Russiasonyawalger

July,2020WorkingPaperNo.
20-032INTERGENERATIONALMOBILITYOFIMMIGRANTSINTHEUSOVERTWOCENTURIESRanAbramitzkyStanfordUniversity&NBERLeahBoustanPrincetonUniversity&NBERElisaJácomePrincetonUniversitySantiagoPérezUCDavis&NBERIntergenerationalMobilityofImmigrantsintheUSoverTwoCenturiesRanAbramitzkyLeahBoustanStanfordUniversityandNBERPrincetonUniversityandNBERElisaJácomeSantiagoPérezPrincetonUniversityUCDavisandNBERJune2020Abstract:Usingmillionsoffather-sonpairsspanningmorethan100yearsofUShistory,wefindthatchildrenofimmigrantsfromnearlyeverysendingcountryhavehigherratesofupwardmobilitythanchildrenoftheUS-born.
Immigrants'advantageissimilarhistoricallyandtodaydespitedramaticshiftsinsendingcountriesandUSimmigrationpolicy.
Immigrantsachievethisadvantageinpartbychoosingtosettleinlocationsthatofferbetterprospectsfortheirchildren.
Thesefindingsareconsistentwiththe"AmericanDream"viewthatevenpoorerimmigrantscanimprovetheirchildren'sprospects.
Keywords:Intergenerationalmobility,immigration*Wethanktheeditor,EstherDuflo,forhermostvaluableguidance,andthreeanonymousrefereesfortheirveryusefulfeedback.
WealsothankthefeedbackofChrisBecker,AlvaroCalderón,BrunoCaprettini,RajChetty,JenniferHunt,MicheleRosenberg,MartinSaavedra,KjellSalvanes,seminarparticipantsatCopenhagenBusinessSchool,CopenhagenUniversity,Munich,Princeton,NorwegianSchoolofEconomicsinBergen,Rotterdam,Stanford,Tilburg,andUCDavis,aswellasconferenceparticipantsatthe12thInternationalconferenceonMigrationandDevelopmentinMadrid,the"IntergenerationalMobility,Gender,andFamilyFormationintheLongRun"conferenceinOslo,theSocialScienceHistoryAssociationannualmeeting,andSITEconferenceonimmigrationatStanford.
AlongerversionofthepaperincludingadditionalresultsisavailableasNBERworkingpaper#26408.
11.
IntroductionImmigrantswhomovetotheUnitedStatesaspiretoofferabetterfuturefortheirchildren.
Bothtodayandinthepast,manyimmigrantsearnlessthanUS-bornworkersuponfirstarrivalanddonotcompletelycatchupwithinasinglegeneration.
1However,adefiningfeatureofthe"AmericanDream"istheviewthatevenimmigrantswhocometotheUnitedStateswithfewresourcesandlittleskillshavearealchanceatimprovingtheirchildren'sprospects.
Thispaperstudiestheintergenerationalmobilityofthechildrenofimmigrants(the"second-generation").
WeaskwhetherthesonsofimmigrantsachieveearningsparitywiththesonsoftheUS-born,andhowtheirrelativeintergenerationalmobilitychangedbetweenthelate19thcenturyandtoday.
2Ontheonehand,childrenofimmigrantsmightbeinaparticularlygoodpositiontomoveuptheladder,giventhattheirparentsmayearnlessthantheirtruetalentorabilitywouldsuggest(forexampleiftheyhadlittleexposuretoUSeducationorfaceddiscriminationinthelabormarket).
ImmigrantfamilieswhorecentlyarrivedintheUSmightalsobemorefootloose,thusallowingthemtosettleinareaswithbetterprospectsfortheirchildren.
Ontheotherhand,childrenofimmigrantsmightgrowupinsegregatedneighborhoods,sufferfromdiscriminationthemselves,andotherwiseenjoymorelimitedopportunitiesthanthechildrenoftheUS-born.
Wefindthat,bothhistoricallyandtoday,childrenofimmigrantsatthebottomoftheincomedistributionhavehigherratesofupwardmobilitythanchildrenofsimilarlyrankedUS-bornfathers,andtoastrikinglysimilardegreeineachtimeperiod.
Second-generationimmigrantsgrowingupatthe25thpercentileoftheincomedistributionendup5–6percentilerankpointshigherthanthechildrenoftheUS-bornwhowerealsoraisedatthesameincomerank.
Second-generationimmigrantstodayexhibitasimilardegreeofupwardmobility,despitenotableshiftsincountriesoforiginawayfromEuropetowardLatinAmericaandAsia,aswellasmajorchangesinUSimmigrationpolicyfromaregimeofnearlyopenborders(toEuropeanimmigrants)tooneofsubstantialrestrictions.
Ouranalysisspans130yearsofUShistoryandfocusesonthreecohortsofimmigrantswhoenteredtheUSduringthe19thor20thcenturies.
Thefirsttwocohortsconsistoffourmillionfirst-1SeeAbramitzky,Boustan,andEriksson(2014)onearningsconvergenceforimmigrantsinthepastandLubotsky(2007)formorerecentimmigrants.
2Inhistoricaldata,weareonlyabletolinksonsacrossCensuswaves,giventhatdaughtersoftenchangetheirnamesatmarriage.
Thus,forconsistency,wemostlyfocusonfather-sonpairsthroughoutouranalysis.
2generationimmigrantsobservedwiththeirchildrenintheUSineitherthe1880or1910Censuses.
3The1880cohortcontainsmostlyimmigrantsfromNorthernandWesternEurope(e.
g.
,Ireland,Germany,andtheUK),whereasthe1910cohortincludesmoreimmigrantsfromSouthernandEasternEuropethoughttohavefacedgreaterinitialdisadvantagesintheUSlabormarket.
4Wefollowthechildrenoftheseimmigrantstothe1910and1940Censuses,respectively—usinginformationontheirname,yearofbirth,andbirthplace—andcomparetheiradultoutcomestothoseofthechildrenofUS-bornwhites.
5BecausetheCensusesdidnotcollectincomedatabefore1940,ouranalysisofthesecohortsreliesoncomputinganumberofalternativeproxiesforindividual-levelincome("incomescores")basedonaperson'sdetailedoccupation,age,andstateofresidence.
Thethirdcohortincludeschildrenofimmigrantsbornaround1980.
Unlikethehistoricalcohorts,immigrantsinthismoderncohortenteredtheUSduringaneraofsubstantialimmigrationpolicyrestrictionsandcamemainlyfrompoorerandmoreethnicallydiversecountriesinLatinAmericaandAsia.
Tostudythiscohort,weuseaggregateadministrativedatamadepublicbytheOpportunityInsightsproject,whichisbasedonlinksofparentsandtheirnearlysixmillionchildren(Chettyetal.
2018a,2018b).
Inthesedata,weobservechildren'soutcomesin2014–15andtheoutcomesoftheirparentsin1994–2000.
Inaddition,weusedatafromtheGeneralSocialSurveys(GSS),inwhichweobservetheoutcomesofchildrenofimmigrantsandtheUS-bornin2000–2018.
TheGSShavesmallersamplesizesbutincludesomechildrenofundocumentedimmigrantsaswellasinformationonoccupationstofacilitatecomparisonswiththehistoricaldata.
WestartbydocumentingincomescoregapsbetweenimmigrantsandUS-bornworkers.
Inboththepastandpresent,thereisawidevariationinlabormarketoutcomesamongfirst-generationimmigrantsfromdifferentcountries.
ImmigrantsfromcountrieslikeFinlandandNorwayinthepastandVietnamandtheDominicanRepublictodayearnbelowtheUS-bornonaverage,whereasimmigrantsfromcountrieslikeEngland(historical)andIndia(today)out-earntheirUS-born3ImmigrantsinthesecohortsenteredtheUSbeforetheimpositionofstrictquotasbasedoncountryoforiginimposedin1921,whichendedtheeraofnearlyopenbordersforEuropeanmigrantsanddramaticallyreducedtheinflowofimmigrantsfromEasternandSouthernEurope.
4FormorebackgroundonimmigrationduringtheAgeofMassMigration,seeAbramitzkyandBoustan(2017).
5Morethan95percentofimmigrantsinthisperiodwerewhite.
FocusingonUS-bornwhitesasthecomparisongroupensuresthatthehighermobilityofsecond-generationimmigrantsthatweobserveisnotduetoblack-whitedifferencesinmobility.
CollinsandWanamaker(2017)findthatAfricanAmericanshadmuchlowerratesofupwardmobilitythanwhiteshistorically.
Thus,includingAfricanAmericansinthesamplewouldmakeimmigrants'mobilityadvantageappearevenlarger.
3counterparts.
Withafewexceptions,earningsdifferencesforimmigrantgroupsrangebetween+20and-40logpoints.
However,formostsendingcountriesinwhichfirst-generationimmigrantsearnedlessthantheUS-born,wefindthatsecond-generationimmigrantscatchuporevenovertaketheearningsofthechildrenoftheUS-born.
Ourmainanalysisthenusesfather-sonlinkstocomparetheaveragerankthatchildrencanexpecttoreachinthenationalincomedistribution,conditionalontheirparent'sincomeorincome-scorerankandimmigrationstatus(asimilarapproachtoChettyetal.
2014andChettyetal.
2018a).
Inboththepastandpresent,wefindthatthechildrenofimmigrantsaremoreupwardlymobilethanthechildrenoftheUS-born:conditionalontherankoftheirparents,childrenofimmigrantshaveahigherexpectedrankinadulthood.
Thehigherlevelofupwardmobilityamongchildrenofimmigrantsisespeciallymeaningfulinrelativelypoorfamilies:Theestimatedgapsimplythatchildrenofimmigrantswithparentsinthe25thpercentilehaveasimilarexpectedrankaschildrenofUS-bornindividualswhoseparentswererankedabout15–20percentileshigher,whichisaboutonehalfofthecontemporaneousgapbetweenblacksandwhitesintheUS(Chettyetal.
2018a).
WefindasmallermobilitygapbetweenchildrenofimmigrantsandchildrenoftheUS-bornatthetopoftheincomedistribution.
Ourdataenableustoestimateratesofintergenerationalmobilityseparatelybyanindividual'scountryoforigin.
Whendoingso,wefindthat,bothinthepastandtoday,immigrantsatthebottomoftheincomedistributionfromnearlyeverysendingcountryhavehigherratesofupwardmobilitythanthechildrenoftheUS-born.
Moreover,despiteallthechangesthathaveoccurredoverthelastcentury—morerestrictedborders,ashifttopoorersendingcountriesrelativetotheUS,thegrowingimportanceofservicesrelativetomanufacturing—childrenofimmigrantstodayachievestrikinglysimilarratesofupwardmobilitytothoseofthepast.
Forexample,childrenofpoorMexicanandDominicanimmigrantstodayhaveasimilaradvantagerelativetothechildrenoftheUS-bornasdidthechildrenofpoorDanishorSwissimmigrantsin1910.
ThisfindingsuggeststhatthedescendantsofimmigrantsofallbackgroundscaneventuallyintegrateintotheUSeconomy.
Inthelastpartofthepaper,weexplorethequestionofwhychildrenofimmigrantsaremoreupwardlymobile,focusingprimarilyonthehistoricaldatabutcomparingwiththemoderndatawheneverpossible.
First,geographymatters.
Bothtodayandinthepast,wefindthatimmigrantparentsweremorelikelythanUS-bornparentstomovetoareasofferingbetterprospectsfortheir4children.
Inthehistoricalcohorts,theintergenerationalgapbetweenimmigrantsandtheUS-bornisreducedbyabout50percentwhencomparingchildrengrowingupinthesameUSregion(Censusdivision)anddisappearsentirelywhencomparingchildrengrowingupinthesamecounty.
Inthemoderncohort,theimmigrantmobilityadvantageis25percentsmallerintheaveragecountythanatthenationallevel.
Finally,wediscusstheroleofimmigrants'self-selection,andprovidesuggestiveevidencethatimmigrantfatherswere"under-placed"intheincomedistribution(thatis,theirearningsdidnotfullyreflecttheirabilities),therebygivingtheirchildrenmoreroomtoimprove.
2.
LiteratureOurpapercontributestotheunderstandingofimmigrantassimilationandintergenerationalmobilityintheUS.
ClosesttoourpaperisCard,DiNardo,andEstes(2000),whichusescross-sectionaldatatocompareaverageoutcomesbycountryoforiginforfirst-andsecond-generationimmigrantsinthelabormarketfrom1940untilthemid-1990s.
Likethispaper,wealsofindthatthechildrenofimmigrantsfarebetteronaveragethanthechildrenoftheUS-bornconditionalontheaverageoutcomesofthefirstgeneration.
Ourpaperbroadenstheanalysistostudyimmigrantsandtheirchildrenovermorethanacentury(1880to2015).
