This decline is surprising given that 7 coups have happened between January 2008 and December 2010 bringing the last decades totaltoalmostthreedozenOneexplanationforthelackofcoupresearchistheabsenceofatemporallyandspatially comprehensive dataset to ID: 44114
Download Pdf The PPT/PDF document "Special Data Feature Global instances of..." 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-civilianmembersofthemilitaryandsecurityser- vices,orcivilianmembersofgovernment. Thisapproachhasthreeprimaryadvantages.First,it avoidsconflatingcoupswithotherformsofanti-regime activity,whichistheprimaryproblemwithbroader approaches.Taylor&Jodice(1983),forinstance,claim thatirregularexecutivetransferscanbeperpetratedby groups,cliques,cabals,partiesorfactionseitheroutside orinsideagovernment,anditsagencies ... rebellious minorities ... themilitary,orconspiratorsbackedbyfor- eignpowers.Thisrisksconflatinganumberofanti- regimeactionsthataredistinctfromcoups. 2 Forexample, civilwarscommonlyincludevastsegmentsofthegeneral populationthatneednothaveanytiestothegovernment (e.g.Small&Singer,1982:210),whilerevolutionsor popularmovementscouldalsobeclassifiedascoups underbroaderdefinitions(Goodwin,2001:9).Though manyfeaturesofcoupsarepresentincivilwarsandrevo- lutions,thelattersexecutionbythemassesrequirestheir distinctclassification. 3 Second,allowingnon-militaryelitestoperpetrate coupsisnecessarybecausetheinitialinstigationofacoup attemptfrequentlyinvolvescivilianmembersofthegov- ernmentalone,withthemilitaryplayingalaterrolein decidingwhethertheputschwillbesuccessful.Focusing exclusivelyonthemilitarywouldlikelybiasourdataset towardsselectingonlysuccessfulcoupattempts.For example,the1962coupattemptledbySenegalesePrime MinisterMamadouDiafailedbecausehewasunableto gainsupportthroughoutthemilitary.Whilethisis clearlyacoupattempt,thecasewouldnothavemetthe morestringent,military-onlydefinitions. Third,therequirementthatcoupperpetratorscome fromwithinthestatedifferentiatescoupsfromexecutive changesbroughtaboutbyinternationalforce.While coupscanbebackedbyforeignpowers,theyareonly includedifforeignersactedinasupportingrole.For example,the1953Iraniancoupwasunlikelytohave occurredwithoutCIAinfluence,butitwasultimately IranianactorswhooverthrewMosaddeq.Incontrast,the fallofIdiAminatthehandsoftheTanzanianmilitaryin 1979doesnotconstituteacoupbecauseforeignpowers weretheprimaryactors. Tactics Twofactorsmustbeconsideredinregardtothetactics usedtooverthrowthechiefexecutive.First,theactivity mustbe illegal .ThisdiffersfromFiners(1988:3)early definitionofacoup,whichisinterestedmoregenerally inthemilitarysinterventionintocivilianaffairs.Others areconsistentwithourcriterion,includingMcGowans (2003)claimthatcoupsmustbeillegallyundertaken. Theillegaldistinctionisimportantbecauseitdifferenti- atescoupsfrompoliticalpressure,whichiscommon wheneverpeoplehavefreedomtoorganize.Massivepro- testsinThailandpromptedGeneralAnupongtobluntly adviseThaiPrimeMinisterSomchaitostepdownin 2008,forexample.Thisattempttoinfluencepoliticswas perfectlylegaland,therefore,doesnotconstituteacoup. Second,anear-universalcriterionforcoupsisthatvio- lencedoesnothavetobepresent.Finer(1988:23) includeseventsinwhichnobloodhasbeenshed,while othershaveusedthethreatofforceorviolenceinaddi- tiontoitsovertuseasacriterion.Weremainconsistent withthisdistinctioninproviding nominimaldeath criter- ionforaneventtobeconsideredacoup.Thisruleavoids conflatingcoupswithcivilwars.Whilemanycoup attemptshavesparkedcivilwars,mostarebloodless. Plotsandrumors Welchscontentionthatcoupsareclearevents,easyto dateandpossibletodocumentincludedaqualifier hewasspeakingonlyofsuccessfulcoups.Changesina stateschiefexecutiveareeasytospot.Itbecomes increasinglydifficulttoidentifymoreambiguousforms ofcoupactivity,suchascoupfailures,plots,andrumors. Webeginbyeliminatingtheleastreliableevents:coup plotsandrumors.Weexcludetheseeventsfortworea- sons.First,theyareoftentoomundanetobepicked upbylargemediasources,whichmakesthemdifficult toreliablycode.Second,governmentsoftenhavean incentivetofabricateoroverstatecoupplotsandrumors tojustifyrepression.Kebschull(1994:568),forexample, notedthatplotscouldbedeliberatelycontrivednon- sense,putforwardtoservetheregimespurposeofiniti- atingemergencyrule,suppressingaparticulargroup,or justifyingotheractionssoughtbytheregime.Giventhe difficultiesinattemptingtoverifywhetherclaimsregard- ingplotsorrumorsarelegitimate,weopttofollow Finers(1988:307)criteriaincodingonlycaseswhere 2 Taylor&Jodice(1983)notethattheirmeasuredoesnotcoincide withtheclassicaldefinitionofacoup.However,theirmeasurehas beenutilizedintestingcouptheoriesonmultipleoccasions(e.g. Londregan&Poole,1990;Galetovic&Sanhueza,2000). 