THE INFLUENCE OF ANCESTRAL LIFEWAYS ON INDIVIDUAL ECONOMIC OUTCOMES IN D o w SUB-SAHARAN AFRICA nlo a d e d fro m h ttp s SteliosMichalopoulos LouisPutterman ://a c BrownUniversity BrownUniversity a d e m DavidN.Weil ic .o BrownUniversity u p .c o m /je e a /a Abstract d v Does a person’s historical lineage influence his or her current economic status? Motivated by a an c large literature in the social sciences stressing the effect of an early transition to agriculture on e -a currenteconomicperformanceatthecountrylevel,weexaminetherelativecontemporarystatusof rtic individualsasafunctionofhowmuchtheirancestorsreliedonagricultureduringthepreindustrial le era.WefocusonAfrica,where—bycombininganthropologicalrecordsofgroupswithindividual- -a b levelsurveydata—wecanexploretheeffectofthehistoricallifewaysofone’sforefathers.Within stra enumerationareas(typicallyasinglevillageorgroupofvillagesinthecountrysideandacityblock c inurbanareas)aswellasoccupationalgroups,wefindthatindividualsfromethnicitiesthatderived t/d o a larger share of subsistence from agriculture in the precolonial era are today more educated and i/1 0 wealthy.Atentativeexplorationofchannelssuggeststhatdifferencesinattitudesandbeliefsaswell .1 0 as differential treatment by others, including differential political power, may contribute to these 9 3 divergentoutcomes.(JEL:O15,N37,N97,J6,Z1) /je e a /jv y 0 1. Introduction 2 9 /5 1 0 Economistsgenerallyagreethathistorymattersinexplainingvariationsinthestandards 75 4 of living among people. But what aspects of history should we be looking at? Two 7 b y S c ie n c TheeditorinchargeofthispaperwasDirkKrueger. es L Acknowledgments:Wewouldliketothanktheeditorandtwoanonymousrefereesfortheirveryuseful ibra comments.DanielPrinz,DevinStreur,YoungMinKim,PhilipTrammell,NickReynolds,AnnaCroley, ry andespeciallyYunanJiprovidedsuperlativeresearchassistance.WearealsogratefultoBillEasterly,Filipe u s Campante,NicoVoigtlaender,andseminarparticipantsatCambridgeUniversity,OxfordUniversity,and er o theUniversityofCopenhagenandtoparticipantsattheconference“Deep-RootedFactorsinComparative n EconomicDevelopment”atBrownUniversity,theCESifoworkshoponDemographicChangeandLong 08 RunDevelopment,theBCEPconferenceatBerkeley,theNYUDRI2015AnnualConference,andthe O c NBERPoliticalEconomygroupforhelpfulcomments.MichalopoulosisaResearchAffiliateatCEPR to b andaResearchAssociateatNBER.WeilisaResearchAssociateatNBER. e r 2 E-mail:Stylianos [email protected](Michalopoulos); 0 1 Louis [email protected](Putterman);David [email protected](Weil) 8 JournaloftheEuropeanEconomicAssociation 2018 0(0):1–59 DOI:10.1093/jeea/jvy029 (cid:2)c TheAuthor(s)2018.PublishedbyOxfordUniversityPressonbehalfofEuropeanEconomicAssociation. Allrightsreserved.Forpermissions,pleasee-mail:[email protected] 2 JournaloftheEuropeanEconomicAssociation of the most important are the history of the place where the individual lives and the historyofhisorherownlineage. D Of these two branches, the study of how historical events in a given place shape ow n economicoutcomesisthebetterdeveloped.Inlargepart,thisisbecauseitisrelatively lo a easy to map the locations of historical events to modern-day territories. If we know de d that something happened in one place and not another—for example, on one side of fro m a border but not the other—we can compare contemporary outcomes of these two h places, and thus learn about the role of whatever it was that differed. This strategy ttp s has been particularly fruitful in examining the role of institutions, which have the ://a c helpful property of tending to stay put in physical locations.1 A slight variation on a d e thisliteratureonpersistenceinplacesallowsformovementsoflargegroupsofpeople m ic from