UCP PHOS article Taking Absurd Theories Seriously Economics and the Case of Rational - PDF document

04353UCP:PHOSarticle#710302 TakingAbsurdTheoriesSeriously:EconomicsandtheCaseofRationalAddictionTheoriesOleRogebergRationaladdictiontheoriesillustratehowabsurdchoicetheoriesineconomicsgettakenseriouslyaspossiblytrueexplanationsandtoolsforwelfareanalysisdespitebeingpoorlyinterpreted,empiricallyunfalsiÞable,andbasedonwildlyinaccurateassumptionsselectivelyjustiÞedbyad-hocstories.Thelackoftransparencyintroducedbypoorlyanchoredmath-ematicalmodels,thepsychologicalpersuasivenessofstories,andthewaytheprofessionneglectsrelevantissuesaresuggestedasexplanationsforhowwhatweperhapsshouldseeasdisplaysoftechnicalskillandingenuityareallowedtoblurthelinesbetweenscienceand1.Introduction.Howcaneconomiststakeabsurdtheoriesofchoiceseriously?And,whyarenÕtthesetheoriesexposedtorelevantcriticism?Idiscussthesequestionsthroughacase-studyofrationaladdictiontheory,currentlyÔÔoneofthestandardtoolsintheeconomicanalysisofthemarketsfordrugs,alcohol,tobaccoandotherpotentiallyaddictivegoodsÕÕ(Ferguson2000,587).Thesetheoriesareregularlyfeaturedintheleadingeconomicsjournals,andtheirimpactissuchthattheissuesin-volvedcannotbebrushedoffasÔÔmereÕÕmethodologybyeconomists:Theirstandingintheliteraturemeansthattheyinßuencethethinkingofeconomistsonaddiction,turnupinpolicydiscussionssuchasMironandZwiebel1995,Þlterouttonon-economiststhroughinterdisciplinaryPhilosophyofScience,71(July2004)pp.263Ð285.0031-8248/2004/7103-0002$10.002004bythePhilosophyofScienceAssociation.Allrightsreserved.*ReceivedAugust2003;revisedOctober2003.TocontacttheauthorwritetoRagnarFrischCentreforEconomicResearch,Gaustadalle21,N-0349Oslo,Norway;e-mail:o.j.rogeberg@frisch.uio.no.ThisworkwasÞnancedbytheNorwegianResearchCouncilthroughaPh.D.grantfromHERO(HealthEconomicsResearchProgrammeattheUniversityofOslo).Thankstotwoanonymousreferees,mythesissupervisors(ProfessorsMichaelHoelandAanundHyllandattheUniversityofOslo)andMortenNordbergforcomments. 04353UCP:PHOSarticle#710302workssuchasElsterandSkog1999andElster1999a,andmakeeconomictheoryseemirrelevantandabsurd(seeforinstanceLevine2000).DevelopedbyNobelLaureateGaryBeckerandKevinMurphy,thetheoryofrationaladdictionanditsvariantsmodelwell-informedandforwardlookingrationaladdictswithstablepreferences,whoseaddic-tionsaretime-consistent,optimalconsumptionplans.Thetheoryisclaimedtoexplainthebehaviorofrealaddicts,i.e.identifythereasonswhytheydowhattheydo.Ithasalsobeenusedforwelfareanalysis.focusontwomajorrationaladdictiontheories,butthesefeaturesarecommonalsotolatervariantssuchasLaibson2001.Theseclaims,takentogether,committhesetheoriststorealismÑtheclaimthattheirtheoriesaretrue,thatindividualsarerationalinthesensediscussedandthatthewelfareeffectsobservedinthetheoryareinformativeaboutthewelfareeffectsthatactuallywouldfollowforrealindividuals.Iarguethatthisisanabsurdclaim.Thesillinessoftheoriessuchasrationaladdictiontheoryraisesthequestionofhowtheycanbetakenseriously.Isuggestanexplanationbasedontherelativeimpenetrabilityofmathematicalargumentsascomparedtoverbalarguments,thepsychologicalfeelingofunderstandingtriggeredbythestoriesincludedinarticles,alongwiththeneglectofsomeoftheunderlyingissuesbytheeconomicsprofession.Toavoidmisunderstanding:Thereisnothingwrongwithmathematicalmodelingperse.Mathematicshasprovenitselfaninvaluabletoolinavarietyofsciences.Theargumentthatfollowssimplystatesthatmathe-maticsÑlikeanyothertoolÑcanbebothappliedandmisapplied.Therealproblemwithrationaladdictiontheoryisnotthatitusesmathemat-icalmodeling,butthatitisempiricallyunfalsiÞable,basedonwildly1.Examplesfromthetwoarticlesusedinthefollowing:BeckerandMurphy1988statesthatÔÔrationalchoicetheorycanexplainawidevarietyofaddictivebehaviorÕÕ(676)andthatÔÔatheoryofrationaladdictiondoesexplainwell-knownfeaturesofaddictionsandappearstohavearichersetofadditionalimplicationsaboutaddictivebehaviorthanotherapproachesÕÕ(695).OrphanidesandZervos(1995)writethattheirtheorysubstantiallyimprovesourÔÔunderstandingofthedeterminantsofaddictionÕÕ(740).Ofthegeneralframeworkinvolved,BeckerhasstatedthatnoÔÔalternativeapproachÑbeitfoundedonÔÔcultural,ÕÕÔÔbiological,ÕÕorÔÔpsychologicalÕÕforcesÑcomesclosetoprovidingcomparableinsightsandexplanatorypowerÕÕintoÔÔawideclassofbehavior,includinghabitual,social,andpoliticalbehaviorÕÕ(Becker1996,4).2.BeckerandMurphy(1988)statethatÔÔaddictions,evenstrongones,areusuallyrationalinthesenseofinvolvingforward-lookingmaximizationwithstablepreferencesÕÕ(675)andthattheirmodelimpliesthataddictsÔÔwouldbeevenmoreunhappyiftheywerepreventedfromconsumingtheaddictivegoodsÕÕ(691).OrphanidesandZervos(1995)usetheirmodelasthebasisforawelfareanalysisofinformation,educationandpublicpolicy(751Ð754).SeealsotheviewsonwelfareanalysisinBecker1996,20.olerogeberg 04353UCP:PHOSarticle#710302inaccurateassumptions,andpoorlyinterpretedinaselectivewaythatmakesitseemplausibletosomeeconomists.Mathematicsentersonlybymakingthispossibleinpractice.Putdifferently,Iamnotsayingeconomistsshouldstopusingmathe-matics,butthattheyshouldcreateasystemofchecksandbalancesbyintroducingfurthermethods:Mathematicsallowingforconstraintsandcomplexinterdependencieswhereappropriate;afull,verbalexplanationofthetheorymakingiteasiertothinkofcontradictoryevidence,weak-nessesinassumptionsandfurtherfactorsthataffectthemechanismorprocessdiscussed;statisticalevidenceusedtoestablishthevalidityofempiricalassertions.ThisissimilartoothereconomistÕsdesireformeth-odologicalpluralismandtailoringthemethodtothesubjectmatter(e.g.,Chick1998).Finally,Ialsodiscussthepossibilitythattheeconomistswhoauthorthesetheoriesdonotreallytakethemseriouslyinthewayindicatedbytheirwrittenclaims.Instead,theymaybeplayingatheory-gamethatblursthelinesbetweeneconomicscienceandgame-playing,causingconfusionandmisunderstanding.2.UseandAbuseofMathematicsinEconomics.Mostofthetopicsonwhicheconomistsholdviews[that]arebothdifferentfromÔcommonsenseÕandunambiguouslyclosertothetruththanpopularbeliefsinvolvesomeformofadding-upconstraint,in-directchainofcausation,feedbackeffect,etc.Whycaneconomistskeepsuchthingsstraightwhenevenhighlyintelligentnon-economistscannot?Becausetheyhaveusedmathematicalmodelstohelpfocusandformtheirintuition.