Parametric Estimation of Jump-Diffusions（519-530）

《Parametric Estimation of Jump-Diffusions（519-530）》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。

1、FinancialEconometrics24ParametricEstimationofJump-Diffusions*Keywords:maximumlikelihoodestimation,jump-diffusions,MonteCarlosimulation,termstructureofinterestratesWeconsidertheproblemofestimatingthecoefficientsofdiffusionprocesseswithjumps.Theseprocesseshavebeenwidelyemployedfordescribingsudden

2、changesbothinstockpriceandininterestratedynamicsandreproducinghigh-ordermo-mentsofsampledata.Inmostcases,thelackofanalyticalexpressionsfortransitiondensitiesmakestheuseofmaximumlikelihoodmethodsaverychallengingtask.Thiscasedescribesandteststhesimulatedmaximumlikelihood(SML)methodintroducedbyPe

3、dersen(1995)andBrandtandSantaClara(2002).Thistechniqueallowsfortheestimationofawidevarietyofdiffusionprocesses,includingthosewhichlackclosed-formexpressionsforthetransitiondensity.WeillustratetheSMLmethodbydevelopingalgorithmsfortheestimationofbothcontinuousandmixed-jump-diffusionprocesses.The

4、latterprocesshasbeenemployedininterestratemod-ellingbyPiazzesi(2001).ThemainideaunderlyingSMListonumericallyevaluatethetransitionproba-bilitiesoftheprocesscorrespondingtoallpairsofvaluestakenbythestatevariableatconsecutivetimes.Ifadiscretizationofthetime–spaceaxesisproperlyrefined,theresultingt

5、ransitiondensityapproachesaGaussiandistribution.Thelikelihoodestimatorbecomesareliableapproximationoftheexactmaximumlikelihoodesti-mator,namelytheonestemmingfromtheexact,yetunknown,transitiondensityoftheprocess.Sections24.1and24.2introducetheestimationproblemandillustratethegen-eralmethodology

6、,respectively.Section24.3detailsthealgorithm,whileSect.24.4providestwoapplications.First,weconsidertheCox,IngersollandRoss(1985)model(CIR)forshort-terminterestratedynamics.Sincethetransitiondensityfor∗withPiergiacomoSabino.52024ParametricEstimationofJump-Diffusionsthismodelisknowninclosed-form

7、,parameterscanbeestimatedusingtheexactmaximumlikelihoodmethod.Consequently,wecanusethismodelasabenchmarkagainstwhichweevaluatetherelativeperformanceofSML.Second,weadaptSMLtojump-diffusionprocessesandimplementtheresultingalgorithmtothege