Opportunity Engines: Middle-Class Mobility in Higher Education...• Middle-Class Mobility varies...
Transcript of Opportunity Engines: Middle-Class Mobility in Higher Education...• Middle-Class Mobility varies...
Opportunity Engines: Middle-Class Mobility in Higher Education
SarahReberRubensteinFellow,BrookingsInstitution
ChenoahSinclairResearchAssistant,BrookingsInstitution
May 2020
Introduction InequalityintheUnitedStateshasbeenrisinginrecentdecades,whileintergenerationalmobilityremainslow.Thismeansthatabsolutemobility—theextenttowhichchildrenareeconomicallybetteroffthantheirparents—isdeclining,andintergenerationalinequalityisincreasinglyentrenched.Alongliteraturesuggestslargereturnstoattendingcollegeandpointstotheimportanceofhighereducationforintergenerationalmobility.RecentworkbyOpportunityInsightsexploresinmoredetailtherolethatdifferentcollegesplayinpromotingupwardmobility,pointingtosignificantdifferencesacrosscollegesintheextenttowhichtheyenrollstudentsfromlow-incomefamilieswhohavehighearningsasadults.Here,weusethedataproducedbyOpportunityInsightstofocusspecificallyonstudentsfrommiddle-classfamiliestounderstandpatternsofattendanceandupwardmobilityforthemiddleclass.
BuildingontheCollegeScorecardDatacompiledbytheU.S.DepartmentofEducation,theOpportunityInsightsMobilityReportCardsusede-identifieddatafromtaxreturns,linkedtoinformationaboutcolleges,toconstructapubliclyavailabledatabaseforcollegesintheUnitedStates.AlthoughtheyprovidemanymeasuresintheMobilityReportCards,theOpportunityInsightsteamfocusesprimarilyonupwardmobilityforthelowest-incomestudents—thosewhoseparents’incomefallsinthebottomquintile—andonthelikelihoodthatthosestudentsmakeitallthewaytothetopearningsquintileasadults.This“Bottom-to-Top"measureofmobilityisimportant,butitisnotthewholestory.Itisimportanttoalsoconsidertheprospectsforstudentsfromabroaderrangeoftheincomedistribution,andinparticularfrommiddle-classfamilies.
TheBottom-to-TopMobility(BTM)measuredependsonbothhowmanybottom-quintilestudentsacollegeenrolls(bottom-quintile“access”)andwhatproportionofthoseenrolledbottom-quintilestudentsreachthetopearningsquintileinadulthood(“success”).Collegeswithlowaccessforbottom-quintilestudentshaverelativelylowBTMevenifthosewhodoenrollhavehighupwardmobility;conversely,collegeswherelow-incomestudentsarenotupwardlymobilehavelowBTMeveniftheyenrollmanylow-incomestudents.
WedevelopananalogousmeasureofMiddle-ClassMobility(MCM),focusingonupwardmobilityforstudentsfrommiddle-incomefamilies.Ourmeasureistheshareofstudentsatacollegewhocomefromthemiddlequintileoftheparentalincomedistribution—andmoveupatleastonequintileinadulthood.(Notethatthroughoutthispaper,weusetheterm“middleclass”torefertothemiddlequintile.)Wealsoadjustthismeasuretoaccountfordownwardmobilityandfortypicalmobilityamongstudentswhodonotattendcollegeatall,discussedfurtherbelow.
Key Findings • Middle-ClassMobilityvariessubstantiallyacrosscolleges.
• CollegeswithhighBottom-to-TopMobilitydonotnecessarilyhavehighMiddle-ClassMobility,andvice-versa;thecorrelationbetweenthetwomeasuresisjust0.26.
• Selectivefour-yearcollegeshavethehighestratesofMiddle-ClassMobilityonaverage,followedbynonselectiveandhighlyselectivefour-yearcolleges,thentwo-yearandfor-profitcolleges.
• Publicandprivatefour-yearcollegeshavesimilaraverageMiddle-ClassMobilityrates,butpublicfour-yearscontributesubstantiallymoretoupwardmobilityoverallbecausetheyenrollmanymorestudents.
• Two-yearcollegesaccountforasmallershareoftotalMiddle-ClassMobility(31percent)thantheirshareofmiddle-classenrollment(43percent).Butper-studentinstructionalspendingislowerandstudentsinthesecollegesspendfeweryearsincollege,onaverage,sothesectornonethelessaccountsforalargeshareofnetupwardmobilitycomparedtoitsshareoftotalspending,just11percent.
• Conversely,highlyselectivecollegesaccountforadisproportionateshareofnetupwardmobilitycomparedtoenrollment.Buttheyhavehighper-studentinstructionalexpenditure,sotheyaccountforalargershareofspendingthanMiddle-ClassMobility.
• Selectivefour-yearcollegesaretheworkhorsesofupwardmobilityforthemiddleclass,accountingfor34percentofmiddle-classenrollment,50percentofspendingonmiddle-classstudents,and43percentofMiddle-ClassMobility.
Background Intergenerational mobility is low in the United States Thefactthatchildrenfrompoorfamiliesarelikelytobepooradults,whilechildrenluckyenoughtobeborntowell-offparentstendtogrowuptobewell-offadultsiswell-established.Figure1showsthispatternusingtheOpportunityInsightsdataforthecohortsanalyzedinthisreport.Thefigureshowstheadultearningsquintileforchildrenwhostartineachquintileofparentalincome.Ifadultearningsdidnotdependonparentalincome—thatis,ifintergenerationalmobilitywerehigh—eachbarwouldlookthesame,showing20percentineachadultearningsquintileregardlessofparentalincomequintile.Infact,thebarslookquitedifferent,suggestingsignificantpersistenceineconomicoutcomesacrossgenerations.Forexample,thefirstbarshowsthatamongthosegrowingupwithparentswhohadthelowestincome(thebottom20percentofthedistribution),only9percentreachedthetop20percentoftheearningsdistributionasadults,and31percentdidnotexperienceanyupwardmobility,earninginthebottom20percentasadults.Bycontrast,therightmostbarshowsthatforchildrengrowingupwithparentsinthetop20percentoftheparentalincomedistribution,37percentwereinthetop20percentoftheearningsdistributionasadults.Andonly13percentofchildrengrowingupwiththemost-affluentparentsendedupinthebottomearningsquintileasadults.Americans’chancestomakeit
tothetopoftheeconomicheap,oreventhemiddle,dependsignificantlyontheincomeoftheirparents.