(Krugman1998,1834)Mathematicalmethodshelpusaccuratelyformulate,reason,andex-aminecertaintypesofargumentswearepoorathandlingunaided.This,asPaulKrugmannotes,isastrongargumentinfavorofmathematics,thoughitalsosuggestscautionwhenapplyingsuchtechniquestoindi-vidualdecision-making.Themathematicalmodelofdecisionmakingwillbewell-equippedtohandlesophisticated,precisetrade-offscausedbythemanywaysthevalueofgoodsmightdifferwithourconsumptionofothergoodspast,presentandinplansforthefuture.ByKrugmanÕsargument,evenintelligentpeoplewouldhandlesuchtradeoffspoorly.WewouldbeusingamethodthatexcelsatXtomodelsomethingthatispooratX.AsecondimplicationofKrugmanÕsargumentisthatourintuitionandcommonsensewillhaveahardtimefollowingandcriticallyassessingtheexplanationofferedinmathematicalmodels,preciselybecausethesedealwithdeductionsandimplicationsthatarehardtomakeinotherways.ItbecomesÔÔpossible,underthecoverofacarefulformalism,tomaketakingabsurdtheoriesseriously 04353UCP:PHOSarticle#710302statementswhich,ifexpressedinplainlanguage,themindwouldim-mediatelyrepudiateÕÕ(KeynesquotedinChick1998,1865).Inadditiontomakingithardforustoseetheassumptionsandimplicationsofthemathematicalmodel,humancognitivelimitationsalsointroduceatendencytobeoverlyswayedbystories.Asrecentlydis-cussedinTrout2002,oursubjectivefeelingofunderstandingdrivesourÔÔjudgmentsoftheplausibilityand,ultimately,theacceptability,ofanexplanationÕÕdespitebeingÔÔinparttheroutineconsequenceoftwowell-documentedbiasesincognitivepsychology:overconÞdenceandhind-sight.ÕÕRoughly,ifwearepresentedwithevidence,itwilloftenfailtohaveanyinßuenceintheabsenceofagoodstory,andifwearepresentedwithagoodstoryitwilloftenhaveanexcessiveinßuencedespitealackofevidence(cf.Dawes1999).Iwouldarguethatthereislittlebeyondadhocstoriesinthewayrationaladdictiontheoriesaresupported.Or,putdifferently,thatstoriescanbetoldtosupportselectedassump-tionsandresultswithsuchease,thatthesetheoriesalltooeasilycre-ateamistakenimpressionofhavingaccountedforawidevarietyoffacts.Finally,theclimateintheprofessionmakesithardtopointouttheabsurditiesmadepossiblebytheabovefactors.EconomicsseemstohaveembracedandidentiÞeditselfwithmathematicalmethodstotheextentthateconomistsinarecentÔÔcontroversyÕÕsectionofatop-rankingjournalfelttheneedtoarguethat,while[f]ormaltechniquesarepowerfultools,[]theycanalsobedan-gerous;theproblemistoidentifyapplicationswheretheycanbeusedsafely.[]Mypurposeistomakethecaseforsuchadebatetobeopened.(Chick1998,1859)Thesurprisingthingisthatevensuch(bland?)statementsarenotuncontroversial:Inthesameissue,KrugmandeniesßatlythatÔÔexcessiveformalism,ÕÕÔÔexcessiverelianceonthedeductivemethod,ÕÕandÔÔarcanealgebrathathasnorelationshiptorealityÕÕisaproblemineconomics.Theverybesteconomists,heclaims,aswellasmostofthesuccessfulones,arefreeofsucherrors.Economistswhothinkformalismisaproblemaremistaken;outsiderswhothinksojustdislikesimplemaththatrefutetheirpetdoctrines(Krugman1998).Whilemyremarksinsupportofmethodologicalpluralisminthein-troductionmaynotdiffermuchfromthemethodologicalguidelinesKrugmanendsupsupportinginhispiece,IbelieveÑunlikehimÑthattheseareviolatedinpracticebysuccessfuleconomists,withabsurdtheoriesdiscussedinfullseriousnessastheresult.Rationaladdictiontheoryhopefullyillustratesthis.olerogeberg 04353UCP:PHOSarticle#7103023.Story-tellinginRationalAddictionTheories.3.1.Instrumentalism,Realism,andIdealization.Theoriesofrationaladdictionmakeassumptionsconcerningthechoicerule,preferencesandbeliefsofpeople,andderivetheresultingconsumptionplans.Thechoiceruleisassumedtoberational,meaningthatyouuseyourbeliefstoidentifyandchoosetheoptionbestsatisfyingthepreferences.Sincetherearenoclearorobviouslimitsonpreferencesandbeliefs,thesetofpossiblerationalconsumptionpatternsremainstobeidentiÞed.Currentlyderivedpatternsbyrationaladdictiontheoriesincluderising,falling,shifting,cyclic,stable,stochastic,andchaotic,aswellasendogenousquittingofbothÔÔcoldturkeyÕÕandgradualkind.Thisßexibilityisastrengthinthattheframeworkhasaprimafaciepossibilityofexplainingthesedifferentconsumptionpatterns.Ontheotherhand,itmakesiteasytoseeminglyexplainsomethingbyjustmimickingit.Thiscasecouldbecalled‘mockexplanations’,inthatsuchtheoriesÔÔexplainÕÕbyrepro-ducingobservableregularitieswithoutcapturingthecausalmechanismsthatproducetheseregularities,i.e.describingratherthanexplaining.Someeconomistsviewsuchdistinctionsasunimportant,claimingsupportfromthefamousas-ifmethodologyofFriedman1953ÑusuallyinterpretedasinstrumentalismÑwhichexplicitlyidentiÞespredictionastheonlyaimofÔÔpositiveeconomics.ÕÕSeenasinstrumentalist,thiswouldarguethatgoodas-iftheoriesarepredictivedevices,notexplanations.Granted,mockexplanationsmayproveexcellentpredictivedevicesaslongastheempiricalregularitiestheydescriberemainstable,andtheas-ifstorieshelpourunderstandinggraspandmanipulatewhatwouldother-wisebedisconnectedfactsandabstractrelationships.AsnotedinRosenberg1992,160,though,thisdefensecomesatacost:Sincesomeoneactingas-ifhewasrationallysolvingsomedecisionproblemwouldnotbehaveoptimallyunlessthiswastheactualdecisionproblemhefaced,assumptionsmatterwhenweturntowelfareanalysis.Norcanas-iforiesclaimtoexplaininthesenseofdescribingthemechanismorcausalprocessunderlyingaphenomena,theiraimisjusttorelateobservablequantitiesinthesimplest,mostempiricallysuccessfulway.TouseametaphorfromHausman1992,anas-iftheoryofacarwouldbeofnohelptoamechanicifthecarbrokedown.Nor,itwouldseem,couldanas-iftheoryofaddictiongiveinsightintoandunderstandingofaddictions.3.TheseexamplesarefromFerguson2000(rising,falling,stable,andshiftingwhennewinfo),Feichtinger,Hommesetal.1997(chaotic),BeckerandMurphy1988(rising,falling,quittingcoldturkey,shifting,cyclic),Laibson2001(triggeredbystochasticcues),DocknerandFeichtinger1993(cyclic),AddaandLechene2001(rising,falling,endogenousquit-tingÑallwithinasinglemodelbychangingparametervalues).takingabsurdtheoriesseriously 04353UCP:PHOSarticle#710302NotethatFriedmanexplicitlyrestrictshimselftopredictionandelse-whereexplicitlywarnsagainstdrawinginferencesaboutwelfarefromasuccessfullypredictingtheory(FriedmanandSavage1952,473).Sincerationaladdictiontheoriesareclaimedtoexplainandprovideabasisforwelfareanalysis,Iinferthattheyarenotjustas-iftheories.Whilethisdisposesoftheas-ifdefense,thequestionremainswhetherthereareotherwaystoclaimvalidconclusionsfrominvalidpremises.Putdifferently:Iassumeinwhatfollowsthatthevalidityofyourconclusionsderivesfromthevalidityofyourreasoningandthesoundnessofyourassumptions.EconomistsclaimthattheirrationalchoicetheoriesareavalidbasisforwelfareanalysisandprovidevalidexplanationsofrealpeopleÕsbehavior.ItakethisasaclaimthattheirassumptionsareÔÔgoodenoughÕÕapproximationstothetruth.Thishas,indeed,beenthedefenseofeconomicmodelsinvokedbysomeeconomists(see,forinstance,GibbardandVarian1978orKrugman1998),andIattempttoshowthatthisdefenseisinvalidforrationaladdictiontheories.Thediscussionisrelatedtoidealization,ÔÔaliterallyfalseexaggerationthatservesanabstractiveorisolatingtheoreticalpurposeÕÕ(Hausman1992,132).Theoriessuchasrationaladdictiontheoriescouldbeseen,insuchalight,aspresentingidealizedrepresentationsofrealdecisionmakers.Sinceweknowthatouridealizedindividualisafarcryfromrealindividuals,thisrequiresthatwehaveasubsidiarytheoryconcerninghowdeviationsfromthisidealwillinßuencetheoutcomeofourmodel.AnexampleofsuchanattemptistheworkofpsychoeconomistAlanNelson,whointroducedadistinctionbetweencompetenceandperformancealongthelinesoflinguistictheory:Discrepanciesbetweenthecompetenceofeconomicagentsbasedupontheiridealabilitytomaximizetheutilityfunctionwhichwenowassumetheyreallydopossess,andtheactualperformancetheseagentsinthemarketplace(oreveninthelaboratory)canbeaccountedforintermsofinterferingfactorswhichareirrelevanttoabstractmicroeconomictheory.Itcanbehypothesizedthatthisin-terferencewilleventuallybesatisfactorilyexplainedasresultingfromthingslikemisinformation,imperfectmemory,limitedcalculatingabilities,etc.