3 Previousscholarshavepaidcloseattentiontothisdistinctionto differentiatecoupsfromcivilwars.Morrison&Stevenson(1972: 128)requiredcoupstobeundertakenwithoutovertmass participation,aqualifierlaterincorporatedbyJackman(1978), Ferguson(1987:13),McGowan(2003),andMarshall&Marshall (2007:1). Powell&Thyne 251 SpecialDataFeature Globalinstancesofcoupsfrom 1950to2010:Anewdataset JonathanMPowell&ClaytonLThyne DepartmentofPoliticalScience,UniversityofKentucky Abstract Onceconsideredahottopicamongscholars,researchoncoupsde ´ tathaswanedinrecentyears.Thisdeclineis surprisinggiventhat7coupshavehappenedbetweenJanuary2008andDecember2010,bringingthelastdecades totaltoalmostthreedozen.Oneexplanationforthelackofcoupresearchistheabsenceofatemporallyandspatially comprehensivedatasettotesttheories.Alsoabsentisadiscussionofwhatmakescoupsdistinctfromotherformsof anti-regimeactivity.Thisarticleseekstoremedytheseproblems.Theauthorspresentanewdatasetoncoupsfrom 1950to2010.Theybeginbyexplainingtheirtheoreticaldefinitionandcodingprocedures.Next,theyexaminegen- eraltrendsinthedataacrosstimeandspace.Theauthorsconcludebyexplainingwhyscholarsstudyingavarietyof topics,includingcivilwars,regimestability,anddemocratization,wouldbenefitbypayingcloserattentiontocoups. Keywords civilwars,codingprocedures,coupde ´ tat,newdataset Introduction Whileresearchonviolentformsofanti-regimeactivity hasflourishedinrecentyears,thevolumeofcross- nationalresearchoncoupsisrelativelythin.Earlyefforts tounderstandcoupsmostoftenfocusedonspecific regions,LatinAmericaandAfricainparticular,while recentstudieshavetakenamoregeneralapproach. 1 WorkfromOKane(1987)andLondregan&Poole (1990)wereamongthefirsttouseglobalcoupdata,and subsequentcross-nationalresearchoffersmoregeneral theoryandcomprehensiveempiricaltests(e.g.Alesina etal.,1996;Galetovic&Sanhueza,2000;Belkin& Schofer,2003).Whilethisbodyofworkprovidesauseful startingpointtounderstandthecausesandeffectsof coups,therehasbeensurprisinglylittlediscussionasto whatacoupactuallyis.Wealsocontinuetolackabasic understandingofhowcoupsmighthaveabroadimpact onarangeoftopics.TherecentcoupsinThailandand Mauritania,forexample,suggestthatcoupscanquickly derailtheprocessofdemocraticconsolidation.Coupsalso seemtobelinkedwithcivilwars,includingthe1975coup thatsparked25yearsofviolenceintheBangladesh. Whiletheseexamplessuggestthatcoupsarerelevant formanyareasofstudy,severalbarriersinhibitourabil- itytoevaluatethesequestionsempirically.Themost obviousbarriersarethelackofadefinitionforacoup thathasbeenwidelydiscussedandacceptedbyscholars, andadiscussionofwhatdifferentiatescoupsfromother formsofanti-regimeactivity.Thepurposeofthisarticle istoovercomethesebarriers. Aworkingdefinition Welch(1970:1)hasclaimedthatacoupde ´ tatisa sharp,clearevent,easytodateand(ifsuccessful)possible todocument.SubsequentresearchsuggeststhatWelch wasperhapstoooptimistic.Whilesomeagreementatthe conceptuallevelhasemerged,acriticalexaminationof howscholarshaveoperationalizedcoupsrevealsseveral topicsdeservingoffurtherconversation.Webeginby 1 Forworkonspecificregions,seeFossum,1967;Dix,1994; Jackman,1978;Kposowa&Jenkins,1993;Lunde,1991; McGowan,2003;Agyeman-Duah,1990;Decalo,1990. Correspondingauthor: clayton.thyne@uky.edu JournalofPeaceResearch 48(2)249 259 ª TheAuthor(s)2011 Reprintsandpermission: sagepub.co.uk/journalsPermissions.nav DOI:10.1177/0022343310397436 jpr.sagepub.com p eace R ESEARCH journal of TableII.Datasummaryandcodingdiscrepancies Whole sampleBanks Belkin& SchoferFergusonFinerJanowitzKennedyLundeLuttwak Marshall& MarshallMcGowan Moreno etal.OKane Taylor& JodiceThompson Datasummary No.eventscoded12592313622951291102639784434199218134584215 No.falsepositives (percentofN) 811 (64.5) 88 (38.1) 141 (39.0) 116 (39.3) 51 (39.5) 18 (16.4) 126 (47.9) 19 (19.6) 20 (23.8) 124 (28.6) 45 (22.6) 107 (49.1) 42 (31.3) 276 (47.3) 45 (20.9) Accuracy*35.771.071.873.174.077.568.079.778.279.883.967.773.273.088.3 Distributionoffalsepositivesbyerrorcategory(percentages) Noevidenceof anti-regimeactivity 21.12.39.313.90.044.422.421.126.321.015.625.711.914.26.7 Noovertattemptto seizecontrol 