksbetweengrandfathersandadultgrandchil-dren’ssocialclass(Erola&Moisio,2007).
Unlikeresearchestimatingtheeffectsofgrandparentsonoldergrandchildren,studiesofgrandparentsandtheiryoungergrandchildrenprovidemoreconsistentfindings.Althoughnottheirexplicitfocus—grandparentattributeswereemployedlargely
218J.L.Ferguson,D.D.Ready/EarlyChildhoodResearchQuarterly26(2011)216–226
ascovariates—authorshavefoundpositiveassociationsbetweengrandparents’education,income,andwealthandgrandchildren’sstandardizedtestscores,evenaftercontrollingforsimilarparentalcharacteristics(Grant,2005;Hill&O’Neill,1994;Yeung&Conley,2008).Thesegrandparenteffectsmayoperatethroughthesym-bolic,direct,andindirectsupportgrandparentsafford.Symbolicsupportcanrepresentapsychologicalbufferthatprovidesfami-liespeaceofmindthroughtheknowledgethatgrandparentsarethereifneeded(Denham&Smith,1989).Grandparenteffectsarealsomanifestthroughmoredirectsupport.Importanttoourfocushere,educatedgrandparentsaremorelikelytomentorandsupportgrandchildren,discusstheirplanswiththem,andimpartspecificskills(King&Elder,1997,1998).Theintergenerationalinfluenceofgrandparenteducationpotentiallyevenoperatesthroughparent-ingpractices,withthechildrenofmore-educatedparentsgenerallyapplyingparentingpracticeswiththeirownchildrenthatarepos-itivelyassociatedwithchildren’scognitivedevelopment(Phillips,Brooks-Gunn,Duncan,Klebanov,&Crane,1998).Finally,grandpar-entsoftendirectlysupportfamiliesascoresidents(Jendrek,1993;Moyi,Pong,&Frick,2004)orprimarycaregivers(Bowers&Myers,1999;Bryson&Casper,1999;Jendrek,1993;Kropf&Burnette,2003),especiallywhenchildrenareyoungorduringtimesoffamil-ialeconomicoremotionalcrisis.
Perhapsthemosttangiblemanifestationofgrandparentinvolvementisfinancialassistance.Importantly,theabilitytopro-videsuchsupportisassociatedwithgrandparents’educationalattainment.Authorshavenotedthatwealthismorereadilypassedontosubsequentgenerationsthanisincome,asitiseasiertotrans-fermaterialassetsthanoccupationsandtheirattendantwages(Becker&Tomes,1986;Wolff,2002).OneexampleiswhatShapiro(2004)aptlycalls“transformativeassets,”whichoftenmanifestastransfersofwealthfromparentstotheiryoungadultchildrenintheformofpaymentsforcollegetuitionorhelpwithadownpay-mentforafirsthome.Saddledwithlessdebtandinfusedwithstart-upcapital,thesefinancialgiftsfromgrandparentsarecru-cialinhelpingyoungfamiliessecuremiddle-classlifestylesthatmightotherwisebebeyondtheirreach.Keytothefocusofourcurrentstudy,these“transformativeassets”oftenallowrecipientfamiliestoliveinmorestableandresource-richneighborhoods(seeDurlauf,1996).Therefore,grandparents’fiscalresourceslikelyrepresentauniquesourceofvariationingrandchildren’scognitiveoutcomes.
4.Challengesofestimatinggrandparenteffects
InadditiontoOLSregressiontechniques,thisstudyalsoemployspropensityscorematching,whichaddressesconcernsregardingtheendogeneityinherentinourresearchquestions.Althoughrandomizationarguablyproducesthemostrobusttreatmentesti-mates,psychologistsandsocialscientistsofteninvestigatetopicsthatfortemporal,logistical,and/orethicalreasonsdonotlendthemselveswelltorandomassignment.Thisiscertainlythecaseinthecontextofourcurrentstudy,asresearchersareclearlyunabletorandomlyassignchildrentograndparentswithvaryinglevelsofeducation.Intheabsenceofrandomization,researchershaveturnedtomultipleregressionapproacheswiththeaimofstatisticallycontrollingforcharacteristicsthatarerelatedtoboththeoutcomeandthelikelihoodofreceivingaparticulartreatment.However,concernsregardingunmeasuredselectionbiascallintoquestiontheabilityoftraditionalregres-siontechniquestoisolateuniquetreatmenteffects(Rubin,1997).Alternately,researchershaveemployedmatchingtechniques,butmatchingcasesacrossmanyvariables—particularlycontinuousvariables—canbequitecumbersome(Rosenbaum&Rubin,1983).Morerecently,researchershaveusedpropensityscorematchingtechniquesasatooltobetterapproximaterandomization.The
propensityscoreissimplythepredictedprobabilityofreceivingthetreatmentbasedonahostofcovariates(Shadish,Cook,&Campbell,2002).Thesescoresarethenusedtocreatematchedpairsoftreatmentandcontrolcaseswhoareessentiallyidenticalacrossobservedcharacteristics.Inthisstudy,eachchildwithcollege-educatedgrandparentsispairedwithachildwhosegrandparentsdidnotattendcollege,butwhohadthesameprobability(propen-sity)ofhavingagrandparentwithacollegedegree.Oneadvantageofthisapproachisthatcasescanbematchedonmanyvariablessimultaneously,whichmakesfarfewerparametricassumptionsthantraditionalregressionapproaches.Wediscussourspecificapproachtopropensityscorematchingingreaterdetailbelow.5.Researchfocus
Theimportanceofparentalsocialclasstochildoutcomesiswelldocumented.Morerecently,however,researchershavebeguntoexpandtheirscopebeyondtwogenerationstoinvestigatethedirectandindirectrolegrandparentsplayintheirgrandchildren’ssocialandacademicdevelopment.Clearly,grandparentinfluencesareimportanttoconsiderforboththeirmicro-leveleffectsonchildren’scognitivegrowthandtheirmacro-levelimplicationsforsocioeconomicreproduction.Asmorechildrenmovetowardstronger,longer-lastingtieswiththeirgrandparents,itwillbecrucialtounderstandtheserelationships,theirnetbenefits,andhowfamiliesandsocialpolicyimpacttheseincreasinglyinfluentialbonds.Thisstudyaimstoexpandtraditionalconceptualizationsofsocialclassbeyondtheparent–childdyadtoexaminehownatu-rallyoccurringvariabilityingrandparenteducationalattainmentinfluencesgrandchildren’scognitiveability.Weusealargeandnationallyrepresentativesampleofover13,000childrenfromtheEarlyChildhoodLongitudinalStudy,KindergartenCohort(NationalCenterforEducationalStatistics,2004a)toaddressesthefollowingquestions:
•ResearchQuestion1:Howcanwecharacterizetheassocia-tionsbetweengrandparents’educationandtheirgrandchildren’ssocio-demographiccharacteristics?Towhatextentaredis-paritiesineducationalattainmentmanifestassocioeconomicinequalitiestwogenerationslater?
•ResearchQuestion2:Afteraccountingforchildren’ssocio-demographicbackgrounds—includingtheirparents’socioe-conomicstatus—istherearesidualrelationshipbetweengrandparents’educationandtheirgrandchildren’scognitiveabil-ity?Inotherwords,isthereanintergenerationallinkbetweengrandparents’educationandtheirgrandchildren’sacademicskills?
•ResearchQuestion3:Giventhatthisstudyemploysanon-experimentaldesign,howrobustareourfindings?Dopropensityscoretechniquesandtraditionalregressionapproachesproducesimilarestimatesoftheassociationsbetweengrandparents’edu-cationandtheirgrandchildren’scognitiveabilities?6.Method6.1.Data
ThisstudyemployeddatafromtheEarlyChildhoodLongitudi-nalStudy,KindergartenCohort(ECLS-K),sponsoredbytheNationalCenterofEducationStatistics(NationalCenterforEducationalStatistics,2004b).TheECLS-Kcollectionofbase-year(1998)datafollowedastratifiedrandomdesignstructure.Theprimarysam-plingunitsweregeographicareasconsistingofcountiesorgroupsofcountiesfromwhichabout1000publicandprivateschoolsofferingkindergartenprogramswereselected.Atargetsampleof
J.L.Ferguson,D.D.Ready/EarlyChildhoodResearchQuarterly26(2011)216–226219
roughlytwenty-fourchildrenwasthendrawnfromeachschool.FromthefullECLS-Ksample,weconstructedouranalyticsampleinthreestages.First,weselectedchildrenwhosebiologicalmotherorfatherwasidentifiedastherespondentfortheparentsurvey.Wethenselectedchildrenforwhombothmathematicsandreadingscoreswerereportedfortheinitialdatacollectionwave(thefallofkindergarten).Finally,weselectedchildrenwhohadanon-missingbase-yeardesignweight.