(quotedinRosenberg1992,137)AsRosenberg(1992)notesinhisdiscussionofNelson,traditionaleconomistsshowlittleifanyinterestinsuchwork.ThislackofinterestisapparentintherationaladdictionÞeld,wherea)directvalidityisclaimed4.Theargumentintermsofidealizationowesalargedebttooneoftheanonymousolerogeberg 04353UCP:PHOSarticle#710302forexplanationandwelfareanalysis,b)nodiscussionisincludedofwhereandinwhatwaystheidealizationsemployedmisrepresentreality,c)nointerestisshowninandnomethodpresentedforexaminingthedependenceofconclusionsontheinaccuraciespresent.Inotherwords:Yes,clearlyuntruetheoriesmighthaveexplanatoryvalueandmaketrueclaims,butthisneedstobeestablishedandshown,andthathasnotbeendoneintherationaladdictionliterature.Itisofcourseconceivablethatfurtherargumentsmightbefoundthatwouldmaketheoriesofthediscussedkindvalidfortheirproposedusesdespitetheirabsurdity,butgoodreasonsforfaithinthispossibilityareneithercommonknowledgenorgivenorreferencedbyarticlesofthekindunderdiscussion.3.2.EarlierCriticismoftheRationalAddictionApproach.ThemaincriticofrationaladdictiontheoryinthepasthasbeenphilosopherJonElster,whosechiefclaimhasbeenthatBeckerÕstheoriesareconceptuallyincoherent(seeElster1997,1999b,2000b,2000a).Thisclaimhasbeencontestedbyothers(e.g.Rogeberg2003).ThesociologistOle-J¿rgenSkog(1999)hasraisedanumberofgoodpointsconcerningBeckerandMurphy1988thatwerepartlytobereadasalistofsuggestionsastohowtheirtheorycouldbedevelopedintoamoreelaboratetheory[thoughalso]someproblemsofamorefundamentalnature,whicharelesseasilyre-solved.(Skog1999,188)SkogÕsmorefundamentalproblemsmainlyconcerntherationalityassumed,whichhasalsobeencriticizedbrießyinWinston1980,302.Wereturntothislater.Whileotherkindsofcriticismofthetheoryorearliervariantsoftheparentframeworkhavebeenmade(e.g.,Cowen1989,Rosenberg1979,1981,1992),noneofthesehavefocusedonbringingouttheabsurdityofthetheoriesandthewaymathematicsisinvolvedinhidingit.3.3.Problemswiththe‘‘Story-telling’’Approach.AsimplewaytothinkofthemathematicalmodelofrationaladdictiontheoryisasaformalsystemÑasetofvariablesrelatedtoeachotherbyasetofequationsÑthathopefullyÔÔmirrorsÕÕsomerealworldsystemmechanismprocess.Ourgoalisatheorythatcapturessomerealworldsystemmechanismprocesswellenoughthatwecandrawvalidinferencesfromthetheory(e.g.,userationaladdictiontheoryforwelfareanalysis)andexplaintherealphenomenon(e.g.,understandaddicts).takingabsurdtheoriesseriously 04353UCP:PHOSarticle#710302Atheorycanobviouslynotcaptureeverything.Forinstance,econo-mistsoftenclaimthatmathematicalchoicemodelsdonotneedtomirrorprocessbywhichhumansidentifythesolutiontotheirdecisionproblemaslongastheycanspecifytheoutcomethatthehumandeci-sionprocesswillarriveat.Ifhumansmakeoptimalchoices,anydecisionprocessabletoidentifythisglobaloptimumcanbeusedbyourtheory(sometimesphrasedbystatingthatproceduralrationalityisunimportantifpeoplearesubstantivelyrational).Whilethisargumentsuggeststhatourassumedsolutionmethodmaybeunimportant,itdoesrequirethatsomesolutionmethodequivalenttothatusedbyourtheoreticalmodelisavailabletorealpeople.Moreimportantly,itrequiresthatthedecisionproblemitselfiscorrectlyspeciÞed.TosaythatpeoplechooseYbecauseitisas-iftheyweresolvingproblemXisinsufÞcient.ToclaimthatXexplainsthechoiceofYorestablishesitsoptimalityorwelfaremaxi-mizingproperty,wemustestablishthatpeoplefaceaproblemsuchasX.Todothisweneedto‘anchor’theassumptionsinourformalsystemtotheirempiricalcounterparts.Weinterpretthemodel,statehowitrelatestotheworld,identifyvariableswithempiricalcounterpartsandgiveev-idencethatrelationshipsspeciÞedbetweentheoreticalvariablesmirrorrelationshipsbetweentheempiricalphenomena.Unanchoredassumptionsareimportanttonote,sinceaßexibleframeworkmaybemadeconsistentwithobservationssimplybyalteringthese.Thisalsobringsoutandcatalogsinaccuraciesintroducedbyouridealizationofdecisionmakers.Inpractice,theanchorsthattietherationaladdictiontheoriestorealityarestories.TheserangefromincentivesandmechanismssuggestedbyresearchÞndingstoanecdotalevidenceandcommonsense,andaimatestablishingtheassumptionsasintuitivelyplausibleandacceptable.Thelooselystructuredwaythisisdoneinpracticeopensforthreeproblems:Weakanchors.Storiesneedtoseemplausibleandrelevantsothattheyarerhetoricallypersuasive.Apartfromthis,therearenorulesforÔÔgoodÕÕstories.Theymaysuggestqualitativeincentiveswithoutindi-catingeithertheirstrengthorwhichgroupsofpeopleweshouldexpecttobeawareofthem.InsteadofsupportingsomestrangeassumptionXwithastory,youmaysupportsomereasonableaspectofXorevensomedifferentbutrelatedassumptionY.Unanchoredassumptions.Youdonotneedtoprovideexplicitinter-pretationsortellstoriestosupportthestrangerassumptionsofyourtheory.Weakanchorscanmakeitseemasthoughtheyhavebeensupported.Unanchoredfacts.Astoryfocusesourattentiononanincentiveorsomeaspectofreality.Thesetofpossiblestoriesthatmightberelevantis5.Iamnotclaimingthisissufcienttoestablishthatthetheory-explanationiscorrect,onlythatitisforittobeaplausiblealternative.olerogeberg 04353UCP:PHOSarticle#710302thereforeimmense,makingiteasytotellanon-representativesetofstories.Ifstoriescontradictassumptions,theycanbeleftout.Storiesthatgivethedesiredqualitativeeffectscanbepresentedasthoughrepresen-tative.Anomaliescanbeexplainedwithinthetheorybytellingfurtherstories.4.ProblemsinPracticeandTheirImportance.4.1.RationalAddictionTheories.Drugusersoftenexperimentwithdrugsbeforegraduallyincreasingconsumptionastheygethooked.Rationaladdictiontheoriesusestablepreferencestoexplainthisunstableconsumptionbyviewingpresentconsumptionpartlyasinvestmentbe-havior.Bysmokingtodayyouincreaseyoursmokingstock.StablepreferencesaredeÞnedoversuchstocksinadditiontotheconsumptiongoods.Theseextendedpreferencesallowforvariousinteractions.Forexam-ple,ahighersmokingstockisundesirableinitself(reduceshealth,pro-duceswithdrawal,etc.).Italsomakessomeothergoodsmoreenjoyable,e.g.cigarettes(relieffromwithdrawal,learningtoenjoyeffects)andlunchwithsmokingcolleagues(reducedsensitivitytotobaccosmoke).Othersmaybecomelessenjoyable,suchasexercise(reducedcardiovascularÞtness)andfamilyparties(anti-smokingrelatives).AccordingtoBeckerandMurphy(1988),consumptionofheroin,to-bacco,jogging,andtelevisionwatchingareallcharacterizedbyhavingsucheffectsonfuturelevelsofvariousstocks.Individualshaveprecise,quantitativeestimatesbothoftheseeffectsandofallfuturelevelsofÔÔexogenousÕÕfactorssuchasprices,normsandpublicregulations.Todecidewhetherornottosmoketobacco,cigarettes,heroin,orcrack,whetherornottoturnonthetelevision,goforajogoreatachocolatebar,individualsexploittheseeffectsanddesignadetailedconsumptionplanfortheirfuturelifeexhaustingallgainsfromalltrade-offsacrosstimeandgoodsconditionalonopportunitiesandexogenousfactors;i.e.theymapoutÑinprecisedetailÑtheiroptimal,lifetimeplansofmomenttomo-mentconsumptionofthesegoods.4.2.Detailed,Forward-looking,Long-termPlanning.ThedeÞningas-sumptionofrationaladdictiontheoriesisthatindividualsmakedetailed,6.Thefollowingisthesimplest,mostplausibleinterpretationofrationaladdictiontheory.InRogeberg2003Itriedtointerprettheentireextendedutilityframework,includingtheendogenoustimepreferencesinBeckerandMulligan1997andOrphanidesandZervos1998,butsuchinterpretationsseemlessopentoevidenceandargueforagnosticismasregardsresultsofwelfareanalysis.takingabsurdtheoriesseriously 