20.65.813.613.03.95.630.426.310.529.826.76.70.027.413.3 Rebelorinsurgent fighting 18.916.39.37.03.90.012.826.321.116.113.342.92.421.533.3 Unclear11.059.352.953.978.45.64.80.021.10.00.01.081.00.70.0 Seizureofpowerby chiefexecutive 6.41.22.11.70.016.76.45.35.39.711.111.40.09.16.7 Challengenotagainst executive 5.70.01.41.73.90.04.00.00.04.86.72.90.08.815.6 Legalattemptto remove executive 5.67.05.03.55.922.24.85.35.35.70.02.94.85.815.6 Popularprotests5.34.74.32.63.95.69.65.30.07.32.24.80.04.06.7 Assassination attempts 3.11.21.41.70.00.02.410.510.51.617.80.00.05.50.0 Effortsbyforeigngovs/ mercenaries 2.22.30.70.90.00.02.40.00.04.06.71.90.02.92.2 *Accuracy ¼ (truepositives þ truenegatives)/(truepositives þ truenegatives þ falsepositives þ falsenegatives). discrepanciesemerge,theseaggregatescoresgiveus confidencethatthesedatasetsareprovidinganexcellent setofpotentialcoupstobeginbuildingourdataset. Focusingonthefirstcolumn,weseethatthemost commoncodingdiscrepancycamefromourinability toverifythatthecodedeventshappened(n ¼ 171). Ahandfuloftheseeventsappearedtocomefromthe scholarsinputtingthedataincorrectly(miscodingcoun- triesoryears).Othersarelikelyduetotheauthorsrelying ontheirspecializedknowledgeofforeigncountriesor languagestofindsourcesoutsideofourreach.Luttwak, forinstance,likelyreliedonhisextensiveforeignpolicy experiencewhencodingcoupattempts,whichcould explainwhy26.3 % ofhisfalsepositivesfallinthiscate- gory.Wereturntothiscategoryinanalyzingpotential biasesinourfinaldataset. Severalauthorsalsocodecoupattemptswhenthere werenoovertchallengestotheexecutivesauthority(n ¼ 167).Eventsinthiscategoryweredominatedbycoup plotsandrumors.TheHaitiangovernmentfoiledan allegedcoupplotinOctober2000,forexample,arrest- ingeightseniorpoliceofficers.However,giventhatthere wasnoactualattempttotakeoverthegovernment,itis unclearwhethertheallegedplotwouldhaveeverbeen attempted.Whileourearlierdiscussionjustifiestheir exclusionfromourdataset,weseethattheirinclusion byseveralauthorsyieldshighinstancesoffalsepositives. Thenextmostcommondiscrepancycomesfromcon- flatinginsurgentfightingwithcoupattempts(n ¼ 153). Rebelleadersurgedthemilitarytodefectwhenthe FMLNlauncheditsfirstmajormilitaryoffensiveagainst theSalvadoriangovernmentin1981,forexample.This eventshouldnotbeconsideredacoupbecausenosector ofthegovernmentaidedtheinsurgency.Thisandsimilar eventsexplainthebulkoffalsepositivesinMoreno etal.sdataset,forexample,andrepresentahighpercent- ageamongseveralothersources.Similardiscrepancies comefromconflatingpopularprotestswithcoup attempts(n ¼ 43).Thousandsofstudentsprotestedand riotedagainsttherulingSouthKoreangovernmentprior toRheeselectioncampaigninMarch1960,forexample. Thoughtheseeffortseventuallyledtohisresignation, therewasnoattemptfromwithinthegovernmenttoseize power. Thenextmostcommoncategory,unclear,comes fromourinabilitytomatchcoupeventswithotherpoten- tialcoupswithinthesamecountry/year(n ¼ 139).This happenedwhenthesourcesfailedtoincludethedayand/ ormonthintheirdatasetswhenthestatehadmultiplecoup attemptsinthesameyear.Thismadeitimpossibletoknow towhichcoupeventtheauthorwasreferring.Forexample, BankscodesonesuccessfulcoupintheDominican Republicin1961withoutreportingthemonthorday. Othersourcescodecoupsonthreedifferentspecificdates inthesameyear.ThismadeitimpossibletomatchBankss eventwithanyoftheotherthreeeventsreported.The failuretoreportthemonthand/ordaymakesupthebulk ofthefalsepositivesinfourotherdatasets. Manypotentialcoupsalsofailtomeetourdefinition becausetheydonotchallengethechiefexecutivesposi- tion(n ¼ 46).Militaryleaderspurgedmuchofthe SouthVietnamesegovernmentinDecember1964,for example,dissolvingtheHighNationalCouncilandthe provisionallegislature,andarrestingpoliticalleaders. Theydidnot,however,attempttoremovePresident PhanKhacSuufrompower. Manyeventsalsofailtomeetourillegalcriterion (n ¼ 45).Forexample,in1962PremierKhaled el-Azmcameintopowerwithavoteofconfidencefrom theSyrianAssemblyafterPremierBashiral-Azmeh announcedhiscabinetsresignation.Hequicklydis- solvedparliamentandbeganarulebydecree.Though thesemovescertainlyincreasedhispower,el-Azms moveswerewithintherulesoftheconstitution.After comingtopower,hisattempttoseizeextraconstitutional controllikewisedoesnotmeetourdefinitionofacoup becausetheeffortscamefromtheleaderhimself.We eliminated52similarevents,whicharecommonly referredtoas autogolpes (self-coups).Weseegenerally