one place to another, recognizing that when these large-scale migrations take .o u place, people maybring withthem muchofwhateveritis—culture, institutions, and p.c o soforth—thatwasfoundintheirplaceoforigin.2 m Among the various place-based determinants of comparative development, the /jee a transition to agriculture is often cited as being of paramount importance in fostering /a d v thedevelopmentofmodernurbancivilizations.Thisideaisattheheartofavenerable a n c line of research among anthropologists and historians embracing social evolutionary e -a schema.EconomistsHibbsandOlsson(2004),Putterman(2008),andBorcan,Olsson, rtic and Putterman (2018) establish empirically a positive influence of early agricultural le-a b transitiononstateformationandcontemporaryincomesacrosscountries.Thisplace- s tra centered perspective on history points naturally toward thinking about aggregate or c t/d averageincomesinaparticularcountryorregioninthemodernworld. o i/1 The other approach to quantifying the role of history looks at heterogeneity in 0 .1 outcomeswithinapopulation.Thefocushereisonthelineageofanindividualandhow 0 9 3 this contributes to his or her relative economic standing and cultural attributes today /je e (Fenske2013;Alesina,Giuliano,andNunn2013).Overperiodsofafewgenerations, a /jv the effects of one’s lineage on current outcomes are addressed under the heading of y 0 2 intergenerational mobility. However, economists are increasingly realizing that there 9 /5 areelementsoflineagethatareimportantbeyondwhatcanbeunderstoodfrom,say, 10 7 aonegenerationaltransitionmatrix.Recentattemptstolengthentheintergenerational 54 7 horizon include work by Clark (2015) and Guell et al. (2015) who use surnames b y to track family-level economic performance over several generations, finding that S c ie intergenerationalmobilityisratherlow. n c e Inthispaper,ourgoalistoapplythissecond,lineage-basedapproach,inacontext s L where we can link lineages all the way back to variation in ancestral “lifeways”, ib ra that is, forms of economic support before the advent of the modern industrial era. In ry u particular, the question we ask is whether tracing one’s lineage to predominantly se farming communities in the preindustrial period is beneficial in the modern era, r o n 0 8 O c to b 1. Acemoglu,Johnson,andRobinson(2001,2002),BanerjeeandIyer(2005),Iyer(2010),Dell(2012), e andMichalopoulosandPapaioannou(2013,2014). r 2 0 1 2. PuttermanandWeil(2010)andAbramitzky,Boustan,andEriksson(2014). 8 Michalopoulos,Putterman,andWeil InfluenceofAncestralLifeways 3 specifically the world of urban and rural sub-Saharan Africa. Notably, since we are looking at variation in individual economic outcomes within specific locations, the D stories about how agricultural history shapes place-based differences in economic ow n performance,discussedpreviously,arelessapplicableinourframework. lo a Awell-knownsocialevolutionaryapproachholdsthathumansocietiesprogressed de d fromhunter-gathereroriginstoindustrialmodernityviathedevelopmentofsedentary fro m agricultureanditsmaturationintostate-level,partiallyurbansocieties.Althoughthere h isnoevidencethatpastoralismpredatedagriculturechronologically,itslesssedentary ttp s characteroftenleadstoitscharacterizationasifreflectingaregressionbackwardfrom ://a c settled society. As Kra¨tli (2001) writes, “At the core of the public representation of a d e pastoralism is the idea that “pastoralism” and “modern life” are mutually exclusive, m ic as two successive stages of human development in a unique line that goes from .o u nature to civilization, passing from sedentary life and agriculture. This frame offers p.c o no ground on which pastoralism and [the] modern world could meet: one