,Vol.81,No.5(September,2013),2033–2053REVOLTONTHENILE:ECONOMICSHO - PDF document

,Vol.81,No.5(September,2013),2033–2053REVOLTONTHENILE:ECONOMICSHOCKS,RELIGION,ANDPOLITICALPOWERHarvardUniversity,Cambridge,MA02138,U.S.A. ThecopyrighttothisArticleisheldbytheEconometricSociety.Itmaybedownloaded,printedandreproducedonlyforeducationalorresearchpurposes,includinguseincoursepacks.NodownloadingorcopyingmaybedoneforanycommercialpurposewithouttheexplicitpermissionoftheEconometricSociety.ForsuchcommercialpurposescontacttheOfceoftheEconometricSociety(contactinformationmaybefoundatthewebsitehttp://www.econometricsociety.orgorinthebackcoverof).ThisstatementmustbeincludedonallcopiesofthisArticlethataremadeavailableelectronicallyorinanyother ,Vol.81,No.5(September,2013),2033–2053REVOLTONTHENILE:ECONOMICSHOCKS,RELIGION,ANDPOLITICALPOWERUsingcenturiesofNileooddata,IdocumentthatduringdeviantNileoods,Egypt’shighest-rankingreligiousauthoritywaslesslikelytobereplacedandrelativeallocationstoreligiousstructuresincreased.Thesendingsareconsistentwithhistori-calevidencethatNileshocksincreasedthisauthority’spoliticalinuencebyraisingtheprobabilityhecouldcoordinatearevolt.Indthattheavailabledataprovidesupportforthisinterpretationandweighagainstsomeofthemostplausiblealternatives.Forexample,IshowthatwhileNileshocksincreasedhistoricalreferencestosocialunrest,deviantoodsdidnotincreaseaproxyforpopularreligiosity.Together,theresultssug-gestanincreaseinthepoliticalpowerofreligiousleadersduringperiodsofeconomic:Politicaleconomy,religion,economicshocks,socialconict,politicalECENTRESEARCHHASSHOWNthateconomicdownturnsincreasetheproba-bilityofdemocraticchange(AcemogluandRobinson(2006)BurkeandLeighBrücknerandCiccone(2011)).Formuchofhistory,however,democ-ratizationdidnotfolloweconomicshocks.AtleastsincetheEnlightenment,scholarshaveattributedthehistoricresilienceofautocraticregimestotheac-tionsofreligiousleadersworkingin“alliancewithsocialhierarchyandoppres-sivegovernment”(,p.103)).Althoughthinkersincluding(1776[2009])Marx(1844[1982]),and,morerecently,North,Wallis,andWeingast,p.39)havestressedtheimportanceofreligiousauthoritiesindetermininginstitutionalequilibria,welackempiricalinvestigationsoftheextenttowhichreligiousleaderswieldpoliticalpower.ThispaperdocumentsanegativerelationshipbetweenarguablyexogenousNileshocksandtheprobabilityofchangeinIslamicEgypt’shighest-rankingreligiousauthority—knownastheheadjudge—usingcenturiesofNileooddata.Inotherwords,IndthattheheadjudgewaslesslikelytobedismissedbythesovereignortoresignhispostwhentheannualNileoodwasabnormallyhighorlow.Inaddition,Ishowthatduringsuchoods,theconstructionofreligiousstructuresrelativetosecularstructuresincreased.IthankCREIandIAS,Princetonforhostingmewhilepartofthisprojectwascarriedout,andIthanknumerousindividualsandseminarparticipantsforhelpfulsuggestions.Iparticu-larlythankaco-editorandfouranonymousrefereesfordetailedsuggestionsandcommentsthatsignicantlyimprovedthepaper.ThelibrarystaffattheBibliothèqueNationaledeFrance,ManuscriptsOrientauxgreatlyfacilitatedtheNiledatatranscription.Anappendixthatprovidesadetaileddescriptionofthedata,additionalresults,andsensitivitychecksisavailableintheSupplementalMaterial(Chaney(2013)).Iamresponsibleforanyremainingerrors.Throughoutthispaper,IfollowAcemogluandRobinson,p.173)anddenepoliticalpoweras“ameasureofhowinuentialaparticulargroup(orindividual)isinthepoliticalarenawhenthereisconictoverwhichpolicyshouldbeimplemented.”©2013TheEconometricSociety10.3982/ECTA10233 ERICCHANEYTheseempiricalpatternsareinterestingbecausetheyareconsistentwithhistoricalevidencethatNileshocksincreasedthepoliticalpowerofreligiousleaders.Scholarssuggestthattheheadjudgeenjoyedwidespreadpopularin-uenceandthathispoliticalpowerstemmedfromthisinuence.Duringperi-odsofsocialunrest—suchasthosecreatedbydeviantNileoods—thejudge’spoliticalinuenceisbelievedtohaveincreased.Socialunrestisthoughttohavestrengthenedthejudge’sbargainingpositionbyrenderingtheprobabil-ityofsuccessofajudge-coordinatedrevolt(whichgenerallyremained“offtheequilibriumpath”)abnormallyhigh.Igatheradditionaldatatofurtherinvestigatetheempiricalrelevanceofthisconceptualframework.Forexample,IprovideevidencethatduringNileshocks,thepricesoffoodstuffswerehigherandtheincidenceofsocialconictwasgreaterthanduringotherperiods.TheseresultsareconsistentwiththeclaimthatNileshocksincreasedthepropensityforsocialunrest.Additionalresultsweighagainstsomeofthemostplausiblealternativeex-planations.Forexample,IshowthathistoricalreferencestoreligiouspracticesdidnotincreaseduringNileshocks.ThisresultcastsdoubtonexplanationsthatstresstheimportanceofNile-inducedincreasesinreligiosityorthepro-visionofreligiousservices.IalsoshowthatNileshocksdidnotaffectthere-placementprobabilityofjudgeswithoutsizeablefollowings.ThisresultisnotconsistentwithexplanationsthatsuggestthatNileshocksdecreasedthere-placementprobabilitiesofallbureaucraticorreligiousofcials.WhilethisevidenceisconsistentwithaNile-inducedincreaseinthejudge’spoliticalpower,datalimitationsmakeitimpossibletoruleoutcompletelyallalternativeinterpretations.Forexample,Nile-inducedunrestmighthaveledthepoliticalleaderwhoappointedtheheadjudgetodeferjudgereplacementandtoincreaserelativeallocationstoreligiousstructurestochannelaidtotheMyinterpretationoftheresultsiscloselyrelatedtogrowingevidencethateconomiccrisesincreasetheprobabilityofthecollapseofautocraticregimesbytemporarilyalteringthebalanceofpoliticalpower(AcemogluandRobin-son(2001)BergerandSpoerer(2001)Acemoglu,Johnson,Robinson,andYared(2005)AcemogluandRobinson(2006)BurkeandLeigh(2010)BrücknerandCiccone(2011)).Thispapersuggeststherelevanceofthislit-eratureinthepre-modernerabyprovidingevidencethatNileshocksalteredthebalanceofpoliticalpowerinfavorofEgypt’shighest-rankingreligiousau-thority.Tothebestofmyknowledge,thisistherstempiricalevidencethatreligiousleadershistoricallywieldedpoliticalpower.Noneoftheshocksdocumentedinthispaperledtothecollapseofpre-modernEgypt’sautocraticinstitutionalarrangements.Somehistorianshavesuggestedthatreligiousleaderscontributedtothisinstitutionalresiliencebyusingtheirpopularinuencetomitigatethethreatof“revolutionarydevel- REVOLTONTHENILEopments”duringperiodsofcrisis(Lapidus,p.153)).EvidencethatthepoliticalpowerofEgypt’shighest-rankingreligiousleaderincreasedduringpe-riodsofeconomicdownturnisbothconsistentwiththisviewandsuggeststhatfutureresearchinvestigatingthepoliticalroleofreligiousleadersmayenhanceourunderstandingofthehistoricresilienceofautocraticinstitutions.Theremainderofthepaperproceedsasfollows.Therstsectionprovidesanaccountoftheinstitutionalframeworkinpre-modernEgyptandtheeffectsofdeviantNileoodsonthebalanceofpoliticalpowerwithinthisframework.Thesecondsectiondescribesthedata,documentsthenegativerelationshipbetweenheadjudgereplacementandNileshocks,andinvestigatestheeffectsofNileshocksontheconstructionofreligiousstructures.Thethirdsectionempiricallyinvestigatespotentialcausalchannels.Afourthsectionconcludes.HISTORICALBACKGROUNDInstitutionsandPoliticalPowerinIslamicEgyptAuniqueinstitutionalframeworkcharacterizedbytheintroductionofforeign-bornmilitaryslavesemergedintheareasconqueredbyArabarmiesduringthecenturiesafterthedeathoftheprophetMuhammad(BlaydesandChaney(2013)Aspre-Islamicelitesdisappeared,Is-lamicreligiousassociationsbecamethebasisforcommunalorganizationandreligiousleadersemergedastheprimarycheckonthepowerofthesovereignbackedbyhisarmyofslaves(Lapidus,p.189)).Althoughmilitaryslaveswereintroducedacrosstheregionfromanearlydate,thetransformationofpre-IslamicsocialstructurestookcenturiesLapidus,pp.183,189)).ManyoftheseinstitutionaltransformationsarethoughttohavebeenrelatedtotheprocessofpopularconversiontoIslamandtheconcomitantincreaseinthepoliticalpowerofMuslimreligiousleaders,p.45)).Whilethesedevelopmentsoccurredatdifferentdatesacrosstheregion,itisgenerallyagreedthattheriseoftheAyyubiddy-nastyin1169CEmarksthecompletionofthisprocessinEgypt(Lapidusp.279)).Afterthisdate,thebalanceofpoliticalpowerbetweenreligiousandmilitaryleaders—andtheaccompanyinginstitutionalequilibrium—isbelievedtohaveremainedbroadlyconstantintothemodernera(pp.211–212)).Politicalscientistsincluding,p.46)havealsosuggestedthatinthepast,religionwasusedtosupportnondemocraticrule.Foraninvestigationofwhysomemodernauthoritarianregimesaremoreresilienttoeco-nomicshocksthanothers,seeGeddes(1999).Forworksuggestingalinkbetweenreligionandpoliticalstructure,seeLipset(1994)Barro(1999)IslamemergedintheArabianPeninsulaintheseventhcenturyandArabMuslimsbythemiddleoftheeighthcenturycontrolledterritoriesstretchingfromtheIberianPeninsulatotheIndusValley.Consequently,throughoutthepaper,IdrawonevidencefromEgypt’sIslamichistoryfollow-ingtheriseoftheAyyubiddynastyin1169andpriorto1805unlessotherwisenoted.Inaddition, ERICCHANEYConsistentwiththepremiseunderlyingtheabovehypothesis,scholarshipsuggeststhatthepoliticalpowerofreligiousleadersowedfromtheircon-troloverpopularsupport.Forexample,Al-SayyidMarsot,pp.133–134)notedthat“[i]twasthroughusingthethreatofrousingthemobthat[religiousleaders]couldrestraintheauthorities,whorecognizedthedangersbehindthethreat.”Napoleonnotedthisduringhisshort-livedconquestofEgyptin1798.Heremarkedthatreligiousleaderswerethe“naturalleaders”ofnativeEgyptiansandappointedthemtohigh-rankingpositionsinhisrulingcouncil,p.173)).AftertheAyyubidconquestin1169andpriortotheOttomanconquestofEgyptintheearlysixteenthcentury,theheadjudgewasEgypt’smostinu-entialreligiousleader(Petry,p.231)).Thisleaderisbelievedtohaveenjoyed“vastpersonalauthority”andisthoughttohavebeenabletoprovide“massivepopularbacking”(Lapidus,pp.134,136)).Thesovereigngen-erallyappointedheadjudgesfromalistofcandidatespresentedbyacouncilofreligiousleadersorsoughtoutleaderswithsubstantialpopularfollowings.Rulersdidnotgenerallyappointservilecandidatesbecausetheywereunlikelytoretainpopularinuence.Religiousleaderssuchastheheadjudgeenjoyedpopularinuencebe-causetheyactedasintermediariesbetweenthepopulaceandthesovereign.Inreturnforpopularsupport,religiousleadersextractedresourcesfromthesovereignthattheythendistributedtoboththemselvesandtheir“constituen-cies”(Lapidus,p.189)).Inadditiontosuchpatronage,itisalsoundoubt-edlytruethatideology(i.e.,religiousbeliefs)contributedtotheirinuence.Giventhisintermediaryrole,itisnotsurprisingthattheheadjudgeoftenconictedwiththesovereignoverpolicyimplementation.JudgesadvocatedfortheimplementationofIslamiclaw(theSharia),whichwasinterpretedineachperiodwiththeinterestsofboththepopulaceandreligiousleadersinmind.Inparticular,theheadjudge“resisted,haggledanddefendedagainstabuse”and“stoodagainstconscationswithoutconsentandpropercompen-sation”(Lapidus,p.140)).Thisreligiousleaderalsofavoredprohibitionsof“deviant”behavior(suchaswinedrinkingandprostitution)andadvocatedfortheconstructionofreligiousstructures.Thesovereigngenerallyaimedtominimizesuchconcessionssincetheyreducedtheamountofresourcesavail-abletohimselfandthemilitary.Forexample,resourcesallocatedtoreligiousstructureswereheldinreligiousendowments()thatthemilitarycouldnottax.Similarly,taverns,prostitution,andhashishconsumptionallyieldedsizeabletaxrevenues.Thesetaxrevenuessharplydeclinedifsuchactivitieswereprohibited(Lapidus,p.172)). attimesIsubstitutesovereignformilitary/andheadjudgefor(religiousscholars)forexpositionalease.Petry,p.315)and1441[1997],II,p.228;III,pp.238–239;IV,p.101).See1441[1997],III,pp.353,383,384)fortheimportanceofajudge’spopularfollowing.1441[1997],VII,p.450)andPetry,p.320). REVOLTONTHENILEWhentheheadjudgeandsovereignconictedoverpolicyimplementation,theheadjudgewasmorelikelytoresignhispostortobedismissedbytheWhydidtheheadjudgeoftenchoosetoresignorbedismissedratherthanendorseconscationsandotherviolationsofIslamiclaw?Asanintermediarybetweenthesovereignandthepopulace,theheadjudge’sinu-encewasrootedinhiscontroloverpopularsupport.SinceaheadjudgewhoregularlysanctionedviolationsoftheShariawaslikelytoseehispopularin-uencedecline,judgesoftenpreferredtobereplacedwhentheywereunabletopreventthesovereignfromimplementingpoliciesthatwentagainstthein-terestsoftheir“constituencies.”Suchjudgesseemtohaveretainedorevenincreasedtheirpopularstandingandweresometimesreappointedtotheheadjudgeshipatalaterdate.NileShocksandtheBalanceofPoliticalPowerItiswellknownthatpre-modernEgypt’seconomicactivitydependedheav-ilyontheNile’sannualood.Acomplexsystemofdikesandirrigationnet-workshelpedharnesstheood’sagriculturalpotential,makingEgyptianagri-culturalyieldssomeofthehighestinthepre-modernworld.However,whentheNile’sannualoodsignicantlydeviatedaboveorbelowitsoptimallevel(theseepisodesarereferredtothroughoutthetextasdeviantNileoods,Nileshocks,orNilefailures),agriculturaloutputdroppedsharplywhencropswereharvestedthefollowingspring.SincesharpNiledeviationscouldleadtowidespreadfamine,EgyptianscloselymonitoredtheNile’srateofriseduringthesummer.Asloworexces-sivelyrapidriseledtoabruptincreasesinthepricesoffoodstuffsasindividualshoardedsuppliesinpreparationforshortagesthefollowingyear.Thissharpriseinthepriceoffoodstuffsresultedinlocalizedriotsinwhichpeoplesackedbakeriesandlootedstores(,p.473)).HistoricalevidencesuggeststwochannelsthroughwhichNileshocksmayhaveincreasedthepoliticalpoweroftheheadjudge.First,withouttheaidoftheheadjudge,coordinationoflocalizedriotsintoawidespreaduprisingisthoughttohavebeen“impossible”(Lapidus,p.191)).SinceNileshocksincreasedthelikelihoodofapopularrevolt,theymadetheheadjudge’sabil-itytocontrolandchannelsuchriotsintoactionsthatdidnotthreatenthesovereignmorevaluable.Second,itispossiblethatNileshocksincreasedcon-ictsbetweenfactionsintheEgyptianmilitary.ItisbelievedthatduringsuchForafewillustrativeexamples,seeIbnHajar1449[1998],pp.231,246,392).IbnHajar1449[1998],pp.29,231)forexamplesofsuchjudges.Thisevidencesuggeststhepre-modernrelevanceoftheinsightthat“pooreconomicperfor-mancereducesthebargainingpowerofauthoritarianincumbentsandincreasesthestrengthofoppositions”(HaggardandKaufman,p.267)).ForevidenceofsuchNile-inducedconictduringtheFatimiddynasty(969–1169),seeLev,pp.14,76). ERICCHANEYperiodsofmilitaryconict,theheadjudge’ssupport“andwithitmassivepop-ularbacking,wasoftenofdecisiveimportance”indeterminingthefateofthesovereign(Lapidus,p.134)).DuringperiodsofsocialunrestsuchasthoseinducedbydeviantNileoods,scholarshavenotedthatit“wasnotunusualforthe[sovereign]tobegthe[headjudge]tohelp[him]preservethepeaceandkeepthepopulationtractable”Al-SayyidMarsot,p.153)).ThisevidencesuggeststhatNileshocksincreasedtheheadjudge’sleverageoverthesovereignbecausetheseshocksmadetheprobabilityofsuccessofajudge-coordinatedrevolt(whichgenerallyremainedofftheequilibriumpath)abnormallyhigh.AdditionalqualitativeevidenceisconsistentwiththeclaimthatNileshocksledtoanincreaseintheheadjudge’spoliticalpower.Forexample,sovereignsseemtohaveincreasedtheimplementationoftheheadjudges’preferredpoli-ciesduringNilefailures.Onehistoriansummedupthispolicyshiftbynot-ingthatduringperiodsofNilefailure,“thesultanwouldbowto[...]pres-sure[fromtheheadjudge]andenforcedecreesagainst[...]prostitution,hashisheating,beerdrinking,thewearingofimmodestorover-luxuriousdress[or]ChristianandJewishfunctionarieslordingitoverMuslims”(p.50)).Thehistoricalrecordalsoprovidesevidencethatthesovereignin-creasedpaymentsandperquisitestoreligiousleadersduringNilefailures.Importantlyfortheempiricalanalysisbelow,thehistoricalrecordincludesexamplesofthesovereignincreasingallocationstoreligiousstructuresdur-ingNileshocks.Onesuchexampleoccurredin1420whentheNileroseatanabnormallyslowrate.Inresponse,thesovereignorderedtheconstructionofamosqueandtherepairofanotherreligiousstructure(TaghriBirdiBirdi,III,pp.77–79)).Similarly,duringthedeviantoodspanningtheNileyear1415–1416thesovereigngave“thetreasureralargesumandorderedhimtogodowntoCairoanddistributeitamongthemosques”(TaghriBirdiBirdi,III,p.38)).Suchallocationsparticularlybenettedtheheadjudgeandotherreligiousauthorities,sincetheyreceivedstipendsinreturnforoverseeingandstafngthesestructures.Howdidtheheadjudgecontrolthecoordinationofpopularsupport?Theheadjudgecouldissuejudicialopinionsthatjustiedrebellions.Suchrulingswouldbepubliclyannouncedfromthepulpitofmosquesandcouldresultinacoordinatedpopularuprising.Theheadjudge’sin-uencewithlower-rankingreligiousleaderswasalsoimportant.Theseleadershadbothlocalknowledgeandinuence,andcouldhelporganizethepopulace.ForanexampleofincreasedpaymentstoreligiousleadersfollowingaNileshock,see1441[1997],IV,p.269).Foranexampleofincreasedperquisites,seetheeffectofNilefailureontheabilityofreligiousleaderstoridehorsesinTaghriBirdi1468[1990],pp.220,238).Forevidencethatreligiousstructuresdisproportionatelybenettedtheheadjudgeandotherreligiousleaders,see,pp.113,181).Forevidencethattheydecreasedresourcesavail-abletothemilitary,see,p.279). REVOLTONTHENILENILEFLOODS,JUDGEREPLACEMENT,ANDRELIGIOUSSTRUCTURESTheevidencepresentedintheprevioussectionsuggeststhatduringperiodsofsocialunrest,thepoliticalpoweroftheheadjudgeincreased,leadingtoanincreaseintheimplementationofhispreferredpolicies.Thisincrease,inturn,lessenedconictbetweenthesovereignandtheheadjudge,resultinginanequilibriumdecreaseinthejudgereplacementprobability.Inthissection,Iinvestigatetheextenttowhichthedataareconsistentwiththesepredictions.NileFloodandJudgeReplacementDataEgyptianshavemeasuredtheheightoftheannualNileoodsinceveryearlytimes.The“nilometer”ontheislandofRaudainCairoconsistedofarectan-gularwellthatwasconnectedtotheNilebythreeconduits.AverticalcolumninthemiddleofthewellservedasagaugeandwasusedtomeasuretheheightoftheNile.TheNileooddataprovidedbyIbnal-Hijazi(1470