lowvaluesforeachofthesecategoriesamongour sources.Janowitzisanexception.Hisfalsepositivescan beattributedtohisratherbroaddefinitionofcoups, whichwouldallowforinclusionofbothlegalinterven- tionsand autogolpes . Thefinaltwodiscrepanciesincludeassassination attempts(n ¼ 25)andexternaleffortsbyforeigngovern- mentsormercenaries(n ¼ 18).Coupattemptsfre- quentlyincludeeffortstokilltheexecutive,suchasthe assassinationofFaisalIIofIraqin1958.Suchkillingsare rightlyconsideredtobecoupattemptswhentheassassin comesfromaconspiracywithinthestateapparatus. However,assassinationattemptsareexcludedwhenthey comefromoutsidethegovernment,suchastheassassi- nationofRwandanPresidentHabyarimanain1994. Whiletherearefewfalsepositivesforexternalinvasions, weseehandfulofhighervaluesforassassinationattempts, including17.8 % ofMcGowansfalsepositives.These discrepanciescomefromourjudgmentaboutthepurpose oftheassassinationattemptandthepeopleinvolved.We excludeattemptsifthereisnoevidenceof(1)anattempt totakeoverthegovernmentand(2)alargerconspiracy. McGowanislessconstrictiveinhiscodingdecisions. Powell&Thyne 253 Ourfinalstepwastocombthroughallinstanceswhere coupswerementionedinmajormediasourcesfrom1950 to2010,addingtheseeventstoourdatatoprovidethe mostcurrentlistpossible.Giventheexhaustivecoverage inourcandidatedataset,onlysevencompletelynewcases werecoded:Seychelles(1977),Russia(1991),Fiji(2000), Guinea,Guinea-BissauandMauritania(2008),Mada- gascarandHonduras(2009),andMadagascarandNiger (2010).Ourfinaldatasetyields457coupattemptsfrom 1950to2010,ofwhich227(49.7 % )weresuccessfuland 230(50.3 % )wereunsuccessful. Descriptivestatistics Coupattemptshappenedin94statesduringourtime period.Figure1presentstheinstancesofcoupsbystate, revealingthatwhilecoupsseemtobeclustered,fewareas havebeenimmune.Coupshavebeenmostcommonin AfricaandtheAmericas(36.5 % and31.9 % ,respec- tively).AsiaandtheMiddleEasthaveexperienced 13.1 % and15.8 % oftotalglobalcoups,respectively. Europehasexperiencedbyfarthefewestnumberofcoup attempts:2.6 % . Wealsoseesomeinterestingtrendsinthefrequency ofcoupattemptsovertime.AsshowninFigure2,there isafairlycleardeclineinthetotalfrequencyofcoup attemptsovertime. 6 Thehighpointforcoupattempts cameinthemid-1960s,followedbytwomorebubbles inthemid-1970sandtheearly1990s.Thenumberof 0 1-3 4-6 7-9 10+ Africa 169 coups (51.5% successful) Americas 145 coups (48.3% successful) Europe 12 coups (33.3% successful) Asia 59 coups (55.9% successful) Middle East 72 coups (45.8% successful) Figure1.Instancesofcoupattempts,1950to2010 Percent successful 0 5 10 15 0 20 40 60 80 100 Frequency 1950 1960 1970 1980 1990 2000 2010 year Total coupsSuccessful coups % Successful Figure2.Couptrendsovertime 6 ThedatainFigure2aredisplayedasa5-yearmovingaverage. Powell&Thyne 255 coupattemptswereovert(therehasbeenavisible movementtoclaimpower)andactual(theeventsare notalleged expostfacto insomekindoftrialproceeding). Tosummarize,ourdefinitionofacoupattempt includes illegalandovertattemptsbythemilitaryorother eliteswithinthestateapparatustounseatthesittingexecutive . ThisfinaldefinitioncoincidesmostcloselywithOKane (1987)andMarshall&Marshall(2007),whobothcode theworldacrossalargetimespan.Othersprovidebroader definitions,particularlyfortactics.Aswewillsee,weuse thisbreadthtoouradvantageincreatingourdataset. Successandfailure Wenowdifferentiatebetweenfailedandsuccessfulefforts. Manycoupattemptsarequicklyputdownbythegovern- ment,makingthemeasytocodeasfailures.Othersare muchmoreambiguous.Leadersofthe1991Sovietcoup attemptmanagedtodeposePresidentGorbachevforthree days,butitwouldbedifficulttocallthisasuccessful attemptoverall.ManyscholarshavefollowedThompson (1973)inconsideringacouptobesuccessfulifthepost- couprulingarrangementremainsinplaceforatleasta week(e.g.Jackman,1978;McGowan,2003).Ourdefini- tionremainsconsistentwiththisone-weekthreshold.A coupattemptisthusdefinedassuccessfuliftheperpetrators seizeandholdpowerforatleastsevendays . WhilealmostallscholarslistedinTableIdifferentiate successandfailureinasimilarmanner,theone-week thresholdissomewhatarbitrary.Atheoreticallystronger approachwouldbetofollowOKane(1987:37 38),who considersacouptobesuccessfulifitleadstotheinstalla- tioninpowerofagovernmentoftheconspiratorsown choosing.Thisdefinitionisproblematicbecauseit assumestheplottersultimatelyseekpower,whenthey maymerelybeattemptingtochangethepoliticalsystem