Moreimportantly,ourpaperaddsactualparent-childlinkages,enablingustoassesstheviewthatevenpoorimmigrantscansubstantiallyimprovetheirchildren'sprospects,aswellasexplorethechannelsthroughwhichimmigrantfamiliesachievetheirmobilityadvantage.
MostexistingworkoneconomicoutcomesduringtheAgeofMassMigrationfocusesonassimilationwithinageneration(seeforinstance,Abramitzky,Boustan,andEriksson2014;Ferrie1997;Hatton1997;andMinns2000).
OneexceptionisWard(2019),whodocumentspersistentoutcomesacrossgenerationsbyancestry(nationaloriginofgrandfather)intheearly20thcentury.
WegobeyondWardbycontrastingmobilityratesforthechildrenofimmigrantswiththoseoftheUS-born,andbycomparingmobilityratesinboththepastandpresent.
66Borjas(1993)usescohort-leveldatafromthe1940and1970CensusestostudytheintergenerationalmobilityofimmigrantsintheUS.
PerlmannandWaldinger(1997)andPerlmann(2005)comparetheoutcomesofsecond-generationimmigrantsinthepastandthepresentbutdonothaveparent-childlinkages.
Ourworkalsorelatestostudiesthatfocusontheintergenerationalprogressofspecificimmigrantgroups,includingtheIrishfamineimmigrants(CollinsandZimran2019),andMexicanAmericans(KosackandWard2018;Duncanetal.
2017).
5Finally,ourpaperaddstothebroaderliteratureonintergenerationalmobilityintheUnitedStates.
Anumberofpapersestimatecontemporarylevelsofintergenerationalmobility(e.
g.
,Chettyetal.
2014,2017;DavisandMazumder2019;Hilger2016;LeeandSolon2009;Mazumder2015).
7Aseriesofrelatedstudiesdocumenthistoricalratesofintergenerationalmobility,andcomparethesehistoricalratestopresent-daylevels(e.
g.
,Card,Domnisoru,andTaylor2018;Derenoncourt2019;Feigenbaum2015,2018;Ferrieetal.
2016;LongandFerrie2013;OlivettiandPaserman2015;Olivetti,Paserman,andSalisbury2018;Tan2018;Pérez2019).
Borjas(1992)developsatheoryofethniccapital,whichsuggeststhatthesocialmobilityprocessmaydifferbetweenimmigrantsandtheUS-born.
83.
DataA.
HistoricalDatasetsforStudyingIntergenerationalMobilityWemeasurehistoricalmobilityratesusingtwonewdatasetsoflinkedCensusrecords.
Thefirstdatasetlinkssonsobservedinthe1880Censustothe1910Census,andthesecondlinkssonsobservedinthe1910Censustothe1940Census.
9Thesedataallowustoobserveanindividual'sownlabormarketoutcomesduringadulthoodandhisfather'slabormarketoutcomesduringhischildhood.
WomencannotbesystematicallymatchedacrosshistoricalCensusesbecausetheytypicallychangetheirlastnamesaftermarriage,sothepast-presentcomparisonfocusesonmen.
Becausethemoderndatadefineimmigrantstatusbasedonfather'sbirthplace,weusethesamedefinitioninthehistoricaldataforcomparability.
10i.
LinkingFathersandSons7BhattacharyaandMazumder(2011),Chettyetal.
(2018a)andCollinsandWanamaker(2017)focusonblack-whitedifferencesinmobility,andHilger(2017)focusesonAsianAmericans.
8OutsideoftheUS,Aydemir,Chen,andCorak(2009)focusontheintergenerationalmobilityofimmigrantstoCanada,BolotnyyandBratu(2018)onimmigrantstoSweden,andGielenandWebbink(2019)ontheNetherlands.
9Forreasonsofbothdataavailabilityandhistoricalinterest,ouranalysisdoesnotincludeimmigrantsinthe1940–1970cohorts.
First,the1940CensusisthelastCensusforwhichinformationonnamesispubliclyavailable,allowingforbackwardbutnotforwardlinksfrom1940.
Second,neitherofthesourcesthatweuseforthemodernperiodincludeimmigrantsinthesecohorts:theGSSstartedcollectingdatain1972andtheOpportunityInsightsdatacoverschildrenofimmigrantsbornafter1978.
Inaddition,thesecohortscoincidewithaperiodoflimitedmigrationintotheUS:in1970,only4.
7percentoftheUSpopulationwasbornabroad(comparedtonearly15percentinthepast[1880–1920]andtoday).
10AppendixAandtheOnlineAppendixintheNBERworkingversionofthispapershowthattheresultsinboththepastandtodayaresimilarwhendefiningimmigrantstatususingthemother'splaceofbirth(orusingbothparents'birthplacesinthehistoricaldatawhensuchinformationisavailable).
6Tocreateeachlinkedsample,wematchedallmalesaged0–16inachildhoodCensus(thatis,eitherin1880orin1910)toalaterCensus(1910or1940,respectively)usinginformationonfirstandlastnames,age,andstateofbirth.
WeusethelinkingalgorithmdevelopedinAbramitzky,Boustan,andEriksson(2012,2014),whichisexplainedandevaluatedinAbramitzkyetal.
(2019).
Werestricttheanalysistofather-sonpairsinwhich:(1)boththefatherandthesonwerewhite,(2)thesonwaslivingwithhisfatheratthetimeoftheearliestCensus(sothatwecanobserveafather'slabormarketoutcomes),11(3)thefatherwasbornintheUSorinoneofthe17largestsendingcountriesduringthistimeperiod,12and(4)boththefatherandthesonwereaged30–50bythetimewemeasuretheirlabormarketoutcomes.
13Finally,weexcludefather-sonpairswithmissinginformationonoccupation.
14AppendixAprovidesfurtherdetailsonthematchingprocedureaswellassensitivitychecksforpossible"falsepositives"andlackofrepresentativeness.
15AppendixTableA1showsthesamplesizeineachofthehistoricalcohortsasweimposetheserestrictions.
Inourbaselinesamples,wematch23percentofindividualsinthe1880Censusto1910and29percentofindividualsinthe1910Censusto1940,standardmatchratesforhistoricalsamplesgiventhepresenceofcommonnamesandvariouscausesofnon-matches(mortality,returnmigration,under-enumeration,transcriptionerror;seeAbramitzkyetal.
2019).
Afterimposingthesamplerestrictions,ourbaselinesamplesinclude1.
3millionmeninthe1880–1910(30percentofwhomaresecond-generationimmigrants)and2.
7millionmenin1910–1940(24percentofwhomaresecond-generationimmigrants).
ThelargestorigincountriesareGermany,Ireland,andEnglandinthefirstcohort,andGermany,Russia,andCanadainthesecond.
11Mostchildreninthisagegroupwerelivingwiththeirfatherregardlessofimmigrationstatus:88and90percentamongchildrenwithaUS-bornfatherinthe1880and1910Census,respectively,and89and90percentofchildrenwithaforeign-bornfather.
12ThecountriesweconsiderareAustria,Belgium,Canada,Denmark,England,Finland,France,Germany,Ireland,Italy,Norway,Portugal,Russia,Scotland,Sweden,Switzerland,andWales.
Thesecountriesarethelargestsourceofimmigrantsinthe1850–1910period,accountingfor93and88percentoftheforeign-bornpopulationinthe1880and1910Censuses,respectively.
Werestricttheanalysistolargesendingcountriessothatwehaveenoughobservationstoestimateourcountry-by-countryresults.
Althoughwerefertothechildrenofimmigrantsas"thesecondgeneration,"wedonotdistinguishbetweensonsborninsideandoutsideoftheUS,becausetheOpportunityInsightDatadoesnotincludesuchinformation(thatis,weincludeboththe"1.
5generation"andthe"secondgeneration"inoursamplesofchildren).
ThevastmajorityofsonsinourdataarebornintheUS(98percentinthehistoricalcohortsand95percentintheGSSmoderncohort).
13Werestrictoursampletomenages30–50inordertomeasurefathersandsonsatthesamemomentintheearningslifecycle,andtoavoidmeasuringincomewhenchildrenorparentsaretooyoungorold(Mazumder2005).
14Ourfindingsaresimilarwhenweassignanincomeofzerotothosewithmissingoccupation(AppendixFigureB2).
15IntheOnlineAppendixoftheNBERversionofthepaperweshowthattheresultsaresimilarwhenweconstructthehistoricalsamplestomatchthecharacteristicsoftheOpportunityInsightssample.
7ii.
AssigningIncomeMeasuresinHistoricalDataBecausethe1940CensusisthefirstUSCensustoincludeinformationonindividualearnings,weneedtoconstructproxiesofindividualincomeforfathersandsonsinourpre-1940historicalsamples(werefertotheseproxiesas"incomescores").
16Ourpreferredapproachistouseastatisticalmodeltopredictincomefromarichsetofcovariatesforwhitemenaged30–50inthe1940Census,andthenusethismodeltopredictincomeformeninearlieryears(wediscussfarmersbelow,whomwehandleseparatelybecausethereislimitedinformationonfarmincomeinthe1940Census).
Inparticular,weregresslogincomein1940onanumberoffixedeffectsandcompletesetofinteractiontermsusing3-digitoccupation,age,andcurrentstateofresidenceasourexplanatoryvariables.
17The1940incomevariableexcludesincomefromself-employment.
Becausethevastmajorityoffarmersareself-employed,wecomputeincomeforfarmersusingamethoddevelopedbyCollinsandWanamaker(2017).
Specifically,wemakeuseofthefactsthatthe1940Censusrecordstheincomesoffarmlaborers,andthatlaterCensusyearsrecordhowmuchfarmersearnrelativetofarmlaborers.
Wethuscomputefarmerincomesbymultiplyingtheincomeoffarmlaborersin1940withtheratioofearningsforfarmersversusfarmlaborersinthe1960Census,byregionandimmigrationstatus.
18Onelimitationofusingthe1940CensustocomputeincomescoresforindividualsinearlierCensusesisthatweassumethattherelativerankbetweenincomecellsisstablefrom1880to1940.
AppendixBshowsthatresultsarerobusttousingmorecontemporaneoussourcestocompute16Inourbaselineanalysis,wealsousepredicted(ratherthanactual)incomeforthesonsweobserveinthe1940Census(soastouseaconsistentmeasureacrossallofourhistoricalsamples).
However,theresultsaresimilarwhenweinsteaduseindividualincomedatafromthe1940Census(seeAppendixFigureB2).
17InAppendixB,weshowthatresultsaresimilarwhenweuseaperson'scountryofbirthasanadditionalvariableintheincomeprediction.
Inallinteractionterms,weinteractcovariateswith1-digitCensusoccupationsandwithCensusregion,insteadofstate.
Thismethodissimilartothemachine-learningapproachforcomputingincomescoresproposedbySaavedraandTwinam(2018).
Indeed,thecorrelationbetweenourpredictedearningsandthoseestimatedusingtheirapproachis0.
99forfathersinthe1910–1940cohort.
18Whenperformingthiscalculation,wefocusonUS-bornwhitesandimmigrantsfromthesendingcountriesweincludeinthepaper.
WenotethatCollinsandWanamakerusethe1960Census(ratherthan1950)becausethe1950Censusonlyaskedtheincomequestiontoasmallsubsetofthepopulationandisthereforenotusefulforthisexercise.
Inadditiontoadjustingforfarmerincome,wealsofollowCollinsandWanamaker(2017)andscaleupfarmmanagers'andfarmlaborers'incomestoaccountforin-kindtransfers.
Finally,whenusingactualratherthanpredictedincomeforthesonsweobserveinthe1940Census(secondrowinpanel(b)ofAppendixFigureB2),wealsoadjusttheincomeofself-employednon-farmersonsfollowingCollinsandWanamaker(2017).
Todoso,wecalculatetheratioofmeanearningsforself-employednon-wageworkersinanoccupationtothemeanearningsofwageworkersinthesameoccupationfromthe1960Censusandthenusethisratiotoscalewageearningsoftheself-employedinthe1940Census.
8incomescores,namelythe1901CostofLivingSurveyandthe1900CensusofAgriculture.
Theappendixalsodocumentsthatimmigrantshaveasimilarmobilityadvantagewhenusingasampleoffather-sonpairsforwhomwecanobservefathersinthe1940Censusandthuscanmeasureactual,ratherthanonlypredicted,incomeforbothfatherandson.
19Overall,ourconclusionsareunaffectedbythespecificincomeproxiesweuse.
20AppendixBprovidesfurtherdetailsontheconstructionoftheincomescoresandsensitivitychecks.
B.
ModernDatasetsforStudyingIntergenerationalMobilityOurlinkedCensusdatacannotbeextendedbeyond1940becauseconfidentialityrestrictionsdictatethatindividuals'namesareonlyreleasedtothepublic72yearsafteraCensusistaken.
Instead,werelyontwoothersourcesofdatatofollowfather-sonpairsinthemodernlabormarket.
