1EvidencetobearuponthemeritocraticfeaturesofChinascivilexamsystemcanbefoundintheEuropeandiscourseamongthelikesofVoltaireQuesnayandChristianWol11allofwhomvieweditasasuperioralternativetothetradi ID: 824835
Download Pdf The PPT/PDF document "1IntroductionGrowingevidencesuggeststhat..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
1IntroductionGrowingevidencesuggeststhat
1IntroductionGrowingevidencesuggeststhathistoricalinstitutionscangeneratelong-lastingeectsonmoderneconomicdevelopmenteitherdirectlyorviatheculturaltraitstheyfoster(Acemoglu,Johnson,andRobinson,2001;Dell,2010;Dell,Lane,andQuerubin,2018;Guiso,Sapienza,andZingales,2016;Nunn,2008;NunnandWantchekon,2011;VoigtlanderandVoth,2012,amongothers).China'scivilexamsystemorkejuisonesuchinstitution.Astheearliestmeritocraticinstitutionintheworld(sincetheSongdynasty,c.960-1276),thecivilexamsystemwasaimedatrecruitinglearnedindividualstoserveinthebureaucracy.1Asascholarattainingthehighestqualicationinthecivilexam,thejinshiwouldbemadeascholar-ocial;apositionassociatedwithconsiderableprestige,andreceivedgenerouspecuniaryrewards.Becauseofthis,kejuattractedmillionsofzealousscholars|afervor
intensiedbythelackofrestrictionspla
intensiedbythelackofrestrictionsplaceduponacandidate'ssocialbackgroundandage.2Overtime,thecivilexaminstitutioncreatedadistinctgroupoflocaleliteswithdeeprespectforlearningandacademicachievements(Russell,1922)|a\culturaltrait"thathaspersistedevenlongaftertheexam'sabolishment.Inthispaper,weexaminethepersistentimpactofthecivilexamsystem|aninstitutionofrst-orderimportanceinChinesehistory|ontoday'shumancapitaloutcomes.Ourstudyisinspiredbythestrongpositiverelationshipbetweenaprefecture'ssuccessinthecivilexamasmeasuredbythenumberofjinshiintheMingandQingdynastiesandhumancapitaloutcomesasmeasuredbytheyearsofschoolingtoday(Figure1).3Ourspecicgoalsaretwofold.First,weattempttoverifywhetherthisrelationshipiscausal.Second,andconditionaluponprovingthecausalrelationship,weendeavortoi
dentifythevariouspossiblechannelsthrough
dentifythevariouspossiblechannelsthroughwhichthiseectendurestothisday,includingaculturalnormofvaluingeducation,schoolsoreducationalinfrastructuremoregenerally,socialcapitalviaclansandothersocialorganizations,andpoliticalelites.[Figure1abouthere]Usingthenumberofjinshiacrossthe278Chineseprefectures(normalizedbytheirpop-ulation)asameasureofthedegreeofsuccessinkejuexam(hereafterjinshidensity),thebaselineordinaryleastsquares(OLS)resultshowsthat,forevery10,000people,adoublingofjinshidensityisassociatedwithanincreaseinyearsofschoolingin2010of7%,whichis1EvidencetobearuponthemeritocraticfeaturesofChina'scivilexamsystemcanbefoundintheEuropeandiscourseamongthelikesofVoltaire,Quesnay,andChristianWol,allofwhomvieweditasasuperioralternativetothetraditionalEuropeanaristocracyintermsofg
overnance(Ford,1992).Foranempiricalanaly
overnance(Ford,1992).ForanempiricalanalysisoftheemergenceoftheSongcivilexamsystemseeChenandKung(2019).2Women,however,werenotallowedtotakepartinthecivilexam.3AprefecturewasanadministrativeunitbetweentheprovinceandthecountyinimperialChina.1notsmallinlightofthehugegeographicvariationsinjinshidensityacrosshistoricalChina.Thisresultisrobusttotheinclusionofarichgamutofcovariatesincludingaprefecture'seconomicprosperitybothintheMing-Qingperiod(populationdensity,urbanizationrate,andagriculturalsuitability)andtoday(usingnighttimelightsin2010asproxy),andnottheleastgeography(terrainruggednessanddistancetocoast);allwithprovincexedeects.ThevariationinjinshidensityintheMing-QingperiodacrosstheChineseprefecturesisclearlysubjecttomanyendogenousforces,withomittedvariablebiasbeingourprimar
yconcerns.Toaddresstheseissues,weconstru
yconcerns.Toaddresstheseissues,weconstructaninstrumentalvariableusingaprefecture'sshortestriverdistancetoitsnearestsitesofpineandbamboo|thetwokeyingredientsrequiredforproducinginkandpaperinwoodblockprinting.4Distancetothesetworawingredientsisimportantbecausetextbooksandexamaids(referencebooks),whichcontainnuanced,authoritativeinterpretationsoftheConfucianclassics,werecrucialtokejuexamsuccess.Wefurthershowthatourinstrumentalvariableisorthogonaltoahostofmeasuresofhistoricaleconomicprosperity,agriculturalsuitability,geography,andnighttimelightsin2010,anddemonstratethatthelocationsofpineandbambooforests,asnaturalhabitats,areexogenous.Thetwo-stageleastsquares(2SLS)regressionsproduceanestimateconsistentwiththatoftheOLSbutwithasomewhatlargerestimate(byabout24%).Thistimeadoublingofthenumbe
rofjinshi(per10,000people)isassociatedwi
rofjinshi(per10,000people)isassociatedwitha8.7%increaseintheaverageyearsofschoolingtoday.Thisimpliesthatanadditionaljinshiper10,000peopleleadstoanincreaseinschoolingof0.8yearswhenevaluatedatthemeanof8.712.5Inaccountingforthepersistentimpactofthecivilexamwersttestthehypothesisthattheexamhadbredacultureofvaluingeducationwhichwasinternalizedbyatleasttheeducatedelitefamilies(andaccordinglytransmittedacrossgenerations).Todosowetakeadvantageofanationallyrepresentativesocialsurvey,the2010ChineseFamilyPanelSurvey(CFPS),andconstructvariablestomeasuretheeectoftherespondents'ancestralachievementsinthejinshiexamontheirattitudestowardtheimportanceoflearningandeducationasadeterminantofsocialstatus.Aftercontrollingforbothgeneticandculturaltransmissionoffamilyhumancapital(asmeasuredbyvar
iousIQtestscores,yearsofed-ucation,andin
