When 100 taxation of pro t is available taxes on services supplied to rms should be deductible from pro t implying the optimality of a VATtype tax As for the rate of tax in the steady state an optimal arrange ment is to set it equal to the rate of t ID: 3918
Download Pdf The PPT/PDF document "How Should Financial Intermediation Serv..." 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.
HowShouldFinancialIntermediationServicesbe Taxed? ¤ BenLockwood y Firstversion:May312010 Thisversion:30September2013 Abstract Thispaperconsiderstheoptimaltaxationofsavingsintermediationservicesin adynamicgeneralequilibriumsetting,whenthegovernmentcanalsouseconsump- tion,incomeandpro ttaxes.When100%taxationofpro tisavailable,taxeson servicessuppliedto rmsshouldbedeductiblefrompro t,implyingtheoptimality ofaVAT-typetax.Asforthe rate oftax,inthesteadystate,anoptimalarrange- mentistosetitequaltotherateoftaxoncapitalincome, not consumption.In turn,thecapitalincometaxiszerowhenthewhenanunrestrictedpro ttaxis available,butinthemorerealisticcasewhensuchataxisnotavailable,thisrate canbepositiveornegative,butgenerallydi¤erenttotheoptimalrateoftaxon consumption. JELClassi cation :G21,H21,H25 Keywords : nancialintermediationservices,taxdesign,banks,paymentservices ¤ Thispaperisasubstantiallyrevisedversionofpartofanearlierpaperofthesametitle,Lock- wood(2010).IwouldliketothankSteveBond,MichaelDevereux,ClemensFuest,MiltosMakris,Michael McMahon,RuuddeMooij,DavidRonayneandseminarparticipantsattheUniversityofSouthampton, GREQAM,the2010CBTSummerSymposium,andthe2011IIPFConferenceforhelpfulcommentson anearlierdraft.IalsogratefullyackowledgesupportfromtheESRCgrantRES-060-25-0033,"Business, TaxandWelfare". y CBT,CEPRandDepartmentofEconomics,UniversityofWarwick,CoventryCV47AL,England; Email:B.Lockwood@warwick.ac.uk 1.Introduction Financialintermediationservicesincludesuchimportantservicesasintermediationbe- tweenborrowersandlenders,insurance,andpaymentservices(e.g.creditanddebitcard services).Theseservicescompriseasigni cantandgrowingpartofthenationaleconomy. Forexample, nancialintermediationservicesasconventionallyde nedinthenational accountsincludesactivitiessuchasthetakingofdepositsandthegrantingofcredit, nancialleasing,investmentinsecuritiesandpropertiesby nancialintermediaries,in- suranceandpensionfunding,andservicesancillaryto nancialintermediation 1 .TheEU KLEMSdatabaseshowsthatthissectorcomprised6.5%ofnationaloutputintheUKin 1997,increasingto7.9%by2007.The guresfortheUS,usingthesamede nitionof nancialintermediationservices,are7.3%in1997,risingto8.6%andfortheEurozone, 4.8%,risingto5.3% 2 .Evenexcludinginsurance-whichisbeyondthescopeofthispaper - nancialintermediationisquantitativelyimportantinOECDcountries 3 . Thequestionofwhether,andhow, nancialintermediationservicesshouldbetaxed isacontentiousone 4 .Forexample,withinEuropeanUnioncountries,most nancial servicesarecurrentlyexemptfromVAT,andthereisconsiderabledebateaboutthe possiblebene tsfrombringingthemintotheVATsystem(delaFeriaandLockwood (2010)).Also,therecentIMFproposalsfora"banktax"tocoverthecostofgovernment interventionsinthebankingsystemincludeaFinancialActivitiesTaxleviedonbank pro tsandremuneration,oneversionofwhich-FAT1-whichwouldworkverymuchlike aVAT,leviedusingtheadditionmethod(IMF(2010)). Inthepolicyliteratureonthistopic,itislargelyassumedthatwithinaconsump- tiontaxsystem,suchasaVAT,itisdesirabletotax nancialservicessuppliedatthe standardrateofVAT,andallowprovidersofintermediationservicestoclaimbackVAT theypayoninputs:seee.g.Ebril,Keen,Bodin,andSummers(2001)).However,this policyprescriptionisatvariancewithasmallacademicliteratureonthistopic(Grubert andMackie(1999),Jack(1999),andBoadwayandKeen(2003)),whichsuggeststhatwhile 1 Financialintermediationcomprisesactivities65,66,67intheISIC/NACEsystemofnationalaccounts. Thede nitionoftheseactivitiescanbefoundforexample,inthehandbookNACE:REV.1,published byEurostat. 2 Authors calculations: nancialintermediationcompriseslinesJ65-67intheEUKLEMSGrowth andProductivityAccounts(http://www.euklems.net/index.html). 3 Datashowingthesizeofintermediationservicesrelatingtothetakingofdepositsandgrantingof credit only arepresentedfortheUKinSection1.2below. 4 Therearetechnicaldi¢cultiesintaxing nancialintermediationservices;however,thesedi¢culties arenotinsurmountable-seeSection1.1below. 2 paymentservicesshouldbetaxedatthesamerateasconsumption,intermediationbe- tweenborrowerandlendershouldnotbetaxedatall.However,thisliteratureisbasedon rst-bestargumentsi.e. ndingthetaxarrangementthatdoesnotdriveawedgebetween thehouseholdmarginalrateofsubstitutionandthemarginalrateoftransformationin production. Theobjectiveofthispaperistotakeafreshlookatthisquestion,fromataxdesign pointofview.Ourfocusonthemostimportantintermediationservice-intermediation betweenborrowersandlenders 5 .Wesetupandsolvethetaxdesignprobleminadynamic generalequilibriummodeloftheChamley(1986)type,wherethegovernmentchoosesa taxonsavingsintermediation,aswellastheusualtaxesonconsumption(orequivalently, wageincome)andincomefromcapital,to nanceapublicgood,andwhere nancial intermediaries,intheformofbanks,areexplicitlymodelled.Realistically,weassume thatsavingsintermediationisnotexplicitlypriced,butchargedforviaaspreadbetween borrowingandlendingratessetbycompetitivebanks.Thisspreadcanbetaxed,atarate