First,weusepubliclyavailableaggregatedatafromtheOpportunityInsightsproject(Chettyetal.
2018a).
Thesedataarebasedonnearlysixmillionchildrenbornbetween1978and1983,outofwhom310,000haveanimmigrantfatherborninoneofthe21largestsendingcountriestotheUS.
21ThesedataarebuiltbytheOpportunityInsightsteamintwosteps:Inthefirststep,the2000USCensusandthe2005–2015AmericanCommunitySurvey(ACS)arelinkedtofederalincometaxreturns(usingapersonidentifierassignedbytheCensusBureau).
Thisprocedureenablestheresearcherstoobserveaperson'sincomefromthetaxrecordsaswellashisorherraceandplaceofbirthfromtheCensusortheACS.
Inthesecondstep,theOpportunityInsightsteamobtainedinformationonthefamilyincomeofanindividual'sparentsduringhis/herchildhood.
Aparentisdefinedasthefirstpersonwhoclaimsanindividualasadependentinthefederalincometaxrecordsbetweentheyears1994and2015.
NotethatthisprocedureforlinkingparentstochildrenwillexcludepairsinwhicheitherthechildortheparentslackaSocialSecurityNumber.
Asa19WedosoinAppendixBbyconsideringthefatherswhoareyoungenoughin1910thattheycanbelinkedtothe1940Census.
20TheoneexceptiontothisclaimisthesetofresultsbasedontheIPUMSoccupationalincomescore,whichmapsoccupationstomedianincomeinthe1950Census.
Relianceonthe1950incomedistributionisproblematicbecausetheincomesoffarmersdeterioratedbetween1880and1950duetosecularchangesintheagriculturalsector,leadingfarmincomederivedfrom1950datatobetoolowforfathersin1880or1910.
FigureB2inAppendixBshowsthatoncefarmers'incomesintheIPUMSoccupationalscoreareadjustedsothatfarmershavethesameaveragerankasinourbaselinescore,weagainseealargeimmigrantmobilityadvantage.
21Weusethe21countriesforwhichtheOpportunityInsightsprojectreportschildrenoutcomesbyventileoftheparentalincomedistribution.
ThecountriesareCanada,China,Colombia,Cuba,theDominicanRepublic,Ecuador,ElSalvador,Germany,Greece,Guatemala,Haiti,India,Israel,Italy,Jamaica,Japan,Mexico,thePhilippines,SouthKorea,theUK,andVietnam.
9consequence,thissampleonlyincludeschildrenwhoareeitherUScitizensorauthorizedimmigrants,andwhoseparentsarealsoUScitizensorauthorizedimmigrants.
Incomeforthiscohortismeasuredastheaverageannualincomefrom2014to2015,whenthesechildrenwere31to37yearsold,whereasparentalincomeismeasuredastheaveragehouseholdincomefrom1994to2000(father'sincomeisnotreportedseparatelyinthisdataset).
Weusethepubliclyavailableversionofthedata,whichreportsinformationcollapsedbyventilebininthenationalincomedistributionforimmigrantsfrom21countriesoforigin.
Hence,foreachparentalranking,weobservetheaverageincomerankofchildrenwhoseparentsfallintothatventile,byafather'scountryoforigin.
22Whenfocusingoncountry-by-countryoutcomesatthe25thand75thpercentiles(ratherthanonoutcomesthroughoutthefullparentalincomedistribution),weareabletoexpandouranalysisto51countriesoforigin.
TheOpportunityInsightsdatasethasseveralattractivefeaturesforthestudyoftheintergenerationalmobilityofimmigrants:Itisverylargerelativetootherpaneldatasets(nootherlongitudinaldatasetwouldenableustolookatmobilitydifferencesbycountryoforigininthemodernperiod),andincludesadministrativeincomedatabothforparentsandtheirchildren.
23However,thedatasethastwolimitationsforourpurposes.
First,itdoesnotincludeinformationonindividual-leveloccupations—themainlabormarketoutcomeweobserveinthehistoricaldata—thuscomplicatingourpast-presentcomparisonsofintergenerationalmobility.
Second,becauseunauthorizedimmigrantsdonothaveaSocialSecurityNumber(SSN),thesedataexcludethechildrenofunauthorizedimmigrants.
ThislimitationaffectsHispanicimmigrantsmorethanothergroups.
Specifically,Chettyetal.
(2018a)comparestheOpportunityInsightsdatatotheAmericanCommunitySurveyandreportsthattheyincludeabout79percentofHispanics,whereasthecoverageiscloseto100percentforotherethnicgroups.
WeadddatafromtheGeneralSocialSurvey(describedbelow)aswellascross-sectionaldatafromtheCurrentPopulationSurvey,bothofwhichincludeunauthorizedimmigrants,topartiallyaddressthisconcern.
WealsonotethattheOpportunityInsightsdatadoincludefirst-generationmigrants(andtheirchildren)whowereat22Thesedatacorrespondtothe"Non-ParametricEstimatesofIncomeRanksforSecondGenerationImmigrantChildrenbyParentIncome,CountryofOrigin,andGender"and"ParametricEstimatesofIncomeRanksforSecondGenerationImmigrantChildrenbyParentIncome,CountryofOrigin,andGender"filesavailableontheOpportunityInsightswebsiteathttps://opportunityinsights.
org/data/.
23Forinstance,thePSID(whichincludesinformationonparentalincome)startedin1968andthusdoesnothavemuchcoverageofimmigrantfamilieswhoarrivedintheUSaftertheborderreopenedtoimmigrationin1965.
Otherpaneldatasets(forinstance,theNLSYusedinBorjas1993)includeretrospectivequestions(askedofthechildren)abouttheoccupationandeducationintheparent'sgeneration,butdonotcontaininformationonparentalincome.
10somepointundocumented,butwholateronbecameauthorizedandhencereceivedaSocialSecurityNumber.
Forexample,thedataincludefirst-generationimmigrantswhowereinthecountryby1982andbenefitedfromthe1986ImmigrationReformandControlActamnestyof2.
7millionpreviouslyunauthorizedimmigrants,anumberthatrepresentsnearlyalloftheundocumentedimmigrantslivingintheUSatthetime(Passel,1986).
24WesupplementouranalysisoftheOpportunityInsightsdatawithdatafromtheGeneralSocialSurvey(GSS).
AlthoughtheGSShassubstantiallysmallersamplesizes,itincludesquestionsonoccupationandcoversunauthorizedimmigrants,bothofwhichfacilitatecomparisonwiththehistoricaldata.
Thesurveyincludesinformationonarespondent'soccupation,parents'placesofbirth(intheUSorabroad),andretrospectivequestionsonparents'occupationswhentherespondentwas16yearsold.
Weusethisinformationtoassignincomeproxiesforfathersandtheirsonsusingananalogousproceduretotheoneweimplementwiththehistoricaldata.
25Tomakethissampleascomparableaspossibletothehistoricaldata,wefocusonrespondentswho:(1)aremales,(2)livedintheUSbyage16,(3)were30to50yearsoldatthetimeofthesurvey,and(4)haveavailableinformationontheirandtheirfathers'occupations.
Finally,tomaximizecomparabilitywiththeOpportunityInsightssamplewhileretainingenoughobservations,werestricttheanalysistosurveyyears2000to2018.
Afterimposingthesesamplerestrictions,ourGSSsampleincludes3,100respondents,ninepercentofwhomaresecond-generationimmigrants.
4.
PreliminaryEvidence:EarningGapsbetweenImmigrantsandUS-bornsince1880WestartbyestimatingearninggapsbetweenUS-bornmenandfirst-andsecond-generationimmigrantsbycountryoforigin.
Forthisexercise,weusecross-sectionalsamples,ratherthan24Conceptually,wemightexpectthechildrenofundocumentedimmigrantstobeevenmoreupwardlymobilethanthechildrenofauthorizedmigrants.
Aswedocumentbelow,lackofUS-specifichumancapitalamongfirst-generationimmigrantspartlyexplainsthehighmobilityamongtheirchildren(i.
e.
,first-generationimmigrantsare"under-placed"intheUSincomedistribution).
Giventherestrictionsthatunauthorizedimmigrantsfaceinthelabormarket,wemightexpectthemtobeevenmore"under-placed"intheUSlabormarketrelativetotheirtrueabilities,makingtheirchildren(whowilloftenbebornintheUSandhavecitizenship)especiallylikelytobeupwardlymobile.
25Weusethe2006–2015CPStocreateincomeproxiesforfathersintheGSSsample.
Todoso,weregressloggedtotalearnedincomeonanumberoffixedeffectsandcompletesetofinteractiontermsusingoccupation,age,region,andimmigrantstatus(andtheirinteractionterms)asourexplanatoryvariables.
WethenusethisregressiontopredictincomescoresforGSSrespondentsandtheirfathers.
WenotethattheGSSreportsparentalimmigrantstatusbutdoesnotrecordcountryoforiginandthatthemostprecisemeasureofgeographyintheGSSisregion.
Inaddition,theGSSdoesnotincludeinformationonparentalage.
11father-sonpairs.
26Ourinterestisindocumentingtheaverageearningsdisadvantagefacedbyfirst-generationimmigrantsbycountryoforigin,andthentoaskwhethertheaveragechildoftheseimmigrantswasabletoerasesomeorallofthesegapsbythenextgeneration.
Specifically,weestimatethefollowingequation,firstforthefather'sgenerationandthenfortheson's:(1)$#%&Countrycwhere!
!
"#istheloggedincomescoreoffatherorson*ofage+fromsendingcountry,.
The&#coefficientsrevealtheaveragedifferenceinincomescorebetweenUS-bornindividualsandimmigrantsfromcountry,.
Allregressionsincludeaquadraticinfather'sorson'sage(".
Figure1plotsthe&#coefficientsforfirst-generation(inblack)andsecond-generation(ingray)immigrantsineachcohort.
Formostcountriesoforiginandtimeperiods,thegapsinincomescoresbetweenfirst-generationimmigrantsandUS-bornmenrangebetween-40logpointsand+20logpoints.
Inthemoderncohorts,fourcountriesfacedlargergapsofabout-60logpoints(theDominicanRepublic,Haiti,Mexico,andVietnam).
Inthepast,immigrantsfromScandinaviancountriesfacedthelargestearningspenalties.
Inallcohorts,immigrantsfromasetofsendingcountriesoutperformtheUS-borninthelabormarket.
Today,thesehigh-earningcountriesincludeCanada,Germany,India,Japan,andtheUnitedKingdom,whereasinthepastthehighest-earningimmigrantswerefromcountriessuchasEnglandandScotland.
Ineachofthethreecohorts,second-generationimmigrantsfromcountriesinwhichfirst-generationimmigrantsearnedlessthantheUS-borncloseorevenreversethecorrespondingearningsgapinthesecondgeneration,leadingchildrenofimmigrantstoout-earnchildrenofUS-bornindividuals.
Thereareonlyafewcasesinwhichthegapremainedabove5logpointsbythesecondgeneration(Norwegiansin1910and1940,Finnsin1940,andJamaicans,Haitians,and26Forourhistoricalcohorts,weusethe1880,1910,and1940CensusesofPopulation.
Forthemoderncohort,weusethe1980Censusforthefathers'generationandtheCPS'sAnnualSocialandEconomicSupplementfor2006–2015forthesons(Floodetal.
2018,Rugglesetal.
2018).
Werestrictthe1980Censussampletomenaged30–50,whohaveasonpresentinthehousehold,andwhowerebornintheUSorinoneofthetopsendingcountriesidentifiedintheOpportunityInsightssample.
WerestricttheCPSsampletoUS-bornmenaged30–50andwhosefatherswereeitherbornintheUSorinoneofthetopsendingcountries.
Forbothofthesedatasets,weconstructoccupation-basedincomescoresusingananalogousproceduretotheonethatweuseinboththeGSSandthehistoricaldata.
IntheappendixtotheNBERversionofthepaper,weshowthattheresultsaresimilarwhenusingactualincomewheneverpossible(FigureA1)andwhenexcludingtheUSSouthfromthehistoricalsamples(FigureA4).
12Mexicansinthemoderndata),butformostsendingcountries,theearningsgapissubstantiallysmallerbythesecondgeneration.
Forcountrieswhosefirst-generationimmigrantsalreadyout-earnedUS-bornindividualsinthelabormarket,thepatternismoremixed,withthecorrespondingearningsadvantagebecomingsmallerforsomecountries,remainingsimilarinsizeforothers,orevengrowingbythesecondgenerationinsomecases.
AlthoughearningsgapsforcountriesthatstartbelowtheUS-borndiminishsubstantiallyacrossgenerations,itisstillthecasethatthechildrenoflow-earningimmigrantstendtoearnlessthanthechildrenofhigh-earningimmigrants,consistentwiththepersistenceresultsinAbramitzky,Boustan,andEriksson(2014)andWard(Forthcoming).
5.