Figure2showstheintergenerationaltransitionmatricesforthosewhoneverattendedcollege,attendedcollegelaterinlife,andattendedcollege“on-time”(betweenages19and22).Itshowssignificantlymoreupwardmobilityforchildrenfromlow-incomehouseholdswhoattendedcollege.Forexample,42percentofpeoplewhoseparentswereinthebottomquintileandwhodidnotgotocollegehadadultearningsinthebottomquintile,comparedtojust24percentwhoattendedcollegelaterinlifeand16percentforthosewhoattendedcollegeon-time.Forthosestartinginthemiddleoftheparentalincomedistribution,collegeattendanceisassociatedwithahigherlikelihoodofstayinginthemiddleoftheearningsdistributionormovingup.Thisisconsistentwithalonglineofresearchshowingthatobtainingacollegeeducationboostsearnings.
Whenitcomestoreachingthetop20percentoftheearningsdistribution,however,theimportanceofbeinganon-timeattenderbecomesclear.Amongpeoplewhoseparentswereinthebottomquintile,18percentofon-timeattendersreachedthetopquintileinadulthood,comparedto5percentoflate-attendersand4percentofnon-attenders.Asimilarpatterncanbeseenforpeoplefrommiddle-classfamilies:25percentofon-timeattendersreachedthetopquintileinadulthood,comparedto9percentoflate-attendersand8percentofnon-attenders.Whilethoseattendingcollegelaterinlifehaveadultearningsoutcomesmoresimilartonever-attendersthanon-timeattenders,late-attendersaresubstantiallylesslikelytoendupinthelowestadultearningsquintile.Thesepatternsarepartiallyattributabletodifferencesinothercharacteristicsofthosewhoattendcollegeon-time,laterinlife,ornotatall,butattendingcollegelaterinlifemayoffersomeprotectionagainstlowearnings,evenifitisunlikelytolaunchastudenttothetopoftheearningsdistribution.
Unfortunately,youngadultswhogrowupinlower-andmiddle-classfamiliesaremuchlesslikelytoattendcollegeon-timeoratall,comparedtotheirmoreaffluentpeers.Figure3showsthedistributionofyoungadultsacrosscollegeattendancecategoriesbyparental
income.Therelationshipbetweenparentalincomeandcollegeattendanceratesisstrong:onlyone-thirdofthosefromthelowest-incomefamiliesattendcollegeon-time.Attendanceratesincreasesharplywithparentalincome,withjustoverhalfofchildrenofmiddle-quintileparentsheadingofftocollegeon-time,comparedto87percentforthosefromthetop-quintile.Someofthisgapismadeupslightlylaterinlife,aschildrenoflower-incomeparentsaremorelikelytoattendcollegelaterinadulthood.Still,largesharesoflow-andmiddle-incomechildrennevergoontoattendcollege,andFigure2suggestslateattendershavebetteroutcomesthannever-attenders,butdoworsethanthosewhoenrollincollegeshortlyafterhighschoolgraduation.Researchsuggeststhatthesegapsarenotfullyexplainedbydifferencesinacademicpreparation.Infact,low-scoringstudentsfromaffluentbackgroundsareaboutaslikelytograduatefromcollegeashigh-scoringstudentsfromlow-incomefamilies.
Moreover,evenamongstudentswhodoenrollincollegesoonafterhighschoolgraduation,theselectivityandtypeofcollegeattendeddependsstronglyonparentalincome.Figure4showsthatlow-andmiddle-incomestudentsarepredominatelyservedbytwo-yearcollegesandnonselectivefour-yearcolleges,whereashigher-incomestudentsaremorelikelytoattendselectiveandhighlyselectivefour-yearcolleges.Thisisconsistentwiththe
researchliteratureshowingthatchildrenofhigher-incomefamiliesaremorelikelytoattendcollege,andtoattendmore-selectivecolleges,comparedtotheirmiddle-classandlow-incomecounterparts.
A Middle-Class Mobility Measure WedevelopameasureofmobilitysimilarinspirittoChettyetal.’sBottom-to-TopMobility(BTM)measurebutadaptittofocusonstudentsfromthemiddleoftheparentalincomedistribution.WebeginbydescribingthedataChettyetal.useandhowtheyconstructBTMbeforeturningtohowwemodifythisapproachtoconstructourownMiddle-ClassMobility(MCM)measure.
Data Wedonothaveaccesstotheindividual-level,de-identifiedtaxdatausedbyChettyetal.intheiranalysis,sowearelimitedtothedatatheymakepublicthroughOpportunityInsights.Fortunately,thepublic-usedatasetincludesawiderangeofinformationaboutcollegesthatwecanuseinthisanalysis.ThissectiondetailshowtheOpportunityInsightsteam
constructedthecollege-levelvariablesusedinthisanalysis;interestedreaderscanconsultthefulldocumentationandthispaperformoreinformation.
Sample TheOpportunityInsightsteamconstructedcollege-levelstatisticsusingde-identifiedtaxfilings.ThesampleincludedallindividualsintheU.S.whohaveavalidSocialSecurityNumberorIndividualTaxpayerIdentificationNumber,werebornbetween1980and1982,andcouldbelinkedtoatleastoneparentinthetaxdata.Childrenwerelinkedtoparentsbasedonthemostrecenttaxfilerstoclaimthechildasadependentduringtheperiodwhenthechildwas12–17yearsold.Ifthechildwasclaimedbyasinglefiler,thechildisdefinedashavingasingleparent.About2percentofchildrenwereneverclaimedasdependentsandwereconsequentlyexcludedfromtheanalysis.1
Ouranalysisnecessarilyfocusesonoldercohortsbecauseweneedtimetoseehowtheyfareinthelabormarket.Thecohortsrepresentedinthesedatalargelyattendedcollegepriortorecentexpansionsofthefor-profitsector,2whichwasaccompaniedbyincreasinglypredatorypractices.Mobilitymeasuresforthefor-profitsectorshouldthereforebeinterpretedwithcaution.
