Corporate tax evasion with agency costs

www.elsevier.com/locate/econbase

Corporatetaxevasionwithagencycosts

KeithJ.Crockera,JoelSlemrodb,*

bThePennsylvaniaStateUniversity,UniversityPark,PA16802-3003

UniversityofMichiganBusinessSchool,701TappanStreet,Rm.A2120,AnnArbor,MI48109-1234Received15February2004;receivedinrevisedform27July2004;accepted9August2004

Availableonline18November2004

aAbstract

Thispaperexaminescorporatetaxevasioninthecontextofthecontractualrelationshipbetweentheshareholdersofafirmandataxmanagerwhopossessesprivateinformationregardingtheextentoflegallypermissiblereductionsintaxableincome,andwhomayalsoundertakeillegaltaxevasion.Usingacostlystatefalsificationframework,wecharacterizeformallytheoptimalincentivecompensationcontractforthetaxmanagerand,inparticular,howtheformofthatcontractchangesinresponsetoalternativeenforcementpoliciesimposedbythetaxingauthority.Theoptimalcontractmayadjusttooffset,atleastpartially,theeffectofsanctionsagainstillegalevasion,andwefindanewandpolicy-relevantnon-equivalenceresult:penaltiesimposedonthetaxmanageraremoreeffectiveinreducingevasionthanarethoseimposedonshareholders.D2004ElsevierB.V.Allrightsreserved.

JELclassification:H25;H26;D82

Keywords:Taxevasion;Enforcement;Agencycosts

1.Introduction

Recenthigh-profilecasesofcorporateaccountingfraudandtaxevasionhavegeneratedsubstantialchangesintherulesgoverningcorporategovernanceandaccounting,andintheprocessspawnedanextensivedebateregardingchangesintaxrulesandtheenforcement

*Correspondingauthor.Tel.:+17349363914;fax:+17347634032.E-mailaddress:jslemrod@umich.edu(J.Slemrod).

0047-2727/$-seefrontmatterD2004ElsevierB.V.Allrightsreserved.doi:10.1016/j.jpubeco.2004.08.003

1594K.J.Crocker,J.Slemrod/JournalofPublicEconomics(2005)1593–1610

ofexistingtaxlaw.1Intheaccountingsphere,theSarbanes–OxleybillpassedonJuly31,2002establishedaregulatoryboardappointedbytheSECtooverseetheaccountingindustry,creatednewlegalstandardsforprosecutingcorporatewrongdoing,requiredtopmanagementtocertifytheirfirms’financialstatementsandinternalcontrols,setforthlongprisonsentencesforexecutivesconvictedoffraud,andgavenewprotectionstocorporatewhistleblowers.2Inaddition,corporatetaxnoncomplianceisbynomeansatrivialissue.In1998,ayearinwhich(federal)corporatetaxreceiptswereUS$204.2billion,theInternalRevenueService(IRS)hasestimatedthatcorporateunderreportingwasUS$37.5billion.3Recently,anIRScontractorestimatedthetaxrevenuelossfromabusivetaxsheltersin1999tobebetweenUS$14.5toUS$18.4billion,whichis50%higherthanthelevelfor1993.4Inspiteoftheserecentpolicydevelopmentsandtheapparentincreasesincorporatetaxevasion,thereislittletheoreticalguidanceastotheimpactofalternativepenaltystructures,ortheappropriatestructureofpenalties,foreitheraccountingmisconductortaxevasion.

Thispaperexaminescorporatetaxevasioninthecontextofthecontractualrelationshipbetweentheshareholdersofafirmandthechieffinancialofficer(CFO),whodeterminesthefirm’sdeductionsfromtaxablecorporateincome.TheCFOisassumedtopossessprivateinformationregardingtheextentoflegallypermissiblereductionsintaxableincome,andmayalsoinflatethesizeofthefirm’staxshieldthroughillegalevasion.TheincentivesoftheCFOtoengageintaxevasionareaffectedbythenatureofhercompensationarrangement.Usingacostlystatefalsificationframework,wecharacterizeformallytheoptimal(informationallyconstrained)incentivecompensationcontractfortheCFOand,inparticular,howtheformofthatcontractchangesinresponsetoalternativeenforcementpoliciesimposedbythetaxingauthority.WefindthatpenaltiesimposedontheCFOdirectlyaremoreeffectiveinreducingevasionthanarethoseimposedonshareholders,andthattheoptimalcontractmayadjusttooffset,atleastpartially,theincentivesgeneratedbyincreasedsanctionsagainstillegalevasion.

Akeydifficultyinachievingasolidtheoreticalunderstandingofthedeterminantsofcorporatetaxevasionistheflexiblecontractualrelationshipthataffectsthebehaviorofthecorporatemanagers.Inacorporation,theshareholders,ortheBoardofDirectorsactingontheshareholders’behalf,willstructurethecompensationpackagesofmanagerstoprovideincentivesforthemtoactintheinterestofshareholders.Consider,forexample,thecompensationcontractfortheofficerinchargeofcorporatetaxes.Itisintheshareholders’interestforthetaxdirectortoreducethecompany’seffectivetaxburden,netofanycostsofdoingso,whichwouldincludeanyexpectedpenaltiesincurredduetodetectedtax

Throughoutthepaper,weusethetermbevasionQtorefertocorporatetaxreportingbehaviorthatwould,ifdiscovered,besubjecttocivilorcriminalsanctions.2Inthetaxarea,theInternalRevenueServiceannouncedin2002thatitwillreallocatemoreenforcementresourcestowardwealthytaxpayerssuspectedofhidingincomefromtheirbusinesses,partnerships,andinvestments.3Underreportingisonlyoneofthethreecomponentsofthetotaltaxgap,theothertwobeingnonfilingandunderpayment.Thereisnoestimateforcorporatenonfiling,andunderpaymentisaquitedifferentissue.4USGeneralAccountingOffice(2003,p.13).TheGAO(2003,p.1)definesabusiveshelterstobebverycomplicatedtransactionspromotedtocorporationsandwealthyindividualstoexploittaxloopholesandprovidelarge,unintendedtaxbenefits.Q

1K.J.Crocker,J.Slemrod/JournalofPublicEconomics(2005)1593–16101595

evasion.Toalignincentives,itmaybeappropriateforthetaxofficer’ssalarytodepend(inversely)ontheeffectivetaxrateachieved.Designingthiscontractiscomplicatedbythefactthatthetaxdirectorislikelytohaveprivateinformationabouttheavailabilityoflegalavenuesfortaxreduction,andmayalsolowertheeffectivetaxratethroughillegaltaxevasion.Ifdetectedbythetaxingauthority,however,suchillegalevasionmygeneratecoststoboththetaxofficerandthecorporation.

