DepartmentofPoliticsandWoodrowWilsonSchoolofPublicandInternationalA airsPrincetonUniversityEmail cboixprincetonedu yDepartmentofPoliticalScienceUniversityofIllinoisatUrbanaChampaignEmail ms ID: 201419
Download Pdf The PPT/PDF document "Non-tyrannicalAutocraciesCarlesBoixandM..." 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.
Non-tyrannicalAutocraciesCarlesBoixandMilanSvolikApril2007AbstractWhydosomedictatorshipsestablishinstitutionstypicallyassociatedwithdemoc-racy,suchaselectionsandlegislatures?Weproposeanewtheoreticalmodelofnon-tyrannicalorelectoralautocracies.Electoralautocraciesarepolitiesgovernedbyanunelectedexecutivethatsharespowerwithanelectedlegislature.Weexplainwhylegislaturesandelectionsenhancethecredibilityanddurabilityofpower-sharinginautocracies.Ourtheorygeneratesasetofnewpredictionsabouttheempiricalrela-tionshipbetweenpoliticalinstitutions,economicdevelopment,andregimesurvivalindictatorships. DepartmentofPoliticsandWoodrowWilsonSchoolofPublicandInternationalA airs,PrincetonUniversity.Email: cboix@princeton.edu .yDepartmentofPoliticalScience,UniversityofIllinoisatUrbana-Champaign.Email: msvolik@uiuc.edu . 1IntroductionTyranny,thatistheunconstrainedruleofapolitybyoneperson,haslongbeentreatedasthestandard,almoststereotypicaltypeofdictatorship.Theclassicalliteratureondictator-shipsmainlyinvestigatedpersonalautocraciesandthemechanismsemployedbydictatorstogovernandsecuretheacquiescenceofhissubjects( Xenophon/Strauss1961 ; Machiavelli1513/1985 ).Thepostwarliteratureondictatorshipsfocusedonthephenomenonoftotali-tarianismandonthemeansthroughwhichthetotalitarianleaderandhispartyexercisedabsolutecontroloversociety( Arendt1973 ; FriedrichandBrzezinski1965 ; Linz1975 , 2000 ; Neumann1957 ). 1 Theformalliteraturethathasreplacedthatdescriptivebodyofworkdidnotabandonthebasicpointofdepartureofthetraditionalresearchonautocracies.Dicta-torshipscontinuetobemodeledaspoliticalregimesinwhichasingletyrantgovernsalonethroughtheuseofcrediblethreats(andsometimesthedistributionofselectivebene ts),exploitstheinabilityofhissubjectstocoordinateagainsthim,andisnotsubjecttoanyexternalconstraintorin\ruence( Haber2007 ; Kuran1991 ; Tullock1987 ; Wintrobe1998 ).Yet,foralltheirhistoricalandtheoreticalimportance,single-rulerautocraciesconstituteaminorityoftheuniverseofauthoritarianregimes.SinceWorldWarII,lessthanafourthofallauthoritarianregimes(andhenceonlyaboutatenthofallcurrentpoliticalregimes)havebeengovernedbyasingleruler.Furthermore,afterpeakingat43percentinthemid-1970s,theproportionofsingle-rulerautocraciesamongallauthoritarianregimeshasdeclinedtolessthan20percentin1999.Incontrast,roughlythree-fourthsofallauthoritarianregimesinthelastsixtyyearshavehadanelectedlegislature,and,withit,aninstitutionalizedbodythatmayservetobalanceagainstthedictatoraswellastosustainlawsandregularizedpatternsofpoliticalinteractionwhichmayrestrian,inturn,thetyrannicaltendenciesofany 1 Linz ( 1975 , 2000 )isapartialexceptioninthathealsoexaminesnon-totalitarianregimes.1 singleruler.However,theoreticalresearchonnon-tyrannicaldictatorshipshasbeenscant.Animpor-tantexceptionisLinz'sanalysisofauthoritarianregimes( Linz1975 , 2000 ),buthisworkremainsmostlytypologicalinnature.Morerecently,severalscholarshaveexaminedauthor-itarianregimesinwhichatleastonepoliticalinstitution,thelegislature,ischosenthroughsomekindofelectoralprocess( Brownlee2004 ; GandhiandPrzeworski2006 ; LevitskyandWay2003 ; Lust-Okar2006 ; Magaloni2006 ).Thisbodyofworkisalreadymakingimpor-tantcontributionstothestudyofdictatorships.Yetthisliteratureremainsincompleteintwoimportantways.First,mostofitfocusesnarrowlyontheanalysisofparticularcases.Second,themoretheoreticalworkpaystoomuchattentiontotheelectoralgamebetweenvotersandautocraticelitesandfailstomodelthestructure,incentivesandbehavioroftheauthoritarianeliteitself.Theresultisafragmentedexplanationofwhymanyauthoritar-ianregimeshaveinstitutionsendowedwithatleastsomeindependentpowersandthereforememberswhocanautonomouslycoordinatetheiractionsvis-a-visthedictator.Inthispaper,wedevelopamorecompletetheoryofnon-tyrannicalautocracies.Wede ne\non-tyrannicalautocracies"asauthoritarianregimesthatpresenttwocharacteristics.First,thereisadictator,thatis,anunelectedexecutivewithlegislativepowers.Second,thereisalegislaturethatisselectedindependentlyfromthedictator. 2 Withinnon-tyrannicalautocracies,wefurtherde nethemorespeci ctypeof\electoralautocracies",inwhichthelegislatureiselectedthroughgeneralelections,eventhoughthesemaynotbeconductedaccordingtofreeandfairprocedures.Letusclarifythisde nitionbycomparingelectoralautocracieswithothernon-democraticregimes.Inpuretyrannies,thedictatorshareshisauthoritywithnoone.Hisauthorityis 2Thusthelegislatureisnotamereadvisorybodydirectlyappointedbythedictator2 thereforebasedonthesubordinationofallsubjects.Withintheratherlooselimitsimposedbythestructureofsupporthehas,thetyrantiscompletelysovereign.Bycontrast,non-tyrannicalautocracies(themostcommonofwhichareelectoralautocracies)arecharacterizedbytheruleofanauthoritarianelite,thatis,byseveralindividualssharingpowerwitheachother.Conceptually,then,non-tyrannicalautocraciesaredictatorshipscharacterizedbyatwo-levelgame.Ontheonehand,thereisagameplayedwithintheauthoritarianeliteitself.Thedictatorandothermembersoftheeliteneedtopreservea\deal"togovernjointlyaccordingtosomepower-sharingagreement.Preservingapower-sharingdealamongtheeliteguardsagainstthepossibilitythatoneofthemwouldgainabsolutepreeminenceovertherest,henceestablishingatyranny.Ontheotherhand,weobserveagameinwhichthisautocraticeliteexcludestherestofthepopulationfrompower(andwherethegameisplayedinsuchawaythatreinforcessomeoftheconditionsofthewithin{elitebalanceofpower).Ourtheoreticalargumentstartswiththerecognitionthatinordertodeterthreatsagainsthisrule,adictatormayseekthesupportofeliteornotableswithlocalin\ruence.Inreturnfortheirhelp,thedictatormaypromisethenotablesa(greater)shareofthenationalbudgetorotherpoliticalbene ts.Inturn,thenotablesmaydemandaguaranteethatthepromisedbene tswilllastbeyondtheemergencythatcompelledthedictatortoseektheirsupportinthe rstplace.However,resortingtothird-partyenforcementofanypower-sharingar-rangementbetweenthedictatorandtheenlistednotablesisnotpossible,sinceitwouldimplysurrenderingtheverypowersthatbothwishpreserve.Hence,thecentraldilemmaofanynon-tyrannicalautocracyistoestablishtheappropriatemechanismsthatwillallowthedictatortogovernjointlywiththenotables,whileminimizingthepossibilitythateither3 partywilldeviatefromsuchapower-sharingarrangement. 