robust estimation of style analysis coefficients

《robust estimation of style analysis coefficients》由会员上传分享，免费在线阅读，更多相关内容在工程资料-天天文库。

1、RobustestimationofstyleanalysiscoefficientsMicheleLaRoccaandDomenicoVistoccoAbstract.Styleanalysis,asoriginallyproposedbySharpe,isanassetclassfactormodelaimedatobtaininginformationontheinternalallocationofafinancialportfolioandatcomparingportfolioswithsimilarinvestmentst

2、rategies.Theclassicalapproachisbasedonaconstrainedlinearregressionmodelandthecoefficientsareusuallyestimatedexploitingaleastsquaresprocedure.Thissolutionclearlysuffersfromthepresenceofoutlyingobservations.Theaimofthepaperistoinvestigatetheuseofarobustestimatorforstyleco

3、efficientsbasedonconstrainedquantileregression.TheperformanceofthenovelprocedureisevaluatedbymeansofaMonteCarlostudywheredifferentsetsofoutliers(bothintheconstituentreturnsandintheportfolioreturns)havebeenconsidered.Keywords:styleanalysis,quantileregression,subsampling1

4、IntroductionStyleanalysis,aswidelydescribedbyHorstetal.[12],isapopularandimportanttoolinportfoliomanagement.Firstly,itcanbeusedtoestimatetherelevantfactorexposureofafinancialportfolio.Secondly,itcanbeavaluabletoolinperformancemeasurementsincethestyleportfoliocanbeusedas

5、abenchmarkinevaluatingtheportfolioperformance.Finally,itcanbeusedtogainhighlyaccuratefutureportfo-lioreturnpredictionssinceitiswellknownfromempiricalstudies[12]thatfactorexposuresseemtobemorerelevantthanactualportfolioholdings.Themethod,originallyproposedbySharpe[25],i

6、sareturn-basedanalysisaimedatdecomposingportfolioperformancewithrespecttothecontributionofdifferentconstituentscomposingtheportfolio.Eachsectorisrepresentedbyanindexwhosereturnsareavailable.Themodelregressesportfolioreturnsonconstituentreturnsinordertodecomposetheportf

7、olioperformancewithrespecttoeachconstituent.Indeed,intheframeworkofclassicalregression,theestimatedcoefficientsmeanthesensitivityofportfolioexpectedreturnstoconstituentreturns.Theclassicalapproachisbasedonalinearregressionmodel,estimatedbyusingleastsquares,butdifferentc

8、onstraintscanbeimposedonthecoefficients.M.Corazzaetal.(eds.),MathematicalandStatisticalMethodsforActuarialSciencesandF