Thereisabundantevidencethatthefocusofcorporatetaxdepartmentshaschangedfromthatofpassivecompliancewiththetaxlawstoactive,aggressive,andoftenarguablyillegaltaxplanning.Forbesmagazinerecentlyreportedthatbwithencouragementfromshelterhustlers,anewattitudeisspreading:thatthecorporatetaxdepartmentisaprofitcenterallitsown,andthatahigheffectivetaxrateisasignofweakness.dApotentialclientoncesaidhewouldhirethefirmifwecouldgettheirtaxratedown,becauseitwashigherthantheircompetitors’andtheywereembarrassed,Tsaysoneaccountant.Q5Theideaoftaxdepartmentschangingfrombeingtoleratedasnecessarycostcenterstobeingcountedonasinnovativeprofitcentersisconsistentwithevidencefroma2001surveyofcorporatetaxdepartmentsinthemanufacturingsector.6Ofthevariousmeasuresusedtoevaluatetheperformanceoftaxdepartments,themostoftencitedwasthesavings,orvalueadded,theyprovided:86%citedthisperformancemeasure,upfrom75%in1997.Ofthose86%,63%saidthatthismeasureaffectedthecompensationoftaxdepartmentpersonnel.Theeffectivetaxraterelativetogoalwascitedasameasureusedtoevaluateperformanceby58%,upfrom48%in1997;ofthosethatusetheeffectivetaxratetoevaluateperformance,58%saiditaffectscompensation.ThenumberofmentionsofeachofthreepossibleperformancemeasuresthatincludedthewordbaccuracyQdeclinedsubstantiallybetween1997and2001.7Thegovernment,actingintheinterestofthetaxpayersgenerally,isanotherpartywithaclearstakeintaxcompliance.TheInternalRevenueServiceformulatesanenforcementpolicyregardingtaxevasionandfraudwiththeobjectiveofmaximizingitsperceptionofthepublicinterest.Thegoalofabenevolentgovernmentwouldbe,ceterisparibus,tominimizetherealresourcesusedupinthecomplianceprocess,andwouldvaluerevenuecollectionsatashadowpricereflectingthefactthattheyrepresenttransfersratherthanthecreationofresources.Theenforcementpolicywilllikelyinvolvepenalties,directedeitheratthecompanyorthecorporateofficer,fordetectednoncompliance.

ThemodernliteratureontaxevasionbeganwithAllinghamandSandmo(1972),whomodel(individual)taxpayersascompletelyamoral,decidingwhetherandhowmuchtoevadetaxesinthesamewaytheywouldapproachanyriskydecisionorgamble,asanexpectedutilitymaximizingchoice.Successfultaxevasionbenefitsthetaxpayerbecauseitsavesontaxes,butdetectedtaxevasionresultsinapenalty.Optimaltaxevasion,fromtheindividual’sstandpoint,dependsonthe(assumedtobefixed)chanceofgettingcaughtandpenalized,thesizeofthepenaltyforevasion,andtheindividual’sdegreeofriskaversion.Thirtyyearsofsubsequentanalysishasextendedthismodelinanumberofdimensions,

NovackandSaunders(1998).Hollingsworth(2002,pp.67–68).7SeeDouglasetal.(1996)foranearliersurveyoftaxdepartmentexecutivesthatfindsthateffectivetaxratesareoneofthemostfrequentlyusedformalperformancemeasurements.

651596K.J.Crocker,J.Slemrod/JournalofPublicEconomics(2005)1593–1610

includingallowinganendogenousprobabilityofdetection,analyzingevasionjointlywiththelaborsupplydecision,incorporatingsourcesofuncertaintyotherthanthechanceofaudit,andgeneralequilibriumconsiderations.

TheoverwhelmingmajorityoftaxevasionresearchhasfollowedAllinghamandSandmobyaddressingtaxevasiondecisionsmadebyindividuals.Thereis,tobesure,amuchsmallerliteraturethataddressestaxcompliancebyfirms,thebasicframeworkofwhichistheAllingham–Sandmosetupwithaunitarydecisionmaker.8Thenewtwistinthisstrandoftheliteratureistoexaminehowthetaxrate,probabilityofdetectionandpenaltyrateaffectthetwochoicesofevasion(usuallyexpressedasthefractionofsalesconcealedfromanoutputtax)andoutput,whenthereisacostlyconcealmenttechnology.Someofthemodelsinthistraditionassumethatthefirmisriskaverse,whileinothersthefirmisassumedtobehaveinarisk-neutralway.9Morerecently,ChenandChu(2002)investigatecorporatetaxevasionwithastandardprincipal-agentmodelinwhicharisk-neutralownerofafirmhiresarisk-aversemanager.10Theyfocusontheefficiencylossduetotheseparationofmanagementandcontrol,anddonotaddresstherelativeefficacyofpenaltiesontheprincipalandagent,akeyfocusonthemodelwepresentbelow.AsidefromChenandChu(2002),alloftheprecedingliteratureassumesthatthefirmowner,orresidualclaimant,makesthetaxreportingdecisionwithnoagencyconsid-erations.11Thisassumptionmakessensewhenoneisanalyzingsmall,closelyheldbusinesses.However,inalarge,publiclyheldcorporation,decisionsabouttaxes(andaccounting)arenotmadebytheshareholdersdirectlybut,rather,bytheiragents,whetherthatisthechieffinancialofficerorthevicepresidentfortaxation.Inordertoaligntheincentivesofthedecisionmakersandtheshareholders,thecorporationhastheincentivetotietheagent’scompensationtoobservableoutcomesthataffectafter-taxcorporationprofitability.

InthissettingtheinsightsgeneratedbytheAllingham–Sandmomodelmaynotapply.Forexample,ifpenaltiesforevasionapplytotheagent,theprincipalcanalteritscompensationcontractwiththeagent,possiblyoffsettingtheintendedconsequencesoftheIRSpolicy.Moregenerally,enforcementstrategiesdirectedatthetaxdirectorandatthe

ThisliteratureisnicelyreviewedinCowell(2004).

Inthelattercase,Lee(1998)showsthatprofittaxesneednotbeneutralinthepresenceoftaxevasion.Thefirmmaybeoperatinginaperfectlycompetitivemodel,orbeamonopoly;inthelattercase,undersomeconditions,therecanbeovershifting,wheretheproductpricerisesmorethantheincreaseintax.10ThebasicframeworkisthatofHolmstrom(1979)inwhichtheowner(principal)designsawagecontracttoincentivizethemanager(agent)toengagein(privately)costlyeffortthatincreasesfirmprofit.Thetraditionalmodelismodifiedtopermittheownertoengageintaxevasion.Theirprimaryconclusionisthat,whenthemanagerispenalizedfor(detected)taxevasion,theoptimalwagecontractresultsinaninefficient(relativetothesecond-bestbenchmarkinwhichevasionisnotpossible)levelofmanagerialeffort.Thisdistortionarisesbecauseofthemaintainedassumptionthatthemanager’scompensationcannotbebasedontheowner’schosenlevelofevasion,sincesuchacontractwouldbepredicatedonanillegalactand,therefore,wouldnotbeenforceableincourt.Desaietal.(2003)analyzeanotherinteractionbetweencorporatetaxationandcorporategovernance.Theypositthatunreportedincomeismoreeasilyexpropriatedbymanagers,andshowthatinthissituationahighertaxratemayincreasethelevelofmanagerialdiversion,whilestrongertaxenforcementmayreduceit.11Aliteratureinthelawandeconomicsfieldinvestigatesthesociallyoptimaldivisionofsanctionsoncorporationsandindividualemployeesforsocialharmsgenerally.Forexample,Kraakman(1984)emphasizesthepossibilitythatthecorporation’sassetsareinadequatetopayfortheharm.PolinskyandShavell(1993)arguethatthetotalmagnitudeofpublicsanctionsmayexceedthesanctionsthatafirmcanimposeonitsemployees.

98K.J.Crocker,J.Slemrod/JournalofPublicEconomics(2005)1593–16101597

corporationitselfmayhavedifferentimpactsoncorporatebehavior.Becauseeachofthesepoliciesisavailabletothegovernment,12itisvaluabletoknowwhetherthereisatheoreticalreasontopreferonetotheother.Themodelwedevelopbelowprovidesaframeworkforanalyzingthisandotherrelatedquestions.