3 Thetwokeyinstitutionalfea-turesofelectoralautocracies,alegislativebodyandpublicelectionsofthatlegislature,aretheinstrumentsdevelopedtosolvethisdilemmaofautocraticpower-sharingbetweenthedictatorandtheelites.Howexactlydothesetwoinstitutionssolvethedilemmaofautocraticpower-sharing?Inthe rstplace,alegislatureprovidesthenotablesaforumwithinwhichtheymayshareinformationandallowsthemtomonitorwhetherthedictatorhonorstheiragreementtosharepower.Italsofacilitatestheirabilitytocoordinateonjointlywithdrawingsupportfromthedictatorshouldherenegeonhispromisetosharepowerwithanyofthem,therebyleavingthedictatorvulnerabletoanythreatstohispositioninpower.Thusinadditiontoitspuremonitoringfunction,thelegislaturealsoperformsanenforcementfunction.Intuitively,anyonenotableistooweaktocompelthedictatortocomplywithapower-sharingarrangement.However,asagroup,notablescanthreatentosupportsomechallengetothedictator'spowerifherenegesonthepower-sharingarrangementinplace.Putsimply,alegislatureisaninstitutionalizedforumwhichreducesnotables'coordinationcostsandtherebymakestheautocraticdealmorestable.Ontheotherhand,wearguethatelectionsallowthedictatortoensurethathein-cludesonlyimportantorin\ruentialnotablesinthepower-sharingarrangement.Speci cally,electionsallowthedictatortomonitorwhetherthelocalin\ruenceofindividualnotablesissucienttowarranttheircontinuinginclusioninthepower-sharingarrangement.Electionsserveasanimperfectyetpublicsignalofwhetherthein\ruenceofindividualnotablesispreserved,eveniftheyarepatronage-riddenorpartiallyfraudulent.Beforethecontemporaryperiodthevastmajorityoflegislaturesoperatingunderauthor- 3Startingwithasimilarpoliticaldilemma, Myerson ( 2006 )investigatesthefoundationsofpoliticallead-ershipsandconstitutions.4 itarian(monarchical)regimeswerenotselectedinelections.Instead,theyweresimplyagatheringofpowerfulnobleswhohadaninterestincheckingthepowersofthemonarchs. NorthandWeingast ( 1989 ),forinstance,explainhowtheGloriousRevolutionof1688inwhichnoblesturnedagainsttheexploitativeStuartsledtothestrengtheningofthePar-liamentthatprotectedwealthyelitesagainstarbitraryexpropriationbytheCrown.Theuseofelections(withbroadelectorates)wasminimalduetotheeconomicstructureofthepre-modernera.Wealthwasmainlybasedonlandand,asaresult,landholdingswereaverygoodindicatorofnotables'in\ruence.Landholdingshavebecomelessinformativeovertime.Moderneconomicdevelopmentimpliesthatanincreasingfractionofnotables'politicalin\ruencederivesfromwealthaccu-mulatedthroughtradeandservices.Tradeandservicesrelylargelyonastockofcontractualrelationshipsthatcannotbemeasuredbyaneasilyobservablefactor,suchaslandholdings.Hencemodernautocracieshavesolvedthismonitoringproblemviatheinstitutionofelec-tions.Power-sharingatthenationallevelendowseachnotablewithpatronageresourcesthathecanusetocultivatehislocalpoliticalin\ruenceandtheresultsofelectionsareanimperfectbutpublicsignalofthatin\ruence.Althoughourmodelisbuilttoexplainnon-tyrannicaldictatorships,itcanbeemployedtoexaminethecircumstancesunderwhichanelectoralautocracywillnotbeafeasiblepower-sharingarrangement.Recallthattheoriginalreasonwhythedictatorconsiderssharingpowerwiththenotablesisthepresenceoftemporarythreatstohisrule.We ndthatwhensuchthreatstothedictatorareinsigni cant,hedoesnothavesucientincentivestosharepowerbeyondthemomentwhenthosethreatsareimminent.Thenotablesunderstandthisandwillthereforenotsupportthedictatorwhenthreatstohispowerarenotsigni cant.Interestingly,thedictatoralsolackstheappropriateincentivestosharepowerwithnotables5 whenthreatstohisruleareoverwhelming.Inthatcase,hewillnotevenbotherwithenlistingnotablesthatmaysupporthimandinsteadgovernsalone\whileitlasts."Ourtheorythereforegeneratesasetofnewpredictionsabouttheempiricalrelationshipbetweenpoliticalinstitutions,economicdevelopmentandregimesurvivalindictatorships.First,theaboveargumentsuggeststhattyranniesmaybeeithershort-livedorlong-lived.Thiswillbethecasewhenthreatstothedictatorareoverwhelmingorinsigni cant,re-spectively.Ontheotherhand,electoralautocraciesoccupythemiddlegroundintermsofregimedurability.Finally,whilehistoricallylegislaturesappeartobeakeyinstitutionalfeatureofe ectivepower-sharingindictatorships,ourtheorysuggeststhatelectionswillbeanotherfeatureofsuchinstitutionalizedarrangementsonlywhenasigni cantfractionofeconomicproductionderivesfromtradeandservices.Thusweexpectthenumberofelectoralautocraciestoincreasewitheconomicdevelopment.Thepaperisorganizedasfollows.InSection2wepresentsomedataonthedistributionoftypesofdictatorshipsovertimeandbrie\rydiscussthecurrenttheoriesofelectoralorvotingautocracies.InSections3to5wedevelopourtheoryofelectoralautocraciesintwosteps.Aftercharacterizingthenatureofthe\autocraticdeal,"wedescribewhicharetheinstitutionalmechanisms(mainly,alegislaturecomposedof\independent"notables)employedtosolvethedilemmasembeddedinthatdeal.Wearetheninapositiontoexplorewhatstrategiesareemployedbytheautocraticelite(andtheparticularinstitutionalsolutionithaschosen)tosustaintheircontrolovertherestofthepopulation.Section6concludes.(Inworkinprogressnotincludedinthispaperweconsidertheoriginsofdi erenttypesofdictatorshipsaswellastheirimpactonthepoliticaleconomyoftheircountries.)6 2DescriptiveDataInthedataweexamine,weemploythefollowingde nitions.Adictatorshipisanyregimewhereoneofthetwofollowingconditionsisnotmet:freeandcompetitiveelectionsandanexecutiveaccountabletocitizens(eitherthroughelectionsinpresidentialsystemsortothelegislativepowerinparliamentaryregimes). 