The Relationship Between Financial Risk
Premia and Macroeconomic Volatility: Issues
and Perspectives on the Run-Up to the Turmoil
Massimiliano Marzo
Zhoushi Liu
Paolo Zagaglia
THERELATIONSHIPBET
THERELATIONSHIPBET
THERELATIONSHIPBET
THERELATIONSHIPBETWEENFINANCIALRISK
WEENFINANCIALRISK
WEENFINANCIALRISK
WEENFINANCIALRISKPREMIAAND
PREMIAAND
PREMIAAND
PREMIAAND
MACROECONOMICVOLATI
MACROECONOMICVOLATI
MACROECONOMICVOLATI
MACROECONOMICVOLATILITY
LITY
LITY
LITY
::::
ISSUESANDPERSPECTI
ISSUESANDPERSPECTI
ISSUESANDPERSPECTI
ISSUESANDPERSPECTIVES
VES
VES
VESONTHERUN
ONTHERUN
ONTHERUN
ONTHERUN
----
UPTO
UPTO
UPTO
UPTOTHETURMOIL
THETURMOIL
THETURMOIL
THETURMOIL
M MM
MASSIMILIANOASSIMILIANOASSIMILIANOASSIMILIANOMMMMARZOARZOARZOARZO,,,,ZZZZHOUSHIHOUSHIHOUSHIHOUSHILLLLIUANDIUANDIUANDIUANDPPPPAOLOAOLOAOLOAOLOZZZZAGAGLIAGAGLIAGAGLIAGAGLIAAAAiiii
ABSTRACT
ABSTRACTABSTRACT
ABSTRACT
This note sketches the issues that arise while interpreting the relation between macroeconomic volatility and financial risk premia from the perspective of the standard consumption-based asset pricingmodel.Therelationarisesfrom thefactthatall assetsarepriced bythe same ‘pricing kernel’, givenbytheinter-temporalmarginalrateofsubstitutioninconsumptionoftherepresentativeinvestor. Since the pricing kernel is a function of aggregate consumption, financial risk premia are positively related to consumption growth volatility. Therefore, from the perspective of this workhorse often employed in the academic debate, the persistent reduction in macroeconomic volatility can be consideredacauseforthelowaverageriskpremiaprevailingduringtheso-calledGreatModeration, namely the period preceding the recent turmoil in financial markets. We challenge this view by shedding light on the issues that generate an inconsistent interpretation of the model outcomes. In particular, since the consumption-based model is geared towards asset prices consistent with macroeconomic fundamentals, we argue that it is not suited for interpreting current developments where underestimation of risk may have subsidized asset prices. In particular, according to the evidencefortheGreatModeration,themodelviewsuffersfromobservationalequivalence. Keywords: Keywords:Keywords: Keywords:assetpricing,macroeconomicvolatility,GreatModeration. JELClassification: JELClassification:JELClassification: JELClassification:E43,G12. iMarzo:DepartmentofEconomics,UniversityofBologna;massimiliano.marzo@unibo.it; Liu:FinancialEconomicApplicationsTeam,Barrie&Hibberg;zhoushi.liu@barrhibb.com; Zagaglia:DepartmentofEconomics,UniversityofBologna;paolo.zagaglia@gmail.com.
This material is for academic purposes only. The views expressed herein are solely those of the authors and shouldnotbeattributedtoanypublicorprivatesectorentity.
1.
1.1.
