VaR值计算-参数、半参、非参.doc

《VaR值计算-参数、半参、非参.doc》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。

1、CalculateVaRusingthreemethods:Parametric,NonparametricandSemiparametricAbstractValueatrisk(VAR)isastatisticalmeasurementoftotalportfoliorisk,takenastheworstlossataspecificconfidenceleveloverthehorizon.VaRisthedollarorpercentagelossinportfolio(asset)valuethatwillbeequaledorexceededonlyXpercentofthe

2、time.Inotherwords,thereisanXpercentprobabilitythatthelossinportfoliovaluewillbeequaltoorgreaterthantheVaRmeasure.TocalculatetheVaR,therearethreemethods:parametric,nonparametricandsemiparametric.CalculatingparametricVaRisasimplematterbutrequiresassumingthatassetreturnsconformtoastandardnormaldistri

3、bution.WeusuallyusehistoricalsimulationmethodasnonparametricmethodtoestimateVaR.Thismethodjustusehistoricaldata,andwedon’tneedtoknowthedistributionoftheassetreturn.Thedistributionisinthedata.Semiparametricmethodisahybridmethod.ItusebothparametricandnonparametricmethodstoestimateVaR.Afterusingnonpa

4、rametricestimatingVaR,weassumethattheassetreturndensityhasapolynomialefttail,andthetailindexisa.thesemiparametricmodelcombingparametricandnonparametriccomponents.Threemethodshavetheiradvantagesanddisadvantages,wecalculateVaRusingthesemethodsandcontrasttheresultswithoutbacktesting.Butweuset-testtof

5、indifthereexistasignificantdifferencebetweenthethree.Andwefoundthatthesethreeshavenosignificantdifference.Keywords:VaR,parametric,nonparametric,semiparametric1.IntroductionInthisdecadeValueatRisk(VaR)hasbecameaverypopularmeasureofmarketrisk.VaRisthelossontheportfoliothatwillnotbeexceededwithaspeci

6、fiedprobabilityoveraspecifiedtimehorizon.VaRisanextremelypowerfulriskmeasure,becauseitcanbecalculatedassuminganykindofdistributionsofportfolioreturns.VaRwasdevelopedasanefficient,inexpensivemethodtodetermineeconomicriskexposureofbankswithcomplexdiversifiedassetsholding.ThemethodtoestimateVaRareoft

7、enasfollowing:parametric,nonprarmetric,semiparametric.ParametricsimulationforestimatingVaRrequirestheassumptionofanormaldistribution.Thisisbecausethemethodutilizestheexpectedreturnandstandarddeviationreturns.Nonp