4 Atyrannyisadictatorshipwithoutalegislature(orwithalegislativecouncilappointedbytheexecutive).Wecallanon-tyrannicalautocracyor,forthesakeofbrevity,anautocracyanydictatorshipthathasalegislature(withexclusionofadvisoryorlegislativebodiesappointedbythedictator).Figure 1 and 2 showthenumberandtheproportionoftyranniesandautocraciesintheworldfrom1951to1999.Weemploytwodatasetstodeterminethedistributionoftypesofdictatorshipsrespectively.Weusethedatasetdevelopedby Przeworskietal. ( 2000 )totracktheevolutionofdictatorshipswithandwithoutlegislaturesfrom1951to1990.WerelyonKeefer's\DatabaseofPoliticalInstitutions"( Keefer2002 )toconstructthesametwoseriesfrom1975to1999.Fortheperiodinwhichwehaveoverlappingdatafromthetwosources,thatis,1975to1990,thereisamismatchbetweenPrzeworski'sandKeefer'sdatasets,particularlyforthenumberoftyrannies.Almostallofthedi erencehastodo,however,withthefactthatthePrzeworskidatasetdoesnotreportoil-exportingcountries.Figure 1 and 2 abouthere.Thetotalnumberofdictatorshipsgrewsteadilyinthepostwarperiodfromabout40toapeakof108in1978.Thenumberthendeclinedtoabout90bytheturnofthecentury.Thenumberoftyranniesroseuntilthemid1970s:in1976therewere47dictatorshipswithoutlegislativebodies.Autocraciesalsospreadovertime.Theymultipliedsharplyintheearly1960sfollowingtheprocessofcolonization.Theirnumberstayedputforabout fteenyears 4Thede nitionandthecodingistakenfrom BoixandRosato ( 2001 ).7 andthengrewagainafterthelate1970s.Bytheearly1980therewerearoundseventyauthoritarianregimesthathadlegislatures.AsshowninFigure 2 ,inspiteofthegrowthintheoverallnumberofdictatorships,theproportionbetweentyranniesandautocracieshasbeenconsiderablystablesincetheearly1950s.Between70and80percentofallauthor-itarianregimeshavehadanelectedlegislatureinthelast50years.Duringtheseventiesthisproportionfellexceptionallytolessthan60percent,followingadramaticexplosionoftyrannicalrulers.Table 1 abouthere.Table 1 reportsthenumberoftransitionstodictatorialregimesatthetimeofindepen-denceandfromalreadysovereigndemocracies.Italsocomputesthenumberoftransitionsfromtyranniestoautocraciesandviceversa.Mostregimetransitionsoccurredbefore1975.Thirty-onecountriesbecameautocraticatthetimeofindependence.Anotherthreefellintoatyrannicalsystemastheybecamesovereign.Inaddition,wecounted19democraciesshiftingtotyrannyand5toautocracy.Similarly,theturnoverwithindictatorshipswassubstantial:26tyranniesintroducedlegislaturesand37autocracieslostthem.After1975,however,changewasmoresubdued.Onlyahandfulofauthoritarianregimesresultedatthetimeofindependenceandmostoftheregimetransitionsconsistedindemocraticbreakdownsleadingtoelectoralautocracies(15cases).Figure 3 and 4 abouthere.TakingadvantageofKeefer's ne-grainedinformationonthemechanismsemployedtoselectthelegislature,Figure 3 and 4 displaythenumberandproportionoftyrannies(thosedictatorshipswitheithernolegislatureorwithunelectedlegislatures),\hard"autocracies(authoritarianregimeswithelectedlegislaturesinwhichonlyonepartyhasgainedanyseats)and\soft"autocracies(authoritarianregimeswithelectedlegislaturesinwhichatleasttwo8 partyhaveseats).Asnotedbefore,thenumberandproportionoftyrannicalregimes,whichpeakedinthemid1970s,hasexperiencedadeclineto guressimilartothoseatthebeginningofthepostwarperiod.Withinautocracies,however,therewasanotablechangeinthenatureoftheirlegislatures.Untilthelate1980slessthanathirdofalllegislatureshadmorethanoneparty.Bythemid1990sthisproportionhadjustreversed.Table 2 showsthatthegrowthinthenumberofsoftautocraciescamefromtwomainsources: rst,thecollapseofdemocraticregimes(15events)andsecond,athawingofhardautocracies(30cases,countered,however,by18episodesoftransitionsfromsofttohardautocracies).Table 2 abouthere.3AutocraciesandtheAutocratic\Deal"Aspointedoutabove,whereasinatyrannytheruler(asingleindividual)governsinanunconstrained,absolutemanner,allautocracies,thatis,allauthoritarianregimesgovernedbyseveralindividualssharinginpowerwitheachother,arebasedona(sometimesexplicityetmoreoften)tacit'deal'amongthemembersoftheautocraticorrulingelitebothtogovernjointlyaccordingtosomepredeterminedruleandtorespectsucha'power-sharing'agreement.Innon-tyrannicalautocracies(whicharethemostpredominantvarietyofautocraciesintoday'sworld),theautocraticdealorpower-sharingagreementtakestheformofastructureinwhichthereisanationalleader(adictator),governinginapreeminentorcommandingposition,surroundedbyacertainnumberof(local)notables,whoparticipate(sometimesindirectly)inanationallegislatureelectedthoughgeneralyetnotcompetitiveelections.Thenationalleadercontrolsthecentralgovernmentand,withit,theultimateinstrumentsofstatepower:thearmy,theinternalpolice,ataxcollectionsystem(eitherindependent9 oftheclassofnotablesormerelyresultingfromthecashin\rowsorservicespaidbyhisnotables).Thenotablesaccepttheauthorityoftheleaderorpresident,defertohisdecisionsandsupporthim nanciallyandevenmilitarily.Still,thenotablescontrol,eachoneofthem,somepartofthenationalterritoryoragivenportionofeconomicresourcesinastablemannerandthereforeretainsomeindependencefromthenationalleader.Inautocraciesthesourcesofauthorityoftheleaderandthenotablesareindependent.Thenotablesdonotelecttheleader(incontrasttowhatmayhappeninmorehorizontaloligarchies).Theleaderreliesforhissupport(and\legitimacy")onsourcesotherthanthedirectopinionandconsentoftheclassofnotables.Sometimeshehasearnedhispositionbymeansofacoup(a'soft'coupsincehehasnottransgressedcompletelythejurisdictionofthenotables).Alternatively,hehasinheritedhispositionfromaparentorrelative.Orhehasbeenappointedbythepreviousleader(\dedocracia").Theseproceduresmaketheleaderunaccountabletoorindependentfromthelocalnotables.Inturn,thedictatorornationalleaderdoesnotchoosethenotables(oratleastdoesnotchoosethemarbitrarily).Thenotableshavesomeindependentauthorityorjurisdic-tionalpowers,recognizedbythenationalleader,overtheirrespectivelocalarea(andthecorrespondingpopulation).Thenatureoftheirauthorityvarieswiththeformsofdomina-tionthenotablesemploy(whichinturndependonthetypeofeconomyandprobablythekindofstateinplace).Inmostnon-tyrannicalautocracies(basically`electoralautocracies'),theprovincialnotablesarepatronswhomanageawell-oilednetworkofclientsthroughtheprovisionofprivategoodsandbylinkingthemtothestateandthejobsandservicesthestatemachineryo ers.Therelativeindependenceoflocalnotablesdi