)coverthein-in-6221469]CEandcomefromacopyofthemeasurementstakenatthisnilometer(Popper,p.91)).Scholarsagreethatthesedataprovidecred-ibleestimatesoftheNile’sannualoodlevelandhavebeenusedextensivelyintheclimatologyliterature.ThemonthandyearoftheheadjudgechangesusedinthispaperaretakenIbnHajar(1449[1998])andcovertheintervalinterval6411438]CE.IusetheNileyear(whichrunsfromJulytoJuly)astheunitofobservationandomitNileyearsthatareincompletelycoveredbythejudgedata.Whenmerged,theooddataandheadjudgedatacovertheNileyearsyears6411437].Inthebaselinesample,IdropallobservationspriortothestartoftheAyyubiddynastyin1169andallobservationsafter1425.Idroptheearlyob-servationsbecause,ingeneral,theEgyptianheadjudgehadaninsignicantpopularfollowingbefore1169.Theearliestsuchjudgeshavebeendescribedasgovernment-appointedadministratorsandas“anintegralpartoftherul-ingclass”(,p.39)).AlthoughEgyptianheadjudgesincreasinglydevelopedanindependentpowerbaseaslocalpopulationsconvertedtoIs-lam,thesedevelopmentswereinterruptedbytheFatimidconquestofEgyptLapidus,p.279)).DuringtheFatimidperiod(969–1169),theShiaheadjudgerevertedtotheroleofgovernment-appointedadministratorandhadanegligiblepopularfollowingamongthemainlySunniMuslimpopulation.Thus,itisonlyaftertheriseoftheAyyubidsin1169thattheheadjudgeisbelievedtohaveregularlycommandedsignicantpopularsupport.IomitAlthoughtheFatimiddynastywasShiaandappointedaShiaheadjudge,itislikelythattheFatimidsprovidedpatronagetoinuentialSunnireligiousleaders(seeLapidus,p.281)).ThisremarkrefersprimarilytotheShaiheadjudge,sincethemajorityofEgypt’sMuslimpopulationinthisperiodbelongedtotheShaischool.Whenfourheadjudgesfromdifferentschoolsweresimultaneouslyappointedafter1265,itseemslikelythatthosewhohadinsignif-icantfollowingsweremoregovernment-appointedadministratorsthanintermediariesbetweentherulerandthepopulace. ERICCHANEY 1.—Nileshocksandjudgereplacementonthebaselinesample.Graydotsatthetopofthegraphdenoteyearswithjudgereplacement;blackdotsdenoteNileshocks.ThetimeseriesplotoftheNile’sannualmaximumonthebaselinesampleisalsoprovidedforcomparison.yearsafter1425becausetherewasanabnormallyintenseconcentrationofNileshocksafterthisdate.graphicallypresentsthejudgereplacementresultsinthebaselinesample.TheblackdotsatthetopofthegraphdenoteNileshocks,whereasthegraydotsdenoteyearswithatleastonejudgereplacement.Theverticalaxisdetailstheoodlevelinmetersabovesealevel.TheNiletimetrendisplottedinblackandshowsthattheNileoodlevelslowlytrendedupwardovertime.ThisupwardtrendwascausedbysedimentaccumulationthatledtoanincreaseinboththeleveloftheNilebedandthesurroundingland(Popper,pp.241–247),Borsch(2000)).Giventhisfact,IusealineartrendasaproxyfortheoptimalNileoodlevelorthelevelthatmaximizedagriculturalproductionduringtheharvestthefollowingspring.IthensettheNileshockvariableequalto1forNileooddeviationsthatareinthebottomortop5%ofthedeviationdistribution(thesecalculationsweremadeusingtheentiresampletofacilitatecomparisonsbetweensubsamples).ThecutoffsthatdeneNileshocksaremarkedbythetwograytrendlinesinFigureTheshockvariableusedthroughoutthepaperincludesbothapredictable(inastatisticalsense)nonlineartrendand“surprise”changesintheNilelevel. REVOLTONTHENILETABLEIROBABILITYOFEPLACEMENTUMMARYTATISTICS AllShockNonshock (2) PanelA:BaselineProbability22180PanelB:Before1169Probability1989166720 Columns1–3present100timestheprobabilityofjudgereplacementinallyears,yearsofNileshock,andnonshockyears,respectively.Standarddeviationsarepre-sentedbelowtherelevantprobabilitiesinparentheses.AlthoughitisplausiblethatmorepredictableNileshockshaddifferenteffectsthanthosethatweremoreofasurprise(perhapsbecauseelitesand/orpopu-lacecouldbetterprepareformorepredictableshocks),empiricallimitationsmakeitimpossibletoinvestigatethishypothesiswithanyreasonabledegreeofcertainty.However,closeinspectionofFiguresuggeststhatNileshockswereprimarilydrivenbyhigh-frequencyorsurprisevariationintheoodlevel.ThiscanbedemonstratedinamoreformalmannerbydecomposingNileshocksintothepartattributabletodeviationsoftheHodrick–Prescott(HP)trendfromthelineartrendandthepartattributabletodeviationsfromtheHPtrend.ThelattersourceaccountsforoverhalftheNile’sdeviationsfromthelineartrendduringshockyears.InTable,Ipresentsummarystatistics.Inthebaselinesample,theproba-bilityofajudgereplacementis22%inallyears,0%inyearsofNileshocks,and24%innonshockyears.Priorto1169CE,Nileshocksappeartohavehadasmallereffectontheprobabilityofjudgereplacement.JudgeReplacementToformallyestimatetherelationshipbetweenjudgereplacementandNileshocks,Iestimateaspecicationoftheformisadummyequalto1iftheincumbentjudgeatthestartofNilewasnolongerinofceattheendofthatyear,isanindicatorvariableequalto1iftheNileooddeviationisinthetoporbottom5%ofthedeviationdistribution,andisavectorofcontrols.Notethatthisspecicationiscloselylinkedtobinaryresponsemodelsforgroupeddurationdata(e.g., ERICCHANEYSueyoshi(1995)).Toaccountforpossibleserialcorrelation,theerrorstructureisassumedtobeautocorrelatedupto10lagsandheteroscedastic(NeweyandWest(1987)).Forexpositionalease,inthetablesIreport100timesTheestimatedobtainedomittingthevectorofcontrolsispresentedincol-umn1ofTableandprovidesthestandarderrorforthedifferencebetweencolumns3and2ofpanelAinTable.Thesecondandthirdcolumnsshowthattheseresultsdonotqualitativelychangewhendynastyanddecadedum-miesareincluded.IntherowlabelledAR,Iprovidethe-valuefromtheBreusch–Godfreytestwiththenullhypothesisofnoautocorrelationintheer-rortermupto10lags.Incolumns1–3,thistestrejectsthenullhypothesisthattheerrorsarenotautocorrelatedatvaryinglevelsofstatisticalsignicance.TofurtherexaminetheextenttowhichNileshockscausedjudgechanges,incolumn4,Iaddvelagsandleadsoftheshockvariable.Inaddition,Iincludetwolagsofthedependentvariabletoaddressautocorrelationintheerrorterm.Thedatadonotrejectthenullhypothesisthatthecoefcientsonthelagsandleadsare,respectively