Identifying where students attend college OpportunityInsightsprovidescollege-levelestimatesofattendancederivedfromtwosources:federaltaxrecordsandDepartmentofEducationrecordsfrom1999to2013.TheyusedatafromForm1098-T—aninformationalreturnfiledbycollegesforeachenrolledstudentforthepurposesofreportingtuitionpayments—andPellgrantrecordstodeterminewhetherandwhereeachpersonenrolledincollege.Astudentisconsideredtohaveattendedacollegeinaschoolyearifshehasa1098-TfiledbythecollegeorreceivedaPellgrantthatyear.Studentswhoattendedmorethanonecollegeareassignedtothecollegeattendedmostfrequentlybetweentheagesof19and22;ifastudentattendedtwoormorecollegesforthesamenumberofyears,sheisassignedtothefirstcollegeattended.
The1098-TformsidentifyeachcollegebyitsEmployerIdentificationNumber(EIN)andZIPcode.Importantly,somecollegesfiletheseformsformultiplecampusesusingasingleEIN-ZIP,inwhichcasewecannotdistinguishbetweencampusesinthesamesystem.Forexample,theUniversityofMarylandreportsunderasingleEIN-Zipcode,lumpingtogether
1Theshareofchildrenwhocanbesuccessfullylinkedtoparentsdropssignificantlywhenlookingatbirthcohortsbefore1980.ThisisbecausethefederaltaxdatausedbyChettyetal.tomeasureparents’andchildren’sincomesbeginsin1996,andmanychildrenleavethehouseholdstartingatage17.Consequently,Chettyandcolleagueslimittheiranalysissampletochildrenborninorafter1980.Thisrangeisrestrictedfurtherinthepublic-accessdata,wheredataareonlyavailableforchildreninthe1980-‘82birthcohorts,withsupplementarydataprovidedfromthe1983and1984cohortswheninformationforacollegeforoneofthesecohortsisincomplete.
2Between1986and2009,enrollmentwithinfor-profithighereducationinstitutions(FPHEIs)grewfrom2%tomorethan10%ofallstudentsenrolledininstitutionsofhighereducation(Liu,2011).Furthermore,thepercentageofundergraduatesattendingFPHEIsmorethandoubledbetween1995and2012,from5%to13%overallandfrom1%to17%in4-yearFPHEIs(NationalCenterforEducationStatistics,2017).
studentswhoattendanyofits15campuses.OtherlargesystemstowhichthislimitationappliesincludetheUniversityofTennessee,UniversityofIllinois,andUniversityofMinnesota. Weexcludetheseaggregatedcampusesfromtheanalysis.
Measuring parental income and adult earnings Childrenintherelevantcohortsarelinkedtoparentswhoclaimedthemasadependent,asdiscussedabove.DataontheincomeofparentsandtheearningsofchildrencomefromfederalincometaxreturnsandW-2formstocaptureincomeforthosewhodonotfiletaxreturns.Parentincomeisdefinedastotalpre-taxincomeforthehousehold,whichincludesbothearningsandotherformsofincome(suchasinterestpaymentsandgovernmentbenefits).Theyaverageparentalincomeforthefiveyearswhenthechildisaged15–19toobtainameasureofresourcesavailablewhencollegeattendancedecisionsaretypicallybeingmade.Childrenarethenassignedtoparentalincomequintilesbyrankingthemonthismeasurerelativetootherchildreninthesamebirthcohort.
Earningsinadulthoodforthesechildrenaredefinedastotalpre-taxindividualearnings.Unliketheincomemeasureforparents—whichaccountsfornon-wageformsofincomelikeunemploymentbenefitsandinterestpayments—thismeasureisrestrictedtoearnings,definedasthesumofwagesandnetself-employmentearnings.Theearningsofchildrenweremeasuredin2014whentheywerebetween32and34yearsold,lateenoughtofinishtheirschoolingandgainsomeexperienceinthelabormarket.Theywereassignedtoearningsquintilesbyrankingthemrelativetoothersinthesamebirthcohort,regardlessofcollegeattendancestatus.
TheOpportunityInsightsteamusethesedatatocountthenumberofstudentsineachcollegewhocomefromeachparentalincomequintileandreacheachadultearningsquintile.Weusethesecollege-levelcountsinouranalysis.FollowingChettyetal.,werestricttheanalysistocollegeswithanaveragecohortsizeofatleast200forthe1980to1982birthcohorts;thisexcludes416institutionsthataccountfor2.8percentofallstudents.
College Characteristics
WeusedatafromtheIntegratedPostsecondaryEducationDataSystem(IPEDS)database3andtheCollegeScorecardmadeavailablethroughOpportunityInsights.4UsingtheBarron’sselectivityrankingprovidedintheCollegeScorecarddataasaguideline,wegroupcollegesintothefollowing5selectivitytiers:
3IPEDSisasystemofinterrelatedsurveysconductedannuallybytheU.S.DepartmentofEducation.Thesurveyscontaindataonenrollment,programcompletion,graduationrates,facultyandstaff,finances,institutionalprices,selectivity,andstudentfinancialaidforallinstitutionsthatparticipateinfederalstudentaidprograms.
4Weusethe“selectedcharacteristics”file.
• HighlySelectiveFour-Year:Four-yearcollegesbelongingtotheeliteandhighlyselectiveBarron’sselectivitycategories.
• SelectiveFour-Year:Four-yearcollegesbelongingtothe3rdand4thBarron’sselectivitycategories.
• NonselectiveFour-Year:Four-yearcollegesbelongingtothe5thandnonselectiveBarron’sselectivitycategories,aswellasfour-yearcollegeswithmissingBarron’srankings.
• Two-Year:Alltwo-yearcollegesandlessthantwo-yearcollegesthatarenotfor-profit.
• For-Profit:Alltwo-yearandfour-yearfor-profitcolleges.WeusetheIPEDSdatatoclassifycollegesintosectorsbasedonhighestdegreeofferedandwhethertheyarepublic,privatenon-profit,orfor-profit:
• PrivateFour-Year:Private,non-profitcollegesthatofferafour-yeardegree.• PublicFour-Year:Publiccollegesthatofferafour-yeardegree.• Two-Year:Publicandnon-profitcollegesthatdonotofferafour-yeardegree,
includingthosethatonlyoffercertificateprograms.Inpractice,thevastmajorityofthesecollegesarepubliccommunitycollegesthatoffertwo-yeardegrees.
