duringaboomigovernmentspendingasashareofGDPshouldgodownbecauseofautoma - PDF document

1 duringaboom:(i)governmentspendingasashar
duringaboom:(i)governmentspendingasashareofGDPshouldgodownbecauseofautomaticstabilizers(ifdiscretionaryspendingremainedconstantinrealterms,theeectwouldbereinforced);(ii)withconstanttaxratesandsomedegreeofprogressivity,governmentrevenuesasashareofGDPshouldgoup(theeectwouldbereinforcedbytaxcutsinrecessionsandtaxin-creasesinboom.);(iii)asaresult,budgetsurplusesasashareofGDPshouldincrease.Theoppositeshouldoccurinrecessions.Inpractice,inmanydevelopingcountriesscalpolicyhastheoppositeproperties:itisprocyclical.Inparticular,governmentspendingasashareofGDPgoesupduringboomsanddowninrecessions,whiledecitsincreaseinboomsanddecreaseinrecessions.InOECDcountries,instead,scalpolicyisgenerallycounter-cyclical.GavinandPerotti(1997)werethersttopointoutthatinLatinAmericascalpolicyisprocyclical,butTalviandVegh(2005),CataoandSutton(2002)andKaminski,Reinhart,andVegh(2004)notedthatthisisnotaLatinAmericanphenomenononly:procyclicalityofscalpolicyiscommoninmany—thoughnotall—developingcountries.Whydomanycountriesfollowseeminglysuboptimalprocyclicalscalpoliciesthataddtomacroeconomicinstability?Acommonanswerhastodowiththesupplyofcredit.Inbadtimesmanydevelopingcountriescannotborrow,orcandosoonlyatveryhighinterestrates,thereforetheycannotrundecitsandhavetocutspending;inboomstheycanborrowmoreeasilyandchoosetodoso,increasingpublicspending(cf.GavinandPerotti1997,CataoandSutton2001andKaminski,Reinha

2 rt,andVegh2004).Thisargumentisincomplete
rt,andVegh2004).Thisargumentisincomplete,however,sinceitbegstwocriticalques-tions.First,whydon’tthesecountriesself-insurebyaccumulatingreservesingoodtimes,sothattheyarelesslikelytofacebindingcreditconstraintsinrecessions?Second,whywouldlendersnotprovidefundstocountrieseveninrecessions,iftheywereconvincedthattheborrowingwouldoptimally InlightofthecarefuldiscussionofKaminski,ReinhartandVegh(2004)wewanttobeclearregardingourchoiceofwords.Wedeneascounter-cyclicalapolicythatfollowsthetaxsmoothingprincipleofholdingconstanttaxratesanddiscretionarygovernmentspendingasafractionofGDPoverthecycle.Theydenesuchpolicyas“acyclical”.BothweandtheywoulddeneasprocyclicalapolicyinwhichtaxratesgodowninboomsandupinrecessionsandspendingoverGDPgoesupinbooms.Asthoseauthorsthemselvesnote,ourdenitionisthemostcommonintheliterature.Somecountriesbelongingtobothgroupshaveaccumulatedlargeamountsofpublicdebt.Forareviewofmodelsthatexplainexcessivedecits,seeAlesinaandPerotti(1995)andPerssonandTabellini(2000).OnthecyclicalpropertyofscalpolicyinOECDcountriesseePerotti(2004) smoothoutthecycle?Toanswerbothquestionsoneneedstoconsiderthepoliticalarena,andthisiswhatwedointhispaper.Wearguethatprocyclicalandmyopicscalpolicystemsfromapoliticalagencyproblem.Votersfacecorruptgov-ernmentsthatcanappropriatepartoftaxrevenuesforunproductivepublicconsumption,i.e.politicalrents.Rentscanbethoughtofasdirectappro-priatio

3 n(stealing)oftaxrevenuesbygovernmentocia
n(stealing)oftaxrevenuesbygovernmentocials,butalsofavorspaidtospecialinterestssuchaspublicemployeesor“friends”ofthegovern-ment,oftenidentiedalongethnic,orreligiouslinesetc.Voterscanreplaceagovernmentthatabusesofhispowers,butinequilibriumtheygenerallycannotpushrentsallthewaytozero.Thisagencyprobleminteractswithlackofinformation:votersobservethestateoftheeconomy,buttheycannotobservegovernmentborrowing,atleastnotatthemargin;forinstance,thegovernmentcanaccumulatehiddeno-balance-sheetliabilities.Hence,whenvotersseetheeconomybooming,theydemandhigherutilityforthemselves(intheformoflowertaxes,orbetterpublicgoods),inawaythatresem-blesthe“starvetheLeviathan”argument.Thisforcesthegovernmenttoimpartaprocyclicalbiastoscalpolicy,andtoborrowtoomuch.Thus,procyclicalandmyopicscalpolicy(i.e.anincreaseingovernmentspend-ingduringboomsandexcessivegovernmentborrowing)arisesfromvoters’demands—notethatvotersdonotdemandirrationalpolicies,theyarejustpoorlyinformedabouteconomicpolicy.Theydemandandobtain,throughareelectionconstraintonthegovernment,asecond-bestsolutiontoanagencyprobleminanenvironmentofcorruptionandimperfectinformation.For-mally,themodelextendstoadynamicenvironmentwithpublicdebtamodelofmoralhazardandpoliticalaccountabilityoriginallyformulatedbyBarro(1973)andFerejohn(1986),andadaptedtopublicnancebyPerssonandTabellini(2000).Wethentakethetheoreticalpredictionstothedata.First,wecon

4 previousevidenceonthewidespreadprocyclic
previousevidenceonthewidespreadprocyclicalityofscalpolicy.Thisismainlyduetogovernmentspending,thatgoesupinboomsanddowninrecessionasashareofGDP.Second,weshowastrongpositivecorrelationbetweenprocyclicalityandmeasuresofcorruption:morecorruptcountriesdisplayamoreprocyclicalscalpolicy.Third,thecorrelationbetweencor-ruptionandprocyclicalityisonlyormainlypresentindemocracies,conrm-ingthetheoreticalideathatprocyclicalityemergesbecausevoterstrytoholdcorruptgovernmentsaccountable.Finally,weaskhowrobustisthecorrela-tionbetweencorruptionandprocyclicalitywhenalsotakingintoaccountthe evidenceonborrowingconstraints.Thisisnoteasy,becausemorecorruptgovernmentsmightalsofacemorebindingcreditconstraints.Asaresult,manyofthesamevariablesthatinuencepoliticalcorruptionarealsolikelytoaecttheseverityofborrowingconstraints—indeed,corruptionishighlycorrelatedwithcreditratingsinthedata.Nevertheless,wepresentsomesuggestiveevidencethatpoliticalagencyproblemsindemocracies,ratherthancreditmarketimperfections,aretheunderlyingcauseofprocyclicalcalpolicy.Ourmainconclusionisthatprocyclicalityofscalpolicyresultsfromagovernmentfailure,notamarketfailure,andtakesplaceirrespectiveofwhetherornotthegovernmentisupagainstacreditlimit.Fiscalpolicyisprocyclicalbecauserationalbutuninformedvoters“starvetheLeviathan”anddemandmoreingoodtimesthaninbadtimes.Wearenotthersttosuggestapoliticalexplanationtotheprocyclicali

5 tyscalpolicy.InTalviandVegh(2005),thepre
tyscalpolicy.InTalviandVegh(2005),thepresenceofsurplusesincreasesthegovernmentpropensitytospend;thisdistortionisassumed,however,ratherthanderivedfromanexplicitpoliticalmodel.Analternativepo-liticalexplanationisthe“voracityeect”ofTornellandLane(1999)andLaneandTornell(1998):whenmoreresourcesareavailable(i.e.inbooms),thecommonpoolproblemismoresevereandtheghtovercommonre-sourcesintensies,leadingtobudgetdecits.Butwedonotknowofotherpapersthatlinkscalmyopiaandprocyclicalitytoapoliticalagencyprob-lem.Finally,theideathatvotersinducedebtaccumulationtodisciplinegovernmentsthattheydonotrustisrelatedtoJensenandMeckling(1976).Thatseminalcontributionshowsthatdebtnancing(asopposedtoexternalequitynancing)canmitigatetheagencyprobleminsidetherm;butofcourse,themechanismthroughwhichthishappensinourpoliticalcontextisdierent.Ourideathatpoliticalagencycanleadtoexcessivedebtaccumulationwhenvotersareuninformeddiersfromtwootherpoliticalmodelsofgovern-mentborrowingintheliterature.Thestrategicdebtargument(AlesinaandTabellini(1990),PerssonandSvensson(1989),TabelliniandAlesina(1990))doesnotrelyonanagencyproblem:votersarenotuninformedaboutscalpolicyandtheresultsaredrivenbydierentpreferencesamongstpoliticalpartiesorgroupsofvoters.Intherationalbudgetcyclesliterature(Ro- SatyanathandSubramanian(2004)showempiricalevidencethatdemocraticfailureexplainsmacroeconomicinstability;buttheyfocusonthedistincti

6 onbetweendemocra-ciesvsnon-democracies,w
onbetweendemocra-ciesvsnon-democracies,whereaswearguethatprocyclicalscalpolicystemsfromtheinteractionofdemocraticaccountabilityandpoliticalcorruption. andSiebert(1989)andRogo(1990)),votersfaceanadverseselectionproblemandthisleadstodistortedscalpolicybeforetheelection.Theassumptionaboutvoters’informationissimilartoours,butheretheincen-tiveproblemisoneofmoralhazard,notadverseselection.Moreover,thosepapersdonotdiscussthereactionofeconomicpolicytoexternalshocks,nordotheyallowforastatevariablelikegovernmentdebt.Thepaperisorganizedasfollows.Insection2welayoutthemodel.Insection3wederivetheeconomicandpoliticalequilibrium.Section4discussestheempiricalevidence;thelastsectionconcludes.2TheModel2.1TheeconomyConsiderasmallopeneconomywithaninnitehorizon.Theprivatesec-torconsistsofarepresentativeconsumerthatmaximizesthepresenteddis-countedvalueofexpectedutilityfromprivateconsumption:(1)wheredenotesconsumptioninperiodt,Eistheexpectationsoperator,isasmoothandstrictlyconcaveincreasingfunction.Forsimplicity,weneglecttheeconomicchoicesoftheprivatesector,andonlyfocusonitspoliticalroleofcontrollingthegovernmentagencyproblem.Thus,weassumethatprivateconsumptionineachperiodisjustgivenbyendowmentincome()netoftaxes(Themodelismeaningfulonlyifgovernmentdebtisnon-neutralandthisisthesimplestwaytogetthatproperty.Incomeisani.i.d.randomvariable,drawneachperiodfromadistributionwithboundedsupportover

