1SeeNunn2009forareview2SeeamongstotherstudiesAcemogluetal20012002GlaeserandShleifer2002BanerjeeandIyer2005Nunn20082011andDell20103TedMiguelSatyanathShankerandErnestSergenti200 ID: 121830
Download Pdf The PPT/PDF document "1IntroductionEmpiricalevidencedocumentst..." 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.
1IntroductionEmpiricalevidencedocumentsthatmajorhistoricalepisodescanhavepersistenteconomicconsequences.1Forexample,recentstudieshaveshownthatcolonialismandtheAtlanticslavetradefundamentallyin uencedsocietalorganizationcenturiesago,placinginmotionforceswhosee ectshavepersistedlongaftertheinitialinstitutionalcauseswereremoved.2Thisstudy,incontrast,examineshowasmaller,typicallyinconsequentialeventcanexertpersistentconsequencesbecauseofitstiming.Specically,ittestshowamoderatedroughtinearly20thcenturyMexicoa ectedinsurgencyduringtheMexicanRevolution,inturnin uencinglong-runeconomicandpoliticaldevelopment.TheMexicanRevolutionwasamulti-sidedcivilwarthatbeganin1910withtheover-throwoflong-timeautocratPorrioDaz.Atitsheartweredisputesaboutlanddistributionandthedegreetowhichpoliticalpowershouldbecentralized,andtheRevolutionwasulti-matelywonbyafactionfavoringstatecentralization.Whenghtingabatedin1918,rampantcon ictsoverlandremainedunresolvedandthecentralstatehadlimitedauthorityoutsidethenationalcapital.Thisstudyquantiestherelationshipbetweenidiosyncraticrainfallshocksandrevolu-tionaryinsurgency,whichisdenedasthesustaineduseofviolentforcebylocalresidentstosubvertrepresentativesofthegovernment.Itthenexamineswhethereconomicdevelopmentandpoliticssubsequentlydi eredinmunicipalitieswithandwithoutinsurgentactivity,us-inganinstrumentalvariablesstrategythatemploysplausiblyexogenousvariationinrainfalltoidentifyinsurgency'se ects.3Allestimatesareconditionalonstatexede ectsandtimeinvariantgeographiccontrols.Thereisastrongcorrelationbetweendroughtseverityandinsurgency.Movingfromhalfoflong-runaverageprecipitation-aseveredrought-toaverageprecipitationdecreasestheprobabilityofinsurgentactivitybyaround38percentagepoints,andtheF-statisticonthisrelationshipis19.Incontrast,droughtseverityisuncorrelatedwithahostofcharacteristicsmeasuredin1900.\Landandliberty"wasthebattlecryoftheRevolution,andinthedecadesthatfollowedit,Mexicoredistributedoverhalfofitssurfaceareaintheformofejidos:farmscomprisedofindividualandcommunalplotsthatweregrantedtoagroupofpetitioners.Instrumentalvariablesestimatesdocumentthatmunicipalitieswithrevolutionaryinsurgencyhadaround 1SeeNunn,2009forareview.2See,amongstotherstudies,Acemogluetal.(2001,2002),GlaeserandShleifer(2002),BanerjeeandIyer(2005),Nunn(2008,2011),andDell(2010).3TedMiguel,SatyanathShanker,andErnestSergenti(2004)pioneeredtheuseofrainfallshockstoempiricallyexaminecon ict.Theyutilizerainfallasaninstrumentforgrowthinordertoidentifythecausale ectofgrowthoncon ictinAfrica.1 boomedduringthe1920s.Incontrast,existinghistoricalandquantitativeevidencesuggestthatlandreformcouldbeaplausibleandimportantchannellinkingtheRevolutiontolong-rundevelopment.IhypothesizethattheMexicanstatepromotedstabilityininsurgentregionsbyimple-mentinglarge-scaleagrarianreformandthatthisinturnloweredindustrialization,income,andpoliticalcompetitioninthelongrun.Empiricallyidentifyinghowpropertyrightshavea ectedlong-rundevelopmentinMexicoisasignicantresearchagendathatrequiresalterna-tivesamplesandidenticationapproaches.Hencespaceconstraintsrequirethatexaminationofthedetailedmechanismslinkinglandreformtolong-rundevelopmentbepursuedinasep-aratestudy.Nevertheless,theexistingevidencestronglysuggestslandreformasaplausiblemechanismlinkingdroughtandrevolutiontoeconomicandpoliticaloutcomestoday(seeDeiningerandBresciani,2001forareviewofMexico'sejidalsector).MostcloselyrelatedtothisstudyisworkbyBeatrizMagaloni,BarryWeingast,andAlbertoDiaz-Cayeros(2008)thatusesastandardgrowthregressionframeworkappliedtoMexico's31statesbetween1950and1995tomeasuretheeconomicimpactoflandreform.TheyestimatethatMexicanGDPpercapitawouldhavebeen124percenthigherin1995hadtherebeennolandreform.TheyalsoprovideevidencethatthedistributionofejidallandswasakeypatronageinstrumentforgeneratingsupportforthePRI(InstitutionalizedRevolutionaryParty),whichdominatedMexicanpoliticsfromthe1920sthroughtheendofthe20thcentury.Additionally,becauseindividualsdonotownejidallands,theycouldnotusethemforcollateralandinsteadcouldonlyborrowfromthehighlycorruptstatebank(DeiningerandBresciani,2001;Benjamin,1989;deJanvryet.al,1997;Ronfeldt,1973).Finally,inongoingworkIshowthatlandreformloweredlong-runindustrializationbypreventingtheestablishmentofagribusiness.Landdistributionremainscentralinmanycon ictstoday,includingthoseinAfghanistan,Iraq,Uganda,Guatemala,Indonesia,Cambodia,andelsewhere.TheMexicangovernment'suseofagrarianreformtopromotestabilityandextenditscontrolovertheagriculturalsectorisbynomeansuniqueamongstlandreformsoccurringoverthecourseofthepastcentury.Forexample,parallelsexistbetweentheMexicancaseandtheIraqilandreformsofthe1950sandearly1960s,inwhichthestatedidnotfullydistributeconscatedlandstoprivateholdersbutratherbecame\averylargeabsenteelandowner"(Warriner,1969,p.92).CommunistbloccountriessuchasChina,Cuba,andVietnamwentevenfurther,withrevolutionaryregimescollectivizingfarming(King,1977).Thisstudy'sempiricalresultshighlightthepotentialforpersistenteconomicinecienciestoarisewhenreformstoresolvelanddisputesextendthestate'spowerbyreplacingmarketinteractionswithpoliticalpatronage.Whilethe1900censussuggeststhatmunicipalitiesthatwouldlaterexperiencevaryingdegreesofdroughtwereinitiallysimilar,thefortunesofthesemunicipalitieslaterdiverged,plausibly3 inpartbecauseofdi erencesinhowpropertyrightsweresubsequentlyorganized.Inthenextsection,IprovidehistoricalbackgroundontheMexicanRevolution,andSection3examinestherelationshipbetweendroughtandrevolutionaryinsurgency.Section4testswhetherinsurgencyimpactedagrarianreformandlong-runeconomicdevelopmentbyusingdroughtasaninstrumentforinsurgency.Section5discussestherelationshipsbetweenhistoricalinsurgency,landreform,andlong-rundevelopment.Finally,Section6o