• For-Profit:Allfor-profitcolleges,regardlessofhighestdegreeoffered.Opportunity Insights’ Bottom-to-Top Mobility measure OurMCMmeasurebuildsconceptuallyontheBTMmeasureemphasizedbyOpportunityInsights.BTMistheshareofacollege’senrollmentthatisbothfromthebottomquintileofparentalincomeandinthetopquintileofadultearnings.Ithastwocomponents:
• Accessistheshareofacollege’senrollmentthatcomesfromthebottomquintile.Ittellsushowmanylow-incomestudentsacollegeserves,comparedtototalenrollment.
• Successistheshareofbottom-quintilestudentsatacollegewhomakeittothetopquintileoftheadultearningsdistribution.Ittellsushowlow-incomestudentsdo,conditionalonenrollinginacollege.
• MobilityisequaltoAccessmultipliedbySuccess.
Table1illustratesthiscalculationforseveralcolleges.Forexample,atHarvard,only3.0percentofenrollmentisfromthebottomquintileoftheparentalincomedistribution(bottom-quintileaccess).Amongthosestudents,57.7percenthaveadultearningsthatputtheminthetopquintileoftheadultearningsdistributionfortheircohort(bottom-to-topsuccess).Thisimpliesthat1.8percentofHarvard’stotalenrollmentcomesfromthebottomparentalincomequintileandisinthetopearningsquintileinadulthood.SUNY-StonyBrookhasaslightlylowersuccessrateof51.2percent,butbottom-quintilestudentscompriseamuchlargershareofenrollment(16.4percent),yieldingaBTMrateof8.4percent.WrightCareerCollegeprimarilyserveslow-incomestudents,soithasahighbottom-quintileaccessrate(42.1percent).However,only1.1percentofbottom-quintile
studentswhoattendWrightmakeittothetopquintileoftheadultearningsdistribution,soWrighthasamobilityrateofjust0.5percent.
Table1.CalculatingBottom-to-TopMobilityforselectcolleges Q1Access
↓ P(Parentin
Q1)
X
SuccessRate↓
P(ChildinQ5|ParentinQ1)
=
UnadjustedMobilityRate↓
P(ChildinQ5&ParentinQ1)
HarvardUniversity
3.0% X 57.7% = 1.8%
SUNY-StonyBrook
16.4% X 51.2% = 8.4%
WrightCareerCollege
42.1% X 1.1% = 0.5%
A new measure of Middle-Class Mobility ToconstructourmeasureofMiddle-ClassMobility,weusethesameapproachdescribedabove,butwedefineaccessandsuccessinwaysthataremorerelevanttothemiddleclass.AlthoughtheFutureoftheMiddleClassInitiativedefinesthemiddleclassasthosefallinginthemiddle-threequintilesoftheincomedistribution,wefocusexclusivelyonthethirdquintile.Thissimplifiestheanalysiscomparedtoconsideringquintiles2,3,and4,andproducesasimilarpicture.5Webeginbydefiningaccessandsuccessforthemiddleclass:
• Accessistheshareofacollege’senrollmentcomingfromthethirdquintileoftheparentalincomedistribution.
• Successistheshareofthird-quintilestudentsattendingacollegewhomoveupatleastonequintilenettheshareofstudentswhofallatleastoneearningsquintile.Bythismeasure,acollegehasapositive“success”measureifmoreofitsthird-quintilestudentsmoveupthanmovedown.
Somechildrenfromlow-incomefamilieswillbeupwardlymobiledespitenotattendingcollege;thosewhoseparentsareinthebottomincomequintilecannotbedownwardlymobilebydefinition,andsomewillevenmakeittothetopearningsquintileinadulthoodwithoutattendingcollege.Figure2showsthat4percentofnon-attendersfromthebottomparentalincomequintilewereinthetopearningsquintileasadults.AsChettyetal.note,ahypotheticalcollegethatenrolledonlystudentsfromthebottomquintile(bottom-quintile
5Interpretingbothaccessandsuccessmeasures—andthereforemobilitymeasures—forthemiddle-threeincomequintilesproveddifficult.Forexample,acollegemighthavelowermiddle-three-quintileaccessbecauseitenrollsmanybottom-quintilestudentsorbecauseitenrollsmanytop-quintilestudents,andtheequityimplicationsofthosealternativesarequitedifferent.Mechanically,thepossibilityofupwardordownwardeconomicmobilityalsovaryconsiderablyacrossthemiddle-threequintiles(childrenfromthesecondquintilehavemoreroomtomoveupthanchildrenfromfourth-quintilefamilies),anditisnotobvioushowtoweightdifferentquintiletransitionstomakeasingle“success”measure.FocusingonthethirdquintilesimplifiestheinterpretationoftheMCMmeasure.
accessof100percent),4percentofwhomhadadultearningsinthetopquintile(bottom-to-topsuccessof4percent),wouldhaveaBTMrateof4percent,wellabovetheaverage,eventhoughbottom-quintilestudentsweredoingnobetterthanstudentswhodidnotattendcollegeatall.Theyadjustforthisbysubtractingthebottom-to-top“success”ofnon-attendersfromeachcollege’ssuccessrate.Bythismeasure,acollegeonlygets“credit”forupwardmobilitythatexceedstheaveragemobilityamongnon-attenders.
Chettyandcolleaguesusethisadjustedmeasureinrobustnesschecksbutfinditmakeslittledifferencebecausecollegeswithveryhighaccessratesforbottom-quintilestudentsarerare.However,thisadjustmentmattersmoreformeasuringMiddle-ClassMobility.Onaverage,studentsfromthethirdquintileofparentalincomewhoneverattendcollegefallintheearningsdistributionrelativetotheirparents’positionintheincomedistribution;theaveragenetmobilityrateformiddle-classnon-attendersisabout–29percent.Thatis,substantiallymorenon-attendersfrommiddle-classfamiliesaredownwardlymobilethanupwardlymobile.Followingthelogicdescribedabove,we“netout”theaveragemobilityofnon-attendersinourmeasureofsuccessforthemiddleclass.Forcomparability,weusetheadjustedmeasureofBTM,ratherthanthemainmeasureChettyandcolleaguesuseintheiranalysis.Themeasureswithandwithoutthisadjustmentarehighlycorrelated,buttheadjustmentaffectsourcomparisonsacrosssectorsandselectivitytiers,particularlyforthemiddleclass,soweusetheadjustedversioninouranalysis.