Ourfinalanalyticsampleincludesasocio-demographicallydiversegroupof13,003children,whichissmallerthantheover-allECLS-Ksampleduelargelytothefactthatchildrendeemednon-Englishproficientwerenotadministeredtheliteracyassess-ments.ChildrenidentifiedasEnglishLanguageLearners(ELLs)wereadministeredtheOralLanguageDevelopmentScale(OLDS).ApproximatelyhalfofthesechildrenwereidentifiedashavingsufficientEnglishskillstocompletethestandardliteracyassess-ment.Assuch,readersshouldbearinmindthatourindicatorofnon-EnglishhomelanguagedoesnotnecessarilyindicatechildrenwhopossessnoEnglishlanguageskills.Moreover,duetotheseselectioncriteria,oursampleissomewhatmoresocioeconomicallyadvantagedthanthefullECLS-Ksample,andcontainsfewerHis-panicandAsianchildren.Oneimplicationisthatourestimatesofthelinksbetweengrandparenteducationandtheirgrandchil-dren’sacademicskillsarelikelytobesomewhatconservative(i.e.,theintergenerationalinequalitywereportbelowmayactuallybesomewhatgreaterintheactualpopulation).6.2.Measures
6.2.1.Dependentvariables
Asprimaryoutcomes,theseanalysesusedmeasuresofchil-dren’smathematicsandliteracyskillsatkindergartenentry.TheECLS-Kcognitiveassessmentswereconductedone-on-onebetweenadesignatedassessorandeachchild.Thoughun-timed,eachassessmentlastedbetween50and70min.Theliteracyassess-menttestedchildren’sabilitiesinthreekeyareas:basicskills(printfamiliarity,letterrecognition,beginningandendingsounds,rhymingsounds,wordrecognition),vocabulary(receptivevocab-ulary),andcomprehension(listeningcomprehension,wordsincontext).Thereliabilityofthereadingassessmentwasquitestrong(0.93;NationalCenterforEducationalStatistics,2004b).Themathematicsassessmentitems,whichmeasuredconceptualandproceduralknowledgeandproblemsolving,canbeorganizedintofiveproficiencyareas:(1)identifyingandcountingnumbersandgeometricshapes;(2)readingone-digitnumbersandcount-ingbeyond10,recognizingsequentialpatterns,andcomparingobjectsizeusingnonstandardobjects;(3)readingtwo-digitnum-bers,identifyingnumbersequencesandordinalpositions,solvingsimplewordproblems;(4)simpleadditionandsubtraction,and;(5)simplemultiplicationanddivision,andrecognizingmorecom-plexpatterns(NationalCenterforEducationalStatistics,2004b).Thereliabilityofthemathematicsassessmentwasalsostrong(0.92forfall;NationalCenterforEducationalStatistics,2004b).Weusedthesetestscoresasz-scores(M=0,SD=1)inouranalyses,whichpermitsthediscussionofresultsineffectsize(SD)units.Thisisimportantgivenourlargesamplesizeandthestatisticalpoweritaffords(seeCohen,1988).
6.2.2.Socioeconomicstatusmeasures
Thefocusofthispaperisthelinkbetweengrandparents’edu-cationandtheirgrandchildren’scognitiveskills.DuringtheECLS-Kparentinterviews,therespondentbiologicalparentwasaskedforhisorhermotherandfather’slevelofeducation.Possibleresponsesincludedallprimaryandsecondaryschoolgrades,highschooldiploma,vocationalschoolwithouthighschooldiploma,voca-tionalschoolwithhighschooldiploma,somecollege,associates
degree,collegedegree,somegraduate/professionalschool,Mas-ter’sdegree,Doctorate,graduate/professionaldegree.Duetothenon-intervalnatureofthemeasure,wecollapsedtheseresponsesintothreecategoriesthatdefinethehighestlevelofeducationattainedbyeitherofthetwograndparents:highschoolorbelow,somecollege,andcollegeormore.Weenteredseparatedummyvariablesforhighschoolorbelow(yes=1,else=0)andcollegeormore(yes=1,else=0)intotheregressionmodels,withsomecollegeormoreservingastheuncodedcomparisongroup.
WealsocontrolledforthreeseparateindicatorsofparentalSES—parentaleducation,income,andoccupationalprestige.Respondentparentsreportedontheirowneducation,aswellasthatofthechild’sotherparent(ifapplicable).Similartograndpar-ents’levelofeducation,weorganizedthehighestlevelofeducationbetweenthetwoparentsintothreegroups:highschoolorbelow,somecollege,andcollegeormore;weincludeddummyvariablesforhighschoolorbelowandcollegeormoreintothemodels,withsomecollegeservingasthecomparisongroup.Tocorrectforthemeasure’spositiveskew,weemployedalog-transformedindicatorofhouseholdincome.Ourmodelsalsoincludedseparatecontinuousmeasuresofmothersandfathers’occupationalpres-tige(z-scored;M=0,SD=1),derivedfromthe1998GeneralSocialSurveyoccupationalprestigescores.
6.2.3.Demographiccharacteristics
Evidencesuggeststhatanumberofotherchildcharacteris-ticsareassociatedwiththeintergenerationaltransferofbothsocioeconomicstatusandchildren’scognitiveabilities.Tocapturerace/ethnicity,weusedaseriesofdummy-codedmeasuresindicat-ingwhetherthechildwasAsian,Hispanic,Black,NativeAmerican,ormultiracial,withWhitesservingastheuncodedcomparisongroup.Familycompositionmeasuresincludedadummy-codedindicatorofsingleparentstatus(yes=1;no=0)andameasureofthenumberofsiblings.Wealsoaccountedfortheprimarylanguagethechildusedathome(non-English=1;English=0);gender(female=1,male=0);whetherthechildrepeatedkinder-garten(yes=1;no=0);child’sage(inmonths,centered);andmother’sageatfirstbirth(inyears,centered).Weconsideredneigh-borhoodurbanicitywithdummy-codedindicatorsoflargecity,mid-sizecity,andsmall-townorrural,withsuburbancommu-nitiesastheuncodedcomparisongroup.Wealsocontrolledforpre-kindergartendaycaretypewithindicatorsforrelativecare,non-relativecare,Headstart,center-basedcare,andvariedcare,withparentalcareservingasthecomparisongroupinthemul-tivariatemodels.Table1providesdescriptiveinformationforallmeasures.
Modelsnotpresentedherealsocontrolledforthenumberoflivinggrandparents,whetherthegrandparentsreportedwerematernalorpaternal,andwhethertheylivedinproximity.Thesemeasures,however,werenotincludedinthefinalmodelsasnoneweresignificant(allp>.05).Theextantliteraturefurthersug-geststhattheeffectsofgrandparents’schoolingongrandchildren’scognitiveandeducationaloutcomesmaydifferbygrandparentandchildgender(Hill&O’Neill,1994;Loury,2006),co-residentversuscaregiverstatus(Aquilino,1996;DeLeire&Kalil,2002;Edwards&Daire,2006;Gordon,1999;Moyietal.,2004),andrace/ethnicity(Dunifon&Kowaleski-Jones,2007).Wetestedthesepotentialassociationsusinginteractionterms,butfoundnonethatweresignificant(p>.05).Onlyoneinteractiontermwassignifi-cantatthep<.10level—college-educatedgrandparentsbyBlackchild.Thesignofthiscoefficientwasnegative,whichindicatesthatBlackchildrenwithgrandparentswithcollegedegreesscoresomewhatlowerthanWhitechildrenwithgrandparentswithonlysomecollege.Thisresultmayreflecthistoricracialdisparitiesinpost-secondaryeducationalqualitybetweenBlackandWhitegrandparents.
220J.L.Ferguson,D.D.Ready/EarlyChildhoodResearchQuarterly26(2011)216–226
Table1
Variablemeansandstandarddeviations(n=13,003).