ersconcludingremarks.2HistoricalBackground2.1TheMexicanRevolutionTheMexicanRevolutionbeganin1910withtheoverthrowofautocratPorrioDaz,whorstascendedtothepresidencyofMexicoin1876.WhileMexico'sconstitutionstipulatedademocraticgovernmentwithnore-electionofthepresident,DazrepeatedlyclaimedthatMexicowasnotyetreadyfordemocracyandremainedinpowerthroughriggedelections.His35yeartenurewasmarkedbyindustrialization,adramaticincreaseinlandconcentration,andthecentralizationofpoliticalpower.Ina1908interviewwiththeU.S.journalistJamesCreelman,Daz-thenapproachingeighty-statedthathewouldretireandallowothercandidatestocompeteforthepresidency.However,Dazultimatelychangedhismindandranagainforre-electionin1910,allowingnorthernoppositionpoliticianFranciscoMaderotorunagainsthim.DespitewidespreadpopularsupportforMadero,Dazclaimedtohavebeenre-electedalmostunanimouslyandhadMaderojailed.Maderosubsequentlyissuedaletterfromjailcallingforpopularrevolt,andhisvaguepromisesofagrarianreformattractedpeasantsthroughoutMexico,leadingtonumerousrebellions.Thetimewasparticularlyripeforsuccessfulrebellion,giventheadvancedageofDazandmanyofhismilitaryallies,andDazandthefederalarmyweredefeatedinMayof1911.Maderowaselectedtothepresidency,butprovedanunpopularleader.Heangeredthepeasantrevolutionarieswhohadbroughthimtopowerbyfailingtoimplementagrarianre-formandbyincreasinglyallyingwithmembersofthePorrianelitetoquellunrestamongstthosedemandingradicalchange(Knight,1986).Whilehefacedarmedoppositionfromtheleft,Maderowasultimatelyoverthrownina1913coupbycounter-revolutionaryGeneralVic-torianoHuerta.NumerousrevolutionarymovementsagainstHuerta'smilitarygovernmentarosein1913,occurringlargelyinthesameplacesthathadwitnessedrevolutionaryactivityin1910-1911.ThesedisparatemovementswereabletouniteintheiroppositiontoHuerta4 andoverthrowhisregimeinJulyof1914.However,theirdi erencesprovedirreconcilableandsoonafterHuerta'sdefeatthecon ictdeterioratedintoamulti-sidedcivilwar.Theultimatelyvictoriousfaction,referredtoasConstitutionalism,emphasizedeconomicmodernizationandstatecentralizationunderapoliticalclassthatwasdi erentandsome-whatbroaderthanthePorrianelite.Constitutionalismwascentralizedunderasinglemilitaryandpoliticalcommandandunambiguouslysoughtnationalpower.Themove-ment'sbackbonewasinthenorthernMexicanstatesofSonoraandCoahuila,whichwererelativelyprosperous.Itgarnereditsprimarysupportfrommiddleclass,urban,andindus-trialinterestsandalsogainedtheocialrecognitionoftheU.S.government.ThevictoriousConstitutionalistsultimatelyformedtheInstitutionalizedRevolutionaryParty(PRI),whichdominatedMexicoasasinglepartypoliticalsystemformostofthetwentiethcentury.Incontrast,themostwidespreadtypeofrevolutionarymovementsoughttodefendlocalpoliticalstructuresagainstincursionsbythecentralgovernment.ManyofthesemovementscalledforagrarianreformtoreturnlandsconscatedbylargeestatesduringtheDazregimetothepeasantswhohadpreviouslyheldthem,andtheyweretypicallylocalintheirdemands,scope,andpoliticalaspirations.5TherebellionsledbyPanchoVillainnorthernMexicoandEmilianoZapataincentralMexicoarethelargestandmostwell-knownofthesemovements.TheRevolutionwitnessedsometraditionalpitchedbattles,foughtprimarilybetweentheConstitutionalistsandPanchoVillainnorthernandnorth-centralMexico,butmuchoftheghtingconsistedofguerrillawarfare.Bytheendof1915,VillahadbeenreducedtoguerrillatacticsinhishomebaseofChihuahua,andittooktheConstitutionalistarmyseveralmoreyearsofghtingtodefeathimandthelargenumberofotherlocalguerrillamovements,oneofthemosttenaciousofwhichwastheZapatistasinMorelos.Purgesoflocalrebelleaderscontinuedthroughoutthe1920sandinsomecasesintothe1930s.2.2BringinginsurgentregionsunderthecontrolofthestateTherallyingbattlecryoftheRevolutionwas\landandliberty"(tierraylibertad),andwhilenotallinsurgentgroupswereghtingforlandredistribution,thiswasacentralde-mandofmanyinsurgentmovements(Knight,1986).TowardsthecloseoftheRevolution,Mexicoratiedanewconstitutionthatstatedthatcentersofpopulationthatlackedaccesstoadequatelandwouldbegrantedlandinsucientquantitiesfortheirinhabitants'needs.AccordingtoArticle27,estateswhosesizeexceededamaximumlimitcouldbeexpropri- 5ProminentexamplesofmovementscallingforradicalagrarianreformincludetheZapatistasoperatinginMorelosaswellaspartsofPuebla,Mexicostate,andsomeotherregions;theCedillobrothersinSanLuisPotosi;andCalixtoContrerasinDurango.ExamplesofmovementsthatdidnotpursueagrariangoalsincludingrebellionsintheMisantlaandHuatusco-CordobaregionsofVeracruzledbylargelandowners,theManuelPelaezrebellionintheHuasteca,andtheNatividadbrothersinTepic(Knight,1986).5 ated;andreligiousinstitutionscouldnothold,administer,oracquireland.AllpropertiesbelongingtotheCatholicChurch,whichheldextensiveproperty,revertedtothestate.InthedecadesfollowingtheRevolution,thegovernmentoperationalizedArticle27throughthecreationofstate-ownedpropertiescalledejidos.Ejidosconsistedofcommu-nalandindividualplotsthatweregrantedbypresidentialdecreetoagroupofpetitioners.Communalplotsweredevotedtopurposessuchasgrazingandrewood,whereascropsweretypicallycultivatedonindividualejidalplots.Ejidomembersreceivedusufructrightstotheirplots,butownershiprightswereheldbythecommunityandultimatelycontrolledbythecentralstate.Whiletherightstotheproducefromone'splotwereinalienableaslongasoneremainedontheland,theycouldnotbetransferred,andtherentalofejidallandandhiringoflabortoworkejidalplotswerealsoprohibited.EjidoswerecentraltoMexicanagriculture,asnearlyhalfofMexico'ssurfaceareaenteredtheejidalsectorduringthe1920sand1930s.Thestateobtainedthelandfortheejdialsectorfromestatesthathadbeenconscatedfromthepre-revolutionaryeliteduringtheRevolution,fromtheextensiveholdingsconscatedfromtheChurch,andfromthepurchaseoflandsfromlargelandownerswhoseestatesexceededthelegalmaximumsize.EjidosmetthedemandsofMexico'speasantrevolutionariesforaccesstoland,andsimul-taneouslyservedasacentralvehicleforthestatetoextenditscontrolintothecountryside.Thecentralstatecontrolledaccesstoessentialinputs,suchaswaterresources(includingwellsonejidalproperties)andcredit.Ejidoproducerscouldnotusetheirlandascollateraltoaccessprivatecreditmarketssincethestateownedtheland,andthuswerereliantonthestateforthecreditnecessarytobuyseeds,fertilizer,andotherinputs.Pervasivecorruptioninthestatebankservingtheejidalsectorhasbeenwell-documented(DeWalt,1979;Wilkie,1971).Ejidalelitesexerteddisproportionatein