Thepaperproceedsasfollows.Webegininthenextsectionbyprovidingaformalcharacterizationofthecontractthattheshareholderswillwanttooffertothetaxofficerwithinthecorporation.OurmodeldrawsonthecostlystatefalsificationframeworkthatwasfirstconsideredbyLackerandWeinberg(19),anddevelopedbyCrockerandMorgan(1998)tocharacterizetheoptimalinsurancecontractsinasettingwhereclaimantscanengageinfraudulentbehaviortoinflatetheirapparentlosses.13Next,weshowhowthe

TheUStaxlawhasprovisionsthatimposecriminalandcivilpenaltiesonboththecorporationandcorporateofficers.IndividualswhosignfalsetaxreturnsmaybesubjecttocriminalpenaltiesunderCodesection7206(1).Section7206(1)imposescriminalliabilityinthecaseofawillfullyfalsedeclaration.Becausesections6062and6065combinetorequireacorporateofficertosignthecorporation’staxreturnunderpenaltiesofperjury,section7206(1)meansthatacorporateofficermayfacecriminalliabilityforsigningataxreturnthatisinaccurateiftheofficerknowsthereturnisinaccurateandintendstoviolatethelawinsigningthereturn.Criminaltaxprosecutionsare,however,fairlyrare.Infiscalyear,therewere512suchprosecutions,abouthalfofthe1993total.SeparateInternalRevenueCodeprovisionsimposecivilliabilityforviolationsoftaxreportingrules.Section6694providesformonetarypenaltiesforreturnpreparerswhofilereturns(i)showinganunderstatementoftaxliabilitythathasnorealisticpossibilityofbeingsustainedonthemeritsor(ii)withthewillfulattempttounderstatetaxliabilityortherecklessorintentionaldisregardoftaxrules.Sections6700and6701imposepenaltiesforpromotingabusivetaxsheltersandaidingandabettingunderstatementsoftaxliability.Section6707imposescivilfinesonpersonswhoarerequiredbutfailtoregisterordisclosetaxsheltertransactions.Corporateofficersintheorycouldbesubjecttoanyofthesepenalties,butthepenaltiesaremorelikelytobeimposedonoutsideprofessionals.Morebroadly,taxpayersthatunderstatetheirtaxliabilitymaybesubjecttopenaltiesfortheunderstatement.Section6662oftheCodeimposesapenaltyontaxpayersfor,amongotherthings,anybsubstantialunderstatementQofincometax.Forcorporatetaxpayersasubstantialunderstatementtypicallyisanunderstatementthatexceeds10percentofthetaxrequiredtobeshownonthereturnfortheyearinquestion.Thepenaltyforasubstantialunderstatementis20percentoftheamountoftheunderstatement.Thesubstantialunderstatementpenaltycanthereforebesignificant,butthepenaltyisnotimposedoncorporateofficers,butratheroncorporationsthemselves.13Intheinsuranceenvironment,theoptimalindemnitycontractresultsinunderpaymentofclaimsatthemargin,andreflectsatensionbetweenthebenefitsofunderpayment(reducedincentivesoftheinsuredtoexpendresourcesinclaimsinflation)andthecosts(lessincomesmoothingfortheriskaverseinsured).Fraudhasvalueintheoptimalcontractbecauseitprovidesamechanismbywhichtheinformationallyconstrainedinsurercansorttheclaimantsbasedontheirunderlyingprivateinformation(thatis,themagnitudeofthelossactuallysuffered).CrockerandMorganalsoconsideramodelofcontractingbetweenariskneutralprincipal,andariskneutralagentwhomayengageincostlystatefalsification.Theprecisesettingconsideredisthatofsharecropping,wheretheabsenteelandlord(principal)contractswiththesharecropper(agent)overthedivisionofacrop.Costlystatefalsificationarisesbecausethesharecroppercan,atsomeresourcecosttohimself,hidepartofthecropfromtheabsenteelandlord.Inthissetting,theincentivestoengageinfalsificationmaybemitigatedbybflatteningQthesharecroppingcontract(asafunctionoftheobservedcrop),whichreducesthemarginalreturntothesharecropperoffalsification.Thisstrategy,however,hinderstheprincipal’sabilitytoextractsurplusfromtherelationship,sincetheoptimalcontractmustrespectthesharecropper’sbankruptcyconstraint.Thus,theoptimalcontractreflectsatradeoffbetweenmitigatingfalsification(whichincreasestotalsurplus)andsurplusextractionbytheprincipal.Inataxationsetting,theCrockerandMorganmodelisequivalenttotheproblemfacingarisk-neutraltaxingauthorityattemptingtomaximizetherevenueyieldfromarisk-neutraltaxpayerwhomayexpendresourcesinhidingincomethroughtaxevasion.Ourconcerninthispaper,however,isdifferent,asitconcernsthemodelingoftheemploymentrelationshipbetweentheshareholdersofthefirm,andtheadministrativeofficerhandlingthefirm’staxavoidancestrategy,whohastheabilitytoevadetaxesthroughcostlystatefalsification.

121598K.J.Crocker,J.Slemrod/JournalofPublicEconomics(2005)1593–1610

formofthatcontractwillchangeinresponsetoIRSenforcementpolicies,andexaminetowhatextentthatresponsemaycounteracttheintentionoftheIRS.Wedothisfirstinafairlygeneralsettinginwhichthetaxofficerhasaccesstoacostlyconcealmenttechnologyandthen,inthefollowingsection,inthecontextofaspecificparameterizationoftheproblem.Wefindthat,asaconsequenceoftheinformationalasymmetryenjoyedbytheCFO,penaltieslevieddirectlyonthetaxofficeraremorelikelytobeeffectiveinreducingtaxevasionthanpenaltiesdirectedatthecorporation’sshareholders.Theintuitionbehindthisbnon-equivalencyQresultisthattheincentiveeffectsofsanctionsleviedontheshareholdersareimperfectlytransmittedtothetaxofficerthroughasecond-bestcompensationcontract,sothatpenaltiesdirectlyappliedtothetaxofficerhavemoreimpact.Thefinalsectioncontainsconcludingremarks.

2.Themodel

Weconsideranenvironmentinwhichtherisk-neutralownersofafirm(thebshareholdersQ)contractwiththevicepresidentfortaxationorchieffinancialofficer(bCFOQ)tomanagethefirm’staxliability.14TheCFO,whoisalsoriskneutral,isassumedtopossessprivateinformationregardingthelevelofpermissiblelegalreductionsintaxableincome,x.15Theshareholdersknowonlythattheactualvalueofxisdistributedaccordingtothedensityf,andthecumulativedistributionF,ontheinterval[x,x¯].Whilexisnotobservabletotheshareholders,theydoknowtheactualreductionsintaxableincome,R,claimedbytheCFO,whomayengageintaxevasionbyclaimingillegaldeductions,R-x.16Suchtaxevasionmayresultinpenaltiesleviedoneithertheshareholdersofthefirm,ontheCFOpersonally,oronboth,inafashionthatwewilldescribebelow.

