Professional Documents
Culture Documents
Reardon District Ses and Achievement Discussion Draft April2016 PDF
Reardon District Ses and Achievement Discussion Draft April2016 PDF
SeanF.Reardon
StanfordUniversity
PreliminaryDraft,fordiscussion
Version:April,2016
TheresearchdescribedherewassupportedbygrantsfromtheInstituteofEducationSciences
(R305D110018),theSpencerFoundation,andtheWilliamT.GrantFoundation.Thepaperwouldnothave
beenpossiblewithouttheassistanceofRossSanty,MichaelHawes,andMarilynSeastrom,who
facilitatedaccesstotheEdFactsdata.Thispaperbenefittedsubstantiallyfromongoingcollaborationwith
AndrewHo,DemetraKalogrides,KennethShores,ErinFahle,andBenShear.Someofthedatausedinthis
paperwereprovidedbytheNationalCenterforEducationStatistics(NCES).Theopinionsexpressedhere
aremyownanddonotrepresentviewsofNCES,theInstituteofEducationSciences,theSpencer
Foundation,theWilliamT.GrantFoundation,ortheU.S.DepartmentofEducation.Direct
correspondenceandcommentstoSeanF.Reardon,sean.reardon@stanford.edu,520CERASBuilding
#526,StanfordUniversity,Stanford,CA94305.
SchoolDistrictSocioeconomicStatus,Race,andAcademicAchievement
HowmuchdoesacademicperformancevaryamongschooldistrictsandcommunitiesintheU.S.?
Howmuchofthatvariationisduetothesocioeconomiccontextoftheschoolsandthesocioeconomic
backgroundofthestudents?Howdotestscoresvarybyracewithinandbetweendistricts?
ThisshortreportusesnewdatafromtheStanfordEducationDataArchivetoinvestigatethesequestions.
Itisaworkingdraftthatwillbeupdatedinthenextfewweeks.Theintendedpurposeistohighlightkey
patternsofacademicachievementacrossthecountry;infuturepapersIwilladdadditionalmoredetailed
analyses.
TestScoreData
ThetestscoredatausedherecomefromtheStanfordEducationDataArchive(SEDA),which
includesestimatesoftheaveragetestscoresofstudentsinalmosteverypublicschooldistrictinthe
UnitedStates(seda.stanford.edu).Theseestimatesarebasedonroughly215millionstateaccountability
testscores(representingmorethan40millionstudents)onmathandEnglishLanguageArts(ELA)testsin
grades38intheyears20092013ineverypublicschooldistrictintheUnitedStates.Detailsonthe
sourceandconstructionoftheestimatesisavailableontheSEDAwebsite.
Thescoresareplacedonacommonscaleacrossstates,grades,andyears,sothatperformance
canbemeaningfullycomparedacrossplaces.ThescaleIusehereisoneinwhichtestscoresare
standardizedinrelationtotheaveragenationalperformanceforeachgradelevel.Althoughscoresare
availableforeachschooldistrictineachyearfrom2009to2013andeachgradefrom3to8,Iusea
versionoftheSEDAdatathataveragesthesescoresacrossgradesandyearstocreateasingleaverage
scoreforeachdistrict;IalsopoolthescoresacrossmathandELAforpresentationalsimplicity,sincethe
twoareveryhighlycorrelatedatthedistrictlevel.Theresultingmeasureisscaledsothatavalueof0
indicatesthatstudentsinadistrictscore,onaverage,atthenationalaverageofstudentsattheirgrade
level;aunitdifferenceinthescalecorrespondstothenationalaveragedifferenceinscoresbetween
studentsinadjacentgrades.
Ialsoestimatetheaveragewithingrade(acrosscohort)changeinscoresforeachdistrictandthe
averagewithincohort(acrossgrades)changeinscoreswithineachdistrict.Themeasureoftherateof
changeacrossgradeshasameanof1byconstruction(studentsscoresincreasebyonegradelevelon
averageeachyear);largervaluesindicatestudentsinadistrictmakefasterthanaveragegrowth(soa
growthrateof1.1wouldindicatethatstudentscoresinthatdistrictgrow10%fasterthanaverage,or
aboutahalfgrademorethanaveragefrom3rdto8thgrade).