uenceoverdecisionsaboutthereallocationofvacatedejidallands,withplotsbecameconcentratedinthehandsofindividualswhoheldpoliticalpositionsintheejidos(DeWalt,1979;Wilkie,1971).ManydecisionsaboutlandallocationandcredithadtobecountersignedbystatepoliticiansfromMexico'ssinglepartybureaucracy(DeiningerandBresciani,2001;Benjamin,1989;deJanvryet.al,1997;Ronfeldt,1973).Ejidostodayaccountfor54%ofMexico'slandarea,andabouthalfofitsruralpopulation.In1992,MexicoreformedArticle27oftheconstitutionsothatthestatenolongerhasanobligationtoprovideejidos.ThestatebegananejidotitlingprogramcalledPROCEDEinthesameyear,seekingtoresolvecon ictsoverplotboundarieswithinandbetweenejidosandtofacilitateinvestmentandmarkets.ThroughPROCEDE,anejidoassemblyselectswhichpartsoftheejidowillbedesignatedforcommonuseandwhichpartswillbedesignatedforprivateplots.Boundariesaredelineatedandrentaloftheplotsdesignatedforprivate6 useispermitted.Oncetheindividualplotshavebeendelineated,producershavetheoptionofregisteringtheirplotintheprivatedomain-allowingittobeboughtandsoldinlandmarkets-ifasuper-majorityoftheejidomembersagree.Todate,approximately90%ofejidoshavehadtheirinternalandexternalboundariesdelineatedthroughPROCEDE,andaround2.5%ofthesurfaceareaintheejidalsectorhasenteredtheprivatedomain.Therearevariousmechanismsthroughwhichagrarianorganizationcoulda ectlong-rundevelopment.ThesewillbediscussedinmoredetailinSection5,aftertherelationshipsbe-tweeninsurgencyandlandreformandbetweeninsurgencyandlong-runeconomicoutcomeshavebeenexamined.3DroughtandInsurgencyInordertoexaminetheimpactofinsurgencyonsubsequentstatepoliciesandlong-runeconomicdevelopment,Iusedroughtseveritytoinstrumentforinsurgentactivity.Scholarshavearguedthatamajordroughtoccurringbetween1907and1910wascentralinspurringrevolutionaryactivity(seeTutino,1981;LaFrance,1990),butthishypothesishasnotbeentestedempirically.Inthissection,Iexploretherststagerelationshipbetweendroughtseverityandinsurgency,conditionalonstatexede ects.Ialsotestwhetherdroughtiscorrelatedwithanumberofimportantpre-characteristicsandexaminewhetherthesizeoftherststagerelationshipdi ersacrossvarioussub-groups.3.1IdenticationStrategyItestwhetherdroughtseverityintheyearsleadinguptotheRevolutiona ectedinsurgencybyrunningthefollowingregression:insurgencyms= 0+ 1droughtms+X0ms+ s+ms(1)whereinsurgencymsisadummyvariableequalto1ifthecitizensofmunicipalitym-duringtheperiodbetween1910and1918-usedviolentforceinasustainedattempttosubvertrepresentativesoftheMexicangovernment(i.e.localauthoritiesandthemilitary)ortoconscateothers'property.droughtmsmeasurestheseverityofdroughtduringthe1906-1910period,Xmscontainsavectoroftimeinvariantgeographiccharacteristics,and sisastatexede ect.Allvariablesaredescribedinmoredetailinthefollowingsection.Ithenusedroughtseverityasaninstrumentforinsurgencyinthefollowingregression:yms=0+1insurgencyms+Xms+ s+ms(2)7 whereymsistheoutcomeofinterest.Thisinstrumentalvariablesapproachrequiresthetwofollowingassumptions(Angrist,2009).First,droughtmustbecorrelatedwithinsurgency.Ifthiscorrelationisonlymarginallydi erentfromzero,theresultinginstrumentalvariablesestimatesareunlikelytobeinformative.Second,droughtmustbeuncorrelatedwithanyotherdeterminantsoftheoutcomesofinterest:inotherwords,corr(droughtms;ms)=0.Thisconditionisreferredtoastheexclusionrestriction.Itwillobtainifdroughtisasgoodasrandomlyassigned,conditionalonstatexede ects,andifdroughthasnoe ectonlong-runeconomicandpoliticaloutcomesotherthanthroughtheinsurgencychannel.Whiletheexclusionrestrictionreliesontheinstrumentbeinguncorrelatedwithunob-serveddeterminantsoftheoutcomesandhenceisuntestable,Ishedlightonitsplausibilitybyrunningtwosetsofplacebochecks.First,Itestwhetherdroughtisuncorrelatedwithanumberofimportantobservablecharacteristicsmeasuredin1900.Second,Iexaminewhetherdroughtinotherveyearperiodsexertspersistente ectsonlong-rundevelopment.Totheextentthatsimilardroughtsinotherperiodsdonothavepersistente ects,thiswouldincreaseourcondencethatanylong-rune ectsofthe1906-1910droughtareactingprimarilythroughitsimpactsoninsurgency.3.2DataMonthlyveyearaveragesofprecipitationduring1906-1910areavailablefor217municipal-ities,locatedindistrictseatsthroughoutMexico.6ThesedataarepreservedinagovernmentpublicationAtlastermopluviometricathatwassenttotheWorld'sFairandisnowheldbyTulaneUniversity.Imeasuredroughtseverityastheratioofaveragemonthlyprecipitationin1906-1910tolong-runaveragemonthlyprecipitation,whichittakenfromWorldClim4'smonthlylong-runaverageprecipitationrasters.Motivatedbythehistoricalandagronomicliterature(i.e.HollingerandChangnon,1993),Icensorthemeasureatone7Ifocusonrain-fallduringnon-harvestmonthsforcorn-Mexico'smainstaplecropin1910.Whiledroughtisharmfulduringmostoftheyear-loweringsoilmoisturecontentandreducingplantgrowth-itisbenecialduringtheharvestseason.Resultsaregenerallyrobusttousingdroughtseverityovertheyearasawhole,buttherststageisweaker.DroughtseverityisplottedinFigure1. 