being m thought to begin where the other is supposed to end”. This view of herders as less /jee a civilized than agriculturalists, or as a dead end branch line off the main path from /a d v agriculturetocivilization,echoesmillennia-oldChinese,Persian,andEgyptianviews a n c ofsteppeanddesertnomads.Motivatedbythissweepingnarrative,ourpaperexplores e -a whether a similar evolutionary approach can also be traced in the current economic rtic outcomesofdescendantsofgroupsthatpracticeddifferentsubsistencepatternsduring le-a b the preindustrial era. Our study is the first of which we are aware that explores the s tra impactofeconomicculture,asidentifiedbytheprimarysourceofsubsistence,atthe c individuallevel.3 t/do i/1 WestudyAfricaforseveralreasons.First,itisaplacewherethetransitionaway 0 .1 from historical lifeways took place only recently. On the eve of the “Scramble for 0 9 3 Africa” in the late 19th century, the continent was replete with examples of almost /je e every kind of preindustrial subsistence economy, from hunter-gatherers, to nomadic a /jv pastoralists,toshiftingandintensiveagriculturalists.Second,Africapresentsasetting y 0 2 in which it is relatively easy to match individuals with the economic lifeway of their 9 /5 preindustrial ancestors. In brief, lifeways can be associated with ethnic groups, and 10 7 giventheratherlimitedmatingacrossethniclines,modernindividualscanusuallybe 54 7 identified with a single tribe, and thus a particular historical lifeway.4 Finally, in the b y modern African setting, we can identify individuals with different ancestral lifeways S c ie living in the same location, thus allowing us to study lineage-based historical effects n c e inisolation,thatis,purgedfromtheeffectsoftheplace-basedhistory. s L ib ra ry u s e r o 3. Fortheroleofeconomicculture(asreflectedinthedependenceonfishing)onregionaleconomic n performance,seeDalgaardetal.(2015). 08 O 4. Thelimiteddegreeofinterethnicmarriagesisevidentinoursampleofhouseholdssurveyedatthe c turn of the 21st century. Within an average household in the Demographic and Health Surveys, there tob isa71%probabilitythattheethnicidentityofthewifeisidenticaltothatofherhusband,despitethe er 2 considerable ethnic heterogeneity of many of today’s urban centers. In absence of historical data, we 0 1 believethatintermarriagerateswerefarlowerinpreviousgenerations. 8 4 JournaloftheEuropeanEconomicAssociation The channels by which lineage—and in particular the premodern economic lifeways of an individual’s ancestors—can affect modern outcomes are not the same D asthosechannelsthatwouldbeoperativeattheleveloflocations.Mostsignificantly, ow n institutions are generally associated with places, and thus are unlikely to explain lo a heterogeneity of outcomes within a region. Hence, culture is a natural suspect, as de d it is something that can vary among individuals in a given location based on their fro m lineage. As we discuss in what follows, there are particular cultural traits associated h with increased dependence on agriculture that one would expect to yield differential ttp s benefitsinamoderneconomicsetting.Ourapproachofidentifyingtheeffectofculture ://a c onoutcomesattheindividuallevelbyfocusingonpeoplewithdifferentoriginsliving a d e inthesameplacefollowswhatFernandez(2011)callsthe“epidemiologicalapproach” m ic tostudyingculture.Thishasbeenimplementedusinginternationalmigrantstostudy .o u culture’seffectonfertility(FernandezandFogli2009)andsavingrates(Carroll,Rhee, p.c o andRhee1994),amongotherthings. m Pursuing our inquiry requires being able to associate individuals in a modern /jee a data set with historical characteristics of the groups from which they are descended. /a d v The Demographic and Health Surveys