orleadership.The2005coupinMauritania,forexample, ledtoelectionsinwhichcoupleadersbannedmembersof themilitaryfromrunningforoffice.Whilethemilitary wasinstrumentalinthedemiseoftheTayaregime,itdid notinstallagovernmentofitsownchoosing.Itisdifficult toclassifysuccessinthismannerwithoutknowingthe exactmotivationoftheconspirators somethingthatis almostimpossibletoevaluatebecausecoupleadersalmost inevitablycouchtheirgoalsintermsofdemocracypromo- tionandfreedom.Thus,weopttouseamoreempirically- verifiabletemporalrequirement. Codingprocedures Wedrawonthe14datasetslistedinTableIasastarting pointtobuildacandidatedataset.Threeofthese datasetsarelimitedgeographically,whilefourothersend priorto1980.Theselimitationsconstrictourabilityto understandcoupsacrosstheglobeandindifferenttime periods.Allothersarequiteremarkableinboththeirspa- tialandtemporalcoverage.However,thedifferencesin thecodingcriteriawouldmakeitproblematictosimply mergetheinformationandconsiderallobservationsas similarphenomena.Moreimportantly,adetailedexam- inationofmanyeventswithineachdatasetrevealspecu- liaritiesdeservingofmorecarefulconsideration.Coups areattimesconflatedwithcivilwars,mutinies,andassas- sinationattempts,forexample.Thedatasetsalsooften lackspecificdatesforcoupevents,whichlimitsthepre- cisionofanalyses.Severalauthorsalsocodeonlysuccess- fulcoups,whichlimitsourabilitytoanalyzethefactors thatmightcauseacouptofailonceattempted. Giventheseconsiderations,ourcodingproceduresfol- lowedtwomainsteps.First,wemergedtheinformation fromthe14sourcesintoasingledataset.Thisresulted in2,141potentialcoupeventsfrom1950to2006.Many eventsweremissinginformationonthemonthordayof thecoupattempt.Wewereabletomerge882ofthese eventswithsourcesthatincludedthisinformation.We thensearchedthroughdozensofsourcesforinformation abouteachoftheremaining1,259allegedcoups. 4 Twoor moresourcesagreedontheexactdateandoutcomeof 377coupevents,whichmadeitquiteeasytoverifythe eventwithhistoricaldocuments.Amongtheremaining 882questionableevents,wewereabletoverify73as codedcorrectly.Oursearchthroughthehistorical accountsoftheremaining809allegedcoupattempts revealedseveralcommondiscrepancies. Wesummarizethecodingdiscrepancies(falseposi- tives)inTableII.Thefirstcolumnshowstheentiresam- ple,whilethelatter14columnsexaminediscrepancies amongoursourcedatasets. 5 Webeginbynotingthe fairlyhighlevelofaccuracyacrossallsources(row3, rangingfrom67.7 % to88.3 % ).Whilemany 4 Themostcommonlyusedsourceswere TheHistoricalNewYork Times , Keesings ,and Lexis-Nexis .Thelattersourcewasparticularly usefulingivingusaccesstoforeignmedia.Commoninternational newsoutletsincluded Agence-FrancePresse , DeutschePresseAgentur , Xinhua ,andthe BBC .Inthedatasetwehavegivenpreferencetocit- ingthe NewYorkTimes asadefaultduetoitsaccessibility(77 % of sourceslistedcomefromthe NewYorkTimes ).Exceptionsinclude alternativesourcesthatbetterillustratetheeventsapplicabilitytoour definition.Thefulllistisavailableonlineinourcandidatedataset. 5 Weshouldbeclearthattheseeventsareonlyfalseorerrorsinthe sensethattheydonotcoincidewithourdefinitionofcoups.Mostof theexcludedeventsfitperfectlywellwiththeoriginalauthors intentions. 252 journalof P EACE R ESEARCH 48(2) successfulcoupshaslikewisedecreasedovertime.We saw12successfulcoupsinboth1963and1966.The mid-tolate-1970salsosawabriefburstofsuccessful coups(rangingfrom3to9foreachyear).Aninteresting trendemergeswhenwelookatthepercentageofcoup attemptsthatresultedinsuccessfulregimechanges, whichweplotontherightsideoftheY-axis.Themean successrateis48 % duringtheentiretimespan.Thisrate sawearlypeaksaround1970and1980,andthena declineuntiltheturnofthecentury.However,wesee anotherspikeinthesuccessratestartingin2003.Twelve ofthe18(67 % )coupattemptssincethenhavebeensuc- cessful,andonlyoneofthemostrecentfourcoup attemptshasfailed.Whilecoupshavecertainlywaned overtime,therecentsuccessofcoupplotterssuggests thatcoupsremainakeyelementofgovernmental instability. Pitfalls Afewwordsofcautionareneeded.First,ourobserva- tionsarerarelyoriginalinthesensethattheyhavenot beenpreviouslycoded.Thisisunsurprisinggiventhat wecomparedthehistoricalrecordto14earlierscholarly sourcesoncoups.Oureffortsutilityisnotsolely reflectedbythediscoveryofnewcases.Rather,itisin verifyingthatcodedeventscoincidewithanacceptable andconsistentdefinition.Indoingso,wehaveexcluded avarietyofeventsthatmightbeofinteresttoscholars. Thoseseekingtostudyotherformsofuncommonpower transferswouldbenefitfromutilizingdatathatarebuilt tospecificallyaddressthosetypesofevents.Scholars seekingtoexplainirregularpowertransfersingeneral, forexample,shouldfindtheArchigos(Goemans, Gleditsch&Chiozza,2009)orWorldHandbook (Taylor&Jodice,1983)datasetsparticularlyuseful. Banks(2001)providesdataforpopularprotestsand riots,whileGleditschetal.