AcompensationagreementbetweentheshareholdersandtheCFOisacontractCu{S,R}consistingofthepaymentSandthelevelofdeductionsfromtaxableincomeR,bothofwhichmaybeconditioneduponxifcertainincentiveconditionsaresatisfied.17Wemaywritetheprofitaccruingtoshareholdersas

jðCjxÞuIÀtðIÀRÞÀSÀahðRÀxÞ

14ð1Þ

OnecanimagineamorecomplicatedmodelinwhichtheshareholdershiretheCEO,whointurnhirestheCFO,withtheCFOpossessinginformationthatishiddentotheCEOandtheCEOpossessinginformationthatishiddentotheshareholders.15TheassumptionofriskneutralityoftheCFOisonekeydifferencebetweenthismodelandtheoneofferedbyChenandChu(2002),whoseresultsrequireriskaversiononthepartofthefirm’smanager.Weconsidertherisk-neutralitycasenotasasimplifyingassumptionbut,rather,todemonstratethatexplicitdifferencesinriskpreferencesbetweentheprincipalandtheagentarenotnecessaryinorderforarationalagenttoengageintaxevasion.16Tobeprecise,thedistinctionbetweenxandR-xisthatthelatteraresubjecttopenaltiesandtheformerarenot.Notethat,inreality,eventaxableincomeandtaxpaymentarenotgenerallyknowntoshareholdersoflargepubliccorporationswhodonothaveaccesstothetaxreturnsthemselves;thefinancialstatementsdonotrevealthisinformation.SeeHanlon(2003)foradiscussionofthisissue.Whatisessentialforthismodelisthattherearecontractableelementsoftaxreportingbehavior.17NotethatweareultimatelyinterestedintheformoftheoptimalCFOcompensationcontract,S(R).Sincethisisanenvironmentwithprivateinformation,however,wewillproceedintheusualfashiontocharacterizetheoptimalbtruth-tellingQcontract{S(x),R(x)}thatsatisfiestheincentiveconstraint[Eq.(3)]below,fromwhichS(R)mayberecoveredbyinvertingRtoobtainx,andsubstitutingtheresultintoS.

K.J.Crocker,J.Slemrod/JournalofPublicEconomics(2005)1593–16101599

whereIisthe(observable)incomeofthefirmbeforeeitheravoidanceorevasion,tisthecorporatetaxrate,andahembodiestheexpectedcoststoshareholdersofillegaltaxevasion.18ThepayofftotheCFOisdenoted

VðCjxÞuSÀbgðRÀxÞ

ð2Þ

wherebgaretheexpectedcoststotheCFOofengagingintaxevasion.Weassumethat,foreachparty,tocommitnoillegaltaxevasion(R=x)iscostless,andthatthepenaltiesforsuchevasionareincreasingintheamountofevasion,R-x.19Assumption1.h(0)=g(0)=hV(0)=gV(0)=0;andhV,gV,hW,gWN0forR-xN0.

Intheanalysisthatfollows,wewillassumethatthepenaltiesfortaxevasionaresymmetric,sog(d)=h(d).Thisassumptionallowsustofocusourattentionontheimplicationsofdifferentialseverityofthetwotypesofpenalties,andnotbedistractedbyissuesoftheoptimalfunctionalformofthepenaltystructure.Moreover,wewillinterprettheparametersaandb,eachofwhichis,withoutlossofgenerality,restrictedtobeintheunitinterval[0,1],tobebenforcementparametersQthatareultimatelyselectedbythetaxingauthority.20Sincethisisanenvironmentofprivateinformation,therevelationprinciple(Myerson,1979)impliesthatanyimplementablecontractmustsatisfytheincentiveconstraint

ˆÞ;RðxˆÞjxÞVðSðxÞ;RðxÞjxÞzVðSðx

ð3Þ

ˆa[x,xforeveryx,x¯]inordertogeneratetruthfulrevelationbytheCFO.Whenthis

incentiveconstraintissatisfied,aCFOwhopossessestheprivateinformationxwould

Notethatweareassumingthatthelevelofpermissiblereductions,x,isindependentofincomeI.ThisassumptionisinnocuoussincewehavealsoassumedthatIhasbeenrealized(soanyproductionhasalreadyoccurred)priortotheCFOmakingthetaxevasiondecision.WealsoaddressherethepointmadefirstbyYitzhaki(1974)thatifthepenaltyisproportionaltothetaxevadedratherthan,ashere,theincomeevaded,theimplicationsoftheeffectofataxratechangeonevasionarequalitativelydifferent.Theeffectofthetaxrateonevasionisnot,though,afocusofthisresearch.WeabstractfromthefactthatSwouldgenerallybedeductiblefromthetaxableincomeofthefirmandtaxabletothetaxofficer.(Penaltiesarenot,though,deductiblebusinessexpenses.)Finally,notethatthepenaltyfunctionscouldbegeneralized,sothatforexampleh(R-x)couldbeh(I,R,x).Thiswouldallow,interalia,theexpectedcosttodependontheratioofevasiontotrueincome,leadingtoapredictionthatbiggerfirmswouldevademore.Ourbasicconclusionswouldbeneitheraffectednorenrichedbythisgeneralization.SeeSlemrod(2001)foradiscussionoftheimportanceofthefunctionalformofthecostfunction,orthetechnologyoftaxavoidanceandevasion,forunderstandingtheimpactonbehaviorofchangesinthetaxrate.19ThisisasettingofcostlystatefalsificationbecausetheCFOpossessesanimmutableinformationalasymmetry(theknowledgeoftheactualvaluetakenbyx),andcanbmisrepresentQtheallowabletaxdeductions(byselectingRNx)byincurringtheprivatecostbg.ThisisincontrasttothecostlystateverificationapproachthatwasfirstdevelopedbyTownsend(1979).Inthatsetting,theuninformedagentcanperforma(privatelycostly)audittoobtaintheprivateinformation.20Oneinterpretationwouldbetothinkofa(b)asbeingtheprobabilitythattaxevasionisdetectedandthepenaltyh(g)isleviedonshareholders(theCFO).Thesizeoftheparametersaandbisdeterminedbythelevelof(costly)enforcementactivitiesselectedbythetaxingauthority.Therelativeeffectivenessofthesetwotoolsistheobjectofouranalysisbelow.Notethatwealsoabstractfromwhetheritisoptimalforthetaxenforcementagencytoconditiontheprobabilityofdetection(i.e.,theparametersaandb)onthereportedincomeofthecorporation.

181600K.J.Crocker,J.Slemrod/JournalofPublicEconomics(2005)1593–1610

alwayspreferthecontract{S(x),R(x)}overthealternatives{S(xˆ),R(xˆ)}foreveryxˆpx.Thisobtainswhenthefirst-orderconditionforacontracttobeincentive-compatibleissatisfied,whichisgivenby

ˆÞ;RðxˆÞjxÞdVðSðx

¼VSSVþVRRV¼0

dˆx

ð4Þ

atxˆ=x.TotaldifferentiationofVwithrespecttox,andsubstitutingfrom[Eq.(4)],yields

theresultthat

dV

¼Vx:dx

ð5Þ

AnefficientcontractbetweentheshareholdersandtheCFOisasolutiontotheproblemthatmaximizestheexpectedpayofftotheshareholders

Zx¯

PðSðxÞ;RðxÞjxÞfðxÞdxð6Þmax

SðxÞ;RðxÞ

x

subjecttotheincentivecompatibilityconstraint[Eq.(3)]andaparticipationconstraintfor

theCFO,whichrequires

VðSðxÞ;RðxÞjxÞz0

ð7Þ

21foreveryxa[x,x¯].

Beforeproceedingtocharacterizetheoptimalcontractinthisinformationallyconstrainedenvironment,itisinsightfultoconsiderfirstthebaselinecaseinwhichthereisnoincentiveproblemcausedbyprivateinformation.ThiswouldbethecaseweretheshareholderstoobservexandselecttheirpreferredlevelofR,subjecttotheparticipationconstraint[Eq.(7)].Inthiscase,thetax-reportingbehaviorwouldnotbeconstrainedbytheincentivecondition[Eq.(3)],andthecompensationoftheCFOwouldbealump-sumpaymentthatsatisfiedtheparticipationconstraint[Eq.(7)].Underthesecircumstances,taxevasionbehaviorwouldmaximizeshareholderprofitPforeveryx,subjectto[Eq.(7)],andtheresultingfullinformationoptimalamountoftaxevasionischaracterizedbythecondition

t¼ðaþbÞgVðRÀxÞ:ð8Þ

Weshalldenoteasolutionto[Eq.(8)]byR*(x).