on which we mostly rely contain data on the a n c ethnicityofindividuals.WematchthisdatawithinformationfromMurdock’s(1967) e -a EthnographicAtlasonhistoricalcharacteristicsofethnicgroupsaswellasinformation rtic from Murdock (1959) on the geographical regions historically inhabited by these le-a b ethnicities. Matching these two data sets required the construction of a concordance s tra ofethnicities,thedetailsofwhicharediscussedinwhatfollows.Weexpectthatthis c t/d concordancewillhavegreatusefulnessbeyondthecurrentstudy. o i/1 Ourmainfindingisthatthehighertheshareoftheirsubsistenceaperson’sancestors 0 .1 obtained from agriculture, the higher are his or her education and wealth levels 0 9 3 today. This result holds not only when comparing the descendants of pastoralists /je e to the descendants of agriculturalists, but also when comparing the descendants of a /jv nonpastoralist groups that varied in the degree to which they relied on agriculture. y 0 2 Importantly, this pattern continues to hold even when we restrict our attention to 9 /5 individualslivingoutsidetheirgroup’sancestralhomelands,toresidentswithinurban 10 7 places, and to individuals engaged in occupations other than agriculture and animal 54 7 husbandry. b y The rest of this paper is structured as follows. Section 2 discusses the related S c ie literatureonthehistoricaldeterminantsofmoderneconomicandpoliticaloutcomes, n c e withanemphasisonAfrica.InSection3,weintroducetheethnicitydatafromtheDHS s L that we use, and discuss the matching of modern ethnicity to historical groups, their ib ra ancestrallocations,andprecolonialcharacteristics.InSection4,wedescribethedata ry u onthehistoricalmeansofsubsistenceofAfricangroupsandestimateempiricalmodels se linkingeducationandwealthtotheancestrallifewaycharacteristicsofanindividual’s r o n group,controllingforthecurrentlocationofresidence.Weexperimentwithsplitting 08 O thesamplebyoccupationandurban/rurallocation,withtheinclusionoflocationfixed c to effects, and also assess the role of selection into migration. In Section 5, we explore b e the determinants of ancestral lifeways themselves, in particular, the degree to which r 2 0 1 dependenceonagricultureisafunctionofland’sagriculturalquality.Wethenuseland 8 Michalopoulos,Putterman,andWeil InfluenceofAncestralLifeways 5 qualityasaninstrumentforancestralagriculturaldependenceinourbasicregression setting, finding broadly similar estimates. In Section 6, we investigate whether the D identified pattern is robust to exploiting variation in the mode of subsistence within ow n linguisticorethnicfamilies.InSection7,weexaminethepotentialchannelsatwork, lo a investigating how the inclusion of precolonial and colonial-era variables influences de d ourbasicresults.Wealsoreportexercisesexploringtherolesofdifferentialtreatment fro m bythecentralgovernment,aswellaswhetherpersonalitytraitsrelatedtoproclivityto h violence,impatience,andcooperationmighthelpexplainthelessfavorableoutcomes ttp s of descendants of communities relying less on agriculture in the precolonial era. We ://a c bringtobeardatafromtheAfrobarometersurveys,tosupplementtheDHS.Although a d e these exercises yield some suggestive results, we emphasize their provisional nature m ic and secondary importance relative to our core finding that premodern lifeway is a .o u robustcorrelateofeconomicoutcomesgenerationslater.Section8concludes. p.c o m /je e a 2. RelatedLiterature /a d v a n c A growing body of work examines the historical origins and political economy of e -a Africandevelopment.Broadlyspeakingthemainargumentsthathavebeenproposedin rtic thisliteraturerefertothreedifferentperiodsinAfricanhistory.Inreversechronological le-a b order, the first category includes an