Table2showshowthisadjustmentaffectsthecalculationsinTable1.Theaccessmeasureisthesame,butthesuccessmeasureisreducedby3.9percentagepoints—theaverage“success”forthosewhodon’tattendcollege.Themobilityrateisalsoreduced.
Table2.CalculatingadjustedBottom-to-TopMobilityrateforselectcolleges Q1
Access
X
AdjustedBottom-to-TopSuccessRate
=
AdjustedBottom-to-TopMobilityRate
HarvardUniversity 3.0%
X 53.9%
= 1.6%
SUNY-StonyBrook 16.4%
X 47.4%
= 7.8%
WrightCareerCollege 42.1%
X -2.7%
= -1.1%
Table3illustratesthecalculationofourMiddle-ClassMobilityrate.AtHarvard,studentsfromthethirdquintileoftheparentalincomedistributioncomprise8.1percentoftotalenrollment(middle-classaccess).Amongthosestudents,56.6percentmoremoveupatleastonequintilethanmovedownatleastonequintile.Butamongstudentswhodon’tattendcollegeatall,successbythismeasureisnegative28.9percent(moremovedownthanup),soweadd28.9to56.6percenttogettheadjustedmiddle-classsuccessrateforHarvard.Thiscanbeinterpretedasnetupwardmobilityforthird-quintilestudentswhoattendHarvard,comparedtonetupwardmobilityforthird-quintilestudentswhonever
attendcollege.MultiplyingaccessbythismeasureofsuccessyieldstheMiddle-ClassMobilityrate.
Table3.CalculatingMiddle-ClassMobilityRateforselectcolleges Middle-Class
Access
X
Middle-ClassSuccessRate
=
Middle-ClassMobilityRate
HarvardUniversity 8.1%
X 85.5%=
(56.6%+28.9%)
= 6.9%
SUNY-StonyBrook 16.1%
X 83.5%=
(54.6%+28.9%)
= 13.5%
WrightCareerCollege 16.5%
X -10.4%=
(-39.3%+28.9%)
= -1.7%
Figure5showshowthe(adjusted)Bottom-to-TopMobilitymeasurerelatestothenewMiddle-ClassMobilitymeasureforallcolleges.CollegeswithhigherMiddle-ClassMobilityalsohavehigherBottom-to-TopMobility,onaverage,buttherelationshipisnotverystrong.Thecorrelationis0.26.Figure5alsoshowssubstantialvariationinMCMacrosscolleges.Aninteractiveversionofthisfigureallowstheusertofindaspecificcollege,orfilteroncategoriesofcolleges,toseehowcollegesstackuponthesemeasures.
The“success”measures—andthereforethemobilityrates—reflectdifferencesacrosscollegesbothinwhoattendsthecollegeandthecausalimpactofthecollegeitself.Implicitly,thesuccessmeasurescompareoutcomesforstudentswhoattendedaparticularcollegetooutcomesfortheaveragestudentwhoneverattendedcollege.Variationinthismeasureacrosscollegescouldbeduetodifferencesinwhichstudentscollegesacceptandenrollordifferencesintheeffectofattendance.ThecalculationsinTable3showthatmiddle-classstudentswhogotoHarvardhavesimilarupwardmobilitytothosewhoattendStonyBrook,butthosewhoattendWrightdomuchworse;infact,theyhaveworseoutcomesthanstudentswhoneverattendcollege.PartofthedifferenceinoutcomesforstudentsattendingHarvardandStonyBrook,ontheonehand,andWrightontheother,isalmostcertainlypartiallyduetodifferencesincharacteristicsofstudentswhoattendeachcollege,suchasacademicpreparationorstudyskills.Intheiranalysis,Chettyandcolleaguesshowthat,conditionalonattendingthesamecollege,outcomesareremarkablysimilaracrossthedistributionofparentalincome;theyarguethedatasuggestasubstantialshareofvariationinsuccessacrosscollegesisduetodifferencesinthecausalimpactofcollegesratherthanselection.Nevertheless,itisimportanttokeepinmindthatdifferencesacrosscollegesinsuccess—andthereforemobility—reflectdifferencesinbothselectionandimpact.
Differences in mobility across different types of colleges Figure5showsthatMCMvariesconsiderablyacrosscollegesandisonlyweaklycorrelatedwithBTM. Inotherwords,thecollegeswiththemoststudentsmovingfromthebottomrungoftheeconomicladdertothetoparenotnecessarilythesameastheoneswiththemoststudentsmovingupfromthemiddlequintile.
Dosomecollegeshavesystematicallyhigherorlowermobilitydependingonhowselectivetheyareortheirsector?Weaddressthisquestionbyanalyzingaveragemobilitybythesecharacteristics;wealsoexaminethetwocomponentsofmobility—accessandsuccess—tobetterunderstandthereasonsfordifferencesinmobilityratesfordifferenttypesofcolleges.
Middle-Class Mobility by college selectivity tier Studentsfromlow-incomefamiliesattendcollegesthatarelessselective,onaverage,comparedtomiddle-andhigher-incomestudents(Figure4).Figure6showshowaccess,success,andmobilityforbottom-quintileandmiddle-classstudentsvarydependingoncollegeselectivity.Thefirstpanelshowsaveragebottom-quintileandmiddle-quintileaccess—theshareoftotalenrollmentthatcomesfromthosequintiles—foreachselectivitytier.Thesecondpanelshowsaveragesuccess,andthefinalpanelshowsmobility —whichincorporatesaccessandsuccess.(NotethatFigure6showstheproportionofstudentsfromeachquintileenrolledininstitutionsineachcategory,whileFigure4showstheproportionfromeachquintileenrolledineachtypeofinstitution.)
Theleftpanelshowsthatforstudentsfromthebottomquintile,accessincreasesasselectivitydeclines.Onaverage,lessthan5percentofstudentsinhighlyselectivecollegescomefromfamiliesinthebottomquintile,comparedto22percentinfor-profitcolleges.Asimilarrelationshipholdsforstudentsfromthemiddleclass,thoughitislesspronounced.Onaverage,about11percentofstudentsinhighlyselectiveinstitutionscomefromthemiddleclass,comparedto22percentinfor-profits.