(2002)capturemoreviolent rebellions.Externalinvasionscanbecapturedwiththe MIDdataset(Jones,Bremer&Singer,1996),andthe PolityIV(Marshall&Jaggers,2000)datasetcaptures seizuresofpowermadebytheexecutivesthemselves. Second,despiteconsiderableeffortstoverifycoups andcoupattempts,reportingbiascouldhavelimitedour abilitytoidentifyevents.Thisisparticularlytruefor coupattemptsinstateswithasmallWesternmediapres- enceandeventsthathappenedintheearliesttimeperiod underinvestigation.Onewaytoanalyzethesepotential biasesistoexamineinstanceswherescholarsidentified coupswithintheoriginalcandidatedatasetsforwhich wecouldfindnoevidence(seeTableII,row4).This isusefulbecausetheoriginalsourceslikelyhadaccess toavarietyofnon-Englishsources,olderdocuments, orfirst-handknowledgeoftheeventstheycoded.Biases canbeidentifiedbyanalyzingwhetherthenon-verifiable casesvarysystematicallybyregionortime. AsshowninFigure3,thebreakdownbyregionshows someevidenceofbias,withthehighestvaluesinregions thatareleastapttobecoveredbyWesternmedia sources.However,across-tabulationbetweenregionand percentnoevidenceisstatisticallyinsignificant( w 2 ¼ 4.27,p.370),whichgivesusconfidencethatourreli- anceonWesternmediasourcesisnotprovidingsystema- ticbiasinourdataset.Thetemporalbreakdownsimilarly showslittleevidenceofbias.Asonewouldexpect,there arefewunverifiablecasesinthe2000s,butthenext smallestcategoryistheoldestinourdataset:the 1950s.Thecorrelationcoefficientbetweennoevidence andyearisalsoinsignificant(r ¼ 0.15,p.250),again suggestinglittletemporalbiasinsourcecoverage. Athirdpotentialpitfallliesnotinourdataset,butin waysinwhichscholarsarelikelytouseourdataset.Coup attemptsarefrequentlylumpedinwithmoreaggregate measuresofanti-regimeactivity,suchasBankss (2001)annualcountofriots,protests,etc.,and Uppsala/PRIOsarmedconflictdataset(Gleditsch etal.,2002),particularlyifdeathsexceedthe 25battle-deaththreshold.Scholarsbuildingempirical modelsshouldbeawareofpotentialcross-codingofthe sameeventsinordertoavoidcollinearityorendogeneity. Whilewehavegonetogreatlengthstoassurethatcoups arenotconflatedwithotherformsofanti-regimeactivity inourdataset,thereisnoguaranteethatscholarscoding othereventshaveexcludedcoupsintheirs.Forexample, wehaveidentified38eventswithintheUppsala/PRIO datasetand4eventsintheCorrelatesofWar(Sarkees, 2000)internalconflictdatasetthatarebestdescribed ascoups.Scholarswhousethesedatasetstooperationa- lizecivilwarsshouldbewaryofincludingtheseevents. 7 Conclusion Thepurposeofthisarticlewastoprovideasoundtheo- reticaloperationalizationofcoupsandtointroducethe newest,mostcomprehensivedatasetoncoupstodate. Weconcludeherebysuggestingtwobroadareasof researchthatmightfindthesedatauseful.Theburgeon- ingcivilwarliteratureisonefruitfulavenue.Scholars studyingcivilwaronsetmightbeinterestedinthe 7 Alistofthesecasescanbefoundinouronlineappendix. 256 journalof P EACE R ESEARCH 48(2) Theymighthavetheoppositeeffectbyremovinglong- standingauthoritarianleadersfrompower.Recentrum- blingsbythemilitaryinZimbabwe,forexample,suggest thatPresidentMugabeslong-awaitedexitislikelytobe theresultofacoup.Eachofthesepossibilitiesisimme- diatelyrelevanttopolicymakers,particularlygiventhe well-documentedeffortstofomentcoupsinforeign statesinthepast.USeffortsagainstCha ´ vezinVenezuela (2002)andHusseininIraq(1991),forexample,were basedontheviewthatcoupswerethemostpeacefuland directwaystobringaboutpositiveregimechanges (Thyne,2010).Policymakerswouldlikelybesimilarly interestedinknowingwhichoftheirfavoredleadershave ahighlikelihoodofbeingoverthrowninacoup,which couldbeforecastedwithourdata.Wehopethatthe datasetpreviewedinthisarticlewillprovidethetools necessarytotacklesuchinterestingresearchquestions. Replicationdata Replicationdataandappendicesareavailableathttp:// www.uky.edu/ * clthyn2orhttp://www.prio.no/jpr/ datasets/. Acknowledgments WewishtothankCameronThies,AaronBelkin,Patrick JMcGowan,MagnusO ¨ berg,andananonymous reviewerfortheirhelpfulcomments.Wearealso indebtedtoPatrickMcGowan,AaronBelkin,Evan Schofer,andErikaMorenoformakingtheirdataavail- abletous. References