thatmaybedi¤erentfromthetaxonconsumption,andthetaxsystemisparameterized sothatsomefraction ofthetaxpaidby rmson nancialintermediationinputscanbe creditedagainsttheconsumptiontaxchangedby rms.InthecaseofaVAT, =1 Themainresultsareasfollows.First,thetaxpaidby rmson nancialintermediation inputsshouldbefullycreditedi.e.thetaxshouldbeaVAT,when100%taxationofpro t ispossible 6 .ThisisanexampleofthegeneralDiamond-Mirrleesproductione¢ciency resultthatintermediateinputsshouldnotbetaxedundertheseconditions.But,atwhat rateshouldtheVATbeset?Here,therearetwomain ndings.First,theoptimaltax structureisindeterminate,becausethegovernmenthastwoinstruments,acapitalincome taxanda nancialintermediationtax,tocontrolthemarginalrateofsubstitutionbetween consumptioninsuccessiveperiods.However,itturnsoutthatfromaninformationalpoint ofview,thesimplestoptimaltaxstructureiswherecapitalincometaxanda nancial intermediationtaxareequal. Inparticular,when100%taxationofpurepro tispossible,thesimplestoptimaltax structureistosetboththetaxrateoncapitalincome,andthetaxrateon nancial intermediationservicesequaltozero.Inthemorerealisticcasewhenthereisanupper boundontherateofpro toflessthan100%,weshowthatasimpleoptimaltaxstructure isagaintosetthetwotaxesatthesamerate.Thesignofthiscommonratethendepends 5 Istudytaxationofpaymentservicesinacompanionpaper,LockwoodandYerushalmi(2013). 6 Firmsmustmakepurepro tinequilibrium,becausetheymusthavedecreasingreturns,inturn becausetheypossiblyfacedi¤erentborrowingcosts. 3 onthepropertiesoftheproductionfunction;itcanbepositiveornegative.Moreover, thiscommonrateisgenerallydi¤erentfromtheoptimaltaxrateonconsumption. Theseresultsarequitedi¤erentfromtheexistingliterature(seeSection1.2)which generally ndsthat nancialintermediationservicesshouldbeuntaxed.However,these are rst-bestmodels,wherethereis(implicitly)norevenuerequirement.There,the optimalityofleaving nancialintermediationservicesuntaxedisderivedjustfromthe conditionthatthemarginalrateofsubstitutioninconsumptionisequaltothemarginal rateoftransformation.Ourresultsalsodi¤erfromAuerbachandGordon(2002),which nds,usingaratherdi¤erentargument,that nancialintermediationservicesshouldbe taxedatthesamerateasconsumption. Theremainderofthepaperisorganizedasfollows.Section1.1discussessomebasic factsaboutthetaxationof nancialintermediation,andSection1.2discussesrelated literature.Section2outlinesthemodelandSection3presentsthemainresults.Section 4considersthecasewithout100%pro ttaxation,Section5considersotherextensions, andSection6concludes. 1.1.TheSizeandTaxTreatmentofFinancialIntermediationServices The guresquotedatthebeginningofthepapermeasuretheoverallsizeofthe nancial intermediationsector.Togetanideaofthevalueof nancialintermediationassociated withthetakingofdepositsandthegrantofcreditonly,wecanlookatFISIM 7 .FISIM iscomputedfromthetransactionsbetweenthebankingsectorandothersectorsofthe economy(forexample,non- nancial rmsandhouseholds).Foreachofthesesectors, theloansfrom,anddepositswith,thebankingsector,aremeasuredandthemargins madebythebankingsectorontheseactivitiesarecalculated.Speci cally,themarginper currencyunitdepositedisareferencerateminustheaveragerateofinterestondeposits, andthemarginpercurrencyunitlentistheaveragerateofinterestonloans,minusa referencerate(Akritidis(2007)).So,FISIMcanbecalculatedbysector,andalsoonloans anddepositsseparately.Asourfocusisprimarilyontaxationofthehouseholdsector,we showconsumptionofFISIMbyhouseholdsandnon-pro tinstitutionsservinghouseholds fortheUKovertheperiod1997to2012. 7 Thisisanacronymfor" nancialintermediationservicesindirectlymeasured". 4 Figure1:FISIMConsumedbyHouseholdsasa PercentageofHouseholdConsumptionExpenditure,UK .0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 1997199819992000200120022003200420052006200720082009201020112012 Source:O¢ceofNationalStatistics,UK.ThechartshowsFISIMonloans(seriesIV8X) anddeposits(seriesIV8W)consumedbyhouseholdsandnon-pro tinstitutions servinghouseholdsasapercentageofaggregateconsumption(seriesRPQM) ThisshowsthatconsumptionofFISIMbyhouseholdsisbetween3%and4%oftotal householdexpenditureintheUKoverthisperiod;notlarge,butnotanegligiblefraction, either.Alsothe nancialcrisishashadaspositiveimpactonthis gure,asbankshave increasedtheirspreadsonloanstorepairtheirbalancesheets.So,overall,itcanbeseen that nancialintermediationservicesareasigni cantandgrowingpartoftheeconomy intheUK.ThepictureisthesameforotherOECDcountrieswithadeveloped nancial sector. Asregardsthetaxationof nancialintermediationservices,inourtheoreticalanalysis below,weassumethatintermediationservicescanbetaxed;speci cally,thatbankscan chargetaxestohouseholdsand rmsseparatelyontheirconsumptionofintermediation services.Itisrecognizedthatinpracticetherearetechnicaldi¢cultieswhenthoseservices arenotexplicitlypriced(so-calledmargin-basedservices),becauseitisnotstraightforward todividethe"value-added"betweenborrowerandlender.Inparticular,thisraisesa problemfortheuseofaVATviatheusualinvoice-creditmethod(Ebril,Keen,Bodin andSummers(2001)).Asaresult,thestatusquoinmostcountriesisthatawiderange 5 of nancialintermediationservicesarenottaxed.Forexample,intheEU,manysuch servicesareexemptasaresultofthe6thVATdirective 8 . However,conceptually,theproblemscanbesolvedinseveraldi¤erentways.One administrativelystraightforwardsystemwouldbetozero-ratesalestoVAT-registered entitiesandtaxsalestonon-registeredentitiese.g.householdsonanaggregatebasis (Huizinga(2002)).Alternatively,PoddarandEnglish(1997)haveproposedacash- ow VATwithtaxcalculationaccounts;thisisadministrativelymorecomplex,buttheincreas- ingsophisticationofbanks ITsystemsmeansthatthissolutionisbecomingpractical.A recentstudybytheEuropeanCommissioncalculatesthatEU-27taxrevenuemightrise byaround£15billionEuroifintermediationserviceswerebroughtintotheVATsystem, andtaxedatthestandardrate(EuropeanCommission(2011)). 