Themiddlepanelshowsaverage“success”byselectivitytier.Recallthatforthebottomquintile,successisdefinedasmakingittothetopquintileoftheadultearningsdistribution(asinChettyetal.);forthemiddleclass,wedefinesuccessasnetupwardmobility;inbothcases,wenetoutthesuccessrateofnon-attenders,asdescribedabove.Forstudentsfromthebottomquintile,themore-selectiveschoolshavesignificantlyhighersuccessrates.Amonglow-incomestudentswhoattendhighlyselectivefour-yearcolleges,44percentmakeittothetopearningsquintileinadulthood.Asimilarpatternisobservedformiddle-classstudents.
Thelastpanelshowsthemobilitymeasures.Amongfour-yearcolleges,BTMisrelativelyconstantacrossselectivitytiers.Thisisbecausedifferencesinaccessandsuccessoffset
eachother.Forexample,selectiveandnonselectivefour-yearshavesimilarmobilityrates(.017and.016,respectively),butselectivefour-yearcollegeshavehighersuccessandloweraccess,whereasnonselectivefour-yearcollegeshavehigheraccessandlowersuccess.Amongtwo-yearcollegesandfor-profits,averageBTMratesareslightlylower,whichislargelyattributabletotheirweaksuccessrates.
IntermsofMCM,selectivefour-yearcollegeshavethehighestmobilityrates,onaverage,followedbynonselectiveandhighlyselectivefour-yearcolleges,thentwo-yearcollegesandfor-profits.AswithBTM,accessandsuccessaresomewhatoffsetting—butnottoquitethesamedegree.Selectivefour-years,forexample,havebothhighaccessandhighsuccess,explainingtheirhighaveragerateofMiddle-ClassMobility.
ThefirsttwopanelsinFigure6underscorethatthesimilarityofmobilityratesacrossselectivitytiersobscuresthedifferingcontributionsofaccessandsuccess.Unfortunately,thetypesofinstitutionswherelow-incomeandmiddle-classstudentsaremostlikelytobesuccessful—highlyselectiveandselectivefour-years—aretheleastaccessible,whereastheschoolsatwhichthesestudentscaneasilyenrollareassociatedwiththepoorestlabor-marketoutcomes.
Middle-Class Mobility by college sector Figure7showshowaccess,success,andmobilityvarybysector,definedas:
• Private(non-profit)four-year• Publicfour-year• Two-year6• For-profit(includingtwo-yearandfour-year)
ThefirstpanelofFigure7showsthat,withtheexceptionoffor-profits,studentsfromthebottomquintilemakeupasmallershareofenrollmentacrossallsectors,comparedtomiddle-classstudents.Accessforbottom-quintilestudentsissignificantlyloweratfour-yearcolleges,particularlythosethatareprivate,comparedtotwo-yearorfor-profitcolleges.Asabove,differencesinaccessandsuccessforbottom-quintilestudentsarelargelyoffsetting,soBTMrates(thirdpanel)aresimilaramongfour-yearcollegesandtwo-yearsandfor-profits.
6Private,non-profittwo-yearcollegesarerare(lessthan2percentoftwo-yearcolleges),sowecombinethemwithpublictwo-years.
Formiddle-classstudents,accessfollowsasimilar,butlesssteeppattern,andsuccessforbothpublicandprivatefour-yearcollegesissubstantiallyhigherthanfortwo-yearandfor-profitcolleges,onaverage.MCMrates(lastpanel)varymoreacrosssectorsthandoBTMrates.AswithBTM,four-yearcollegeshavehighermobilitythantwo-yearsorfor-profits,butthefour-yearadvantageislargerforthemiddleclass,comparedtothebottomquintile.
Accounting for upward mobility TheresultspresentedinFigures6and7showhowaccess,success,andmobilityratesvarydependingoncollegecharacteristics.TheMiddle-ClassMobilitymeasureindicateswhatshareofacollege’senrollmentisbothfromthemiddleclassandupwardlymobile.However, holdingconstantacollege’smobilityrate,acollegethatenrollsmorestudentsoverallwillcontributemoretoupwardmobilitythanacollegethatenrollsfewerstudents.
Weaccountforthesedifferencesincollegesizebycalculatingthenumberofupwardlymobilestudentswhoattendeachcollegeandexamininghowtotalupwardmobilityis
distributedacrosscollegesofdifferenttypes.7Thetotalnumberof(net)upwardlymobilestudentsfromthemiddlequintilebyourdefinitionaveragedapproximately160,000percohort.Wethencalculatethecontributionofeachcollegetothattotal,asdemonstratedinTable4:Harvardhas1,609studentspercohort,8.1percentofwhomcomefromthethirdparentalincomequintile—30students.Amongthird-quintilestudentsatHarvard,netupwardmobilityis85.5percentagepointshigherthannetupwardmobilityfornon-attenders.Harvardthereforeaccountsfor111ofthe160,000upwardlymobilemiddle-classstudentsbythismeasure.StonyBrookhasbothhigheraccessforthemiddlequintileandhigherenrollmentoverall,soitcontributesmoretoupwardmobility,despitehavingaslightlylowermiddle-classsuccessrate.Wrighthasanegativesuccessrate—meaningstudentsenrolledtheredoworsethanstudentswhodon’tattendcollegeatall—buttheydon’tenrollmanystudents,sotheir(negative)contributionissmall.
Table4.Calculatingcolleges’contributiontooverallMiddle-ClassMobility
Q3Access
X Cohortsize
= NumberofQ3
students
X AdjustedMiddle-ClassSuccessRate
= Numberof
SuccessesHarvardUniversity
8.1%
X 1,609 = 130 85.5%
X 111
SUNY-StonyBrook
16.1% X 2,070 = 333 83.5% X 278
WrightCareerCollege
16.5% X 249 = 41 -10.4% X -4
Thecostsassociatedwithdifferenttypesofcollegesalsovarysubstantially,bothbecauseper-pupil,per-yearinstructionalspendingissignificantlyhigherinmore-selectiveinstitutions,onaverage,andbecausestudentswhoenrollintwo-yearcollegesenrollforfeweryears,onaverage,thanthoseenrolledinfour-yearcolleges.Wecalculatetheestimatedcostassociatedwithbottom-quintileandmiddle-classstudentsateachcollege;thiscalculationisnecessarilyapproximatebutneverthelessallowsforaroughcomparisonoftotalspendingtoenrollmentandupwardmobilitybytypeofcollege.8
7Asabove,wenetouttheaverageupwardmobilityrateforstudentswhodon’tattendcollege.Thenumberofupwardlymobilemiddle-classstudentsatacollegeisthesuccessmeasuretimesthenumberofmiddle-classstudentsthecollegeenrolls.Inotherwords,itisthenumberofadditionalstudentswhomovedupatleastonequintilecomparedtowhatwouldbeexpectedbasedontransitionsamongnon-attenders.