MeasuringAverageSocioeconomicStatusAmongStudentsEnrolledinaSchoolDistrict
Inordertomeasurethesocioeconomiccharacteristicsofthefamiliesofchildren,Iusedatafrom
theAmericanCommunitySurvey(ACS).TheACSincludesdetailedsociodemographicdataforfamilies
livingineachschooldistrictintheU.S.;thesetabulationsareavailablethroughtheSchoolDistrict
DemographicSystem(SDDS).Iusedatafromthe200610SDDStabulationsbecausetheyinclude
tabulationsoffamilycharacteristicsamongfamilieswithschoolagechildrenenrolledinpublicschools.
Inparticular,Iusesixmeasuresofthesocioeconomiccompositionoffamilieslivinginadistrict
withchildrenenrolledinpublicschools:1)medianfamilyincome;2)percentofadultswithabachelors
degreeorhigherdegree;3)povertyrate;4)unemploymentrate;5)SNAPeligibilityrate;6)thepercentof
familiesheadedbyasingleparent.Eachoftheseisavailableseparatelybyrace/ethnicity(forracial/ethnic
groupsofsufficientlocalpopulationsize).
Iconstructameasureofeachdistrictsaveragesocioeconomicstatusasthefirstprincipal
componentofthesixmeasuresabove.Thismeasureisstandardizedtohaveameanofzeroanda
standarddeviationof1.Togiveasenseofhowthismeasureisscaled,Table1describestheaverage
characteristicsofschooldistrictsatvariousvaluesoftheSEScomposite.
Table1:
AverageFamilySocioeconomicCharacteristics,atVariousDistrictSESCompositeValues
SESComposite
3 2 1 0 1 2
MedianFamilyIncome $24,038 $31,026 $39,634 $53,029 $78,644 $136,804
%WithBAorHigher 13.5% 14.9% 14.6% 18.3% 32.3% 62.4%
PovertyRate 48.0% 37.6% 25.9% 14.7% 6.0% 1.6%
SNAPEligibilityRate 50.0% 39.9% 27.6% 15.5% 5.6% 0.2%
UnemploymentRate 10.5% 8.0% 6.0% 4.5% 3.4% 2.6%
SingleParentFamilyRate 51.9% 41.9% 31.7% 22.2% 14.6% 10.0%
Ialsoconstructthesamesocioeconomicstatuscompositeseparatelybyrace.Idothisby
applyingthesamescoringequationtotheracespecificsocioeconomicdistrictcharacteristics.Thismeans
thatthescaleofsocioeconomicstatusisdirectlycomparableacrossrace/ethnicgroups.Icomputerace
specificSESonlywheretheSDDStabulationsofACSdataincluderacespecificmeasuresforall6ofthe
variablesinthecomposite(whenagivenracialgroupistoosmallwithinaschooldistrict,theSDDS
tabulationsarenotavailable).
ThedataIusehereincludes11,280schooldistrictsforwhichIamabletocomputea
socioeconomicstatusvariableandforwhichtheSEDAdataincludemeasuresofacademicachievement.
Districtsnotincludedinthesamplearepredominantlyverysmalldistrictsforwhichsamplesaretoosmall
forSDDStoreportsocioeconomiccharacteristicsorthathavefewerthan20studentstotalpergrade(in
whichcasetheSEDAdatadonotincludeestimatesofaveragetestscores).The11,280districts
collectivelyenroll3.7millionstudentspergrade(over90%ofallpublicschoolstudentsintheU.S.)
Whenreportingracespecificaveragetestscores,Ilimitthesampletodistrictbyrace
combinationsforwhichtherewereatleast200studentspergradeandatleast100studentsofagiven
racepergradeinthedistrict.Thisrestrictionensuresthattheracespecificmeansarehighlyreliable.The
sampleincludes1,514districtsforwhichSEDAincludesaveragetestscoresforwhitestudents;946with
scoresforblackstudents;and1,115withscoresforHispanicstudents.Thesedistrictsinclude46%ofall
whitestudents;78%ofallblackstudents;and74%ofallHispanicstudents.
Table2belowdescribestheaveragesocioeconomiccompositevaluesamongdistricts,weighted
byeachrace/ethnicgroupsenrollment.ThetoprowindicatestheaverageSESamongall11,280districts
inoursample,weightedbyracial/ethnicenrollment.Theaveragewhitestudentisenrolledinadistrict
withSESlevels0.94standarddeviationshigherthantheaverageblackstudent;and0.63standard
deviationshigherthantheaverageHispanicstudent.Amongthe1,514districtsforwhichIhavewhite
meanscores,theaveragewhitestudentisinadistrictwheretheaverageSESis0.17,butwherethe
averageSESamongwhitestudentsis0.56.Inotherwords,whitestudentstypicallyattendschooldistricts
wheretheirwhitepeersarefrommoreadvantagedfamiliesthantheirnonwhitepeers.Thepatternis
theoppositeforblackandHispanicstudents.Theaverageblackstudentattendsschoolinadistrictwhere
herblackpeersarefarpoorerthanhernonblackpeers;thesameholdsforHispanicsthoughthe
discrepancyisnotaslarge.