iousIQtestscores,yearsofed-ucation,andincome),wendthatancestraljinshidensity(measuredseparatelyalongthepatrilinealandmatrilineallines)hasasignicantlypositiveeectontheospring'sattitudetowardtheimportanceofeducationindeterminingsocialstatus.Moreover,theserespon-dentsalsowantthegovernmenttospendmoreoneducation.Thoserespondentswhowere4Ourinstrumentthustakestheaverageofthetwo(river)distancestoaprefecture'snearestavailablepineandbamboolocations.5AccordingtotheUnitedNations,thedierenceintheyearsofschoolingbetweenthelow-andmiddle-incomecountriesin2010wasamere1.4years,yettheydierinannualincomebymorethanthreetimes(2,109versus6,452USD).2parentsexpectedtheirchildrentoreceivemoreeducationingeneral,anddevotedmoretimetosupervisingtheirchildren'shomework(by,forexamp
le,watchinglessTVthemselves).Perhapsbeca
le,watchinglessTVthemselves).Perhapsbecauseofthisstrongerparentalethic,theirchildrenperformedbetterinclass,hadlowerratesofabsenteeism,andspentmoretimestudying.Alloftheseaugurwellwiththehypothesisregardingthetransmissionofacultureofvaluingeducationwithinfamilies.Evenifourinstrumentisvalidforidentifyingthecausaleectofhistoricaljinshi,chancesareitwouldhavemissedtheotherpossiblechannels.Thereasonisthatprefectureswithstrongerexamachievementswerelikelycorrelatedwiththefundingofbetterschoolsoreducationalinfrastructuremoregenerally,betterorganizedintheirprovisionoflocalpublicgoods,andmorecapableofproducingaclassofpoliticalelites,allofwhichaectedtheincentivestoacquireeducation.Itisthusalsoourgoaltoidentifytheseadditionalchannels.First,totheextentthatsuccessinthecivilexamdependedon
thequalityandquantityofschools,teachers,
thequalityandquantityofschools,teachers,andbooks,andthatprefectureswithgreatersuccessinthecivilexamhistoricallywouldbemoreeectiveinbuildingagoodeducationalinfrastructure,thelatterissurelyanotherlikelychannelofthepersistenteectdocumentedhere.Indeed,wendthatjinshidensityhasapositiveeectnotjustonthenumberofConfucianacademiesintheMing-Qingperiod,butalsoontheprimaryandsecondaryschoolsestablishedattheturnofthe20thcentury,andmoresignicantlyontheuniversitiesestablishedbothbeforethefoundingofthePeople'sRepublic(in1947)andinthemorerecenttimes(e.g.,2010).Anotherobviouschannelissocialcapital.Afterall,thosewhopassedthejinshiexaminvariablyenteredocialdomtojoinasmallgroupofeliteswhowouldlikelyinteractwithoneanotherinavarietyofsocialcontextsrangingfromclanandlineagea&
#11;airstocharita-bleorganizations(Elman
#11;airstocharita-bleorganizations(Elman,2000;McDermott,1997).Ashasbeenwelldocumented,socialcapitalisconducivetopublicgoodsprovision(Dell,Lane,andQuerubin,2018;Satyanath,Voigtlander,andVoth,2017).Usingmostlyhistoricalbutalsoasmallamountofcontempo-rarydata,wendthatjinshidensityhasasignicantlypositiveeectonthestrengthofclanorlineageorganizationasmeasuredbythenumberofgenealogyeditions(thestrongertheclanthemoreeditions).Likewise,ahigherjinshidensityisalsofoundtobeassociatedwithalargernumberofcharitableorganizationsin1840,andavarietyofsocialorganizationsincludingfarmers'associations,laborunions,chambersofcommerce,women'sassociation-s,educationalandstudentbodies,religiousassociations,charityorganizations,etc.,inthesurveyyearsofboth1935and2008.Allofthesespeaktothepossiblee
xistenceofasocialcapitalchannel.Byvirtue
xistenceofasocialcapitalchannel.Byvirtueoftheirentryintoocialdomuponexamsuccessthejinshiwerealsopoliticalelites.Asmanyofthemwouldreturntotheirhometownsandinvestintheeducationalin-frastructure,theyperpetuatedthesuccessoftheirdescendants.Usingdataonthreedierent3periodsinChinesehistory,wendthatjinshidensityhasasignicantlypositiveeectonthenumberofhigh-rankingocialsinboththelateQingandRepublicaneras,suggestingthepersistencealsoofapoliticalcapitalchannel.ButthiseectvanishedwhenChinaturnedtocommunismandthepoliticaleliteswerenolongerconnedtothosecomingfromaliteratibackground,therebyendingthetransmission.Inthelightofthesemultiplechannelsandthepossibilityoftheirwideningtheinequalitygapsviaunequalaccesstoeducation,wegobeyondthepersistenteectofthec
ivilexaminstitutiononeducationalattainme
ivilexaminstitutiononeducationalattainmentbyexaminingalsoitseectonsocioeconomicinequal-ityoverthelongrun.Usingsurveyevidence,wendthatjinshidensityactuallyhasasignicantlynegativeeectonsocioeconomicinequalitybetweenchildrenandtheirparents'generationintermsofyearsofschoolingandincome.Presumablybecauseastrongerlo-calkejuculturehashadthevirtuouseectofencouragingcompetitioninaprefecture,thesameinstitutionisrenderedmore\inclusive"inthatprefecturethaninoneswherethesameculturaltraitisnotasstrong.6ByexaminingthepersistentimpactofahugelyimportantChineseinstitutiononhumancapitaloutcomestoday,ourstudycontributesto,butinfactgoesbeyond,theliteraturethatidentiesimportantpatternsofculturalpersistencestemmingfromvariationsinhistor-icalinstitutions(e.g.,AlesinaandFuchs-S
chundeln,2007;Dell,Lane,andQuerubin,201
chundeln,2007;Dell,Lane,andQuerubin,2018;Guiso,Sapienza,andZingales,2016),aswellastotheliteraturethatdelvesintothehistor-icaloriginofculturaltraitsandtheirpersistence(e.g.,Alesina,Giuliano,andNunn,2013;Becker,Boeckh,Hainz,andWoessmann,2016;Grosjean,2014;NunnandWantchekon,2011;VoigtlanderandVoth,2012).Thatis,insteadoffocusingsolelyontheculturalchannel(asGuisoetal.'s[2016]havedoneinanexcellentmanner)butinadierenthistoricalcontext,wewentfurthertoidentifyanumberofadditionalmechanismsthatmightalsoexplaintheobservedpersistenceinquestion.Ourclaimofacontributionnotwithstanding,thisstudydoeshaveamajorlimitation.Asourcausallinkbetweenthecivilexaminstitutionanditsattendantculturaltraitofvaluingeducationisdrawnfromtheregionalvariationincivilexamsuccess,andbecauseboththecauseandthee
ectaremostsalientatthenationallevel,our&