Agyeman-Duah,Baffour(1990)Militarycoups,regime change,andinterstateconflictsinWestAfrica. Armed Forces&Society 16(4):547 570. Alesina,Alberta;SuleOlzer,NourielBoubini&Phillip Swagel(1996)Politicalinstabilityandeconomic growth. JournalofEconomicGrowth 1:189 211. Banks,Arthur(2001)Cross-NationalTime-SeriesData Archive.Binghamton,NY:ComputerSystems Unlimited. Belkin,Aaron&EvanSchofer(2003)Towardastruc- turalunderstandingofcouprisk. JournalofConflict Resolution 47(5):594 620. Belkin,Aaron&EvanSchofer(2005)Couprisk, counterbalancing,andinternationalconflict. Security Studies 14(1):140 177. Decalo,Samuel(1990) CoupsandArmyRuleinAfrica: MotivationsandConstraints .NewHaven,CT:Yale UniversityPress. Dix,RobertH(1994)Militarycoupsandmilitaryrule inLatinAmerica. ArmedForces&Society 20(3): 439 456. Ferguson,Gregor(1987) Coupde ´ tat:APracticalMan- ual .NewYork:Sterling. Finer,Samuel(1988) TheManonHorseback:TheRoleof theMilitaryinPolitics ,2ndedn.Boulder,CO: Westview. Fossum,Egil(1967)Factorsinfluencingtheoccurrence ofmilitarycoupsde ´ tatinLatinAmerica. Journalof PeaceResearch 4(3):228 251. Galetovic,Alexander&RicardoSanhueza(2000)Citi- zens,autocrats,andplotters:Amodelandnewevi- denceoncoupsde ´ tat. EconomicsandPolitics 12(2): 183 204. Gleditsch,NilsPetter;PeterWallensteen,MikaelEriksson, MargaretaSollenberg&HåvardStrand(2002)Armed Conflict1946 2001:Anewdataset. JournalofPeace Research 39:615 637. Goemans,HenkE;KristianSkredeGleditsch& GiacomoChiozza(2009)IntroducingArchigos:A datasetofpoliticalleaders. JournalofPeaceResearch 46(2):269 283. Goodwin,Jeff(2001) NoOtherWayOut:StatesandRev- olutionaryMovements,1945 1991 .Cambridge:Cam- bridgeUniversityPress. Jackman,RobertW(1978)Thepredictabilityofcoups de ´ tat:AmodelwithAfricandata. AmericanPolitical ScienceReview 72:1262 1275. Janowitz,Morris(1977) MilitaryInstitutionsandCoercion intheDevelopingCountries .Chicago,IL:Universityof ChicagoPress. Jones,DanielM;StuartABremer&JDavidSinger (1996)Militarizedinterstatedisputes,1816 1992: Rationale,codingrules,andempiricalpatterns. Con- flictManagementandPeaceScience 15(2):163 215. Kebschull,HarveyG(1994)Operationjust-missed: Lessonsfromfailedcoupattempts. ArmedForcesand Society 20(4):565 579. Kennedy,Gavin(1974) TheMilitaryintheThirdWorld . NewYork:CharlesScribnersSons. Kposowa,AugustineJ&JCraigJenkins(1993)The structuralsourcesofmilitarycoupsinpostcolonial Africa,1957 1984. AmericanJournalofSociology 99(1):126 163. Londregan,JohnB&KeithTPoole(1990)Poverty,the couptrap,andtheseizureofexecutivepower. World Politics 42(2):151 183. 258 journalof P EACE R ESEARCH 48(2) impactofcoups,perhapsexpectingthemtosparkcivil conflictsbyweakeningthestate,astheyseemedtoin Afghanistan(1978)andNepal(1960).Coupsmightalso haveaninterestingimpactforstudiesofcivilwardura- tionandoutcome.Theymightlengthencivilconflicts byinstigatingfurthergovernmentalinstability,asevi- dencedbythe1989coupinSudan.Theymightalterna- tivelyshortencivilconflictsbybringingmore conciliatoryleaderstopower.Pinillasousterof ColombiandictatorGo ´ mezin1953,forexample,led tocivilpeaceandapower-sharingagreementbetween liberalandconservativeparties.Methodologically,scho- larsstudyingthistopicshouldconsidereitherexcluding orcontrollingforcoupsintheirempiricalmodels. Coupsshouldalsobeofinteresttothosestudying democracy.Coupsarelikelytoderailtheprocessof democratizationand/ordemocraticconsolidation,as theyseemtohaverecentlyinThailandandMauritania. 0 10 20 30 Percent by region AfricaAmericasAsiaEuropeMiddle East 0 10 20 30 Percent by year 1950 1960 1970 1980 1990 2000 Note: Y axes = frequency of events coded as no evidence within each region or year total events coded within each region of year Figure3.Coupeventscodedasnoevidence Powell&Thyne 257 summarizing,inTableI,how14studiescapturecoups andthenwalkthroughcommontrendsandambiguities. Ourgoalistoarriveatafinaldefinitionthatbestcap- turescoupsasdistinctfromotherpoliticalphenomena. Targets Thefirstfactorinarrivingatadefinitionisindeciding who maybetargeted.Wefindlittledebateonthissubject, withmostscholarsfocusingonthestatesprimaryleader. Weremainconsistentwithpreviousresearchbyconsider- ingonlyattemptstooverthrowthe chiefexecutive .This ruleseparatescoupsfromlessextrememechanismsof pressuringtheleadership.Forexample,ahandfulof Nepalesepolicerecentlyrevoltedagainstitsofficersover inferiorfoodrations.Thiseventisbestdescribedasa mutinybecausePrimeMinisterKoiralawasneverdirectly