influential body of research that stresses how s tra the institutions established by European powers during colonization persisted after c t/d independence and continue to shape contemporary economic performance (e.g., La o i/1 Portaetal.1997,1998;Acemoglu,Johnson,andRobinson2001,2002;Michalopoulos 0 .1 and Papaioannou 2014). The second set of studies focuses on events that took place 0 9 3 during the colonial period itself. Huillery (2009), for example, quantifies the long- /je e run effects of colonial investments whereas recent works shed light on the negative a /jv effectsoftheimpropercolonialborderdesignduringtheScrambleforAfrica.5Finally, y0 2 several recent studies highlight the persistent legacy of the precolonial era. Nunn 9 /5 (2008)andNunnandWantchekon(2011),forexample,stresstheroleofslavetrades 10 7 whereas Michalopoulos and Papaioannou (2013) demonstrate the beneficial role of 54 7 ethnicpoliticalcentralizationonregionalAfricandevelopment. b y Our study belongs to the latter strand by establishing that descendants of S c ie predominantly agricultural groups today outperform economically individuals from n c e groupsofdifferentprecolonialoccupationalbackgrounds.Thisfindingcontributesto s L our understanding of the legacy of ethnicity in Africa and sheds light on the sources ib ra ofethnicinequality,afeaturethathasbeenlinkedtounderdevelopment(seeAlesina, ry u Michalopoulos,andPapaioannou2016). se More generally, our work relates to the literature on the cultural origins of r o n comparative development, adding to a vibrant body of research that examines the 08 O within-country impact of various historical legacies on economic performance. By c to b e r 2 0 1 5. Englebert,Tarango,andCarter(2002)andMichalopoulosandPapaioannou(2016). 8 6 JournaloftheEuropeanEconomicAssociation utilizingindividual-levelvariation,weovercomesomeoftheidentificationproblems inherent in cross-country or cross-regional analyses. First, it allows us to quantify D how much of the individual-level variation in economic outcomes may be attributed ow n to one’s ethnic identity. Second, we can account for location-specific traits. This is lo a feasiblebecauseweobservepeoplefromdifferentethnicgroupsresidinginthesame de d enumerationareas.(EnumerationareasasusedbytheDHSarecountingunitscreated fro m fornationalpopulationcensuses.Theyaretypicallyasinglevillageorgroupofvillages h inthecountrysideandacityblockorapartmentbuildinginurbanareas,andcontain ttp s 100–300 households, of which 20–30 are randomly selected for survey participation ://a c SeeBurgertetal.2013). a d e The introduction of location fixed effects is crucial, since it allows us to absorb m ic characteristicsrelatedtothegeographic,ecological,andinstitutionalenvironmentofa .o u givenregionthatrecentstudieshavehighlightedasimportantdeterminantsofregional p.c o African development.6 Moreover, it allows us to uncover the importance of portable m ethnic-specifictraitswhoseinfluenceisnotlimitedtotheancestralhomelandofagiven /jee a group.ThismethodologyissimilartoNunnandWantchekon(2011),whoinvestigate /a d v the impact of slavery on individual trust among respondents residing outside their a n c ethnicenclaves. e -a Ourfindingthatdescendantsofgroupsthatintheprecolonialeraderivedalarger rtic shareofsubsistencefromagriculturearetodaymoreeducatedandmorewealthybrings le-a b totheforegroundthepersistentroleoftraitsverticallytransmittedwithingroupsover s tra time.Inthisrespect,ourstudycontributestoanemergingbodyofworkthatemphasizes c t/d the importance of cultural norms, historical persistence, and human and geographic o traitsforcomparativedevelopment.7 i/10 .1 0 9 3 /je e 3. EthnicityandModernOutcomes a /jv y 0 2 3.1. EthnicityData 9/5 1 0 7 5 