Thesebetweenandwithindistrictdiscrepanciesareimportantforunderstandingbetweenand
withindistrictracialdisparitiesinacademicachievement,asweshallseebelow.
Table2:
AverageDistrictSocioeconomicCharacteristics,byRace/EthnicGroup
Group
White Black Hispanic
AverageDistrictSES(inallDistricts) 0.30 0.64 0.33
AverageDistrictSES(inRaceSampleDistricts 0.17 0.63 0.28
AverageSESofSameRaceinDistrict 0.56 1.99 0.96
N(districts) 1514 946 1115
Note:Meansareweightedbyracespecificdistrictenrollment.
AverageAcademicPerformance,bySchoolDistrictSocioeconomicStatus
Ibeginbyexaminingthepatternofassociationbetweendistrictsocioeconomicstatusand
averageacademicachievement.Beforedoingso,itisimportanttonotethataveragetestscoresina
districtshouldnotbeinterpretedasameasureofschoolquality.Testscoresandacademicperformance
moregenerallyareshapedbymanyfactorsotherthanschools.Theyareshapedbychildrensfamilies,
theirhomeenvironments,theirneighborhoodcontexts,theirchildcareandpreschoolexperiences,
afterschoolexperiences,andbytheirschools.Knowingthatchildreninaparticularcommunityscored
higher,onaverage,thanthoseinanothercommunitydoesnottellusthattheschoolswerebetterinthat
community.Averagetestscoresaremoreappropriatelyinterpretedasameasureoftheeducational
opportunitiesavailabletochildrenlivingwithinadistrict.Moreover,whilemathandELAtestscoresarea
proxyforthedesiredoutcomesofschooling;theydonotmeasureallaspectsofchilddevelopmentthat
students,parents,andsocietyvalue.
Withthatinmind,Figure1plotstheaveragetestscoresineachofthe11,280schooldistricts,
Figure1
AcademicAchievementandSocioeconomicStatus
USSchoolDistricts,20092013
AverageAchievement(GradeLevels)
4
5
4 3 2 1 0 1 2 3
<Poor/DisadvantagedAffluent/Advantaged>
againsttheaverageSESlevelofthestudentsenrolledinthedistrict.Thebubblesintheplotareweighted
bythesizeofthedistrict.
Severalthingsarestrikinginthisfigure.Firstistheobvious,andverystrong,relationshipbetween
districtsocioeconomicstatusandaverageacademicachievement.Thedistrictlevelcorrelationbetween
thetwois0.78;thestudentlevelcorrelation(thatis,thecorrelationweightedbydistrictenrollment),is
0.84.Studentsinmanyofthemostadvantagedschooldistrictshavetestscoresthataremorethanfour
gradelevelsabovethoseofstudentsinthemostdisadvantageddistricts.Thesocioeconomiccontextofa
schooldistrictisaverypowerfulpredictorofstudentsacademicperformance
Ofthe1,000poorestdistrictsintheU.S.,only68(6.8%)havemeantestscoresatorabovethe
nationalaverage.Thesearemostlysmalldistricts;theycollectivelyenrollabout7,000studentspergrade
(lessthantwotenthsofonepercentofallstudentsintheU.S.).Likewise,ofthe1,000mostaffluent
districtsintheU.S.,only16(1.6%ofdistricts,collectivelyenrollingfewerthan1,000studentspergrade)
havemeantestscoresatorbelowthenationalaverage.Inotherwords,wehavelittleevidencethatwe
knowhowtoprovideadequateeducationalopportunitiesforchildrengrowingupinlowincome
communities.
Secondisthefactthat,despitethisstrongassociationbetweenSESandaverageacademic
performance,thereareschooldistrictswiththesamelevelofSESbutwithmeaningfullydifferentaverage
testscores.TheresidualstandarddeviationaroundthefittedregressioncurveinFigure1is0.64.Roughly
5percentofdistrictshaveaveragescoresmorethanonegradelevelabovewhatwewouldpredictbased
ontheirsocioeconomicstatus;another5percenthaveaveragescoresmorethanonegradelevelbelow
whatIwouldpredict.Thisvariationisnotconfinedtosmalldistricts;evenamonglargeschooldistricts
(highlightedinFigure2),thereissubstantialvariationinaveragetestscoreamongschooldistrictswith
similarsocioeconomicprofiles.