8Wemultiplyenrollmentintherelevantquintile(bottomorthird)byper-pupilinstructionalspendingin2000asreportedinIPEDS;wescalethatby1.5for2-yearcollegesand4for4-yearcollegestoaccountforthedifferentaveragetimesstudentsspendincollegeineachsector.Theseestimatesarenecessarilycrude,asthe
Figure8showstheshareofenrollment,spending,andmobilityacrossselectivitytiersforbothbottom-quintileandmiddle-classstudents.Forexample,highlyselectivefour-yearcollegesaccountfor5percentofbottom-quintileenrollment,21percentofestimatedinstructionalspending,and17percentofBottom-to-TopMobility.Forbottom-quintilestudents,two-yearcollegesaccountfor49percentofenrollment,15percentofestimatedspending,and28percentofupwardmobility.
Thesecondsetofbarsshowsthesameestimatesformiddle-classstudents.Forthemiddleclass,selectivefour-yearcollegesaccountforthelargestshareofupwardmobility–43percent–whichisdisproportionaterelativetotheirenrollmentshareof34percent,butslightlylessthantheir50percentshareofestimatedspending.Two-yearcollegesaccountforthelargestshareofenrollment—43percent—butonly11percentofspendingand31percentofupwardmobility.
typicalnumberofyearsstudentsattendmayvaryacrosscollegesandtheinstructionalspendingistheaverage,notmarginal,cost.Nevertheless,thisback-of-the-envelopecalculationcangiveussomesenseofhowspendingisdistributedacrosstypesofcolleges.
ThepatternsinFigure8forbottom-quintileandmiddle-classstudentsaresimilar,thoughmiddle-classstudentsaresomewhatmorelikelytoattendselectiveandhighlyselectivefour-yearcollegesandlesslikelytoattendtwo-yearcolleges.Highlyselectivecolleges’contributiontoupwardmobilityisdisproportionaterelativetotheirenrollment,butstillsmallduetotheirsmallenrollmentshare.Highlyselectivecollegesspendmoreperstudent,sotheirshareofmobilityisalsolowerthantheirshareofspending.Two-yearcollegesaccountforlargesharesofenrollmentbutsmallsharesofspending—asstudentsspendlesstimeenrolledandspendingperstudentislower—andmoderatesharesofupwardmobility.Selectivefour-yearcollegesaccountforthelion’sshareofupwardmobility—butanevenlargershareofspending.
Figure9showsthesameanalysisbysector.Thetwo-yearandfor-profitcollegesaredefinedasinFigure8,butherewedividefour-yearcollegesnotbyselectivitybutbywhethertheyarepublicorprivate.Publiccolleges,includingtwo-yearandfour-yearcolleges,accountforalmost80percentofupwardmobilityonbothmeasures.Thisfollowsfromthesimplefactthatmoststudentsattendpubliccolleges.Privatecollegesaccountforadisproportionatelylargeshareofupwardmobilityrelativetotheirenrollment,buttheyalsohavehighinstructionalspending.Publicfour-yearcollegesaretheworkhorsesofupwardmobility,accountingforlargesharesofenrollment,spending,andupwardmobility.Buttwo-yearcollegesprovidealotof“bangforthebuck”,sincetheyhavelowspending.For-profitsofferupwardmobilityroughlyproportionaltotheirenrollment,thoughtheydidn’tenrollmanystudentsinthesecohorts.Again,itisimportanttonotethatthesecohortspre-dateaconsiderableexpansioninthefor-profitsectorandincreasingconcernsaboutpredatorypractices.
Discussion Attendingcollegeisassociatedwithupwardmobilityforstudentswhogrowupinlow-incomeandmiddle-classfamilies.Butnotallcollegesofferthesameopportunitiesforupwardmobility,andwhetherandwhereyoungadultsattendcollegedependsheavilyontheirparents’income.Inthispaper,wedevelopanewmeasureofMiddle-ClassMobility,incorporatingmeasuresofbothaccessforthemiddleclassandupwardmobilityconditionalonattendance.WeshowthattheextenttowhichdifferentcollegescontributetoMiddle-ClassMobilityvariesconsiderablyacrossselectivitytierandsector,butalsowithineachcategory.TheaccompanyinginteractiveshowsMCMandBTMmobilityratesforindividualinstitutions.Althoughouranalysisdoesnotidentifycausalpathwaysthatcanbemanipulatedtoimproveoutcomesformiddle-classstudents,itdoeshighlightseveralpatternsofinteresttopolicymakers.
Thecollegeswiththebestoutcomesformiddle-classstudents,namelyselectiveandhighlyselectivefour-years,aretheleastaccessibletypesofinstitutionsforthesestudents.Ontheotherhand,two-yearcollegesarequiteaccessible—enrollingnearlyhalfofmiddle-classstudents—buttheirstudentsdonotexperiencethesameupwardmobility,comparedtothoseenrollinginfour-yearcolleges.NewresearchfromOpportunityInsightsshowsthatthesedifferencesarenotfullyexplainedbydifferencesinSAT/ACTscores.Low-andmiddle-incomestudentsattendless-selectivecollegesthantheirhigherincomepeersevenwhentheyhavethesametestscores.Whilethebetteroutcomesenjoyedbystudentswhoattendmore-selective,four-yearcollegesaredueinparttodifferencesinacademicpreparation,researchsuggeststhatatleastpartofthedifferenceisacausaleffectofcolleges:studentsbenefitfromaccesstomore-selectivecolleges.