,jointlyequaltozero.Inaddition,whenthesecontrolsareadded,theBreusch–Godfreytestdoesnotrejectthenullhypothesisofnoautocorrelationintheerrorterm.ThisresultsuggeststhattheeffectsofNileshockswereconcentratedinthe“impact”year.Incolumn5,Iperformthesameexerciseincluding10leadsandlags.Theresultsare,ingeneral,similar,withtheexceptionthatthedatarejectthenullhypothesisthatthelagsarejointlyequaltozero.Thisrobustresultisdrivenbyapositivecorrelationbetweentheeighthshocklagandjudgereplacements.Oneinterpretationofthisresultisthatitindicatesan“unwinding”ofconces-sionsgiventotheheadjudgeduringNileoodsafterthethreatofNile-inducedunresthadsubsided.Alternatively,itispossiblethatthisresultisafalsepos-ReligiousStructuresThehistoricalevidencepresentedintherstsectionsuggeststhattheNile-induceddecreaseintheheadjudge’sreplacementprobabilitymaybeindica-tiveofanincreaseintheimplementationofhispreferredpolicies.Inthissec-tion,IinvestigatetheeffectsofNileshocksonrelativeallocationstoreligiousstructures.Therelevanceofthisexerciseisstressedbystudiesthatsuggestthatsuchallocationsare“themostsignicantindicatoroftherelativeposi-tionsofthereligiousandmilitaryestablishments”(,p.120))andFortheresultspresentedinTable,thecoefcientsonNiledroughtsandoods(i.e.,ab-normallylowandhighoods)aregenerallysimilarinmagnitude.Consequently,intheseregressions,Ireportheteroscedasticity-robuststandarderrors.Forexample,thetheoreticalframeworkinAcemogluandRobinson(2006)suggeststhesovereignwouldhavebeenlikelytorenegeonNile-inducedconcessionsoncethethreatofrevolthadpassed.WhytheheadjudgedidnotnditoptimaltooverthrowthesovereignandestablisharuleofjudgestoaddressthiscommitmentproblemwasdiscussedbyLapidus,p.153). REVOLTONTHENILETABLEIIEPLACEMENTAND DependentVariable:JudgeReplacedononttStandardizedMonuments (2)(4)(6)(7)(8)(9)(10) 33***8788**8621**9549*7032*8181005][009][000][093][072][001][002][000][064][045]p-Value(5leads)leads)020][091]p-Value(5lags))045][066]p-Value(10leads))047][088]p-Value(10lags)lags)000][087]Dynastydummies?NoYesNoYesYesNoYesNoYesYesDecadedummies?NoNoYesNoNoNoNoYesNoNo257257257257257257257257257257 Thedependentvariableincolumns1–5isadummyequalto1iftheincumbentjudgeatstartofNileyearisreplacedinthefollowingyear,whereasthedependentvariableincolumns6–10isastandardizedmeasureoftherelativeallocationofnewconstructionstoreligiousstructuresasexplainedinthetext.Incolumns1–5,Ireport100timestheestimatedcoefcient.Shockisanindicatorvariableequalto1iftheoodresidualisintheupper5%orlower5%oftheooddistribution.Therowlabel-valuefortheBreusch–Godfreytestwiththenullhypothesisofnoautocorrelationupto10lags.Therows-Valueprovidethe-valueforthetestofthenullhypothesisthatthecoefcientsonthestatednumberofleadsandlagsofNileshocksarejointlyequalto0.Standarderrors,assumingtheerrorstructureisautocorrelatedupto10lagsandheteroscedastic,arepresentedinparentheses,asidefromthoseincolumns4,5,9,and10.whicharerobusttoheteroscedasticity.***,**,and*indicatesignicanceatthe1%,5%,and10%levels. ERICCHANEYstemsfromthefactthatresourcesallocatedtoreligiousstructuresbenettedtheheadjudgewhiledetractingfromtheresourcesavailabletothesovereign.IuseCreswell(1919)tocompileadatasetoftheconstructiondatesofmajorsecularandreligiousstructures,andcreateavariablethatmeasuresthetotalnumberofsecularandreligiousstructuresonwhichconstructionbeganineachlunaryear.ThesestructuresincludeCairo’smostimportantreligiousmonu-ments(e.g.,mosquesormadrasas)aswellassecularstructures(e.g.,palacesorcitywalls).ThesovereignandothermilitaryelitesfundedalmostallthestructuresprovidedbyCreswellandthisdatasetlikelyprovidesareasonableapproximationtothetotalnumberofimportantstructuresbuiltoverthepe-riodcoveredbythedata.ToinvestigatetheeffectofNileooddeviationsontherateofconstructionofreligiousstructures,Iconstructameasureofthenumberofreligiousstruc-turesbuiltineachNileyear.Sincetheconstructiondatesaregivenbylunaryear,ItaketheweightedaverageofthetotalnumberofreligiousandsecularbuildingsbeguninthelunaryearsthatoverlappedwithatleastpartofagivenNileyear(wheretheweightsarethepercentageoftheNileyearoccupiedbytherespectivelunaryear).Ithentakethedifferencebetweenthestandardizednumberofreligiousandsecularbuildingsandmultiplythisdifferenceby100.IpresentregressionresultsoftheinvestigationoftheeffectofNileshocksonrelativeallocationstoreligiousstructuresincolumns6–10ofTable.Re-sultsincolumn6omitcontrolsandshowthatallocationstoreligiousstructuresincreasedbyroughly0.88standarddeviationsduringNileshocks.Incolumns7and8,Ishowthatthisresultremainsrobusttotheintroductionofdynastyanddecadedummies.Incolumns9and10,Iincludetwolagsofthedependentvariable,andlagsandleadsofNileshocks.Theseresultssuggestthatincreasesinrelativeallocationstoreligiousstructureswerelimitedtothe“impact”year.POSSIBLECAUSALCHANNELSIntheprevioussection,Ipresentedresultsconsistentwithhistoricalevi-dencethatNileshocksincreasedthepoliticalpoweroftheheadjudge.Inthissection,Iinvestigatetheextenttowhichsomeofthemostplausibleinterpre-tationsoftheresultsareconsistentwiththeavailableempiricalevidence.IncreasesinReligiosityItispossiblethattheempiricalresultsaredrivenbyaNile-inducedincreaseinthepopulace’sreligiosity.Tomeettheincreaseddemandforreligiousser-vices,thesovereignmayhavebeenmorelikelytobothretaintheheadjudgeanddivertresourcestoreligiousstructures.Toinvestigatetheextenttowhichanincreaseinreligiositycanexplaintheresults,Iconstructameasureofreligiosityusingthechronicleli-Ma’arifatDuwalal-MulcomposedbytheEgyptianhistorianal-Maqrizi