Intheenvironmentconsideredinthispaper,however,theselectionofRmustbedelegatedtotheCFO,whopossessesprivateinformationregardingthelegallypermissible

ThisconstraintrequiresthatthecontractrespecttheparticipationconstraintforanypossibleprivateRx¯

information,x,thattheCFOmayhave.Asanalternative,onecouldimposetheweakerconstraintxVfdxz0,sothattheCFO,priortoreceivingherprivateinformation,expectstosatisfytheparticipationconstraint[Eq.(7)].Suchanapproachwouldpermittheattainmentoffullinformationoptimallevelsoftaxevasion(seethediscussionbelow),sincethefirmcouldbeeffectivelybsoldQtotheCFO,andalltheexpectedsurplusextractedbyalump-sumtransfertotheshareholders.Theproblemwiththisapproachisthat,sinceVxN0,theresultingcontractwouldprovidetheCFOwiththeincentivetoexittheagreementforlowervaluesoftheprivateinformationparameter,x.

21K.J.Crocker,J.Slemrod/JournalofPublicEconomics(2005)1593–16101601

deductionsfromtaxableincome,x.Sincetheactualvalueofxisunknowntotheshareholders,thefullinformationoutcomeviolatestheincentiveconstraint[Eq.(3)]and,therefore,cannotbeachieved.Toseethis,notethat,sinceevasiongeneratescoststotheCFO,withthelump-sumcontractdescribedabovetheCFOwouldengageinnoevasionatall.Alternatively,theshareholderscouldinducetheCFOtochooseR*byofferingthecompensationplanS(R)=tR-ah(R-x),whichisthenetgaintotheshareholdersfromevasion,butsuchanapproachwoulddrainallofthesurplusfromtheshareholdersandleavethemwithafter-taxprofitsofonly(1-t)I.22Theshareholderscoulddobetterbyofferingtheoptimalincentive-compatiblecontractthatwecharacterizeintheanalysisbelow,butundersuchacontractR*isgenerallynotattainable.Theoptimalincentive-compatiblecontractmaximizestheexpectedafter-taxpayofftotheshareholders,subjecttotheconditionthatthecontractmustbeincentive-compatible,andthattheCFOmustreceiveatleastherreservationlevelofutilityinanystateoftheworld,x.

TheHamiltonianexpressionassociatedwiththeproblemofmaximizing[Eq.(6)]subjectto[Eq.(3)]and[Eq.(7)]maybewrittenas

H¼PðS;RjxÞfþ/ðxÞVx;

wherewedefinethestatevariabletobeV(x)uV(S(x),R(x)|x),thecontrolvariabletobeR(x),and/(x)tobethecostatevariablefortheequationofmotionVx.

Theorem1.Anoptimalcontractmustsatisfythefollowing(necessary)conditions:f[t-(a+b)g’]+/bgW=0;andR¯x

S(x)=bg+xbgV(R(m)Àm)dm,

¯

where/=F-1.

˙=BH/BV,whichyieldProof.ThePontryaginconditionsforamaximumareHR=0and/

󰀃󰀄󰀃󰀄

BSBS

þPRþ/VxRþVxS¼0;andð9ÞfPs

BRBR˙¼ÀðfPsþ/VxSÞBS;/

BVð10Þ

(a)(b)

respectively.Since,bydefinition,V(S,R|x)-Vu0,weknowthatBS/BR=ÀVR/VSandBS/BV=1/VS.Takingtheappropriatederivativesfrom[Eq.(1)]and[Eq.(2)],andsubstituting

˙=f,andf[t-(a+b)gV]+/gW=0.Sincethetransversalityconditionistheresults,yields/

/(x¯)=0,weobtain/=F-1,whichyieldscondition(a).

22Ofcourse,anycontractoftheformS(R)=tR-ah-K,whereKisa(lumpsum)constantwouldinducetheCFOtochoosethefullinformationoptimalleveloftaxevasion.SubstitutingthiscontractintotheCFOutilityyieldsV=tR-(a+b)g(R-x)-K,andtheutilitymaximizingchoiceofRsatisfiesVR=0.SincedV

dx¼VRRxþVx¼VxN0,constraint[Eq.(7)]impliesthatthemaximumvaluethatcanbetakenbyKisthesurplusgeneratedatx,andtheCFOwouldretainalloftheincrementalsurplusforlargervaluesofx.Asweshallseebelow,theoptimalcontractinthissettingisnecessarilysecond-best,sothattheshareholdersarebetteroffbysacrificingabitofefficiencyatthemargininordertoextractmoreofthesurplusfromtheCFO.

1602K.J.Crocker,J.Slemrod/JournalofPublicEconomics(2005)1593–1610

Finally,thetotalsurplusfromthecontract,whichisI(1-t)+tR-(a+b)g,isdividedbetweentheshareholdersandtheCFO.Theshareholders’portionofthesurplusisthenwrittenas

Zx

PðS;RjxÞ¼Ið1ÀtÞþtRÀðaþbÞgÀVxðmÞdm:ð11Þ

x

NotingthatVx=bg’,andsubstitutingfrom[Eq.(1)],yieldscondition(b).5

Part(a)ofthetheoremcharacterizestheoptimalreductionsintaxableincome,R,asafunctionofthepermissibledeductions,x.Sincethecostatevariable/representstheshadowcostoftheconstraintimposedbytheinformationalasymmetry,intheabsenceofanyprivateinformationaboutxwewouldhave/=0andtheoptimalcontractwouldresultinthefullinformationleveloftaxevasioncharacterizedby[Eq.(8)].WhentheCFOenjoysaninformationaladvantageregardingthepermissiblelevelofdeductions,however,/isnonzeroandtheresultingcontractisnecessarilysecondbest.

Theconditionsofthetheoremcharacterizeanoptimumaslongasthecontractisimplementable,whichinthissettingrequiresthatR’(x)z0.23WhilethemonotonicityofRisnotguaranteed,thefollowingconditionsthatplacerestrictionsonthehigher-orderderivativesofgandonthehazardrate,f/(1-F),aresufficientforthisresult.24ÈfÉ

Assumption2.gjz0andddx1ÀFz0:Underthisassumption,thefunctionR(x)isnecessarilyinvertible,andwewillusethenotationx(R)toindicatethisinversefunction.Theoptimalincentivecontract,asafunctionoftheobservedreductionsintaxableincome,isS(x(R)),whichwewilldenoteasS(R)inthediscussionthatfollows.TheoptimalincentivecontractS(R)isdepictedasthedottedlineinFig.1.

Theamountofreporteddeductionsfromtaxableincome,R,varieswithx,asdoesthecompensation,S,receivedbytheCFO.SinceVxN0,theparticipationconstraintisbindingonlyatx,andisslackforlargervaluesofx.TheCFOreceivesalevelofcompensationhigherthanherreservationwageifsheinheritsalargedrawofx,becausetheshareholderscannotdistinguishbetweenafavorabletaxsituationduetocircumstancesoutoftheCFO’scontrolfromafavorabletaxsituationduetotheCFO’swillingnesstoincurcosts.Atthelowestpossibledrawofx,theCFOreceivesnomorethanherreservationwage,so[Eq.(7)]holdswithequalityatx.25󰀇󰀈BVRdR

AcontractisimplementableifBxVSdxz0(GuesnerieandLaffont,1984,Theorem1).TakingtheappropriatederivativesyieldstheimplementabilityconditionbgWRVz0,whichrequiresthatRVN0.24SeeFudenbergandTirole(1991,pp.263–267)foradiscussionofthosesufficiencyconditions.Thefirstcondition(gjz0)correspondstotheirassumption(A8),whilethesecondisreferredtointheliteratureasabmonotonehazardrate.QIntheabsenceofAssumption2,optimalcontractsmaystillbecharacterizedbyusingthebironingQtechniquethatwasfirstdiscussedinMussaandRosen(1978),anddescribedindetailbyGuesnerieandLaffont(1984).25Notethatweareformallymodelingtheefficientcontractinasingle-periodenvironmentinwhichthereisonlyuncertaintyregardingasinglebdrawQoftheinformationalparameter,x.Thisapproachwould,however,beequallyapplicabletoamultiperiodsettinginwhichtheCFOreceivedanewvalueofxineachperiod,drawnindependentlyfromthedistributionF.