Mean
SD
Mathachievement20.17.3Readingachievement
22.58.4Grandparents’highestlevelofeducation(%)Highschoolorless67.3–Somecollege
11.9–Bachelor’sormore
20.8–Parents’highestlevelofeducation(%)Highschoolorless36.3–Somecollege
35.8–Bachelor’sormore
27.9–Parentalincome($)
55,072.7
56,062.1
Mother’soccupationalprestige43.711.2Father’soccupationalprestige43.4
11.0
Race/ethnicity(%)White69.5–Black11.6–Hispanic12.4–Asian
2.1–NativeAmerican1.9–Multiracial
2.4–Female(%)
48.9Singleparentfamily(%)
20.1–Non-Englishspeakinghousehold(%)5.4–Kindergartenrepeater(%)3.8–Numberofsiblings1.41.1Child’sage(inmonths)
65.74.3Mother’sageatfirstbirth(inyears)23.95.4Location(%)Largecity13.7–Mid-sizecity19.4–Suburb
44.8–Small-town/rural
22.1–Pre-KCareType(%)Pre-kcarebyparent(s)16.5–Pre-kcarebyrelative
12.9–Pre-kcarebynon-relative11.6–Headstart
7.3–Pre-kcenter-basedcare46.7–Multiple/variouspre-kcare
5.0
–
6.2.4.Designweights
BecauseECLS-Kusedamultistagestratifiedsamplingdesign,thedataincludeaseriesofdesignweights.AswithotherlongitudinalNCESdatasets,analysesusingECLS-Krequiretheuseofweightstocompensateforunequalprobabilitiesofselectionwithinandbetweenschools(e.g.,theintentionaloversamplingofAsianandPacificIslanderchildren)andnon-responseeffects.Ourdescrip-tiveandanalyticanalysesusedthebase-yearchild-levelweight(C1CW0),normalizedtoameanof1toreflecttheactual(smaller)samplesizeforstatisticaltesting.6.3.Analyticapproach
Webeginbydescribingtheassociationsbetweengrandpar-enteducationalattainmentandtheirgrandchildren’scognitiveandsocio-demographicbackgrounds,whichhighlightsthestrongdegreeofintergenerationaltransferofadvantagethatcharacter-izesU.S.society.Totestforstatisticallysignificantgroupmeandifferences,weusedone-wayanalysisofvariance(ANOVA)forcon-tinuousmeasures(withShefféposthoctests)andchi-squaresforcategoricalvariables.Wethenemployedaseriesofmultipleregres-sionstoinvestigatetheassociationsbetweenyoungchildren’scognitiveskillsandtheirgrandparents’educationalattainment.Weranseparateanalysesforboththemathematicsandliteracyoutcomes.Thesemodelswereconstructedinahierarchicalfash-
ioninfoursteps(seeCohen,Cohen,West,&Aiken,2003).Model1establishedtheunadjustedassociationbetweengrandparents’educationandtheirgrandchildren’smeasuredcognitiveabilities.Model2introducedparentaleducationtogaugetheextenttowhichthelinkbetweengrandparenteducationandgrandchildren’scognitiveskillsoperatedthroughparentaleducation.Model3addedadditionalparentalsocioeconomicstatusmeasures,includ-inghouseholdincomeandoccupationalprestige.Inthefourthandfinalmodel,wecontrolledforawidevarietyofcovariatesthatcorrelatedwithbothchildren’sacademicskillsandgrandparents’education.Ourfocusthroughoutwastheadjusted,residualassoci-ationsbetweengrandparents’schoolingandchildren’stestscores.
Ourthirdanalyticapproachentailedpropensityscorematch-ingtochecktherobustnessoftheregressionfindings.Asnotedabove,propensityscorematchingcreatesmatchedpairsoftreat-mentandcontrolcaseswhoareessentiallyidenticalonallobservedcharacteristics,butone—inthisinstancegrandparents’educationalattainment(seeGelman&Hill,2007;Rubin,1997;Rubin&Thomas,1996).Thus,childrenareconsidered“identical”basedonthepre-dictedprobability(propensity)ofreceivingthetreatment—havingacollege-educatedgrandparent—givenobservedcovariates.Sincethemultivariatefindingsforhighschoolversussomecollegewerenon-significantacrossallmodels,wecollapsedthesetwocate-goriesandcreatedadichotomoustreatmentvariable(grandparentswithcollegeormore=1,grandparentswithlessthancollege=0).Asiscustomary,weusedprobitmodelstoestimatethepropensityscores.Tomatchthetreatedandcontrolgroups,weusednearest-neighbormatchingwithreplacementandcommonsupport,whichrestrictedthesampletocasesthatdidnothavepropensityscoresbeloworabovetherangeofpropensityscoresforthecontrolcases(Brand&Halaby,2005).
Oneuniqueaspectofthe“treatment”herewasitsduration.Inthecontextofourcurrentstudy,theinfluenceofgrandparentedu-cationalattainmenttheoreticallystretchedfromthechild’sbirthtokindergartenentry.Akeyidentificationrequirementforpropen-sityscorematchingisthatallcovariatesusedintheprobitmodeltoestimatethepropensityscoresmustbepre-treatment,withtheexceptionoftime-invariantcovariates,suchasrace/ethnicity(Morgan&Harding,2006).Thisrequiredustoemploycharacteris-ticsassociatedwiththechildortherespondingparentbeforeoratthetimeofthechild’sbirth.Accordingly,toestimatethepropensityscores,weusedmother’sageatfirstbirth,lowbirthweight(lessthan5.5pounds),prematurebirth(morethantwoweeksearly),motheronwelfareattimeofbirth,mothermarriedattimeofchild’sbirth,respondent’sparticipationinanacademicprograminhighschool(versuscommercialorvocational),whethertherespondenttooktrigonometry,calculus,and/orphysicsinhighschool,whetherEnglishwasasecondlanguageforthechild,andrace/ethnicityofthechild(time-invariant).7.Results
7.1.1.Descriptiveresults
Table2presentsinformationonchildren’scognitiveandsocio-demographicbackgroundsorganizedbytheirgrandparents’educationalattainment.Thesedescriptiveresultsaddressthefirstofourthreeresearchquestions.Astronglinearrelationshipisevidentbetweenchildren’scognitiveskillsandtheirgrandpar-ents’levelofeducation.Aroughlyone-thirdstandarddeviationgapininitialmathematicsandliteracyskillsseparateschildrenwithcollege-educatedgrandparentsfromthosewhosegrandpar-entshadattendedcollege,butdidnotgraduate.Similarly,childrenwithgrandparentswhoattainedatmostahighschooldegreeper-formedroughlyone-fifthstandarddeviationbelowthosewhosegrandparentshadsomecollegeeducation.Moreover,thegaps
J.L.Ferguson,D.D.Ready/EarlyChildhoodResearchQuarterly26(2011)216–226
221
Table2
Childsocio-demographicandacademiccharacteristicsbygrandparenteducationalattainment(n=13,003).
Grandparents’highestlevelofeducationHighschoolorSomecollegeCollegeormorebelow(n=6955)(n=1775)
(n=2634)Mathematicsachievement,−0.15*0.030.37*MaSD(0.93)(0.97)(1.06)Readingachievement,Ma−0.15*0.020.34*SD
(0.90)(0.99)(1.11)Race(%)White64.1*70.777.0*Black15.1*12.58.9*Hispanic14.8*9.47.3*Asian
2.22.33.5*NativeAmerican1.61.71.0*Multiracial
2.2*
3.32.4*Parents’highestlevelofeducation(%)Highschoolorbelow40.8*18.79.4*Somecollege36.5*42.626.3*Collegeormore2.7*38.864.2Parentalincome($),M48,037*57,48280,138*SD
(44,847)(50,832)(78,329)Mother’soccupational−0.12*0.160.44*prestige,MaSD
(0.92)(1.03)(1.13)Father’soccupational−0.14*0.110.40*prestige,MaSD
(0.91)(1.03)(1.16)Singleparentfamily(%)23.122.114.7*Non-Englishspeaking6.3*2.54.4*household(%)
Kindergartenrepeater(%)4.5*3.03.0Numberofsiblings,M1.46*1.341.39SD
(1.10)(1.02)(1.00)Child’sage(inmonths),M65.6865.7465.69SD
(4.26)(4.18)(4.25)Mother’sageatfirstbirth23.26*24.2526.31*(inyears),MSD
(5.27)(5.34)(5.21)Location(%)Large-city13.013.113.9Mid-city
18.5*20.620.8Large/midsuburb39.442.047.6Small-townorrural29.1*24.317.8*Pre-KCareType(%)Pre-kcarebyparent(s)18.8*14.412.1*Pre-kcarebyrelative
14.8*12.88.7*Pre-kcarebynon-relative10.8*12.513.0Headstart
10.3*7.03.1*Pre-kcenter-basedcare40.1*48.458.2*Multiple/variouspre-kcare
5.2
4.9
4.9
*p<.05.“Somecollege”isthecomparisongroupforsignificancetesting.a
Measureisz-scored(M=0;SD=1).
betweenchildrenwithonlyhigh-school-educatedgrandparentsandthosewithcollege-educatedgrandparentswereconsiderablylarger(roughlyone-halfstandarddeviation).