ectaremostsalientatthenationallevel,ourndings|whilerobust|cannotbeestablishedasdirectevidencetobearuponhowthecivilexamsystemaectedChinaasawhole.7Theremainderofthispaperproceedsasfollows.Thenextsection(Section2)providesahistoricalbackgroundonthecivilexaminthecontextoflateimperialChina.InSection3weexaminethepersistenteectofkejuonyearsofschoolingtoday,whereasinSection46Weborrowtheterm\inclusive"fromAcemogluandRobinson(2012).7Wethankananonymousrefereeforthisinsight.4weidentifythecausaleectofkejuusinganinstrumentalvariableapproach.Section5goesbeyondtheculturaltransmissionchannelandexaminesanumberofmechanismsthroughwhichthelong-defunctkejuhasimpactedoncontemporaryhumancapitaloutcomes,whereasSection6examineskeju'seectonsocioeconomicinequalityoverthelongrun.Secti
on7oersaconclusion.2HistoricalConte
on7oersaconclusion.2HistoricalContext2.1CivilExamSuccessinLateImperialChinaEventhoughkejuwasalreadyconsolidatedintheSongdynasty(c.960-1276),itwasheldsparinglybytheMongolsintheYuandynasty(amere16times),hencethecivilexamsystemdidnotbecomefullyinstitutionalizeduntiltheMingdynasty(c.1368-1643)whentheHanresumedsovereignty.Thesystemthenlasteduntil1905|afewyearsbeforethelastimperialrule(theQingdynasty,c.1644-1911)cametoanend.China'scivilexamconsistedofthreelevels.Attheentrylevelwastheprefecturalexam,successinwhichledtothequalicationofashengyuan.Thenextlevelupwastheprovincialexam,whichonlythequaliedshengyuan(earnedbypassingaqualifyingexam)wereeligibletotake.Inthefortunateeventthattheypassedtheywereawardedajuren.Finally,onlythosewithajurenqualicationcouldtakethejinshiexa
m|thenalstageofthecivilexam.Jinshih
m|thenalstageofthecivilexam.Jinshiholderswereguaranteedapositioninmid-to-high-levelgovernmentadministrationand,assuchbecameamandarin|theenvyofmany.8FigureA1intheAppendixsummarizesthehierarchyofChina'scivilexam.RegardlessofitslevelChina'scivilexamwasessentiallyregulatedbyaquotasystem.Atthelowestlevel,theshengyuanquotawasapportionedtothecountiesandprefecturesonthebasisof:(1)thesizeofthecountyorprefectureocialschool,(2)populationsize,(3)taxobligations,and(4)pastexamachievements(Chang,1955).EssentiallytheQingdynastyinheritedtheshengyuanquotadistributionfromtheMingdynasty,withtheQingEmperorKangximakingonlyminoradjustmentstoitin1670afterwhichitremainedunchangedforsometime.Itwasonlyafterthe1850s,when,inanattempttoencouragetheregionalgovernmentstohelpsuppresstheTaipingrebels(c.18
50-1864),thatEmperorXianfengincreasedthe
50-1864),thatEmperorXianfengincreasedthequotatothosewhocontributed(Chang,1955).Whiletherewerealsoquotas8Forinstance,Ho(1962)showsthat,fromtheMingdynastyonwardsthejinshiqualication\automat-icallyplacedapersoninthemiddlestratumoftheocialdom"(p.26andp.120),whereasthejurencouldonlybeplaced\inthepostsofcountymagistrates,directorsandsubdirectorsofschools(attheprefecturalandcountylevels),andothercomparableoces"(p.27).BymiddlestratumHo(1962)referstoappoint-mentsrankedbetweenfourthandsixthoutofnineranks,whereasmagistratesandschooldirectorsatthecountyleveltypicallycommensuratewithranksseventhtoninth.5forthejurenandjinshi,unliketheshengyuantheywereapportionedtotheprovinces,which,aswiththecaseoftheshengyuanhadalsochangedlittleovertime(Shang,2004).Asthequotasystemwasstrictlydet
erminedbytheabovementionedfactorsand,inv
erminedbytheabovementionedfactorsand,inviewoftheirstabilityovermostoftheMing-Qingperiod,itwasunlikelythataprovinceorprefecturelobbiedformorequota.China'scivilexamhadthreedistinctcharacteristics.Foremostwasitsopenness.Kejuwasopentoallmalesregardlessofsocialbackground.Thismeansthata\commoner"|someonewhoseancestorshadneverpassedeventhelowestleveloftheexam|waseligibletositforthecivilexamsolongashepassedeachleveloftheexamintheestablishedsequence.9Ping-tiHo,aneminenthistorianofChina,indeedndsthatintheQingdynastyasmanyas45.1%ofjurenand37.2%ofjinshicamefromacommonerbackground(Ho,1962,pp.114-116).10Second,kejuwasrelativelyfreeofcorruption.Forinstance,topreventexaminersfromrecognizingaparticularcandidatethroughhishandwriting,allexamscriptswerehandcopiedrstandgradedbyasmanyaseightex
aminerswhowereoblivioustotheidentityofth
aminerswhowereoblivioustotheidentityofthecandidates(thecandidates'nameswereconcealed).Moreover,theexaminerswouldberemovedfromoceiftheywerefoundtohavefavoredaparticularcandidateintheirgrading,orevenfacedapossibledeathpenaltyifthisoccurredinthenalstageoftheexam(Shang,2004).Theseverityofthepenaltyislikelypowerfulenoughtodetercorruption.Lastbutnotleast,giventhatonewasalsoallowedtotaketheexamrepeatedly(inadditiontoitsopenness),China'scivilexamsystemwasextremelycompetitive.Passingratesfortheshengyuan,forexample,areestimatedatonly1-1.5%(Ji,2006),whereasthoseforthejurenandjinshiwereabout6%and17.7%,respectively.11Hence,inthemid-Qingthechancesthatsomeoneattemptingtheshengyuanexamwouldeventuallybecomeajurenandjinshiwereamere0.09%(1.5%*6%)and0.016%(1.5%*6%*17.7%),respectively|verysli
mindeed!Moreover,competitionhadintensi
mindeed!Moreover,competitionhadintensiedovertime,thankstotheexplosivegrowthinpopulationfromapproximately110millionintheMingto400millionintheQing(Ho,9Thankstoitsopennessandpositivepublicperceptions,atleasttwomillionmen(about2.5%ofthemalepopulationagedbetween15and49inthemid-Qingera)registeredforeachprefecture-levelexam,withatotalquotaforeachexamofaround30,000(BaiandJia,2016:684-685).Anotherestimatehasputthetotalnumberofliterati(dushuren)ataround20million,roughly5%ofthetotalpopulationinmid-Qing(Ji,2006).10Theextenttowhichsocialmobilitywasactuallyachievedunderthecivilexamsystemisasubjectofintensedebate,withsomescholarsarguingthatonlythosefrom\literatiormerchantfamilies,lineagesorclanswithsucientlinguisticandculturalresources"wereabletocompeteandsucceedinthecivilexam(Elman,20
00).Herewerefrainfromjoiningthisdebatebu