04353UCP:PHOSarticle#710302forward-lookingplans.Thisistheengineofthetheories,neededforconsumptiontorationallychangeovertimedespitestablepreferencesandnonewinformation.ItisalsothefeaturethatisusedinBeckerandMurphy1988,691Ð692todistinguishthistheoryfromothertheoriesofaddiction:ÔÔ[I]nourmodel,bothpresentandfuturebehaviorarepartofaconsistent,maximizingplan.ÕÕIntermsofKukla1991,thisrequiresamaximalcriterionofrationality,alogicalomnisciencewhereweimmediatelyseeallimplicationsofourknowledge,anassumptionthatpeoplealwaysreasoninthemannerKrugmandeniedweareabletodowithouttheuseofmathematics.ToÞndthesolution,theeconomiststurntomathematicaltechniquessuchasthetheoryofoptimalcontrolandderivepresentconsumptionastheÞrststepinawell-designed,coherentplan.Sinceyourequireyourlifetimeplantoknowhowmuchtoconsumetoday,changesinbeliefsaboutfuturepriceshaveimmediateeffectsonpresentconsumption.ÔÔAproposedhikeinthetobaccotaxmaykickinnextyear,soIÕmsmokingacigarettelessdailytoadjustmystockpath.ÕÕThiseffecthasbeenamajorfocusoftherelatedeconometricliterature.ItÕsacommonsayingthatextraordinaryclaimsrequireextraordinaryevidence,sowemightexpectthiscentralassumptiontobewellsup-ported.FehrandZych(1998)foundthatpeoplearepooratÞndingthecorrectsolutiontosuchdecisionproblemsinexperimentalsettings.Skog(1999)seemstoviewthechoiceruleasdescriptivelyinaccurate,butanidealtowardswhichwemightstrive.Othersviewtheassumptionofdetailedplanningoflifetimeconsumptionpathsassoobviouslyfalsethereisnoneedtodisproveit.Anoft-quotedexampleisWinston1980,302.SinceWinstonÕsarticleiscitedforotherreasonsbyBeckerandMurphy(1988),theypresumablyknewthisdecisionrulewasseenasimplausible.Howdotheyjustifyit?TheansweristhattheydonÕt.TheystatethattheirÔÔpaperreliesonaweakconceptofrationalityÕÕsinceindividualsmaystronglydiscountfutureutility(BeckerandMurphy1988,683).Thisisaweakanchor,sinceitdoesnotrelatetotheassumptionofintricate,long-rangeplanningthattheirrationalityassumptionrepresents.Theyalsonotethatsmokersdonotneglectthefuture,sincesmokingwasreducedwhenhealthdangersbecameknown(687).Thistoo,isaweakanchor,sincethereisavastoceanbetweencompletemyopiaandlogicalomniscience.ThisstrangeÔÔunlessyouareobliviousofthefuture,youwillbemakingintricate,longrangeplansÕÕthinkingisnotauniquefeatureofBeckerandMurphy,butalsoappearsinthereviewofÔÔTheEconomicsofSmoking,ÕÕwhose7.SeeFerguson2000foraclearexpositionofthemathinvolved.olerogeberg 04353UCP:PHOSarticle#710302authorssupporttheassumptionofforward-lookingrationalitybystatingthatmyopicbehaviorimpliesaninÞnitediscountingofthefuture,whilerationalbehaviorimpliesthatfutureimplicationsareconsidered,whilenotrulingoutarelativelyhighdiscountrate.(ChaloupkaandWarner2000,1559)Thereisaninterestingneglectofnegativeevidencehere:Theas-sumptionshouldbeeasytosupportifcorrect.Sincerationaladdictshavemadeforwardlookingplans,theyareawarewhentheyÞrsttakeupsmokingthatÔÔIÕmsmokingtwoadaynow,nextweekIÕmgonnasmokefour,inamonthIÕllbeuptotenaday,andfromOctobernextyearIÕllbesmokingÞftyadaytillIquitatage34.ÕÕAsecondpieceofnegativeevidenceconcernstheproblemssolvedandpresentedineconomicjournalswhichseem,despitetheirsophistication,verysimplecomparedtothefullcomplexityarealpersonÕsdecisionproblemwouldhave.Thisisstrange:Ifrealpeopleroutinelysolvethefull,complexproblems,wouldwenothavesomewayofcommunicatingthereasoningbehindouractions?WhywouldweÞnddescriptionsofthedecisionproblemanditssolutionsobizarre?WhywouldsimpliÞedversionsseemcomplex?TopointtoamarketÔÔanomalyÕÕ:Ifpeopleroutinelysolveandimplementthemostfantasticallysophisticatedplansderivedfromaccurateknowledgeandclarityofpreferences,whywouldtherebesuchaplethoraofbooksattemptingtoteachtime-management,personaldecisionmaking,goalandprioritysetting,valueclariÞcation,etc.(abibliographyoftimemanagementbooksinCovey,Merrilletal.[1994]2003listswelloverahundred)?Whilenotsupported,theunrealisticplanningandperfectforesightoftheoriginalrationaladdictionmodelwasdifÞculttooverlookandhardtoswallow.TheseaspectsthereforebecamepartofthemotivationforOrphanidesandZervos(1995),whosetheoryofÔÔRationalAddictionwithLearningandRegretÕÕstatedthatÔÔthebulkofobjectionsconcerningearlierrationalmodelscanbeattributednottorationaldecisionmaking,butrathertothecommonimplicitassumptionofperfectforesightÕÕ(740).TheauthorsthereforeintroduceuncertaintythroughthreeÔÔfundamentalpostulatesÕÕ:[T]hatconsumptionoftheaddictivegoodisnotequallyharmfultoallindividuals,thateachindividualpossessesasubjectivebeliefstruc-tureconcerninghispotentialtobecomeaddicted,andthatthisbeliefstructureisoptimallyupdatedwithinformationgainedthroughconsumption,viaaBayesianlearningprocess.(740)takingabsurdtheoriesseriously 04353UCP:PHOSarticle#710302Theirmodelassumesthatpeopleareeithervulnerableorinvulnerabletoharmsfromaddictivegoods.Theywouldwanttousedrugsatlowlevelsiftheyknewtheywereinvulnerable,notatalliftheyknewwerevulnerable.Todiscoveryourtypeyouhavetoconsumedrugsandnotewhetherharmsfollow.Yourpriorbeliefsaboutyourtypedeterminewhetheritisrationaltoexperimentwithlowlevelconsumptionofdrugs.Ifharmfollows,theuncertaintyisresolved.Sinceharmsoccurstochas-ticallyevenforvulnerables,avulnerablemighthavebeenconsumingdrugsforsometimebeforehelearnshistruetype.Ifhehasconsumedatalowlevelforalongtime,theconsumptionstockhasincreasedtoalevelmakingitoptimaltoimplementanincreasetowardsahighlevelofconsumptionandthusbeanÔÔaddict.ÕÕIftheharmhappensafterashortperiodoflowlevelconsumptionitisrationaltoquit.AsanattempttoavoidtheunrealisticforesightassumedinBeckerandMurphyÕsmodelthisisafailure:Thereasonpeopledonotwishtobeaddictedisnowthattheyhavecalculatedthepresentvalueofimple-mentingtheconsumptionpathswecalladdictive,andfoundthesetohavelowpresentvaluescomparedtoalternatives.Inotherwords,merelytodislikeaddictionsinthismodelrequirestheforesightandplanningabilitythemodelseeminglydoesawaywith.Also:Unstableconsumptionisstilltheimplementationofarationalanddetailed,forward-lookingplan.Suchplansareimplementedwhentheuncertaintyisresolvedandthepersondiscoversthatheisavulnerable.