TheremainderofTable2providesclearindicationthatmulti-generationalhumancapitaltendstobeconcentratedacrossgen-erationswithinparticularsocio-demographicsubgroups.Fewerthan9%ofchildrenwithatleastonecollege-educatedgrandparentwereBlack,andfewerthan8%wereHispanic.Atthesametime,over15%ofchildrenwithgrandparentswhodidnotgobeyondhighschoolwereBlackandnearly15%wereHispanic.Conversely,whileWhitechildrenconstituted69.5%ofthesample,theyrep-resented77.0%ofchildrenwithcollege-educatedgrandparents.In
short,almostone-thirdofchildrenwiththeleast-educatedgrand-parentswereBlackorHispanic,whereaslessthan17%ofchildrenwiththemost-educatedgrandparentssharedtheseracial/ethnicbackgrounds.Similarpatternsemergedwithparentaleducation,highlightingtherealitythatdisparitiesinhumancapitalarefirstpassedfromgrandparentstoparents.Fewerthan10%ofchildrenwithcollege-educatedgrandparentshadparentswhofailedtogobeyondhighschool,whileover40%ofchildrenwithgrandparentswhodidnotmakeitbeyondhighschoolhadsimilarlyeducatedpar-ents.Morestrikingstill,lessthanone-quarterofchildrenwhosegrandparentsneverwenttocollegehadparentswithacollegedegree.Thispreliminaryevidenceofthenatureandscopeoftheintergenerationaltransmissionofeducationalattainmentsuggeststhecentralrole—eitherdirectlyorindirectly—grandparentsplayintheirgrandchildren’scognitiveskills.
Reflectingadditionalinequalitiesassociatedwithgrandpar-ents’educationandtheintergenerationaltransferofeconomicresources,wefoundstarkdifferencesinparentalincomebygrand-parents’education.Theaverageannualincomeofafamilywithgrandparentswhoseeducationdidnotextendbeyondhighschoolwaslessthan$50,000,comparedtoover$80,000forafamilywithcollege-educatedgrandparents.Similardifferenceswereevidentintermsofparentaloccupationalprestige,withoverhalfastan-darddeviationgapinparentaloccupationalprestigeseparatingthelowestandhighestlevelsofgrandparenteducation.
Othersocio-demographiccharacteristics,suchassingle-parenthouseholds,Englishlanguageuse,andnumberofsiblingswerealsostronglyassociatedwithgrandparents’education.Fewerthan15%ofchildrenwithcollege-educatedgrandparentswereraisedinsingle-parenthouseholds,whereasover23%ofchildrenwhosegrandparentswerenoteducatedbeyondhighschoolwereraisedinsimilarfamilies.Theaveragenumberofsiblingsdecreasedslightlyasgrandparents’educationalattainmentincreased,whilemoth-erswhoseparentshadattainedahighschooldegreeorlesswereonaveragethreeyearsyoungeratthebirthoftheirfirstchildcomparedtothosewithcollege-educatedparents.Anoticeabledifferencealsoexistedbetweenchildrenoftheleast-andmost-educatedgrandparentsinkindergartenrepetitionrates.Roughly4.5%ofchildrenwiththeleast-educatedgrandparentsrepeatedkindergarten,comparedtoonly3%ofthosewithcollege-educatedgrandparents.
Severalimportantoverarchingpatternsemergefromthesedescriptiveresults.Childrenwithcollege-educatedgrandparentsbeganformalschoolingwithconsiderablystrongercognitiveskills.Importantly,BlackandHispanicchildrenarefarlesslikelythanWhitestohavehadcollege-educatedgrandparents.Manyofthesesystematicracial/ethnicdifferencesingrandparents’educationalprofilespointtohistoricaldisparitiesinaccesstohigheredu-cation.Otherdifferencesingrandparents’educationreflecttheconcentrationofhumancapitalwithinfamiliesovermanygener-ations:over60%ofchildrenwithcollege-educatedgrandparentsalsohadcollege-educatedparents.Conversely,lessthanaquar-terofchildrenwiththeleast-educatedgrandparentshadparentswhosuccessfullymadeitthroughcollege.Better-educatedgrand-parentsalsohadgrandchildrenwhosefamilieswerebetterofffinancially.Themeanhouseholdincomeofchildrenwithcollege-educatedgrandparentswasmorethanoneandahalftimestheincomeofchildrenwiththeleast-educatedgrandparents.Thesesystematicdifferencessuggestthatathree-generationapproachprovidesamorenuancedunderstandingofinequalityinearlychildhood.
8.Analyticresults
Theprevioussectionpresenteddescriptiveanalysesexploringtheassociationsbetweengrandparents’educationandtheirgrand-
222J.L.Ferguson,D.D.Ready/EarlyChildhoodResearchQuarterly26(2011)216–226
Table3
Grandparenteducationalattainmentandgrandchildren’smathskills(n=13,003).
Model1
Model2Model3Model4Grandparents’educationaHighschoolorbelow0.000.050.040.01Collegeormore0.51*0.24*0.20*0.15*Parents’educationHighschoolorbelow−0.33*
−0.21*
−0.14*
Collegeormore
0.50*
0.28*0.18*AdditionalparentalSESmeasuresLogincome
0.18*0.09*Mother’soccupationalprestige0.07*0.06*Father’soccupationalprestige0.06*
0.06*DemographiccontrolsBlackb−0.22*Hispanic−0.19*Asian
0.14*NativeAmerican−0.29*Multiracial
−0.13Female
0.01Singleparent
−0.08*Non-Englishspeakinghousehold−0.13*Kindergartenrepeater0.57*NumberofSiblings−0.03*Child’sagec
0.06*Mother’sageatfirstbirthc
0.02*Mother’sageatfirstbirth,squaredc0.00Largecityd0.02Mid-sizecity
−0.03Small-townorrural−0.13*Pre-kcarebyrelativee
−0.01Pre-kcarebynon-relative0.06Headstart
−0.03Pre-kcenter-basedcare0.17*Multiple/variouspre-kcare0.08
Intercept−0.15*−0.17*−1.99*−0.94*R20.05
0.15*0.18*0.28*R2
0.10*
0.03*
0.10*
*p<.05.
a
Grandparentandparenteducationcomparedtosomecollege.b
Allracial/ethnicgroupscomparedtoWhites.c
Measureiscenteredaboutitsmean.d
Large-city,mid-city,andsmall-town/ruralarecomparedtolargeandmid-sizesuburbs.e
Pre-kcaremeasurescomparedtoparentalcare.children’scognitiveprofilesandsocio-demographicbackgrounds.Theaimhereistoinvestigatethesecomplexandinterrelatedassociationswithinamultivariateframeworkthatidentifiestheuniquelinksbetweengrandparenteducationalattainmentandtheirgrandchildren’scognitiveoutcomes.Tables3and4presentmultivariateregressionmodelspredictingyoungchildren’smath-ematicsandreadingskills.Weemployedahierarchicalregressionapproachtohighlightthechangesinthegrandparents’educationestimatesassubsequentmodelsintroducedadditionaldemo-graphiccontrols.8.1.Mathematicskills
Model1inTable3establishestheunadjustedrelationshipbetweengrandparents’educationandchildren’smathematicsskills.Asourdescriptiveresultssuggested,thechildrenofcollege-educatedgrandparentstypicallybegankindergartenwithstrongermathematicsskillsthanchildrenwhosegrandparentshadobtainedonlysomehighereducation(ES[effectsize]=0.51).Incontrast,childrenwithonlyhigh-schooleducatedgrandparentsscorednodifferentfromthosewhosegrandparentshadonlysomehighereducation(p>.05).
Table4
Grandparenteducationalattainmentandgrandchildren’sliteracyskills(n=13,003).