Our starting point is data from the Demographic and Health Surveys (DHS) for 26 4 7 countriesinwhichanethnicityvariablewascollectedaspartofthesurvey.Weusethe b y mostrecentDHSwaveforwhichbothethnicityinformationandlocationcoordinates Sc ie areavailable.Thisreducesthesampleto21countriessincefor5outof26countrieswe n c e donothavecoordinateinformationfromtheDHS.Thesamplesizewithinformation s L onbothethnicityandenumerationareacoordinatesrangesfrom3040individualsfor ib ra theIvoryCoastto48,871forNigeria,totaling337,382respondents.InourfinalDHS ry u s e r o n 0 8 O c 6. Alsan(2015)andFenske(2013). to b 7. SeeDiamond(1997),EasterlyandLevine(1997,2012),SpolaoreandWacziarg(2013),Putterman er 2 andWeil(2010),AshrafandGalor(2013),andMichalopoulos,Naghavi,andPrarolo(2016,2017)among 0 1 others. 8 Michalopoulos,Putterman,andWeil InfluenceofAncestralLifeways 7 sample,thereare492ethnicity-countrygroups,wherethesameethnicityappearingin twodifferentcountriesiscountedastwodifferentgroups.8 D o w n 3.1.1. Matching Modern Ethnicities to Ancestral Groups, Historical Locations, and lo a d Group-SpecificPrecolonialTraits. Theinformationontribalprecolonialtraitscomes e d fromGray’s(1999)compilationofMurdock’s(1967)EthnographicAtlaswhereasthe fro m spatial information on the homeland of a group in the beginning of the colonial era h comesfromMurdock’s(1959)Map.TheAtlasisbasedonadistillationbyMurdock ttp s fromalmostthefullcorpusofethnographicmaterialson1167societies,inaseriesof ://a c installments appearing in the journal Ethnology between 1962 and 1980. Following ad e further editing and coding by anthropologists Herbert Barry, Douglas R. White, m ic Gregory F. Truex, and Michael Fischer, it was compiled in 1999 by anthropologist .o u p PatrickGrayintotheversionusedhereandbyeconomistsbeginningwithGennaioli .c o andRainer(2007). m /je Themap,separatelypublishedbyMurdock(1959),doesnotalwaysfullycoincide e a with the Atlas from the standpoint of ethnicity names, and it suffers from potential /a d v problemsofprecision,whichweaddress,alongwithadditionalinformationaboutthe a n c Atlas,inAppendixA.Inbrief,welinkedtheethnicityasreportedbyeachrespondent e -a in the DHS to both Murdock’s (1967) list and Murdock’s (1959) Map. Whenever rtic le possible we used the concordance constructed by Fenske (2013) and Michalopoulos -a b andPapaioannou(2013)toassociatethegroupsinMurdock’sMap(1959)tothegroups s tra inMurdock’sAtlas(1967).SeeAppendixAforfurtherdetails. c t/d Atotalof287,433individualswerematchedtoaMurdockAtlasgroupandassigned o i/1 characteristicsofthecorrespondingethnicgroupintheEthnographicAtlas.Aslightly 0 .1 largernumber,292,942,werematchedtogroupsincludedintheMurdockmap. 09 3 Our matching procedure was as follows. We constructed a series of ten possible /je e methods for matching ethnicities in the DHS to ethnicities in one of the Murdock a/jv datasets.Thesemethodswereorderedfrombesttoworstintermsofourassessmentof y0 2 theirlikelyaccuracy.Wethenproceededdownthelist,usingforeachDHSethnicity 9/5 1 the first method for which we were able to achieve a match. Matching was done 0 7 5 separately for the ethnicities included in the Atlas and Map, respectively. In the text 4 7 inwhatfollows,wedescribethemostimportantmethods.InPanelA,wedescribeall b y S tenmethodsandgivethefractionsofcasesmatchedusingeachone. c ie Themethodatthetopofourlistwas“directmatch”,inwhichthesamenamewas n c e used in the DHS and the Murdock source. We were able to directly match 58.7% of s L observations to Atlas ethnicities and 67.0% to the ethnicities on the Murdock map. ib ra The second method on our list was “Afrobarometer match”, in which we applied to ry u theethnicitynamesthatappearintheDHStheconcordanceconstructedbyNunnand se r o n 0 8 O 8. Thesurveyroundsintherespectivecountriesare:BF6(BurkinaFaso),BJ4(Benin),CD5(Congo c DemocraticRepublic),CF3(CentralAfricanRepublic),CM4(Cameroon),ET6(Ethiopia),GH5(Ghana), tob GN4(Guinea),(CI3)IvoryCoast,KE5(Kenya),ML5(Mali),MW5(Malawi),MZ6(Mozambique),NG5 er 2 (Nigeria),NI3(Niger),NM4(Namibia),SL5(SierraLeone),SN6(Senegal),TG4(Togo),UG6(Uganda), 0 1 andZM5(Zambia). 