ÔÔIhavenÕtsmokedmuchinthepastbutIhadasmokersÕcoughsobadmyvisionblurredthismorning.Conse-quently,Iplantoreachfortyadaybytheendofnextmonth.ÕÕWhilenotdenied,theseforward-lookingplanningaspectsofthemodelarenotemphasizedorsupportedbystoriesinthenon-technicalcom-ments.Theverbaldiscussioninsteadsuggeststhenotionthatpeopleareafraidofaddiction,thephenomenonleftuninterpretedandperhapsseenasanundesirableillness.This,andthenotionthatpeoplehaveasub-jectiveestimateoftheirriskofbecomingaddicted,seemsmoreplausibleandintuitivetomany.PhilosopherJonElster,forinstance,statesthatthemodelÔÔcapturestheinvoluntaryaspectofaddictionthatiscentralinreal-lifecasesÕÕ(Elster1999b,185).TheauthorsthemselvesseemtoimplythisinterpretationwhentheywriteofpeoplebeingÔÔdrawnintoaddic-tionÕÕ(OrphanidesandZervos1995,741),ofaddictionbeingÔÔtheun-intendedoccasionaloutcomeofexperimentingwithanaddictivegoodÕÕ(741)andofaddictionneverbeingÔÔtheresultofadeliberateplanbutratherofwhatprovestohavebeenanincorrectassessmentofthepos-sibilityofbecominganaddictÕÕ(751).Thislastoneisinterestinginthataddictionsareadeliberateplanintheirmodel,addictsjustratedtheprobabilityofthisplanbecomingoptimalaslow.NotealsothatthesecondÔÔfundamentalpostulateÕÕquotedstatesthatpeopleestimatetheirolerogeberg 04353UCP:PHOSarticle#710302ÔÔpotentialtobecomeaddicted,ÕÕwhereasinthemodeltheyestimatetheirpotentialforexperiencingharms.Thedifferenceisimportant,asweseewhenOrphanidesandZervos[c]onsideranaddictiontoalcohol.Tomostpeople,currentcon-sumptionofalcoholinmoderationprovidesenjoymentwithoutanyserioussideeffects.Toothers,thesamestreamofconsumptionmayleadtoadependenceonalcohol,aharmfuladdiction.Apredispo-sitiontoaddiction,however,cannotbedetectedwithouttheexperi-encegainedfromrepeatedconsumptionofalcohol.(741)Rewritingthisinlinewiththemodelmakesitnotonlylesspersuasivebutcounterintuitive:Therationalresponsetoexperiencingseriousharmsfrommoderateconsumptionofalcoholisoftentoplanandimplementan increaseindrinking.Thisdetailedplanforfuturedrinkingcanbecalledan‘‘addiction,’’andthelongeryouhavebeenanormal,moderatedrinkerbeforeyouexperienceharm,themorelikelyitisthatyouwillwanttoincreaseyourconsumptiongraduallytowardsaddictivelevelswhen,forinstance,yourlivergivesout.Again,hadtheassumptionsbeencorrect,betterstorieswouldhavebeenavailable.Thetheoryclaimsthata)peopleconsumelargeamountsbecausetheyexperiencedharmsfromlowamounts,b)theyincreasetheirconsumptiononlyiftheharmsÞrstappearafterconsumingatnormallevelsforarelativelylongtime,andc)thoseexperiencingharmsatearlierstagesquitconsuming.Theseimplicationsshouldbeeasytoexamineempirically.Especiallyclaima)seemstoßyinthefaceofotherviewsonaddiction,wherepeopleexperienceharmsbecausetheyconsumeathighlevels,andarecalledaddictedbecausetheyconsumeatthesehighlevelsinspiteoftheharm.Secondly,wewouldnothavethephenomenonofÔÔdenialÕÕwherepeopledenythattheirhigh-levelconsumptioncausesthemharm.Instead,peopleshouldsayÔÔofcourseIamexperiencingproblemsfrommyconsumption.WhyelsewouldIbeconsumingmoreandmore?ÕÕFinally,asElster(1999b,183Ð185)notes,peopleshouldwanttogetanduseinformationonriskfactorsforaddictions(suchassocialbackground,personalcharacteristics,etc.)toimprovetheaccuracyoftheirbeliefs.Thismeansthatthosewithfewriskfactorsshouldex-perimentmorethanthosewithmany,otherthingsequal.Inotherwords,thefactorspredictinguseshouldbetheÔÔoppositeÕÕofthosepredictingaddiction.8.SincethenumberofÔÔaddictsÕÕisthenumberofexperimenterstimesriskofbecominganaddict,thenumberofrationallowriskuserscouldbecamesomuchlargerthanthenumbertakingabsurdtheoriesseriously 04353UCP:PHOSarticle#7103024.3.ASimple,Stable,andTransparentWorld.Theplanninginvolvedinrationaladdictiontheoriesrequiresbeliefsofakindwearenotawarethatwepossess,andthatwewouldnothaveevidenceforifwedidpossessthem.Thebeliefsinvolvedarethekindrequiredtoposetheforward-lookingdecisionproblemonthemathematicalformchosenbyeconomists,andmightthereforebeviewedasamathematicalmetaphortakentoolit-erally.AsKeynesputit,Mucheconomictheorizingto-daysuffers,Ithink,becauseitattemptstoapplyhighlypreciseandmathematicalmethodstomaterialwhichisitselfmuchtoovaguetosupportsuchtreatment.(quotedinChick1998,1864)ThedangeristhatthemetaphormaybemisleadingÑinourcasebyassumingprecise,quantitativelyspeciÞedrelationshipsbetweenvaria-blesÑinwaysthatplaytoolargeapartinthemathematicalargumentÑinourcaseenablingtheidentiÞcationofaprecise,quantitativelyspeciÞedconsumptionpaththatmakesnosenseoutsideofthemathematicalmet-aphorwehavecreated.Toseethis,notethatthemathematicalsolutionsusedbythetheoriesrequirethatthestocklevelsandinvestmentequationsareknown:YoucanÕtdesignanoptimalinvestmentplanifyouractionscauseyoutoinvestunknownamountsintoanunknownassetwithunknownrateofreturnthatstartsfromanunknownlevel.Thiscreatesanumberofproblems.First,sinceaddictionsmayinvolvevariousdrugsaswellasÔÔwork,eating,music,television,theirstandardofliving,otherpeople,religion,andmanyotheractivitiesÕÕ(BeckerandMurphy1988,676),constructingthestocksishard:Anythingwithsimilarinßuencesonwelfareandincentivesshouldbesummarizedbyasinglestock;e.g.,ifthedeathofawifeÞllsahusbandwithgriefandincreasesthevalueofescapethroughdrugs,thisisashocktotheÔaddictionstockÕthehusbandcouldseeasequivalenttoahistoryofheroinuse(BeckerandMurphy1988,690Ð691).HavingidentiÞedthestocks,theyshouldbeoperationalized,sothatinitiallevelscanbeidentiÞedandchangesmonitored.E.g.whathistoryofheroinusewouldequalthedeathofaspousewouldequalwhatgeneswouldequalwhatattitudesandexpectationsexpressedinspeciÞedways ofrationalhighriskusersthatthisoutweighedthedifferenceinrisk.Mostaddictswouldthenbefromthelowriskgroup,andifriskestimateswerenotconditionalonuse,thismightbemisinterpretedtosuggestthathighriskuserswerelowriskusersandviceversa.ThiscouldpossiblygenerateÔÔinterestingdynamics.ÕÕolerogeberg 04353UCP:PHOSarticle#710302byvariouspeersetc.?Asmanagerssay:IfyoucanÕtmeasureit,youcanÕtmanageit.Thirdly,beliefsinquantitativetermsaboutthedynamicsarerequired.E.g.asmokingteenagerneedsapreciseestimateofhowmuchandinwhatwaybothhisnextcigaretteandthepassingoftimewillinßuencehisbodilyorgans,personalhabits,futureexperienceofsmokingandwithdrawal,relationshipswithvariouspeers,familymembers,teachers,etc.Healsoneedssuchbeliefsaboutfutureprices,tobaccopolicies,norms,etc.Noteespeciallythebeliefsthatshouldbeformedconcerninghowyourtasteswillchangeinthefutureasaresultofpresentexperiencesandchoices.Conscioustaste-planningisanimportantpartofthemotivationforrationaladdicts.Unfortunately,asshowninLoewensteinandAdler1995,peoplefailtodisplayawarenessofeventheÔÔsimplecaseÕÕin-volvingthesudden,robust,andhighlysystematictastechangescausedbyacquiringanobject(ÔÔtheendowmenteffectÕÕ).Concerningtheprecisionrequiredofthebeliefs:PlansofthekinddiscussedwouldbedifÞcultwithÔÔqualitativebeliefsÕÕ(ÔÔsmokingdam-agesfuturehealthÕÕ)inplaceofnumericallyspeciÞedeffects,sincethetrade-offsbecomeunclear.Attemptingtoseequalitativebeliefsasamorecomplicatedformofquantitativebeliefs(withasubjectiveprobabilitydistributiondeÞnedoverdifferentquantitativebeliefs)willworsentheproblemratherthanavoidit(seeSimon1979,505).Evenassumingquantitativebeliefs,thesewillnotbewell-informedatthelevelofprecisionanddetailrequireddespitethefactthatÔÔsome70,000scientiÞcarticleshaveimplicatedsmokinginawidevarietyofailments,constitutingthelargestandbestdocumentedliteraturelinkinganybehaviortodiseaseinhumansÕÕ(ChaloupkaandWarner2000,1545).Thereasonisthatitisstillunclearhowsmokingpatternstranslateintohealthrisksattheindividuallevel.Smokinglittleisbetterthansmokinglots(LeffondreAbrahamowiczetal.2002),butreducingyoursmokingmaynotreduceyourrisk(Godtfredsen,Holstetal.2002).