Model1
Model2Model3Model4Grandparents’educationaHighschoolorbelow−0.010.030.020.01Collegeormore0.47*
0.22*0.18*0.14*Parents’educationHighschoolorbelow−0.29*−0.18*−0.11*Collegeormore
0.48*
0.27*0.17*AdditionalparentalSESmeasuresLogincome
0.14*0.07*Mother’soccupationalprestige0.07*0.06*Father’soccupationalprestige0.07*
0.07*DemographiccontrolsBlackb0.00Hispanic−0.10*Asian
0.28*NativeAmerican−0.21*Multiracial
−0.03Female
0.16*Singleparent
−0.10*Non-Englishspeakinghousehold−0.16*Kindergartenrepeater0.56*Numberofsiblings−0.08*Child’sagec
0.04*Mother’sageatfirstbirthc
0.02*Mother’sageatfirstbirth,squaredc0.00Largecityd0.01Mid-sizecity
−0.05Small-townorrural−0.15*Pre-kcarebyrelativee
−0.02Pre-kcarebynon-relative0.00Headstart
−0.06Pre-kcenter-basedcare0.17*Multiple/variouspre-kcare0.04
Intercept−0.13*0.17*−1.65*−0.78*R20.04*
0.13*0.16*0.24*R2
0.09*
0.03*
0.08*
*p<.05.
a
Grandparentandparenteducationcomparedtosomecollege.b
Allracial/ethnicgroupscomparedtoWhites.c
Measureiscenteredaboutitsmean.d
Large-city,mid-city,andsmall-town/ruralarecomparedtolargeandmid-sizesuburbs.e
Pre-kcaremeasurescomparedtoparentalcare.
InModel2,weaccountedforparentaleducationandfoundthattheestimateassociatedwithcollege-educatedgrandparentswasreducedbyroughly50%(i.e.,from0.51to0.24SD).Asonemightexpect,parentaleducationaccountedforagreatdealoftheassociationbetweengrandparents’educationandgrandchil-dren’scognitiveskills.Onaverage,childrenwithcollege-educatedparentsoutperformedthosewhoseparentsattainedonlysomecollegebyone-halfstandarddeviation,anestimatevirtuallyiden-ticaltothecollege-educatedgrandparentcoefficientinModel1.Alsoimportantistheoverthree-quarterSDgapbetweenthechil-drenofparentswithahighschooldegreeorlessandthosewithcollege-educatedparents(ES=−0.33vs.0.50).Despitethesecon-trolsforparentaleducation,grandparentpossessionofaBAormoreremainedpositivelyassociatedwithchildren’smathematicsskills(ES=0.24).
AsindicatedinModel3,parentalincomewasalsopositivelyrelatedtochildren’smeasuredacademicabilities,asweremotherandfather’soccupationalprestige(ES=0.07and0.06,respectively).Theseassociationsaccountedforsomeofthelinkbetweenpar-ents’educationandtheirchildren’scognitiveskills.Theeffectsofparentalcollegeattendancewerereducedbyroughly40%,andthenegativeeffectsassociatedwithattainingonlyhighschoolor
J.L.Ferguson,D.D.Ready/EarlyChildhoodResearchQuarterly26(2011)216–226223
belowwerereducedby36%.Centraltothefocusofthisstudy,theeffectofcollegeattainmentforgrandparentsdroppedbyonly22%,andremainedasignificantpredictorofchildren’sacademicskills(ES=0.20).
Model4controlledforadditionalsocio-demographiccharacter-istics,includingrace/ethnicity,gender,familycomposition,child’sage,mother’sageatfirstbirth,urbanicity,andpre-kindergartencare.Thecoefficientassociatedwithcollege-educatedparentswasreducedfurtherwhenweincludedtheseadditionalcovariatesinthemodel.However,evenafteraccountingforthesecovariates,wefoundthatchildrenwithcollege-educatedgrandparentsbegankindergartenwithsomewhatstrongermathematicsskillsthanchildrenwhosegrandparentshadonlysomecollege(ES=0.15).Becausethesedataarelimitedtograndparents’education,theapparentrelationshipwithgrandchildren’searlycognitiveabilitymaybeconflatedwithothergrandparentsocioeconomiccharac-teristics,particularly,income,wealth,andoccupationalprestige.ButregardlessofwhichaspectsofSESthegrandparenteducationcoefficientisreflecting,thefactthattheeducationofchildren’sgrandparentsremainedassociatedwiththeircognitiveskillssug-geststheinfluencemultigenerationalsocioeconomicstatushasoneventheyoungestchildren.8.2.Readingskills
Aswithchildren’smathematicsskills,grandparents’educa-tionalattainmentwasassociatedwithyoungchildren’sliteracyskills.Model1inTable4indicatesthatthechildrenofcollege-educatedgrandparentsscoredconsiderablyabovetheirpeerswhohadgrandparentswithonlysomecollegeeducation(ES=0.47).Theredoesnotappear,however,tohavebeenasignificantben-efitofhavingagrandparentwithsomecollegeeducationversusonewhodidnotprogressbeyondhighschool(p>.05).Afterincor-poratingparents’educationalattainmentinModel2,theeffectofhavingcollege-educatedgrandparentswasreducedbyoverhalf(from0.46to0.22SD).Similartoourfindingswithmathematics,theaveragechildwithatleastonecollege-educatedparentpossessedliteracyskillsnearlyone-halfSDgreaterthanonewhoseparentsattendedbutdidnotgraduatefromcollege(ES=0.48).Despitetheinclusionofparents’education,havingacollege-educatedgrand-parentremainedsignificantlyrelatedtochildren’sliteracyskills(ES=0.22).
Model3addedcontrolsforparentalincomeandoccupationalprestige,andreducedthecollege-educatedparenteffectbyover40%(from0.48to0.27SD).Childrenwithonlyhigh-schooledu-catedparentsscoredalmostone-fifthstandarddeviationbelowthosewithparentswithsomehighereducation.Notsurprisingly,householdincomewaspositivelyassociatedwithliteracyskills,aswerebothmotherandfather’soccupationalprestige(ES=0.07and0.06,respectively).Moreover,evenafteradjustingfortheserel-evantsocio-demographiccharacteristics,havingcollege-educatedgrandparentsremainedpositivelyassociatedwithchildren’scog-nitiveskills(ES=0.18).Thefinalmodelaccountedforabroadrangeofadditionaldemographiccharacteristics.Centraltoourfocushereisthattheeffectassociatedwithcollege-educatedgrandparentsremainedsignificant(althoughsmall).
9.Regression-adjustedpropensitymatchedestimates
Webeganthisstageoftheanalysesbyconstructingasam-plecontainingcomparabletreatmentandcontrolgroups.Table5,whichdisplaystheresultsfortheprobitestimation,indicatesthatparents’highschooleducationalbackgroundandwelfarereceiptwerestrongpredictorsofgrandparenteducationalattainment.Thenextstepinthepropensityanalysisrequiredmatchingthetreated
Table5
Probitmodelestimationsofhavingoneormorecollege-educatedgrandparents.
Coefficients
Mother’sageatfirstbirth
0.00Parentenrolledinacademicprograminhighschoola0.41*Parenttooktrigonometryinhighschool0.14*Parenttookcalculusinhighschool0.10Parenttookphysicsinhighschool−0.10*Childisnon-nativeEnglishspeaker−0.06ChildisBlackb−0.02ChildisHispanic−0.15ChildisAsian
0.00ChildisNativeAmerican0.02Childismultiracial
0.01Childhadlowbirthweight−0.03Childwaspremature
−0.01Motheronwelfareatbirth−0.35*Mothermarriedatbirth−0.05Constant
−1.15*
*p<.05.
aAcademicprogramiscomparedtocommercialorvocationalprogram.b
Allracial/ethnicgroupsarecomparedtoWhites.
andcontrolgroups.Eachindividualchildinthetreatmentgroupwaspairedwithachildinthecontrolgroupwhohadthesameprobabilityofhavingagrandparentwithacollegedegreeormore.Recallthatthegoalwastohavetreatmentandcontrolgroupsthatdidnotdiffersystematicallyonobservedcovariatesorcontrols.Analysesnotpresentedhereindicatedgoodcovariatebalance,asthemeandifferencesbetweenthetreatmentandcontrolgroupswereneitherlargenorstatisticallysignificant.
Wethenestimatedtheassociationsbetweengrandparents’educationandtheirgrandchildren’senteringkindergartenmathe-maticsandreadingscoresusingtheregression-adjustedpropensitymatchedestimates.Theseanalysesincludedallcovariatesthatwereusedintheprobitmodeltoestimatethepropensityscores.Thisvariantofpropensityscorematching(regression-adjusted)allowedustousethepropensityscoreweightsandrestrictthesampletothosecaseswithsimilarpropensityscores(i.e.,tothosecaseswithoverlap).Theseestimatesareconsideredtobeunbiasedprovidedthat,basedontherestrictionofthesampleusingcommonsupportandmatching,nosignificantcorrelationsexistbetweentheindependentvariablesandtheerrorterm.Uponsatisfyingthiscon-dition,thegrandparents’educationcoefficientcanbeseenastheaveragetreatmenteffect.
Table6
Grandparenteducationalattainmentandgrandchildren’sliteracyskills:propensitymatchedestimates.