00).Herewerefrainfromjoiningthisdebatebutinsteadfocusonthefactthatsocialmobility,howeversmallitschancemaybe,wasindeedfeasible.11Ateachexamabout1,241jurenwouldbeselectedoutof20,600shengyuan,andabout220jinshiwouldbeselectedoutof1,250juren(calculatedbasedonourowndata).61962;Kuhn,1978).122.2RewardsfortheJinshiUnderthelastingin uenceofConfucianismtheocials(shi)inimperialChinawereheldinthehighestregard,andcivilexamwastheonlyroadtoocialdomforcommoners.13Whilemadeuponly2%ofthepopulation,thecivilexamscholarsaccountedforalmostaquarter(24%)ofthenation'sincome(Chang,1962).Thisexplainswhytheirsalariedincomewasabout16timesthatofacommoner.Butinrealitythedierencewassomuchgreater,assalariesaccountedforbutatinyportionofexamscholars'actualincomes.Forinstance,manywerefoundtohaveinveste
dinavarietyofbusinesseslikerealestate,ba
dinavarietyofbusinesseslikerealestate,banks,jewelryshops,andeventhemonopolytradeofsalt.Thus,thejinshiwas,forthemajority,theultimatequalicationtoachieve.Inadditiontothelucrativeeconomicrewardsthatcametheirwayasalearnedclass,additionalmotivationmayalsocomefromthevanityofholdingadegreeandthevarietyofritualisticrecognitionsbythecommunityandthenationatlarge.Forinstance,thenameofajinshiwouldberecordedinthelocalgazetteer,carvedonthemonumentofthelocalschoolandeventheConfucianTempleinthenationalcapital,andarches,gatewaysandtempleswouldbeerectedinhisname(Ho,1962).Forallthesereasons,incentiveswerestrongforonetoclimbthesocialladderinlaterimperialChinesesocietyandthecivilexamprovidedjustsuchapossibility.14Theseextraordinaryreturnstothejinshiqualicationforaslongasnearly600yearshavemost
likelyfosteredacultureofvaluingeducation
likelyfosteredacultureofvaluingeducationamongtheChineseovertime.Thedesiretosucceedinthecivilexamwassostrongthatpreparationforitbeganatthetenderageof6-7yearsold,whenchildrenweremadetorecitenolessthan2,000charactersfromtheThreeCharacterPrimer(sanzijing)|anancientConfuciantextbookforchildren|afterjustoneyearofstudy(Rawski,1979).12Thepercentagesofcandidateswhoobtainedtheshengyuan,jurenandjinshiqualicationsintheearlyQingwere5%,0.3%and0.053%,respectively(Ji,2006).13Thesocialhierarchywhichexistedbackthenconsistedoftheshi(ocials)attheverytop,followedbythenong(farmers),thegong(artisans),andtheshang(merchants)inthatorder.14Theyweresostrongthatevenrichmerchantswouldallegedlysendtheirsonsto\enterthecivilexam-inationandtorisehighinthebureaucracy"(Needham,1969,p.202).72015)|theonlyCh
inesehistoriantohaveprovidedpopulationda
inesehistoriantohaveprovidedpopulationdataattheprefecturelevelforvarioustimepointsspanningboththeMingandQingdynasties.16Toreduceskewnesswealsoraisethenumberofjinshidensitytothenaturallog.Toavoidlosingobservationswithavalueofzero,weadd1tothenumberofjinshi,i.e.,ln(1+jinshi/population).FigureA2intheAppendixshowsthegeographicdistributionofthejinshi,juren,andshengyuanquotadensityinthe278historicalprefectures.3.2BaselineControlsItisnecessarytocontrolforanumberofcovariatesthatarelikelytoimpactupontheyearsofschoolingtoday.Weclassifythemintotwocategories.Therstisavectorofbaselinecontrolvariables,denotedbyXci,inequation(1).NighttimeLights.Wecontrolforcontemporaryeconomicprosperityasmeasuredbytheaveragesatellitelightdensityatnighttimein2010attheprefecturallevel(Henderson,Storeygard,andWe
il,2012).17Geography.Wealsocontrolforsev
il,2012).17Geography.Wealsocontrolforseveralkeyfeaturesofgeography,mostnotablydistancetocoastandterrainruggedness.Distancetocoastisimportantbecauseprefectureslocat-edonthecoastwerelikelyearlybeneciariesofWesterntechnology,knowledgeandtrade,whereasterrainruggednesscanhaveaprofoundlylastingeectonlong-termeconomicdevel-opmenteitherthroughitsdirectimpactorthroughitsinteractionwithkeyhistoricalevents(NunnandPuga,2012).Distancetothecoastismeasuredasthedistancebetweenapre-fecture'scentroidtotheclosestpointonthecoast,whereasanindexofterrainruggednessisconstructedbycalculatingthedierenceinelevationbetweenadjacentcellgridsusingdataprovidedbytheUnitedStatesGeographicService(USGS).18ProvinceFixedEects.Tocontrolfortheeectoftheprovincialquotasofjinshi(andotherunobservede
ectsassociatedwiththeprovinces)onji
ectsassociatedwiththeprovinces)onjinshidensity,weincludeprovin-cialdummiesintheregressions.16ThedatapointsofpopulationinCao(2000,2015)are1391(Ming),1580(Ming),1776,1820,1851,1880,and1910.Wenormalizethenumberofjinshiineachprefecturebytakingtheaverageofthesesevendatapoints.Forrobustness,wealsonormalizethenumberofjinshibyaprefecture'slandarea(per10,000km2)andusetheactualnumberofjinshiasalternativemeasures.Bothyieldsimilarresults(notreported)tothejinshi/populationmeasure.17TheactualdataareobtainedfromtheGlobalDMSP-OLSNighttimeLightsprovidedbytheEarthObservationGroupsintheNationalCentersforEnvironmentalInformation.18TheDigitalElevationModel(DEM)istypicallyspacedatthe90square-metercellgridsacrosstheentiresurfaceoftheearthonageographicallyprojectedmap.9variousdensitiesofkejudegreeh
oldersaswedidinTable2,andobtainstrikingl
oldersaswedidinTable2,andobtainstrikinglyconsistentresults(columns(5)-(8)).[Table3abouthere]Alloftheseresultsclearlysuggestthat,ofthethreemeasuresofkejuculturejinshiisthemostappropriate,possiblybecauseitsculturalin uencesarethemostfar-reaching.23Infact,theresultsinTable3canbeveriedgraphicallyinFigureA2intheAppendix.Forinstance,thegeographicdistributionofjinshiisstrikinglysimilartothatofjuren(comparePanelAwithPanelB),butthatoftheshengyuanquotaisnot(PanelC);asmentionedearliertheshengyuanquotawasgeographicallydistributedtomaintainsocialandpoliticalstabilityinthoseregions|mostnotablythoseinthesouthwest|thattraditionallywerenotacradleofcivilexamsuccess(Ho,1962).4CausalIdenticationAprefecture'ssuccessinkejuexamwaslikelyassociatedwithacomplexmyriadoffactors.Althoughwehavealready
controlledformanypossibleconfoundingones