1. IntroductionIntroductionIntroductionIntroduction
Itisafactthata’GreatRecession’hasfolloweda‘GreatModeration’intheworldeconomy(e.g.,see Bernanke,2004).Thishascoincidedwithaboom-bustcycleinfinancialmarketsworldwide.Inother words,therewasabuild-upoftensionsthattookplaceinthemarketsuntil2007.Theaccumulationof excessiverisk-takinginthemarketseruptedinAugust2007intheformoffreezeininterbanklending. Interestingly,theacademicworldhasviewedtheGreatModerationatthemacroeconomiclevelas providingthefoundationsforamoderationofriskpricinginfinancialmarkets.Thislineofargument arisesfromtheinterpretationofaworkhorseforassetpricingthatisembeddedinthecurrent generationofframeworksusedformacroeconomicanalysis,theso-called‘generalequilibriummodels’. Inaconsumption-basedmodel,assetsarepricedbydiscountingfutureyields.Sincetheprocessof portfolioallocationaffectstheabilityofhouseholdstoconsume,pricingassetsinvolvesaconsistent planningofconsumptionacrossdifferentpointsintime.Asaresult,thediscountfactorforassetsisa functionofconsumption.Astandardresultisthatassetreturnsinexcessoftherisk-freeratearerelated toconsumptionvolatility.Theeconomicinterpretationisthatinvestorsrequireasystematic compensationfortheriskarisingfrommacroeconomicvolatility.LikeduringtheGreatmoderation, whenassetpricevolatilityhasbeenlow,ashavebeenexcessreturnsandriskpremia. Wearguethatthereareseveralissuesthatpreventaconsumption-basedmodelfromgeneratinga reliableinterpretationofassetpricemovements.Inparticular,wesuggestthattheinterpretationsuffers fromobservationalequivalence.Theexperienceoftheperiodleadingtotheturmoilhasbeen characterizedbyasystematicunderestimationofriskbecauseofanoveralllackofmarkettransparency. Thishascoexistedwithastablemacroeconomicoutlook. Thisnoteisorganizedasfollows.Section2reviewsthefoundationsofconsumption-basedasset pricing.Section3proposesourcriticallineofargument. 2. 2. 2.
2. FinancialriskpremiaintheconsFinancialriskpremiaintheconsFinancialriskpremiaintheconsFinancialriskpremiaintheconsumptionumptionumption----basedassetpricingmodelumptionbasedassetpricingmodelbasedassetpricingmodelbasedassetpricingmodel Therepresentativeconsumer/investormaximizesexpectedintertemporalutilityeveryperiodby splittinghisbudgetbetweencurrentconsumptionandinvestmentindifferentassetclassesforthe purposeoffutureconsumption.Thereisonerisk-freeasset,andanumberofriskyassetswithrandom
payoffXt+1,iinthenextperiodandcurrentpricePt,i.Thisstructurecaninprincipleapplytoallassets,
suchasbonds,stocks,houses,derivativesorportfolios.Forstocks,forinstance,Pt,iwoulddenotethe
currentex-dividendprice,andnextperiod’spayoffisgivenbythesumofthedividendDt+1,iandthe
then-prevailingex-dividendpricePt+1,i.The(gross)returnsoftheseassetsareRt+1,i=Xt+1,i/Pt,i
;therisk-freeassethasreturnRt,f(risk-freeinterestrate).
