EmpiricalevidenceontheCAPMCAPMassumption:Investorscanbuyandsellallsecuritiesatcompetitivemarketpriceswithoutincurringtaxesortransactionscostsandcanborowandlendattherisk-freeinterestrateInvestorsholdonlyefficientportfoliooftradedsecurities—portfoliothatyieldthemaximumexpectedreturnforagivenlevelofvolatility.Investorshavehomogeneousexpectationsregardingthevolatilities,correlationsandexpectedreturnsofsecurities.LimitaionsofCAPMBetasusedbyCAPMtomeasuresystematicriskarenotobserved:CAPMisaone-periodmodelthattreatsasecurity’sbetaasaconstant,butresearchersfindbetacanbechangedinrespondtofirms’investmentinnewindustry,changeincapitalstructureandsoon.Ifbetaschangeovertime,simplehistoricalestimatesofbetaarenotlikelytobeaccurate.Mismeasuringofbetaswillnotreflectstocks’systematicrisk,sotheCAPMdoesnotcomputetheriskpremiumcorrectly.PredictionabouttheexpectedreturnonanassetbyCAPMisinherentlyunobservable:Thereisnodatabasethatmeasuresthereturnsthatinvestorsusetheassumptionofrationalexpectations,whichmeansthatevenifinvestorsmakemistakeswhentheyformassessmentsaboutexpectedreturns,theirerrorsarenotsystematic.Onlyifassumethatinvestorshaverationalexpectations,researcherscanuserealizedhistoricalreturnsasaproxyforexpectedreturns.Historically,researchershavefoundthatexpectedreturnswererelatedtobetasaspredictedbyCAPMratherthantoothermeasuresofrisksuchasthesecurity’svolatility.ThestudyconductedbyBlack,JensenandScholes(1972)foundapositiverelationshipbetweenbetaandreturnsasCAPMpredicted.However,thedifferenceinreturnsbetweenhigh-betaandlow-betastockswasnotaslargeastheyexpected.Inotherwords,theSMLwastooflat.EugeneFamaandKenFrenchevenarguethereareotherfactorsratherthanjustbetarelatedtothereturnsuchasfirmsizeandbook-to-marketvalueratio.Accordingtotheirtheory,theSMLwasnotjusttooflat,itwascompletelyflat.Problemrelatedtomarketportfolio:formoststocksthestandarerrorsofthealphaestimatesarelarge,sothealphasareobviouslydifferentfromzero.Thenon-zeroalphacontaisriskthattheCAMPdoesnotcapture,whichindi