erentiatesanelectoralautocracyfromatyranny.Inatyrannythedictatorgovernsunchecked.Accountabletonoone,heappointshisservants,rotatesthemandeliminatesthematwill.Hisonlyconcern10 istopreserveintactthegeneralizedfearandthecoordinationequilibriumthatallowhimtocommand.Bycontrast,inanautocracytheleaderrespectstheprovincialorsectorialpowerofeachnotable{althoughtheextentofthisrespectvaries,ofcourse,withthevariantofelectoralautocracy.Giventhestructureofpowerinelectoralautocracies,theseregimes(asanyotherau-thoritarianregimeunderthecontrolofanauthoritarianclique)shouldbebedeviledbyextraordinarypoliticaltensions.Toobtainhigherrents,tolowertheprobabilityofrebel-lionsamongambitiousnotablesortobuildupamorecohesivecountrythatcanbethenemploytoexpandabroad,thedictatorwillstrategizetogetridoforatleastsubjectwithamuch rmergrasptheclassofnotables.Theleadermaytrytocheck,imprisonorkillallthenotablesatthesametimeandthenproclaimhimselfabsoluterulerormonarch.Alterna-tively,hemayusemoregradualtactics.Hemaygetridofafewnotablesatatimewithouttheresttakingnotice,proceedtoabsorbthefortunesandpowerofthosehejusteliminatedandusehisgrowingpowertofurtherpurgemorenotables.Inthathemaybeaidedbyafactionofloyalsupporters(withinthegeneralclassofnotables)whichwouldthengivehimtheextraresourcesneededtooverwhelmeveryoneelse.Naturally,afewsequentialroundsofsuchtypeof'salami'tacticswilltransformhimintoatyrant.Inturn,oneorseverallocalnotablesmayaswelltrytoupsettheexistingbalanceofpower.Theymayattempttoaccumulateresources,organizeparticularnetworksofsupportersorreshapetheclassofnotablesitselftoconstrainthedictatororto rehim. 5 5Inthefaceofapotentiallyoverpoweringleader,capableofexploitinghispreeminentpositiontobecomeatyrantorabsoluteruler,manynotablesmaylong(andaccordinglyact)tobecomefullyindependent.Suchanoptionmaynotbeavailable,however,forthreereasons.Inthe rstplace,notablesneedadictatortoruleovereverynotable,toadjudicateandrefereeintheirdisputesandtoensureastateofpeace.Withoutanationalleader,countriesbecomeengulfedincivilwarsamongterritorialnotables.Ortheysplitintounitsthataretoosmalltodefendthemselvessuccessfully.Thisleadstothesecondmotivetoenterandsustainanautocracy.Itcongealsastableallianceofminorpowerstocopewithanyexternalenemies.Finally,thelocalnotablesmaysometimesneedprotectionagainsttheirownpopulace.Again,thedictatorisinstrumental11 Giventhosesourcesofcon\rictandstability,thecentraldilemmaorchallengeofanyautocracywillalsoconsistinestablishingtheappropriatemechanismstosustainanauto-craticdeal.Althoughtheautocraticcliquehastoorganizethestateintheproperwaytoexcludetherestofthepopulationfromgovernment(and,asweexaminelater,thisgoalisaccomplishedinelectoralautocracythroughaparticularsystemoforganizingelections),thegoverningclasshastomakesure,aboveall,thatthedealtogovernjointlyremainsunchanged,thatis,thatnopartyamongthemselveswillgaintheabilitytoalteritinitsfavor.Inotherwords,intheprocessofruling,theautocraticelitehastomakesurethatnoonewithintherulingclasswilleithercreatefactionsorinternalnetworks,useexogenousresourcesorstirthesupportofthosethatareexcludedfromgovernmenttochallengethestatusquo,thatis,toshifttheinternalbalanceofpower(oftheautocraticelite)andtorecon gurethecompositionofthegoverningelite.Totheextentthattheeliteissuccessfulinpreventinganyonefromrenegingfromtheautocraticdeal,thenon-tyrannicalautocracywillendure.Otherwise,theelectoralautocracywillcollapseandgivewaytoadi erentpoliticalarrangement.Now,theautocraticdealcannotbesimplygroundedonaverbalpactamongautocrats{anorallygivenpromisetorespectthepositionorstatusofeveryoneandtoconsulteveryoneinformallytodecideoveranyissue.Itcannotbebasedeitheronamerewrittenagreement-intheformofacontractacceptedandsignedbyeveryone.Asinanypact,theautocraticdealneedstorelyonsomeexternalguarantees,thatis,onsomeguaranteesthatgobeyondthestrictpromisetobehavewellandkeeptheagreement.Unlesstherearesomemechanismstomonitorthebehaviorofthepartiesoftheagreementandunlesstherearesomemeanstoenforcethepact,anyonemayhavethetemptationtorenegefromitandtoundothe inthispactofmutualassistance:hecansendhisowntroopsorengineercoalitionsofnotablestohelpinsuppressinglocalrevolts.12 agreementtojointlyshareinruling.Theautocratic\pact"requirestheexistenceofsomepublicbodyorinstitutionthatcon- rmsthenatureofthedealmadeamongautocratsandthatguaranteesitsmaintenance.Naturally,thisbody(ororganization)cannotbeathirdpartyindependentfromtheau-tocraticelitetowhomthelatterentruststheenforcementofthepact.Ifthatbodywereexternaltotherulingclique,thesovereigntyofthestatewouldbeactuallylyinginitshandsandnotunderthecontrolofthegoverningelite.Theinstitution(orinstitutions)thatem-bodies(andpreserves)thepactmustbesomestructurewherethemembersoftheelitearerepresentedorinwhichtheyparticipateandwhichre\rectsthenatureofthepactamongthemembersoftheelite(eitherahorizontalpactamongequalsorahierarchicalpactinwhichthemembersoftheeliteparticipateinthegovernmentindi erentdegrees).Thiscollectivebody(orsetofbodies)mustbesuchthatitservesbothtomonitortheautocraticdeal(bygivingtherighttypeofinformationaboutthecurrentbalanceofpowerandaboutthepossibleattemptstoalterit)andtoenforceit(bypunishingorcrediblythreatentopunishthosethatmaydeviatefromthepower-sharingagreement).Inelectoralautocracies,theinstitutionalsolutiontakestheformofacompoundedstruc-tureofgovernment.First,thereisthedictatoratthefrontofthestate.Second,thenotablesmeetwithsomeregularityinanassembly.Finally,thedictatorandthenotables(inassembly)meettogethertorearmthetermsandvalidityoftheirrelationship,toex-changeinformationontheiractionsandplans,toeasesuspicions,releaseanytensionsandsolveanydisputes.Thiscompoundedstructureofgovernmentful llsinanappropriatemannerthetwocentralfunctionsofmonitoringandenforcementneededtosustaintheautocraticdeal.Itgiveseveryoneenoughinformationaboutthestatusofalltheinvolvedparties.Inmeeting13 togetherasaseparategroup,eachnotablechecksthattheirequalsarestillaliveandthattheyhavenotexperiencedarbitrarychanges(broughtonbytheleader)inpower,assetsandstatus.Similarly,everynotableobservesthenature,sizeandstabilityoftheexistingfactionsinthecountry.Indoingso,heveri