MainResults:IntergenerationalMobilityofImmigrantssince1880Wehavesofarconsideredaverageearningsgaps,withoutanyregardfortheparents'rankintheincomedistribution.
ThehigherearningsgrowthforthechildrenofimmigrantsrelativetothechildrenofUS-bornparentsinFigure1couldbedrivenbyhigherratesofupwardmobilityfromthebottomoftheincomedistribution,lowerratesofdownwardmobilityfromthetopoftheincomedistribution,orboth.
TheessenceoftheAmericanDreamisthatevenimmigrantsatthebottomoftheincomedistributioncanhavechildrenthatsucceed.
Thus,weturntothelinkeddata,bothhistoricalandmodern,tostudytheintergenerationalmobilityofchildrenofimmigrantsandUS-bornindividualswhoseparentshadcomparablelabormarketoutcomes,asmeasuredbytheirrankinthenationalincomedistribution(followingChettyetal.
2014andChettyetal.
2018a).
Todoso,wefirstrankeachsonbasedonhisincomescore(oractualincomewhenusingtheOpportunityInsightsdata),relativetoothersonsborninthesameyear.
Similarly,werankfathersrelativetoallotherfatherswithsonsborninthesameyear.
27Bothwhenrankingsonsandwhenrankingfathers,theranksarebasedonaperson'spositioninthenationalincomedistribution.
28Wethenestimateamodelinwhichweregressason'srankonhisfather'srank,allowingboththeslopeandintercepttodifferforsonsofimmigrantsandUS-bornindividuals:27IntheOpportunityInsightsdata,weusefamilyincomerankbecausewedonothaveseparateinformationonafather'sincomerank.
28Ifthereareindividualsinthesamplewhohavethesamefather,weassignthefathertheaveragerankacrosshissons.
DuetotheGSS'smuchmorelimitedsamplesizecomparedtothehistoricallinkeddataandtheOpportunityInsightsdata,wepoolallsurveyyears(2000to2018)andincludesurvey-yearfixedeffectsintheregressions.
13(2)-+.
/01.
2.
345.
7+85:+&(2.
345.
7+85:+).
Theconstantterm#measuresabsoluterankmobilityforthechildrenoftheUS-born,thatis,theexpectedrankofchildrenofUS-bornindividualswithfathersattheverybottomoftheincomedistribution.
Theparameter&'capturesthedegreetowhichthisexpectedrankisdifferentforthechildrenofimmigrants.
Theparameter&&measurestherateofrelativemobilityofchildrenofUS-bornindividuals,ortheassociationbetweentheranksofchildrenandthoseoftheirfathers.
Finally,&(measuresthedegreetowhichthisassociationisdifferentforimmigrantsandtheirchildren.
WecomparetheeconomicmobilityofthesonsofimmigrantstothesonsoftheUS-borninFigure2.
Todoso,weplottheregressionlinescorrespondingtoequation(2),separatelyforeachofourcohorts,aswellasabinnedscatterplotshowingthemeanincomerankofthechildrenofimmigrantsandoftheUS-bornbyparentalincomeventilerank.
Panels(a)and(b)suggestthatinthepast,thesonsofimmigrantsonaverageout-earnedcomparablesonsofUS-bornindividualsthroughouttheincomedistribution,andparticularlyforsonsoffatherswithbelow-medianrank.
Second-generationimmigrantsonswithfathersatthe25thpercentilein1880and1910reachedanaverageincomerankthatwas6percentilesabovetheaveragerankreachedbysonsoftheUS-bornwithsimilarfathers'incomes.
29Panels(c)and(d)showthemobilitypatternsforthemoderncohortwhenusingtheGSSandOpportunityInsightsdata,respectively.
Today,thesonsofimmigrantsremainmoreupwardlymobilethantheircounterpartswithUS-bornparents,andthispatternalsomainlyholdsinthebottomhalfoftheincomedistribution.
Inparticular,andsimilartothehistoricalresults,sonsofimmigrantsinthe25thpercentilerankabout5percentilepointshigherthansonsofUS-bornindividuals,regardlessofwhetherweusetheGSSortheOpportunityInsightsdata.
Table1reportstheestimatedvaluesofthecoefficientscorrespondingtoequation(2)foreachofthesamples.
Forallthreecohorts,wefindahigherintercept(apositive&')forthesonsofimmigrantsrelativetothechildrenoftheUS-born,indicatinghigherlevelsofabsolutemobility.
Thedifferenceininterceptsrangesfrom7to8percentilesinallcohorts.
Theslopesforimmigrant29Wealsofindthatthesonsofunskilledimmigrantfathersmovedintowhite-collarandskilledblue-collarjobsathigherratesthanthesonsofunskilledUS-bornfathers,whoweremorelikelytotransitionintofarming(seetheOnlineAppendixoftheNBERversionofthepaper).
14familiesareinallcasessmallerthantheslopesfortheUS-born(anegative&(),suggestingaweakerassociationbetweenafather'sandhisson'srankamongimmigrantfamilies.
InFigure3,weallowtheinterceptandtheslopeoftherank-rankrelationshiptodifferbycountryoforigin,bothinthehistoricalandthemoderndata.
Wethenusetheseestimatestocomputetheexpectedincomerankforsonsbornatthe25thpercentileforeachcountryoforigin.
30Thefigureshowsthatsonsfromnearlyallsendingcountriesinthepastandthepresent(exceptNorwayandBelgiumin1880,Norwayin1910,andHaiti,TrinidadandTobago,andJamaicainthemoderncohort)hadahigherexpectedrankthanthechildrenofUS-bornindividualswithparentsatthe25thpercentile.
31ImmigrantsfromAsiancountries,notablyChinaandIndia,appearespeciallyupwardlymobileinthecurrentperiod.
32Figure4showsthatthesonsofimmigrantsandtheUS-bornappearmoresimilarwhenlookingatthosewhogrewupwithparentsinthe75thpercentileoftheincomedistribution,particularlyinthemodernperiod.
Inthepast,sonsofimmigrantshadslightlyhigherexpectedranks,andtodaythetwogroupsachievesimilarexpectedranks.
Atthetopoftheincomedistribution,sonsofimmigrantsfromsomesendingcountrieslikePortugalandBelgiuminthepastandJamaica,ElSalvador,andMexicotodaydidworsethanthesonsoftheUS-born,whereasthosefromothercountrieslikeRussiaandIrelandinthepastandIndiaandChinatodaydidbetter.
Themoderndataalsoenableustolookattheintergenerationalmobilityofthedaughtersofimmigrants.
Panel(a)ofFigure5suggeststhatdaughtersofimmigrantsonaverageout-earned30Inthisexercise,wefocusontheOpportunityInsightsforthemoderndatabecausetheGSSdatadoesnotrecordthefather'scountryoforigin(onlywhetherthefatherisforeign-born).
31OnenoticeablepatternfromFigure3isthattheonlythreecountrieswheresonsofimmigrantsatthe25thpercentilehavealowerexpectedrankthanthesonsoftheUS-bornarecountrieswhereasignificantshareofthepopulationisblack.
Thispatternisconsistentwithabroaderpatternofblack-whitemobilitygapsintheUS(seeforexampleChettyetal.
2018a).
InFigure5oftheNBERversionofthispaper,wealsoshowmobilitypatternsbyimmigrantraceandethnicity.
Incontrast,whenfocusingondaughters,wenolongerobservethatthechildrenofimmigrantsfromCaribbeancountrieswithasignificantblackpopulationhavealowerexpectedrankthanthechildrenoftheUS-born(seeFigure5inthispaper).
32Whilethispapercannotexplainwhyimmigrantsfromsomecountriesaremoremobilethanothers,itisnaturaltoaskiffactorssuchaslinguistic,physical,andreligiousdistancefromtheUSplayaroleinimmigrants'mobility.
Foreachofthecountriesinourdata,wecalculated:(1)linguisticdistancetotheUSusingthemeasurefromChiswickandMiller(2005),(2)physicaldistancetotheUS,and(3)anindicatorthattakesavalueofoneforpredominantlyProtestantcountries.
Inthemoderndata,thereisaclearpositivecorrelationbetweenalldistancemeasuresandmobility.
ThiscorrelationislargelydrivenbyAsiancountries,whicharebothverydistancedinallmeasuresfromtheUSandwhoseimmigrantsareveryupwardlymobile.
Incontrast,inthepastthereislittlecorrelationbetweenthevariousdistancemeasuresandmobility.
Thisismainlybecauseinthepastallsendingcountries(besidesCanada)wereinEurope,sothereisverylittlevariationintheirphysicaldistancetotheUS.
ThemainvariationinlinguisticdistanceinthepastisdrivenbywhetherthecountryisEnglishspeaking,butimmigrantsfromEnglish-speakingcountriesdidnotseemtobemoreorlessupwardlymobilethantheaverageimmigrant.
15daughtersofUS-bornindividualsthroughouttheincomedistribution.
Panel(b)showstheexpectedincomerankfordaughtersbornatthe25thpercentileforeachcountryoforigin.
Asforsons,wefindthatdaughtersofimmigrantsfromnearlyeverysendingcountryachievehigherlevelsofupwardmobilitythanthedaughtersoftheUS-born.
Overall,weconcludethatbothinthepastandtoday,second-generationimmigrantsaremoreupwardlymobilethanthechildrenoftheUS-born.
Thisadvantageisnotdrivenbychildrenofimmigrantsfromcertainethnicitiesorcountry-of-origingroups:thereishigherupwardmobilityamongchildrenofimmigrantsfromnearlyeverysendingcountry,andparticularlyatthebottomoftheincomedistribution.
6.
WhyarechildrenofimmigrantsmoreupwardlymobileA.
LocationChoicesintheFirstGenerationFigure6documentsthatbothhistoricallyandtoday,immigrantstendtosettleinareasthatofferhigherlevelsofupwardmobilityfortheirchildren.
Specifically,thecorrelationbetweentheshareofmigrantsinanareaandtheshareofupwardlymobilesonsis0.
36forthe1880cohort,0.
51forthe1910cohort,and0.
23forthemostrecentcohort(upwardmobilityisdefinedhereastheshareofsonsofUS-bornfathersinthebottomfifthofthenationalincomedistributionwhoreachthetopfifthofthedistribution).
33Immigrantsinthemoderncohortcontinuetosettleinplacesthatofferstrongprospectsforupwardmobility,buttherearealsoanumberofhigh-mobilitylocations(e.
g.
,WyomingandUtah)thatdonotattractmanyimmigrantstoday.
WenextillustratemoreexplicitlythatdifferencesinthegeographicdistributionofimmigrantsandUS-bornindividualsmightexplainthemobilitygapbetweenthetwogroups.
Todoso,were-runourrank-basedspecificationfromequation(2),andsequentiallyaddfixedeffectsfortheUSSouth,Censusdivision,state,state-by-urbanstatus,andcounty,allofwhicharebasedonanindividual'schildhoodlocation(weonlyhavedatatorunthesespecificationsforthehistoricalcohorts).
RegressionswithoutlocationfixedeffectsrevealthemobilitygapbetweenchildrenofimmigrantsandchildrenoftheUS-bornatthenationallevel.
Regressionswithlocationfixedeffectsmeasurethegapbetweenchildrenwhogrewupinthesamelocation.
Ifimmigrantfathers33Ratesofintergenerationalmobilitybothhistoricallyandtodayvarygreatlybyregion(Chettyetal.
2017,Tan2018,Connor2019).
Previousresearchhasalsofoundthatfirst-generationimmigrantsinthemodernperiodaremorelikelythantheUS-borntosettleinregionsthatofferthemhigherwages(Borjas2001,CadenaandKovak2016).
16chosetosettleinareasthatofferedhighprospectsformobilityoverall,weexpectthegaptobesmallerintheregressionswithlocationfixedeffects.
Figure7showsthatindeedincludingfixedeffectsfortheUSSouthorforCensusdivisiondiminishestheintergenerationalgapsatthe25thpercentilebyroughly50percentinbothhistoricalcohorts.
34Theintergenerationalgapsarefurtherreducedwhenweaddchildhoodstatefixedeffects(fourthbarofeachgraph),andtheyfullyclosewhenwefurtherincludeeitherchildhoodstate-by-urbanstatus(fifthbar)orchildhoodcountyfixedeffects(sixthbar).
Wethusconcludethatimmigrantparents'locationchoicewasanimportantdriveroftheimmigrantmobilityadvantageweseeinthepast.
Inthemoderndata,wearemorelimitedinourabilitytoillustratetheroleofgeographyinexplainingthemobilityadvantageforthechildrenofimmigrants.
Specifically,theGSSsurveyonlycontainsinformationonCensusdivisionofresidenceatage16andtheOpportunityInsightsdataallowsforsomeaggregatecomparisonsbycountyofresidence.
Unlikeinthepast,theintergenerationalgapisunchangedwhencomparingchildrengrowingupinthesameregionintheGSS(Figure7,panel(c)).