controlledformanypossibleconfoundingones,theremaystillbeomittedvariables|variablesthataresimultaneouslyassociatedwithbothhistoricaljinshidensityandyearsofschoolingtoday.Forinstance,prefecturesthathadproducedmorejinshimaybeassociatedwithunobserved(naturalorgenetic)endowments.Todealwiththeseconcerns,weemployaninstrumentalvariableapproach.4.1DistancetothePrintingIngredients(PineandBamboo)astheInstrumentalVariableofKejuTheFourBooksandtheFiveClassicswerethelynchpinofChina'scivilexam.Foronetosucceedinitonemustnotonlymemorizeitscontentsbutbeabletodemonstrateasolidunderstandingofthenuanced,authoritativeinterpretationsofthesetexts.Thisrequirednotjusttheavailabilityoftherelevanttextsbut,24importantly,alsoalargeclusterofreferencesthatexplainedthesenuancesandtaughtthetricksoftheeight-leggedess
ay(Ho,1962;McDermott,2006).25However,the
ay(Ho,1962;McDermott,2006).25However,thefactthatonly19printingcentersweredistributedacross23Forthatreason,wewillemployonlyjinshidensityasourmeasureofkejucultureintheremaininganalysis.24CountyandprefectureschoolsreliedonthecentralgovernmentforcopiesoftheFourBooksandFiveClassics,butstudentshadtoresorttotheprivatebooksellersfortheirowncopies(McDermott,2006).25Sishujicheng(CollectedCommentariesontheFourBooks),forexample,was\anessentialtextforexaminationstudy"(Brokaw,2007,pp.97-98).AnotherpopularreferencewasSishubeizhi(FullPurportoftheFourBooks).13the278prefectures,andthatthese19centersaccountedfor80%ofthe13,050textspublishedduringthatperiod(ZhangandHan,2006),26meantthataccesstoreferencebooksvariedenormouslyfromoneprefecturetoanotherinviewofprohibitiveoverlandtransportcosts.27Whatthen
determinedthegeographicavailabilityofboo
determinedthegeographicavailabilityofbooks?Toaddressthisquestion,weturntothelocationofthetwomainingredientsrequiredfortraditionalChinese(woodblock)printing|pineandbamboo|forclues.Fromaroundthe14thcenturyonwardstheprevailingprintingtechnologyinChinareliedprimarilyonpineandbambooforproducinginkandpaper(Brokaw,2007;Tsien,1987;ZhangandHan,2006).Toeconomizeontransportcosts,majorprintingcenterswerelikelylocatedincloseproximitytothepineandbamboohabitats(Figure3A).28Forinstance,whileTingzhouandJianningprefecturesofFujianProvincedidnothavetheirownpineandbambooforests,thematerialscouldbesourcedfromthenearbyprefectures(Brokaw,2007),asillustratedinFigure3B.Againstthisstylizedfact,distancetothenearestpineandbamboositeswouldmakeareasonableinstrumentinourcase.Moreover,thankstothewell-connectedriv
ertributariesinthelowerYangtzedeltaregio
ertributariesinthelowerYangtzedeltaregion,themainingredientsrequiredformakinginkandpaperweremostlyshippedalongtheriver(Figure3C).29[Figure3abouthere]Thusthelogicbehindourinstrumentisasfollows.Totheextentthataprefecture'ssuccessinthecivilexamwascloselyrelatedtoprintingoraccesstobooks,thatthemajorprintingcenterswerelocatedincloseproximitytothepineandbamboohabitats,andthattheingredientsrequiredforprintingweretransportedthroughthemainnavigablerivertributaries,aprefecture'sriverdistancetoitsnearestbambooandpinehabitatswouldbea26PrintedbooksrefertothetotalnumberofbookcataloguesprintedinaprefectureduringtheentireMing-Qingperiod.ThedataareobtainedfromDuandDu(2001,2009).Theremaining20%ofbookswerepublishedinnearbylocalitieswiththepublishershiringbothtechnicalpersonnelandprintingmachinesfrom
thesecenters(ZhangandHan,2006).27Toillus
thesecenters(ZhangandHan,2006).27Toillustrate,thesamebooksintheimperialcapitalofBeijingfetched1.7timeshigherpricesthaninthelowerYangtze,and3.3timeshigherpricesthaninFujianProvince(McDermott,2006).Eventhoughbookscouldbeshippedfromthesouthallthewayuptothenationalcapitalinthenorth(HuzhouGazetteer),eectivelydistributingthemposedanotherobstacle.28Outofthe19printingcentersonlytwo|HangzhouandNingboinsoutheasternChina|hadbothpineandbamboohabitatsintheirownprefectures.Oftheremaining17printingcenters,11hadneitherpinenorbamboo;theremainingsixhadeitherpine(4)orbamboo(2),butnotboth.Inshort,all17ofthemhadtorelyonothernearbyprefecturestosupplythemwitheitheroneifnotbothofthenecessaryingredientsforprinting.29Forinstance,Yasushi(2014)ndsthat,afterfellingthepinetreesintheforeststhemerchants
inHuizhouprefectureofAnhuiProvinceshippe
inHuizhouprefectureofAnhuiProvinceshippedtheirtimbertotheprintingcentersalongtherivertributariesofthelowerYangtzedelta.Likewise,inwesternFujianProvincelocalmerchants\cutandprocessedthewoodand oateditdowntheYinRiverforsaleinChaozhou,Foshanzhen,andGuangzhou"(p.43),and\eachboattypicallycarriedthirty'loads'(dan)ofpaper(thatis,about30,000sheets,weighing1.82metrictons)"(Brokaw,2007,p.116).14feasiblevariabletoinstrumenttheendogenousjinshidensity.InformationonthegeographiclocationsofpineandbambooforestsisobtainedfromtheProvincialGazetteers(tongzhi)compiledduringtheMing-Qingperiod.Thepublicationcontainsachapteronthelocalproduce(wuchan)whichincludedspeciesoflocalplantsandanimalsaswellasmajorhandicrafts.Weconsideraprefectureashavingapineand/orbamboohabitatonlyifitissucientlysizeabletobe
mentionedinthelocalproducechapterinthePr
mentionedinthelocalproducechapterintheProvincialGazetteers.Thisrestrictsthenumberofpinehabitatstojust27andbambooto65amongthe278Ming-Qingprefectures.30UsinghistoricalGISdataonthemajorinlandriversidentiedbyMatsuura(2009)astheonesusedforshippingmerchandisesinQingChinafromHarvardCHGIS,31wecomputeaprefecture'sshortestriverdistancestoitsnearestbambooandpineforests.Wethentaketheaverageofthetwodistancesasourinstrumentalvariable.Weexaminetheimportanceofpineandbambooforprintingandaccordinglyapre-fecture'ssuccessinthecivilexaminTable4.Webeginbyregressingjinshidensityonthenumberofprintedbooksrst,andndthemtobehighlysignicant(columns(1)and(2)).32Wethenexaminewhetherriverdistancetothetwomainingredientsforprintingissignif-icantlycorrelatedwithwhetheraprefecturehadamajorprinting
center(columns(3)and(4))andthenumberofpr