Foreachasseti,pricesaregivenasexpectationsoffuturepayoffs,discountedbytheinvestor’s
marginalrateofinter-temporalsubstitutioninconsumption,Mt+1=βU’(Ct+1)/U’(Ct),
whereU’ismarginalutility,βistheinvestor’stimediscountfactorandCdenotesconsumption.This followsasanoptimalityconditionfromtheconsumption/investmentproblem.Theinterpretationisthat theexpectedmarginalbenefitofdevotingonemarginalunitofresourcestoassetiequalsthecurrent marginalutilitylossofforegoingoneunitofcurrentconsumption Rewriting(1)providesanexpressionforriskpremia,i.e.expectedexcessreturnsofriskyassetsover therisk-freerate (2) ] [ ] , [ ] [ 1 , 1 1 , , 1 + + + + − =− t t i t t t f t i t t M E R M Cov R R E . Theimportantinsightofthisrelationisthatwhatmattersfortheriskpremiumofanassetisthe
covarianceofthatasset’sreturnRt+1,iwithconsumer’smarginalrateofsubstitutionMt+1:thehigher
thecovarianceofreturnswiththemarginalrateofsubstitution,thelowertheriskpremium.Assetsthat tendtogivehighreturnswhentheyarereallyappreciated–namelywhenmarginalutilityoffuture consumptionishigh,i.e.consumptiongrowthislow–areattractiveassets,forwhichinvestorsrequire onlyasmallriskpremium.2Incontrast,assetsthattendtopayoffpoorlywhenconsumptiongrowthis low,i.e.marginalutilityishigh(and–viceversa–thattendtopayoffwell,whenconsumptiongrowth ishigh),requirealargeexpectedexcessreturnsuchthatinvestorsarewillingtoholdthem. AsthemarginalrateofsubstitutionMactsasa‘pricingkernel’forallassets,itsownvolatilityrestricts thepossiblemagnitudeoffinancialriskpremia.SinceMisequaltotheconsumer’smarginalrateof substitutionandthusitisafunctionofconsumptiongrowth,thelinkbetweenconsumptionvolatility andfinancialriskpremiabecomesapparent.Toseethisnexusmostclearly,considerthe‘Sharperatio’, i.e.theexpectedexcessreturnperunitoftheassetreturn’sconditionalvolatilityσt[Rt+1,i] (3) ] [ ] [ ] [ ] [ 1 , 1 , , 1 , , 1 + + + + − = − t t i t t M i i t t f t i t t M E M R R R E
σ
ρ
σ
.Itdependsonthecorrelation
ρ
i,Moftheasset’sreturnwiththepricingkernelM,aswellasthefirstandsecondmomentofthelatter. Sincethecorrelationcoefficient
ρ
i,Mrangesbetween-1and1,(3)impliesthat (4) ] [ ] [ ] [ ] [ 1 , 1 , 1 , , 1 + + + + ≤ − t t i t t i t t f t i t t M E M R R R Eσ
σ
. Thelinktoconsumptionisdirectlyapparentfortheprominentexampleofthepowerutilityfunction U(Ct)=(1-γ)-1Ct1-γ,forwhich(4)becomesapproximately (5) [ log ] ] [ ] [ 1 , 1 , , 1 + + + ∆ ⋅ ≤ − t t i t t f t i t t C R R R Eσ
γ
σ
. 2Extremeexamplesareinsurancecontractsthatpayoffinverybadstatesoftheworld: they canbeviewedas assetsforwhichinvestorsarewillingtoacceptanegativeexpectedexcessreturn.
Equation(5)showsthatthemagnitudeoffinancialriskpremiaisapositivefunctionoftheamountof
macroeconomicriskgivenbythevolatilityofconsumptiongrowth
σ
t[∆logCt+1]–aproxyforthequantumofnon-diversifiablemacroeconomicrisk–andthewillingnessofconsumerstobearthatrisk givenbytherelativeriskaversionparameterγ–theindividualpriceattachedtothemacroeconomic risk.Therelationholdswithequalityforthoseassetsofwhichthereturnsareperfectlynegatively correlatedwiththepricingkernel.Thisholdsforso-calledefficientportfoliosthatgivethemaximum expectedreturnforagivenvariance. 1. 1. 1.