TheOpportunityInsightsdatareporttheaverageranksofchildrenwhogrewupatthevariouspercentilesoftheincomedistributionbyparentalnativityfor1,053counties.
Foreachofthesecounties,wecalculatedthegapinexpectedrankbetweenchildrenofimmigrantsandchildrenoftheUS-bornatthe25thpercentilewhogrewupinthesamecounty.
Wefindthattheaveragegapis4.
1percentilesinthese1,053counties,whichissmallerthanthenationalgapof5.
8percentiles,suggestingthatlocationchoicestillmatterstodaybuttoasomewhatlesserdegreethaninthepast.
Overall,ourfindingssuggestanimportantroleformigrants'locationchoicesinexplainingthehighermobilityoftheirchildren.
Immigrantsweresavvyenoughtoavoidthoseareasofthecountrythatdidnotofferstrongprospectsformobility.
Forexample,inthepastimmigrantsavoidedtheUSSouth,whilemanyUS-bornwhiteSouthernersdidnotmoveawayfromthislow-mobilityregion.
Byleavingtheirhomecountry,immigrantshavealreadyrevealedthemselvestobewillingtomove,whichmightprovidethemwithanadvantageoverUS-bornindividualswhomightbelesswillingtoleavearegionwithlow-mobilityprospects.
B.
Father'sincomenotreflectingearningspotential34Thegapsalsodiminishsubstantiallywhenwefocusonchildrenatthe50thand75thpercentileoftheincomedistribution(seeAppendixFigureA12intheNBERWPversion).
17AnotherfactorthatcouldexplainthehigherratesofupwardmobilityforthechildrenofimmigrantsisthatimmigrantfathersdidnotearnuptotheirpotentialintheUSlabormarket,therebygivingtheirchildrenmoreroomtoimprove.
Inotherwords,ifimmigrantfathersarerankedlowerinthedistributionthantheirskillsorabilitywouldsuggest(e.
g.
,duetoalackoflanguageability,discrimination,ormorelimitedlabormarketnetworks),thenthechildrenoftheseimmigrantsareexpectedtoperformbetterthantheirfathers'outcomeswouldpredict.
OnewaytotestthismechanismistocomparethechildrenofimmigrantfatherswhoarrivedintheUSatdifferentages.
Immigrantfatherswhoarrivedaschildrenlikelyassimilatedintothelabormarketmoreeasily(forinstance,bybeingexposedtoUSeducationfromanearlyage),sothattheirobservedincomeislikelyclosertotheirpotentialearnings.
35Panel(a)ofFigure8showsthatthemobilitygapatthe25thpercentileisindeedfourtimeslargerforthechildrenofimmigrantswhoarrivedasadults(17yearsoldorolder)thanforthechildrenofimmigrantswhoarrivedbeforeschoolingage(7yearsoldoryounger).
Indeed,thelattergroupexhibitsonlyasmallmobilitygaprelativetothechildrenoftheUS-born.
36Panels(b)and(c)ofFigure8showthatthemobilitygapgrowswithfathers'ageofarrival,butespeciallyforindividualswithfathersfromnon-English-speakingcountries.
Thisfindingisconsistentwiththeideathatimmigrantfathersfromnon-EnglishspeakingcountriesenduplowerintheincomedistributionpartlyduetoanimperfectcommandofEnglish,thusleavingtheirchildrenwithmoreopportunitytomoveuptheincomedistribution(seealsoBleakleyandChin2010).
ThechildrenofUS-bornfatherswhomigratedinternallywithintheUSareausefulcomparisongroupthatcanshedlightontherelativerolesoflocationchoicesandfather'sunder-placementintheincomedistribution.
Ontheonehand,internalmigrantswerelikelytosortintoareaswithbetterprospectsfortheirchildren.
Ontheotherhand,thesefatherswereraisedintheUSandsowouldnotlikelysufferfromunder-placementintheincomedistributionintheirnewlocation.
Indeed,AppendixFigureC2showsthattheupwardmobilityofchildrenofUS-bornfatherswhomigratedinternallyfallsbetweenthatofchildrenofimmigrantsandchildrenofUS-bornfatherswho35WardandAlexander(2018)showsthatageofarrivalwasakeydriverofimmigrantearningsduringtheAgeofMassMigration,withimmigrantsarrivingatyoungeragesoutearningthosearrivinglaterinlife.
36Wefocusthisanalysisonthe1910–1940cohortbecausethe1880CensusdidnotincludeinformationonyearofarrivaltotheUS.
ParentalageofarrivaltotheUSisalsonotavailableintheGSS.
Wealsofindevidencethatfathersarelessunder-placedastheyspendmoretimeintheUSlabormarket.
Specifically,AppendixFigureC1showsthattheimmigrant-nativegapatthe25thpercentileisabout25%(40%)smallerforchildrenwhoseimmigrantfathersareaged40–50relativetochildrenwhoseimmigrantfathersareaged30–40inthe1880–1910(1910–1940)cohort.
18remainedintheirstateofbirth,suggestinganimportantroleforbothgeographicchoicesandunder-placement.
C.
Immigrantself-selectionAstudyofthelong-termimplicationsofimmigrationforthedestinationcountryrequiresafocusontheindividualswhochosetomigrateandraisechildreninthehostcountry.
IndividualswhochoosetomovetotheUSareself-selected,whichcouldbeonereasonwhythechildrenofimmigrantsaremoreupwardlymobile(forexample,ifimmigrantsaremoreforwardlookingorcaremoreabouttheupwardmobilityoftheirchildren,andarethusmorelikelytoinvestinthem).
Interestingly,wefindthattheimmigrantmobilityadvantageispresentevenforchildrenofimmigrantsfromcountrieswheremostofthemigrationflowiscomprisedofrefugees,forwhomselectionisarguablylessimportant(forinstance,theVietnameseinthemoderndata).
Yetimmigrantselectionmayhavemanifesteditselfinmanyways.
Ourhistoricaldataallowustostudyoneaspectofself-selectionasanexplanationforimmigrants'higherupwardmobility:namely,thatimmigrantfamiliesinvestedmoreintheirchildren'seducationalattainment,whichinturnservedasavehicletowardachievingbetterlabormarketoutcomes.
Figure9considersthe1910–1940cohort—theonlycohortforwhichwehavedataoncompletedyearsofschoolingforthesons—andshowsthat,ifanything,sonsofUS-bornfathersweremorelikelytograduatefromhighschool(panel(a))andreachhighergradesinschool(panel(b)).
Similarly,panel(c)showsthatsonsofUS-bornfatherswhowereaged12–16weremorelikelytobeattendingschoolin1910thanthesonsofimmigrantfathers.
37Hence,weconcludethathistorically,highereducationalinvestmentsamongimmigrantfamilieswereunlikelytobetheexplanationfortheirhigherupwardmobility,althougheducationmayplayalargerroletoday.
387.
Conclusions37Similarly,usingthe10percentsampleofthe1880Census,weseethatsonsofimmigrantfathersaged12–16are8percentagepointslesslikelytobeattendingschoolthansonsofUS-bornfathers(57versus65percent).
38Despitetheselowerlevelsofeducationalattainment,panel(d)confirmsthatthechildrenofimmigrantsenjoyedhigherincomemobilitybecausetheyearnedahigherincomeatanygivenlevelofeducation.
Thefactthateducationaldifferencesdonotexplainthemobilitygapisnotentirelysurprisingbecause,asGoldin(1998)shows,thereturnstoschoolingin1940werelowerthaninrecentyears.
However,wealsonotethatthechildrenofimmigrantsenjoyedafasterrateofeducationalmobility:Foragivenleveloffather'seducation,thechildrenofimmigrantsinthebottomhalfofthedistributiontendedtoachievemoreyearsofschoolingthanthechildrenoftheUS-born,consistentwiththefindingbyCard,Domnisoru,andTaylor2018(seepanel(e)).
19WeusenewlyconstructedlinkeddatasetsoffathersandsonstostudywhetherthechildrenofimmigrantsachieveearningsparitywiththechildrenoftheUS-born,andhowtheintergenerationalmobilityofimmigrantshaschangedoverthelasttwocenturies.
Wefindthat,bothinthepastandtoday,childrenofimmigrants,especiallythosegrowingupinpoorerfamilies,hadgreaterchancesofmovingupintheincomedistributionrelativetothechildrenofUS-bornparents.
Thisfindingisnotdrivenbyimmigrantsfromanyparticularcountryorethnicorigin.
Rather,whenestimatingmobilityratesbycountryoforigin,wefindthatchildrenofimmigrantsfromnearlyeverysendingcountryhavehigherratesofupwardmobilitythanthechildrenoftheUS-born.
Thatimmigrantsweresimilarlyupwardlymobilebothahundredyearsagoandtodayisremarkable.
Forexample,whereasimmigrantsinthepastpredominantlyhailedfromEurope,immigrantstodayhavemorediverseethnicbackgrounds,comingfromvariousLatinAmericanandAsiancountries.
Totheextentthatmembersofethnicminoritiesmighthavedifferentlevelsofintergenerationalmobility(asshowninanumberofstudiesincludingChettyetal.
2018a),wemighthaveexpectedthischangeintheethnicmixtohaveaninfluenceontheratesofintergenerationalassimilation.
Thepast-presentsimilarityisalsoremarkablegiventhatimmigrantstodaycomefromcountrieswithlowerincomelevelsrelativetotheUS,andgiventhedramaticchangesinUSimmigrationpolicyoverthelastcentury(theendingofopenbordersforEuropeanimmigrantsandtheimpositionofadditionalregulationsforimmigrantentry).
39Overall,ourfindingsstandincontrasttothenostalgicviewthatitwaseasierforimmigrantsinthepasttointegrateintotheUSeconomyandsociety.
WefindthatanimportantexplanationforwhythechildrenofimmigrantsaremoreupwardlymobileisthatimmigrantfamiliesaremorelikelythantheUS-borntomovetoareasthatofferbetterprospectsfortheirchildren.
Thisfindingisespeciallyrelevantinlightofthedeclineinbothgeographical(Molloy,Smith,andWozniak2011)andeconomic(Chettyetal.
2017)mobilitythathastakenplaceinrecentdecadesintheUS.
Itsuggeststhattotheextentthat"pocketsofopportunity"remainavailableintheUS,immigrantsmightbeabletoenjoyhighlevelsofmobilityevenifoverallmobilitydeclines.
Italsohighlightsoneuniqueaspectofimmigrants:namely,thatbyleavingtheirhomecountry,theyhaverevealedthemselvestobewillingtomove.
39USGDPpercapitawasmorethan7timeshigherthaninMexicoorChinawhenmigrationflowsfromthesecountriestookoffin1970,whereasitthatwasonly2–3timeshigherthaninEuropeansendingcountriescirca1900.
Lowerinitialincomelevelofacountry-of-origingroupmightreducetheirmobilityifagroup'saverage"ethniccapital"affectstheirupwardmobility(Borjas1992).
20Thesuccessofsecond-generationimmigrantsgivesamoreoptimisticviewofimmigrantassimilationthanpreviousstudiesthathavefocusedonlyonthefirstgeneration.
Indeed,wefindthatsecond-generationimmigrantsovertake,ratherthanjustcatchupwith,thechildrenoftheUS-bornwithcomparablefamilyincomes.
Althoughsomepoliticianshaveashort-termperspectiveonimmigrantassimilation,ourfindingssuggestthatthisviewmightunderestimatethelong-runsuccessofimmigrants.
Ourfindingsaremoreconsistentwiththeideaofthe"AmericanDream,"bywhichevenimmigrantswhocometotheUnitedStateswithfewresourcesandlittleskillshavearealchanceatimprovingtheirchildren'sprospects.
21ReferencesAbramitzky,Ran,andLeahBoustan.
2017.
"ImmigrationinAmericanEconomicHistory.
"JournalofEconomicLiterature55(4):1311–1345.
Abramitzky,Ran,LeahBoustan,KatherineEriksson,JamesFeigenbaum,andSantiagoPérez.
2019.
"AutomatedLinkingofHistoricalData.
"WorkingPaper.
Abramitzky,Ran,LeahPlattBoustan,andKatherineEriksson.
2012.
"Europe'sTired,Poor,HuddledMasses:Self-SelectionandEconomicOutcomesintheAgeofMassMigration.
"AmericanEconomicReview102(5):1832–1856.
Abramitzky,Ran,LeahPlattBoustan,andKatherineEriksson.
2014.
"ANationofImmigrants:AssimilationandEconomicOutcomesintheAgeofMassMigration.
"JournalofPoliticalEconomy122(3):467–506.
Aydemir,Abdurrahman,Wen-HaoChen,andMilesCorak.
2009.
"IntergenerationalEarningsMobilityAmongtheChildrenofCanadianImmigrants.
"TheReviewofEconomicsandStatistics91(2):377–397.
Bhattacharya,DebopamandBhashkarMazumder.