ÔÔ[S]mokingstatus,intensity,duration,cigarette-years,ageatinitiation,andtimesincecessationÕÕareallpossiblyimportantvariablesinmodelingtheimpactofsmoking(Leffondre,Abrahamowiczetal.2002,813),butnoconsensusabouttheireffectsexist.Eventhesizeandspeedofhealthimprovementsforquittersisofuncertainmagnitude,withpersistenceofelevatedriskofcoronaryheartdiseaseforex-smokersbeingestimatedatboth5yearsand30years(Godtfredsen,Holstetal.2002,999).Threefurtherproblemsfollowfromthis:Theactualcausalmecha-nismswillbefarmorecomplexthanwhatrationaladdictiontheoriesassume,andnotjustfortobacco.Skog(1999,188Ð189)arguesthatthiswouldalsobethecaseforalcohol,andthereisnoreasontobelieveothertakingabsurdtheoriesseriously 04353UCP:PHOSarticle#710302setsofcausalmechanismswillbemoreaccuratelysummarizedbythesimpleandstabledifferentialstockequationscommonlyassumed.Sinceforwardlookingplanningissensitiveinthatitgeneratesdifferentcon-sumptionpatternsfordifferentbeliefsaboutcausalmechanisms,thismeansthatÔÔpredictivesuccessÕÕofrationaladdictiontheorywouldonlyestablishthatrealpeopleactinawaythatwouldbeoptimaliftheywerefacingaproblemweknowtheyarenot.Second,thelackofwell-known,detailedknowledgeabouteffectsmakesitprobablethatdifferentpeoplewillhavedifferentbeliefsconcerningthenumberofeffects,theirstrength,lags,etc.Thismeansthatwemightexpectlargebehavioraldifferencesduetounobservabledif-ferencesinbeliefs,whichisproblematicinthatitbecomesdifÞcultorevenimpossibletogeneratepotentiallyfalsiÞablepredictions.Someonewhoexpectssuchbeliefstovaryalotmightevenseesimilarbehaviorpatternsacrosspeopleandtimeasananomalyforthetheory.Finally,thelackofpreciseknowledgemeansthatthemodelismis-speciÞed:Well-informed,rationalpeoplewouldrealizetheirignoranceofthetruestructureoftheirdecisionproblem.TheirresponseshouldbetodeÞneaprobabilitydistributionoverallpossiblespeciÞcationsofthedecisionproblem.Theresultmightbearobustsolutionthatdoestoler-ablywellacrossmanyspeciÞcations,takingintoaccountthatdifferentplansleadtodifferingamountsofcorrectivefeedbackabouttheworld.OrphanidesandZervos(1995)areontherighttrackinthisregard,buttheirindividualsknowthattheirdecisionproblemisoneofonlytwopossible.Inreality,constructingthesetofallpossiblespeciÞcationsofthedecisionproblemseemsdaunting,muchlesssolvingthem.4.4.AssumeAway,Mr.Storyteller.Theprevioustwosectionsconcen-tratedonbringingoutthemostlyundiscussedandhiddenassumptionsmadeinthemathematicsofrationaladdictiontheories.Theremainingquestioniswhyeconomiststakethesetheoriesseriouslyaspossiblyvalidexplanationswithapossiblevalidityforwelfareanalysis.Theanswerseemstobethatthetheoriesseemconsistentwiththerationalchoiceframeworkthateconomistsareusedtoseeing,whilepresentingstoriesthatgivefeelingsofinsightandthattriggerfeelingsofunderstanding.Agoodexampletoillustratehowtheseassumptionsaresupportedconcernsthenecessaryconditionsforaharmful,rationaladdictioninBeckerandMurphy1988:Themoreyousmoke,snort,inject,orother-wiseconsumeofyouraddictivegoodtoday,theworseoffyouwillbetomorrow(theleveleffectofthestock)butthemorerewardingthead-dictiveconsumptionwillbecome(thehigherthestock,thehigherthemarginalutilityoftheaddictivecommodity).Manystoriescanbeusedtoillustratetheseeffects:Theycanbeinterpretedaswithdrawal,laggedolerogeberg 04353UCP:PHOSarticle#710302healtheffects,etc.(Skog1999,175),theyseemplausible,relevant,andimportant.Theproblemisthatthesestoriesareselectedad-hocpreciselyforthisreason,andthatcontradictoryassumptionscouldbeequallywellsupportedbyadifferentselectionofstories.Forinstance,wecouldtellstoriesthatgoagainstthenecessaryas-sumptions:Ayoungboycanbecomesomuchbetteroffthroughraisedself-esteem,status,andarichersocialnetworkwhensmokingallowshimtojointhein-crowdathisschool,thattheleveleffectofsmokingbecomespositiveandsmokingbecomesabeneÞcialaddiction.Asmokingadultcanreceivesuchpangsofremorsefromeachcigaretteifheknowsthatsmokingbothdisappointshisnon-smokingfriends,colleagues,andfamilymembersandmakesthemseehimasapatheticaddict,thattheeffectofthestockonthemarginalutilityofcigarettesisreversed.Accordingtothetheory,anaddictionwouldbeimpossible.TodisregardsuchstoriessimplybecausenorationalindividualwouldconsumeaddictivelyunlessthenecessaryconditionswerefulÞlledwouldbetobegthequestionbyusingthetheorytoconstraintthestoriesaboutrealityweaccept.Theeaseoftellingstoriessuggeststhatalmostanyassumptionsofqualitativeeffectscanbesupportedbycreativethought.Forinstance:Justasstoriescanbefoundimplyingthatrealaddictsfailthenecessaryconditionsforrationaladdictions,otherstoriesimplythatactivitiesnotusuallyconsideredaddictivemaysatisfythese:Scratchinganitchincreasesfutureitching(leveleffect),butalsoreliefprovidedbyeachfuturescratching(marginaleffect).Therationalscratchingaddictmapsouthislifetimescratchingplanandscratcheswithincreasingintensityuntilhereachesahigh,stablelevel.IfweaddinaÔÔskinsorenessstockÕÕwemightsupportatheoryofrationalcyclicalscratching.Withstochasticeffectsandpeoplewiththickandthinskinwecouldgetrationalscratchingaddictionswithlearningandregret.Ofcourse,sometimessuchstoriesaremoreeasilyfoundthanatothertimes.OrphanidesandZervos(1995),forinstance,lackgoodstoriesfortheassumptionsthatthereareinvulnerables(or,moregenerally,initiallyinvisibledifferencesinvulnerability).Cansomepeoplesmokewithoutriskinglungcancerorbotheringnon-smokers?Getdrunkwithouthavingtheirspeechslurred,theirthoughtconfusedand,withlongtermcon-sumption,liver,andbraindamaged?Aswehaveseen,theirverbalcommentsprovidestoriesconcerningdifferencesinsomeunspeciÞedaddictivetendencyratherthandifferencesinharmvulnerability.Bytellingstorieswecanalsoridourtheoryofanomalies;e.g.,oldpeopledonotneedtoworryaboutlongtermconsequences.SinceBeckerandMurphydoubttheimplicationthattheelderlywillbeaddictsmoreoftenthanothers,furtherstoriesareintroduced:takingabsurdtheoriesseriously 04353UCP:PHOSarticle#710302Ofcourse,otherthingsarenotusuallythesame:olderpeoplearelesshealthyandsubjecttodifferentlifecycleeventsthanyoungerpeople.Moreover,peoplewhomanagetobecomeoldarelesslikelytobestronglyaddictedtoharmfulgoods.