1.2.RelatedLiterature Thereisasmallliteraturedirectlyaddressingtheoptimaltaxationofborrower-lender intermediationandpaymentservices,GrubertandMackie(1999),Jack(1999),andBoad- wayandKeen(2003).Usingasimpletwo-periodconsumption-savingsmodel,thesepapers agreeonapolicyprescription 9 .Givenaconsumptiontaxthatisuniformovertime,pay- mentservicesshouldbetaxedatthisuniformrate,butsavingsintermediationshouldbe leftuntaxed.Theargumentusedtoestablishthisissimple;inatwo-periodconsumption- savingsmodelwiththesame,exogenously xed,taxonconsumptioninbothperiods, thisarrangementleavesthemarginalrateofsubstitutionbetweencurrentandfuturecon- sumptionundistortedi.e.equaltothemarginalrateoftransformationinproduction. Usingadi¤erentapproach,AuerbachandGordon(2002)donotmakeasharpdistinc- tionbetweenpaymentservicesandsavingsintermediation,andarguethatbothactivities shouldbetaxedatthesamerateasconsumption 10 .Moreprecisely,theyshowthata wagetaxisequivalenttoauniformtaxonconsumptionandintermediationservices. 8 TheSixthVATDirectiveandsubsequentlegislationexemptsawiderangeof nancialservicesfrom VAT,includinginsuranceandreinsurancetransactions,thegrantingandthenegotiationofcredit,trans- actionsconcerningdepositandcurrentaccounts,payments,transfers,debts,cheques,currency,bank notesandcoinsusedaslegaltenderetc.(CouncilDirective2006/112/ECof28November2006,Article 135). 9 ChiaandWhalley(1999),usingacomputationalapproach,reachtheratherdi¤erentconclusionthat paymentservicesshouldbeuntaxed,butbuttheirmodelisnotdirectlycomparabletotheseothers,as theintermediationcostsareassemedtobeproportionaltothe price ofthegoodsbeingtransacted. 10 AuerbachandGordon(2002)state:"transactionscostscanincludetherealresources...lostwhen investingthesefundssothattheywillbeavailableinalaterperiod"(AuerbachandGordon(2002), p412). 6 However,onecanmakethreecriticismsofthisliterature.First,taxesaretakenas given.Inparticular,consumptiontaxesareassumedequalinbothperiods,andcapital incometaxesaresettozero.Combinedwiththe(implicit)assumptionof xedlabour supplyinthosemodels 11 ,thetaxsystem(otherthantaxeson nancialservices)amounts toanon-distortionarytaxonlabourincome.Inthissetting,itisofcourse,optimal formarginalrateofsubstitutioninconsumptiontobeequaltothemarginalrateof transformationinproduction.Itisthennotverysurprisingthatthetaxonborrower- lenderintermediationshouldbezero.Second,astaxesarenotdistortionary,thereisno second-besttaxdesignproblem,sothequestionoftradingo¤distortionsgeneratedby ataxon nancialservicesagainstotherdistortionarytaxesdoesnotarise.Third,the productionsideoftheeconomyisnotexplicitlymodelled,sothatquestionsofdistortions ininputpricescausedbytaxeson nancialservicescannotbeaddressed. Second,thereisalsoalesscloselyrelatedliteratureontheuseoftaxationtocontrol "badbanks".Theideahereisthatwhilebanksmayengageinsociallyundesirableactiv- itiesonbothlendinganddeposit-takingmargins,theseshouldbecorrectedbyPigouvian taxes(orregulations)thatapplydirectlytothesedecisionmargins.Therehasrecently beensurgeofliteratureonsuchPigouviantaxes;seee.g.Acharyaet.al.(2010),Bianchi andMendoza(2010),Keen(2010),PerrottiandSuarez(2011),Coulter,Mayer,andVick- ers(2012).Inourmodel,banksaremerelyproducersthatpriceintermediationservices atmarginalcost,sothereisnoroleforPigouviantaxes. 2.TheModel 2.1.Households ThemodelisaversionofAtkeson,ChariandKehoe(1999)withsavingsintermediation bybanksaddedtothebasicstructure.Thereisasinglein nitelylivedhouseholdwith preferencesoverlevelsofasingleconsumption-capitalgood,leisure,andapublicgoodin eachperiod =0 1 oftheform 1 X =0 ( ( )+ ( )) (2.1) where isthelevelofconsumptioninperiod 2 [0 1] isthesupplyoflabourhours, and ispublicgoodprovision.Utility ( ) isstrictlyincreasingin strictlydecreasing 11 TheexceptionhereisAuerbachandGordon(2002),wherelaboursupplyisvariable.However,in theirmodel,theconsumptiontaxisjustassumedtobeuniform,notoptimised. 7 in andstrictlyconcave,and ( ) isstrictlyincreasingandstrictlyconcavein Finally, 0 1 isadiscountfactor. Inanyperiod thehouseholdisassumedtopayanadvaloremtax on andalso paysproportionaltaxesonlabourandcapitalincome.Usingthewell-knownfactthat aconsumptiontaxisequivalenttoawagetax,weassumew.l.o.g.thatthewagetaxis zero Finally,forthemoment,wesupposethatthehouseholdhasnopro tincomeinany period: rmsgeneratepurepro ts(forreasonsexplainedinSection3.2below),butthese aretaxedat100%.So,inanyperiod,thebudgetconstraintis (1+ )+ +1 = +(1+ ) where isthepost-taxreturntothehouseholdonsavings,and isthewage,and +1 issavings.Finally, =(1 ¡ ) where isthepre-taxreturnonsavingsforthe household,determinedbelow,and isthecapitalincometax. So,followingAtkeson,ChariandKehoe(1999),thepresentvaluebudgetconstraintof thehouseholdcanbeobtainedbyaggregatingoverper-periodbudgetconstraints: 1 X =0 ( (1+ )+ +1 )= 1 X =0 ( +(1+ ) ) (2.2) where isthepriceofoutputinperiod Wenormalizebysetting 0 =1 andassume forconveniencethat 0 =0 i.e.initialcapitaliszero 12 The rst-orderconditionsfora maximumof(2.1)subjectto(2.2)withrespectto +1 respectivelyare: = (1+ ) (2.3) ¡ = (2.4) =(1+ +1 ) +1 (2.5) where isthemultiplieron(2.2),andweuse(hereandbelow)thenotationthatfor anyanyfunction ( ) ,thepartialderivativeof withrespectto is the cross-derivativeis etc. 2.2.Banks Forsimplicity,weassume100%depreciationofthecapitalgood.So,inthestandard versionofthismodel,without nancialintermediation,thehouseholdprovides unitsof 12 Thisimpliesthatthegovernmentcannotseta rst-periodcapitallevyon xedcapital 0 andthus simpli estheanalysis(seeAtkeson,ChariandKehoe(1999)). 