AR -人与机器：机器分析师与传统研究分析师投资建议的比较-英-25页

THEACCOUNTINGREVIEWAmericanAccountingAssociationVol.97,No.5DOI:10.2308/TAR-2020-0096September2022pp.221–244HumanVersusMachine:AComparisonofRobo-AnalystandTraditionalResearchAnalystInvestmentRecommendationsBraidenColemanUniversityofGeorgiaKennethMerkleyIndianaUniversityJosephPacelliHarvardUniversityABSTRACT:Weprovidethefirstcomprehensiveanalysisofthepropertiesofinvestmentrecommendationsgeneratedby‘‘Robo-Analysts,’’whicharehumananalyst-assistedcomputerprogramsconductingautomatedresearchanalysis.OurresultsindicatethatRobo-Analystrecommendationsdifferfromthoseproducedbytraditional‘‘human’’researchanalystsacrossseveralimportantdimensions.First,Robo-Analystsproduceamorebalanceddistributionofbuy,hold,andsellrecommendationsthandohumananalystsandarelesslikelytorecommend‘‘glamour’’stocksandfirmswithprospectiveinvestmentbankingbusiness.Second,automationallowsRobo-Analyststorevisetheirrecommendationsmorefrequentlythanhumananalystsandincorporateinformationfromcomplexperiodicfilings.Third,whileRobo-Analysts’recommendationsexhibitweakshort-windowreturnreactions,theyhavelong-terminvestmentvalue.Specifically,portfoliosformedbasedonthebuyrecommendationsofRobo-Analystssignificantlyoutperformthoseofhumananalysts.Overall,ourresultssuggestthatautomationinthesell-sideresearchindustrycanbenefitinvestors.JELClassificati...