(BeckerandMurphy1988,684)Now,assumethatthepredictionhadbeenwrongintheoppositedi-rection:Ofcourse,oldpeoplehavefrailbodiesthatarepoorat‘pro-ducingutility’fromphysicalactivities(sports,dancing,sex,etc.).Theycompensatebyshiftingtowardsmentalexperiencesprovidedbydrugs.Ratherthanlyingalonewaitinginvainforfamilyvisits,theyrelaxinthesoft,sweetdazeofheroinortakedazzlingtripsonLSD.Finally,wemaynotethatstoriesonlyprovidequalitativeeffectsandthatthedifferentstorieswecollectmayapplytodifferentindividuals.Ourendresultmaythusbearepresentativeagentmodelthatdoesnotrep-resentanyone,withwhateverqualitativeeffectswedesire,andwithquantitativemagnitudesofeffectsopenformanipulation.4.5.WhoorWhatAreWeTalkingAbout?ManyofthestoriesusedtosupportthespeciÞcationofrationaladdictiontheoriesrefertointuitionandcommonsenseandpresupposethatintrospectioncanprovidevalidinsightsintohowpeoplemakedecisions,i.e.thatweareawareoforcanuncoverthereasoningbehindourchoices.Sincetheassumptionsconcernbelievedconsequencesandperceivedincentives,theuseofintrospectionandcommonsenseseemshardtoavoid.Introspection,however,providesevenstrongerargumentsagainstthesetheories:ÔÔTV-addicts,ÕÕbingeeaters,orchainsmokingtobaccousershardlyÞtourideaofperfected,deliberate,consciousdecisionmakers.WeÞnditabsurdtothinktheyhavemappedtheirfuturelivesinalldetails.Wehavenoawarenessofprecise,quantitative,andwell-informedbeliefsconcerningthecomplexlawsofnatureandsocietyrequiredforstatingtheoptimalcontrolprob-lemrationaladdictiontheoriesclaimthatwesolve.Addinthatweshouldtakethesebeliefsintoaccountinacomplicatedwaythatwouldbafßeanyonewithoutextensivemathematicaltraining,thatweshoulddothisevenwhentheproblemuniquelyfacingusismorecomplexthanourformalmathematicscanhandleinpractice,andalltoyieldplansthatfewifanyofusareawarethatwepossess,andthetheorybecomeshardtotakeseriously.Especiallywhenwerecallthatthewhole,theoreticaled-iÞcehasbeenraisedtoexplainwhypeoplearelazy,failtotakecareoftheirhealth,consumeharmfulsubstancesthatarepleasurablehereandnow,etc.Itisaparadoxthattheverytheoriesclaimingthataddictionsarewellinformed,rational,forwardlookingdecisionsclashstronglywithhowweperceiveourowndeliberateandconsciousdecisionprocesses.Thiscanolerogeberg 04353UCP:PHOSarticle#710302beavoidedbydenyingthatchoicesaremadeconsciously.InanearlierapproachBeckerdidthisbystatingexplicitlythatthedecisionproblemcouldbeformulatedandsolvedatsomenot-clearly-identiÞedandnot-necessarily-consciouslevel,claimingsupportfrompsychologyÕsbeliefinthesubconscious(referencesinRosenberg1979,513).AsnotedbyRosenbergsuchadefenseofthetheorymeansthatÔÔtheordinaryinter-pretationofthetheorywhichcommon-sensespeciÞcationseemstosuggestisruledoutÕÕ(Rosenberg1979,525).Inotherwords:IfenjoymentBeckerhasadifferentmeaningfromenjoy-mentcommonusage,thenthefactthatsomeoneconsciouslyenjoyscommonusagetheeffectsofcannabisbecauseitenhancestheirpleasurecommonusagefromfood,sexormusichasnoimplicationforhowcannabisaffectstheirpleasureBeckerfromfood,sexormusicandwhethertheseeffectsareenjoyedBeckerornot.WhyshouldevengoodsBeckerbeassumedtoinvolvethesamecategoriesasgoodscommonusage?Thepointissimple:WhenwebeginreferringtoXbyanameweuseforsomethingelse,thisdoesnotteachusanythingaboutX.Denyingthatchoicesareconsciousmeansthatintuition,plausibility,andcommonsensenowhasnovalueforspecifyingthetheoryorjusti-fyingitsassumptions.Weneedtospecifywhatthedecisionmakingunitactuallyis,andwhatitmaximizes:DowewantanextremeselÞshgenetheorythatexplainsallbehavior(includingvoluntarychastity,suicide,homosexuality,art,science,drugtakingetc.)aspurelyattemptstomaximizethespreadoftheorganismÕsowngenes?DowewantaFreudiantheorywhereunobservable,postulatedforcesofId,Ego,andSuper-egoclashandconßictcausingbehaviorrichlysymbolicofre-pressedchildhoodmemoriesandsexualconßicts?Ordowepostulatetheunconsciousassomeguardian-angelsupercomputerthatsolvesproblemstheconsciousnessfacesbutisunabletoformulateandunderstanditself?Howdoesitdecide,plan,getinformation,updatebeliefs,etc.?Andonwhatpossibleevidencecouldwebaseourassumptionsaboutthesematters?Finally,ifthedecisionmakerisnottheconsciousperson,howdoweweightheÔÔwelfareÕÕ(orÔÔoverallevaluationÕÕifyouwill)ofthedecidingsubconsciousagainstthewelfareofthedeludedandpassivelyexperi-encingalong-for-the-rideconsciousness?Shouldwecare,inapolicycontext,abouthowÔÔsatisÞedÕÕthesubconsciouswaswithitsattempttomaximizethespreadofitsgenes?Ifwecouldthwartthesubconsciousinawaythatmadetheconsciouspersonfeelmorehappy,satisÞedand9.ThisextremeselÞshgeneviewisnotintendedasarepresentationoftheviewsofRichardDawkins,coinerofthephraseÔÔselÞshgene,ÕÕwhoacceptsthatwemayconsciouslychooseactionsthatgoagainsttheÔÔwishesÕÕofourgenes.takingabsurdtheoriesseriously 04353UCP:PHOSarticle#710302valuable,shouldwedoit?Thisrelatestotheearlierdiscussionofwhetherthewelfareclaimscanbevalidwhentheassumptionsarenot:Thetheoriesofrationaladdictionareclearlynottrueattheconsciouslevel.Iftheyaretrueatsomesubconsciouslevel,shouldwestillcareaboutthewelfareanalysisthatßowsfromit?Ifthetheorywascorrectwhateveranswerwegivetothesequestionswouldactuallyserveoursubconscious,soIÕllleavethematterhere.5.Conclusion.Ihavehereusedrationaladdictiontheoryasanexampleofchoicetheoriesthatareabsurdwhenunderstoodintherealistsenserequiredbytheclaimsmadeforthem.Thesetheoriesshowhowaloose,unstructuredapproachtoexplainingandjustifyingamathematicalmodelallowsonetohideproblematicassumptionsevenwhenthesearecentraltotheargumentmade,whileprovidingad-hocillustrationsthattriggerfeelingsofunderstandingandinsightthoughneitherjustifyingtheassumptionsnorprovidinganadequateexplanationinobjectivisticterms.Readersagreeingatthispointthatthetheoriesareabsurdmightstillwonderwhethereconomistsreallytakethesetheoriesseriously.Myan-swerissplit.Ontheonehand:Yes.Theytakethemseriouslyintheirwritings.Theyselltheirarticlesbyclaimingrealworldrelevancefortheirexplanationsandthederivedwelfareanalysis.Forinstance,intheirsecondaddictionarticleOrphanidesandZervosconcludethattheiranalysissuggeststhatharmfuladdictivebehaviorislargelyconsistentwiththestandardaxiomsofrational,forwardlookingutilitymaxi-mizationandshouldnolongerbeconsideredachallengetostandardeconomicanalysis.Rather,aremainingchallengeistoemploystandardeconomicanalysisforfurtherexaminationofthewelfareimplicationsofaddictiveconsumptionandprovidebetterguidanceforpublicpolicydesign.