2011.
"ANonparametricAnalysisofBlack-WhiteDifferencesinIntergenerationalIncomeMobilityintheUnitedStates"QuantitativeEconomics2:335–379.
Bleakley,HoytandAimeeChin.
2010.
"AgeatArrival,EnglishProficiency,andSocialAssimilationamongUSImmigrants.
"AmericanEconomicJournal:AppliedEconomics2(1):165–92.
Bolotnyy,ValentinandCristinaBratu.
2018.
"TheIntergenerationalMobilityofImmigrantsandtheNative-Born:EvidencefromSweden.
"WorkingPaper.
Borjas,George.
1992.
"EthnicCapitalandIntergenerationalMobility.
"TheQuarterlyJournalofEconomics107(1):123–150.
Borjas,George.
1993.
"TheIntergenerationalMobilityofImmigrants.
"JournalofLaborEconomics11(1):113–135.
Cadena,BrianC.
andBrianK.
Kovak.
"ImmigrantsEquilibrateLocalLaborMarkets:EvidencefromtheGreatRecession.
"AmericanEconomicJournal:AppliedEconomics8(1):257–290.
Card,David,JohnDiNardo,andEugeniaEstes.
2000.
"TheMoreThingsChange:ImmigrantsandtheChildrenofImmigrantsinthe1940s,the1970s,andthe1990s.
"inGeorgeJ.
Borjas(Ed.
),IssuesintheEconomicsofImmigration.
Chicago:UniversityofChicagoPressCard,David,CiprianDomnisoru,andLowellTaylor.
2018.
"TheIntergenerationalTransmissionofHumanCapital:EvidencefromtheGoldenAgeofUpwardMobility.
"WorkingPaper25000,NationalBureauofEconomicResearch.
22Chetty,Raj,DavidGrusky,NathanielHendren,MaximilianHell,RobertManduca,andJimmyNarang.
2017.
"TheFadingAmericanDream:TrendsinAbsoluteIncomeMobilitySince1940.
"Science.
Chetty,Raj,JohnN.
Friedman,NathanielHendren,MaggieR.
Jones,andSonyaR.
Porter.
2018a.
"RaceandEconomicOpportunityintheUnitedStates:AnIntergenerationalPerspective.
"WorkingPaper24441,NationalBureauofEconomicResearch.
Chetty,Raj,JohnN.
Friedman,NathanielHendren,MaggieR.
Jones,andSonyaR.
Porter.
2018b.
"TheOpportunityAtlas:MappingtheChildhoodRootsofSocialMobility.
"WorkingPaper25147,NationalBureauofEconomicResearch.
Chetty,Raj,NathanielHendren,PatrickKline,andEmmanuelSaez.
2014.
"WhereisthelandofOpportunityTheGeographyofIntergenerationalMobilityintheUnitedStates".
QuarterlyJournalofEconomics,129(4):1553–1623.
Collins,WilliamJ.
andMarianneH.
Wanamaker.
2017.
"UpfromSlaveryAfricanAmericanIntergenerationalEconomicMobilitySince1880.
"WorkingPaper23395,NationalBureauofEconomicResearch.
Collins,WilliamJ.
andAriellZimran.
2019.
"TheEconomicAssimilationofIrishFamineMigrantstotheUnitedStates.
"WorkingPaper25287,NationalBureauofEconomicResearch.
Connor,DylanShane.
2019.
"ClassBackground,ReceptionContext,andIntergenerationalMobility:ARecordLinkageandSurnameAnalysisoftheChildrenofIrishImmigrants.
"InternationalMigrationReview.
Derenoncourt,Ellora.
2019.
"CanyoumovetoopportunityEvidencefromtheGreatMigration.
"WorkingPaper.
Davis,JonathanandBhashMazumder.
2019.
"TheDeclineinIntergenerationalMobilityAfter1980.
"FederalReserveBankofChicagoWorkingPaperNo.
2017-05.
Duncan,Brian,JeffreyGrogger,AnaSofiaLeonandStephenJ.
Trejo.
2017.
"NewEvidenceofGenerationalProgressforMexicanAmericans.
"WorkingPaper24067,NationalBureauofEconomicResearch.
Feigenbaum,JamesJ.
2018.
"MultipleMeasuresofHistoricalIntergenerationalMobility:Iowa1915to1940.
"EconomicJournal128(612):F446–F481.
Feigenbaum,JamesJ.
2015.
"IntergenerationalMobilityduringtheGreatDepression.
"WorkingPaper.
Ferrie,JosephP.
1997.
"TheEntryIntoTheUSLaborMarketOfAntebellumEuropeanImmigrants,1840–1860.
"ExplorationsinEconomicHistory31(3):295–330.
23Ferrie,Joseph,CatherineMassey,andJonathanRothbaum.
2016.
"DoGrandparentsandGreat-GrandparentsMatterMultigenerationalMobilityintheUS,1910–2013.
"WorkingPaper22635,NationalBureauofEconomicResearch.
Flood,Sarah,MiriamKing,RenaeRodgers,StevenRuggles,andJ.
RobertWarren.
IntegratedPublicUseMicrodataSeries,CurrentPopulationSurvey:Version6.
0[dataset].
Minneapolis,MN:IPUMS,2018.
https://doi.
org/10.
18128/D030.
V6.
0.
Gielen,AnneandDinandWebbink.
2019.
"MigrationSelectionandAssimilation:EvidencefromtheGreatMigrationfromSurinamtotheNetherlands"WorkingPaper.
Goldin,Claudia.
1998.
"America'sGraduationfromHighSchool:TheEvolutionandSpreadofSecondarySchoolingintheTwentiethCentury.
"JournalofEconomicHistory58(2):345–374.
Goldenweiser,E.
A.
1916.
"TheFarmer'sIncome.
"AmericanEconomicReview6(1):42–48.
Hatton,Timothy.
1997.
"TheImmigrantAssimilationPuzzleinLateNineteenthCenturyAmerica.
"JournalofEconomicHistory57(1):34–62.
Hilger,Nathaniel.
2016.
"TheGreatEscape:IntergenerationalMobilityintheUnitedStatesSince1940.
"WorkingPaper21217,NationalBureauofEconomicResearch.
Hilger,Nathaniel.
2017.
"UpwardMobilityandDiscrimination:TheCaseofAsianAmericans.
"WorkingPaper22748,NationalBureauofEconomicResearch.
Kosack,EdwardandZacharyWard.
2018.
"ElsueoamericanoThegenerationalprogressofMexicanAmericanspriortoWorldWarII.
"WorkingPaper.
Lee,Chul-InandGarySolon.
2009.
"TrendsinIntergenerationalIncomeMobility.
"ReviewofEconomicsandStatistics91(4):766–772.
Lubotsky,Darren.
2007.
"ChutesorLaddersALongitudinalAnalysisofImmigrantEarnings.
"JournalofPoliticalEconomy115(5):820–867.
Manson,Steven,JonathanSchroeder,DavidVanRiper,andStevenRuggles.
2019.
IPUMSNationalHistoricalGeographicInformationSystem:Version14.
0[Database].
Minneapolis,MN:IPUMS.
http://doi.
org/10.
18128/D050.
V14.
0Mazumder,Bhashkar.
2005.
"FortunateSons:NewEstimatesofIntergenerationalMobilityintheUnitedStatesUsingSocialSecurityEarningsData.
"TheReviewofEconomicsandStatistics87(2):235–255.
Mazumder,Bhashkar.
2015.
"EstimatingtheIntergenerationalElasticityandRankAssociationintheUS:OvercomingtheCurrentLimitationsofTaxData.
"FederalReserveBankofChicagoWorkingPaperNo.
2015-04.
24Minns,Chris.
2000.
"Income,CohortEffectsandOccupationalMobility:ANewLookatImmigrationtotheUnitedStatesattheTurnofthe20thCentury.
"ExplorationsEconomicHistory37(4):326–50.
Molloy,Raven,ChristopherL.
Smith,andAbigailWozniak.
"InternalmigrationintheUnitedStates.
"JournalofEconomicperspectives25,no.
3(2011):173-96.
Olivetti,ClaudiaandM.
DanielePaserman.
2015.
"IntheNameoftheSon(andtheDaughter):IntergenerationalMobilityintheUnitedStates,1850-1940.
"AmericanEconomicReview105(8):2695–2724.
Olivetti,Claudia,M.
DanielePaserman,andLauraSalisbury.
2018.
"Three-generationMobilityintheUnitedStates,1850–1940:TheRoleofMaternalandPaternalGrandparents.
"ExplorationsinEconomicHistoryVol.
70:73–90.
Passel,Jeffrey.
1986.
"EstimatingtheNumberofUndocumentedAliens.
"MonthlyLaborReview,September.
WashingtonDC:BureauofLaborStatistics.
Perlmann,Joel.
2005.
ItaliansThen,MexicansNow:ImmigrantOriginsandSecond-GenerationProgress,1890–2000.
RussellSageFoundation.
Perlmann,JoelandRogerWaldinger.
1997.
"SecondgenerationdeclineChildrenofimmigrants,pastandpresent--areconsideration.
"TheInternationalMigrationReview31(4):893–922.
Pérez,Santiago.
2019.
"IntergenerationalOccupationalMobilityAcrossThreeContinents.
"TheJournalofEconomicHistory79(2)383–416.
Preston,SamuelH.
,andMichaelR.
Haines.
1991.
FatalYears:ChildMortalityinLateNineteenthCenturyAmerica.
Princeton,NJ:PrincetonUniversityPressRuggles,Steven,SarahFlood,RonaldGoeken,JosiahGrover,ErinMeyer,JosePacas,andMatthewSobek.
IPUMSUSA:Version8.
0[dataset].
Minneapolis,MN:IPUMS,2018.
https://doi.
org/10.
18128/D010.
V8.
0Saavedra,MartinandTateTwinam.
2018.
"AMachineLearningApproachtoImprovingOccupationalIncomeScores.
"WorkingPaper.
Tan,HuiRen.
2018.
"ADifferentLandofOpportunity:TheGeographyofIntergenerationalMobilityintheEarly20th-CenturyUS.
"WorkingPaper.
Ward,Zachary.
Forthcoming.
"TheNot-So-HotMeltingPot:ThePersistenceofOutcomesforDescendantsoftheAgeofMassMigration.
"AmericanEconomicJournal:AppliedEconomics.
Ward,ZacharyandRohanAlexander.
2018.
"AgeatArrivalandAssimilationduringtheAgeofMassMigration.
"TheJournalofEconomicHistory78(3)904–937.
25Ward,Zachary.
2019.
"TheLowReturntoEnglishFluencyduringtheAgeofMassMigration.
"WorkingPaper.
FiguresandTablesFigure1:Cross-SectionalEarningGapsforFirst-andSecond-GenerationImmigrantsrelativetoUS-Born,byCountryofOrigin(a)1880–1910Cohort(b)1910–1940Cohort(c)1980–2010CohortNote:Thehistoricalsamplesusethe1880,1910,and1940Censuses.
Thehistoricalsamplesofrst-generationimmigrantsandtheUSbornarerestrictedtowhitemenintheearlierCensuswhoareaged30–50,whohaveachildinthehousehold,andwhowereeitherbornintheUSorinoneofthe17sourcecountries.
Thesecond-generationsamplesarerestrictedtowhitemeninthelaterCensuswhoareaged30–50andwhosefatherswereeitherbornintheUSorinoneofthe17sourcecountries.
Forthemostrecentcohort,weusethe1980Censusandthe2006–2015CurrentPopulationSurvey(CPS)'sAnnualSocialandEconomicSupplement.
Werestrictthe1980sampleofrst-generationimmigrantsandtheircounterpartstomenaged30–50,whohaveachildpresentinthehousehold,andwhowerebornintheUSorinoneofthetopsendingcountriesidentiedintheOpportunityInsightssample.
WerestricttheCPSsampleofsecond-generationimmigrantsandtheircounterpartstomenaged30–50whosefatherswereeitherbornintheUSorinoneofthetopsendingcountries.
Forallofthesecohorts,weusepredictedincome(asdescribedinthemaintext)asthemeasureofincome.
1Figure2:IntergenerationalMobilityofImmigrantsandtheUSborn,Rank-RankCorrelations(a)1880–1910Cohort(b)1910–to1940Cohort(c)GSS:1984–2006Cohort(d)OpportunityInsights:1997–2015CohortNote:Sonsareassignedpercentileranksrelativetoallotherrespondentsborninthesamebirthyear.
Fathersarerankedrelativetoallfatherswithchildrenborninthesamebirthyear.
Theguresplotthemeanincomerankofchildrenbyfather'sincomerank,forsonswithandwithoutforeign-bornfathers,aswellasthecorrespondingregressionlinesusingequation(2).
IntheGSSsample,theoutcomesforsonsaremeasuredbetween2000and2018(withthemedianyearbeing2006).