esthatnosectionofthenotableclassistooloyaltotheleaderor,inotherwords,too\monarchical."Theveryroutineofmeetinginanassemblyalsoservesasayardsticktomeasuretheintentionsoftheleader.Anyattemptbythenationalleadertoblockornotconvenehislords'assemblyisasignalthatheisindeedintentondisruptingtheoldbalanceofpowerandshouldthereforetriggeranimmediatebacklashfromhisnotables.Itisalsoanadequate(evenifitisnotabullet-proof)mechanismtoenforcethepact.Inanautocracythenationalleaderisstrongerthananyindividualnotable(orthanafractionofthenobility)yetweakerthanthewholepopulationofnotables.Thisis,asamatteroffact,notunlikeatyrannicalgovernment,wherethetyrantalwayssuccumbstothecoordinatedantagonismofhissubjects.Butautocraciesandtyranniesdi erinakeysense.Inatyranny,thetyrant'sgoalistopreventtheverypossibilityofcoordination.Inanautocracythelegislature(andotherorganizationsorbodiessuchasregularmeetingsofpartynotables-conditionalonthelatterhavingsomeautonomousbasisofsupport)makethepossibilityofcoordinationavailableinapublicmanner.Inanautocracy,thenationalleaderhasalreadyconcededtothelocalorregionalnotablesthepossibilityofcoordination.Notablesmeet,withsome(notnecessarilyhigh)regularity,andthismeetingmaybecomeanoccasiontocoordinateagainsttheoverlord.Toputitinotherwords,theassemblyofnotablesreducesthecoordinationcostsoftheagentsoftheoverlordandhencetipsthebalanceawayfromwhatotherwisewouldbecomeatyranny.Tosumup,acompoundsystemofgovernment,withapreeminentleaderandsome14 sortofassembly(whosepowersmayextendfrompureconsultationtolegislativecapacity),stabilizestheautocraticdeal.Theexistenceofanassembly(anditsinteractionwiththenationalleader)actsastheenforcerofthedeal.Withoutanassemblyorafocalpointofcoordination,andeveniftherewerestrongnotablesorfactionsaroundthecountry,wecouldnottalkaboutanautocraticgovernment(agovernmentbyapluralityofindividuals).Thenationalleader(orevenanentrepreneurialnotable)wouldemploythelackofhorizontalcoordinationtodestroyhisenemies,tocreatenetworksofprivilegedandto nallygovernabsolutely.Allthisdiscussiondoesnotentail,however,thatautocraciesandtheirassembliesshouldbesmoothbusinesses.Iftheywere,wewouldbeinthepresenceofadefactotyranny.Inanassembly,thenotableshaveanopportunitytovoicetheirgrievances,suggestalternativecoursesofactionanddisruptanytyrannicaldesignsoftheautocraticleader.Sporadicclashesbetweenthenationalleaderandtheassemblyofnotablesshouldbeexpectedbecausetheyconstitutetheverymechanismsthroughwhicheachsidereevaluatesthestrengthoftheotherparty.Still,thoseclashes(ortemporarydeviationsfromacooperativeequilibrium)areveryunlikelytoresultingeneralizeddisorderandinthedethronementofthenationalleaderfortworeasons.Deposingthedictatormaybedonewithtwoalternativegoalsinmind.Theclassofnotablesmaytakeadvantageofitsmeetingtodeposetheleadertoestablishanewformofgovernment{suchasarepublicofnotablesoroligarchs(wheretheleaderisdirectlyelectedbythelatterandthereforehasbecomeamuchweaker gure).Instead,theymaypursuealessradicalgoal{maintainingtheelectoralautocracywhiletransferringtheleadershiptoanewindividual.Letusstartdiscussingwhythesetwopossibilitiesareveryrareeventsbylookingatthe15 lattercase(achangeinleader) rst.Replacingleaderscomeswithahighprize:violence,thelikelyemergenceofnew(andpermanent)factionsthattransformthecurrentdistributionofbene tswithintheelite,and,evenmoreworryingly,thepossibilitythatthechangeofleaderisthedoingofafactionthathas,asitsgoal,theconstructionofanetworkthatwillresultinthecleansingofpartofthecurrentcliqueofnotables.Yetthisprizeispaidwithoutchanging,inanyrelevantway,thestructureofdecision-makingandtheincentivemechanismsthatareembeddedinthecompoundedsystemofgovernmentwearedescribing.Hence,theclassofnotableswillbeextremelyreluctanttochangeleaders.Itwillonlyreplacethedictatorwithanewleaderifitbelievesthecurrentoneisbecomingatyrannicalautocrat.4PoliticalInstitutionsandPower-SharinginAutocra-ciesApossiblewaytoformalizetheaboveintuitionmaybedonebyconsideringthefollowingsetting.Inperiodone,thenationalleaderfacesathreattohispositionfromachallengerorarevolutionaryaction.Assumethatthethreattothenationalleaderisofmagnitudem�0.Inordertodefeatthisthreat,theleadermayenlistthesupportofsomenumberofnotables.Weassumethatnotablesmayenjoysigni cantin\ruencelocallybutthatthepowerofanysinglenotableisofverylittleconsequenceatthenationallevel.Tocapturethatidea,supposethereisalargenumberofidenticalnotableswhosejointpoweratthenationallevelismeasuredbyn0,whilethepowerofanysinglenotableisclosetozeroatthenationallevel.Inordertounderstandtheconsequencesofthethreatfortheleader,weassumethattheleaderdefeatsthethreatwithprobability(m;n),whichisalogisticfunctionofthe16 di erencebetweenthejointpowerofnotablesnandthemagnitudeofthethreatm,(m;n)=1 1+e (n m).Intuitively,(m;n)isincreasinginm n,(m;n)=1 2whenm=n,andextremeimbalancesinpowerbetweentheleader'scoalitionandthethreatimplyalmostcertainvictoryordefeat. 6 Recruitingnotablesis,however,costly.Inordertorecruitthenotables,theleaderagreestosupplytheminthepresentandeachfollowingperiodwithresourcesthateachnotablecanusetocultivatehislocalpoliticalin\ruence.Suppose,therefore,thatenlistingnotableswiththejointpowerofonecoststheleaderc�0.Ontheotherhand,theleader'spayo isidenticaltothesizeofthenationalbudgetb�2cwhenheisinpower,andweassumethattheleaderreceivesthepayo ofzeroifheisdefeatedbythethreat. 7 Theninperiodone,theexpectedpayo totheleaderwhenherecruitsnotablesofmagnituden,whilehefacesathreatofmagnitudemis(m;n)( cn)+[1 (m;n)]0.(1)Theleaderrecruitsanoptimalnumberofnotablesnwhenhechoosesn0thatmaximizeshisexpectedpayo in( 1 ).Takingthederivativeof( 1 )withrespectton,theleaderobtainsamaximumexpectedpayo whenever( c cn)em cen=0, 6 Skaperdas ( 1996 )foradiscussionofcontestsuccessfunctions.Notethat(m;0)declinesfrom1 2tozeroasmincreasestoin nity.Thisallowsforthepossibilitythataleaderthatdoesnotenlistanynotablessurviveswithapositiveprobabilityifthethreatisofasmallmagnitude.7Theassumptionb�2ensuresthatthebudgetbissucienttoenlistapositivenumberofnotablesandthatn0,seethede nitionnbelow.17 orequivalentlywhenevern= c 1 W(eb c m 1).