(1998,89)Ontheotherhand:No.Iampersonallyunabletobelievetheseeconomistsreallytakesuchtheoriesseriously.Iftheytrulybelievedthesetheoriesheldvalueforexplanationandwelfareanalysis,thenIbelievetheywouldhaveexplicitlyandindetailjustiÞedthissurprisingbelieftoconvinceskeptics.Untilagoodcasefortakingthesetheoriesseriouslyhasbeenmade,Ibelieveitmightmakesensetostopseeingthemasfailedtheoriesabouttheworld.Instead,theycouldbeseenassuccessfulmovesinagamewherethegoalispeacockdisplaysoftechnicalingenuityandskillwhileexplainingallbehaviorasthechoicesofwell-informed,utilitymaximizingindividuals,withhighermarksthelessplausibletheseassumptionswouldseeminitially.Oneoftherefereestothispapersharedolerogeberg 04353UCP:PHOSarticle#710302astorythatwouldÞtsuchaninterpretation,wheretheauthorsintheirworkonlypretendedtobelieveinrationalchoicetheoriesbecausetherulesofthegamedemandedit.TheirworkÔÔsolvedÕÕapuzzletherefereepointedoutwouldpuzzleonlysomeoneignorantofwell-establishedÞndingsfrompsychology.ÔÔTheauthorsofthepaperÑbothsmartecon-omistsdoingveryinterestingworkthatspansdisciplinesÑrepliedthattheyrecognizetheirpositionispsychologicallyinaccurateanddeliber-atelynaõ¬ve,buttheyaretryingtopublishthepaperinaneconomicsjournal,andthedominantassumptionsintheÞeldrequireit.ÕÕOnemightofcoursefeelthatadultsshouldbeallowedtoplaywhat-evergamestheylikeaslongastheyareconsensual.TheproblemwiththechoicetheorygameÑifitisindeedonlyagameÑisthatitisplayedasthoughitwasascientiÞcdisciplinedoingresearchrelevanttotherealworld.Apparently,therulesprohibitstatingwhenyouorothersarejustplayingthegame,thusblurringtheboundariesbetweenthegameandthescienceofeconomics.TheauthorsmightknowwhattheyÕredoing;thereadersmightnotbeinonthejoke.Someofthemgetfooledbystoriesintheorieslooselyreadorbrießyreferredtointothinkingthateconomistshaveactuallydonevalidresearchrevealing,forinstance,drugaddictstoberationalandaddictionsunproblematicfromawelfareperspective.OtherswriteofflargeportionsoftheeconomicsdisciplineÑincludingitssensiblepartsÑasnonsense.Bothconsequencesareundesirable,andbothcouldbeavoidedifeconomistswouldadmittoplayingtheory-gamesifandwhenthatisindeedwhattheyaredoing.Or,alternatively,iftheycouldtelluswhyweshouldbelievethesurprisingclaimstheykeepmaking.Adda,Jeme,andValerieLechene(2001),SmokingandEndogenousMortality:DoesHeterogeneityinLifeExpectancyExplainDifferencesinSmokingBehaviour?DiscussionPaper77,DepartmentofEconomicsOxford:UniversityofOxford.Becker,GaryS.(1996),ÔÔPreferencesandValuesÕÕ,inAccountingforTaste.MA:HarvardUniversityPress,3Ð23.Becker,GaryS.,andCaseyB.Mulligan(1997),ÔÔTheEndogenousDeterminationofTimePreferenceÕÕ,QuarterlyJournalofEconomics112(3):729Ð758.Becker,GaryS.,andKevinM.Murphy(1988),ÔÔATheoryofRationalAddictionÕÕ,ofPoliticalEconomy96(4):675Ð700.Chaloupka,FrankJ.,andKennethE.Warner(2000),ÔÔTheEconomicsofSmokingÕÕ,inHandbookofHealthEconomics,Amsterdam:Elsevier,1539Ð1627.Chick,Victoria(1998),ÔÔOnKnowingOneÕsPlace:TheRoleofFormalisminEconomicsÕÕ,EconomicJournal108(November):1859Ð1869.Covey,StephenR.,A.RogerMerrill,andRebeccaR.Merrill([1994]2003),FirstThingsNewYork:FreePress.Cowen,Tyler(1989),ÔÔAreAllTastesConstantandIdentical?ACritiqueofStiglerandBeckerÕÕ,JournalofEconomicBehaviorandOrganization11(1):127Ð135.takingabsurdtheoriesseriously 04353UCP:PHOSarticle#710302Dawes,RobynM.(1999),ÔÔAMessagefromPsychologiststoEconomists:MerePredict-abilityDoesnÕtMatterLikeItShould(withoutaGoodStoryAppendedtoIt)ÕÕ,ofEconomicBehaviorandOrganization39:29Ð40.Dockner,EngelbertJ.,andGustavFeichtinger(1993),ÔÔCyclicalConsumptionPatternsandRationalAddictionÕÕ,AmericanEconomicReview83(1):256Ð263.Elster,Jon(1997),ÔÔMoreThanEnoughÕÕ,reviewofAccountingforTastes,byGaryS.Becker.UniversityofChicagoLawReview64(2):749Ð764.ÑÑÑ(ed.)(1999a),Addiction:EntriesandExits.NewYork:RussellSageFoundation.ÑÑÑ(1999b),StrongFeelings:Emotion,Addiction,andHumanBehavior.MA:MITPress.ÑÑÑ(2000a),AlchemiesoftheMind:RationalityandtheEmotions.CambridgeUniversityPress.ÑÑÑ(2000b),UlyssesUnbound:StudiesinRationality,Precommitment,andConstraints.Cambridge:CambridgeUniversityPress.Elster,Jon,andOle-J¿rgenSkog(eds.)(1999),GettingHooked:RationalityandAddiction.Cambridge:CambridgeUniversityPress.Fehr,Ernst,andPeterK.Zych(1998),ÔÔDoAddictsBehaveRationally?ÕÕJournalofEconomics100(3):643Ð662.Feichtinger,Gustav,CarsH.Hommes,andAlexandraMilik(1997),ÔÔChaoticConsumptionPatternsinaSimple2-DAddictionModelÕÕ,EconomicTheory10(1):147Ð173.Ferguson,BrianS.(2000),ÔÔInterpretingtheRationalAddictionModelÕÕ,HealthEconomics9:587Ð598.Friedman,Milton(1953),ÔÔTheMethodologyofPositiveEconomicsÕÕ,inEssaysinPositiveChicago:UniversityofChicagoPress,3Ð43.Friedman,Milton,andLeonardJ.Savage(1952),ÔÔTheExpected-UtilityHypothesisandtheMeasurabilityofUtilityÕÕ,TheJournalofPoliticalEconomy60(6):463Ð474.Gibbard,Allan,andHalR.Varian(1978),ÔÔEconomicModelsÕÕ,TheJournalofPhilosophy75(11):664Ð677.Godtfredsen,NinaS.,ClausHolst,EvaPrescott,J¿rgenVestbo,andMereteOsler(2002),ÔÔSmokingReduction,SmokingCessation,andMortality:A16-YearFollow-upof19,732MenandWomenfromtheCopenhagenCentreforProspectivePopulationStudiesÕÕ,AmericanJournalofEpidemiology156(11):994Ð1001.Hausman,DanielM.(1992),TheInexactandSeparateScienceofEconomics.CambridgeUniversityPress.Krugman,Paul(1998),ÔÔTwoCheersforFormalismÕÕ,EconomicJournal1829Ð1836.Kukla,Andre(1991),ÔÔCriteriaofRationalityandtheProblemofLogicalSlothÕÕ,ophyofScience58(3):486Ð490.Laibson,David(2001),ÔÔACue-TheoryofConsumptionÕÕ,QuarterlyJournalofEconomics116(1):81Ð119.Leffondre,Karen,MichalAbrahamowicz,JackSiemiatycki,andBernardRachet(2002),ÔÔModelingSmokingHistory:AComparisonofDifferentApproachesÕÕ,JournalofEpidemiology156(9):813Ð823.Levine,HarryG.(2000),reviewofÔÔGettingHooked:RationalityandAddictionÕÕ,95(3):449Ð452.Loewenstein,George,andDanielAdler(1995),ÔÔABiasinthePredictionofTastesÕÕ,EconomicJournal105(431):929Ð937.Miron,JeffreyA.,andJeffreyZwiebel(1995),ÔÔTheEconomicCaseagainstDrugProhi-bitionÕÕ,TheJournalofEconomicPerspectives9(4):175Ð192.Orphanides,Athanasios,andDavidZervos(1995),ÔÔRationalAddictionwithLearningandRegretÕÕ,JournalofPoliticalEconomy103(4):739Ð758.ÑÑÑ(1998),ÔÔMyopiaandAddictiveBehaviourÕÕ,EconomicJournal108(446):75Ð91.Rogeberg,Ole(2003),ÔÔPreferences,RationalityandWelfareinBeckerÕsExtendedUtilityApproachÕÕ,RationalityandSociety15(3):283Ð323.Rosenberg,Alexander(1979),ÔÔCanEconomicTheoryExplainEverything?ÕÕPhilosophyofSocialScience9:509Ð529.ÑÑÑ(1981),SociobiologyandthePreemptionofSocialScience,Oxford:BasilBlackwell.olerogeberg 04353UCP:PHOSarticle#710302ÑÑÑ(1992),Economics—MathematicalPoliticsorScienceofDiminishingReturns,Chicago:UniversityofChicagoPress.Simon,HerbertA.(1979),ÔÔRationalDecisionMakinginBusinessOrganizationsÕÕ,AmericanEconomicReview69(4):493Ð513.Skog,Ole-J¿rgen(1999),ÔÔRationality,IrrationalityandAddiction-NotesonBeckerandMurphyÕsTheoryofAddictionÕÕ,inJ.ElsterandO.J.Skog(eds.),GettingHooked.Cambridge:CambridgeUniversityPress,173Ð207.TroutJ.D.(2002),ÔÔScientiÞcExplanationandtheSenseofUnderstandingÕÕ,Philosophyof69:212Ð233.Winston,GordonC.(1980),ÔÔAddictionandBacksliding-aTheoryofCompulsiveCon-sumptionÕÕ,JournalofEconomicBehaviorandOrganization1:295Ð324.takingabsurdtheoriesseriously