center(columns(3)and(4))andthenumberofprintedbooks(columns(5)and(6)),andconrmtheirnegativeandsignicantrelationships.Finally,weshowthatourinstrumentisnegativelyandsignicantlycorrelatedwithjinshidensity(columns(7)and(8)).Takentogether,theseresultsconrmthehypothesizedimportanceofpineandbambooindeterminingtheavailabilityofbooksand,asacorollary,aprefecture'ssuccessinkejuexams.[Table4abouthere]4.2ExclusionRestrictionsAuniqueadvantageofusingthelocationsofbambooandpineasinstrumentisthat,ashabitatstheirgeographicdistributionswereexogenouslygiven.Historiansndlittleifanyevidenceofplantingpineandbamboointentionallyforthepurposeofcommercialprinting(Elvin,2004).33Whilethelocationsofpineandbambooforestswereexogenouslydeter-30Whilepineandbamboohabitatsmayalsoexistinprefec
turesotherthantheonesdocumentedinthegaze
turesotherthantheonesdocumentedinthegazetteers,theyweretoosmallforsupplyingtheprintingindustry.31Hissourcesincludehistoricalandgovernmentarchives,travelers'notesaswellasbusinessguidebooks.32Whileprintingwascrucialforexamsuccess,itwasunlikelytohavebecomeaspopularasitdidhadtherebeennocivilexam,orhadthecivilexambeenanylessimportant.TheimportanceofkejuforprintingiswellrecognizedbyChinesehistorians(see,forexample,Cao,2013;Zhang,2010).33EventhemostsuccessfulwoodmerchantsinHuizhouprefectureofAnhuiProvincepreferredtolumbertreesfromprimordialforestsratherthanplantingtheirown(Li,2000).15Toruleoutthatpossibilitywecontrolforaprefecture'sshortestdistancetoitsnearestmajornavigableriver(denedasthegreat-circledistancemeasuredfromaprefecture'scentroidtothenearestpointonthemajornavigableriver)
intheMing-Qingperiod,togetherwiththefull
intheMing-Qingperiod,togetherwiththefullsetofcontrolvariablesincolumn(2).Inbothcolumns,wendthatthereducedformestimatesaresignicantlynegative;ashypothesized,thefartherawayaprefecturewasfromthebambooandpinehabitatsthelowertheyearsofschoolingtoday.Incolumns(3)and(4)weemploytheshareofpopulationwith\highschool"and\universityandabove"educationasthedependentvariableandndconsistentresults.Wenowreportthe2SLSestimatesincolumns(5)and(6)ofTable5.Theresultsrevealthattherst-stageIV-estimatesareallsignicant.TheF-statisticsuggeststhatourinstrumentalvariableisbynomeansaweakinstrument.Moreimportantlythesecond-stageIVestimatesarealsosignicant.Usingtheaverageofthetwoshortestriverdistancestopineandbambooasinstrument,wendthatjinshidensitysignicantlyandpositivelyex
plainsyearsofschooling.35Theinstrumented
plainsyearsofschooling.35Theinstrumentedjinshicoecient(0.087,column(6)ofTable5)issomewhatlargerthantheOLSestimate(0.07,column(3)ofTable2),suggestingthattheendogenousjinshimeasurelikelyhastheeectofbiasingtheestimationtowardszero,andthusunderestimatedthelong-termeectofkejuoncontemporaryhumancapitaloutcomes.Theinstrumentedresultsuggeststhatadoublingofjinshidensity(per10,000people)isassociatedwithan8.7%increaseinaverageyearsofschoolingin2010.Thisimpliesthatanadditionaljinshiper10,000peopleleadstoanincreaseinyearsofschoolingof0.8whenevaluatedatthemeanof8.712years.ComparedwiththeOLSestimatesofTable2whenthedependentvariabletakesontheshareofpopulationwith\highschool"and\universityandabove"education(columns(7)and(8)),theIV-estimatesareconsistentlysignicant,withslightlyla
rgermagnitudes(byabout14%and29%).[Table5
rgermagnitudes(byabout14%and29%).[Table5abouthere]5AccountingfortheChannelsofKejuPersistence5.1TransmissionofKejuCultureCultureisinherentlystable,andassuchitislikelytotransmitfromonegenerationtothenextoveralongperiodoftime(AlseinaandGiuliano,2015;BoydandRicherson,35Inlightofthefactthatprintingcentersinthenorthmayrelylessonrivertransport,wealsousedgreat-circledistanceinplaceofriverdistanceasinstrumenttoruntheregressionsagainandobtainedsimilarresults(notreported).171985;GiulianoandNunn,2017;VoigtlanderandVoth,2012).Premisedonthisreasoning,althoughkejuhaslongbeenabolished,thecultureofvaluingeducationthatitfosteredhaslikelypersisted.BertrandRussell's(1922)insightfulremarkinTheProblemofChinahintedpreciselyatthat:\atanyrate,forgoodorevil,theexaminationsystemprofoundlyaectedtheciv
ilizationofChina.Amongitsgoodeectsw
ilizationofChina.Amongitsgoodeectswereawidely-diusedrespectforlearning...".Conceptually,economistshavedevelopedaframeworkforanalyzinghowculturaltraitscouldbetransmittedverticallyacrossgenerationswithinthefamilycontextthroughparentalindoctrinationandinput(Becker,1991;DoepkeandZilibotti,2014;Guryan,Hurst,andKearny,2008).36InthespeciccontextofChina,the\widely-diusedrespectforlearning"iscommonlyunderstoodas\jiaxueyunyuan"or\shuxiangmendi",meaning\alongtraditionoflearningintheliteratifamilies".WithinEastAsiawherethesameConfuciannormissimilarlyprevalent,evidencesuggeststhatstudentsspentsignicantlymoretimestudyingthaneithertheirAmericanorEuropeancounterparts(Rozman,2014;StevensonandLee,1990)andinvestedmoreresourcesinprivatetutoring(ChaoandTseng,2002).Totestwhether
thelong-termpersistenteectofkejuisd
thelong-termpersistenteectofkejuisdue(atleastinpart)totheverticaltransmissionofculturewithintheeducatedelitefamilies,weturntothe2010ChineseFamilyPanelSurvey(CFPS),anationallyrepresentativesurveyconductedbytheInstituteofSocialScienceSurveyofPekingUniversity.37Auniqueadvantageofthissurveyisthatitconsistsofbothanadultsample,whichinturnhasasubsampleconsistingofonlythosewhoareparents,andachildsampleofthoseaged16andunderbutover7.Inourempiricalanalysiswedrawuponallthreesamplesinordertomoreaccuratelycapturetheculturaltransmissionchannel.Werstmakeuseoftwoattitudinalquestionsintheoverall(adult)sampletomeasurethestrengthofcultureofvaluingeducationamongthe42,590respondents.Therstisadummyvariableindicatingwhetherarespondentregardseducationas\themostimportantdeterminantofsocialsta
tus",whereasthesecondisalsoadummyvariabl