6Districtsareapoliticalunitlargerthanmunicipalitiesbutsmallerthanstatesthatwereabolishedin1916.7Ineldexperiments,agronomistshavefoundonlymarginalbenetstocornyieldsfromaugmentingprecipitationabovelong-runaveragelevels.Ontheotherhand,additionalprecipitationisbenecialindryyears(seeforexampleHolliingerandChangnon,1993).8 way.SummarystatisticsforthedroughtandinsurgencyvariablesarepresentedinTable1.3.3ResultsTable2documentstherststagerelationshipbetween1906-1910droughtseverityandin-surgencyduringtheMexicanRevolutionandexamineswhetherthisrelationshipisrobusttodi erentmeasurementsofdroughtseverity.RobuststandarderrorsareinparenthesesandConleystandarderrorsthatcorrectforspatialcorrelationareinbrackets.Allcolumnsincludecontrolsfortime-invariantmunicipalgeographiccharacteristics:long-runaverageprecipitation(1950-2000),longrunaverageminimumandmaximumtemperature(1950-2000),elevation,andslope,aswellasstatexede ects.9Column1usesthepercentnormalmeasureofdroughtseverityinnon-harvestmonths,themeasurethatconcursmostcloselywiththebiologyofplantgrowth.Thismeasuredividesprecipitationin1906-1910bylong-runaverageprecipitationandisavailableforthefullsampleofmunicipalitiesforwhichwehavedataon1906-1910precipitation.Column2examinesthepercentnormalmeasurecalculatedovertheentireyear.Inbothcases,thecorrelationbetweendroughtseverityandrevolutionisstatisticallysignicantattheonepercentlevel,withasomewhatstrongerrelationshipwhenthenon-harvestdroughtmeasureisused(rststageF-statisticof19.1)thanwhenthefullyeardroughtmeasureisused(rststageF-statisticof9.9).Movingfromhalfoflong-runaverageprecipitation-aseveredrought-toaverageprecipitationdecreasestheprobabilityofinsurgentactivitybyaround38percentagepoints.Inthesampleasawhole,theprobabilityofinsurgencyis59%.Columns3through6examinetherelationshipbetweenstandardizedprecipitationin-dicesandinsurgency.Recallthatthestandardizedprecipitationindexsubtractsthelong-runmonthlymeanfromthe1906-1910monthlymeanandthedividesbythelong-runstandard-izeddeviation,averagingthesestandardizedmonthlydeviationsacrosstheyear.Columns3and4constructthestandardizedprecipitationindexbyusingweatherstationswithin25kmofa1906-1910weatherstationtocalculatethelong-runstandarddeviationofprecipitation.Columns5and6usedatafromweatherstationswithin10kmofa1906-1910weathersta-tiontocalculatethelong-runstandarddeviation.In182municipalities,aweatherstationappearingforatleast25yearsintheweatherrecordislocatedwithin25kilometersofthe1906-1910weatherstation,and120municipalitieshaveoneofthesestationswithin10kilo-metersoftheir1906-1910weatherstation.Thesesamplescomparetothefullsampleof210municipalitiesforwhich1906-1910precipitationdataareavailable.Columns3and5calcu-latethestandardizedprecipitationindexaveragedovernon-harvestmonths,andcolumns4 9Results(notshown)aresimilarwhenthegeographiccontrolsareexcludedorwhenmunicipal-levelcharacteristicsmeasuredin1900areincluded.10 and6averagethemonthlystandardizedprecipitationindexoverallmonthsintheyear.Thereistypicallyarststagerelationshipbetweenthedroughtmeasureandinsurgency,buttherststageF-statisticsaresmallerthantheyareincolumns1and2.Movingfromonestandarddeviationbelownormalprecipitationtonormalprecipitationdecreasestheprobabilityofinsurgencybyaround17percentagepoints.Becausetherststageisnotasstrongforthestandardizedindicesandthesamplesizeissmaller,Iusethepercentnormaldroughtmeasurefortheremainderoftheanalysis.Resultsusingthestandardizedprecipitationindexasaninstrumentforinsurgencycanbefoundintheappendix.Theinstrumentalvariablesapproachwillestimatetheimpactofinsurgencyonvariousoutcomesforthosemunicipalitiesthatwereinducedbydroughttoparticipateininsurgentactivity.Whilewecannotobservewhethercitizensinagivenmunicipalitytookuparmsinresponsetodrought,Table3shedslightonwhichsortsofmunicipalitieswerein uencedbythedroughtbyexaminingthesizeoftherststagefordi erentsub-populations.Column1reportsthebaselinerststagerelationshipfromthefullsample,reproducingcolumn1ofTable2forcomparisonpurposes.Column2limitsthesampletomunicipalitiesinstatesthatarecloserthanthemediandistancetotheU.S.,andcolumn3limitsthesampletomunicipalitiesinstatesthatarefartherthanthemediandistancefromtheU.S.Thecoecientondroughtseverityinthefullsampleis�0:770(s.e.=0:254).Thiscoecientis�0:650forplacesnearertheU.S.and�0:917(s.e.=0:224)forplacesfurtherfromtheU.S.Botharestatisticallydi erentfromzero,andtheyarenotstatisticallydistinguishablefromeachother.Columns4and5dividethesamplebywhetherthemunicipalityhadahigherpercentageofitsmalepopulationworkinginagriculturalin1900thanthemedianmunicipality.Thecorrelationbetweendroughtseverityandinsurgencyisstatisticallysignicantinbothsam-plesbutissubstantiallylargerinthemoreagriculturalsample,at-1.288,thaninthelessagriculturalsample.Inthemoreagriculturalsample,movingfromhalfoflong-runaveragerainfallin1906-1910(aseveredrought)tolong-runaveragerainfalldecreasestheprobabilityofinsurgencybyaround64percentagepoints.Next,columns6and7dividethesamplebywhetherthemunicipalityhadmoreorlessofitspopulationlivinginanhaciendain1900relativetothemedianmunicipality.Therelationshipbetweendroughtseverityandinsur-gencyislargeandhighlystatisticallysignicantforbothsamples.Finally,columns8and9dividethesamplebywhetherahigherorlowerpercentageofthemunicipality'spopulationspokeanindigenouslanguagein1900thaninthemedianmunicipality.Therelationshipbetweendroughtandinsurgencyislargeandstatisticallysignicantinbothsub-samples.Overall,theseresultsdocumentthatdroughtledtoinsurgencyinmunicipalitieswithawidevarietyofinitialcharacteristics,andthisrelationshipwasparticularlypronouncedinmore11 agriculturalmunicipalities.Identicationrequiresdroughtseveritybetween1906and1910tobeasifrandomlyassigned.Inotherwords,intheabsenceofdi erencesinrainfallduringtheseyears,munic-ipalitiesthatsu ereddroughtwouldnothavebeendi erentonaveragefrommunicipalitiesthatreceivedaboveaveragerainfall.Toshedlightontheplausibilityofthisassumption,Table4regressesavarietyofoutcomesfromthe1900MexicanPopulationCensusondroughtseverity,measuredasthepercentofnormalprecipitationinnon-harvestmonths.Thesamplesizesareslightlydi