8 theconsumption-capitalgoodtothe rmsinperiod ¡ 1 ,and rms(inaggregate)repay the unitsofconsumption-capitalgoodtothehouseholdinperiod plusinterest In ourversionofthemodel,the unitsoftheconsumption-capitalgoodaredepositedwith abankat ¡ 1 whocanthenprovidethisstock to rmsasaninputtoproduction. The rmsthenrepay tothebankat ,plusinterest,and nally,thebankrepays to thehouseholds,plusinterest. Thecostofintermediatingoneunitofsavingsbetweenthehouseholdand rm is ~ unitsoflabour.Notethatwetake ~ as xed,butpossiblyvaryingbetween rms.This isrealistic;lendingisacomplexprocessinvolvinginitialassessmentoftheborrowervia e.g.creditscoring,structuringandpricingtheloan,andmonitoringcompliancewithloan covenants(GupandKolari(2005,chapter9). Wealsosupposethatthetotalintermediationcost ~ canbedividedbythebank betweenthecostofservicesprovidedtothehousehold, (e.g.safekeepingofdeposits, liquidity)andcostofservicesprovidedto rm (e.g.monitoring), i.e. ~ = + .Thisisarealisticassumption,becauseinpractice,anyscheme(suchasthetax calculationaccountsofPoddarandEnglish(1997))whichimplementsaVATonmargin- based nancialintermediationservicesmustsplitthe"value-added"betweenborrower andlender.Weassumethatbankscanborrowandlendata"pure"rateofinterest whichwilleventuallybedeterminedinequilibriuminthecapitalmarket.Finally,we assumethatbanksareperfectlycompetitive,whichcombinedwiththeconstantreturns intermediationtechnology,impliesthattheymakezeropro tinequilibrium. Finally,weassumethatataxonintermediationservices,orspreadtax,isinoper- ationatrate whichcanbedi¤erentfromtherate on nalconsumption.Banks mustbreakevenoneachofthetwoactivitiesofprovidingservicestohouseholds,and to rmsseparately,otherwiseacompetitorbankcouldpro tablyundercutthem.So, inequilibrium,thecostsofintermediation plusthetaxpaid,areequaltothe spreads ¡ ¡ respectively,where istherateatwhich rm borrows,and istherateofreturnonsavingsforthehousehold.Thatis: ¡ = (1+ ) ¡ = (1+ ) (2.6) Thisisintuitive:thespreadonbothborrowingandlendingisequaltotherealresource costplusbythetax. 9 2.3.Firms Thereare rms =1 whichproducethehomogenousconsumptiongoodineachpe- riod 13 Firm producesoutputfromlabourandcapitalviathestrictlyconcaveproduction function ( ) where arecapitalandlabourinputs 14 .Because rmsmaydi¤er inintermediationcosts rmsfacedi¤erencesinthecostofcapitali.e. rm must repay 1+ perunitofcapitalborrowedfromthebank.So,thepro tof rm is ( ) ¡ ¡ (1+ ) + where isthetotaltaxpaidonintermediationservicesbythe rm,and isthe fractionofthetaxpaidonintermediationservicesthatthe rmcanclaimagainstthe taxpaidonoutput.Ofcourse,intheusualVATsystem, =1 So,substituting from (2.6)pro tcanbewritten ( ) ¡ ¡ (1+ + (1+(1 ¡ ) )) (2.7) Maximizing(2.7)withrespectto impliesthe rst-orderconditions: ( )= ( )=1+~ (2.8) where ~ = + (1+(1 ¡ ) ) isthecostofcapitalfor rm Finally,thecapital andlabourmarketclearingconditionsare: X =1 = X =1 + + X =1 = (2.9) Theseconditions(2.8),(2.9)jointlydetermine and ,givenhouseholdsavingsand laboursupplydecisions. 2.4.Discussion Theabovemodelprovidesageneralframeworkwhichencompassessomeaspectsofthe speci cmodelsoftaxationof nancialservices(AuerbachandGordon(2002),Boadway andKeen(2003)),Jack(1999),GrubertandMackie(1999))thathavebeendevelopedso 13 Wealsoassumeforconveniencethatoneunitoftheconsumptiongoodcanbetransformedintoone unitofthepublicgood.This xestherelativepre-taxpriceof and atunity. 14 Weassumethat rmsfacedecreasingreturns,becausewithdi¤erentcostsofcapital,andthesame wage,withconstantreturns,onlytheone rmwiththelowestunitcostwouldoperate,andthiscaseis oflimitedinterest. 10 far.Speci cally,ignoringpaymentservices,whicharenotdealtwithinthispaper,Boad- wayandKeen(2003)),Jack(1999),GrubertandMackie(1999)aretwo-periodversionsof theabovemodel 15 ,with xedtaxesand(implicitly) xedlaboursupply.Auerbachand Gordon(2002)isa nite-horizonversionofthemodel,withtheadditionalfeature 16 that thereare consumptiongoodsineachperiod.Paymentservicesaredealtwithina companionpaper,LockwoodandYerushalmi(2013). 