7 .Allvariablesareexpressedinpercapitater
.Allvariablesareexpressedinpercapitaterms.Governmentspendinginperiodonlytakestheformof(non-negative)rents,,thatbenetthegovernmentbutnottheconsumer.Addingpro-ductivegovernmentspendingwouldcomplicatethenotationwithoutchang-ingtheresults:forallpracticalpurposes,wecanthinkofastheutility Alternativebutmorecomplicatedassumptionswouldbetoallowtheconsumertoborroworlendinaneconomywithtaxdistortions,ortomodelexplicitlyaliquidityconstraintonprivateconsumption. accruingtotheprivatesectorfromprivateandpublicconsumption,andofastheadditionaltaxesneededtopayforunproductivespendingthatbenethegovernmentbutnottheconsumersatlarge.Inperiodthegovernmentcanissuepublicdebt,atamarketpriceGovernmentdebtisboughtbyforeignresidentsandthereisfullrepaymentofdebtnextperiod.Thus,exploitingthegovernmentbudgetconstraint,wecanwriteprivateconsumptioninperiodas:(2)Weassumethatthereisalimittohowmuchresourcesagovernmentcanappropriateforhisownexclusivebene.Theupperbounddenoteswhatthegovernmentcanstealfromthepubliccoerswithoutendingupinjail.Weconsidertwoalternativeassumptionsaboutthesimplestcase,itisaconstant:Alternatively,weassumethattheupperboundonrentsisadecreasingandconcavefunctionofpublicdebtoutstanding:with.Thus,ifthepreviouslegislatureaccumulatedalargeamountofgovernmentdebt,thereislessroomtostealtoday.Asdiscussedinthenextsubsection,debtisonlyobservedbythepublicatlargeinthesubsequentperiod,whenithas

8 toberepaid.Thus,thissecondassumptionsays
toberepaid.Thus,thissecondassumptionsaysthat,ifthegovernmentaccumulatedlargeliabilitiesinthepreviouslegislature,itisundermorecarefulscrutinytoday,bothfromthedomesticvotersandinternationalorganizations,andasaresulttheupperboundonrentsismoresevere.Asweshallsee,theassumptionthatthereisanupperboundonrentsplaysaroleevenifthisconstraintisnotbindinginequilibrium,becauseitdeterminesthestrengthofout-of-equilibriumthreats.Butthepolicyresponsetoincomeshocksisthesameirrespectiveofwhethertheupperboundisaconstantorafunctionofdebtoutstanding.Finally,weassumethatgovernmentdebtcanbeissuedonlyuptoamaximumamountUptothisamount,debtisalwaysrepaidinfullandthereisnodefaultrisknoranycreditmarketimperfection.Thisupperlimitongovernmentdebtislowenough(comparedtothepossiblerealizationsofpercapitaincome),sothatthenon-negativityconstraintsonconsumption Theassumptionofasmallopeneconomyisappropriateforourempiricalwork,inwhichweconsiderthiskindofcountries.Withoutdefaultriskthereisnoriskpremium,butinourempiricalanalysisweallowfortheeectsofriskpremiaongovernment-issued andrentsarenotviolatedinequilibrium:(3)(4)Therstinequalityguaranteesthat,eveniftheupperboundonrentsisaconstantandincomeislowest,outstandingdebtcanalwaysberepaidinfullwithoutpushingprivateconsumptiontozero.Thesecondinequalityimpliesthereisalwayssomethingtosteal,eveniftheupperboundonrentsisadecreasingfunctionofdebt,anddebtismaximalThese

9 assumptionsplaynorole,otherthantomakesur
assumptionsplaynorole,otherthantomakesurethattheequilibriumdoesnotviolatesomenon-negativityconstraints.Inthissimpleenvironment,theoptimalpolicyforthevotersisofcourseandadebtpolicythatsmoothsprivateconsumptioninthefaceofincomeshocks.Suchapolicywouldtypicallyinducetheconsumertospendonlyafractionofanyincomeincrease,sincebyassumptionincomeisi.i.d.andhenceincomeincreasesaretemporary.Conversely,negativeincomeshockswouldleadtosomedebtaccumulationandconsumptionwouldfalllessthanoneforonewithincome.2.2ThepoliticalsystemElectionsareheldattheendofeachperiod.Theincumbentgovernmentonlycaresaboutgrabbingrentsforhimself.Thus,hemaximizes:whereitisunderstoodthathecangetrentsonlywhileinoce(ifthein-cumbentisnotreappointed,thenfuturepoliticalrentswillbeenjoyedbyanotherpoliticianinoce).Theutilityfunctionissmooth,increasingandstrictlyconcave.ThepoliticalenvironmentisadaptedfromBarro(1973),Ferejohn(1986)andPerssonandTabellini(2000,ch.4).Specically,governmentpolicyischosenaftertheelections,bytheincumbent,andthereisnocommitmenttoelectoralpromises.Thus,thereisanelementof“contractincompleteness”inthepoliticalenvironment,andthegovernmentcanonlybeheldaccountableex-postthroughbackward-lookingvotingstrategies.Thisaccountabilityis madepossiblebyassumingthat,ateachelection,theincumbentischallengedbyanidenticalopponent,whoseroleistoprovideanalternative.Voterschoosetheoptimalvotingstrategythatminimiz

10 estheirlossofwelfarefromthisagencyproble
estheirlossofwelfarefromthisagencyproblem.RelativetothemodelsofFerejohn(1986)andPerssonandTabellini(2000),wehaveaddedgovernmentdebt.Thismakesthemodeltrulydynamic,whilethepreviousliteratureonpoliticalagencyhadstaticeconomicenvironments.Votersobservetheirowncurrentutilityaswellastheirincome(andhencehowmuchtheyarepayingintaxes).Buttheydonotobservehowmuchgov-ernmentdebtisbeingaccumulatedinthecurrentperiod.Thisisequivalenttosayingthatthegovernmentcanincuro-balance-sheetliabilitieswithwhichtopayforrents.Thesizeoftheseliabilitiesonlybecomesknowntothevotersaftertheelections.Thisassumptionisconsistentwiththevastlit-eraturethathasemphasizedthesizeandsignicanceofcreativeaccountingandlackoftransparencyofthebudgetespeciallyindevelopingcountries;ithasthesameavoroftheinformationassumptionsoftheliteratureorratio-nalpoliticalbusinessandbudgetcycles.Notehowanexpost“discovery”oflargegovernmentliabilitiesmaytriggermorecontroloverthegovernmentandthereforemakeitmoredicultforthegovernmenttoappropriaterentsinthefuture,whichisoneofthecasesweexaminebelow.Thus,thesequenceofeventsisasfollows:(i)Atthestartofeachperiod,beforegovernmentpolicyischosen,votersobservetheirincomebeforetaxesinthecurrentperiods,anddebtoutstanding,;theyselectareservationlevelofprivateconsumption,,andpromisere-electiontotheincumbentconditionalonattainingatleastthatlevelofconsumption.(ii)Thegov-ernmentobservesthereservati

11 onleveldemandedbyvotersaswellastheircurr
onleveldemandedbyvotersaswellastheircurrentincome,andsetspolicyforthecurrentperiod,namelyrents()andgovernmentdebt().(iii)Votersobservetheirutilityfromprivatecon-sumptionandvoteaccordingtotheirpromiseThissequenceofeventsis SeeinparticularVonHagenandHarden(1994),AlesinaandPerotti(1995),MilesiFerretti(2003),amongstothers,ontheroleoflackoftransparencyinthebudgetprocess,andRogo(1990),RogoandSiebert(1990)andPerssonandTabellini(2000)forrationalpoliticalcycles.Notethatthereisanasymmetry:whilevotersdonotobserveuntilperiodforeignlendersdonotlendtothegovernmentpastthepoint;henceinternationalnancialmarketshavebetterinformationaboutthedebtpolicycomparedtonationalvoters.Qualitatively,thisassumptionisnotimplausible,althoughhereforsimplictyitisformulatedinaverystarkform:votersaretotallyignorant,whileforeigninvestorsareperfectlyinformed.Asweshallsee,however,theupperboundplaysnoroleinthecaseinwhichtheceilingonrentsisafunctionofdebtoutstanding: repeatedineachperiod.Anequilibriumisareservationlevelofprivateconsumptionthatisop-timalforthevotersinthecurrentperiod,giventheinitialconditionsandtakingintoaccountsubsequentequilibriumoutcomes,andapolicythatisoptimalforthegovernment,giventhevotingstrategyandsubsequentequi-libriumoutcomes.Notethatthisdenitionofsequentialequilibriumrulesoutpre-commitmentbythevoterstoasequenceofvotingrules.Voterscanpunishthegovernmentforbadbehaviorduringthecurren

12 tlegislature.Butwedonotallowvoterstopuni
tlegislature.Butwedonotallowvoterstopunishthegovernmentforthepolicychosenbeforethepreviouselection,oncetheydiscoverhowmuchpublicdebtwasaccumu-latedduringthepreviouslegislature.Inotherwords,werestrictattentiontoMarkov-perfectequilibria.Sincethegovernmentisfullyinformedandthereisnoasymmetricinformation,rationalvoterscanfullypredictgovernmentpolicy,eveniftheydonotobserveit.Hence,inequilibriumnogovernmentchangeoccursandtheincumbentisalwaysre-elected,althoughthethreatofoutofequilibriumeventsisamajordeterminantofthevoters’andofthegovernment’sdecisions.3EquilibriumpoliciesInthissection,wecharacterizetheequilibriumandthenwediscussitsprop-erties.Sinceitissimpler,westartwiththecaseinwhichtheupperboundonrentsisaconstantirrespectiveofpublicdebtoutstanding:.Theappendixdescribestheequilibriumunderthealternativeassumptionthat.Supposethattheincumbentchoosestoforgore-election.Inthiscase,hewillcertainlygrabasmanyrentsaspossible,andobtainutilityNext,supposethattheincumbentgovernmentseekstopleasethevoters.Letb,|,cbetheincumbent’smaximalutilityinthiscase,givencurrentincome,debtoutstanding,andvotersdemands.Letainfrontofavariabledenotenextperiodvalues.Thenb,|,cisdenedby:b,|,cMa∆r,by:v(r)+βEV(b0,|0)](5) Underourassumptions,governmentdebtpolicyinthisout-of-equilibriumoutcomeisnotwelldened(inthesensethatthegovernmentisindierentabout).Butwedon’tneedtospecifytheout-of-equilibriumdebttodete