8 8 JournaloftheEuropeanEconomicAssociation Wantchekon(2011)relatingethnicitynamesthatappearintheAfrobarometerRound 3datasettoethnicitiesthatappearintheMurdockdataset.Thismatchedafurther4.5% D ofobservationstoAtlasethnicitiesand10.0%ofobservationstotheMap’sethnicities. ow n The next three methods used data on alternate ethnicity names from the Ethnologue lo a ortheJoshuaProject.ThethirdmethodappliedtocaseswheretheDHSandMurdock de d nameswerelistedasalternates;thefourthwhereanamethatappearedintheMurdock fro m source is listed as a superset of the ethnicity that appears in the DHS; and the fifth h where the name that appears in the DHS is listed as a subset of the ethnicity in the ttp s Murdockdata.Together,thesethreemethodsmatched19.1%ofobservationstoAtlas ://a c ethnicitiesand13.2%toMapethnicities. a d e m ic 3.1.2. MoversandAverageDistanceMoved. Asdescribedpreviously,muchofour .o u interestinthispaperiswiththeaspectsofhumancapital(broadlydefined)thatpersist p.c o overgenerationsandareportableacrosslocations.Further,weareinterestedinaspects m ofculturethathavetheiroriginsintheconditions ofparticulargeographic locations. /jee a To the extent that people live in the regions traditionally associated with their kin, it /a d v would not be possible to separately identify the effect of tribal characteristics from a n c geographical characteristics. Thus we have a particular interest in individuals who e -a live outside the territory associated with their group of origin. We follow Nunn and rtic Wantchekon (2011) in calling such individuals “movers”, even though they may not le-a havemovedintheirownlifetimes.9 bs tra TheDHSreportscoordinateinformationforaperson’scurrentresidence.Wecan c t/d thusclassifyindividualsaslivinginsideoroutsidetheirancestralhomeland.Forthose o i/1 livingoutsideoftheirhomeland,wegeneratedavariablemeasuringdistancetotheir 0 .1 homeland. Specifically, this is the distance from the coordinates of an individual’s 0 9 3 currentresidencereportedintheDHSsurveytothenearestborderofhis/herancestral /je e homeland(Murdock’smap).10 a /jv IntheDHSdata,40%ofindividualscurrentlylivewithintheboundariesoftheir y 0 2 ancestralhomelands.Ofthosewhodonot,12%livemorethan500km,36%between 9 /5 100 and 500 km, 43% between 10 and 100 km, and 9% within 10 km of the border. 10 7 Given the imprecise nature of the borders in the Murdock map, the fact that ethnic 54 7 group locations may have some overlap and that DHS coordinates in rural areas are b y perturbed by 5 or 10 km, we are reluctant to assume that members of this last group S c ie areinfactlivingawayfromtheirancestrallands.Hence,wetreatthemasnonmovers n c e in the empirical exercises in what follows. Panel B gives summary statistics for our s L DHSsampleasawhole. ib ra ry u s 9. TheDHSreportswhetheranindividualhasmovedinhis/herlifetimeforasubsetofrespondents. er o Thisquestiondoesnotdistinguishbetweenpeoplethatmovedoutoftheirhomelandorfromsomeother n locationwithintheirhomeland. 08 O 10. Outofthe285,155respondents,wehavedistancetoancestralhomelandfor258,284.Intheremaining c cases, we have matched directly the DHS ethnicity of the respondent to an Atlas group so we have tob informationonitsprecolonialtraitsbutwecouldnotmatchthisAtlasgrouptoanethnicgrouponthe er 2 Murdockmap.Forcaseswhereanindividualismatchedtomorethanoneancestralhomeland,thenearest 0 1 homelandwaspickedtocomputethisdistance. 