3.TaxDesign Wetakeaprimalapproachtothetaxdesignproblem.Inthisapproach,anoptimal policyforthegovernmentisachoiceofalltheprimalvariablesinthemodel,inthiscase © +1 ( ) =1 ª 1 =0 tomaximizeutility(2.1)subjecttothecapitalandlabour marketclearingconditions(2.9),aggregateresource,andimplementabilityconstraints. Wearethusassuming,followingChamley(1986),thatthegovernmentcanpre-committo apolicyat =0 Theaggregateresourceconstraintsaysthattotalproductionmustequal tothesumoftheusestowhichthatproductionisput: + +1 + = X =1 ( ) =0 1 (3.1) Theimplementabilityconstraintensuresthatthegovernment schoicesalsosolvethe householdoptimizationproblem,andisinourmodel,quitestandard.Weobtainit bysubstitutingthehousehold s rst-orderconditions(2.3)-(2.5)in(2.2).Aftersome rearrangement,thisgivesthecondition: 1 X =0 ( + )=0 (3.2) Asisstandardintheprimalapproachtotaxdesign,wecanincorporatetheimple- mentabilityconstraint(3.2)intothegovernment smaximandbywriting = ( )+ ( )+ ( + ) (3.3) where istheLagrangemultiplieron(3.2).If · 0 i.e.consumptionandleisureare complements,itispossibletoshowthat ¸ 0 atthesolutiontothistaxdesignproblem 15 Aminorquali cationhereisthatBoadwayandKeenallowfora xedcostofsavingsintermediation e.g. xedcostsofopeningasavingsaccount.Theseintroduceanon-convexityintohouseholddecision- making,whichgreatlycomplicatestheoptimaltaxproblem,andsoweabstractfromtheseinthispaper. 16 Italsohaslaboursupplyinonlyoneperiod. 11 (seeAppendix).If =0 therevenuefrompro ttaxationissu¢cienttofundthepublic good, Wewillruleoutthisuninterestingcase,andsowillassumethat 0 atthe optimuminwhatfollows Thegovernment schoiceofprimalvariablesmustmaximize P 1 =0 subjectto(3.1) and(2.9).The rst-orderconditionswithrespectto are,respectively ; = (3.4) ¡ = (3.5) = ¡ 1 + (3.6) = (3.7) = + =1 (3.8) = =1 (3.9) where arethemultipliersontheresourceconstraintandthecapitalandlabour marketclearingconditionsattime respectively. Moreover,from(3.3), = (1+ (1+ )) = + (3.10) and = (1+ (1+ )) = + (3.11) Here, aregivenbystandardformulaefound,forexample,intheprimalapproach tothestatictaxdesignproblem(AtkinsonandStiglitz(1980)).Inparticular, ¡ 0 underourassumption · 0 anditsmagnitudemeasuresthedegreeofcomplementarity betweenconsumptionandleisure. Webeginbycharacterizingtherateoftaxontheconsumptiongoodviathefollowing result,whichisprovedintheAppendix: Proposition1. Atanydate =0 1 2 theoptimaltaxon nalconsumptioninad valoremformis 1+ = µ ¡ ¶µ ¡ 1+ ¶ = ¡ (3.12) Notethat(3.12)isaformulaforanoptimalconsumptiontaxthatalsooccursinthe staticoptimaltaxproblem,whentheprimalapproachisused(AtkinsonandStiglitz(1980, p377).Inparticular, isthemarginalbene tof$1tothegovernment,and isa measureofthemarginalutilityof$1tothehousehold,so ¡ isameasureofthesocial gainfromadditionaltaxationatthemargin.Itsaysthatotherthingsequal,thehigher 12 thedegreeofcomplementaritybetweenconsumptionandleisure, ¡ thehigheris Notealsothatbyourassumptionthat · 0 0 Nowweturntoconsiderthequestionofwhethertaxpaidon nancialintermediation shouldbedeductibleby rmsi.e.thechoiceof From(3.8),(3.9),weseethatthe marginalrateofsubstitutionbetweenlabourandcapitalis = + (3.13) whichimpliesthat ¡ = ¡ (3.14) However,fromthe rst-orderconditionsforthe rm,(2.8),weseethat = 1+ + (1+(1 ¡ ) ) (3.15) whichimpliesthat ¡ =( ¡ )(1+(1 ¡ ) ) (3.16) If 6 = and 6 =0 equations(3.14),(3.16)canonlyholdsimultaneouslyif =1 So, wehaveshown: Proposition2. Ifthereisheterogeneityinintermediationcosts, ( 6 = some ) and therateoftaxon nancialservices 6 =0 thenanydate e¢ciencyrequires =1 i.e. fulldeductibilityof by rms. Theintuitionforthisresultisclear.Equation(3.14)saysthatthemarginalproduct ofcapitalnetoftrueintermediationcostsshouldbeequalacross rms,whichofcourseis justtheconditionforcapitaltobeallocatede¢cientlyacross rms.But,condition(3.14) isgenerallynotconsistentwithanon-zero when rmsareheterogenous.Thisisjust aninstanceoftheDiamond-Mirrleesproductione¢ciencytheorem.Ataxonthebank marginisanintermediatetaxontheallocationofcapital,andgivenourassumptions(a fullsetoftaxinstruments,andnopurepro ts),thistaxshouldbesettozero.Note alsothatwhenthereisonlyone rm,thisargumenthasnobite,andthus isleft indeterminate. Wenowturntothequestionofhowthetaxesoncapitalincome, andoninter- mediationservices, shouldbeset.Generally,itcanbeshown,bystraightforward 13 manipulationofthe rst-orderconditionstotheoptimaltaxproblemandthehousehold and rmproblems,that: Proposition3. Atalldates =1 2 solve ¡ 1+ ¡ ¢ =1+(1 ¡ )( ¡ (1+ )) (3.17) where = 1+ (1+ ) 1+ (1+ ¡ 1 ) 1+ 1+ ¡ 1 TheproofofthisisintheAppendix.Clearly, arenotuniquelydetermined fromthissinglecondition.Ultimately,thisisbecausetheplannerhastwoinstruments, forcontrollingthemarginalrateofsubstitutionbetweenconsumptionat and +1 However,wecanusethefollowingcriteriontochoosebetweensolutions.Saythata solution to(3.17)is simple ifitsolves(3.17)independentlyoftheprecisevaluesof theeconomicdata Asthetaxauthorityisunlikelytoknowthesevalues,or atleasttosettaxesconditionalonthem,asimplesolutionisadministrativelyconvenient. Nowconsiderthesteadystate.Then, =1 andtherefore(3.17)becomes ¡ 1+ ¡ ¢ =1+(1 ¡ )( ¡ (1+ )) (3.18) Byinspection,theonlysimplesolutionto(3.18)is = =0 Theresultthattheinterest incometaxiszeroisofcourse,theclassicresultofChamley(1986)andJudd(1985);our newresultisthattherateoftaxon nancialservicesshouldbezero.So,wehaveproved: Proposition4. Atthesteadystate,theonlysimpletaxsystemiswherethetaxon interestincome, andthetaxon nancialintermediationservices, arebothzero. Twocommentscanbemadehere.First,awayfromthesteadystate, = =0 isgenerallynotoptimal.So,Proposition4-alongwithProposition6below-makes precisetheconditionsunderwhichtheresultoftheexistingliteraturethatsavingsinter- mediationshouldnotbetaxedgeneralizestoasecond-bestenvironment;zerotaxationof intermediationservicesrequires(a)unrestrictedtaxationofpro t,and(b)asteadystate. Second,thecelebratedresultofChamley(1986)andJudd(1985)thatinthesteadystate, thetaxoncapitalincomeiszerodoesnotholdpreciselyinourmodel,astheplanner hasanadditionaltaxinstrument, However,zerotaxationofcapitalemergesifwealso imposeadministrativesimplicity. 