Mathematics
ReadingGrandparent(s)–Bachelor’sormore0.17*0.19*Mother’sageatfirstbirth
0.03*0.02*Parentenrolledinacademicprogramin0.03−0.02highschoola
Parenttooktrigonometryinhighschool−0.080.04Parenttookcalculusinhighschool0.01−0.07Parenttookphysicsinhighschool0.120.18*Childisnon-nativeEnglishspeaker−0.38*−0.35*ChildisBlackb−0.070.13ChildisHispanic−0.120.01ChildisAsian
0.160.42*ChildisNativeAmerican−0.22−0.10Childismultiracial
−0.22−0.19Childhadlowbirthweight0.24*0.12Childwaspremature
0.060.10Motheronwelfareatbirth−0.21*−0.28*Mothermarriedatbirth0.070.10Constant
−0.95*
−0.76*
*p<.05.
aAcademicprogramiscomparedtocommercialorvocationalprogram.b
Allracial/ethnicgroupsarecomparedtoWhites.
224J.L.Ferguson,D.D.Ready/EarlyChildhoodResearchQuarterly26(2011)216–226
Table6presentstheresultsofthepropensityscoreapproach.Theseestimatessuggestthathavingatleastonecollege-educatedgrandparentwasassociatedwithanadvantageinmathematicsskillsof0.17SDandasimilarpositivelinkbetweengrandparenteducationalattainmentandtheirgrandchildren’sliteracyabilities.Onaverage,childrenwithcollege-educatedgrandparentsscored0.19SDabovethosewithnon-college-educatedgrandparents.Thesegrandparenteducationestimatesformathematicsandread-inginTable6arequitesimilartotheregressionfindingspresentedinTables3and4.Thisanswersourthirdresearchquestion.Thecoefficientforcollege-educatedgrandparentsinTable6isactuallysomewhatlargerthanthatindicatedbytheregressioninTable3(0.17vs.0.15SD).Similarlyforliteracy,thematchedcaseestimatesfromTable6areslightlystrongerthantheestimatesfromTable4(0.19vs.0.14SD).ThesefindingssuggestthattheregressionresultspresentedinTables3and4oftheassociationbetweengrandpar-ents’educationandgrandchildren’scognitiveskillsarequiterobustandmayevenrepresentsomewhatconservativeestimates.
10.Discussion
Theintergenerationalcrystallizationofsocioeconomicadvan-tagehasgeneratedvigorousdebateamongsocialscientists,policymakers,andthepublic.Ourcurrentstudyconfirmsthathistoricaldisadvantagesineducationalattainmentrepresentanadditionalsourceofvariabilityinchildoutcomes.Acrossmulti-plecognitiveandsocio-demographicdimensions,childrenwithmore-educatedgrandparentsdiffersystematicallyfromthosewithless-educatedgrandparents.Indeed,ourstudysuggestspersistent,multigenerationalinfluencesofsocioeconomicadvan-tage.Evenafteraccountingforawidearrayofsocioeconomicanddemographiccharacteristics,childrenwithcollege-educatedgrandparentstendtobeginformalschoolingwithstrongerliter-acyandmathematicsskills.Althoughtheeffectswereportedherearenotparticularlylarge,evenmodestacademicdisparitiesatsuchanearlyagecanhavesubstantialimplicationsforchildren’sacademicplacementsandexperiencesinschool.Forexample,deci-sionsregardinggraderetention,abilitygrouping,curricula,andassignmenttospecialeducationandESLprogramscanoftencomedowntorelativelysmallacademicdifferencesbetweenchildren(Burkam,LoGerfo,Ready,&Lee,2007;Entwisleetal.,1997;Smith&Shepard,1988).
Ourfirstresearchquestioninquiredintothelinksbetweengrandparents’educationandtheirgrandchildren’ssocio-demographiccharacteristics.Theresultssuggestthatcollege-educatedgrandparentstendtohavegrandchildrenwhoaresimilarlyadvantaged,bothsociallyandacademically.Thegrandchildrenofcollege-educatedgrandparentsbeginkinder-gartenwithmoremathematicsandreadingskills,tendtohavemotherswhowereconsiderablyolderatthebirthoftheirfirstchild,andaremorelikelytocomefromtwo-parent,higher-SEShouseholds.Oursecondresearchquestionconcernedtheresidualassociationsbetweengrandparents’educationandtheirgrandchil-dren’sliteracyandmathematicsskills.Evenaftercontrollingforgrandchildren’ssocioeconomicstatus,theassociationsbetweengrandparents’educationandgrandchildren’scognitiveskillsremain.Thethirdresearchfocustestedtherobustnessoftheseresultsusingregression-adjustedpropensitymatchedestimates.Importantly,theOLSfindingswererobusttoalternatemodelspecificationsandapproaches.Thepropensityscoreanalysessupportedourfindingsregardingthelinksbetweengrandparents’educationandtheirgrandchildren’sacademicabilities.Aspreviousauthorshavenoted,theidentificationofsocialreproductionacrossthreegenerationssuggestsquitedifferentpolicyresponses,asevendramaticreductionsineducationalinequalitywouldtake
morethanonegenerationtoreducehistoriccognitiveinequalities(seePhillipsetal.,1998).
Acrossseveraldimensions,thisstudyaddstothegrowingbodyofresearchontheintergenerationaltransferofsocial,economic,andcognitiveadvantage.Unlikemuchoftheexistingliterature,weexploredintergenerationaleffectsattheverystartofchildren’sformalschooling.Althoughsomestudieshaveaddressedsimilarquestionswithschool-entry-agechildren(e.g.,Hill&O’Neill,1994),othershaveemployedsamplesofchildrenseveralyearsintoele-mentaryschool(e.g.,Grant,2005)orchildrenfrommultipleagespanssimultaneously(e.g.,Yeung&Conley,2008,whosesampleincludeschildrenrangingfromage3to12).Limitingouranalysestochildrenatthestartofkindergartenpermittedtheestimationofgrandparenteducationalattainmenteffectsabsenttheeffectsofformalschooling.Thisisimportant,asschoolqualityisassociatedwithchildren’ssocialclassandinitialacademicskills,potentiallyproducingestimatesthatconflateschool,parental,andgrandpar-entinfluences(Lee&Burkam,2002).Moreover,incontrasttoourcurrentstudy,priorresearchrarelyemployedadvancedcausalinferencetechniquesthataddressissuesofendogeneity.Previouswork,includingthestudiesnotedimmediatelyabove,alsotypicallyusedrelativelysmallsamples(generallyfewerthan1800children).Incontrast,weusedalargeandnationallyrepresentativesampleofover13,000children.
Althoughthisstudyholdsseveraladvantagesovertheextantresearch,italsohasseverallimitations.Duetoconstraintsinher-entinourdata,wecouldnotexaminetheextenttowhichmaternalversuspaternalgrandparentswereassociatedwithearlycogni-tiveoutcomeswithinthesamefamily.Moreover,wewereunabletoidentifytheprecisemechanismsthatfacilitatetheintergenera-tionaltransferofcognitiveability.Futureresearchmightexplorewhetherthecognitivebenefitsofhavingmoreeducatedgrand-parentsreflectgenetictransferversusthehistoricaccumulationofsocial,cultural,andeconomiccapital.Additionally,theassoci-ationswereportedheremaybecontingentuponthenatureoftherelationshipsbetweengrandparentsandgrandchildren.Forexample,itispossiblethatcollege-educatedgrandparentsmorefrequentlyinteractwiththeirgrandchildreninwaysthatpositivelyaffecttheirgrandchildren’sliteracyandmathdevelopment.Otherpotentialcausalmechanismsmaysimplyinvolvegreaterfiscalresourcesamongmore-educatedgrandparents,andanincreasedabilitytoassisttheirchildrenandgrandchildren.Becausemea-suresofsocialclassarelimitedtograndparents’educationwithintheECLS-Kdata,theapparentrelationshipwithgrandchildren’searlycognitiveabilitymaybeconflatedwithothergrandparentsocioeconomiccharacteristics,particularly,income,wealth,andoccupationalprestige.Inshort,theactualmethodoftransferislikelybestcharacterizedasacomplicatednexusinvolvingmultipleinfluencesthatextendacrossgenerations.