REVOLTONTHENILE(1364–1442).Inthischronicle,al-Maqrizi—oneofthemosttrustedsourcesfortheperiodcoveredbyhiswritings—providesayearlydescriptionofeventsinEgypt.ThemetricisinthespiritofBaker,Bloom,andDavis(2011)andis100timesthestandardizedmeasureofthenumberoftimestheword“prayer”occursinthedescriptionofeachyear’sevents.Allthemetricsusingthissourceinthesubsequentanalysisareconstructedinasimilarmanner.Ipresenttheestimatesofregression()usingthismetricasthedependentvariableincolumn1ofTable.Theresultsshowthatal-MaqrizimentionsthewordprayersignicantlylessinhischronicleduringyearsofNileshock.OnewouldnotexpectthisresultifdeviantNileoodsledtoimportantin-creasesinreligiosityorinthedemandforreligiousservices,sinceanincreaseinbothindividualandcollectiveprayersseemsalikelymanifestationofin-creasedreligiosity.WhileitisclearlypossiblethatotherNile-inducedevents“crowdedout”referencestoaveragelevelsofreligiosity(andthustheresultsshouldnotnecessarilybeinterpretedasevidenceofaNile-induceddecreaseinreligiosity),presumablyal-Maqriziwouldhaverecordedeventsrelatedtoim-portantincreasesinreligiosity.Thus,thisresultprovidessomeevidencethatNile-inducedincreasesinreligiosityarenotdrivingtheresults.DecreaseinReplacementProbabilitiesDidNilefailuresleadtoadecreaseinthereplacementprobabilitiesofallappointedofcials?Iempiricallyinvestigatethispossibilityintwosteps.First,IexaminetheeffectsofNileshocksonthereplacementoftheheadjudgepriorto1169.IfNile-induceddecreasesinthereplacementprobabili-tiesofallofcialsaredrivingtheresults,IwouldnotexpecttondsystematicdifferencesintheeffectsofNileshocksonjudgereplacementovertime.Toempiricallyinvestigatetheextenttowhichthedataareconsistentwiththisprediction,Ireportresultsobtainedbyrunningregression()ontheentirepre-1425sampleandallowingtheeffectofNileshockstobedifferentbeforeandafter1169incolumn2ofTable.Thenullhypothesisthatthesecoef-cientareequalisrejected,whichisnotconsistentwiththe“uniformdecreaseinprobabilities”hypothesis.Sinceitispossiblethatthisresultisaproductofthefactthatthedataaremeasuredwithgreatererrorpriorto1169ortoothertime-varyingfactors,incolumn3ofTable,Ilimitthesampletoafter1265andreportresultsfromrunningequation()forthefourheadjudgeswhowereappointedinEgyptafterthisdate.ThesejudgesrepresentedthemainSunnilawschoolsandhaddifferentlevelsofadherenceamongEgyptianMuslims.Asabove,iftheuni-formdecreaseinprobabilitieshypothesisweretrue,IwouldexpectNileshockstoresultinsimilarprobabilitydecreasesacrossthefourschools.Resultsincol-umn3showthatthedatarejectthenullhypothesisthatNileshockshadsimilareffectsonthereplacementprobabilitiesofalltheheadjudges.Inparticular, ERICCHANEYTABLEIIIAUSAL PrayerJudgeCrusadeJudgeHighPricesUnrestSultanSultanJudge (2)(4)(5)(6)(7)(8)(9)(10) 20***8525*1 ( REVOLTONTHENILETABLEIII— PrayerJudgeJudgeCrusadeJudgeHighPricesUnrestSultanSultan (1)(2)(4)(5)(6)(7)(8)(9)(10) (HighPrices)rices)Š5680Š1051]p-Valuee001][000]EstimationOLSOLSSUROLSOLSOLSOLSOLSOLS254785160254254254254528528Maq.Maq.Maq.Maq.Ind.Ind. ThedependentvariableinthecolumnsmarkedJudge(Sultan)isadummyequalto1iftheincumbentjudge(sovereign)atthestartofNileyearisreplacedinthefollowingyear.Inthesecolumns,Ireport100timestheestimatedcoefcient.ThecolumnsPrayer,Crusade,HighPrices,andUnrestdenotetheuseofstandardizedmeasuresoftheextenttowhichprayer,Crusaders,highprices,andunrestarementionedinMaqrizi’schronicleasdescribedinthetextandtheSupplementalMaterial.Shockisanindicatorvariableequalto1iftheoodresidualisintheupper5%orlower5%oftheooddistribution.Shock5isanindicatorvariableequalto1iftheoodresidualisintheupper2.5%orlower2.5%oftheooddistribution.TheentriesMalikiShock,HanaShock,andHanbaliShockprovidethecoefcientonthevariableShockinregression()estimatedusingseeminglyunrelatedregression(SUR)whenheadjudgereplacementsfromtheMaliki,Hana,andHanbalischoolsareusedasthedependentvariable(thecoefcientontheShaiheadjudgeisprovidedintherstrowofcolumn3).Therow-Valueprovidesthe-valuecorrespondingtothetestthatalltheprovidedcoefcientsinthecolumnareequal.Intherowsample,Maq.denotestheyearsinthebaselinesampleinwhichthevariablesconstructedusingMaqrizi’schronicleareavailable([1172,1425]),andInd.denotesyearsinboththeearlyandbaselinesamplesinwhichEgyptwasnotpartofalargerempire.Standarderrorsassumingtheerrorstructureisautocorrelatedupto10lagsandheteroscedasticarepresentedinparenthesesasidefromthoseincolumn3,wheretheyarerobusttoheteroscedasticity.Allregressionsincludedecadedummies.***,**,and*indicatesignicanceatthe1%,5%,and10%levels. ERICCHANEYtheeffectofNileshocksisonlystatisticallysignicantforthejudgeswithsize-ablepopularfollowings(theShaiandMaliki)inthisperiod(p.172)).Takeninunison,theseresultscastdoubtontheuniformdecreaseinprob-abilitiesinterpretation,sinceNileshocksdidnotdecreasethereplacementprobabilitiesofalljudges.Moreover,theseresultsprovidesuggestiveevidencethatthedecreaseinreplacementprobabilitywasrootedinajudge’scontroloverpopularsupport,becauseonlyjudgeswithsizeable“constituencies”wit-nesseddeclinesintheirreplacementprobabilities.ExternalWarsItispossiblethatNileshocksincreasedtheprobabilityofconictwithneigh-boringpowers(perhapsbecauseNileshocksinducedsuchpowerstoinvade)andaffectedjudgereplacementsthroughthischannel.ToexaminewhetherNileshocksaffectedtheprobabilityofmilitaryconictwithneighboringpow-ers,IuseMaqrizi’schronicletoconstructametricofthetimeshementionstheword“Crusader”()ineachyear.Presumably,thewordCrusadershouldappearmoreoftenduringperiodsofconictbetweenEgyptandtheCrusaders.SincetheCrusadersweretheforeignpowerthatwasmostlikelytoattack(orbeattackedby)Egyptformuchofthebaselinesample,thefactthatthepointestimatepresentedincolumn4ofTableisnotstatisticallysignicantpro-videssomeevidenceagainsttheimportanceofNile-inducedexternalwarsingeneratingtheresults.Incolumn5ofTable,IshowthatCrusaderreferencesarenegativelycorrelatedwithjudgereplacements.WhilethisresultshouldbetreatedwithcautiongiventhattheactivitiesoftheEgyptianandCrusaderarmiesmaybeendogenous,thesignofthiscoefcientisconsistentwiththehypothesisthatrulersengagedinwarsneededincreasedjudgesupporttohelppreventsocialunrest,tomaintainorder,and/ortomobilizethepopulaceasmilitaryauxil-iaries(e.g.,Lapidus,p.164)).IncreasedPoliticalPowerofReligiousLeadersAretheresultsindicativeofaNile-inducedincreaseinthejudge’spoliticalpowerrootedinhiscontroloverpopularsupport?Intheidealworld,Iwouldobservetheprobabilityofasuccessfulrevolt,bothwithandwithoutjudgesup-portineachperiod.AlthoughthedatanecessarytoconstructthesemetricsdoInterestingly,thisresultisdrivenbyanegativerelationshipbetweenreplacementsandCru-saderreferencespriortothebattleofAcrein1291,whichmarkedthefallofthelastmajorCrusaderstrongholdinthemainlandMiddleEast.ForarelatedresultinearlymodernSpain,seeVidal-Robert(2011).Foratheorythatlinksdemocratizationtomilitaryconictoverthepasttwocenturies,seeTicchiandVindigni(2008) REVOLTONTHENILEnotexist,inthissection,IinvestigatetheextenttowhichNileshocksincreasedthepropensityforsocialunrest.PerhapsthecleanesttestoftheextenttowhichNileshocksincreasedthepropensityforunrestcomesfromameasureofperiodsofhighpricescon-structedusingMaqrizi’schronicle.IusethissourcetoconstructametricthatmeasurestheextenttowhichMaqrizireferstohighfoodpricesineachyear.Estimatesusingthisvariableasthedependentvariableinequation()arepresentedincolumn6ofTableandshowthatreferencestohighpricesin-creasedbyalittleover1standarddeviationduringNileshocks.Thisresultpro-videsdirectevidencethatNileshocksledtoincreasesinthepriceoffoodstuffsandprovidessupportfortheclaimthateliteswereunabletosystematicallyim-plementpoliciestopreventsuchincreases.Giventhelargeliteraturethatlinksincreasesinthepropensityforpopularunresttohighfoodprices,thisresultalsoprovidesindirectevidencethatNileshocksincreasedsocialunrest.Maqrizi’schroniclecanalsobeusedtoprovidedirectevidencethatNileshocksincreasedsocialunrest.IconstructametrictomeasuretheextenttowhichMaqrizireferencescombat,riots,andlootingineachyear.Regressionoutputusingthismetricasthedependentvariableispresentedincolumn7andshowsthatduringNileshocks,Maqriziincreasedreferencestounrestbyapproximately0.9standarddeviations.TofurtherinvestigatetheextenttowhichNileshocksincreasedunrest,Iin-vestigatetheeffectofNilefailuresonrulerchangesforallyearsinwhichEgyptwasnotaprovinceoflargerempirespriorto1425.Althoughthismetricisatbestanimperfectproxyforunrest,sinceincreasesinunrestoftendidnotleadtorulerchanges,theresultsincolumns8and9showapositivecorrelationbe-tweenNileshocksandrulerchanges.Thepointestimateisveryclosetozeroincolumn8,however,andisonlystatisticallysignicantincolumn9whenNileshocksaredenedusingthetopandbottom2.5%ofooddeviations.Thus,thisresultprovidessomeadditionalevidencethatNileshocksincreasedun-rest,althoughonlythemostsevereshocksappeartohaveresultedinrulerFinally,totheextentthatNileshocksonlyaffectedjudgereplacementbyincreasingtheprobabilityhecouldorganizearevolt,NileshocksshouldbeaOneimplicationoftheconceptualframeworkisthatceterisparibusthereshouldbemoreunrestwhenthesovereignreplacestheheadjudgethanwhenhedoesnotduringNileshocks.Unfortunatelythispredictioncannotbetakentothedata,sinceinthebaselinesamplethejudgeisneverreplacedduringshocks.Inaddition,conditioningonjudgereplacement(whichisaffectedbyNileshocks)introducesaformofselectionbias(see,forexample,AngristandPischke(2009)Forevidenceofapositiverelationshipbetweensocialunrestanddemocratizationsoverthepasttwocenturies,seePrzeworski(2009).Forevidencethateconomiccrisesincreasesuchunrestinmodernautocraticenvironments,seePrzeworski,Alvarez,Cheibub,andLimongi(2000)Giventhefactthattherearefewexamplesofjudgesendorsingrebellionsexante,itispossiblethatthisrelationshipisaproductoftheactionsofrivalmilitaryfactions. ERICCHANEYvalidinstrumentforjudgechangesinaregressionofjudgechangesonunrest.Ipresenttheinstrumentalvariable(IV)relationshipbetweenjudgereplace-mentandthepotentialforunrestproxiedbythehighpricesmetricincolumn10ofTable(therststageisgivenincolumn6).Belowthestandarderror,Iprovidea95%condenceintervalfortheIVcoefcientthatisrobusttobothweakinstrumentsandarbitrarywithin-decadecorrelation.TheIVpointesti-mateisnegativeandstatisticallysignicant,suggestingthatastandarddevia-tionincreaseinreferencestohighpricesledtoa30percentagepointdecreaseinthejudgereplacementprobability.CONCLUSIONDuringarguablyexogenousNileshocks,Egypt’shighest-rankingreligiousauthoritywaslesslikelytobereplacedandrelativeallocationsofresourcestoreligiousstructuresincreased.Usinghistoryasaguide,Inotethatthesend-ingsareconsistentwiththehypothesisthatNileshocksincreasedthereligiousleader’spoliticalinuencebyraisingtheprobabilityhecouldcoordinateare-volt.Iprovideresultsconsistentwiththisconceptualframeworkandadditionalevidenceweighingagainstsomeofthemostplausiblealternativeexplanationsfortheseempiricalpatterns.MyoverallinterpretationoftheresultsiscloselyrelatedtothetheoreticalframeworkdevelopedinAcemogluandRobinson),whoempha-sizedthateconomiccrisesincreasetheprobabilityofthecollapseofautocraticregimesbytemporarilyalteringthebalanceofpoliticalpower.Ontheonehand,theresultsinthispapersuggesttherelevanceofthisliteratureinthepre-modernerabyprovidingevidencethatNileshocksalteredthebalanceofpoliticalpowerinfavorofEgypt’shighest-rankingreligiousauthority.Ontheotherhand,noneofthedocumentedshocksledtothecollapseofpre-modernEgypt’sautocraticarrangements.AtleastsincetheEnlightenment,scholarshavearguedthatreligiouslead-ershistoricallyusedtheirpopularinuencetohelpperpetuateauthoritarianinstitutions.Althoughtheextenttowhichreligiousleadershavehistoricallyworkedtostiedemocraticchangeremainsanareaforfutureresearch,itisinterestingthattheEnlightenmentthinkerscreditedwithlayingtheideologi-calunderpinningsofmoderndemocraticinstitutions(AcemogluandRobinson,p.69),Israel(2011))believedthatdemocracywas“impossible”with-outrst“dismantlingpriestlypower”(,p.103)).Theseandsubse-quentscholarsmaintainedthatreligiousleadershistoricallyusedtheirpopularinuencetomitigatepressuresforinstitutionalchange.Consistentwiththisview,thispaperprovidesevidencethatthepoliticalpowerofEgypt’shighest-rankingreligiousleaderincreasedduringperiodsofeconomicdownturnwhenThiscondenceintervalisderivedfromtheAnderson–Rubin(AR)teststatisticasdescribedFinlayandMagnusson(2009) REVOLTONTHENILEthepressuresforinstitutionalchangesarethoughttohavebeenparticularlyCEMOGLU,D.,J.A.R(2001):“ATheoryofPoliticalTransitions,”AmericanEco-nomicReview,91(4),938–963. EconomicOriginsofDictatorshipandDemocracy.Cambridge:CambridgeUni-versityPress.