erentacrossoutcomes,asportionsofthe1900censusvolumeshavenotbeenpreservedforthestateofOaxaca.Forcomparisonpurposes,column1reportstherststagerelationshipbetweendroughtseverityandinsurgencyfromTable2.Thedependentvariableincolumn2isthepercentageofthemunicipalmalepopulationthatworkedinagriculturalin1900.10Thedependentvariableincolumn3isthepercentageofthepopulationlivinginhaciendas,landedestateswithanattachedlaborforce.Thedependentvariableincolumn4isthepercentageofthepopulationthatprimarilyspokeanindigenouslanguage,incolumn5itisthepercentagethatwasliterate,incolumn6itisthenumberofpublicemployeesper1,000inhabitants,andincolumn7itisthenumberofpoliceper1,000inhabitants.Thecorrelationsbetweentheseoutcomesanddroughtseverityareallstatisticallyin-signicant.Themagnitudesoftheserelationshipalsotendtobesmaller,relativetothesamplemean,thanthemagnitudeoftherelationshipbetweendroughtandinsurgency.Thenextsectionwilldocumentthatsomeoftheseoutcomesdivergeinmunicipalitieswithin-surgentactivityascomparedtomunicipalitiesthatdidnotexperienceinsurgentactivity,intheyearsfollowingtheRevolution.4Insurgency'simpactsonpolicyanddevelopmentThissectionusesaninstrumentalvariablesapproachtotestwhetherinsurgencya ectedtwoofthemostcentralMexicanfederalgovernmentpoliciesintheyearsfollowingtheRevolu-tion:agrarianreformandtheexpansionofthefederalbureaucracy.Duringthe1920sand1930s,thefederalgovernmentredistributedalmosthalfofMexico'ssurfaceareathroughamajoragrarianreformprogram,andthesizeofthefederalbureaucracyincreasedmorethanvefold.Indthatagrarianreformwasconcentratedininsurgentmunicipalities,whereasthereislittlerelationshipbetweeninsurgencyandtheexpansionofthefederalbureaucracy.Afterexamininginsurgency'simpactsonthesemajorpolicies,Itestwhetherithasexerted 10Nearlyallwomenarecountedinthe1900censusashavingadomesticoccupation,andhenceareexcludedfromthismeasure.12 persistente ectsoneconomicoutcomes.Instrumentalvariablesestimatesdocumentthatinsurgencyhasloweredincomeandindustrializationandincreasedthepercentageofthela-borforceinagricultureinthelong-run.Inthenextsection,Iprovideevidencethatagrarianreformisaplausiblecentralmechanismexplaininginsurgency'spersistenteconomicimpacts.4.1DataIobtaineddataonagrarianreformfromMexico'sonlineSistemadeInformaciondelPadronehistorialdenucleosagrarios,whichcompilesinformationonallgovernmentactionsrelatedtoagrarianreformandtitling,attheleveloftheejido,from1916untilthepresent.Dataonover31,000ejidoswereusedtocalculateinformationonmunicipal-levelagrarianreform,titling,andentryofejidalplotsintotheprivatedomain.Dataonincome,thelaborforce,publicemployees,andeducationaretakenfromMexicancensusdatafortheyears1900,1910,1930,1940,1960,2000,and2010.DataonhouseholdaccesstowaterarefromCONAPO(NationalPopulationCouncil)(2005).ElectoraldataarefromElectoral-BanamexandelectoralresultspublishedbytheElectoralTribunalsineachofMexico's31states.ThegeographiccharacteristicsarefromAcemogluandDell(2010).Dataonhomicides(1990-2006)arefromINEGIanddataondrugtrade-relatedviolenceandsubversionarefromcondentialgovernmentsources.DataonmunicipaltaxcollectionarefromSistemadeCuentasMunicipales,INEGI.4.2InsurgencyandgovernmentpoliciesTable5testswhetherinsurgencyin uencedthetargetingofagrarianreform,usingthepercentageofamunicipality'ssurfacearearedistributedbyagrarianreformasthedependentvariable.PanelAreportstheIVestimates,andforcomparisonpurposesPanelBreportstheOLSestimates.Thepercentnormaldroughtmeasureisusedastheinstrument.Robustnesstoalternativemeasuresofdroughtandtotheinclusionofadditionalcontrolsisdocumentedintheappendix.Overall,Table5providesstrongevidencethatinsurgencyledtoincreasedlandreform.TheIVestimatesincolumn1documentthatinthesampleasawhole,insurgencyincreasedthepercentageofmunicipalsurfacearearedistributedasejidosby21.8percentagepoints(s.e.=0:111),relativetothesamplemeanof48.9percent.Whenwefocusattentiononmunicipalitieswhereparticipationinagriculturalwasabovethesamplemedianin1900,theestimatede ectis31percentagepoints,andthise ectisstatisticallysignicantattheonepercentlevel.Incontrast,theimpactofinsurgencyissmallerandnotstatisticallydi erentfromzerointhesamplewithbelowmedianparticipationinagriculturalin1900.Thispattern13 iswhatwewouldexpect,giventhatthedemandforagrarianreformislikelytobehigherinareaswhereagreaterpercentageofthepopulationisemployedinagriculture.Whileproducersnowhavetheoptionofregisteringejidallandsintheprivatedomain,only2.5%ofejidallandshavebeenregisteredthusfar,soagrarianreformnumbersarelargelyre ectiveofhowmuchlandremainsintheejidalsectortoday.Incolumns1and2,theIVestimatesaresomewhatlargerthantheOLSestimates.Thereareanumberofreasonswhythiscouldoccur:becauseofmeasurementerrorintheinsurgencyvariable,becauseofomittedvariablesbiasintheOLSspecication,becausetheIVmeasuresalocalaveragetreatmente ectonmunicipalitiesinducedtotakeuparmsbydroughtseverity(whereastheOLSmeasuresthecorrelationacrossthefullsample),orbecausedroughtseverityviolatestheIVexclusionrestriction.Itisnotpossibletofullydisentangleorexplicitlytestthesepossibilitiessincetheydependoncharacteristicsthere-searcherdoesnotobserve.However,itisunlikelythataviolationoftheexclusionrestrictionistheprimaryreasonwhytheIVestimatesarelargerthantheOLSestimates,sincetheinstrumentisuncorrelatedwithahostofpre-characteristics(Table4).Moreover,whenIexaminetheestimatesacrossthedi erentsub-samplesinTable4,theinsurgencycoecientisnotsubstantiallydi erentinthesub-sampleswithlargerrststages.Thus,itappearsunlikelythattheIVestimatesre ectalocalaveragetreatmente ectthatonlyappliestoasmallsub-population.Incontrast,bothomittedvariablesbiasandmeasurementerrorappearplausible,giventheinherentdicultiesinquantifyinginsurgentactivitiesandthemanyfactorsthatcouldin uencethedecisiontotakeuparms.Forexample,whiletherearesomewell-knownexamplesofindigenousrevolutionaryleaders,onaveragemunicipalitieswithrevolutionaryactivityweresubstantiallylessindigenousin1900.Inturn,itislikelythatmoreindigenousplacesreceivedmorelandreform,sincetheirlandwasmorelikelytobetakenintherstplace,andthiswouldleadtoadownwardbiasinOLScoecientonrevolution.AmajorexpansionofthefederalbureaucracyalsooccurredinthedecadesfollowingtheRevolution.In1900,therewere3.2governmentemployeesper1,000municipalresidentsinthesampleexaminedinthispaper.By1940,thisnumberhadrisento20.Onaverage,16ofthesewerefederalgovernmentemployees.Table6exploreswhetherinsurgencydi erentiallya