13 rminetheequilibriumoutcome,soweleaveitat
rminetheequilibriumoutcome,soweleaveitatthat. subjecttothebudgetconstraint,andtotheupperboundsonrentsandgovernmentdebt.Thefunctionistheequilibriumvalueofreappointmentfortheincumbent,inthefuturestateTheincumbentcanalwayschoosetoforegore-election.Hence,voterscannotpushgovernmentutilitybelowthethreshold(whathecanachievebygrabbingmaximalrentsonce).Inotherwords,foranyvaluesofvoters’demandshavetosatisfythefollowingincentiveconstraint:b,|,c(6)Clearly,itisoptimalforthevoterstodemandprivateconsumptionuptothepointwhere(6)holdsasequality.Notdoingthatwouldsimplyenablethegovernmenttograbmorerentsforitself,withoutincreasingvoters’utilityincurrentandfutureperiods.Hence,equilibriumdemandsbythevoters,areafunctionb,|,denedimplicitlybythecondition:b,|,c(7)Wecanthendenetheequilibriumvalueofreappointment,namelythefunctionb,|introducedabove,as:b,|b,|,c(8)wherethelastequalityfollowsby(7).Since(8)mustholdforanyvaluesofitmustalsoholdinallfutureperiods.Thus,forallpossiblevaluesofBasedon(5)and(8),equilibriumrentsinthecurrentperiodarethenimplicitlydenedbythefollowingcondition:(9)Thelefthandsideof(9)istheincumbent’sutilityifhepleasesthevoters,giventheequilibriumcontinuationvalueofbeingreappointedtomorrow.Therighthandsideishisutilityifhestealsasmuchaspossibletoday,butisthenthrownoutofoce.Inequilibrium,theincumbentmustbeindierentbetweenthesetwooptions.Equation(9)canbeeasilysolvedtoobtainequilibr

14 iumrents: Implictly,weareassumingthatequ
iumrents: Implictly,weareassumingthatequilibriumrentsarestrictlypositive. Itremainstodetermineoptimalpublicdebt,,foragovernmentseekingre-appointment.Usingthepreviousnotation,thisisthesolutiontothefollowingoptimizationproblem:Ma∆:(|−c∗+βb0−b)+βEV(b0,|0)](11)subjecttoTheexpressioninsidetheroundbracketscorrespondstorentsinthecurrentperiod,giventhevotersequilibriumrequest,Thelasttermistheexpectedequilibriumcontinuationvalue(i.e.whatthegovern-mentexpectstogetfromnextperiodonwardsifheisre-appointed).Bytheargumentabove,foranyvalueof.Thismeansthat,fromtheperspectiveofagovernmentseekingreappointment,issuingpublicdebtinthecurrentperiodentailsnofuturecosts.Thecostsarefullybornebytheconsumers.Butbyassumption,consumersdonotobservegovernmentdebt.Hence,theincumbentcanpockettheproceedsfromissuinggovernmentdebtintheformofhigherrents.Indeed,theoptimaldebtpolicythatsolves(11)istoalwaysborrowasmuchaspossible:.Equilibriumprivateconsumptionistheneasilyobtainedbyinsertingalltheseresultsinthebudgetconstraint,(2).Wesummarizeallthisinthefollowing:Proposition1Supposethattheupperboundonrentsisaconstant,Thentheequilibriumstochasticsteadystatehas::v(¯r)(1−β)]b∗=¯bThesteadystateisreachedafteroneperiod.Theappendixsolvesthecaseinwhichtheupperboundonrentsdependsondebtoutstanding:.Thesolutionprocedureisverysimilar,theonlydierencebeingthatissuinggovernmentdebtnowiscostlyforthegovernmen

15 t,becauseitreducesthevalueofitsout-of-eq
t,becauseitreducesthevalueofitsout-of-equilibriumthreatnextperiod.Asaresult,equilibriumdebtisnowataninterioroptimumlowerthantheupperboundandthesteadystateisreachedgraduallyratherthanatonce.Butthroughouttheadjustmenttothesteadystate,andoncethesteadystateisreached,consumptionmovesoneforonewithincome.Equilibriumdebtandrentsarenotaectedbyincomeshocks.Morespecically,theappendixproves: Proposition2Supposethattheupperboundonrentsdependsondebtout-standing,.Thentheequilibriumstochasticsteadystatehas:has:v(R(b∗))(1−β)]andsteadystatedebtisataninterioroptimumUndertheconditionsstatedintheappendix,thesteadystateislocallystable.Duringtheadjust-menttothesteadystate,incomeshocksonlyaectconsumption,thatchangesoneforonewithincome.3.1DiscussionAccordingtoPropositions1and2,incomeshockschangeconsumptiononeforone:positiveincomeshocksarenotsavedthroughthegovernmentbudgettobringabouthigherutilityfortomorrow;andnegativeincomeshocksdonotleadtomoregovernmentborrowing.Thishappensirrespectiveofwhetherthegovernmentisupagainstitsdebtceilingornot.Thisfailuretosmoothincomeshockswithscalpolicyisduetoanagencyproblem,nottoacreditmarketimperfection.Theintuitionisstraightforward:consumersdonotobservedebtaccumulation.Theyalsoknowthattheycannottrustthegovernment.Thus,whentheyseebettermacroeconomicconditions,theydemandhigherutilityforthemselves.Iftheydidnotdothat,thegovernmentwouldsimplyappropriatemor

16 erents,andtheywouldnotreceiveanyhigherco
erents,andtheywouldnotreceiveanyhigherconsumptioninthefutureanyway.Theconversehappenswhenincomeisseentogodown.Tokeepthemodelsimple,herethegovernmentcanonlydeliverutilitytothevotersthroughtaxcuts.Thus,literallythemodelpredictsthatthetaxrevenuetoGDPratiogoesdowninboomsandupinrecessions.Butasalreadynoted,itwouldbeeasytoaddproductivegovernmentspendingtothemodel(thatalsoyieldsutilitytothevoters).Insuchacase,duringaneconomicboomvoterswoulddemandbothhigherprivateandpublicconsumption.Dependingonthespecicassumptions,eithergovernmentspendingwouldgouportaxrateswouldgodownwhenincomeincreases,orboth.Buteitherway,voterswoulddemandimmediatewelfaregains,andwouldnottrustthegovernmenttosaveresourcesforthefuture.Finally,notethatinthemodelaswehaveitwritten,thedegreeof”cor-ruption”isazero-onevariable:eitherthegovernmentcanappropriaterents, inwhichcasescalpolicyisprocyclical,oritcannot,inwhichcasescalpolicyisoptimal.Thisstarkcontrastcomesformthestrongassumptionsongovernmentpreferences.Wealsosolvedatwo-periodversionofthismodelwitharelativeweightcapturinghowmuchthegovernmentcaresaboutrentsrelativetoconsumerwelfare.Insuchamodel,thedegreeofprocyclicalityisafunctionoftheweightgiventoconsumerwelfare:themoregovernmentcaresaboutrents,thelargeristhereactionofprivateconsumptiontoin-comeshocks(i.e.,themoreprocyclicalscalpolicyis,inthesensedescribedabove).4EmpiricalEvidence4.1EstimationstrategyThem

17 odeloftheprevioussectionsexplainsprocycl
odeloftheprevioussectionsexplainsprocyclicalscalpolicyastheresultofapoliticalagencyproblem.Votersdonottrustthegovernment,anddemandhigherutility(intheformoflowertaxesorhigherproductivegovernmentspending)whentheyseeaggregateoutputgoingup.Thus,themodelpredictsthatprocyclicalscalpolicyshouldmainlybeobservedincountrieswherepoliticalcorruptioniswidespread.Incountrieswherethegovernmentissubjecttochecksandbalances,andpoliticalrentsarenegligi-bleordiculttocapture,votershavenoreasontomistrustthegovernmentandimpartaprocyclicalbiastoscalpolicy.Moreover,procyclicalscalpolicyresultsfromacorruptgovernmentwhichisneverthelessaccountabletothevoters.Ifthegovernmentweretotallyunaccountable(sayinastabledictatorship),thencorruptionwouldbehighbutvoterscouldnotinwhatthegovernmentdoes,andhencetheprocyclicalbiasshouldnotbeobservedorbelessprevalent.Totakethesepredictionstothedata,weestimateacross-countryregres-sionofthefollowingtype:BetaControlcorruptionwherethesubscriptdenotescountries,thedependentvariable,Beta,isameasureofcounter-cyclicalityinscalpolicy,thevariableControlcorruptionisameasureof(theabsenceof)politicalcorruption,anddenotesothercontrolvariablesdescribedbelow.Thecoecientofinterestis,whichweexpecttobepositive:scalpolicyismorecounter-cyclicalwhenthereisless corruption.Moreover,weexpecttobepositiveindemocracies,butnotindictatorships:corruptionimpartsaprocyclicalbiasindemocracies(w

18 herethegovernmentisaccountabletothevoter
herethegovernmentisaccountabletothevoters),butnotindictatorships(whereaccountabilityisnotanissue).Finally,itisnotobviousthattherelationshipbetweenpoliticalcorruptionandprocyclicalityinscalpolicyislinearandcontinuous(atinyamountofmorecorruptionleadstoslightlymoreprocyclicality).Onthecontrary,theversionofthemodelwhichwepresentedimpliesthattherearetwogroupsofcountries:thosewherepoliticalchecksandbalancesarenotworkingwellandcorruptionisrampant,andthosetowhichthisdoesnotapply.Theformerismorelikelytobethecaseinlessdevelopedeconomies,giventhewellknownpositivecorrelationbetweenqualityofinstitutionsandpercapitaincome.Infact,previousstudieshaveshownthatprocyclicalscalpolicyisabigprobleminsomedevelopingcountries,butnotamongindustrialdemocracies(cf.Kaminski,ReinhartandVegh2004).Toallowforanon-linearrelationshipbetweencorruptionandprocyclicalityinscalpolicy,wealsoestimateaprobitversionof(12),wherethedependentvariableis1ifBeta0(counter-cyclicalscalpolicy)and0ifBeta0(procyclicalscalpolicy).Thus,weestimatetheprobabilityofobservingprocyclicalscalpolicyincountryTheregressorsarethesameregressorsthatappearontherighthandsideof(12).4.2Data4.2.1CyclicalityofscalpolicyWeconstructameasureofcyclicalityofscalpolicyfollowingCataoandSutton(2002),whointurnadaptGavinandPerotti’s(1997)specication.Ourmeasureofprocyclicalityisthecoecientfromthefollowingre-gressionestimatedseparatelyforeachcountryin