Theoutcomesforparentsaremeasuredviaaretrospectivequestionaboutparents'occupationswhentherespondentwas16yearsold;thedatathereforecorrespondtoparentaloutcomesbetween1966and2004(withthemedianyearbeing1984).
TheOpportunityInsightsDatacomefromChettyetal.
(2018a);son'sincomeismeasuredin2014–2015andparentalincomeismeasuredin1994–2000.
2Figure3:AverageIncomeRankforChildrenBornto25thPercentile,byFather'sBirthplace(a)1880–1910Cohort(b)1910–1940Cohort(c)OpportunityInsights:1997–2015CohortNote:Thisgureplotstheaverageincomerankforchildrenborntothe25thpercentileoftheparentalincomedistribution.
DatafortheOpportunityInsightscohortcomefromChettyetal.
(2018a);son'sincomeismeasuredin2014–2015andparentalincomeismeasuredin1994–2000.
3Figure4:AverageIncomeRankforChildrenBornto75thPercentile,byFather'sBirthplace(a)1880–1910Cohort(b)1910–1940Cohort(c)OpportunityInsights:1997–2015CohortNote:Thisgureplotstheaverageincomerankforchildrenborntothe75thpercentileoftheparentalincomedistribution.
DatafortheOpportunityInsightscohortcomefromChettyetal.
(2018a);son'sincomeismeasuredin2014–2015andparentalincomeismeasuredin1994–2000.
4Figure5:IntergenerationalMobilityofImmigrants'Daughters,1997–2015Cohort(a)Rank-rankCorrelations(b)AverageIncomeRankforDaughtersBornto25thPercentile,byFather'sBirthplaceNote:Panel(a)plotsthemeanincomerankofdaughtersbyparentalincomerank,fordaughterswithandwithoutforeign-bornfathers,aswellasthecorrespondingregressionlinesusingequation(2).
Panel(b)plotstheaverageincomerankfordaughtersborntothe25thpercentileoftheparentalincomedistribution,byfather'scountryoforigin.
DataarefromChettyetal.
(2018a);daughter'sincomeismeasuredin2014–2015andparentalincomeismeasuredin1994–2000.
5Figure6:ShareofImmigrantsandUpwardMobility,byStateEconomicAreaorCommutingZone(a)Shareofsonsthataresecond-generation(1880–1910)(b)UpwardmobilityofsonsofUS-bornfathers(1880–1910)(c)Shareofsonsthataresecond-generation(1910–1940)(d)UpwardmobilityofsonsofUS-bornfathers(1910–1940)(e)Shareofpopulationthatisforeign-born,1980(f)UpwardmobilityofOpp.
InsightscohortNote:Panels(a)and(c)showtheshareoftheindividualsinourlinkedsamplethataresecond-generationimmigrantsineach1880or1910stateeconomicarea,respectively.
Panel(e)showstheshareeachofeach1980county'spopulationthatisforeign-born,usingdatafromMansonetal.
(2019).
Inpanels(b)and(d),upwardmobilityismeasuredastheshareofsonsofUS-bornfathersineachstateeconomicareathatreachedthetopfthofthenationalincomedistribution,conditionalonhavinghadafatherinthebottomfthofthenationalincomedistribution.
Panel(f)showstheshareofnon-HispanicwhitesonswhoareupwardlymobileusingdatafromChettyetal.
(2018b);upwardmobilityismeasuredastheshareofsonsinthetop20percentoftheincomedistribution,conditionalonhavinghadhouseholdincomeinthebottomfthofthenationalincomedistribution.
6Figure7:IntergenerationalGapatthe25thpercentile,ComparingChildreninSimilarChildhoodLocations(a)1880–1910Cohort(b)1910–1940Cohort(c)GSS:1984–2006CohortNote:ThisgraphplotstheintergenerationalgapbetweenchildrenofimmigrantsandchildrenofUS-bornmen,forchildrenborntothe25thpercentileoftheincomedistribution.
Therstbarforeachcohortisestimatedbyregressingson'sincomeranksontheirfathers'ranks,anindicatorforhavingaforeign-bornfather,andtheinteractionofthesetwovariables.
ThesecondandthirdbarsplotthegapafterincludinganindicatorvariableforgrowingupintheUSSouthandchildhoodCensusdivisionxedeects,respectively.
Inthehistoricalcohorts,thefourthbarplotsthegapafterincludingchildhoodstatexedeectsandthefthbarplotsthegapafterincludingstateurbanxedeects.
Finally,thesixthbarplotsthegapafterincludingchildhoodcountyxedeects.
7Figure8:IntergenerationalGapatthe25thpercentilefor1910–1940Cohort,byFather'sAgeofArrivalandLanguageinSendingCountry(a)AllCountries(b)Non-English-SpeakingCountries(c)English-SpeakingCountriesNote:ThisgraphplotstheintergenerationalgapbetweenchildrenofimmigrantsandchildrenofUS-bornmen,forchildrenborntothe25thpercentileoftheincomedistribution.
Theguresrstseparatesonsbasedonthelanguagespokenintheirfather'ssendingcountry(whereEnglish-speakingcountriesareCanada,England,Ireland,Scotland,andWales).
Then,sonsofimmigrantsaredividedintogroupsbasedontheirfather'sageofarrivaltotheUnitedStates.
8Figure9:EducationalAttainment,1910–1940cohort(a)Songraduatedhighschool(b)Son'shighestgradeinschool(c)Son'sschoolattendancein1910(d)Son'sincomerankandeducationalattainment(e)Son'sandfather'seducationalattainmentNote:Fathersarerankedrelativetoallfatherswithchildreninthesamebirthyear.
Panels(a)and(b)plotthemeaneducationalattainmentofchildrenbytheirfather'sincomerank,forsonswithandwithoutforeign-bornfathers,aswellasthecorrespondingregressionlines.
Panel(c)considerslinkedsonsaged12–16intheearlierCensusandplotsschoolattendanceasafunctionofhisfather'sincomerank(aftercontrollingforagexedeects).
Panel(d)plotsason'smeanincomerankasafunctionofcompletedyearsofschooling.
Panel(e)plotsason'smeanyearsofcompletedschoolingasafunctionofhisfather'syearsofschooling(forfatherswhocanbelinkedtothe1940Census).
9Table1:IntergenerationalMobilityEstimates,byCohortIntercept()Immigrantfather(0)Father'srank(1)Immigrantfatherrank(2)1880–1910Cohort30.
988.
370.
36-0.
08[0.
06][0.
12][0.
00][0.
00]1910–1940Cohort31.
006.
730.
36-0.
04[0.
04][0.
09][0.
00][0.
00]GSSCohort35.
426.
760.
29-0.
09[1.
17][3.
06][0.
02][0.
06]Opp.
InsightsCohort36.
647.
420.
33-0.
08[0.
40][0.
60][0.
01][0.
01]Note:Thistablereportsestimatestheestimatedcoecientsfromequation2inthepaper(standarderrorsinbracketsbeloweachestimate).
DatafortheOpportunityInsightscohortcomefromChettyetal.
(2018a).
101AppendixA:SensitivityofResultstoLinkingProcedureandSampleConstructioni.
LinkingAlgorithmAfirstconcernwithCensuslinksbasedontheavailableinformationonnames,age,andplaceofbirthisthepotentialforlinkingchildrentothewrongadults("falsepositives").
AppendixFigureA1showsthattheresultsaresimilarwhenusingmoreconservativeversionsoftheAbramitzky,Boustan,andErikssonlinkingalgorithm,whichreducessamplesizebutalsolikelyreducesmatchingerrors.
First,werestrictthesampletoexactmatchesbyfull(non-standardized)nameandage.
Second,werestrictthesampletoindividualswhosefirstandlastnamesareuniquewithinafive-yearageband.
Third,werequirematchestobebothexactanduniquewithinafive-yearageband.
Asecondconcernistheextenttowhichweareabletogeneraterepresentativesamples.
Itiswell-knownthatlinkedsamplesarenotfullyrepresentativeofthepopulation,asindividualswithuncommonnamesandotherattributesaremorelikelytomatchbetweenCensuswaves(Abramitzkyetal.
2019).
AppendixTableA2comparesthesonsinourlinkedsamplestotherelevantpopulationofUS-bornmeninthe1910and1940Censuses.
1Our1880–1910sampleoflinkedsonshasasimilarshareofsecond-generationimmigrantsasthefullpopulation(26vs.
29percent),andhasanaverageincomescorethatisroughly3percenthigherthanthepopulation.
Our1910–1940linkedsamplehasfewersecond-generationimmigrantsthanthefullpopulation(22vs.
24percent),andhasanaverageincomescorethatis5percenthigherthanthepopulation.
Weaddressthenon-representativenessofourlinkeddatabyrunningaversionofourmainspecificationinwhichwereweightthematcheddatasoastomimicthepopulationoncharacteristicssuchasage,stateofresidence,birthplace,andoccupation.
AppendixFigureA1showsthattheresultswhenreweightingthedataaresimilartothoseinthemainspecification.
21Thatis,wefocusonwhitemenaged30to50whosefatherswereeitherbornintheUSorinoneofthesourcecountriesofinterest.
2Theseweightsarecomputedusingtwoalternativeprocedures.
Inthefirstprocedure,observationsareweightedusingearlierCensus(i.
e.
,childhood)characteristics:Toconstructthisweight,weusethecross-sectionalversionoftheearlierCensusandidentifytheindividuals(i.
e.
,thesons)weareabletolinkforward.
Wethenregresstheindicatorforbeinga"matched"individualonfixedeffectsforthestateofresidence,theson'sage,thebirthplace,aswellasthefather'sbirthplaceandbroadoccupationalcategory.
Inthesecondprocedure,observationsareweightedusinglaterCensus(i.
e.
,adulthood)characteristics:Toconstructthisweight,weusethecross-sectionalversionofthelaterCensusandidentifytheindividuals(i.
e.
,thesons)weareabletolinkbackward.
Wethenregresstheindicatorforbeinga"matched"individualonfixedeffectsfortheson'sstateofresidence,age,birthplace,andbroadoccupationalcategory.
Inbothcases,theweightsweusetore-weightobservationsaretheinverseprobabilitythatanindividualcanbelinkedacrossCensuses.
2ii.
SampleConstructionOurbaselineapproachdefinesason'simmigrantstatususinghisfather'sbirthplace.
InAppendixFigureA1weshowthatourmainresultsremainunchangedwhendefiningimmigrantstatususingmother'sbirthplace,ortheimmigrantstatusofbothparents.
Second,unlikewiththehistoricaldata,ourmainspecificationforthemodernperiodcomparesimmigrantstotheentireUS-bornpopulationregardlessofrace.
3FigureA29intheNBERworkingversionofthispapershowsthatresultsaresimilarwhencomparingwhiteimmigrantstothewhitepopulationinthemodernperiodorcomparingallimmigrantstothewhitepopulationinthemodernperiod.
4AppendixB:SensitivityofResultstoIncomeAssignmentAppendixFigureB1showsthemainresultsforfather-sonpairsforwhomweobserveactualratherthanpredictedincomeinthe1940Census.
Wedosobylinkingfatherswhowererelativelyyoungin1910(20–30yearsold)from1910to1940,whichenablesustoobservethesefathers'actualincomein1940whentheywere50–60yearsold.
5Combiningthissamplewithourlinkedsampleofsons,weareabletoobserveactualincomeofbothfathersandsonsfor62,000pairs,ofwhich7,800areimmigrants(i.
e.
,thosepairsforwhichweareabletolinkboththesonandhisfatherfrom1910to1940).
Forthissetoffather-sonpairs,weshowasimilarimmigrants'mobilityadvantagewhenusingeitheractualincomeorincomescores.
WealsocontinuetofindthatchildrenofimmigrantsfrommostsendingcountriesaremoreupwardlymobilethanthechildrenoftheUS-born(butnaturally,theexactrankingofcountrieschangesoncewedividethe7,800immigrantfathersinto17sendingcountries).
AppendixFigureB2summarizestherobustnessofourestimatesforthe1880-1910and1910-1940cohortstousingvariousproxiesofincome.
Eachrowinthisfigureshowsthevaluesoftheinterceptandtheslopeoftherank-rankrelationshipforbothsonsofimmigrantandUS-bornfathers.
First,weshowthattheresultsaresimilarwhenweuseactualratherthanpredictedincomeforthesonsweobserveinthe1940Census(secondrowofpanel(b)).
Second,wecontinuetofind3Inthepast,immigrantswereoverwhelminglywhite,andsowecompareimmigrantstothewhiteUS-bornpopulationinthehistoricaldata.
Today,incontrast,immigrantsaremoreethnicallydiverse.
4Wenotethatinthiscasewedefinesecond-generationstatusinthehistoricalcohortsusingthemother'sbirthplaceforcomparabilitywiththecorrespondingmoderndata.