(2)Inequation( 2 ),W(x)istheproductlogfunctionde nedastheinversefunctionofx=WeW. 8 Theexpressionin( 2 )hassomeveryintuitiveimplicationsforthenumberofnotablesthattheleaderoptimallyrecruits.Whenconsideringwhetherheshouldrecruitmorenotables,theleaderweightstheincreasedlikelihoodofdefeatingthethreatagainstthecostofmaintainingalargercoalitionofnotables.Theleaderwilloptimallyrecruitmorenotableswhenthethreatmislargerandwhenhisbudgetislarger,buthewillrecruitfewernotablesasthecostcofenlistingtheirsupportincreases.Furthermore,theoptimalnumberofrecruitednotablesnexhibitsdiminishingmarginalreturnsinthemagnitudeofthethreatm.Finally,theleaderneverrecruitsmorenotablesthanhecana ord,evenifhefacesathreatofaverylargemagnitude.Toillustratetheserelationships,weplotnfor=10,c=1,andm2(0;20]inFigure 5 .Figure 5 abouthere.Nowconsidertheleader'sincentivestomaintaintheoptimalcoalitionninthefuturegiventhefollowingsetting.Inperiodtwoandanysubsequentperiod,theleader rstdecideswhatnumbernofnotablestoenlistandthenotablesdecidewhethertosupporttheleader.Second,athreatofmagnitudemarisesagainwithprobabilityp2(0;1)(inwhichcasetheleaderwouldhaveoptimallyenlistednnotablesinordertosurvive.)Wheneverathreatdoesnotarise,whichhappenswithprobability1 p,theleaderremainsinpowerforsure(inwhichcasetheleaderwouldhaveoptimallyenlistedzeronotables.)Finally,afterlearning 8Givenourassumptions,thefunctionW(x)issingle-valued,strictlyincreasing,concave,andsuchthatW(0)=0andlim!1W(x)=.18 whetherathreatispresentandcontingentondefeatingthethreat,theleadermaydeliverthepromisedresourcestothenotablesheenlistedifanyextantthreatissuccessfullydefeated.Iftheleader'scoalitiondefeatsthethreat,thegamemovestothenextperiod;ifinsteadtheleaderisdefeated,thenthegameends.Thenotablesunderstandthatattheendofanyperiod,theleaderwillfacethetemptationtorenegeonthepromisetodeliverthemresourcesinreturnfortheirsupport.Inturn,theywillsupporttheleaderinperiodoneonlyiftheycanexpecthimtodeliverthepromisedbene tsattheendofeachsubsequentperiod.Aswehaveemphasizedpreviously,thenationalleaderandthenotablescannotrelyonanoutsideauthoritytoenforceanagreementbetweenthemsincethisagreementistheveryfoundationoftheirpoliticalauthoritywithinthecountry.Wewillthereforerequirethatanyagreementbetweentheleaderandthenotablesis\self-enforcing"inthesensethatnopartyhasanincentivetodeviatefromitatanytime.Giventhisconsideration,weconsideroptimalbehaviorbytheleaderandthenotablesinturn.Consider rsttheexpecteddiscountedpayo totheleaderfromenlistinganoptimalnumberofnotablesnfromperiodtwoonwardanddenoteitbyVS(Sstandsfor\power-sharing".)Assumeadiscountfactor2(0;1).ThenwehaveVS=(1 p)[ cn+VS]+p(m;n)[ cn+VS],orequivalentlyVS= cn 1 1 p+p(m;n) .(3)Incontrast,whennonotablessupporttheleaderhisexpecteddiscountedpayo fromperiod19 twoonwardisVT=(1 p)[+VT]+p(m;0)[+VT],orequivalentlyVT= 1 1 p+p(m;0) .(4)ComparingVSandVT(whereTstandsfor\tyranny")shedslightontheincentivesoftheleadertomaintainacoalitionofoptimalsize.Thereisathresholdprobabilityofthreatp suchthatwhenpp ,theleaderfeelssosafethathelackssucientincentivestomaintainacoalitionofnnotables,VSVT.Thenotablesunderstandthisandwillthereforenotsupporttheleaderinanyperiod.Thisisalong-livedequilibriumoutcomeandwecallitatyranny.Therewillalsobethresholdsofathreatmagnitude m,cost c,nationalbudget ,anddiscountfactor suchthatifm morc cor or ,theleaderissovulnerabletothethreatofachallengerorarevolutionaryactionthathedoesnotevenbotherrecruitingacoalitionofnotablesandenjoysthebene tsofrulingalone\whileitlasts."Again,thenotablesanticipatethisandwillnotsupporttheleaderinanyperiod.Thisisashort-livedequilibriumandwecallitalame-ducktyranny.Importantly,however,thereisapoliticallyinterestingrangeofparameterswithinwhichpower-sharingbetweentheleaderandthenotablesmaybefeasible:Ifthenotablescoulddeviseproperincentivesfortheleadernottorenegeonthebene tspromisedinreturnfortheirsupport,theleaderwouldactuallyprefertomaintainanoptimalcoalitionofnotablestorulingalone.Thiswillbethecasewhenp�p ,m m,c c,b�b ,and� .Howmightnotablesfostersuchincentives?Weconsidertwopossibilities.The rstisabilateralenforcementmechanismaccordingtowhichanyonenotablethreatenstoper-20 manentlywithdrawhissupportfromtheleaderifheisdeniedthepromisedbene ts.Asecondpossibilityisamultilateralenforcementmechanismaccordingtowhichallnnota-blesthreatentoneveragainsupporttheleader,shouldanynotablenotreceivehispromisedbene ts.Thissecondenforcementmechanismrequiresaforumthroughwhichnotablescanmonitortheleader'srelationshipwitheveryothernotable.Weproposethatanationalleg-islatureprovidessuchaforum.Wethereforecallthe rstsettinganautocracywithoutalegislatureandwecallthesecondanautocracywithalegislature.Weexaminethesesettingsinturnhere.Underautocracywithoutalegislature,theleader'scommitmenttomaintainacoalitionofnisnotcredible.Inperiodtwo,beforehelearnswhethertherewillbeanotherthreat,theleaderwouldideallyformacoalitionwithonly^nnnotablessuchthat^nmaximizeshisexpectedpayo inthatperiod,(1 p)[ cn]+p(m;n)[ cn].Theleaderwillthereforeprefertodismissapositivenumbern ^nofnotablesfromthepower-sharingarrangementdevisedinperiodone.Inotherwords,whatseemedtobeacoalitionoftheoptimalsizeinperiodone(whenthethreatwascertain)becomestoolargeofacoalitioninperiodtwo(whenthethreathappensonlywithprobabilityp.)Thusthethreatofpunishmentbyanyindividualnotableinperiodtwowillnotbesucienttocommittheleadertomaintaintheoriginalcoalitionofsizen.Thiscredibilityproblemthatnotablesfaceisresolvedunderautocracywithalegislature.Theleader'sexpecteddiscountedpayo fromperiodtwoonwardwhennotablesjointlywith-drawtheirsupportisVT.RecallthatVTVS.Thusthepresenceofthelegislaturemakes21 thenotables'threatofpunishmentcredible.Nowtheleaderpreferstomaintainacoalitionofsizenaslongas cn+VC+VP.Althoughalgebraicallytedioustocompute,thereexistsaunique2(0;1)thatsatis estheaboverelationatequality:Whenallnotableshavetheabilitytocoordinateonwithdrawingsupportfromtheleadershouldanyofthembedismissedfromthepower-sharingarrange-ment,theleaderwillmaintaintheoriginalcoalitionofsizenforasucientlyhighdiscountfactor2[;1).Putsimply,crediblepower-sharingbetweentheleaderandnotablesisthereforepossibleonlyunderautocracywithalegislature.Finally,weneedtoverifythatthenotableswouldwanttosupporttheleaderinanyperiodaslongashemaintainsacoalitionofsizenbutwouldwithdrawtheirsupport,shouldanyofthembedeniedthebene