Collegesattendedbylower-andmiddle-incomestudentsalsospendmuchlessperstudent,comparedtothemore-selectivecollegesattendedbytheiraffluentpeers.Instructionalspendinghasbeenshowntohavelarge,positiveimpactsondegreecompletion.Ontheonehand,therelativelylowspendingattwo-yearcollegesandfour-yearpublicsmeansthatthoseinstitutionsoffervalue.Ontheotherhand,thoseinstitutionshavelowcompletionrates,sothelargelylow-andmiddle-incomestudentstheyservewouldlikelybenefitfromhigherspending.PolicymakersshouldbeespeciallymindfultoprotecttheseinstitutionsfromdeepcutsinresponsetotheCOVID-19crisis.
TheteamatOpportunityInsightssuggestchangestoadmissionspoliciesatselectivefour-yearcollegestoimproveaccess.Weagree.Buttheeffectivenessofthisstrategyforimprovingupwardmobilitycouldbelimitediflow-andmiddle-incomestudentsfaceotherbarrierstoenrollmentorifthosecollegesfailtodeliverthesamebenefitwhentheybecomemoreaccessible.Policymakersshouldbeatleastasconcernedaboutboostingsupportforthecollegesalreadyservingmillionsoflow-incomeandmiddle-classstudents,especiallypublictwo-yearandmoderatelyselectivefour-yearcolleges.
Segregationofhighereducation—wherethechildrenfromlow-andmoderate-incomefamiliesattendless-selective,under-resourcedcolleges,andhigher-incomepeersattendbetter-resourcedcolleges,regardlessoftheirtestsscores—actsasadragonintergenerationalmobility.Acombinationofchangesinpoliciestoreducesegregationandinequalityoffundingacrosshighereducationisurgentlyneededtorealizethepromiseofhighereducationasanengineformiddle-classmobility.
References
Bailey,M.J.,&Dynarski,S.M.(2011).Gainsandgaps:ChanginginequalityinUScollegeentryandcompletion.NationalBureauofEconomicResearch(No.w17633).
Björklund,A.,&Jäntti,M.(2009).Intergenerationalincomemobilityandtheroleoffamilybackground.Oxfordhandbookofeconomicinequality,491,521.
Bleemer,Z.(2018).TopPercentPoliciesandtheReturntoPostsecondarySelectivity.AvailableatSSRN3272618.
Chetty,R.,Hendren,N.,Kline,P.,&Saez,E.(2014).Whereisthelandofopportunity?ThegeographyofintergenerationalmobilityintheUnitedStates.TheQuarterlyJournalofEconomics,129(4),1553-1623.
Chetty,R.,Grusky,D.,Hell,M.,Hendren,N.,Manduca,R.,&Narang,J.(2017).ThefadingAmericandream:Trendsinabsoluteincomemobilitysince1940.NationalBureauofEconomicResearch(No.w22910).
Chetty,R.,Friedman,J.N.,Saez,E.,Turner,N.,&Yagan,D.(2017).Mobilityreportcards:Theroleofcollegesinintergenerationalmobility.NationalBureauofEconomicResearch(No.w23618).
Chetty,R.,Friedman,J.,Saez,E.,Turner,N.,&Yagan,D.(2020).TheDeterminantsofIncomeSegregationandIntergenerationalMobility:UsingTestScorestoMeasureUndermatching.NBERWorkingPaper(No.w26748).
Corak,Miles(2006).Dopoorchildrenbecomepooradults?Lessonsfromacrosscountrycomparisonofgenerationalearningsmobility,IZADiscussionPapers,No.1993,InstitutefortheStudyofLabor(IZA),Bonn.
Corak,M.(2013).Incomeinequality,equalityofopportunity,andintergenerationalmobility.JournalofEconomicPerspectives,27(3),79-102.
Deming,D.J.,&Walters,C.R.(2017).TheimpactofpricecapsandspendingcutsonUSpostsecondaryattainment.NationalBureauofEconomicResearch(No.w23736).
Dynarski,S.(2015).ForthePoor,theGraduationGapIsEvenWiderThantheEnrollmentGap.TheNewYorkTimes.https://www.nytimes.com/2015/06/02/upshot/for-the-poor-the-graduation-gap-is-even-wider-than-the-enrollment-gap.html
Goldstein,D.,&Hartocollis,A.(2020).ASimpleWaytoEqualizetheIvies?GiveOtherstheLegacySATBonus.TheNewYorkTimes.https://www.nytimes.com/2020/02/12/us/SAT-bonus-ivy-league.html.
Greenstone,M.,&Looney,A.(2012).Regardlessofthecost,collegestillmatters.BrookingsonJobNumbers,BrookingsInstitution.
https://www.brookings.edu/blog/jobs/2012/10/05/regardless-of-the-cost-college-still-matters/.
Greenstone,M.,Looney,A.,Patashnik,J.,&Yu,M.(2013).Thirteeneconomicfactsaboutsocialmobilityandtheroleofeducation.BrookingsInstitution(report).https://www.brookings.edu/research/thirteen-economic-facts-about-social-mobility-and-the-role-of-education/.
Haskins,R.,Holzer,H.,&Lerman,R.(2009).Promotingeconomicmobilitybyincreasingpostsecondaryeducation.BrookingsInstitution(report).https://www.brookings.edu/research/promoting-economic-mobility-by-increasing-postsecondary-education/.
Holzer,H.J.,&Baum,S.(2017).Makingcollegework:Pathwaystosuccessfordisadvantagedstudents.BrookingsInstitutionPress.
OpportunityInsights(2020).DataLibrary.https://opportunityinsights.org/data/?geographic_level=0&topic=0&paper_id=536#resource-listing.
Reeves,R,&Guyot,K.(2018).Therearemanydefinitionsof“middleclass”—here’sours.BrookingsInstitution(blog).https://www.brookings.edu/blog/up-front/2018/09/04/there-are-many-definitions-of-middle-class-heres-ours/.
Reeves,R,&Guyot,K.(2019)....Andjusticeforall:Communitycollegesservingthemiddleclass.BrookingsInstitution(blog).https://www.brookings.edu/blog/up-front/2019/06/13/and-justice-for-all-community-colleges-serving-the-middle-class/.
Reeves,R,&Venator,Joanna(2014).SavingHoratioAlger:TheDataBehindtheWords(andtheLegoBricks).BrookingsInstitution(blog).https://www.brookings.edu/blog/social-mobility-memos/2014/08/21/saving-horatio-alger-the-data-behind-the-words-and-the-lego-bricks/.
U.S.DepartmentofEducation(2015).TheCollegeScorecard.https://collegescorecard.ed.gov/data/.