Figure2
AcademicAchievementandSocioeconomicStatus
100LargestUSSchoolDistricts,20092013
AverageAchievement(GradeLevels)
4
5
4 3 2 1 0 1 2 3
<Poor/DisadvantagedAffluent/Advantaged>
Racial/EthnicDifferencesinAverageAchievement
Figure1showstheaverageperformanceofallstudentsineachschooldistrict.Butevenwithin
thesamedistrict,therearelargeracialdifferencesinaverageachievement.InFigure3,eachbubble
correspondstoarace/ethnicgroupwithinaspecificschooldistrict.Thepointsareplottedagainstthe
averagedistrictSES(ofallstudents).SoonecanthinkofthisfigureastakingeachpointinFigure1,and
breakingitintomultiplebubbles,eachofwhichisplacedverticallyaboveorbelowtheoveralldistrictdot
inaccordancewiththatgroupsaveragetestscores.
Notethattheonlypointsshownherearethosecaseswherethereareatleast100studentsofa
givenracepergradeinadistrict,andwherethereareatleast200studentspergradeoverallinthe
district.Manydistrictsaretoosmalltomeetthisthreshold,andevenamongthosewithatleast200
students,manydonothavemorethanonegroupwithatleast100studentspergrade.Thereforemany
districtsdonotappearatallinFigure3;someappearonlyforonerace;othersappearfortwoorthree
groups.
Figure3illustratesthattheracialachievementgapsareverylarge,evenamongstudentsof
differentrace/ethnicitieswhoattendschoolswithsimilarsocioeconomicconditions.Theverticaldistance
betweenthethreegroupsisquitelarge.Onaveragewhitestudentsscoreoneandhalformoregrade
levelshigherthanblackandHispanicstudentsenrolledinsocioeconomicallysimilarschooldistricts.
TheotherstrikingfeatureofFigure3isthatthereareveryfewschooldistrictswhereblack
studentsscore,onaverage,atorabovethenationalaverage.Infact,ofthe946schooldistrictswithat
least100blackstudentspergrade,thereareonly18districts,enrollingatotalofroughly3000black
studentspergrade(onehalfofonepercentofallblackstudentsintheU.S.)inwhichblackstudents
averagetestscoresareatorabovethenationalaverage.
Figure3
AcademicAchievementandSocioeconomicStatus,byRace/Ethnicity
USSchoolDistrictsWith100+StudentsofaGivenRace/Ethnicity,20092013
AverageAchievement(GradeLevels)
4
2
WhiteStudents 3
HispanicStudents
4
BlackStudents
5
5 4 3 2 1 0 1 2 3
<Poor/DisadvantagedAffluent/Advantaged>
OverallDistrictSocioeoconomicStatus
Onereasonforthelargeracial/ethnicdisparitiesinachievementmaybethatevenamongthose
wholiveindistrictswithsimilarsocioeconomicconditions,blackandHispanicstudentsarepoorer,on
average,thanwhitestudents(seeTable2above).Figure4presentsthesamesampleofdistrictsasFigure
3,butnoweachrace/ethnicgroupisplacedhorizontallyinrelationtoitsowngroupsaverage
socioeconomicstatuswithinthedistrict.Blackstudentshave,onaverage,lowerlevelsofSESthan
Hispanicstudents,whointurnhavelowerSESthanwhitestudents.Figure4showsthatthesedifferences
explainsome,butnotallofthedifferencesinacademicachievementbetweenwhiteandnonwhite
students.EveninschooldistrictswhereblackstudentshaverelativelyhighSES,theyscorefarbelow
(morethanagradelevelbelow)whitestudentsindistrictswherewhitestudentsthesamesocioeconomic
status.Clearlyracialdifferencesinsocioeconomiccontextandconditionsarepartofthereasonforblack
andHispanicstudentslowerachievementthanwhitestudents,buttheyarenotthewholereason.