Ourresultsalsosuggesttheneedforfutureresearchthatattendstotherelationshipbetweengrandparents’educationandgrand-children’sliteracyandmathskillsasgrandchildrenage.Giventhatchildrenwithcollege-educatedgrandparentsenterkinder-gartenwithmoreadvancedliteracyandmathskillsthantheirpeers,wewonderwhethergrandparents’educationinfluencesgrandchildren’sacademicgrowthonceformalschoolinghasbegun.Dograndchildrenwithcollege-educatedgrandparentslearnmorequicklythantheirpeerswhosegrandparentsdidnotearnpost-secondarydegrees?Ifso,intergenerationaltransferswillproducegreatersocialinequalityasgrandchildrenage.Incontrast,thecognitivedisparitiesindependentlyattributedtograndparents’educationalattainmentmaydisappearasgrandchildrengrowolder.Quitepossibly,norelationshipmayexistbetweengrandpar-ents’educationandgrandchildren’slearningrates.Moreover,theassociationbetweengrandparents’educationandgrandchildren’scognitiveabilitymaydependuponthenatureofthepre-school
J.L.Ferguson,D.D.Ready/EarlyChildhoodResearchQuarterly26(2011)216–226225
relationshipbetweengrandparentsandgrandchildrenandthere-foreweaken—ordisappear—asgrandchildrenageandbecomemoreindependent.
Thevariabilityingrandparents’educationamongdifferentracial/ethnicandsocioeconomicgroupsmayhaveitsbeginningsininequitiesthatoccurredlongago.Moresignificantly,currentinequitiesamongchildrencouldberelatedtopastinequitiesexpe-riencedbytheirgrandparents.Ifwearetofullyunderstandthefactorsthatareassociatedwithandpredictiveofchildren’slifeoutcomes,researchersmustworktocapturemultigenerationalmeasuresofcapital—human,financial,social,andcultural.Todate,moststudiesthatfocusonchildren’sachievementonlyaccountforparentalsocioeconomicstatus.Ourcurrentstudymakesclearthattwogenerations—parentsandchildren—arenotsufficient.Whenavailable,futureresearchshouldstrivetoincludethreegenerationsintheconstructionofmeasuresofSES.Thefindingsreportedhere,combinedwiththehistoricalevidenceregardinggrandparents’unequalaccesstohighereducation,suggestthecrucialintergen-erationalroleofhumancapitalinexacerbatingsocialinequalities.
References
Aquilino,W.(1996).Thelifecourseofchildrenborntounmarriedmothers:Child-hoodlivingarrangementsandyoungadultoutcomes.JournalofMarriageandtheFamily,58,293–310.
Barranti,C.(1985).Thegrandparent/grandchildrelationship:Familyresourceinan
eraofvoluntarybonds.FamilyRelations,34,343–352.
Becker,G.(1991).Thetreatiseonthefamily.Cambridge,MA:HarvardUniversity
Press.
Becker,G.,&Tomes,N.(1986).Humancapitalandtheriseandfallofthefamily.
JournalofLaborEconomics,4(3),576–614.
Blau,P.,&Duncan,O.(1967).TheAmericanoccupationalstructure.NewYork,NY:
Wiley.
Bourdieu,P.(1977).Outlineofatheoryofpractice.Cambridge,UK:CambridgeUni-versityPress.
Bourdieu,P.(1984).Distinction:Asocialcritiqueofthejudgmentoftaste(R.Nice,
Trans.).Cambridge,MA:HarvardUniversityPress(Originalworkpublished1979).
Bourdieu,P.(1985).Thesocialspaceandthegenesisofgroups.TheoryandSociety,
14(6),723–744.
Bowers,B.,&Myers,B.(1999).Grandmothersprovidingcareforgrandchildren:
Consequencesofvariouslevelsofcaregiving.FamilyRelations,48(3),303–311.Bowles,S.,&Gintis,H.(1976).SchoolingincapitalistAmerica:Educationalreformand
thecontradictionsofAmericanlife.NewYork,NY:BasicBooks.
Bowles,S.,&Gintis,H.(2002).SchoolingincapitalistAmericarevisited.Sociologyof
Education,75(2),1–18.
Brand,J.,&Halaby,C.(2005).Regressionandmatchingestimatesofelitecollege
attendanceoneducationadcareerachievement.SocialScienceResearch,35,749–770.
Bronfenbrenner,U.(1979).Theecologyofhumandevelopment.Cambridge,MA:Har-vardUniversityPress.
Bronfenbrenner,U.,&Morris,P.A.(1998).Theecologyofdevelopmentalprocesses.
InW.Damon(Ed.),Handbookofchildpsychology.SixtheditionVol.1Theoreticalmodelsofhumandevelopment(pp.793–828).NewYork,NY:JohnWiley&Sons,Inc.
Brooks-Gunn,J.(1995).Childreninfamiliesincommunities:Riskandintervention
intheBronfenbrennertradition.InP.Moen,G.H.Elder,&K.Luscher(Eds.),Examininglivesincontext:Perspectivesontheecologyofhumandevelopment(pp.467–519).Washington,DC:AmericanPsychologicalAssociation.
Brooks-Gunn,J.,&Duncan,G.(1997).Theeffectsofpovertyonchildren.TheFuture
ofChildren:ChildrenandPoverty,7(2),55–71.
Bryson,K.,&Casper,L.M.(1999).Coresidentgrandparentsandgrandchildren(pp.
23–198).Washington,DC:U.S.CensusBureau.
Burkam,D.T.,LoGerfo,L.F.,Ready,D.D.,&Lee,V.E.(2007).Thedifferentialeffects
ofrepeatingkindergarten.JournalofEducationforStudentsPlacedatRisk,12(2),103–136.
Cohen,J.(1988).Statisticalpoweranalysisforthebehavioralsciences(Rev.ed.).Hills-dale,NJ:Erlbaum.
Cohen,J.,Cohen,P.,West,S.,&Aiken,L.(2003).Appliedmultipleregression/correlation
analysisforthebehavioralsciences(Rev.ed.).Hillsdale,NJ:Erlbaum.
Corcoran,M.(2001).Mobilility,persistence,andtheconsequencesofpovertyfor
children:Child,adultoutcomes.InS.H.,Danzinger&R.H.,Haveman,(Eds.),Understandingpoverty(pp.127–161).NewYork,NY:RussellSageFoundation.DeLeire,T.,&Kalil,A.(2002).Goodthingscomeinthrees:Single-parentmulti-generationalfamilystructureandadolescentadjustment.Demography,39(2),393–413.
Denham,T.,&Smith,C.(1989).Theinfluenceofgrandparentsongrandchildren:A
reviewoftheliteratureandresources.FamilyRelations,38(3),345–350.
Dunifon,R.,&Kowaleski-Jones,L.(2007).Theinfluenceofgrandparentsinsingle-motherfamilies.JournalofMarriageandFamily,69,465–481.
Durlauf,S.N.(1996).Atheoryofpersistentincomeinequality.JournalofEconomic
Growth,1(1),75–93.
Eccles,J.S.(2005).Thepresentandfutureofresearchonactivitysettingsasdevel-opmentalcontexts.InJ.L.Mahoney,R.W.Larson,&J.S.Eccles(Eds.),Organizedactivitiesascontextsofdevelopment:Extracurricularactivities,after-schoolandcommunityprograms(pp.353–371).Mahwah,NJ:Erlbaum.
Edwards,O.,&Daire,A.(2006).Theory-basedinterventionsforschoolchildrencared
forbytheirgrandparents.EducationalPsychologyinPractice,23(2),177–190.Ellwood,D.,&Kane,T.(2000).Whoisgettingacollegeeducation?Familyback-groundandthegrowinggapsinenrollment.InS.Danziger,&J.Waldfogel(Eds.),Securingthefuture(pp.283–324).NewYork,NY:RussellSageFoundation.
Entwisle,D.R.,Alexander,K.L.,&Olson,L.S.(1997).Children,schools,andinequality.
Boulder,CO:WestviewPress.
Erola,J.,&Moisio,P.(2007).SocialmobilityoverthreegenerationsinFinland,1950-2000.EuropeanSociologicalReview,23(2),169–183.
Ganzeboom,H.,Treiman,D.,&Ultee,W.(1991).Comparativeintergenerational
stratificationresearch:Threegenerationsandbeyond.AnnualReviewofSoci-ology,17,277–302.
Gelman,A.,&Hill,J.(2007).Dataanalysisusingregressionandmultilevel/hierarchical
models.NewYork,NY:CambridgeUniversityPress.
Gordon,R.(1999).Multigenerationalcoresidenceandwelfarepolicy.JournalofCom-munityPsychology,27(5),525–549.
Grant,E.(2005).Cumulativeadvantage:Alookattheeffectoffamilysocialclassonchil-dren’seducationalachievementandtheblack-whiteachievementgap.RetrievedfromProQuestDigitalDissertations(AAT3186341).
Hart,B.,&Risley,T.(1992).Americanparentingoflanguage-learningchildren:
Persistingdifferencesinfamily-childinteractionsobservedinnaturalhomeenvironments.DevelopmentalPsychology,28(6),1096–1105.