ectedthenumberofpublicemployeespresentinamunicipalityin1940.Giventhegreateramountofagrarianreform,wemightexpectthatthenumberofgovernmentemployeeswouldbehigherinmunicipalitiesthathadexperiencedinsurgentactivity.Ontheotherhand,sinceamajordemandofmanyinsurgentswasgreaterautonomyfromthecentralgovernment,totheextentthatthesedemandsweremetwewouldexpectlowerfederalbureaucracypres-ence.WeseefromTable5thattheIVdoesnotestimatestatisticallysignicantdi erences14 betweenmunicipalitiesthatexperiencedinsurgentactivityandthosethatdidnot.Ifany-thingthenumberoffederalandstateemployeeswaslowerinmunicipalitiesthatexperiencedinsurgency.Moreover,importantmeasuresofgovernmentbureaucracytodaydonotdi er.Column(5)documentsthatin2005thenumberofschoolteachersper1000schoolagedchildrenwasstatisticallyidenticalinmunicipalitieswithhistoricalinsurgencyandinthosewithouthistoricalinsurgency.Column(6)documentsthatlocaltaxreceiptsperdollarofmunicipalincomein2005alsodidnotdi er.4.3Insurgencyandlong-rundevelopmentTheprevioussectiondocumentedalargeimpactofinsurgencyonsubsequentagrarianre-form.Thissectionexamineswhetherinsurgencyhasexertedlong-runimpactsoneconomicprosperity.First,Table7testswhetherinsurgencyimpactseconomicprosperitytoday.Columns1through4examineincome,usingmicrodatafromthe2000MexicanPopulationCensus.TheIVcoecientreportedincolumn1estimatesthathistoricalinsurgencyhasloweredincomebyaround33%.Columns(2)through(4)dividethesampleintoindividualsworkinginagriculture,inindustry,andinservices.Careshouldbetakenininterpretingtheresultsgiventhatinsurgencymayalsoin uenceselectionintoeconomicsectors,buttheexerciseisneverthelessinformativeaboutwheretheincomee ectsareconcentrated.TheIVcoecientislargeandnegativeinallthreesectors,thoughitisnotstatisticallysignicantforagricul-ture.Thisisnotsurprisinggiventhatincometendstobepoorlymeasuredforhouseholdagriculturalproducers.AsinTable6,theIVcoecientsarelargerthantheOLScoecients,likelyforthesamereasonsthatwerediscussedintheprevioussection.Next,columns5and6examinethepercentageofhouseholdsinamunicipalitythatlackaccesstorunningwaterandtoelectricity.Householdsinmunicipalitieswithinsurgentactivityhistoricallyare14percentagepointsmorelikelytolackaccesstorunningwaterandaroundthreepercentagepointsmorelikelytolackaccesstoelectricity.Thesee ectsarelarge,giventhatinthesampleasawholearound11percentofhouseholdslackaccesstowaterandthreepercentlackaccesstoelectricity.Next,Table8testswhetherinsurgencyhasin uencedthesectoralallocationofthelaborforce.Columns1and2examinethepercentageofthemunicipallaborforceworkinginagricultureandindustryin2010,columns3and4dothesamefor1960,andcolumns5and6examinethepercentageofthemalepopulationworkinginagricultureandindustryin1940.Theinstrumentalvariablesestimatesdocumentthattoday,thepercentageofthelaborforceworkinginagriculturalisaround20percentagepointshigherinmunicipalitiesthat15 economicoutcomesandthenpresentempiricalevidenceonpotentialalternativemechanismsrelatinginsurgencytolong-runeconomicdevelopment.5.1LandreformEmpiricallyidentifyinghowMexicanlandreformhasa ectedlong-rundevelopmentrequiresalternativesamplesandidenticationapproaches,andhencespaceconstraintsrequirethatthisquestionbeexaminedinseparatework.Here,Idiscussexistingempiricalandhistoricalevidencesupportingtheplausibilityofagrarianreformasacentralchannelthroughwhichthee ectsofhistoricalinsurgencypersist.MostcloselyrelatedtothispaperisworkbyBeatrizMagaloni,BarryWeingast,andAl-bertoDiaz-Cayeros(2008)thatempiricallyexaminestheimpactoflandreformoneconomicgrowthandpolitics.Magalonietal.useastandardgrowthregressionframeworkappliedtoMexico's31statesbetween1950and1995tomeasuretheeconomicimpactoflandreform.TheyestimatethatMexicanGDPpercapitawouldhavebeen124percenthigherin1995hadtherebeennolandreforminMexico.TheyalsoprovideevidencethatthedistributionofejidallandswasakeyinstrumentforgeneratingpoliticalsupportforthePRI-Mex-ico'shistoricallydominantparty-andshowthatlandwasdistributedasafunctionofthepresidentialelectioncycleandsocialunrest.Theseresultsareconsistentwithlargeandhighlysignicantmunicipal-levelcorrelationsbetweenlandreformandcontemporaryeco-nomicandpoliticaloutcomes.Theyarealsoconsistentwiththehypothesisthatinsurgencya ectscurrenteconomicandpoliticaloutcomesatleastinpartthroughlandreform.Over-all,thisevidencesuggeststhatwhileMexico'spolicyofcreatingejidoswashighlyinecient,thegovernmentpursueditbecauseitfurtheredthePRI'sgoalsofreducinginstabilityandmaintainingpoliticalcontrol.Additionally,inongoingworkIexaminetheempiricalrelationshipbetweenagrarianstructureandindustrialdevelopment.Ishowthatwhilehigh-productivity,irrigatedagri-culturalareastendedtoindustrialize,thise ecthasbeenmutedinregionswithahighconcentrationofejidallands.Agribusinesseshavepreferredtolocateinareaswheretheycouldverticallyintegratebypurchasingfarms,whichhasnotbeenpossibleinejidalareasduetotheprohibitionsonlandmarkettransactions(Johnstonetal.,1987).Moreover,thefederalgovernment,whichformuchofthe20thcenturycontrollednearlyaquarteroftheMexicanfoodprocessingindustry,tendedtopurchasefoodandlocatestate-ownedprocess-ingindustriesnearwell-connectedprivatefarmers(Ochoa,2000).Giventhatindustryisarelativelyhigh-payingsector,di erencesinindustryarelikelytotranslateintodisparitiesinincome.17 Anotheralternativeinterpretationoftheresultsisthathistoricaldroughtseveritycouldexertdirectimpactsonmodernoutcomes,violatingtheIVexclusionrestriction.Toexaminethispossibility,Table11regressesthestudy'smainoutcomevariablesonseparatemeasuresofdroughtseverityin1906-1910andinallveyearperiodsbetween1960and1995.Thedependentvariableincolumn1isthehistoricalinsurgencyindicator.Thedependentvariableincolumn2isthepercentageofmunicipalsurfaceareaintheejidalsector,thedependentvariableincolumn3isthepercentageofthemunicipallaborforceworkinginindustryin2010,thedependentvariableincolumn4islogincomein2000,andthedependentvariableincolumn5isthepercentageofelectionsheldbetween1974and2009inwhichthepartycontrollingthemayorshipchanged.Thecoecienton1906-1910droughtseverityislargeandstatisticallysignicantinallcolumns.Incontrast,ofthe35droughtvariablesfromotherperiods(7ofwhichentereachregression),onlyoneismarginallysignicant.Thissupportsthestudy'sassumptionthat1906-1910droughtinteractedwithspecichistoricalcircumstancestoproducelong-lastinge