1. AcriticaloverviewAcriticaloverviewAcriticaloverviewAcriticaloverviewfromapractitioner’sperspectivefromapractitioner’sperspectivefromapractitioner’sperspectivefromapractitioner’sperspective
Howdoestheconsumption-basedassetpricingmodelhelptounderstandthebuild-upoftensionsthat ledtotherecentturmoilinfinancialmarkets?Themodelsuggeststhatapermanentreductionof consumptionvolatility–whichistypicallypositivelycorrelatedwithoutputvolatility–isassociatedwith lowerriskpremia.Inthesamefashion,riskpremiafallwhentherelativeriskaversionofarepresentative investorfalls.Alongtheselines,Lettau,LudvigsonandWachter(2008)arguethattherun-upinUS stockmarketvaluationsinthe1990swasdrivenbyareductionintheequityriskpremium,whichwasin turncausedbytheGreatModerationinconsumptiongrowthvolatilitysincethemid1980s.3 Applyingthislogictotheinterpretationofreal-worldeventsrequiresseveralqualifications. Consumption-basedpricingdeliversareturnthatisconsistentwithpricingthesourcesofriskarising fromchangesinmacroeconomicfundamentals.Wecanthinkoftheobservedassetpricesasthesumof thepriceconsistentwithfundamentsandaresidualterm.This‘residual’canmeasureafittingproblem. Itisan‘observation’error,consideringthatthefundamentalpriceisnotknownbythemarket.Itisalso an‘expectation’errorsincethemodeldeliversapredictionforfutureassetprices. Thisperspectiveraisestwoquestions.Thefirstissuehastodowiththereasonforfittingerrorstoarise. Iftheoutcomeofmarketactionsarisesfromanotionofequilibriumconsistentwiththeoneembedded inthemodel,thennosuchadiscrepancyshouldarise.Thesecondquestionisrelatedtohowlargeand howpersistentweshouldexpectthefittingorobservationerrorstobe.Theexperienceofthetransition fromnormaltoturbulentmarketconditionsbetweenFebruaryandAugust2007suggeststhatthese errorscanbesubstantial.Duringthisperiod,episodesofrepricingorriskincertainmarketsegments, suchasthegovernmentbondmarket,correspondedwithastablemacroeconomicoutlook. 3 Weshouldstressthatthisisonlyoneofthepossibleinterpretationsthatconsumption-basedassetpricingcan provideaboutchangesinriskpremia.Forinstance,movementsinthecoefficientofrelativeriskaversioncanplay arole.However,forthemerepurposeofourdiscussion,wefocusonconsumptionvolatility.
Thepointwearetryingtomakeisthattheinterpretationoftheprescriptionfromtheconsumption-basedpricingmodelinvolvesanumberofissueswhichmayleadtoconclusionsinconsistentwiththe veryassumptionsofthemodelitself. Thepracticalapproachtomodeldevelopmentintherisk-managementindustrycanprovideappealing insightsintotheinterpretationofmodellingoutcomes.Theprivateindustryfeaturesseveral characteristicsthataredesirablefromtheperspectiveofabalanceduseofmodels.Acompetitiveoffer ofdifferentmodellingapproachesispervasive.Moreover,differentlyfromtheuseofmodelsproduced byacademicsforthepurposeofacademicdebate,thereisnobuilt-inalignmentofinterestbetween modelproducersandmodelusersforriskmanagementintheprivatesector.Letusdrawananalogy withthecaseofgeneralequilibriummodelsforthesupportofmonetarypolicydecisionsatcentral banks.AstheNewKeynesianparadigmrepresentstheconsensusacademicview,modelsare producedbycentralbankeconomists,usuallywithscrutinyoradvicefromacademics,toprovideeither policyadviceorsupportingarguments.Itmaybearguedthattheprofit-maximizingnatureoftherisk managementindustryprovidescaseswheremodellingsolutionsareprovidedtothirdpartieswithan incentivetohighlightandexplainpossiblemodelshortcomings.Thisexperienceisrelevantforour discussionbecauseassetpricemodellingandforecastingiskeyfortheriskmanagementindustry. Theleft-handsideofequation(3)includesaSharperatio.Standardpracticeintherisk-management