4.LessthanFullTaxationofPurePro t Recallthatwehavetoassumethat rmshavedecreasingreturns,becausetheyface di¤erentcostsofcapital.Therefore,theygeneratepurepro t.Sofar,wehavemade 14 thestrongassumptionthat100%taxationofthispro tispossibleforthegovernment. However,iswell-knownthatthisisakeyassumptionbehindtheclassicalDiamond- Mirrleesresultthatinputsarenottaxedatthesecond-bestoptimum.Here,weinvestigate towhatextentourresultsgeneralizetothemorerealisticcasewherepurepro tcannot betaxedat100%.Ideally,weshouldmodelthisviasomekindofincentiveconstraint formanagersorentrepreneursthatconstrainsapro ttax.However,thatisbeyondthe scopeofthispaper,andfollowingalargeliteratureintaxdesign,wejustassumethatthe pro ttaxis is xedatsome 1 . Themainchangetothetaxdesignproblemisthatnowpost-taxpro tappearsin thebudgetconstraintofthehousehold.Thispost-taxpro tcanbewrittenas (1 ¡ ) where isaggregatepre-taxpro t: = P =1 ( ¡ ¡ ) (4.1) andsoincomeinperiod isnow +(1+ ) +(1 ¡ ) Itcanthenbecheckedthat alsoappearsintheimplementabilityconstraintasfollows: 1 X =0 ( + ( +(1 ¡ ) ))=0 (4.2) From(4.1),andthefactthat = weseethat onlydependson = 1 So,theonly rst-orderconditionstothetaxdesignproblemthatchangeare(3.8), (3.9).Theychangeto: +(1 ¡ ) ( ) = + =1 (4.3) +(1 ¡ ) ( ) = =1 (4.4) The rstquestioniswhetheraggregateproductione¢ciencyholdsi.e.whether =1 Itiseasytocheck 17 that ( ) ( ) cannotbewritten respectivelyfor somecommonconstant Thisimpliesthatwecannotconcludethatitisoptimaltohave (3.13)holdingatthetaxoptimum.Thisimpliesinturn,thatwecannolongershowthat shouldbefullydeductiblei.e. =1 Thisisnotsurprising:generally,without100% taxationofpro t,andheterogenous rms,itiswell-knownthataggregateproduction e¢ciencydoesnothold. 17 Thiscanbeseenfromthefactthat dependson viatheterm ( ¡ )) ¡ ; di¤erentiationofthelatterwithrespectto generateexpressionsthatarenotcommonlyproportional to evenforspecialcasessuchastheCobb-Douglas. 15 But,withone rm,thequestionofwhetherthee¢cientallocationofcapitalacross rmsdoesnotarise.So,tolookatthekeyquestionoftheoptimalratefor weassume justone rm,sothatwecanabstractfromthequestionofwhether canbedeductible from rmcosts.Infact,withoutlossofgeneralitywecanassumefulldeductibility,sothe costofcapitalforthesingle rmis ~ = + wherewereplace by toemphasize thatthereisnowasingle rm.Inthiscase,wecanprovethefollowinganalogueof Proposition3.TheprooffollowstheproofofProposition3closely,exceptthatconditions (4.3),(4.4)replace(3.8),(3.9),andisthusomitted 18 . Proposition5. Atalldates =1 2 solve ¡ 1+ ¡ + ¢ =1+(1 ¡ )( ¡ (1+ )) (4.5) where isde nedinProposition3,and = (1 ¡ ) 1 1+ µ ¡ ¶µ ( + ) ( ) ¡ ( ) ¶ Notethatif100%pro ttaxationisavailablei.e. =1 then(4.5)reducesto(3.17). Asbefore, arenotuniquelydeterminedfromthissinglecondition.However,wecan proceedasabove,byfocussingonasolutionwheretheinterestincometaxandthespread taxareequal.Assumeasteadystate,sothat =1 .Also,assumethatthetwotaxes areequal 19 i.e. = (1+ ) Substitutingthisinto(4.5)gives 1+ = =(1 ¡ ) 1 1+ µ ¡ ¶ µ ( ) ¡ ( + ) ( ) ¶ (4.6) Thisisquiteanintuitiveformula.First,when100%pro ttaxationisavailablei.e. =1 , itreducesto = =0 consistentlywithProposition4.Second, isnon-zeroonly whenitissociallydesirabletotaxmorei.e. ¡ 0 Third,withouttransactionscosts, thesignof isthesameasthesignof ( ) Theintuitionisasfollows.Iftaxationisdistortionaryatthemarginthegovernment wouldliketotaxpro t(more),asitisanon-distortionarysourceoftaxrevenue;more precisely,itwouldliketoreduce asthisrelaxestheimplementabilityconstraint. Itcannotdothisdirectly.However,ifareductionin reduces thiscanbedone indirectlyviataxingcapitalincome.ThisisinlinewithresultsinStiglitzandDas- gupta(1971)forthecaseofastatic(one-period)economy.ItisalsosimilartoCorreia 18 Itisavailableonrequest. 19 Thespreadtaxisexpressedasapercentageofthecostofintermediationservices,grossofthetax, andthusmustbedividedthroughby 1+ tomakeitcomparableto 16 (1996),whoshows,inthecontextoftheChamleymodel,thatifanuntaxed(orincom- pletelytaxed)thirdfactorofproductioniscomplementarytocapitalintheproduction function,thencapitalincomeshouldbetaxedpositively 20 .Withtransactionscosts,an additionalterm ¡ ( + ) ( ) isadded.Fourth,takingtheratioof 1+ in(4.6),and 1+ in(3.12),andcancellingthecommonfactor 1 1+ ³ ¡ ´ wecanseethat and willdi¤er,althoughwecannotsaygenerallywhichwillbelarger. Togetafeelforwhen willbepositive,considertwocases.First,supposethatthe productionfunctionislinearinlabour,i.e. ( )= ( )+ Then,inthiscase,itis easilycheckedthat = ( ) ¡ 0 ( ) so ( ) ¡ ( + ) ( ) = ¡ 00 ( ) 0 andsotheinterestincometaxandspreadtaxarebothpositive.Ontheotherhand,if theproductionfunctionisCobb-Douglasi.e. ( )= + 1 then = (1 ¡ ¡ ) so ( ) ¡ ( + ) ( ) = ¡ ( + )(1 ¡ ¡ ) 0 andsotheinterestincometaxandspreadtaxarebothnegative.Tosummarize: Proposition6. Assumejustone rm,andthattheeconomyisinthesteadystate.Then, anoptimaltaxschemeistotaxinterestincomeand nancialintermediationservicesat thesamerategivenby(4.6).Thiscommontaxcanbepositiveornegative,dependingon thepropertiesoftheproductionfunction. 