19 thesample(subscriptsdenoteyears):βJDPJA
thesample(subscriptsdenoteyears):βJDPJAPγ∗O∗where:referstothecentralgovernment’soverallbudgetsurplusasaper-centageofGDP(obtainedfromtheWorldBank’sGlobalDevelopmentNet-workGrowthDatabase(GDNGD),basedondataoriginallyfromtheInter-nationalMonetaryFund’sGovernmentFinancialStatisticsJDPJAPtheoutputgap,denedasthelogdeviationofGDPfromitsHodrik-Prescott ∗O∗isameasureofthegapintermsoftrade,alsodenedasthelogarithmicdeviationfromaHodrick-Prescott-lteredseries.ThesourcefortheGDPandtermsoftradeseriesistheWorldDevelopmentIndicators(WDI).AnegativecoecientonJDPJAPimpliesthatacyclicalboomisassociatedwithadecreaseinthebudgetsurplus,meaningthatthebehaviorscalpolicyisprocyclical.WealsoconsiderthesamespecicationbutredenethedependentvariableaseitherspendingortaxrevenueasashareofGDP,ratherthanthebudgetsurplus.Inchoosingthespecicationforourcomputationofprocyclicalitywehadtobeparsimonious,tohaveaslargeasampleofcountriesaspossible.Wehavedataon87countries,inasamplethatgoesfrom1960to1999but,despitetheparsimoniousspecication,onlyforasmallnumberofcountrieswehavedataforthewholeperiod,withafewcountrieshavingaslittleasveyearsofdata.Asarobustnesscheck,wehavealsoruntheregressionslimitingthesampletothosecountrieswithatleasttenyearsofdata,andtheresultsaresimilar.Table1reportsourestimateoftheparameterinour87countries;thescalpolicyvariableinthelefthandsideof(13)isthebudgetsurplus.Ourestimatesco

20 nrmpreviousresultsbyothers.IntheOECDcoun
nrmpreviousresultsbyothers.IntheOECDcountriesscalpolicyiscountercyclical:inallbutthreecountriesthecoecientbetahasthecorrect(i.e.positivesign)andsodoestheaveragefortheOECD.In36outof64non-OECDcountriesinthesamplethecoecienthastheincorrectsign(i.e.itisnegative).ThetworegionswiththemostprocyclicalscalpolicyareSub-SaharanAfricaand(especially)LatinAmerica,whichisthe“worst”region.Thisisinfacttheregionthathadoriginallyattractedtheinterestofeconomistsaboutprocyclicalityofscalpolicy(seeGavinandPerotti(1997),forinstance).NotethatmostcountriesinEasternEuropehaveveryfewyearsofobservations,soresultsonthisregionhavetobetakencumgranosalis.Table2reportsthecoecientobtainedinregressionswhereonthelefthandsideof(13)wehavespendingandtaxrevenuesasapercentageofGDP.Here,procyclicalitycorrespondstoforexpenditures,andfor Wehavedroppedfromoursampletwooutliercountries:IsraelandYemen.Thesecountrieshavecoecientsthatexceed1inabsolutevalue(Israelnegative,Yemenposi-tive).NotethatIsraelisacountrywithrelativelylowcorruptionandaveryprocyclicalscalpolicy(accordingtotheestimatedcoecient).ThereforeexcludingIsraelwouldhelpndingtherelationshipbetweenprocyclicalityandcorruptionderivedfromthemodel.ThesameappliestoYemenintheoppositedirection.Includingbothcountriesinoursampledoesnotaecttheresults. revenues.InOECDcountriesmostofthecounter-cyclical“action”isonthespendingside.InLatinAmericaboththespendingsideandt

21 herevenuesideareprocyclical,withthesizeo
herevenuesideareprocyclical,withthesizeofthespendingsidebeingtwiceaslargeinabsolutevalue.InSub-SaharanAfricaalltheprocyclicalitycomesfromthespendingside.Thus,inthesetworegionswhichareespeciallyprocyclicalthe“incorrect”timingofscalpolicyseemstobedrivenalmostexclusivelybyspending.Manyofthecoecientshavelargestandarderrorsandarenotsignif-icantlydierentfromzero.However,inOECDcountriesthepositivecoecientsinthesurplusregressionandthenegativeoneonspendingarealmostallsignicantlydierentfromzero.InLatinAmericanandSub-SaharanAfricaseveralofthecoecientswiththewrongsignarestatisticallydif-ferentfromzero,butamajorityofthecoecientsinnon-OECDcountriesarenotstatisticallydierentfromzero.Tocopewithlikelymeasurementerror,belowwealsoreportregressionswereweweightobservationswiththe(inverseof)theestimatedstandarderrorsofthecoecients.4.2.2ControlofcorruptionTomeasurethedegreeofcorruption,weusetheControlofCorruptionindexfromKaufman,KraayandMastruzzi’s(2004)aggregategovernanceindica-tors,whichaggregatesseveralscoresandratingsfromdierentsourcesonascaleof.Thisindexisdecreasingintheamountofcorruptionanditisavailablefor1996,1998,and2000,andwetaketheaverageofthethreeyears.Werepeatedtheexerciseusingdataonperceptionsfromcor-ruptionfromICRG(availableforthewholeperiodyears1982-97)andfromTransparencyInternational(availablefrom1996onward).Theresultswereverysimilar,sincethesecorruptionindicatorsarehighl

22 ycorrelatedwitheachotherandmoveslowlyove
ycorrelatedwitheachotherandmoveslowlyovertime.Intheend,weonlyreportresultsusingtheKaufman,KraayandMastruzzi(2004)data,sincetheyareavailableforalargernumberofcountries.4.2.3OtherregressorsWealsocontrolforotherregressorsthatmaydeterminethecyclicalresponsescalpolicy(thevariablein(12)).Toallowfordierencesinthelevelofeconomicdevelopment,wecontrolforrealpercapitaincome;wemeasureitasrealGDPpercapitaininternationalprices(PPPadjusted)intherst yearofthesampleoverwhichthemeasureofprocyclicalityofscalpolicyiscomputedforeachcountry.ThesourceistheWDI.ThisvariableiscalledInitialGDP(percapita).Tocapturehowdemocraticacountryis,werelyonthevariablePolity2,whichsubtractsthecountry’sscoreinan”Autocracy”indexfromitsscoreina”Democracy”index(resultinginarangefrom),fromthePolityIVProjectdatabase.Weaveragethisvariableoverthesampleusedtocomputethescalpolicymeasure,andusetheresulttoconstructtwodierentmeasures:acoarserDemocracydummy,whichequalsoneiftheaveragePolity2scoreisstrictlypositive,andzerootherwise;andtheaveragePolity2measureitself.Thecyclicalpropertiesofscalpolicymaydieracrosscountriesbecausethesizeofgovernmentspendingisdierent.Hencewealsocontrolforthisfeatureofacountry’spublicsector,withavariableGovernmentsizenedastheshareofcentralgovernment’sexpenditureinrealGDP,alsofromtheFinally,asothershavenoted(e.g.GavinandPerotti1997),procyclicalscalpolicymayalsoresultfromtightcreditconstraints.

23 Wemakeuseoftwovariablestoproxyforthedegr
Wemakeuseoftwovariablestoproxyforthedegreeofnancialconstraintsfacingacountry’sgovernment.Oneofthemisanaverageoftheexistingsampleofratingsat-tributedbyStandard&Poor’stoacountry’slong-termforeign-denominatedsovereigndebt(S&PRatingWeinterpretthisaverageasaninversemea-sureofthedegreeofnancialconstraintsfacingacountry’sgovernment.Theothervariableisthelogarithmofthespread(inbasispoints)ofacountry’ssovereigndebtoverU.S.Treasurybondsatthetimeofissuance(SpreadwhichcomesfromCapitalDataBondwareandSDCPlatinum.Thiscon-stitutesadirectmeasureofnancialconstraints.Itshouldbenotedthatbothofthesevariablesarelimitedintermsofthenumberofcountriesinoursampleforwhichtheyareavailable(70and47,respectively)andofthe AdaptingCantorandPacker’s(1996)approach,weattributenumbersfrom0to6toS&P’sletter-basedsystem:C(defaultorselectivedefault);B(high-riskobligations);BB(likelytofulllobligations,ongoinguncertainty);BBB(adequatepaymentcapacity);A(strongpaymentcapacity);AA(highquality);AAA(highestquality).CountriesratedatBBBorbetteraresaidtohave”investment-grade”ratings.Sincechangesinratingsoccuratirregularintervals,wecomputedtheaveragebyweigh-ingagivenratingbytherstintegergreaterthanthenumberofyearsoverwhichitwaskept.Similarlytothecaseofratings,theissuanceofnewdebtoccursatirregularintervals.Wethususeasimilarweightingsystemtocomputetheaveragespread,takingintoaccountthelengthofoftheperiodbetweenemissions.

24 periodoverwhichtheyareavailable:formost
periodoverwhichtheyareavailable:formostcountriesthesamplestartsinthe1990s,whichmeansthat,unlikethedemocracyvariables,theaverageisnottakenoverthesampleusedtocomputethescalpolicyvariable.4.3ProcyclicalityandcorruptionWenowexaminethecorrelationbetweenthecyclicalpropertiesofscalpol-icyandpoliticalcorruption.Table3reportssomesimplestatistics.LatinAmericaandSub-SaharanAfricahavethelowestscoreoncontrolofcorrup-tionandtheyarethemostprocyclicalregions.Table4presentssomemultivariateregressionsestimatedbyOLS.ThedependentvariableistheestimatedbetacoecientonthebudgetsurplusreportedinTable1.Columns1and2showastronglysignicantrelationshipbetweenControlofcorruptionandprocyclicality:morecorruptcountriesaremoreprocyclical,evenaftercontrollingforinitialGDPpercapitainlog(col-umn2).Column3showsthattheestimatedcoecientonthevariableDemocracyisnotstatisticallysignicant,butColumn4highlightstheinter-actionbetweenDemocracyControlofcorruption:whenthetwovariablesareinteracted,theestimatedcoecientoncorruptionalonedropsandbe-comesstatisticallyinsignicant,whiletheinteractiontermhasapositiveandsignicantestimatedcoecient(ofaboutthesamesizeasthecoecientofcorruptionaloneinColumn3).Thus,theeectofcorruptiononprocyclical-ityisapparentonlyamongdemocracies,aspredictedbythetheory.Columns5and6repeatthesameexperimentusingthecontinuousPolity2variabletomeasuredemocracy;theresultsareverysimilar.Werepeatedallthe