8 Michalopoulos,Putterman,andWeil InfluenceofAncestralLifeways 9 3.2. EthnicityandModernOutcomes D WefocusontwooutcomesfromtheDHS:educationandwealth.Ourprimarymeasure ow n of education is a variable (mv149—educational attainment) that takes six distinct lo a values(0–5)correspondingtonoeducation,someprimary,exactlyprimarycompleted, de d incomplete secondary, exactly secondary completed, and higher than secondary. fro m Figure1showsahistogramofthedistributionofthismeasureinoursample.Asecond h measureofeducation,yearsofschoolcompleted(mv133—educationinsingleyears), ttp s isavailableforasubsampleofobservations.Weusethismeasureforrobustnesschecks ://a c inwhatfollows.Themedianyearsofschoolingcompletedwithinthesixcategoriesof a d e mv149are0,4,6,9,12,and15years,respectively. m ic Wealthiscodedona1–5scalethatdividesthesampledhouseholdswithinacountry .o u into quintiles. The DHS wealth index is composed taking into account consumer p.c o durables,electricity,toiletfacilities,sourceofdrinkingwater,dwellingcharacteristics, m /je andsomecountry-specificattributessuchaswhetherthereisadomesticservant.Some e a of these components are closer to being measures of consumption flow than wealth /a d v stock. In practice, we use this as a general measure of the standard of living, but a n c follow DHS usage in calling it wealth. The measure is derived by the DHS using e -a principal component analysis to assign indicator weights resulting in a composite rtic standardized index for each country. Since our empirical analysis is at the individual le-a b level, each member of a household is assigned the same level of wealth. Rutstein s tra and Johnson (2004) provide a detailed description of the construction of this index. c t/d The raw correlation between education and wealth in the full sample is 0.45 and the o i/1 correlationsofthesevariableswithanurbanindicatorare0.36and0.60,respectively. 0 .1 Before turning to the role played by ancestral ethnic characteristics, we explore 0 9 3 thepredictivepowerofethnicitymoregenerallyinourdata.Table1reportsR2sfrom /je e regressionsofoureducationandwealthmeasuresondifferentsetsofdummyvariables; a /jv namely,countryfixedeffects,currentethnichomelandfixedeffects,andethnicidentity y0 2 fixed effects. The ethnic homeland fixed effects are dummy variables corresponding 9 /5 1 to the current tribal location of the individual according to the Murdock map. We 0 7 alsoreporttheR2 fromcombiningdifferentgroupsofdummyvariablestogaugethe 54 7 additionalexplanatorypowerofdifferentsetsofdummies. b y The regressions show, first of all, the role of ethnicity in determining outcomes. Sc ie Forexample,oncecountryfixedeffectsareincludedintheregression,addingcountry- n c ethnicity constants raises the R2 for education from 0.159 to 0.281, and for wealth es L from0.013to0.159(resultsformoversareslightlylarger).11 Currentcountry-ethnic ib ra homeland has more predictive power than does ethnic affiliation: for education, the ry u differenceisabout2percentagepoints,whereasforwealththedifferenceisalmost9 se percentagepoints.However,whatismoreimportantforouranalysisisthatevenwhen r o n dummies for current country-ethnic location are included in the regression, there is 08 O c to b e r 2 11. NotethattheR2forthecountry-fixed-effectsregressiononwealthisalmostzerobecausewealthis 0 1 standardizedbycountry. 8 10 JournaloftheEuropeanEconomicAssociation D o w n lo a d e d fro m h ttp s ://a c a d e m ic .o u p .c o m /je e a /a d v a n c e -a rtic le -a b s tra c t/d o i/1 0 .1 0 9 3 /je e a /jv y 0 2 9 /5 1 0 7 5 4 7 b y S c ie n c e s L ib ra ry u s e r o n 0 8 O c to b e r 2 0 1 8 FIGURE1. Populationdistributionacrosseducationalcategories.