tus",whereasthesecondisalsoadummyvariableindicatingwhetherone\prefershis/hergovernmenttoprioritizespendingoneducation"(amongalonglistofpublicexpenditures).38Tocheckrobustnesswealsodrawupontheparentsample;those36Inadditiontotransmittingtheirownhumancapital(genetically)totheirospring,typicallyparentsalsopassontheirownvaluesthroughparentalinput(e.g.,Becker,1991).Forinstance,withapremiumplaceduponliteracy,bookishness,andeducation,theJewshavemostpersuasivelydemonstratedthisprocessofpassingskillsandattitudesontolatergenerations(BotticiniandEckstein,2012).37Thesurveycovered14,960householdsin25provinces.Refertothecenter'swebsite(http://www.isss.edu.cn/cfps/)forfurtherdetails.38Thelistcontainsuptotencategoriesofexpenditure:environmentalprotection,medicalcare,policeandlawenforcement,ed
ucation,sportandrecreationfacilities,pen
ucation,sportandrecreationfacilities,pension,socialinsurance,jobscreation,18respondentswithchildrenaged12andunderwereaskedabouttheirexpectationsfor,andinputin,theirchildren'seducation.Thesequestionsinclude1)\theyearsofschoolingtheirchildrenwere(ideally)expectedtoreceive",2)\whethertheyoftengiveupwatchingTVinordertoaccompanychildren",and3)\thetotalhourstheyspentontutoringchildren'shomeworkperweek".Additionally,theinterviewerwasaskedtoindicateonave-pointscale\whetherparentseectivelycommunicatedwith(their)children".Togaugethefamilialtransmissionofkejuculture,weconstructtwolinearvariables|thepatrilinealandmatrilinealancestors'jinshidensity|toproxyfortheachievements(ifany)ofthesurveyrespondents'ancestorsinagivenpopulationwhohadobtainedajinshiquali-cation.Toidentifythem,w
ematchthesurveyedindividualswiththeirpat
ematchthesurveyedindividualswiththeirpatrilinealancestorsbasedonthesurnameandhometown(prefecture)information,undertheassumptionthatpeopleborninthesameprefectureandsharingthesamesurnamearelikelytoberelatedalongthepatrilinealline(Clark,2014).Toillustratethisidea,supposetherewere90jinshiwiththesurnameKunginSuzhouprefectureintheMing-Qingperiod.GiventhepopulationinthisprefecturewiththesurnameKungtodayis34,000,thenormalizedjinshidensityforpatrilinealancestorshavingthesurnameKunginSuzhouprefectureisthus0.0003.Simi-larly,usingthemaidensurnameandhometowninformationoftherespondent'smother,werepeatthesameproceduretoconstructthevariableonmatrilinealancestors,i.e.,themalejinshifromthemother'sfamily(e.g.thematernalgrandfatheranduncles).Toseparateculturaltransmissionfrominheritedability,wefurther
controlforthere-spondents'memoryandlogic
controlforthere-spondents'memoryandlogictestscores,andtheireducationallevelandincome,alongsideabatteryofcontrolsfortheirage,gender,ethnicityandresidentialstatus(ruralversusurban),inadditiontoalltheprefecturalcontrols.TheresultsarereportedinPanelAofTable6.Ancestraljinshidensityhasasignicantlypositiveeectontherespondents'attitudetowardseducation.Specically,thosewithmorejinshiancestorsaremorelikelytovieweducationas\themostimportantdeterminantofsocialstatus"|resonatingwiththeso-called\exam-to-mobility"normthatundergirdedthekejusystem(column(1)).Consistently,theserespondentsalsopreferthegovernmenttoprioritizespendingoneducation(column(2)).Intheparentsample,ancestraljinshidensityissignicantlycorrelatedwiththenumberofyearsforwhichparentsexpectedtheirchildrentoattendsch
ool(column(3))andtheamountoftimeande
ool(column(3))andtheamountoftimeandeortthatparentsdevotedtosupervisingtheirchildren'shomework(columns(4)-(5)).Ancestraljinshidensityisalsoconsistentlysignicantlyassociatedwiththeinterviewer'ssubjectiveevaluationofhowoftenparentscommunicatedwiththeirchildren(column(6)).unemploymentrelief,andculturalandartisticactivities.19[Table6abouthere]Tofurtherverifytheculturaltransmissionchannel,wedrawuponthechildsampletoexaminetheeectofancestraljinshidensityonchildren'stestscores(wordandmathabilitytests)andtheircognitiveandnon-cognitiveperformance(PanelB).AswiththespecicationinPanelA,wecontrolfortheirinheritedability(memoryandlogictestscores)andfamilyin uence(parents'yearsofeducationandincome).Thebaselineresultsshowthatchildrenwithmorejinshiancestorsindeedperformedbetteri
nboththewordandmathabilitytestsirrespect
nboththewordandmathabilitytestsirrespectiveofthenumberofcontrols(columns(7)-(10)).Concerningcognitiveskills,childrenwithmorejinshiancestorsarealsomorelikelytopasstheirexamswith yingcolors(column(11));andinconnectionwithnon-cognitiveperformance,thesestudentsalsotendtobeabsentfromclasslessoftenandtospendmoretimestudying(columns(12)-(13)).Perhapsaremainingconcernofthisanalysisisthat,comparedwithEnglandandWales,whichaltogetherhad270,000surnamesin2002(ClarkandCummins,2015),theChinesehavefarfewersurnames|approximately4,100excludingtheethnicminoritiesbasedonthe1%minicensusof2005.Forexample,thetopvesurnamesofLi,Wang,Zhang,Liu,andChentogetheralreadyaccountedfor34.09%oftheentirepopulation.Totheextentthatthosesharingthesamesurnameinthesameprefecturemayactuallycomefromdierentlineageg
roups(thesurnameLiisagoodcaseinpoint),th
roups(thesurnameLiisagoodcaseinpoint),thismayrendertheancestralmeasureoffamilyculturaltransmissionmuchlessaccurate.Toaddressthisconcern,werestrictoursampletoonlytheraresurnames,asthelikelihoodofthosepossessingthesamesurnamelivinginthesameprefectureandsharingthesameancestorswouldbedistinctlyhigher.Inthe1%minicensusof2005,wethusemployonly2,265outofthe4,100surnames,whichaccountedforjust20%oftheChinesepopulation.ReportedinAppendixTableA3,theresultsofusingraresurnamesarestrikinglyconsistentwiththoseoftheadultandparentsamples(PanelA).Inthechildsample,theeectofculturaltransmissiononcognitiveperformanceandthatonnon-cognitiveperformancearealsoconsistentlysimilar,whilethoseonthewordandmathabilitytestsarelesspreciselyestimatedduetothelargerstandarderrorsresultingfromamuchsmallersample.5.