industryputsspecialcareingeneratingforecastsforthevolatilityoffuturereturnsthatarereliable.This processinvolvessettingupassumptionsaboutthedistributionofreturnsthatcanhelptocapturethe likelihoodofextremeeventsmaterializing.Inotherwords,itiswellclearintheindustrythatthis involvesforecastingthetailsofthedistributionsofreturns.Typicallyhistoricalinformationisusedto driveapredictionaboutfutureassetprices.Hence,theissueboilsdowntocapturingtheeventsofthe pastthatpresentfeaturesrelevantfortheforecastingexercise.Forinstance,inestimatingtheaverage relationbetweenconsumptiongrowthandtheSharperatio,howfarlongintothepastshouldwego? ShouldwegobacktotheextremeeventsoftheGreatDepression?Andwhatwastherelationbetween consumptiongrowthandtheSharperatioduringtheDepressionperiod?Overlookingtheseissuescan resultinafailuretochoosetheparametersofamodelconsistentlywiththeneedtogenerate conservativeforecasts.Asaresult,theseissuesaretypicallydiscussedatlengthintherisk-management industry. Therearedeeperreasonstomistrustthestandardacademicinterpretationoftheprescriptionfrom consumption-basedpricing.Thetransitiontothecurrentturmoilhasshownthatlowexcessreturnscan beduetoanunderestimationofthecausesofmovementsinassetprices.Inthejargonusedearlier,this translatesintoalargefittingerror.Partoftheriskmanagementindustryalsoreferstothisobservationas arisingfrommodelrisk.Inotherwords,investorshaveusedthewrongkindofmodelforpricing withouttakingintoaccounttheprobabilityofthisissuematerializing.
Theseconsiderationssuggestthattherecanberationalesforthemispricingofrisktobesystematicand persistent.Infact,itcanbearguedthatasoundpricingofriskrequiresfinancialmarketparticipantsto understandthestructureoftradingandthenatureoftherisksthatcanarise.Indeedinvestorsmaybe unableorfailtounderstandthesefactorsforseveralreasons.Marketsmayfunctioninanon-transparentway.ThispointwasalsostressedbyJean-ClaudeTrichetinaspeechdeliveredatthe InternationalSwapsandDerivativesAssociationinApril2007: “Here,itshouldbeacknowledgedthattheopacityofthecreditderivativesmarket,andespeciallyof structuredsyntheticinstruments,isapotentialsourceofconcern.Thecomplexinteractionbetween cashinstrumentsandcreditderivativeshasmadeitincreasinglydifficulttomonitorwheredifferent, possiblysizeable,positionsaretakenandwhererisksareconcentrated.” Inthissense,knowledgeofthemicrostructurefactorscancontributetoabetterunderstandingofthe determinantsofrisk.AsarguedbyDeGrauwe(2008)inanapplicationoftheideaof‘animalspirits’, wavesofoptimismcancontributetothepatternofboom-bustcycleexperiencedduringthemarket crashesoftheparttwentyyears. Summingup,wearguethatthewell-acceptedacademicinterpretationofthemacroeconomiccausesof riskunderestimationduringthebuild-upoftheturmoilisinsightfulabouttheshortcomingsofthe macro-basedapproachtoassetpricing.Thisnoteprovidesadiscussionofhowapractitioner’spointof viewmayhelpuncovertheissuesofanacademicapproach.Inparticular,wearguethatthe interpretationofassetpricemovementsduringtheGreatModerationprovidedbyconsumption-based modelssuffersfromobservationalequivalence. References References References References Bernanke,B.(2004),TheGreatModeration,remarksatthemeetingsofEasternEconomicsAssociation, Washington,D.C.,20February. DeGrauwe,P.(2008),AnimalSpiritsandMonetaryPolicy,CESifoWorkingPaper,2418,October. Lettau,M.,Ludvigson,S.C.,andWachter,J.A.(2008),Thedecliningequitypremium:Whatroledoes macroeconomicriskplay?,ReviewofFinancialStudies,21(4):1653-1687, Trichet,J.C.(2007),SomeReflectionsontheDevelopmentofCreditDerivatives,KeynoteAddress, 22ndAnnualGeneralMeetingoftheInternationalSwapsandDerivativesAssociation,18April.