5.OtherExtensions 5.1.UnitaryTaxationofWageandCapitalIncome Anotherinterestingspecialcaseiswherethereisunitarytaxationofwageandcapital income.Thisisrelevantbecauseinpractice,manycountriestaxwageandnon-wage incomeinaunitaryway,accordingtoasingleprogressiveschedule 21 .Inthissimple 20 Theexactconditionsfor 0 aresomewhatdi¤erentheretoCorreia(1996),asweassumethatthe thirdfactorofproduction,whichgivesrisetopurepro t,itis xedsupply,whereasinCorreia(1996), itisinelasticsupply.Thelatterassumptionimposesanadditionalimplementabilitycontrsaintonthe optimaltaxproblem. 21 Awell-knownexceptionhereisthethedualincometaxsystemwhichleviesaproportionaltaxrate onallnetincome(capital,wageandpensionincomelessdeductions)combinedwithprogressivetaxrates ongrosslabourandpensionincome.Thedualincometaxwas rstimplementedinthefourNordic countries(Denmark,Finland,NorwayandSweden)throughanumberoftaxreformsfrom1987to1993. 17 model,withlineartaxes,unitarytaxationofincomesimplymeanstaxingbothwageand capitalincomeatthesamerate.Here,thereisnoexplicitwageincometax;itisimplicitly de nedviathebudgetconstraintvia = 1+ Thatis,ifthegovernmentreplaceda consumptiontaxatrate byawageincometaxatrate 1+ therealequilibriuminthe modelwouldbeunchanged.So,withunitarytaxation( = ) ,wecanreplace 1 ¡ by 1 1+ in(3.17)toget: (1+ ) ¡ 1+ ¡ ¢ =1+ +( ¡ (1+ )) (5.1) Thisismosteasilyanalyzedinthesteadystatewhen =1 Then,(5.1)canbeeasily solvedfor togivethefollowingresult: Proposition7. Supposethatthereisunitarytaxationofwageandnon-wageincome. Then,atthesteadystate,optimaltaxes satisfy = h 1 ¡ i So,weseethatwithunitaryincometaxation,thetaxrateonintermediationservices isproportionaltothetaxonconsumption,butisatalowerrate,andcouldbenegative. 5.2.EndogenizingSavingsIntermediationServices Wehave,sofar,treatedtheserviceofsavingsintermediationbybanksinrather black box fashion.Inparticular,wehavetreated theamountofintermediationservicesper unitofcapitalsuppliedto rm ,asexogenous.However,itisclearthatbankssupply severaldi¤erentkindsofintermediationservices,notablyliquidityservices(Diamondand Dybvig(1983)),andmonitoringservices(Diamond(1991),BesankoandKanatas(1993), HolmstromandTirole(1997)). Inthisversionofthepaper,wedonotattemptprovideafullymicrofoundedversionof thesekindsofintermediationservices,forseveralreasons.First,itistechnicallydi¢cult toembedsomeexplicitmodelsofintermediationservicesintothedynamicoptimaltax framework.Second,thepayo¤fromdoingsointermsofincreasedinsightsisnotreally proportionatetotheincreasedcomplexity.Intheend,bankintermediationactivity,when explicitlymodelled,may(ormaynot)havespilloversontherestoftheeconomy.Ifthere arespillovers,thentheoptimaltaxisaPiguovianonetointernalisethesespillovers. Ultimately,thisisbecausethegovernmentcanusetheinterestincometaxtocontrolthe household smarginalrateofsubstitutionbetweenpresentandfutureconsumption,and soanytaxonintermediationservicesisafreeinstrumentwhichcanbeusedtointernalize externalitiesarisingfrombankactivity. 18 Thesegeneralpointsareillustratedinapreviousversionofthepaper,Lockwood(2010), where isinterpretedasthelevelofbankmonitoring,alongthelinesofHolmstromand Tirole(1997).Intheirframework,withoutmonitoring,banklendingto rmsisimpossi- ble,becausetheinformationalrenttheydemandissohighthattheresidualreturntothe bankdoesnotcoverthecostofcapital.So,asmonitoringiscostly,thesociallye¢cient levelofmonitoringisthatlevelwhichjustinducestobanktolend.Inthecasewhere thebankiscompetitive,i.e.where rmchoosesthetermsoftheloancontractsubject toabreak-evenconstraintforthebank,anassumptioncommonlymadeinthe nance literature,thisisalsotheequilibriumlevelofmonitoring.Inthiscase,savingsinterme- diationshould not betaxed,becausedoingtowillviolateproductione¢ciency,asinthe casewithheterogenous rmsanda xedamountofintermediationservicesperunitof savings.But,inthecasewherethebankisamonopolisti.e.itchoosesthecontract,it willgenerallychoosea higher levelofmonitoringthanthis,inordertoreducethe rm s informationalrent.So,inthiscase,theoptimaltaxisapositivePigouviantax,setto internalizethisnegativeexternality. 6.Conclusions Thispaperhasconsideredtheoptimaltaxationof nancialintermediationservicesin adynamiceconomy,whenthegovernmentcanalsousewageandcapitalincometaxes. Theobjectiveofthispaperhasbeentotakeafreshlookatthisquestion,fromatax designpointofview.Wesetupandsolvethetaxdesignprobleminadynamicgeneral equilibriummodeloftheChamley(1986)type,wherethegovernmentchoosestaxessavings intermediation,aswellastheusualtaxesonconsumption(orequivalently,wageincome) andincomefromcapital,to nanceapublicgood,andwhere nancialintermediaries,in theformofbanks,areexplicitlymodelled. Ourmain ndingisthatatthesteadystate,theonlyadministrativelysimpleoptimal taxstructureistosetthetaxrateon nancialintermediationservicesequaltothetax rateoncapitalincome.When100%pro ttaxationisavailable,thiscommonrateiszero; inthemorerealisticcaseoflessthan100%pro ttaxation,thiscommonratecanbe positiveornegative,andisgenerallydi¤erentfromtheoptimaltaxonconsumption. Thereareseveralobviouslimitationsoftheanalysis.The rstandmostfundamental, isthattheroleofbanksisnotmicrofounded.Thisisclearlyatopicforfuturework: somepreliminaryresultsinthisdirectionaredescribedinSection5.2.Thesecondis therestrictiontolinearincometaxation.TheclassicresultofAtkinsonandStiglitztells usthatwithnon-linearincometaxation,commoditytaxationisredundant,andmore 19 recently,Golosovet.al.