tspromised.Tomaketheseincentivesassharpaspossible,supposethatwhenevertheleaderhonorshiscommitment,eachnotableinhiscoalitionreceivesbene tsofsizeg�0.Incontrast,iftheleaderrenegesonhiscommitment,eachnotablereceiveszerobene ts.Supportingtheleadercoststhenotablesnothinginperiodswithoutathreat.However,inperiodswhenathreatexists,suchsupportiscostlyanddependsonthewhetherthethreatisdefeated.Thatis,inperiodswhenathreatispresent,eachnotableincursthecostf�0whentheleader'scoalitionfailstodefeatthethreat,andweassumef�g.Ontheotherhand,eachnotable'scostisgwhenthethreatisdefeated.Tosummarize,thepayo toeachnotabledependsonweather(i)heisamemberoftheleader'scoalition,(ii)athreattothedictator'spowerexists,(iii)thatthreatisdefeated,and(iv)theleaderhonorshiscommitmenttothenotableswhosupportedhim.Thesepayo saresummarizedinTable 3 .Weseethatsupportingtheleaderinanyperiodwhenathreatispresentiscostlytothenotables{theydoatleastaswelloutsidetheleader'scoalitioneven22 ifthethreatisdefeated.Butsupportingaleaderwhowillbedefeatedistheworstpossiblechoice;irrespectiveofwhethertheleaderwouldhonorhiscommitment,eachnotable'spayo is f.Thusthenotablesonlyenjoybene tsfromsupportingtheleaderinperiodswhenthreatsareabsentandonlyaftertheyhavedefeatedthreatsinpreviousperiods.Table 3 abouthere.Wecannowverifythatthenotable'schoicesunderautocracywithalegislaturearesub-gameperfect.Anotable'sexpecteddiscountedpayo whenheisamemberoftheleader'scoalitionandtheleaderhonorshiscommitmentsisvS=(1 p)g+vS+p([1 (m;n)]( f)+(m;n)vs),orequivalentlyvS=g(1 p) fp[1 (m;n)] 1 p(m;n).Anotablewillthereforeprefertosupporttheleaderfromperiodoneonwardaslongas(m;n)vS+(1 (m;n))( f)0.(5)Intermsofthevaluesof(m;n)andp,inequality( 5 )impliesthatnotableswillnotsup-portaleaderwhoisabouttolosetoathreat,noraleaderwhoneedshelpdefendingagainstthreatstoooften.Thiswillbethecasewhentheprobability(m;n)istoosmallortheprobabilitypistoohigh,respectively.Assumingthattheseconditionsdonotobtain,nono-tablewillwithdrawhissupportfromtheleaderaslongastheleaderhonorshiscommitmenttomaintainthecoalitionofsizen.Ontheotherhand,oncetheleaderviolateshisagreementbetweenwiththenotables,anynotablewillprefertowithdrawsupportfromtheleaderaslongallothernotableswithdraw23 theirsupportaswell.Thisisbecausebeingtheonlymemberoftheleader'scoalitionmakesitextremelylikelythattheleaderwillbedefeatedbythethreat.Recallourassumptionthatthepowerofanysinglenotableisclosetozeroatthenationallevelanddenoteitby".Thentheexpectedpayo toasinglenotablewhosupportstheleaderwhileallothernotableshavewithdrawntheirsupportis[1 (m;")]( f)+(m;")vs.(6)Thisexpressionisnegativeastheprobability(m;")approacheszeroforasucientlysmall".Thusweseethatgivensomeintuitiveconstraints,crediblepower-sharingbetweentheleaderandnotablescanemergeunderautocracywithalegislature.5LocalNotablesandtheElectoralConnectionAspointedoutintheintroduction,non-tyrannicalautocraciesarecharacterizedbyatwo-levelgame{ rstwithintheeliteandthenbetweeneachnotableandtherestofthepop-ulation.Inthegamebetweeneachnotableandhis\subordinate"population,thenotableisinterestedinsustainingtheresourcesthatallowhimtoparticipate(asanautonomousagent)intheautocraticdealwiththenationalleader.Inturn,thenationalleaderisalsointerestedindeterminingthestrengthofeachnotable,thatis,theformerwillbene tfromconditioningthefuturepower-sharingwithanysinglenotableonthenotable'scontinuingin\ruence.Thelocalin\ruenceofnotablesmaytakeatleasttwobroadforms,whichwecorrelatedwithcurrenteconomicstructuresandtechnologies.Ontheonehand,localnotablesmayderivetheirpositionfromlandholdings(withsomepopulationoflaborers xedtothem).24 Landpossessionsareaclear-cutindicatorofanotable'slocalin\ruence.Electionsarenotnecessarytosustainthenotable'sstatus.Underthosecircumstanceswehave\feudalmonar-chies",inwhichamonarchpreservesapreeminentposition,usessomeclassoflandownerstomaintainhiskingdomandagreestocallwithsomeregularitysomeparliamentorassemblyoflords.Noticethatthissolution(prevalentinmedievalEurope)isclearlydi erentfromthosepalatialregimesinwhichthemonarchisatyrantordespotwithacourtofservants(thecaseofAsiandespotisms,severalabsolutistregimesinmodernEuropeandsoon).Ontheotherhand,asanincreasingfractionofwealthandtheensuingpoliticalin\ruencederivesfromtradeandservices,landholdingsceasetobeasaninformativeindicatorofnotables'in\ruence.Thisisbecausetradeandservicerelytoagreatextentonastockofcontractualrelationshipsthatcannotbemeasuredviaaneasilyobservablefactorsuchaslandholdings.Underthosecircumstances,localnotables'in\ruencederivesfromtheconstructionofanetworkofclientsandthedistributionofrents(rangingfromsmallprivategoodstolegalconcessionstoexploitresources,buildpublicworks,etc).Insuchatypeofeconomy,district-levelresultsofelectionsplayagrowingrolebecausetheybecomeanincreasinglyusefulindicatorofwhetherthenotablemaintainshislocalin\ruence.Supposethereforethatdistrict-levelelectoralresultsareanimperfectsignalofeachno-table'slocalin\ruence. 9 Usingtheargumentsdevelopedin Fudenbergetal. ( 1994 ),wecanshowthatpower-sharingbetweenthenationalleaderandnotablesisfeasibleunderawiderrangeofcircumstanceswhensuchaninformativesignalisusedtoconditionthecontinuationofpower-sharingbetweentheleaderandanyindividualnotable.Wecallthisequilibriumanelectoralautocracywithalegislature. 