The-Barra-US-Equity-Model-(USE4)

msci.comModelInsightTheBarraUSEquityModel(USE4)MethodologyNotesJoseMencheroD.J.OrrJunWangAugust2011MSCIResearchmsci.com©2011MSCIInc.Allrightsreserved.PleaserefertothedisclaimerattheendofthisdocumentRVMay2011ModelInsightUSE4MethodologyAugust20112of44Contents1.Introduction............................................................................31.1.ModelHighlights..............................................................................................31.2.ModernPortfolioTheoryandBarraRiskModels:ABriefHistory.....31.3ForecastingPortfolioRiskwithFactorModels.........................................62.FactorExposures.................................................................72.1GeneralConsiderations..................................................................................72.2.DataQualityandOutlierTreatment...........................................................82.3.StyleExposures...............................................................................................92.4.IndustryFactors...............................................................................................92.5.Multiple-IndustryExposures......................................................................103.FactorReturns....................................................................133.1.CountryFactor..............................................................................................133.2.RelationtoTraditionalApproach.............................................................154.FactorCovarianceMatrix.................................................174.1EstablishedMethods....................................................................................174.2EigenfactorRiskAdjustment......................................................................194.3VolatilityRegimeAdjustment.....................................................................245.SpecificRisk........................................................................285.1EstablishedMethods....................................................................................285.2BayesianShrinkage.....................................................................................295.3VolatilityRegimeAdjustment.....................................................................316.Conclusion...........................................................................35AppendixA:ReviewofBiasStatistics...............................36A1.Single-WindowBiasStatistics...................................................................36A2.Rolling-WindowBiasStatistics..................................................................37AppendixB.EigenfactorRiskAdjustment.......................40REFERENCES.......................................................................43MSCIResearchmsci.com©2011MSCIInc.Allrightsreserved.PleaserefertothedisclaimerattheendofthisdocumentRVMay2011ModelInsightUSE4MethodologyAugust20113of441.Introduction1.1.ModelHighlightsThisdocumentdescribesthenewmethodologiesthatunderpintheUSE4model.Ouraimistoproduceadocumentthatisclearandconcise,yetcomprehensiveaswell.MSCIpridesitselfnotonlyonsettingthestandardforexcellenceinfactorriskmodeling,butalsoonbeingtheindustryleaderinmodeltransparency.Thisdocumentisthecomplementtoacompaniondocument:USE4EmpiricalNotes.Whereasthecurrentdocumentfocusesonmethodology,theEmpiricalNotescontaindetailedinformationaboutUSE4factorstructure,extensiveanalysisontheexplanatorypowerandstatisticalsignificanceofthefactors,andasystematicinvestigationintotheforecastingaccuracyofthemodel.TheEmpiricalNotesalsoprovideathoroughcomparisonwiththeUSE3model.ThemainadvancesofUSE4are:AninnovativeEigenfactorRiskAdjustmentthatimprovesriskforecastsforoptimizedportfoliosbyreducingtheeffectsofsamplingerroronthefactorcovariancematrixAVolatilityRegimeAdjustmentdesignedtocalibratefactorvolatilitiesandspecificriskforecaststocurrentmarketlevelsTheintroductionofacountryfactortoseparatethepureindustryeffectfromtheoverallmarketandprovidetimeliercorrelationforecastsAnewspecificriskmodelbasedondailyasset-levelspecificreturnsABayesianadjustmenttechniquetoreducespecificriskbiasesduetosamplingerrorAuniformresponsivenessforfactorandspecificcomponents,providinggreaterstabilityinsourcesofportfolioriskAsetofmultipleindustryexposuresbasedonGICS®Anindependentvalidationofproductioncodethroughadouble-blinddevelopmentprocesstoassureconsistencyandfidelitybetweenresearchcodeandproductioncodeAdailyupdateforallcomponentsofthemodelTheUSE4modelisofferedinshort-term(USE4S)andlong-term(USE4L)versions.Bothversionshaveidenticalfactorexposuresandfactorreturns,butdifferintheirfactorcovariancematricesandspecificriskforecasts.TheUSE4Smodelisdesignedtobemoreresponsiveandprovidethemostaccurateforecastsatamonthlypredictionhorizon.TheUSE4Lmodelisdesignedforlonger-terminvestorswhoarewillingtotradesomedegreeofaccuracyforgreaterstabilityinriskforecasts.1.2.ModernPortfolioTheoryandBarraRiskModels:ABriefHistoryThepioneeringworkofMarkowitz(195