ectsthroughinsurgency.Ihavealsoexploreddroughtsoccurringbetween1925and1960.Becauseweatherdataforthisperiodisquitelimited,itwouldreducethesamplesizesubstantiallytoincludethesedroughtmeasuresintheregressionanalysisreportedinTable11.However,Ihavelookedattheseperiodsseparatelyanddonotndevidenceofarelationshipbetweendroughtandcurrentoutcomes,eitherforthesub-sampleofmunicipalitiesthatalsohaveweatherdatafor1905-1910orforMexicoasawhole(resultsavailableuponrequest).6ConcludingRemarksThisstudyidentieshowinsurgencyduringtheMexicanRevolutiona ectedsubsequentpoliciesandeconomicdevelopmentbyusingdroughtseveritytoinstrumentforrevolution-aryactivity,conditionalonstatexede ects.InstrumentalvariablesestimatesdocumentthatinsurgentmunicipalitiesreceivedsubstantiallymorelandreformintheyearsfollowingtheRevolutionandaretodaypoorer,moreagricultural,lessindustrial,andlesspoliticallycompetitive.Basedonthequantitativeandhistoricalevidence,IhypothesizethattheMexicanstategainedamonopolyonviolenceinrebelliousregionsthroughimplementinglarge-scaleagrar-ianreformintheyearsfollowingtheRevolution.Agrarianreformimposedconsiderablerestrictionsonredistributedlandsandfosteredaclientalisticpoliticalsystemdominatedbyasinglepartybureaucracy.Thisstudydiscussesevidencethatagrarianreformsti edindus-trializationandeconomicgrowthinplacesthatreceivedmorelandreformrelativetoplacesthatreceivedless.19 Jacobs,I.(1982):RancheroRevolt:TheMexicanRevolutioninGuerrero,UniversityofTexasPress.Janvry,A.,G.Gordillo,andE.Sadoulet(1997):Mexico'ssecondagrarianreform:householdandcommunityresponses,1990-1994.Johnston,B.(1987):US-MexicoRelations:agricultureandruraldevelopment,StanfordUniversityPress.King,R.(1977):Landreform:aworldsurvey,WestviewPressBoulder.Knight,A.(1986a):TheMexicanrevolution:Counterrevolutionandreconstruction,Cam-bridgeUniversityPress.|||(1986b):TheMexicanRevolution:Porrians,Liberals,andPeasants,CambridgeUniversityPress.LaFrance,D.(2003):RevolutioninMexico'sheartland:politics,war,andstatebuildinginPuebla,1913-1920,SrBooks.LaFrance,D.andF.Madero(1989):TheMexicanRevolutioninPuebla,1908-1913:theMaderistamovementandthefailureofliberalreform,SRbooks.MartnezVasquez,V.andF.ChassendeLopez(1985):\LarevolucionenOaxaca1900-1930,"Mexico,InstitutodeAdministracionPublicadeOaxaca.Miguel,E.andG.Roland(2011):\Thelong-runimpactofbombingVietnam,"JournalofDevelopmentEconomics,96,1{15.Miguel,E.,S.Satyanath,andE.Sergenti(2004):\Economicshocksandcivilcon- ict:Aninstrumentalvariablesapproach,"JournalofPoliticalEconomy,112,725{753.Nunn,N.(2009):\TheImportanceofHistoryforEconomicDevelopment,"AnnualReviewofEconomics,1,65{92.Ochoa,E.(2000):FeedingMexico:Thepoliticalusesoffoodsince1910,ScholarlyRe-sourcesInc.Olea,H.(1964):BrevehistoriadelaRevolucionenSinaloa(1910-1917),PatronatodelInstitutoNacionaldeEstudiosHistoricosdelaRevolucionMexicana.Pasztor,S.(2002):TheSpiritofHidalgo:TheMexicanRevolutioninCoahuila,2,Uni-versityofCalgaryPress&MichiganStateUniversityPress.22 Table1:SummaryStatistics MeanS.D.p10p90 (1)(2)(3)(4) Rainfall0.780.210.491.00Insurgency0.590.490.01.0Agrarianreform0.490.260.130.82Publicemployees/1,000inhab.(1940)20.039.52.025.4Logincome(2000)7.70.96.88.8Percentagricultural(2010)0.070.100.020.45Percentindustrial(2010)0.250.080.140.37Percentagricultural(1960)0.260.280.161.00Percentindustrial(1960)0.210.120.010.23Percentagricultural(1940)0.240.190.190.53Percentindustrial(1940)0.120.090.010.14Percentpartyalternations(1974-2009)0.240.190.000.50Percentpartyalternations(1974-1993)0.080.190.000.40Percentpartyalternations(1994-2009)0.370.270.000.80 Notes:Thistableprovidesthemean,standarddeviation,10thpercentile,and90thpercentileforrainfallandforthepaper'smainoutcomevariables.Rainfallis1906-1910precipitationduringnon-harvestmonthsasapercentageoflong-runaverageprecipitation,censoredaboveatone.Insurgencyisadummyequaltooneifthemunicipalityexperiencedinsurgentactivityduring1910-1918andequaltozerootherwise.Agrarianreformisthepercentageofamunicipality'ssurfacearearedistributedthroughagrarianreform.Logincomeisthelogofmeanmunicipalincomein2000.Percentagricultural(2010)andPercentagricultural(1960)arethepercentageofthemunicipallaborforceworkinginagricultureintheyears2010and1960,respectively.Percentindustrial(2010)andPercentindustrial(1960)arethepercentageofthemunicipallaborforceworkinginindustryintheyears2010and1960,respectively.PercentAgricultural(1940)isthepercentageofmunicipalmalepopulationworkinginagriculturein1940,andPercentIndustrial(1940)isthepercentageofmunicipalmalepopulationworkinginindustryin1940.Percentpartyalternationsgivesthepercentageofelectionsinwhichthepartycontrollingthemayorshipchanged,duringthetimeperiodlistedinparentheses.Sourcesforallvariablesareprovidedinthetext.24 Table3:Compliers Dependentvariableisinsurgency.Sampleis: CloseFarMoreLessMoreLessMoreLessBaselineU.S.agriculturalhaciendasindigenous (1)(2)(3)(4)(5)(6)(7)(8)(9) Rainfall-0.770***-0.650**-0.917***-1.288***-0.498*-0.646***-0.935***-0.649*-0.671***(0.176)(0.254)(0.224)(0.293)(0.265)(0.243)(0.268)(0.334)(0.242)R-squared0.4490.3710.5020.5340.4990.5220.4750.5300.434Observations21097113104103106103100105MeanDep.Var.0.590.680.510.580.620.570.620.510.67 Notes:Rainfallmeasures1906-1910precipitationduringnon-harvestmonthsasapercentageoflong-runaverageprecipitation.Thedependentvariableisinsurgency,adummyequaltooneifthemunicipalityexperiencedinsurgentactivityduring1910-1918andequaltozerootherwise.Allcolumnsincludecontrolsforlong-runaverageprecipitation,long-runaverageminimumandmaximumtemperature,slope,andelevation,aswellasstatexede