5Weremindreadersthatthelimitationsofthisexerciseare:(1)thatactualincomeinthe1940Censusisonlyavailableforwageearnersandisnotablymissingforfarmers,and(2)fathersareobservedinadifferentagerangethansons(50–60yearsold,insteadof30–50yearsold).
Whenperformingthisexercise,werankfathersandsonsonlyrelativetootherfathersandsonswhoarealsowageearners.
3animmigrantmobilityadvantagewhenweusealternativemeasuresofincomeforfarmers(secondandthirdrowinpanel(a)andthethirdandfourthrowinpanel(b)).
Specifically,weusethe1900CensusofAgriculturetocalculateaveragefarmers'netearnings(farmincomeminusexpenses)foreachUScounty.
6Alternatively,giventheprominenceoffarmingandthedifficultyinassigningfarmersanincomemeasure,weshowthatthemainresultsdonotchangewhenwedropfathersandsonswhoarefarmersfromthesampleandcomputeranksbasedonthenon-farmingnationalincomedistribution.
7Ourbaselineincomescoreisbasedonthe1940Census.
Weshowthatresultsaresimilarwhenweusemorecontemporaneousincomesourcesfor1880and1910fathersandsons(fourthrowofpanel(a)andfifthrowofpanel(b)).
Specifically,fornon-farmers,weuseinformationonaverageearningsbyoccupationfromthe1901CostofLivingSurvey(PrestonandHaines1991).
8Forfarmers,weusethe1900CensusofAgriculture(asdescribedabove).
Wenextuseameasureofincomethatalsoincorporatesarespondent'scountryofbirthintheincomeprediction(i.
e.
,allowsforanimmigrantpenaltyinearnings),andfindthatsonsofimmigrantfatherscontinuetobemoreupwardlymobilethancomparablechildrenofUS-bornfathers.
9Unlikewhenweuseotherincomescores,childrenofimmigrantsandchildrenoftheUS-bornlooksimilarlymobileifweusetheIPUMSoccupationscore,whichisbasedonassigningeachoccupationitsmedianearningsin1950(sixthandseventhrowsinpanels(a)and(b)ofFigureB2,respectively).
However,theIPUMSoccupationscoreisnotwellsuitedtooursettingmainly6Incomeiscalculatedasthesumofthevalueoffarmoutputnotfedtolivestockandthevalueofhouserentandfood/fuelproducedonfarmandconsumedbythefamily.
Expensesarecalculatedasthesumofexpendituresforfarmlabor;fertilizer;feed,seed,andthreshing;taxes;depreciationofvalueofbuildings;anddepreciationofmachinery.
Formoredetailsonthiscalculation,wereferthereadertoGoldenweiser(1916)andAbramitzkyetal.
(2012).
Wenotethatforfarmerslivingincountiesforwhichwecouldnotcalculatefarmernetearnings,weassignedthemthestateestimateforfarmernetearnings.
Forfarmerslivingincountiesforwhichthisapproachyieldednegativenetearnings,weassignedthemtheminimumpositivevalueofcounty-levelfarmerearningsintheirstate.
Finally,forfarmerslivingincountiesforwhichthisapproachyieldedverylargenetearnings(i.
e.
,obviousoutliersabovethe99thpercentileofthefarmerearningsdistribution),weassignedthemthemedianvalueoffarmerearningsintheirstate.
7Ofcourse,movingintooroutoffarmingwasbyitselfanavenueformobility,whichiswhyweprefertoincludefarmersinourbaselinesampledespitethechallengesinmeasuringfarmerincome.
8The1901CostofLivingSurveyiscloserintimetotheearlierCensusesbutlessdetailedthanourbaselinemeasure(basedonthe1940Census):itincludesroughly150non-agriculturaloccupations,whereasthe1950occupationalclassificationinourlinkedsamplehasaround220occupations.
9TheNBERworkingpapershowsallresultsaresimilarusingthismeasure.
There,wealsopresentaspecificationthatincorporatesthefather'scountryoforiginintotheson'sincomeprediction.
Inthatspecification,theaverageincomerankofchildrenofimmigrantsishigherthantheaverageincomerankofchildrenoftheUS-bornexceptattheverybottomoftheincomedistribution(andasaresult,theestimatedinterceptoftherank-rankrelationshipendsupbeingsimilarforbothgroups).
4becausethescoreforfarmersistoolow:Farmingwasarelativelyhigh-paidoccupationintheearly1900s,butitbecamealow-paidoccupationby1950.
10ThatmeansthattheIPUMSoccupationscoreistoolowforfarmersin1880and1910.
ThisassignmentisproblematicinourcontextbecauseUS-bornindividualsweremuchmorelikelythanimmigrantstobefarmers(andthechildrenofbothUS-bornandimmigrantstendedtoshiftawayfromfarming).
Therefore,assigningfarmersanincomethatistoolowresultsinanunderestimationofincomesofUS-bornfathers,andtoanoverestimationoftheratesatwhichchildrenoftheUS-bornmoveduptheincomeladderrelativetochildrenofimmigrants.
Indeed,ifweusetheIPUMSoccupationscoreandonlyadjustfarmers'incomesin1880and1910tocorrespondtotheirappropriaterankinthe1900s(basedonourbaselineincomescore),weagainseethatimmigrantshaveamobilityadvantageinbothhistoricalcohorts(seventhandeighthrowsinpanels(a)and(b)ofFigureB2,respectively).
10Forinstance,farmerswereatthemedianoftheincomedistributionintheIowa1915Census(whichincludedinformationonfarmer'sincome)butwereataboutthebottom10thpercentilenationallyin1950(Feigenbaum2018).
AppendixFiguresandTablesFigureA1:RobustnesstoWeighting&LinkingAlgorithmandtoSampleConstruction(a)1880–1910Cohort(b)1910–1940CohortNote:Theseguresreporttheslopeandinterceptfromregressionsofson'srankonfather'srank,usingalternativeweightingschemesandlinkingalgorithmsaswellasalternativeapproachesforconstructingthesample.
Therstthreeestimatesineachpanelshowtheresultsfromtherank-rankspecicationsthatusealternativestrategiesforlinkingfathersandsonsacrosstheearlierandlaterCensuses,asdescribedinAppendixA.
"Weighteddata"referstoweighingobservationsbytheinverseprobabilityofbeingalinkedindividual.
Eachcohorthastwopossibleweights:onethatuseschildhoodcharacteristicsandtheearliercross-sectionalCensus,andonethatusesadultcharacteristicsandthelatercross-sectionalCensus.
"Immigrantmother"referstoasamplethatusesthemother'sbirthplacetoclassifymenassonsofimmigrants.
"Immigrantmother&father"referstoasamplethatclassiesindividualsasimmigrantsonlyifboththeirmotherandfatherwerebornoutsideoftheUnitedStates.
1TableA1:SampleSize,byCohortandFather'sCountryofOrigin1880–1910cohort1910–1940cohortUS-bornfatherImmigrantfatherUS-bornfatherImmigrantfatherWhitemen,age0–166,184,2192,814,91010,492,5934,079,624Linkedmen1,432,861602,7263,224,5781,068,064livingwithfather1,287,253544,1972,931,570965,839whitefather,age30–50931,754414,9332,175,175740,814fromlargestsendingcountries931,754397,3142,175,175696,007non-missinglabormarketoutcomes893,046380,5902,054,044657,167Shareofwhitemenage0–16innalsample14.
4413.
5219.
5816.
11Note:Therstrowreferstothenumberofwhitemenages0–16intheearlierCensus.
Rows2through6showthesamplesizeaswerestrictthesamplebasedonthefathers'andsons'characteristics.
"Immigrantfather"referstosonswhosefatherswerebornoutsideoftheUnitedStates,and"largestsendingcountries"referstothe17countriesconsideredinthehistoricalcohorts.
2TableA2:ComparisonofCross-SectionalandLinkedSampleofSons(a)SonsofUS-bornFathers1880-1910cohort1910-1940cohortNon-LinkedLinkedNon-LinkedLinkedAge37.
1137.
2736.
2837.
44Farmer0.
240.
250.
300.
33White-collar0.
320.
330.
140.
14Skilled0.
250.
250.
320.
34Unskilled0.
190.
170.
240.
19Income1,045.
681,077.
481,064.
111,140.
49South0.
350.
290.
330.
31(b)SonsofImmigrantFathers1880-1910cohort1910-1940cohortNon-LinkedLinkedNon-LinkedLinkedImmigrant(fullsample)0.
260.
290.
220.
24Age37.
2837.
7036.
1038.
09Farmer0.
270.
270.
360.
37White-collar0.
190.
220.
080.
10Skilled0.
340.
330.
350.
36Unskilled0.
200.
180.
210.
17Income1,161.
071,164.
521,199.
361,255.
71South0.
080.
080.
050.
06Note:Inthistable,thelinkedsamplehasnotyetbeenrestrictedbasedonthefathers'characteristics(i.
e.
,whetherthefatheriswhite,aged30–50,livingwiththechildintheearlyCensus,andwithoutmissinglabormarketoutcomes).
3FigureB1:RobustnesstoUsingActualIncomeforFathersandSons,1910–1940Cohort(a)Rank-Rank,ActualIncome(b)Avg.
Rankat25thPercentile.
,ActualIncome(c)Rank-Rank,IncomeScore(d)Avg.
Rankat25thPercentile.
,IncomeScoreNote:Theseguresuseasampleoffather-sonpairsforwhichweareabletolinkbothfathersandsonsbetween1910and1940.
Thesampleisrestrictedtofather-sonpairsinwhichboththefatherandthesonreceivedtheirincomefromwages.
Panels(a)and(b)arebasedonindividual-levelincomefromwagesfromthe1940Census,whereaspanels(c)and(d)usethebaselineincomescoresthatweusethroughoutthepaper.
4FigureB2:RobustnesstoUsingAlternativeIncomeMeasures(a)1880–1910Cohort(b)1910–1940CohortNote:Theseguresreporttheslopeandinterceptfromregressionsofson'srankonfather'srank,usingalternativeapproachesformeasuringfathers'andsons'incomes.
"Actualincome"referstospecicationsthatuseactualincomeforthe1940sons,adjustedforself-employmentandfarmincome(notewecanonlyrunthisexerciseforthe1910-1940cohort).
"FarmerAdjustments"keepspredictedincomefornon-farmingoccupations,butadjusts1880and1910farmers'incomeusingcounty-levelmeasuresderivedfromthe1900CensusofAgriculture;1940sonsareadjustedusingthe1950IPUMS"occscore"variable.
"Nofarmers"dropsfathersandsonswhoarefarmersfromthesampleandre-runsthespecicationusingranksfromthenon-farmingincomedistribution.
"1901CostofLiving,FarmerAdj.
"usesthefarmeradjustmentsfromabove,andalsoadjustsnon-farmerincomeforallfathersaswellas1910sonsusingtheaverageearningsinaperson'soccupationfromthe1901CostofLivingSurvey.
"Fathers,ImmigrantPenalty"adjustspredictedincomesothatarespondent'sbirthplaceisincludedintheincomeprediction(orimmigrantstatusforfarmers).
"IPUMS1950Occscore"referstousingtheIPUMS1950"occscore"variabletomeasureincome.
"IPUMS1950Occscore,FarmerAdj.
"referstousingthe1950"occscore"variablefornon-farmers,butscalingup1880and1910farmers'incomesothatfarmersarerankedroughly10percentileshigherontheincomedistribution(atthesamelevelasinourbaselineapproach).
"Includingmissingincome"referstospecicationsinwhichfathersandsonswithmissingoccupationorincomeareassignedzeroincome.
5FigureC1:Rank-rankCorrelations,byImmigrantFathersAge(a)1880–1910Cohort:Rank-rank(b)1910–1940Cohort:Rank-rankNote:Theguresplotthemeanincomerankofchildrenbyfather'sincomerankforeachgroupaswellasthecorrespondingregressionlines.
Thisguredividesimmigrantfathersintotwogroupsbasedonage:fathersaged30to39andfathersaged40to50.
Sonsareassignedpercentileranksrelativetoallotherrespondentsborninthesamebirthyear.
Fathersarealsorankedrelativetoallfatherswithchildreninthesamebirthyear.
FigureC2:Rank-rankCorrelations,byUS-bornFather'sInternalMigrationStatus(a)1880–1910Cohort:Rank-rank(b)1910–1940Cohort:Rank-rankNote:Theguresplotthemeanincomerankofchildrenbyfather'sincomerankforeachgroupaswellasthecorrespondingregressionlines.
ThisguredividessonswithUS-bornfathersintotwogroupsbasedonthefather'smigrantstatus:thosewhosefathersmovedwithintheUS(i.
e.
,theirstateofresidencediersfromtheirstateofbirth)andthosewhodidnotmove.
Sonsareassignedpercentileranksrelativetoallotherrespondentsborninthesamebirthyear.
Fathersarealsorankedrelativetoallfatherswithchildreninthesamebirthyear.
6