23K.J.Crocker,J.Slemrod/JournalofPublicEconomics(2005)1593–16101603

Fig.1.S(R):TheoptimalCFObonusscheme.

Theshareholders’after-taxreturnisalsostatedependent,andishighestwhenx=x¯.Asisgenerallythecaseinthesetypesofproblems,thereisnodistortioninthecontractatx¯[soR(x¯)=R*(x¯)],butR(x)bR*(x)foreveryxbx¯becauseoftheincentiveconstraint[Eq.(3)],sothatthefullinformationoutcomeisachievedonlywhenx=x¯.Finally,theincentiveconstraintguaranteesthat,whenpresentedwiththeincentivecontractS(R),aCFOoftypexprefersthecontractualallocations{S(x),R(x)}overallthealternatives.Putdifferently,

¯(x)istangenttoS(R)atR(x)foreveryxa[x,xanx-type’sindifferencecurveV¯],as

depictedinFig.1.

WeseektocharacterizetheimpactofalternativepenaltiesleviedbythetaxingauthorityonthestructureoftheCFOcompensationarrangement,S,asafunctionoftheobservedreductionsintaxableincome,R.Accordingly,theimpactofchangesintheenforcementparametersaandbontheshapeofthecompensationprofileS(R)isofinterest.Webeginbypresentingapreliminaryresult.Lemma.S’(R)=bg’[R-x(R)].

Proof.Frompart(b)ofthetheorem,anapplicationofLeibnitz’sRuleyields

dSddR

¼bgVðRðxÞÀxÞðRðxÞÀxÞþbgVðRðxÞÀxÞ¼bgVðRÀxÞ:dxdxdx

ð13Þ

Then,afterrecognizingthat(dS/dx)=[(dS/dR)(dR/dx)],substitutiongeneratesthedesiredresult.5

1604K.J.Crocker,J.Slemrod/JournalofPublicEconomics(2005)1593–1610

Wenowpresentourmainresult,whichcharacterizestheimpactofthetaxingauthority’senforcementparametersontheleveloftaxevasion,R,andontheformoftheoptimalCFOcompensationpackage,S(R).

Theorem2.UnderAssumptions(1)and(2),asolutiontoproblem(6)hasthefollowingcharacteristics.(a)(b)(c)

BRðxÞBbBBaBBbðxÞbBRBab0;

ðSVðRÞÞ¼ÀbgW

hBxðRÞBaib0;

ðSVðRÞÞ¼gVÀbgW;

hixðRÞ(d)SWðRÞ¼bgW1ÀddN0:RProof.Recallingthat/=F-1,andtotaldifferentiationofTheorem1(a),yields

dxðD1þD2ÞÀdaðfgVÞÀdb½fgVÀ/gW󰀈ÀdRðD1Þ¼0;where

D1uðaþbÞfgWðRÀxÞÀ/gjðRÀxÞb;andD2ufV½tÀðaþbÞgV󰀈þfbgW:

SolvingTheorem1(a)fort-(a+b)gV,andsubstitutingtheresultintoD2,yields

!fVð1ÀFÞ

þf:D2¼bgW

fBothD1andD2arepositiveunderAssumption(2).From[Eq.(14)],bysettingdxanddbequaltozero,weobtain

dRfgVðRÀxÞ

¼;daÀD1

whilesettingdxanddaequaltozeroyields

dRfgVðRÀxÞÀ/gWðRÀxÞ

¼;dbÀD1

ð16Þð15Þð14Þ

h@xðRÞ@bifromwhichpart(a)ofTheorem2followsdirectly.Toobtainparts(b)and(c)ofTheorem2,firstrecallthattheLemmaimpliesSV(R)=bgV(R-x(R)),anddifferentiatingwithrespecttoaandbyieldstheexpressionsontheright-handsides.Inaddition,notethatsettingdR=db=0in[Eq.(14)]implies

dxfgVðRÀxÞ¼N0;daD1þD2

ð17Þ

K.J.Crocker,J.Slemrod/JournalofPublicEconomics(2005)1593–16101605

whiledR=da=0yields

dxfgVðRÀxÞÀ/gWðRÀxÞ¼N0;ð18ÞdbD1þD2

sothatpart(b)ofTheorem2isnegative,andthesignofpart(c)isindeterminate.

Finally,part(d)ofTheorem2followsdirectlyfromdifferentiationofSV(R)withrespecttoR,andnotingthat

dxD1

¼:dRD1þD2

5

ð19Þ

IncreasingeithertheshareholderortheCFOpenaltywillreducetaxevasionsince,foreverylevelofpermissiblereductions(x),thedeductionsclaimedbytheCFO(R)declineasaandbareincreased.Moreover,part(a)ofTheorem2indicatesthatapenaltyleviedontheCFOthroughtheenforcementparameterbismoreeffectiveinreducingtaxevasionthanwouldbeanequivalentpenaltyassessedontheshareholdersthrougha.Thisnon-equivalencyresult,whichdoesnotdependontheparticularfunctionformsofforg,isadirectconsequenceoftheCFO’sprivateinformation.Thiscanbeseemmosteasilybyobservingthat[Eq.(15)]and[Eq.(16)]differonlywhen/isnonzero.Incontrast,werethisafullinformationsettingsothattheincentiveconstraint[Eq.(3)]didnotapply,then/wouldbezeroandthepenalties(aandb)wouldhaveasymmetriceffectontaxevasion.Theintuitionbehindthisresultisthatthebindingincentiveconstraint[Eq.(3)]resultsinaCFOcompensationcontractS(R)thatisnecessarilysecondbest.Inthissetting,theabilityoftheoptimalcontracttoinfluencetheactionsoftheCFOishinderedbythefactthatthecontractmustalsoelicit,inanincentive-compatiblefashion,theprivateinformationregardingthepermissiblelevelofdeductionsfromtaxableincome.Thus,theincentiveeffectsofanysanctionsleviedontheshareholdersbyaareimperfectlycommunicatedtotheCFOthroughtheincentive-compatiblecontract,sothatthepenaltybdirectlyappliedtotheCFOismoreeffective.26ThisresulthasnormativeimplicationsforIRSenforcementpoliciestargetingtaxevasion.Whiledeterminingtheoptimallevelofevasionfromasocietalperspectivewouldrequiretheevaluationofasocialwelfarefunctionthatincludedthesocialcostsofevasionandenforcementaswellasthesocialbenefitfromtaxrevenues,Theorem1(a)doesprovideguidanceregardingthecost-minimizingwayofachievingtheresultingenforce-menttargets.27Ontheonehand,weretheenforcementparameterstohavethesame

Putdifferently,theoptimalcontractisincomplete,sothatincentives(suchasthoseresultingfromsanctionsbeingimposedontheshareholders)areimperfectlytransmittedthroughtheagreement.ThecontractualincompletenessinourmodelisadirectartifactoftheprivateinformationpossessedbytheCFOregardingthelevelofpermissibledeductions,x.Ofcourse,contractualincompletenesscouldenterthisrelationshipinotherways,involvingdifferenttypesofhiddeninformation,orhiddenactions,thatpermitmanipulationbytheCFOofobservablevariablesthatimpactoncompensation.TotheextentthatthenonequivalencyofpenaltiesinourmodelisaconsequenceofanincompletecontractbetweentheshareholdersandtheCFO,wewouldexpectthatothersourcesofcontractualincompletenessmaygeneratesimilarnon-equivalencyresults.Alternatively,wewouldexpectequivalencyofpenaltiesinenvironmentswithcompletecontracts.27NotethatinthispaperwehaveexaminedtheoptimalleveloftaxevasionfromtheperspectiveoftheCFOandtheshareholders,andhavenotconsideredthelargerquestionofthesociallyoptimalleveloftaxevasion.Thelatterwould,ofcourse,requireanexaminationofboththecostsandthebenefitsofraisingtaxrevenues.