ects.Thesamplesaredenedinthecolumnheadings.Column1examinesthefullsample.MunicipalitiesinstatesthatarecloserthanthemediandistancetotheU.S.areexaminedincolumn2,whereasmunicipalitiesinstatesfartherthanthemediandistancefromtheU.S.areexaminedincolumn3.Municipalitiesthathadahigherpercentageoftheirmalepopulationworkinginagriculturein1900thanthemedianmunicipalityareexaminedincolumn4,whereasmunicipalitiesthathadalowerpercentageoftheirpopulationworkinginagricultureascomparedtothemedianmunicipalityareexaminedincolumn5.Municipalitiesthathadahigherpercentageoftheirpopulationlivinginanhaciendain1900thanthemedianmunicipalityareexaminedincolumn6,whereasmunicipalitiesthathadalowerpercentageoftheirpopulationlivinginanhaciendaascomparedtothemedianmunicipalityareexaminedincolumn7.Municipalitiesinwhichahigherpercentageofthepopulationspokeanindigenouslanguagein1900thaninthemedianmunicipalityareexaminedincolumn8,whereasmunicipalitiesinwhichalowerthanaveragepercentageofthepopulationspokeanindigenouslanguageareexaminedincolumn9.Robuststandarderrorsareinparentheses. Table5:AgrarianReform FullMoreLesssampleagriculturalin1900 (1)(2)(3) PanelA:IVInsurgency0.218*0.311***0.052(0.111)(0.110)(0.234)PanelB:OLSInsurgency0.100***0.146***0.087(0.037)(0.054)(0.060)Observations210104103MeanDep.Var.0.490.520.45 Notes:Insurgencyisadummyequaltooneifthemunicipalityexperiencedinsurgentactivityduring1910-1918andequaltozerootherwise.Thedependentvariableisthepercentageofmunicipalsurfacearearedistributedbyagrarianreform.PanelAreportsinstrumentalvariablesestimates,withinsurgencyinstrumentedby1906-1910precipitationduringnon-harvestmonthsasapercentageoflong-runaverageprecipitation.PanelBreportsordinaryleastsquaresestimates.Allcolumnsincludecontrolsforlong-runaverageprecipitation,long-runaverageminimumandmaximumtemperature,slope,andelevation,aswellasstatexede ects.Thesamplesaredenedinthecolumnheadings.Column1examinesthefullsample.Municipalitiesthathadahigherpercentageoftheirmalepopulationworkinginagriculturein1900thanthemedianmunicipalityareexaminedincolumn2,whereasmunicipalitiesthathadalowerpercentageoftheirpopulationworkinginagricultureascomparedtothemedianmunicipalityareexaminedincolumn3.Robuststandarderrorsareinparentheses.28 Table7:Economicoutcomestoday OverallAgriculturalIndustrialServicesPercentlogwagenowaternoelectricity (1)(2)(3)(4)(5)(6) PanelA:IVInsurgency-0.292**-0.322-0.289*-0.218**14.095**2.922*(0.141)(0.274)(0.169)(0.109)(6.255)(1.657)PanelB:OLSInsurgency-0.109***-0.082*-0.122***-0.086***0.7150.404(0.021)(0.044)(0.021)(0.019)(1.603)(0.465)Observations734,12753,363222,267458,497210210Clusters210210210210210210MeanDep.Var.7.727.137.737.7811.123.32 Notes:Insurgencyisadummyequaltooneifthemunicipalityexperiencedinsurgentactivityduring1910-1918andequaltozerootherwise.PanelAreportsinstrumentalvariablesestimates,withinsurgencyinstrumentedby1906-1910precipitationduringnon-harvestmonthsasapercentageoflong-runaverageprecipitation.PanelBreportsordinaryleastsquaresestimates.Allcolumnsincludecontrolsforlong-runaverageprecipitation,long-runaverageminimumandmaximumtemperature,slope,andelevation,aswellasstatexede ects.Thedependentvariableincolumns1through4islogincomein2000.Thesampleincolumn1includesallindividualsearningpositiveincome,incolumn2itincludesindividualsworkinginagriculture,incolumn3itincludesindividualsworkinginindustry,andincolumn4itincludesindividualsworkinginservices.Incolumn5thedependentvariableisthepercentageofhouseholdsinamunicipalitywholackaccesstorunningwater,andincolumn6itisthepercentageofhouseholdswholackaccesstoelectricity.Robuststandarderrors,clusteredbymunicipality,areinparentheses.30 Table9:Politicalcompetition Percentalternations74-0994-0974-93 (1)(2)(3) PanelA:IVInsurgency-0.327***-0.382***-0.210*(0.101)(0.143)(0.111)PanelB:OLSInsurgency-0.028-0.0500.006(0.031)(0.048)(0.037)Observations205205203Meandep.var0.2440.3690.088 Notes:Insurgencyisadummyequaltooneifthemunicipalityexperiencedinsurgentactivityduring1910-1918andequaltozerootherwise.PanelAreportsinstrumentalvariablesestimates,withinsurgencyinstrumentedby1906-1910precipitationduringnon-harvestmonthsasapercentageoflong-runaverageprecipitation.PanelBreportsordinaryleastsquaresestimates.Allcolumnsincludecontrolsforlong-runaverageprecipitation,long-runaverageminimumandmaximumtemperature,slope,andelevation,aswellasstatexede ects.Thedependentvariableisthepercentageofelectionsinwhichthepartycontrollingthemayorshipchanged,duringthetimeperiodlistedinparenthesesinthecolumnheadings.Robuststandarderrorsareinparentheses.32 Table11:Droughtsinotherperiods Insurgency%ejido%agriculturallogincome%alternations (1)(2)(3)(4)(5) Rainfall1906-0.793***-0.182*-17.490**0.153**0.240***(0.230)(0.089)(8.382)(0.072)(0.072)Rainfall19600.4250.16112.114-0.038-0.082(0.289)(0.186)(12.781)(0.141)(0.148)Rainfall1965-0.0140.13447.1000.075-0.332(0.456)(0.406)(33.486)(0.261)(0.307)Rainfall19700.2470.245-21.345-0.1120.135(0.715)(0.280)(19.224)(0.150)(0.234)Rainfall19750.0720.251-26.5710.412-0.244(0.527)(0.513)(32.687)(0.257)(0.146)Rainfall19800.5860.0773.426-0.411*0.155(1.002)(0.368)(16.466)(0.205)(0.289)Rainfall19850.245-0.299-20.3260.0640.219(0.424)(0.368)(30.188)(0.180)(0.247)Rainfall19900.1160.11229.701-0.079-0.336(0.547)(0.308)(19.970)(0.173)(0.267)Observations210210210733,153205R-squared0.4670.4600.3400.3570.358 Notes:Rainfallmeasuresprecipitationduringnon-harvestmonthsasapercentageoflong-runaverageprecipitationfortherespectiveperiod.Allcolumnsincludecontrolsforlong-runaverageprecipitation,long-runaverageminimumandmaximumtemperature,slope,andelevation,aswellasstatexede ects.Thedependentvariableincolumn1isinsurgency,adummyequaltooneifthemunicipalityexperiencedinsurgentactivityduring1910-1918andequaltozerootherwise.Thedependentvariableincolumn2isthepercentageofmunicipalsurfaceareaintheejidalsector,thedependentvariableincolumn3isthepercentageofthemunicipallaborforceworkinginagriculturein2010,thedependentvariableincolumn4islogincomein2000,andthedependentvariableincolumn5isthepercentageofelectionsheldbetween1974and2009inwhichthepartycontrollingthemayorshipchanged.Robuststandarderrorsareinparentheses.34 Figure2:Insurgency Notes:Insurgencyduring1910-1918isplotted,followingtheschemegiveninthelegend.Stateboundariesareinblack.