25 seregressionsaddingthevariableGovernment
seregressionsaddingthevariableGovernmentsizetherighthandside,wherethisvariableisaveragedfortheperiodinwhichthelefthandsideisdened.Thisnewvariablewasneverstatisticallysig-cant,andthesignandsignicanceofalltheotherestimatedcoecientswasunchangedrelativetoTable4.WealsorantheseregressionsexcludingEasternEuropeancountries,wherethenumberofobservationsavailabletoconstructthebetacoecientswererelativelysmall.Theresultsarequalita-tivelyunchanged.InTable5weusetheProbitmethodandwedenethe InitialGDPismoreexogenoustocyclicalityofscalpolicythanpercapitaincomemeasuredinthemiddleorattheendofthesampleperiod.Inanyevent,werepeatedtheregressionsusingpercapitaincomemeasuredinthemiddleofthesampleperiodavailableforeachcountryandtheresultarevirtuallyunchanged.Alltheseresultsareavailableuponrequest. lefthandvariable(beta)asazero-onevariable(1ifpositive,zeroifnegative).Theresultsofthistableconrmthepreviousone;ifanythingtheymakeitevenclearerthattheeectofcorruptiononprocyclicalityispresentonlyindemocracies,aresultconsistentwiththespiritofthemodel.Ourdependentvariable(theestimatedbetacoecients)islikelytobemeasuredwitherror.Moreover,measurementerrorcouldvaryacrosscoun-tries,andbelargerinthecountrieswhereprocyclicalityisestimatedfromashortertimeseries.Tocopewiththisproblem,inTable6weestimatethesamespecicationsasinTable4,butweweightobservationsbythe(in-verseof)thestandarderroroftheestimatedcoecients

26 .TheestimatedcoecientonControlofcorrupti
.TheestimatedcoecientonControlofcorruptionisnowabitsmaller,butthepatternofsignicanceandtheinferenceswecandrawremainthesameasinTable4.Finally,Table7repeatsthesameexerciseasinTable4,butnowthedependentvariableisthecoecientestimatedfromgovernmentspendingregressions(andreportedinTable2).Hereapositivedenotesprocyclical-ity.HenceweexpectControlofcorruptiontohavetheoppositesignasinTable4:moreControlofcorruptionshouldbeassociatedwithlessprocycli-cality,andhenceweexpectanegativecoecient.Thisisindeedwhatwend.ThepatternofsignicanceisthesameasinTable4.Inparticular,theectofcorruptiononcyclicalityofspendingisstrongerindemocracies(theinteractiontermbetweenDemocracyPolity2Controlofcorruptionhasanegativeandsignicantestimatedcoecient).HerewealsondthattheestimatedcoecientonthevariableDemocracyPolity2)inisolationispositiveandsignicant,suggestingthat,asidefromtheinteractionwithcorruption,democraticgovernmentonitsownleadstomorecountercyclicalgovernmentspending.Thismightalsoreectthecompositionofgovernmentspendingindemocracies,thatislikelytobemoreleaningtowardssocialin-surance.Repeatingthesameexperimentforthecyclicalityoftaxrevenuesleadstoinconclusiveresults.WhenthedependentvariableisdenedascoecientontaxrevenuesinpercentofGDP,norobustpatternofcorrelationsemerges,andtheoftheregressionisoftheorderof0.02:essentiallywecannotexplaintheobservedpatternsofcyclicalityoftaxrev-enues.Theseresultsonta

27 xationandspendingsuggestthattheprocyclic
xationandspendingsuggestthattheprocyclicalityscalpolicyisdrivenbyspending,notbytaxation. 4.4TheroleofborrowingconstraintsThemajoralternativeexplanationofaprocyclicalscalpolicy,dierentfromourown,isthatofborrowingconstraints.Ifacountryiscreditrationed,itcanspendonlyinbooms;moreover,indownturnsthecreditconstraintcouldbecomemorebindingandthegovernmentisforcedtoreininscalpolicy.Howcanwediscriminatebetweenthesetwoexplanations,corruptionvsborrowingconstraints?Thekeydicultyinaddressingthisissueisthatcorruptionandcreditratingsareveryhighlycorrelated.AsshowninTable8,thecorrelationcoecientbetweenthevariablesS&PRatingandControlofcorruption0.92.Controlofcorruptionisalsohighlycorrelatedwithavailabledataoninterestratespreads(Spread),acorrelationof-0.82.Needlesstosaythetwovariablemeasuringcreditrationingarehighlycorrelatedbetweeneachother(-0.89).Infact,someofthesevariablesarecorrelatedbyconstruction.ForinstanceStandardandPoormaylook(directlyorindirectly)toperceptionofcorruptionasoneoftheirinputinassigningratingstocountries.Andperceptionsofcorruptionmaybeinuencedbyforeigners’viewsofacountrycreditworthiness.Thecreditvariableandcorruptionareallcorrelatedwiththeexpectedsignwithourmeasureofprocyclicality.rstobservationrelatestotheresultsontheinteractionbetweendemocracyandcorruption.Asdiscussedintheprevioussubsection,thecorrelationbetweencorruptionandprocyclicalityisstrongerindemocracie

28 s.Thisisadirectimplicationofourmodel.Tob
s.Thisisadirectimplicationofourmodel.Tobealsoconsistentwithaborrow-ingconstraintstory,corruptdemocracieswouldhavetobeworstborrowersthancorruptdictatorships(whilelesscorruptgovernmentswouldbeequallytrustworthyindemocraciesandautocracies).Apriorionecanthinkofmanyreasonwhyitmightbetheopposite,oratleastwhytheinteractionbe-tweendemocracyandcorruptionwouldnotberelevantindeterminingcreditworthiness.Thesecondroughtestweperformedisthefollowing.InTable9wereporttheonthebudgetsurplusestimatedontwosub-samples:preandpost1982,theyearoftheMexicandebtcrisesthatopeneduptwodecadesofdebtcrises,defaultsetc.Itisfairtosaythat,before1982,concernsaboutdefaultandcreditworthinessweremuchlessimportantthanafterwards.In Werepeatedthesameprocedureusingtheprobitspecicationforthelefthandsidevariable.Thegeneralpictureoftheresultsthatweobtainedissimilartotheonepresentedinthetext. generalthistableshowsnoevidencethatprocyclicalityincreasedafter1982.Infacttheoppositeseemstobethecase;inLatinAmericaandSub-SaharanAfrica,thetwomostprocyclicalregions,weobservelessprocyclicalityafter1982thanbefore.Thisseemsratherinconsistentwithapurelyborrowingcostsstory.Thisinterpretationisalsoconrmedbyre-estimatingtheregressionsofprocyclicalityoncorruptionwiththebetacoecientsestimatedbefore1982.AsshowninTable10,Controlofcorruptionretainsapositiveandstatis-ticallyestimatedcoecientevenwhenthemeasureofcyclicalityreferstothepr

29 e-1982period(althoughheretheinteractionb
e-1982period(althoughheretheinteractionbetweenDemocracyandControlofcorruptionisnolongerstatisticallysignicant).Theseregressionsoughttobeinterpretedwithcaution,bothbecausecyclicalityislikelytobemeasuredwitherror(thesampleisrathershortformanycountries),andbecauseourmeasureofcorruptionreferstoaratherdistantperiodintime.Nevertheless,theseestimatessuggestthatthecorrelationbetweenprocycli-calityandcorruptionisarobustfeaturethatpre-datestheyearsofsovereigndebtcrises.WehavealsoaddedthemeasureofStandard&Poor’screditworthinessintoourbasisregressionsofTable4.TheresultsaredisplayedinTable11.InColumn1,wherecontrolofcorruptionandcreditratingareintroducedtogether,theyarebothinsignicantgivenhowhighlycorrelatedtheyare.Column2howeverhighlightsasuggestiveresult:theinteractionofControlofcorruptionwithdemocracyisstillsignicantandpositiveevenwhenwecontrolfortheS&PRatingvariable.Columns3and4repeatthesameex-perimentsofColumns1and2withtwoothermeasureofdemocracy(Polity2)withsimilarresults.Columns5to8repeatthesameregressionsfromColumns1to4usingtheSpreadmeasureofcreditratingratherthantheS&PRatingvariable.WhatwendhereisthattheSpreadmeasureisstatis-ticallysignicantbutsoistheinteractionofdemocracywiththemeasureofcontrolofcorruption.Wehavealsorepeatedthesameregressionsweightingobservationsbythe(inverseof)thestandarderrorsofthecoecients,andusingtheProbitmethod,obtainingsimilarresults.Notealsothati

30 fcreditconstraintswerethemaincauseofproc
fcreditconstraintswerethemaincauseofprocyclicality,thelattershouldbeespeciallyevidentinrecessions,whengovernmentscan- InalltheseregressionswehaveexcludedtheoutliersofIsraelandYemenbutnoneofourresultdependonincludingorexcludingthem.Alsowehaveexcludedanycountrywithfewerthan5observationsusabletocomputethecoecient.ThevariableDemocracyhasbeenredenedtocorrespondtothepre-1982period. notborrowandthushavetotaxmoreorspendlessinordertoovercomerecession-induceddecits.Inboomsinsteadthegovernmentshaveresources,andnotonlycantheybeanticyclical(i.e.reducespendingasashareofGDPorincreasetaxrevenuesasashareofGDP)but,expectingconstraintsinrecessions,theyshouldbeespeciallykeenonaccumulatingsurplusesinbooms;thustheyshouldbeespeciallyanticyclicalinbooms.Itturnsoutthatgovernmentsaremoreprocyclicalinboomsthaninrecessions;infactinmorethan2/3ofthecountriesthatareprocyclicaltheyaremoresoinboomsthaninrecessions.Thisobservationcastssomedoubtsonasimpleborrowingconstraintinterpretationofprocyclicality.Infact,wehavereestimatedequation(13)separatelyforeachcountry,allowingthecoecienttotakeondierentvaluesinpositivevsnegativeoutputgaps.Ouragencymodelpredictsnoasymmetry:thereisnoreasonwhyscalpolicyoughttoreactverydierentlyinboomsvsrecessions.Ac-cordingtothecreditmarketimperfectionhypothesis,instead,ifanythingprocyclicalityshouldbestrongerduringrecessions,asdiscussedabove.Ta-ble12liststhe39countrieswithp