261606K.J.Crocker,J.Slemrod/JournalofPublicEconomics(2005)1593–1610

constantmarginalresourcecosts,sothatcV(a)=cV(b)=c¯,thenthepolicyprescriptionwouldbetouseonlythemoreeffectiveCFOpenalty.Ontheotherhand,weretheenforcementparameterstoexhibitincreasingmarginalcosts,sothatcW(d)N0,thenthecost-minimizingenforcementpolicywouldbetousebothaandb.Theprecisemix,ofcourse,woulddependonthemarginalcostsandmarginaleffectivenessofeachpolicytool.

Part(b)ofTheorem2statesthatincreasingthepenaltiesfortaxevasiononshareholdersbyincreasingahastheeffectofflatteningtheoptimalcompensationcontract,asshareholdersreducetheincentiveoftheCFOtoengageinevasion.Thisisthemechanismbywhichthereductioninevasionisachieved.

Part(c)ofTheorem2demonstratesthattheeffectofincreasingCFOpenaltiesthroughbontheslopeoftheoptimalcompensationarrangementisindeterminate,althoughwedemonstrateinanexamplebelowthatmorepreciseresultscanbeobtainedusingspecificfunctionalformsforgandf.Finally,part(d)ofTheorem2indicatesthattheoptimalcompensationcontractSisastrictlyconvexfunctionoftheobservedreductionsintaxableincome,R.Thus,thecompensationcontractexhibitsprogressivityinthesensethattheCFOispaidatthemarginabonus,SV(R),whichisincreasinginthelevelofdeductionsfromtaxableincome,R.Thisreflectsthefactthatevasioncosts,g,areincreasingintheamountofevasion,sothatincentivizingtheCFOtoincreaseevasionrequiresincreasingbonuses,atthemargin.Thefollowingresultisstraightforward.

h2ibCorollary.Ifgisquadraticandfisuniform,thenSWðRÞ¼aþ2bgW.SincegWisconstantinthecaseofaquadraticfunction,itiseasytodemonstratethattheconvexityofS(R)isincreasing(decreasing)inb(a).Consequently,theoptimalcontractbecomesmoreprogressiveinresponsetoincreasesinthepenaltyb,resultinginlargerbonuses(atthemargin)toincentivizetheCFO.

3.Anexample

Forthepurposesofthisexample,wewillassumethatxisuniformlydistributedontheinterval[0,t],soF(x)=x/tandf(x)=1/t.Inaddition,wewillassumethatthepenaltyfortaxevasionisquadraticandincreasingintheamountofillegalevasion,suchthatg(R-x)=(R-x)2/6.Then,by[Eq.(8)],weknowthatR*(x)=x+3t/(a+b),sothattheoptimalamountofevasion,RÀx,is3t/(a+b).Notsurprisingly,thefull-informationoptimallevelofevasionisindependentoftheactualvalueofx,ishigherwhenthetaxrateishigher,andislowerwhentheexpectedcostsofevasionarehigher.FromTheorem1(a),weobtain

RðxÞ¼R4ðxÞÀ

bðtÀxÞ

¼xþð3tÀbðtÀxÞÞ=ðaþbÞ

aþb

ð20Þ

whichimpliesthatR(x)bR*(x)foreveryxbt,asdepictedinFig.2.Inallbutthecornersituationwherex=t,evasionislowerintheinformationallyconstrainedsituation.

K.J.Crocker,J.Slemrod/JournalofPublicEconomics(2005)1593–16101607

Fig.2.First-BestandSecond-BestReductionsinTaxableIncome.

NotealsothatRV(x)ismonotonic,whichwasassuredbecauseAssumptions(1)and(2)aresatisfiedbyfandg,sothattheconditionsofTheorem1aresufficienttocharacterizeanoptimum.Furthermore,

!btð3ÀbÞþbxVx¼N0ð21Þ

3aþbforbV1,sothattheparticipationconstraintholdswithequality(strictinequality)forx=0

(xN0).

Condition(b)ofTheorem1impliesthat

SðxÞ¼

b½tð3ÀbÞþbx󰀈2

6ðaþbÞ2b½tð3ÀbÞxþbx2=2󰀈

:þ

3ðaþbÞ

ð22Þ

Solving[Eq.(20)]forxandsubstitutingtheresultinto[Eq.(22)]yieldsS(R),andbydifferentiationweobtain

SVðRÞ¼

b½bðRÀtÞþ3t󰀈

N0;

3ðaþ2bÞ

ð23Þ

sinceRztfora,bV1.Differentiationof[Eq.(23)]withrespecttotheparametersaandbyields28dbð3tþbðRÀtÞÞ½SVðRÞ󰀈¼Àb0;andda3ðaþ2bÞ2d2b2ðRÀtÞþað2bðRÀtÞþ3tÞ½SVðRÞ󰀈¼N0:db3ðaþ2bÞ228ð24Þð25Þ

Theresultsthatfollowmayalsobeobtainedthroughadirectapplicationoftheconditionsidentifiedin

Theorem2.

1608K.J.Crocker,J.Slemrod/JournalofPublicEconomics(2005)1593–1610

Expressions(24)and(25)indicatethatincreasingthetaxevasionpenaltyonshare-holders(increasinga)resultsinaflattercompensationcontract.Incontrast,increasingthepenaltybornebytheCFO(increasingb)generatesasteepercompensationcontractthatprovideslargerbonusesatthemargintotheCFOforreducingtaxableincome.Inessence,theresponsebytheshareholderstolargerpenaltiesleviedontheCFOfortaxevasioninthisexampleistoadjustthecompensationarrangementtosharpentheincentivesfortheCFOtoengageinsuchevasion.Putdifferently,anincreaseinbgeneratesarestructuringofS(R)thatpartiallyoffsetsthedisincentivefortheCFOtoengageintaxevasion.

Finally,wecanalsoexaminetheeffectsofthepolicytools(aandb)ontheamountofillegaltaxevasion.Lettinge*uR*-xdenotetheamountofevasionthatresultsfromthefullinformationcontract,itisstraightforwardtodemonstratethat

de4=da¼de4=db¼À3t=ðaþbÞ2b0

ð26Þ

sobothtoolsareequallyeffectiveatdeterringevasionintheabsenceofaninformationalasymmetry.Alternatively,lettingeuR-xdenotetheamountofevasionthatresultsfromtheoptimalinformationallyconstrainedcontract,wefindthat

de=da¼ðÀ3tþbðtÀxÞÞ=ðaþbÞ2Nde4=da;andde=db¼ðÀ3tÀaðtÀxÞÞ=ðaþbÞ2bde4=db

ð27Þð28Þ

sotheb(a)toolbecomesmore(less)effectiveasaresultoftheinformationalasymmetry.Putdifferently,sincede/da-de/db=(a+b)(t-x)N0foreveryxbt,itfollowsthatbhasalargereffectonevasionthandoesa.29ApenaltylevieddirectlyontheCFOwillbemoreeffectiveatreducingevasionthananotherwise-equivalentpenaltyleviedontheshareholders.