challenged.Likewise,militarypressureforcedtheresigna- tionofArgentinePresidentFrondiziscabinetin1959. However,thiseventdoesnotmeritclassificationasacoup becausethechiefexecutiveremainedinoffice. Perpetrators Astartingpointformanystudies,Finer(1988:23)limits perpetratorstothearmedforces.Thisrathernarrow focusisechoedbysixothersourcesinTableI.Others morebroadlyallownon-militaryelites,civiliangroups, andevenmercenariestobeincludedascoupperpetra- tors.Thisbroaddefinitionincludesfoursources,includ- ingJanowitz(1977:49),whoclaimsthatperpetrators needonlybeorganizedfactions.Wetakeamiddle ground.Coupsmaybeundertakenby anyelitewhois partofthestateapparatus .Thesecaninclude TableI.Coverageanddefinitionsofcoupsinpreviousresearch SourceYearscovered States codedTargetPerpetratorTactics Banks(2001:13 14)1946 1999190topgovernment elite ... nations powerstructure undefined,includes revolutions forcedchanges Belkin&Schofer (2005:144) 1945 2000167regimesmallmilitarycoalitionsundefined Ferguson (1987:13 14) 1945 1986164government, establishment soldiers,politicians, mercenaries undefined Finer(1988:23)1958 1974144civilianauthoritiesarmedforcessanction(orthreatof) Janowitz(1977:49)1946 1975151existingregimeorganizedfactionsundefined Kennedy(1974:14)1945 1972142undefinedregulararmedtroopsuseorthreatofuseof armedforces Lunde(1991:18)1955 198447regimemilitaryorsecurityforcesundefined Luttwak(1969:12)1945 1967132governmentsegmentofthestate apparatus conspiracy Marshall&Marshall (2007:1) 1960 2006199executiveauthority andoffice rulingorpoliticalelitesforcefulseizure McGowan(2007:2)1958 200452national government military,security,and/or police violence ... maybe negligible Morenoetal (2004:2) 1950 200019national government militaryleader;executive him/herself militaryforce (orthreatof) OKane (1987:22,37) 1950 1985163governmentcivilormilitarythreatoruseof violence Taylor&Jodice(1983)1948 1982102chiefexecutivegroups,cliques,cabals,parties andfactions threatenedoractual coercion Thompson (1973:6,52) 1946 1970135chiefexecutiveregulararmedforcesuseorthreatofforce 250 journalof P EACE R ESEARCH 48(2) Lunde,TormodK(1991)Modernizationandpolitical instability:Coupsde ´ tatinAfrica,1955 1985. Acta Sociologica 34(1):13 32. Luttwak,Edward(1969) Coupde ´ tat:APracticalHand- book .NewYork:AlfredAKnopf. Marshall,Monty&KeithJaggers(2000) PolityIVData- setandUsersManual:PoliticalRegimeCharacteristics andTransitions,1800 1999 (http://www.bsos.umd. edu/cidcm/polity). Marshall,Monty&DonnaMarshall(2007)Coupde ´ tat events,1960 2006.CenterforSystemicPeace. McGowan,PatrickJ(2003)Africanmilitarycoupsde ´ tat, 1956 2001:Frequency,trendsanddistribution. JournalofModernAfricanStudies 41(3):339 370. Moreno,Erika;MichaelLewis-Beck&JacqueAmoureux (2004)Latinrhythms:CoupcyclesintheAmericas. PaperpresentedattheMidwestPoliticalScience Association,Chicago,IL. Morrison,DonaldG&HughMStevenson(1972) Integrationandinstability:PatternsofAfricanpoliti- caldevelopment. AmericanPoliticalScienceReview 66(3):902 927. OKane,RosemaryHT(1987) TheLikelihoodofCoups . Aldeshot:Avebury. Sarkees,MeredithR(2000)TheCorrelatesofWardata onwar:Anupdateto1997. ConflictManagementand PeaceScience 18(1):123 144. Small,Melvin&JDavidSinger(1982) ResorttoArms: InternationalandCivilWars,1816 1980 .Beverly Hills,CA:Sage. Taylor,CharlesLewis&DavidAJodice(1983) World HandbookofPoliticalandSocialIndicatorsIII, 1948 1982.Part2:DailyPoliticalEventsData .Ann Arbor,MI:Inter-UniversityConsortiumforPolitical andSocialResearch. Thompson,WilliamR(1973) GrievancesofMilitary Coup-Makers .BeverlyHills,CA:Sage. Thyne,ClaytonL(2010)Supporterofstabilityoragent ofagitation?TheeffectofUnitedStatesforeignpolicy oncoupsinLatinAmerica,1960 1999. Journalof PeaceResearch 47(4):449 461. Welch,Claude(1970) SoldiersandStateinAfrica . Evanston,IL:NorthwesternUniversityPress. JONATHANMPOWELL,b.1982,PhDcandidate inPoliticalScience,UniversityofKentucky(2006 ); researchinterestsincludeconflictprocesses,regime transitions,andcivil militaryrelations. CLAYTONLTHYNE,b.1978,PhDinPolitical Science(UniversityofIowa,2007);AssistantProfessor, DepartmentofPoliticalScience,UniversityofKen- tucky(2007 );researchinterestsincludedomestic conflict/instability,regimetypes,andinternational educationpolicy.Mostrecentbook: HowInternational RelationsAffectCivilConflict:CheapSignals,Costly Consequences (LexingtonBooks,2009). Powell&Thyne 259