31 rocyclicalscalpolicy(i.e.withnegativeest
rocyclicalscalpolicy(i.e.withnegativeestimatedcoecientsinTable1).Therstcolumn(BetaRecession)re-portstheestimatedcoecientduringrecessions;thesecondcolumn(BetaBoom)reportstheestimatedcoecientduringbooms.Rememberthatthebetacoecientsshouldbepositivebothinrecessions(GDPGAPneg-ative)andinbooms(GDPGAPpositive).In21outof39countries,BetaBoomisnegativeandBetaRecessionpositive,indicatingthatonlyinboomsweobserveprocyclicality.Moreoverin28outof39countriesBetaBoomisloweralgebraicallythanBetaRecessionInotherwords,onlyin11outof39countriesprocyclicalityseemstobedrivenmostlybyrecessionperi-ods.Giventhelargestandarderrors,thedierencebetweenBetaBoomandBetaRecessionisalmostneverstatisticallysignicant,butneverthelessthisevidenceissuggestive,sincethecreditconstrainthypothesiswouldimplyasignicantdierenceinwhichprocyclicalityemergesespeciallyoronlyinrecessions.Asafurthercheck,inTable13wehaveregressedBetaRecession(i.e.thereactionofthesurplustorecessions)againstourvariablesofinterest.Ac-cordingtothecreditmarketimperfectionhypothesis,thevariablesmeasuringcreditconstraints(S&PRatingandSpread)shouldhavemoreexplanatorypowerintheseregressionsthanagainsttheoverallmeasuresofprocyclicality,andconverselythevariableControlofcorruptionshouldhavelessexplana-torypowerhere:asarguedabove,creditmarketimperfectionsaremorelikely toleadtodistortedscalpolicyatthetimeofrecessionthanduringbooms.Butthisisnotwhatwe

32 nd.Incolumn1,Controlofcorruptionisgenera
nd.Incolumn1,Controlofcorruptionisgenerallyhighlysignicantandwiththeexpectedsign.Columns2and3showthatalsothemeasuresofSpreadS&PRatingaresignicant,butthelasttwocolumnsshowthatwhencontrolofcorruptionandthecreditmarketvariablesareenteredtogethertheControlofCorruptionvariabledominates.Incolumn4itistheonlyonewhichissignicant,inthelastoneitisbor-derlineinsignicantbutmuchclosertosignicancethatthecreditconstraintvariable.Insummary,itisdiculttodisentangleclearlytheeectsofcreditcon-straintspersefrompoliticalvariablessuchasmeasureofcorruption.How-everthetestspresentedabovesuggestatleasttentativelythatcorruptionandpoliticalimperfectionsplayamajorroleinexplainingprocyclicalityofscalpolicy,andperhapsmoresothancreditconstraints.5ConclusionsInmanydevelopingcountriesscalpolicyisprocyclical.Ourexplanationisthatvotersdonottrustcorruptgovernmentwithresourcesanddemandtaxcutsorincreaseinproductivegovernmentspendingortransferswhenposi-tiveshockshittheeconomy.Otherwisetheyfearthattheavailableresourceswouldbe“wasted”inrents.Forthesamereasonsthevotersdonot“allow”governmenttoaccumulatereservesofassets,onthecontrarytheydemandalevelofgovernmentdebtthatforcesthegovernmenttouseresourcestopayinterestratherthanstealthem.Thereforethispoliticaldistortion,relatedtothe“starvetheLeviathan”argument,leadstoahigher-than-rst-bestaccu-mulationofgovernmentdebtandprocyclicalscalpolicy.Creditconstraintscomeint

33 oplayindirectlybecausethepoliticaldistor
oplayindirectlybecausethepoliticaldistor-tionmaypushthegovernmenttowardslevelsofdebtthatareatthelimitofwhattheycanrepayandthereforeatthelimitofwhatborrowerscanlend.Theliteraturethatfocusesoncreditconstrainsreviewedabove,implicitlyorexplicitlysuggeststhatthe“malfunctioning”ofcreditmarketsmakesithardorimpossiblefordevelopingcountriestoborrowexactlywhentheyneeditmore,inbadtimes.Butthisargumentfailstoexplainwhywelfaremaximizinggovernmentsdon’ttakethisintoaccount,buildingupreservesingoodtimes,soastoavoidbeingcreditconstrainedinbadtimes.Moreover,ourevidencesuggeststhatprocyclicalityismoreoftendrivenbyadistorted policyreactiontobooms,ratherthantorecessions.Andwhateverprocycli-calpolicyresponsethereistorecessions,doesnotseemtobeexplainedbyavailablemeasuresofcreditconstraints.6AppendixProofofProposition2Nowconsiderthecaseinwhichtheupperboundonrentsisadecreasingfunctionofdebtoutstanding:withGoingthroughthesamestepsasinsection3,inequilibriumthegovernmentmustbeindientbetweenpleasingthevotersandbeingreappointed(takingintoaccountthefutureequilibriumcontinuationvalue),orgrabbingasmanyrentsaspos-sibletoday.Thisindierencecondition(theanalogueof(9)insection3)herecanbewrittenas:))=Hence,equilibriumrentsaredeterminedjointlywithequilibriumgov-ernmentdebt.RepeatingthestepsofSection3,agovernmentseekingreap-pointmentchoosespublicdebtsoastomaximize(11).Buthere,Hence,equilibriumpublicdebtis

34 determinedbythefollowingop-timalitycondi
determinedbythefollowingop-timalitycondition:thelefthandsideof(15)isthemarginalbenetofborrowing,namelytheadditionalrentsthatthegovernmentcangrabtodaywiththedebtproceeds.Therighthandsideisthemarginalcostofissuingdebt,namelythereductionintheupperboundofrentstomorrow,whichinturnreducesthevalueoftheincumbent’sfutureout-of-equilibriumthreat.Together,(9)and(15)determinetheequilibriumtimepathsofrentsandpublicdebt.Thesteadystateisobtainedimposingin(14),toyieldanex-pressionforequilibriumrentsthatcloselyresemblesequation(10)inSectionBy(16),equilibriumrentsarebelowtheupperboundinthesteadystate:.Withstrictlyconcavepreferences,equation(15)thenimpliesthatthesteadystateisataninterioroptimum(i.e.onlyifIntuitively,forthegovernmenttoborrowlessthanthemaximumthecost ofissuinggovernmentdebtmustbehighenough.With,themarginalutilityofcurrentrentsishigherthanthemarginalutilityofrentsevaluatedattheupperbound;hencethegovernmentndsitoptimalnottoissuemoredebtonlyiftheupperboundonrentsshrinksmorethanoneforoneasmoredebtisissued:.Assumingthatthisconditionholdsforsome,thenthesteadystatecancorrespondtoaninterioroptimumforgovernmentdebt.Wenowshowthatthesteadystateislocallystable(i.e.that inaneighborhoodofthesteadystate).Equation(15)implicitlydeequilibriumrentsasafunctionofgovernmentdebt:.Applyingtheimplicitfunctiontheoremto(15),wealsohave: Replacingin(14),theequilibriumlawofmotionofgovernmentdebtisi

35 mplicitlydenedby:y:F(b0)]++R(b0)]−v[R(
mplicitlydenedby:y:F(b0)]++R(b0)]−v[R(b)]=0Nowuse(18)tocompute inaneighborhoodofthepointAftersomesimplicationswehave: 1 Thus,recallingthat,thatandthat,andusing(17),wehavethat providedthatrrrandthatisnottoocloseto0inabsolutevalue.Finally,notethatinequilibrium(onandothesteadystate)neitherrentsnorpublicdebtdependonincome.Thebudgetconstraintthenimpliesthattemporaryincomeshockschangeconsumptiononeforone.7ReferencesReferences[1]AlesinaAlbertoandRobertoPerotti(1995)“ThePoliticalEconomyofBudgetDecits”IMFStaPapers42:1-31. [2]Alesina,AlbertoandGuidoTabellini(1990)“APositiveTheoryofFiscalcitsandGovernmentDebt”ReviewofEconomicStudies57:403-14.[3]Barro,Robert(1973)“TheControlofPoliticians:AnEconomicModel”,PublicChoice,13:19-42.[4]Cantor,RichardandFrankPacker(1996).“DeterminantsandImpactofSovereignCreditRatings,”FederalReserveBankofNewYorkEconomicPolicyReview,20:37-54.[5]Catao,LuisA.andBennettW.Sutton(2002).“SovereignDefaults:TheRoleofVolatility,”IMFWorkingPapers02/149,InternationalMonetaryFund.[6]Ferejohn,John(1986)“IncumbentPerformanceandElectoralControl”PublicChoice,50:5-25.[7]Gavin,MichaelandRobertoPerotti(1997).“FiscalPolicyinLatinAmerica,”inBernanke,BenandRotemberg,Julio,NBERMacroeco-nomicsAnnual1997,Cambridge,MA:MITPress.[8]GlobalDevelopmentNetworkGrowthDatabase(compiledbyWilliamEasterlyandMirvatSewadeh),availableon:http://www.worldbank.org/research/growth/GDNdata.htm[9

36 ]Jensen,MichaelC.,andWilliamH.Meckling(1
]Jensen,MichaelC.,andWilliamH.Meckling(1976),“TheoryoftheFirm:ManagerialBehavior,AgencyCostsandOwnershipStructure.”JournalofFinancialEconomics,3:305-360.[10]Kaminski,Graciela,CarmenReinhart,andCarlosVegh(2004)“WhenitRainsitPours:ProcyclicalCapitalFlowsandMacroeconomicPoli-cies”inMarkGertlerandKennethRogoNBERMacroeconomicAnnual2004,,Cambridge,MA:MITPress.[11]Kaufman,Daniel,AartKraayandMassimoMastruzzi(2004),“Gover-nanceMattersIII:GovernanceIndicatorsfor1996-2002,”WorldBankPolicyResearchWorkingPaper3106.[12]Lane,PhilipandAaronTornell(1998)“WhyAren’tLatinAmericanSavingRatesProcyclical?”JournalofDevelopmentEconomics,57:185- [13]MilesiFerretti,GianMaria(2003)“GoodBadorUgly?OntheEofFiscalRulesonCreativeAccounting”JournalofPublicEconomics,88:377-94.[14]Perotti,Roberto(2004)“EstimatingtheEectsofFiscalPolicyinOECDCountries”,mimeo,IGIER-Bocconi.[15]Persson,TorstenandLarsSvensson(1989)“Whyastubbornconser-vativewouldrunadecit:Policywithtime-inconsistentpreferences”,QuarterlyJournalofEconomics,104:325-345[16]Persson,TorstenandGuidoTabellini(2000)PoliticalEconomics:Ex-plainingEconomicPolicy,Cambridge,MA:MITPress.[17]PolityIVProject,availableon:http://www.cidcm.umd.edu/inscr/polity/#data[18]Rogo,Kenneth(1990)“EquilibriumPoliticalBudgetCycles”Ameri-canEconomicReview,80:21-36.[19]Rogo,KennethandAnnSiebert(1988),“ElectionsandMacroeconomicPolicyCycles”,ReviewofEconomicStudies55:1-16.