4.Conclusionsandfutureresearchdirections

Taxevasionbylarge,publiccorporationsisapparentlywidespread,andtheappropriatepolicyresponseiswidelydebated.Onesetofresponsesistochangethetaxcodetofacilitatedetectionandsuccessfulprosecutionofcertainclassesofevasion.Anotherpossibleresponseistostrengthenthepenalties—onthecorporationoronthecorporateofficers—forthoseactsofevasionthataredetected.Existingeconomictheoryofferslittleguidancetopolicymakers.Thereis,tobesure,alargepositiveandnormativeliteratureabouttaxevasion,butnearlyallofitpertainstoindividuals.Thesmallliteraturethataddressesfirmsfocusesonthejointoutput-evasiondecision,andignorestheseparationofownershipandcontrolthatcharacterizessuchcorporations.

Thispaperbeginsthetaskofdevelopinganeconomictheoryofcorporatetaxevasionthataddressesthefactthatcorporatetaxofficerswilleitherhaveexplicitcompensation

Recallthatbothde/daandde/dbarenegative,sotheabsolutevalueofthelatterisgreaterthanthatoftheformer.

29K.J.Crocker,J.Slemrod/JournalofPublicEconomics(2005)1593–16101609

contractsinwhichtheirsuccessattaxminimizationisrewarded,orwillhavethissuccessreflectedimplicitlyviaperformancereviewprocedures.Ourresultssuggestthat,inthissetting,theeffectofpoliciesdependsonwhetherthecorporationortheofficerispenalized,andtheextenttowhichthecorporationcanoffsetanypenaltyregimebyrestructuringitscompensationcontractwiththetaxofficer.Fromthetaxagency’spointofview,penaltiesassessedonthetaxofficeraremoreeffectivetoolsagainstevasionbecausetheyexacerbatetheconflictbetweentheshareholdersandthetaxofficer,resultinginwhatisalessefficientoutcomeforthetwotakentogether,butthatreducesevasionmorethanthesamepenaltyassessedonthecorporation.Thenormativeimplicationsofthisfindingarefairlystraightforward.Ifimplementingthetwopenaltieshasequalandconstantmarginalsocialcosts,thenourresultssuggestusingonlytheagentpenalty.Iftheyhaveequalcostfunctionsbutincreasingmarginalcosts,ourresultssuggestusingbothpenalties,butusingtheagentpenaltiesmore.Moregenerally,theresultspushtheoptimalenforcementpolicyfavoragentpenaltiesmorethanwouldbetrueiftheywereequallyeffectiveincombatingcorporatetaxevasion.

Wecanthinkofseveralfascinatingandimportantquestionsthatcanbeaddressedusingtheframeworkweintroducehere.OneistheeffectofthedifferinghorizonsofthetypicalCEOandthetypicaltaxofficer.Giventheoftenseveral-yearlagbetweenataxreportandthecompletionofanaudit,itiscommonthatbythetimeapenaltyisassessedtheCEOisnolongerwiththefirm;thisintroducesanotherconflictofinteresttothetaxevasiondecision.Anotheristheeffectofpublicdisclosureofcorporatetaxableincomeandtaxpayments.Althoughtheadvocatesofthisfavoritbecausetheythinkitwill,throughpublicpressureandcorporateembarrassment,reducetaxavoidance,themodelwehaveoutlinedheresuggeststhatthereisanoffsettingeffect.Note,though,thatpublicdisclosurealsoimpliesdisclosureoftheeffectivetaxrateofafirm’scompetitors.Forthisreason,itmayfacilitatethebenchmarkingoftheperformanceoftheCFO,andenablemoreeffectivecompensationcontracts,whichwouldincreaseevasionandforcedowneffectivetaxrates.30Finally,thistypeofmodelcanbeextendedtoaddressaccountingmisrepre-sentationinadditiontotaxevasion,forwhichtheenforcementagencyistheSecuritiesandExchangeCommission,ratherthantheInternalRevenueService.

References

Allingham,M.,Sandmo,A.,1972.Incometaxevasion:atheoreticalanalysis.JournalofPublicEconomics1,

323–338.

Chen,K.,Chu,C.,2002.Internalcontrolvs.externalmanipulation:amodelofcorporateincometaxevasion.

WorkingPaper.InstituteforSocialSciencesandPhilosophy,AcademiaSinica,Taipei.

Cowell,F.,2004.Sticksandcarrots.In:Aaron,H.,Slemrod,J.(Eds.),TheCrisisinTaxAdministration.Brookings

InstitutionPress,Washington,DC,pp.230–258.

Crocker,K.,Morgan,J.,1998.Ishonestythebestpolicy?Curtailinginsurancefraudthroughoptimalincentive

contracts.JournalofPoliticalEconomy106,355–375.

Desai,M.,Dyck,A.,Zingales,L.,2003.Corporategovernanceandtaxation.HarvardBusinessSchoolWorking

Paper,Cambridge,MA.

30WethankDanSilvermanforsuggestingthisextension.

1610K.J.Crocker,J.Slemrod/JournalofPublicEconomics(2005)1593–1610

Douglas,R.,Ellingsworth,P.,McAndrews,J.,1996.Whatmanagementexpects:evaluationofthetaxdepartment

anditsexecutives.TheTaxExecutive,29–32.

Fudenberg,D.,Tirole,J.,1991.GameTheory.MITPress,Cambridge.

Guesnerie,R.,Laffont,J.,1984.Acompletesolutiontoaclassofprincipal-agentproblemswithanapplicationto

thecontrolofaself-managedfirm.JournalofPublicEconomics25,329–369.

Hanlon,M.,2003.Whatcanweinferaboutafirm’staxableincomefromitsfinancialstatement?NationalTax

Journal56,831–863.

Hollingsworth,T.,2002.ManufacturersAlliance/MAPISurveyofCorporateTaxDepartments—4thEdition

ManufacturersAlliance/MAPI,Arlington,VA.

Holmstrom,B.,1979.Moralhazardandobservability.BellJournalofEconomics10,74–91.

Kraakman,R.,1984.Corporateliabilitystrategiesandthecostsoflegalcontrols.YaleLawJournal93,857–8.Lacker,J.,Weinberg,J.,19.Optimalcontractsundercostlystatefalsification.JournalofPoliticalEconomy97,

1345–1363.

Lee,K.,1998.Taxevasion,monopoly,andnonneutralprofittaxes.NationalTaxJournal51,333–338.Mussa,M.,Rosen,S.,1978.Monopolyandproductquality.JournalofEconomicTheory18,301–317.Myerson,R.,1979.Incentivecompatibilityandthebargainingproblem.Econometrica47,61–73.Novack,J.,Saunders,L.,1998.Thehustlingofx-ratedshelters.Forbes,December14.

Polinsky,A.,Shavell,S.,1993.Shouldemployeesbesubjecttofinesandimprisonmentgiventheexistenceof

corporateliability?InternationalReviewofLawandEconomics13,239–257.

Slemrod,J.,2001.Ageneralmodelofthebehavioralresponsetotaxation.InternationalTaxandPublicFinance8,

125–138.

Townsend,R.,1979.Optimalcontractsandcompetitivemarketswithcostlystateverification.Journalof

EconomicTheory21,265–293.

USGeneralAccountingOffice,1979.ChallengesRemaininCombatingAbusiveTaxShelters.Testimonybefore

theSenateCommitteeonFinance,Washington,DC,October21,2003.

Yitzhaki,S.,1974.Anoteonincometaxevasion:atheoreticalanalysis.JournalofPublicEconomics3,

201–202.