2EducationalInfrastructureInadditiontocu
2EducationalInfrastructureInadditiontoculture,educationalinfrastructuremayrepresentyetanotherchannelthroughwhichkejuproducesapersistenteectoncontemporaryhumancapitaloutcomes.39In39Thepositiverelationshipbetweeneducationalinfrastructureandschoolingoutcomeshasbeenwelldocumented(e.g.,Du o,2001).20becauseitis\essentialtotheexistenceofalineage"(Bol,2008,p.241),somuchsothatresourcefulclanstendtorevisetheirgenealogiesmorefrequentlyinordertostrengthenthesenseofbelongingandhonor(Watson,1982).ThedataongenealogyareobtainedfromZhongguojiapuzongmu(AComprehensiveCatalogueontheChineseGenealogy),whichcontainsthegenealogiesofupto52,401clansandinvolvingmorethan700surnamesacross280prefecturesinChina(ShanghaiLibrary,2009).Weusetwomeasurestoproxyforsocialcapital.Forthelateimperialperiodweenu-mer
atethecharitableorganizationssuchasthose
atethecharitableorganizationssuchasthoseengagedinrelievingfamineandrunningorphanagesintheQingdynastyasproxies.ThedataareobtainedfromtheHarvardChinaMap.42Toconstructtwoseparatemeasuresofsocialorganizations,wealsoenu-merateabroadrangeofnon-protsocialorganizationslikefarmers'associations,laborunions,chambersofcommerce,women'sassociations,educationalandstudentbodies,reli-giousassociations,charitableorganizationsandsoforthinboththeRepublicanperiodandin2008(obtainedrespectivelyfromacensussurveyofsocialorganizationsin1935andtheeconomiccensusconductedin2008).TheresultsarereportedinTable8.First,wendthatjinshidensityhasasignicantlypositiveeectonthenumberofgenealogies(columns(1)and(2)),whichatteststothestrongrelationshipbetweencivilexamsuccessandsocialcapitalembodiedinthec
lanorganization.Likewise,jinshidensityal
lanorganization.Likewise,jinshidensityalsohasasignicantlypositiveeectonthenumberofcharitableorganizationsinthelateQing(columns(3)and(4)),aswellasonthenumberofsocialorganizationsin1935and2008(columns(5)-(8)).Together,theseresultsprovidesuggestiveevidencethatsocialcapitalrepresentsanotherlikelychannelthroughwhichthecivilexamhasimpacteduponhumancapitaloutcomesovertime.[Table8abouthere]5.4PoliticalElitesUponpassingthecivilexamthejinshiwasgrantedentryintoocialdom,withsomebe-cominghigh-levelocials(atthelevelofprovinceandabove).43Likeclans,whichallegedlyhelpedmemberssucceedinthecivilexamviatheunevenresourcestheyprovided,politi-calelitestoocouldbearuponaprefecture'sexamoutcomesbyin uencingthedistributionofeducationalresources(e.g.sizeoftheacademy)totheirhometowns.Toacco
untforthischannel,weconstructameasureofp
untforthischannel,weconstructameasureofpoliticalelitesforeachoftheimperial,Republi-canandcontemporaryperiods.TomeasurethestrengthofpoliticalelitesinlateQing(c.42AnonlineGISdatabasefromhttps://worldmap.harvard.edu/chinamap/.43Attheminimumajinshiholderwouldbecomeacountymagistrate(JiangandKung,2019).23educationalandconsequentlyincomelevelofvarioussocialgroups.45However,thereversecouldhavebeentrueiflegaciesofthecivilexamhaveastrongereectforsome|especiallythewealthier|groupsthanothers.Ifthatisthecase,kejupersistencewouldreinforcetheexistinginequalitybetweenthesesocialgroups,resultingindecreasingsocialmobility.Totesttheeectofkejupersistenceonsocialmobilitywefocusouranalysisoninter-generationalmobility|specicallyeducationalandincomemobilityacrossgenerations.Us-ingthe1%min
icensusof2005asoursample,wecomparetheadu
icensusof2005asoursample,wecomparetheadultrespondents'educationalandincomeattainmentswiththeirparents'attainments.Incolumn(1)ofTable10,wendthatbothjinshidensityandfamilybackground,specicallytheparents'educationalattain-mentandhousingwealthbutnottheirincome,havesignicanteectsonone'seducationaloutcome.Toexaminetheeectofkejuoneducationalmobilityacrossgenerations,wefurtherinteractjinshidensitywithparents'educationalattainmentincolumns(2)to(3).Wendthatinprefectureswithastrongerkejuin uence,parents'educationalbackgroundhaslessstronganeectonanindividual'seducationaloutcome,suggestingthepositiveeectofkejupersistenceonsocialmobilityinthelongrun.Tocheckrobustness,wereplaceeducationalmobilitywithincomemobilityincolumns(4)to(6),andobtainsimilarresults.[T
able10abouthere]7ConclusionsByobservinga
able10abouthere]7ConclusionsByobservingastrong,positiverelationshipbetweenprefectureswithastronghistoricaltra-ditionofexamsuccessandyearsofschoolingtoday,wehaveshownthepersistentimpactofaremarkablylong-livedinstitution|China'scivilexamsystem|onhumancapitaloutcomesorspecicallyeducationalattainmenttoday.Moreover,wehavefurthershownthatthisre-lationshipislikelycausalbyusing,asinstrument,theaverageriverdistancetoaprefecture'snearestpineandbambooforests,whichproducethemainingredientsforprinting.Whilewehaveidentiedimportantpatternsofculturalpersistencestemmingfromvari-ationsinhistoricalinstitutionsintheChinesecontext,ourprimarygoalistoemphasizethemultiplicityofchannelsbeyondthatofculture.Indoingsowehaveprovidedevidencetoshowthateducationalinfrastructure,socialcapital,andtoalesse
rextentpoliticalelitesallfacilitatedthei
rextentpoliticalelitesallfacilitatedtheimpactofChina'scivilexaminstitutionontheculturaltraitofvaluing45ThecivilexamhadallegedlyprovidedsocialmobilityinboththeSongandtheMing-Qingdynasties(Kracke,1947;Ho,1962).TheimportanceofperceivedupwardmobilityprovidedbythecivilexamsystemisemphasizedbyBaiandJia(2016).Byexploitingthevariationsinthequotasonentry-levelexamcandidatesacrossChina,theseauthorsarguethattheprobabilitythatsomeonewouldparticipateinarevolutionin1911aftertheabolitionofthekejuexamsystemwassignicantlyhigherinprefectureswithhigherquotaspercapita.25Figure3B.TwoExamplesofPrintingCenters'ProximitytoPineandBambooLocations35Figure4.EectsofRiverDistancetoPineandBambooLocationsonJinshiDensity,byPeriod37FigureA2.RegionalDistributionsofJinshi,JurenandShengyuanQuotaDensities