9Moreprecisely,weassumethattheyarecorrelatedinthesenseofthemonotonelikelihoodratioproperty.25 6ConclusionsFromahistoricalpointofviewdemocracyhasalwaysconstitutedaveryexceptionalformofgovernment.Untilthelasthundredyears,republicanpolitieswerecon nedtoafewcitiesintheclassicalworldandinmedievalandmodernEurope-andeventhentheirdemocraticinstitutionswereofthemostimperfectsort.Aback-of-the-envelopecalculationshowsthat,sinceitemergedabout100,000yearsago,closetoninetyninepercentofmankindhasbeengovernedbyauthoritarianrulers-tyrants,monarchs,princesandwarlordsofallvenues.Yettheliteratureonnondemocraticregimesisstillatitsinfancyforatleasttworea-sons.First,mostoftheliteratureondictatorshipsmistakenlytreatstheruleofasingleindividualortyrantasthestandardtypeofautocraticgovernment.Second,incontrasttoallthegrowingliteratureonelectionsanddemocraticinstitutions,whichhasbeengraduallytransformedbythereceptionofanalyticaltoolsandthedevelopmentofcausalmodels,theexaminationofauthoritariansystemsisstillweddedtoasociologicalapproachcommittedtotheconstructionanddescriptionofidealtypes,thatis,oftypesdevelopedonthebasisofthehighfrequencyofcertainparticulartraits.Bycontrast,thispaperstartsfromtheassumptionthatausefultheoryofdictatorshipsshoulddepartfromanexplorationofthewayinwhichpowerisdistributedandsustainedbothwithintherulingclassandbetweentherulerandtheruled.Itisthisdistributionwhichdeterminestheincentivesandbehavioroftherulersanditssubordinates.Accordingly,wesuggestthattheappropriatepointofdeparture(inawayfollowingclas-sicalauthors)consistsindistinguishingbetweentyrannies(single-rulerdictatorships)andautocracies(dictatorshipsgovernedbyanauthoritarianclique).Inatyrannythedictatoristhesolebearerofpowerandassuchhisexerciseofpowerisnotsubjecttoanyinstitutional-izedlimitsorconditionsexceptforthosehehimselfchoosestoimpose.In(non-tyrannical)26 autocracies,instead,severalindividualsshareinthesovereigntyoverthestate.Thissystemofsharedsovereigntyconstrainseachoneofthemasfollows.Sincesharingsovereigntycomesfromtheverydecisionofseveralindividualstosubjectthemselvestosomecommonproce-durethatappliestoallofthem,themembersofanautocraticregimecannotunilaterallyreworktherulesofconductthatgovernthem.Withinthecategoryofnon-tyrannicalautocracies,wehavehereinvestigatedthemostcommonsystem,thatis,oneinwhichanationalleaderreliesonasetofnotablesandpartiallysharespowerwiththem. 10 Thecredibilityanddurabilityofthatpower-sharingarrangementissustainedbytwoinstitutionalfeatures:alegislatureandelections.Thetwoinstitutionsperformtwodistinctincentiveroles.Thelegislatureallowsthenotablestomonitorwhethertheleaderhonorstheiragreementtosharepowerand,crucially,tocoordinateonjointlywithdrawingsupportfromtheleadershouldherenegeonhispromisetosharepowerwithanyofthem.Ontheotherhand,electionsallowtheleadertomonitorthelocalin\ruenceofindividualnotablesandthusensurethathemaintainsacoalitionthatiscapableofdefeatinganythreatbyachallengerorarevolutionaryaction.ReferencesArendt,Hannah.1973.TheOriginsofTotalitarianism.HarvestBooks.Boix,CarlesandSebastianRosato.2001.\Acompletedatasetofpoliticalregimes,1800-1999." 10Anotherformofautocraticsystem,notexploredhere,wouldbeanon-hierarchicalpower-sharingagree-ment,thatis,oneinwhichthereisnonationalleaderandallnotablesparticipateinacommitteewithsimilarorequalrights.ThiswouldbethecaseofPoland'searlymodernDietorseveralItalianoligarchicalrepublics.27 Brownlee,JasonM.2004.\RulingPartiesandDurableAuthoritarianism."CenteronDemocracy,DevelopmentandtheRuleofLawWorkingPaperno,23.StanfordInstituteonInternationalStudies.Friedrich,KarlandZbigniewBrzezinski.1965.TotalitarianDictatorshipandAutocracy.Cambridge,MA:HarvardUniversityPress.Fudenberg,Drew,DavidLevine,andEricMaskin.1994.\Thefolktheoremwithimperfectpublicinformation."Econometrica62(5):997{1039.Gandhi,JenniferandAdamPrzeworski.2006.\Cooperation,Cooptation,andRebellionUnderDictatorships."Economics&Politics18(1):1{26.Haber,StephenH.2007.Keefer,Philip.2002.DPI2000:DatabaseofPoliticalInstitutions.Washington,DC:Devel-opmentResearchGroup,TheWorldBank.Kuran,Timur.1991.\NowOutofNever:TheElementofSurpriseintheEastEuropeanRevolutionof1989."WorldPolitics44:7{48.Levitsky,StevenandLucanA.Way.2003.\AutocracybyDemocraticRules:TheDynamicsofCompetitiveAutoritarianisminthePost-ColdWarEra."Unpublishedmanuscript.Linz,Juan.1975.\TotalitarianandAuthoritarianRegimes."InGreenstein,FredandNelsonPolsby(Eds.),\HandbookofPoliticalScience,"vol.3.Reading,Mass.:AddisonWesleyPublishingCompany,pp.191{357.Linz,JuanJ.2000.TotalitarianandAuthoritarianRegimes.Boulder:LynneRienner.28 Lust-Okar,Ellen.2006.\Electionsunderauthoritarianism:PreliminarylessonsfromJor-dan."Democratization13(3):456{71.Machiavelli,Nicolo.1513/1985.ThePrince.Chicago:TheUniversityofChicagoPress.Magaloni,Beatriz.2006.Votingforautocracy:HegemonicpartysurivalanditsdemiseinMexico.NewYork:CambridgeUniversityPress.Myerson,RogerB.2006.\Leadership,trust,andconstitutions."Unpublishedmanuscript.Neumann,Franz.1957.TheDemocraticandAuthoritarianState.NewYork:FreePress.North,DouglassC.andBarryR.Weingast.1989.\ConstitutionsandCommitment:TheEvolutionofInstitutionsGoverningPublicChoiceinSeventeenth-CenturyEngland."JournalofEconomicHistory49(4):803{832.Przeworski,Adam,MichaelE.Alvarez,JoseAntonioCheibub,andFernandoLimongi.2000.DemocracyandDevelopment:PoliticalInstitutionsandWell-BeingintheWorld,1950-1990.NewYork:CambridgeUniversityPress.Skaperdas,Stergios.1996.\ContestSuccessFunctions."EconomicTheory7(2):283{90.Tullock,Gordon.1987.Autocracy.Boston:KluwerAcademicPublishers.Wintrobe,Ronald.1998.Thepoliticaleconomyofdictatorship.Cambridge:CambridgeUniversityPress.Xenophon/Strauss.1961.OnTyranny.Chicago:TheUniversityofChicagoPress.29 Table1:Transitionstodictatorialregimes,1955-99 BecameindependentFromdemocracyFromtyrannyFromautocracyandandtotototoautocracytyrannyautocracytyrannyautocracytyranny 1955-593112241960-6420136651965-6970175121970-7421041316Total1955-7432351926371975-791020001980-840060001985-890120001990-942040001995-99001000Total1975-993115000 Sources: Przeworskietal. ( 2000 )for1955-74, Keefer ( 2002 )for1975-99.30 Table2:Regimetransitionstoautocraticregimes,1975-99 BecameindependentFromdemocracyFromsoftautocracyFromhardautocracyandandtotototosofthardsofthardhardsoftautocracyautocracyautocracyautocracyautocracyautocracy 1975-790120231980-840060121985-890020031990-9402405131995-990010109Total031501830 Sources: Przeworskietal. ( 2000 )for1955-74, Keefer ( 2002 )for1975-99.31 Table3:Payo stonotablesNotable SupportDonotsupportNatureNothreatDefeatthreatLosetothreat Probability1 pp(m;n)p[1 (m;n)] LeaderHonorg0 f0Renege0 g f0 32 Figure1:Numberofdictatorshipswithandwithoutlegislature,1951-1999.33 Figure2:Frequencyofdictatorshipswithandwithoutlegislature,1951-1999.34 Figure3:Numberofdi erenttypesofdictatorships,1975-1999.35 Figure4:Proportionofdi erenttypesofdictatorships,1975-1999.36 Figure5:Leader'soptimalrecruitmentofnotablesnfor=10,c=1,andm2(0;20].37