37 [20]Satyanath,ShankerandArvindSubramania
[20]Satyanath,ShankerandArvindSubramanian(2004)“WhatDeterminesLong-RunMacroeconomicStability?DemocraticInstitutions”,IMFWorkingPaperNo.04/215,InternationalMonetaryFund.[21]StandardandPoor’sCreditRatings,availableon:http://www2.standardandpoors.com/[22]Tabellini,GuidoandAlbertoAlesina(1990),“Votingonthebudgetcit”,AmericanEconomicReview80:37-49.[23]Talvi,ErnestoandCarlosVegh(2005)“TaxBaseVariabilityandPro-cyclicalityofFiscalPolicy”JournalofDevelopmentEconomics,forth-coming.[24]Tornell,AaronandPhilipLane(1999)“VoracityandGrowth”AmericanEconomicReview,89:22-46.[25]VonHagen,JurgenandJanHarden(1994)“NationalBudgetProcessandFiscalPerformance”EuropeanEconomyReportsandStudies3:311- Why is Fiscal Policy Often Procyclical?Alberto Alesina and Guido TabelliniNBER Working Paper No. 11600September 2005JEL No. ABSTRACTMany countries, especially developing ones, follow procyclical fiscal polices, namely spending goesup (taxes go down) in booms and spending goes down (taxes go up) in recessions. We provide anexplanation for this suboptimal fiscal policy based upon political distortions and incentives forless-than-benevolent government to appropriate rents. Voters have incentives similar to the“starving the Leviathan” classic argument, and demand more public goods or fewer taxes to preventgovernments from appropriating rents when the economy is doing well. We test this argumentagai

38 nst more traditional explanations based
nst more traditional explanations based purely on borrowing constraints, with a reasonableamount of success.Alberto AlesinaDepartment of Economics210 Littauer CenterHarvard UniversityCambridge, MA 02138and NBERaalesina@harvard.eduGuido TabelliniIGIER Bocconi UniversityVia Salasco 520136 MilanITALYguido.tabellini@uni-bocconi.it Whyisscalpolicyoftenprocyclical?AlbertoAlesinaandGuidoTabelliniHarvardUniversityandIGIERBocconiFirstdraft:March2005Thisversion:July2005AbstractManycountries,especiallydevelopingones,followprocyclicalpolices,namelyspendinggoesup(taxesgodown)inboomsandspend-inggoesdown(taxesgoup)inrecessions.Weprovideanexplanationforthissuboptimalscalpolicybaseduponpoliticaldistortionsandincentivesforless-than-benevolentgovernmenttoappropriaterents.Votershaveincentivessimilartothe“starvingtheLeviathan”classicargument,anddemandmorepublicgoodsorfewertaxestopreventgovernmentsfromappropriatingrentswhentheeconomyisdoingwell.Wetestthisargumentagainstmoretraditionalexplanationsbasedpurelyonborrowingconstraints,withareasonableamountofsuccess.1IntroductionMosteconomistswouldagreewiththenormativeprescriptionthattaxratesanddiscretionarygovernmentspendingasafractionofGDPoughttoremainconstantoverthebusinesscycle.Ifgovernmentsrespectedtheseprescrip-tions,weshouldobserveacounter-cyclicalpatterninscalpolicy.Namely, WethankElianaLaFerrara,RobertoPerotti,KennethRogo,FrancescoTre

39 bbiandparticipantsinaseminaratHarvardfor
bbiandparticipantsinaseminaratHarvardforusefulsuggestions,AshokaModyandDiegoSaraviaforhelpwithpartofthedatacollection,andFilipeCampanteforoutstandingresearchassistantshipandcomments.WegratefullyacknowledgenancialsupportfromNSFthroughagrantfromNBER(Alesina)andCIAR,MIURandBocconiUniversity(Tabellini) (Dependent variable: Beta on budget surplus in recessions) (Robust standard errors in parenthesis. Intercepts not shown.) *** significant at 1% (1) (2) (3) (4) (5) Control of Corruption (0.109) (0.126) (0.140) (0.186) (0.073) (0.162) (0.148) (0.147) (0.207) (0.165) (0.077) (0.178) R-squared Beta Coefficients - “Recession” and “Boom” (Procyclical Countries only)Country Beta Recession Beta Boom Argentina - 0.175 0.173 Belize - 0.183 - 0.530 Brazil 0.444 - 0.603 Burkina Faso 0.500 - 0.472 Chad 0.733 - 0.376 Chile - 0.019 - 0.095 Congo 0.660 - 0.430 Costa Rica 0.140 - 0.878 Ecuador 0.120 - 0.388 Egypt 0.843 - 0.434 Ethiopia 0.145 - 0.392 Gabon - 0.208 - 0.397 Germany 0.150 - 0.165 Guatemala - 0.200 - 0.791 India 0.025 - 0.126 Ireland - 0.384 - 0.259 Ivory Coast 0.107 - 0.554 Jamaica - 1.354 2.478 Luxembourg - 0.087 - 0.659 Malaysia - 0.148 0.174 Mauritius - 0.461 - 0.263 Mexico - 0.071 - 0.128 Morocco 0.053 - 0.093 Namibia 0.408 - 1.161 Nicaragua - 0.027 - 0.316 Panama - 0.023 - 0.645 Papua New Guinea 0.016 - 0.274 Paraguay 0.040 - 0.078 Peru 0.254 - 0.242 Rwanda -

40 0.037 - 0.033 Seychelles 1.650 - 4.790
0.037 - 0.033 Seychelles 1.650 - 4.790 Singapore 0.355 - 0.607 St. Lucia 0.034 - 0.122 St. Vicent & Grenadines - 0.814 - 0.223 Togo 0.549 - 1.503 Trinidad & Tobago - 2.962 - 0.277 Turkey - 0.473 - 0.031 Uruguay - 0.088 - 0.014 Venezuela - 0.421 0.186 Countries where “Beta Boom” “Beta Recession” in OLS estimates with S&P Ratings and Spread as additional regressors (Dependent variable: Beta on budget surplus) (1) (2) (3) (4) (5) (6) (7) (8) Control of Corruption (per capita) Democracy Democracy*Control of Corruption Polity2 Polity2*Control of Corruption R-squared Robust standard errors in parenthesis; intercepts not reported. *** significant at 1% (Dependent variable: Beta on budget surplus, Pre-1982) (1) (2) (3) (4) (5) (6) Control of Corruption Democracy Democracy*Control of Corruption Corruption (0.038) (0.056) (0.078) (0.057) (0.079) (0.135) (0.078) (0.080) (0.126) (0.088) (0.064) (0.073) (0.012) (0.071) (0.076) (0.011) (0.008) R-squared Robust standard errors in parenthesis; intercepts not reported. *** significant at 1% Country (continued) Beta pre-82 Beta post-82 Nicaragua -0.123 -0.280 Panama - -0.100 Paraguay -0.054 0.174 Peru -0.168 0.032 St. Lucia - -0.442 St. Vicent & Grenadines - -0.535 Trinidad & Tobago -1.350 - Uruguay 0.010 0.013 Venezuela -0.428 -0.020 Latin America and Caribbean -0.204 0.162 (excluding Haiti) “Exclude

41 d” countries are those with a coefficien
d” countries are those with a coefficient with absolute value greater than 1.500. Country (continued) Beta pre-82 Egypt 1.435 -1.135 Iran -0.093 0.497 Morocco -0.409 -0.005 Tunisia 0.203 0.110 Middle East and North Africa 0.284 -0.133 Botswana - 0.329 Burkina Faso -0.221 0.048 Burundi -0.214 0.082 Cameroon -0.020 0.045 Chad - -0.223 Congo - -0.476 Ethiopia - -0.028 Gabon -0.594 -0.157 Gambia -0.034 - Ivory Coast - -0.098 Madagascar - 0.232 Mali 0.338 -0.015 Mauritius -0.160 -0.058 Namibia - -0.243 Rwanda -0.014 - Senegal -0.180 - Seychelles - -0.413 South Africa -0.350 0.525 Togo -1.808 -0.274 Zambia 0.163 2.539 Zimbabwe 1.087 0.204 Sub-Saharan Africa -0.154 0.112 (excluding Zambia) Argentina -0.140 0.079 Belize - -0.345 Bolivia - 1.210 Brazil 0.005 -0.343 Chile -0.100 0.235 Colombia -0.163 0.334 Costa Rica -0.374 -0.206 Dominican Republic -0.035 -0.004 Ecuador 0.005 0.041 El Salvador -0.109 0.097 Guatemala -0.651 -0.064 Haiti - 3.293 Honduras -0.019 0.356 Jamaica 0.248 - Mexico -0.218 0.047 Country Beta pre-82 Beta post-82 Australia 0.015 0.662 Austria 0.135 0.188 Belgium 0.469 -0.320 Canada 0.635 0.174 Denmark 0.379 1.539 Finland 0.056 0.731 France 0.437 0.243 Germany - -0.086 Greece 0.086 -0.896 Iceland 0.329 0.339 Ireland -0.361 -0.037 Italy 0.478 0.086 Japan 0.216 1.232 Luxembourg 0.335 -0.473 Netherlands 0.110 0.147 New Zealand 0.023 0.538 Nor

42 way -0.054 0.787 Portugal -0.081 0.332 S
way -0.054 0.787 Portugal -0.081 0.332 Spain 0.139 1.075 Sweden 0.009 2.128 Switzerland 0.048 0.204 United Kingdom -0.099 0.589 United States 0.426 0.169 OECD 0.170 0.407 (excluding Denmark, Sweden) India -0.027 -0.111 Indonesia 0.002 0.292 Malaysia -0.270 0.101 Nepal -0.062 0.552 Pakistan 0.900 -0.045 Papua New Guinea -0.152 -0.152 Philippines -0.226 0.100 Singapore 0.569 0.080 South Korea 0.181 -0.035 Thailand 0.043 0.163 Asia and Pacific 0.096 Bulgaria - 0.432 Croatia - 0.001 Hungary - 0.254 Romania - 0.140 Turkey - -0.262 Eastern Europe Partial correlation coefficients Beta (surp.) Beta (exp.) Control of Corruption S&P Rating Spread Beta (surp.) 1 Beta (exp.) -0.50 1 Control of Corruption 0.26 -0.39 1 S&P Rating 0.40 -0.45 0.92 1 Spread -0.49 0.39 -0.82 -0.89 1 (Dependent variable: Beta on central government spending) (1) (2) (3) (4) (5) (6) Control of Corruption Democracy Democracy*Control of Corruption Corruption (0.029) (0.041) (0.047) (0.048) (0.044) (0.087) (0.036) (0.046) (0.077) (0.058) (0.052) (0.045) (0.008) (0.037) (0.048) (0.007) (0.005) R-squared Robust standard errors in parenthesis; intercepts not reported. *** significant at 1% (Dependent variable: Beta of budget surplus) (1) (2) (3) (4) (5) (6) Control of Corruption Democracy Democracy*Control of Corruption Corruption (0.021) (0.029) (0.033) (0.030) (0.033) (0.05