Heath,S.(1983).Wayswithwords:Language,life,andworkincommunitiesandclass-rooms.NewYork,NY:CambridgeUniversityPress.
Hertz,T.(2005).Rags,riches,andrace:Theintergenerationaleconomicmobilityof
blackandwhitefamiliesintheUnitedStates.InS.Bowles,H.Gintis,&M.Groves(Eds.),Unequalchances:Familybackgroundandeconomicsuccess(pp.165–191).NewYork,NY:RussellSageFoundation.
Hill,M.,&O’Neill,J.(1994).Familyendowmentsandtheachievementofyoung
childrenwithspecialreferencetotheunderclass.TheJournalofHumanResources,29(4),1064–1100.
Hollingshead,A.(1975).Elmtown’syouthandElmtownrevisted.NewYork,NY:Wiley.Jencks,C.,Smith,M.,Acland,H.,Bane,M.J.,Cohen,D.,Gintis,H.,etal.(1972).Inequal-ity:AreassessmentoftheeffectoffamilyandschoolinginAmerica.NewYork,NY:Harper.
Jendrek,M.(1993).Grandparentswhoparenttheirgrandchildren:Effectson
lifestyle.JournalofMarriageandtheFamily,55,609–621.
King,V.,&Elder,G.(1997).Thelegacyofgrandparenting:Childhoodexperiences
withgrandparentsandcurrentinvolvementwithgrandchildren.JournalofMar-riageandFamily,59(4),848–859.
King,V.,&Elder,G.(1998).Educationandgrandparentingroles.ResearchonAging,
20(4),450–474.
Kropf,N.,&Burnette,D.(2003).Grandparentsasfamilycaregivers:Lessonsfor
intergenerationaleducation.EducationalGerontology,29,361–371.
Lareau,A.(2003).Unequalchildhoods:Class,race,andfamilylife.Berkeley,CA:Uni-versityofCaliforniaPress.
Lee,V.E.,&Burkam,D.(2002).Inequalityatthestartinggate:Socialbackgrounddiffer-encesinachievementaschildrenbeginschool.Washington,DC:EconomicPolicyInstitute.
Loury,L.(2006).Allintheextendedfamily:Effectsofgrandparents,aunts,anduncles
oneducationalattainment.AmericanEconomicReview,96(2),275–278.
Lynd,R.,&Lynd,H.(1929).Middletown:AstudyinAmericanculture.NewYork,NY:
HarcourtBrace.
Mayer,S.E.(1997).Whatmoneycan’tbuy:Familyincomeandchildren’slifechances.
Cambridge,MA:HarvardUniversityPress.
Mazumder,B.(2005).Theapplefallsevenclosertothetreethanwethought:New
andrevisedestimatesoftheintergenerationalinheritanceofearnings.InS.Bowles,H.Gintis,&M.Groves(Eds.),Unequalchances:Familybackgroundandeconomicsuccess(pp.80–99).Princeton,NJ:RussellSageFoundation.
McLoyd,V.(1998).Socioeconomicdisadvantageandchilddevelopment.American
Psychologist,53,185–204.
Morgan,S.,&Harding,D.(2006).Matchingestimatorsofcausaleffects:Prospects
andpitfallsintheoryandpractice.SociologicalMethods&Research,35(1),3–60.Moyi,P.,Pong,S.,&Frick,W.(2004).Co-residentgrandparentsandgrandchildren’s
academicperformance.PopulationResearchInstituteWorkingPaperSeries,PennsylvaniaStateUniversity.
Mulligan,C.(1999).Galtonvs.thehumancapitalapproachtoinheritance.Journalof
PoliticalEconomy,107(6),S184–S224.
Musick,K.,&Mare,R.D.(2006).Recenttrendsintheinheritanceofpovertyand
familystructure.SocialScienceResearch,35,471–499.
NationalCenterforEducationalStatistics.(2004a).Earlychildhoodlongitudinalstudy:
Datafilesandelectroniccodebook,ECLS-Kbaseyearpublic-usechildfile,teacherfile,schoolfile.Washington,DC:U.S.DepartmentofEducation.
NationalCenterforEducationalStatistics.(2004b).User’smanualfortheECLS-Kbase
yearpublic-usedatafilesandelectroniccodebook.Washington,DC:U.S.Depart-mentofEducation.
Phillips,M.,Brooks-Gunn,J.,Duncan,G.,Klebanov,P.,&Crane,J.(1998).Family
background,parentingpractices,andtheblack-whitetestscoregap.InC.Jencks,
226J.L.Ferguson,D.D.Ready/EarlyChildhoodResearchQuarterly26(2011)216–226
&M.Phillips(Eds.),Theblack-whitetestscoregap(pp.103–145).Washington,DC:BrookingsInstitutionPress.
Rosenbaum,P.R.,&Rubin,D.B.(1983).Thecentralroleofthepropensityscorein
observationalstudiesforcausaleffects.Biometrika,70(1),41–55.
Rothstein,R.(2004).Classandschools:Usingsocial,economic,andeducationalreform
toclosetheblack-whiteachievementgap.Washington,DC:EconomicPolicyInsti-tuteandTeachersCollege,ColumbiaUniversity.
Rubin,D.(1997).Estimatingcausaleffectsfromlargedatasetsusingpropensity
scores.AnnalsofInternalMedicine,127,757–763.
Rubin,D.,&Thomas,N.(1996).Matchingusingestimatedpropensityscores:Relat-ingtheorytopractice.Biometrics,52,249–264.
Shadish,W.R.,Cook,T.D.,&Campbell,D.T.(2002).Experimentalandquasi-experimentaldesignsforgeneralizedcausalinference.Boston:HoughtonMifflinCompany.
Shapiro,T.(2004).ThehiddencostofbeingAfricanAmerican:Howwealthperpetuates
inequality.NewYork,NY:OxfordUniversityPress.
Simmons,T.,&Dye,J.L.(2003).Grandparentslivingwithgrandchildren:2000.Wash-ington,DC:U.S.CensusBureau(C2KBR-31).
Smith,J.,Brooks-Gunn,J.,&Klebanov,P.(1997).Consequencesoflivinginpovertyfor
youngchildren’scognitiveandverbalabilityandearlyschoolachievement.InJ.Brooks-Gunn,&G.Duncan(Eds.),Consequencesofgrowinguppoor(pp.132–189).NewYork,NY:RussellSageFoundation.
Smith,M.L.,&Shepard,L.A.(1988).Kindergartenreadinessandretention:Aqual-itativestudyofteachers’beliefsandpractices.AmericanEducationalResearchJournal,25,307–333.
Solon,G.(1999).Intergenerationalincomemobilityinthelabormarket.InO.
Ashenfelter,&D.Card(Eds.),Handbookoflaboreconomics(pp.1761–1800).Amsterdam,TheNetherlands:ElsevierScience.
Szinovacz,M.(1998).Grandparentresearch:Past,present,andfuture.InM.Szi-novacz(Ed.),Handbookongrandparents(pp.1–22).Westport,CT:GreenwoodPress.
Tomlin,A.(1998).Grandparents’influenceongrandchildren.InM.Szinovacz(Ed.),
Handbookongrandparents(pp.159–170).Westport,CT:GreenwoodPress.
Uhlenberg,P.,&Kirby,J.(1998).Grandparenthoodovertime:Historicalanddemo-graphictrends.InM.Szinovacz(Ed.),Handbookongrandparents(pp.23–39).Westport,CT:GreenwoodPress.
U.S.DepartmentofEducation.(2001).Theconditionofeducation2001.Washington,
DC:NationalCenterforEducationStatistics,U.S.GovernmentPrintingOffice,NCES2001072.
Warren,J.,&Hauser,R.(1997).Socialstratificationacrossthreegenerations:New
evidencefromtheWisconsinLongitudinalStudy.AmericanSociologicalReview,63,561–572.
Weber,M.(1946).Class,statusandparty.InH.H.Gerth,&C.WrightMills(Eds.),Max
Weber:Essaysinsociology(pp.180–195).London,UK:OxfordUniversityPress.Willis,P.(1977).Learningtolabor:Howworkingclasskidsgetworkingclassjobs.New
York,NY:ColumbiaUniversityPress.
Wolff,E.(2002).Topheavy:TheincreasinginequalityofwealthinAmericaandwhat
canbedoneaboutit.NewYork,NY:NewPress.
Yeung,W.,&Conley,D.(2008).Black-whiteachievementgapandfamilywealth.
ChildDevelopment,79(2),303–324.
因篇幅问题不能全部显示，请点此查看更多更全内容