[CEMOGLU,D.,S.J,J.A.RP.Y(2005):“IncomeandDemocracy,”WorkingPaper11205,NBER.[AYYID,A.(1972):“TheUlamaofCairointheEighteenthandNineteenthCen-turies,”inScholars,Saints,andSus,ed.byN.Keddie.Berkeley:UniversityofCaliforniaPress. (1973):“ThePoliticalandEconomicFunctionsofthe‘JournaloftheEconomicandSocialHistoryoftheOrient,16(2/3),130–154.130–154.2036]AMIN,M.(1980):Al-Awqafwaal-Hatal-Ijtima’iyafMis.Cairo:Daral-Nahdaal-‘Arabiyya.Arabiyya.2038]ANGRIST,J.,J.-S.PMostlyHarmlessEconometrics,AnEmpiricist’sCompan-.Princeton:PrincetonUniversityPress.[,S.R.,N.BS.J.DAVIS(2011):“MeasuringEconomicPolicyUncertainty,”Report.[,R.(1999):“DeterminantsofDemocracy,”JournalofPoliticalEconomy,107(56),(56),2035]BERGER,H.,M.S(2001):“EconomicCrisesandtheEuropeanRevolutionsofTheJournalofEconomicHistory,61(2),293–326.293–326.2034]BLAYDES,L.,E.C(2013):“TheFeudalRevolutionandEurope’sRise:PoliticalDi-vergenceoftheChristianWestandtheMuslimWorldBefore1500CE,”AmericanPoliticalScienceReview,107(1),16–34.16–34.2035]BORSCH,S.(2000):“NileFloodsandtheIrrigationSysteminFifteenth-CenturyEgypt,”StudiesReview,4,131–145.131–145.2040]BRÜCKNER,M.,A.C(2011):“RainandtheDemocraticWindowofOpportunity,”,79(3),923–947.923–947.2033,2034]BURKE,P.,A.L(2010):“DoOutputContractionsTriggerDemocraticChange?”icanEconomicJournal:Macroneconomics,2,124–157.124–157.2033,2034]CHAMBERLAIN,M.(1998):“TheCrusaderEraandtheAyyubidDynasty,”inTheCambridgeHistoryofEgypt,Vol.1.NewYork:CambridgeUniversityPress,211–241.[,E.(2012):“DemocraticChangeintheArabWorld,PastandPresent,”BrookingsPapersonEconomicActivity,42(1),363–414. (2013):“Supplementto‘RevoltontheNile:EconomicShocks,Religion,andPoliticalPower’,”EconometricaSupplementalMaterial,81,81,2033]CRECELIUS,D.(1972):“NonideologicalResponsesoftheEgyptianUlamatoModernization,”inScholars,Saints,andSus,ed.byN.Keddie.Berkeley:UniversityofCaliforniaPress,167–210.167–210.2036]CRESWELL,K.(1919):“ABriefChronologyoftheMuhammadanMonumentsofEgypttoA.D.Bulletindel’InstitutFrançaisd’ArchéologieOrientale,16,39–164.39–164.2044]FINLAY,K.,L.M.MAGNUSSON(2009):“ImplementingWeak-InstrumentRobustTestsforaGeneralClassofInstrumental-VariablesModels,”TheStataJournal,9(3),398–421.398–421.2050]GEDDES,B.(1999):“WhatDoWeKnowAboutDemocratizationsAfterTwentyYears,”ReviewofPoliticalScience,2,115–144. ERICCHANEYAGGARD,S.,R.R.KAUFMAN(1997):“ThePoliticalEconomyofDemocraticTransitions,”ComparativePolitics,29(3),263–283.263–283.2037]HALLAQ,W.(2009):AnIntroductiontoIslamicLaw.NewYork:CambridgeUniversityPress.ress.2039]HUNTINGTON,S.(1991):TheThirdWave:DemocratizationintheLateTwentiethCentury.Norman:UniversityofOklahomaPress.[BNALIJAZI,S.(1470):“Nailal-Ra’idminal-Nlal-Za’id,”Manuscript2261,BibliothèqueNationaledeFrance,ManuscriptsOrientaux.[AJAR,S.(1449[1998]):Raf’ual-Isr‘anQudatMis.Cairo:Maktabaal-Kh.[RWIN,R.(1986):TheMiddleEastintheMiddleAges:TheEarlyMamlukSultanate.Carbondale:SouthernIllinoisUniversityPress.[SRAEL,J.(2006):EnlightenmentContested.NewYork:OxfordUniversityPress.[ DemocraticEnlightenment:Philosophy,Revolution,andHumanRights1750–.NewYork:OxfordUniversityPress.[,I.(1972):“AyyubidReligiousPolicyandtheDevelopmentoftheSchoolsofLawinCairo,”inColloqueInternationalsurl’HistoireduCaire,27Mars–5Avril1969.Cairo:MinistryofCultureoftheArabRepublicofEgypt,GeneralEgyptianBookOrganization,279–286. MuslimCitiesintheLaterMiddleAges.NewYork:CambridgeUniversityPress. AHistoryofIslamicSocieties.NewYork:CambridgeUniversityPress.[,Y.(1991):StateandSocietyinFatimidEgypt.Leiden:E.J.Brill.[,S.M.(1994):“TheSocialRequisitesofDemocracyRevisited,”AmericanSociologicalReview,59(1),1–22.1–22.2035]MAKDISI,G.(1962):“Ash’arandtheAsh’aritesinIslamicReligiousHistory,”StudiaIslamica17,37–80.37–80.2035]MAQRIZI,T.(1441[1997]):ukli-Ma’arifatDuwalal-Mul.Beirut:Daral-Kutubal-‘Ilmiyya.[,K.(1844[1982]):“ContributiontotheCritiqueofHegel’sPhilosophyofRight,”inReligion.Chico:ScholarsPress.[,W.,K.W(1987):“ASimple,PositiveSemi-Denite,HeteroskedasticityandAutocorrelationConsistentCovarianceMatrix,”,55(3),703–708.703–708.2042]NIELSEN,J.(1984):“Sultanal-ZahirBaybarsandtheAppointmentofFourChiefQStudiaIslamica,60,167–176. SecularJusticeinanIslamicState.Leiden:NederlandsInstituutvoorhetNabijeOosten.[ORTH,D.,J.WB.WEINGASTViolenceandSocialOrders.NewYork:Cam-bridgeUniversityPress.[ETRY,C.(1981):TheCivilianEliteofCairointheLaterMiddleAges.Princeton:PrincetonUni-versityPress.[,W.(1951):TheCairoNilometer.Berkeley:UniversityofCaliforniaPress.[,A.(2009):“ConqueredorGranted?AHistoryofSuffrageExtensions,”JournalofPoliticalScience,39(2),291–321.291–321.2049]PRZEWORSKI,A.,M.ALVAREZ,J.CF.LDemocracyandDevelop-.NewYork:CambridgeUniversityPress.[,B.(1980):“GrainRiotsandthe‘MoralEconomy’:Cairo,1350–1517,”JournalofIn-terdisciplinaryHistory,10(3),459–478.459–478.2037]SMITH,A.(1776[2009]):TheWealthofNations:BooksIV–V.NewYork:ClassicHouseBooks.Books.2033]SUEYOSHI,G.(1995):“AClassofBinaryResponseModelsforGroupedDurationData,”ofAppliedEconometrics,10,411–431.411–431.2042]TAGHRI,A.(1468[1976]):umal-ZahirafMulukMisrwaal-Qahira,Vols.1–7,tr.W.Popper.NewYork:AMSPress.[ REVOLTONTHENILE (1468[1990]):adithal-DuhurfMadaal-Ayyamwaal-Shuh.Cairo:Wizaraal-Awqaf.[,D.,A.V(2008):“WarandEndogenousDemocracy,”Report.[IDALOBERT,J.(2011):“AnEconomicAnalysisoftheSpanishInquisition’sMotivationsandConsequences,”Report.[Dept.ofEconomics,HarvardUniversity,LittauerCenter,Cambridge,MA02138,U.S.A.;ManuscriptreceivedAugust,2011;nalrevisionreceivedMarch,2013.