43 1) (0.060) (0.033) (0.054) (0.066) (0.03
1) (0.060) (0.033) (0.054) (0.066) (0.033) (0.033) (0.005) (0.042) (0.033) (0.005) (0.004) R-squared Robust standard errors in parenthesis; intercepts not reported. The weights are the inverse of the standard errors of the estimated beta coefficients. *** significant at 1% Probit Estimates: cyclicality of budget surplus (Dependent variable: Probability of Beta�0) (1) (2) (3) (4) (5) (6) Control of Corruption Democracy Democracy*Control of Corruption Corruption (0.130) (0.205) (0.236) (0.210) (0.245) (0.371) (0.388) (0.248) (0.382) (0.413) (0.226) (0.254) (0.037) (0.272) (0.256) (0.035) (0.030) Pseudo R-squared Robust standard errors in parenthesis; intercepts not reported. *** significant at 1% (Dependent variable: Beta of budget surplus) (1) (2) (3) (4) (5) (6) Control of Corruption Democracy Democracy*Control of Corruption Corruption (0.027) (0.037) (0.040) (0.040) (0.039) (0.057) (0.034) (0.041) (0.046) (0.046) (0.044) (0.038) (0.006) (0.036) (0.041) (0.006) (0.004) R-squared Robust standard errors in parenthesis; intercepts not reported. *** significant at 1% Region Beta Beta (exp.) Beta (rev.) Control of Corruption OECD 0.261 -0.229 0.047 1.813 Eastern Europe 0.112 0.029 0.183 -0.056 Asia and Pacific 0.032 -0.017 0.122 -0.008 Middle East and North Africa 0.027 -0.133 -0.038 -0.073 Sub-Saharan Africa -0.072 0.132 0.041 -0.339 Latin Am

44 erica and Caribbean -0.132 0.093 -0.047
erica and Caribbean -0.132 0.093 -0.047 -0.237 Country Beta (exp.) Beta (rev.) Burkina Faso -0.039 -0.106 Burundi 0.159 0.061 Cameroon 0.108 0.078 Chad -0.333 -0.078 Congo 0.878 0.152 Ethiopia 0.252 0.172 Gabon 0.200 -0.022 Gambia 0.495 0.154 Ivory Coast 0.105 -0.205 Madagascar -0.161 0.090 Mali -0.321 -0.090 Mauritius 0.136 -0.047 Namibia 0.517 0.172 Rwanda -0.115 0.012 Senegal -0.180 -0.008 Seychelles 0.517 -0.063 South Africa -0.027 0.022 Togo 0.516 0.275 Zambia 0.044 0.243 Zimbabwe -0.122 0.005 Sub-Saharan Africa 0.132 0.041 Argentina 0.016 0.015 Belize 0.179 -0.142 Bolivia 0.389 -0.315 Brazil 0.265 0.161 Chile 0.037 0.047 Colombia -0.211 -0.088 Costa Rica 0.206 0.010 Dominica 3.270 0.061 Dominican Republic 0.056 0.033 Guatemala 0.553 0.080 Haiti 0.155 -0.196 Honduras -0.424 -0.084 Jamaica -0.199 -0.323 Mexico -0.086 -0.121 Nicaragua 0.073 -0.035 Panama 0.203 0.116 Paraguay 0.033 -0.002 Peru -0.081 -0.092 St. Lucia -0.104 -0.094 Trinidad & Tobago 0.482 0.143 Uruguay 0.265 -0.036 Venezuela 0.141 -0.073 Latin America and Caribbean 0.093 -0.047 Country Beta (exp.) Beta (rev.) Australia -0.105 0.104 Austria 0.256 0.260 Belgium -0.488 -0.314 Canada -0.136 0.173 Denmark -0.469 0.354 Finland -0.285 -0.071 France -0.267 0.034 Greece 0.107 -0.067 Iceland 0.105 0.182 Ireland 0.137 -0.193 Italy -0.322 -0.022 Japan 0.154 0.612 Luxembourg -0.

45 643 -0.578 Netherlands -0.528 -0.263 Nor
643 -0.578 Netherlands -0.528 -0.263 Norway -0.299 0.371 Portugal 0.195 0.167 Spain -0.217 -0.127 Sweden -0.760 0.635 Switzerland -0.166 -0.071 United Kingdom -0.803 -0.304 United States -0.270 0.098 OECD -0.229 0.047 Bulgaria 0.544 0.561 Croatia 0.002 0.003 Hungary -0.155 0.207 Romania -0.560 0.158 Turkey 0.316 -0.015 Eastern Europe India -0.029 0.077 Indonesia -0.119 0.029 Malaysia -0.022 0.174 Nepal 0.005 0.025 Pakistan 0.270 0.108 Singapore -0.163 0.245 South Korea 0.013 0.089 Thailand -0.092 0.231 Asia and Pacific -0.017 0.122 Egypt -0.116 0.014 Iran 0.004 0.050 Morocco -0.157 -0.050 Tunisia 0.003 -0.008 Middle East and North Africa -0.066 0.001 Country (continued) Beta Initial Year Final Year Trinidad & Tobago -0.801 1976 1995 Uruguay -0.041 1965 1999 Venezuela -0.144 1974 1999 Latin America and Caribbean -0.132 Regression: OutputGapSurplus Country (continued) Beta Initial Year Final Year Morocco -0.032 1972 1995 Tunisia 0.021 1972 1999 Middle East and North AfricaBotswana 0.224 1982 1996 Burkina Faso -0.071 1973 1993 Burundi 0.016 1974 1999 Cameroon 0.017 1975 1998 Chad -0.044 1972 1991 Congo -0.124 1980 1998 Ethiopia -0.021 1981 1992 -0.384 1973 1991 Gambia 0.012 1966 1993 Ivory Coast -0.219 1979 1998 Madagascar 0.139 1972 1996 Mali 0.064 1977 1988 Mauritius -0.313 1974 1999 Namibia -0.243 1986 1993 Rwanda -0.035 1973

46 1993 Senegal 0.035 1972 1984 Seychelles
1993 Senegal 0.035 1972 1984 Seychelles -0.413 1985 1997 South Africa 0.133 1972 1998 Togo -0.739 1977 1987 Zambia 0.166 1972 1988 Zimbabwe 0.280 1976 1996 Sub-Saharan Africa -0.072 Argentina -0.011 1973 1996 Belize -0.262 1977 1997 Bolivia 0.278 1980 1997 Brazil -0.104 1960 1997 Chile -0.060 1960 1999 Colombia 0.087 1960 1999 -0.264 1970 1997 Dominican Republic 0.048 1960 1998 Ecuador -0.098 1960 1997 El Salvador 0.023 1960 1999 -0.461 1960 1998 Haiti 0.188 1979 1987 Honduras 0.033 1960 1997 Jamaica -0.517 1975 1985 Mexico -0.094 1971 1997 Nicaragua -0.145 1960 1996 -0.164 1980 1998 Paraguay -0.038 1960 1993 Peru -0.037 1965 1999 St. Lucia -0.058 1980 1991 St. Vincent & Grenadines -0.515 1980 1996 Country Beta Initial Year Final Year Australia0.319 1970 1996 Austria 0.192 1970 1996 Belgium 0.120 1970 1996 0.359 1969 1996 0.840 1970 1996 0.264 1960 1996 0.401 1972 1996 Germany -0.086 1991 1996 Greece 0.136 1960 1996 Iceland 0.240 1972 1996 -0.333 1960 1996 Italy 0.419 1960 1996 0.278 1960 1993 Luxembourg -0.267 1975 1996 Netherlands 0.209 1960 1996 New Zealand 0.206 1960 1996 Norway 0.536 1972 1996 Portugal 0.425 1975 1996 Spain 0.265 1962 1996 Sweden 0.788 1960 1996 Switzerland 0.078 1960 1996 United Kingdom 0.216 1960 1996 0.389 1960 1996 Bulgaria 0.432 1988 1997 Croatia 0.001 1991 1996 Hungary 0.248 1981 1997 Romania 0.140 1989 1997 Turkey

47 -0.262 1987 1997 Eastern Europe 0.112
-0.262 1987 1997 Eastern Europe 0.112 India -0.019 1965 1998 Indonesia 0.181 1972 1997 Malaysia -0.003 1960 1997 Nepal 0.013 1973 1999 Pakistan 0.095 1973 1998 Papua New Guinea -0.094 1975 1993 Philippines 0.043 1960 1998 Singapore -0.027 1975 1995 South Korea 0.002 1972 1996 Thailand 0.129 1960 1997 Asia and Pacific 0.032 Egypt -0.028 1975 1997 Iran 0.145 1974 1990 NBER WORKING PAPER SERIESWHY IS FISCAL POLICY OFTEN PROCYCLICAL?Alberto AlesinaGuido TabelliniWorking Paper 11600http://www.nber.org/papers/w11600NATIONAL BUREAU OF ECONOMIC RESEARCH1050 Massachusetts AvenueCambridge, MA 02138September 2005We thank Eliana La Ferrara, Roberto Perotti, Kenneth Rogo, Francesco Trebbi and participants in a seminarat Harvard for useful suggestions, Ashoka Mody and Diego Saravia for help with part of the data collection,and Filipe Campante for outstanding research assistantship and comments. We gratefully acknowledgefinancial support from NSF through a grant from NBER (Alesina) and CIAR, MIUR and Bocconi University(Tabellini) The views expressed herein are those of the author(s) and do not necessarily reflect the viewsof the National Bureau of Economic Research.©2005 by Alberto Alesina and Guido Tabellini. All rights reserved. Short sections of text, not to exceed twoparagraphs, may be quoted without explicit permission provided that full credit, including ©