(2003)hasrecentlyshownthatthisresultgeneralizestoa dynamiceconomy.Theirresultwouldapply,forexample,inaversionofourmodelwhere householdsdi¤erinskilllevels,andwithoutany nancialintermediation.Itisatopicfor futureworktointroduce nancialintermediationinthisenvironment. 7.References Acharya,V.V.,L.H.Pedersen,T.Philippon,andM.Richardson,(2010)"Measuringsys- temicrisk,"WorkingPaper1002,FederalReserveBankofCleveland. Akritidis,L.(2007)"ImprovingthemeasurementofbankingservicesintheUKNa- tionalAccounts", EconomicandLabourMarketReview ,5,29-37 Atkeson,A.,V.V.Chari,andP.J.Kehoe(1999),"Taxingcapitalincome:abadidea", FederalReservebankofMinneapolisQuarterlyReview ,23:3-17 Atkinson,A.andJ.E.Stiglitz(1980), LecturesonPublicEconomics ,MITPress Auerbach,A.andR.H.Gordon(2002), AmericanEconomicReview ,92(Papersand Proceedings):411-416 Besanko,D.andG.Kanatas(1993),"Creditmarketequilibriumwithbankmonitoring andmoralhazard", TheReviewofFinancialStudies ,6:213-232 Bianchi,J.,andE.G.Mendoza,(2010),"Overborrowing,FinancialCrisesand"macro- prudential"taxes",NBERWorkingPaper16091 Boadway,R.andM.Keen(2003),"Theoreticalperspectivesonthetaxationofcapital and nancialservices",inPatrickHonahan(ed.) TheTaxationofFinancialIntermedia- tion (WorldBankandOxfordUniversityPress) 31-80. Chamley,C.(1986),"OptimalTaxationofCapitalIncomeinGeneralEquilibrium withIn niteLives", Econometrica ,54:607-622 Chia,N.C.,&Whalley,J.(1999).Thetaxtreatmentof nancialintermediation. JournalofMoney,Credit,andBanking ,704-719. Correia,I.H.(1996),"Shouldcapitalincomebetaxedinthesteadystate?", Journalof PublicEconomics 60:147-151. Coulter,B.,C.Mayer,andJ.Vickers(2012),"TaxationandRegulationofBanksto ManageSystemicRisk",unpublishedpaper,UniversityofOxford delaFeria,R.,andB.Lockwood(2010),"OptingforOptingIn?AnEvaluationof theEuropeanCommission sProposalsforReformingVATonFinancialServices", Fiscal Studies, 31(2):171-202 Diamond,D.W.(1991),"Monitoringandreputation:thechoicebetweenbankloans anddirectlyplaceddebt", JournalofPoliticalEconomy ,99:689-721 20 Diamond,D.WandDybvig,P.H.(1983),"BankRuns,DepositInsurance,andLiq- uidity," JournalofPoliticalEconomy ,91:401-19 Ebrill,L.,M.Keen,J-P.Bodin,andV.Summers(2001), TheModernVAT (Wash- ingtonD.C.:InternationalMonetaryFund) EuropeanCommission(2011),ImpactAssessmentaccompanyingthedocumentPro- posalforaCouncilDirectiveonacommonsystemof nancialtransactiontaxandamend- ingDirective2008/7/EC,SEC(2011)1102 nal,Vol.6(Annex5). Golosov,M.,N.Kocherlakota,andA.Tsyvinski,(2003),"OptimalIndirectandCapital Taxation", ReviewofEconomicStudies ,70:569-587 Grubert,H.andJ.Mackie(1999),"Must nancialservicesbetaxedunderaconsump- tiontax?", NationalTaxJournal, 53:23-40 Gup,B.E.andJ.W.Kolari(2005), Commercialbanking:TheManagementofRisk , 3rded.(Wiley) Huizinga,H.(2002), FinancialServices VATinEurope? EconomicPolicy ,17: 499-534 Holmstrom,B.andJ.Tirole(1997),"Financialintermediation,loanablefunds,and therealsector", QuarterlyJournalofEconomics ,112:663-691 InternationalMonetaryFund(2010), AFairandSubstantialContributionbytheFi- nancialSector ,FinalReportfortheG20 Judd,KennethL.(1985)"Redistributivetaxationinasimpleperfectforesightmodel." JournalofPublicEconomics 28.1:59-83. Jack,W.(1999),"Thetreatmentof nancialservicesunderabroad-basedconsumption tax", NationalTaxJournal, 53:841-51 Keen,M.(2010),"TaxingandRegulatingBanks",unpublishedpaper,IMF Lockwood,B.(2010),"HowShouldFinancialIntermediationServicesbeTaxed?," CEPRDiscussionPapers8122 Lockwood,B.andE.Yerushalmi,(2013),"TaxationofPaymentServices",unpublished paper,UniversityofWarwick Perrotti,E.andJ.Suarez(2011),"APigovianApproachtoLiquidityRegulation," In- ternationalJournalofCentralBanking ,vol.7(4),pages3-41,December. Poddar,S.N.andM.English(1997), TaxationofFinancialServicesUnderaValue- AddedTax:ApplyingtheCash-FlowApproach NationalTaxJournal, 50:89-111 Stiglitz,J.E.andP.Dasgupta(1971),"Di¤erentialTaxation,PublicGoods,and EconomicE¢ciency" TheReviewofEconomicStudies ,38,151-174. 21 A.Appendix Proofthat ¸ 0 Supposetothecontrarythat 0 attheoptimum.Then,as 0 · 0 0 from(3.10).So,from(3.10), = (1+ (1+ )) (A.1) Butfrom(3.5),(3.7),(3.9),(2.8): ¡ = = = (A.2) So,combining(A.1),(A.2),weseethat ¡ (A.3) But,(A.3)saysthatutilitycouldbeincreasedif1$ofspendingonthepublicgoodwere returnedtothehouseholdasalump-sum,contradictingtheoptimalityofthepolicy. ¤ ProofofProposition1. From(3.4),(3.5),(3.9),(2.8),wehave ¡ = ¡ 1+ (1+ ) 1+ (1+ ) = = 1 = 1 (A.4) And,from(2.3),(2.4): ¡ = 1+ (A.5) So,combining(A.4),(A.5),weget,aftersomerearrangement: 1+ = ( ¡ ) 1+ (1+ ) (A.6) Also,from(A.2),wehave: ¡ = ¡ (1+ (1+ ))= whichimplies = 1 1+ ¡ = ¡ (A.7) Combining(A.6),(A.7)toeliminate andrearranging,weget(3.12)asrequired. ¤ ProofofProposition3. From(3.4),(3.11),weget ¡ 1 ¡ 1 = 1 ~ ¡ 1 = ¡ 1 (A.8) 22 where ~ = (1+ (1+ )) (1+ (1+ ¡ 1 )) .Next,using(4.3),(4.4)toeliminate in(3.6),andalso using(3.7),weget ¡ = ¡ 1 + (A.9) So,thenfrom(A.9), ¡ 1 =( ¡ ( + )) (A.10) = ( ¡ ( + )) Butnow,from(2.8)and(3.15)with =1 wealsohave = 1+ + (A.11) Combining(A.10)and(A.11)gives ¡ 1 = ( 1+ ¡ ) (A.12) Next,combining(A.12)and(A.8),weget: ¡ 1 = ~ ( 1+ ¡ ) (A.13) Finally,from(2.3),(2.5),(2.8),weget: ¡ 1 = (1+ ) 1+ ¡ 1 1+ (A.14) = ¡ 1+(1 ¡ )( ¡ (1+ ¢ ) 1+ ¡ 1 1+ Combining(A.13),(A.14),andeliminating ¡ 1 weget ( 1+ ¡ )=1+(1 ¡ )( ¡ (1+ )) (A.15) where = ~ 1+ 1+ ¡ 1 = (1+ (1+ )) (1+ (1+ ¡ 1 )) 1+ 1+ ¡ 1 asrequired. ¤ 23