Figure4
AcademicAchievementandSocioeconomicStatus,byRace/Ethnicity
USSchoolDistrictsWith100+StudentsofaGivenRace/Ethnicity,20092013
AverageAchievement(GradeLevels)
4
2
WhiteStudents 3
HispanicStudents
4
BlackStudents
5
5 4 3 2 1 0 1 2 3
<Poor/DisadvantagedAffluent/Advantaged>
RaceSpecificDistrictSocioeoconomicStatus
GrowthinAcademicPerformanceAcrossGrades
10
TheSEDAdataincludeatleastoneyearoftestscoresfor10cohortsofstudents,thosewho
enteredkindergartenfromFall2000throughFall2009(theearliestofthesecohortswouldbeexpected
tobein8thgradein2009,thefirstyearoftheSEDAdata;thelatestofthesewouldbeexpectedtobein
3rdgradein2013,thelastyearoftheSEDAdata).Mostofthesecohortsareobservedfor2ormoreyears,
sowecanestimatetheaveragechangeintestscores,withincohorts.Theseestimatesareplottedin
Figure5inrelationtosocioeconomicstatus.
NotablehereistheSESgrowthrategradient.Averagetestscoresgrowmoderatelyfasterin
higherSESschooldistrictsthaninlowSESdistricts.Theslopeofthefittedlineisabout0.04,meaning
thatdistrictsthatdifferby1standarddeviationinSESdifferbyabout4%intheannualgrowthrateof
theirstudentsperformance.Thismeansthatstudentsinthemostaffluentschooldistrictsgainalmost1
yearmoreofacademicperformancegrowthbetweenthirdandeighthgradethandothepoorestschool
districts.
Figure5
AcademicAchievementGrowthandSocioeconomicStatus
100LargestUSSchoolDistricts,20092013
1.50
AverageGrowth(Grades/Year)
1.25
1.00
0.75
0.50
4 3 2 1 0 1 2 3
<Poor/DisadvantagedAffluent/Advantaged>
11
ThismeansthatperformancedisparitybetweenhighandlowSESdistrictsislargerin8thgradethanin3rd
grade.Figure6displaysthis;itshowsthefittedregressionlineoftheassociationbetweenschooldistrict
averagetestscoresandSESforeachgradefrom38.Thelineisroughly20%steeperby8thgrade.
DistrictSESAchievementGradient,byGrade
USSchoolDistricts,20092013
AverageAchievement(GradeLevels)
5
2
Grade3 Grade6
3
Grade4 Grade7
4
Grade5 Grade8
5
4 3 2 1 0 1 2 3
<Poor/DisadvantagedAffluent/Advantaged>
Conclusion
Thedatahereprovidealook,unprecedentedindetail,atthepatternsofacademicachievement
acrossschooldistrictsintheU.S.Theyshowseveralkeythings:
1. Thevariationinacademicachievementamongschooldistrictsisverylarge;studentsinsome
districtshavescoresmorethan4gradelevelshigherthanothers.
2. Thisvariationisveryhighlycorrelatedwiththesocioeconomiccharacteristicsoffamiliesinthe
localcommunity.Itisnotclear,however,howmuchoftheassociationisduetodifferencesinthe
qualityofschooling,orintheopportunitieschildrenhavetolearnanddevelopoutsideofschool,
intheirhomesandneighborhoods.
12
3. Theassociationbetweencommunitysocioeconomicstatusandacademicperformancegrows
steeperaschildrenprogressthroughschool.Again,itisnotclearwhetherthisresultsfrom
differencesinthequalityofschoolsinhighandlowincomecommunities,orbecauseof
differencesinchildrensoutsideofschoolopportunitiestolearn.
4. Racial/ethnicdisparitiesinacademicperformancearelarge,bothoverallandwithinindividual
schooldistricts.TheaveragewithindistrictwhiteblackandwhiteHispanicachievementgapsare
roughly2.0and1.5gradelevels,respectively.Moreover,extremelyfewblackandHispanic
studentsliveinschooldistrictswhereaverageachievementisatthenationalaverageforgrade
level.
5. Partofthewithindistrictracialachievementgapsaretheresultofracial/ethnicdisparitiesin
familysocioeconomicbackground.Buteveninplaceswherewhiteandblackorwhiteand
Hispanicstudentscomefromfamilieswiththesamesocioeconomiccharacteristics,racial/ethnic
achievementgapsarepresent,andsubstantial.
Together,thesefindingssuggestthatsocioeconomiccontextisapowerfulforceshaping
childrenseducationalopportunitiesandsuccess.Butpovertyisnotdestiny;inequalityisnotinevitable.
Thereareplaceswherechildrenofagivensocioeconomicbackgroundperformmuchbetteronteststhan
childreninotherplaceswiththesamebackground.Itisessentialthatwelearnfromsuchplacessothat
wecanimproveeducationalopportunitiesforchildrenwhodonothavetheopportunitytogrowupinan
affluentcommunity.
13