pooling and individual effects estimators for nonstationary panel data

《pooling and individual effects estimators for nonstationary panel data》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、PoolingandIndividualEffectsEstimatorsforNonstationaryPanelData(1)ThepresentpaperisfinanciallysupportedbyMURST60%andCNRfunds.StimatoriPoolingeadEffettiIndividualiperDatiPanelNonStazionariAntoninoDiPinoDipartimentodiEconomia,StatisticaeAnalisidelTerritorio,Univers

2、itàdiMessina,e-mail:dipino@imeuniv.unime.itRiassunto:VienepropostounmetodopercorreggeredaglieffettidellanonstazionarietàlestimeOLSperdatipooled.DallascomposizionedelterminedierroredellastimaOLSsiricavanomomenticampionarimistitemporaliecross-sectiondicui,conl’aiu

3、todialcunesimulazioninonriportateneltesto,èstatariscontratalaconvergenzaaquantitànonnulle.Vengonoproposti,infine,alcunirisultatiasintotici.Keywords:Dynamicpanels,Pooledtimeseries.1.IntroductionRecentcontributionstothedynamicpanelregressioninthepresenceofI(1)pool

4、edseries(PesaranandSmith,1995;Pesaran,ShinandSmith,1999)showthatthetraditionalproceduresforestimatingpooledmodels,suchasindividualeffectsandgeneralizedmethodofmoments(GMM)estimators,canproduceinconsistentestimates.PhillipsandMoon(1999)developalimittheoryforpoole

5、destimationswithnonstationarydata,andshowthatwhensimilarcointegratingrelationsintheindividualsoccurthetraditionalpoolingOLSandtheindividualeffectsOLSestimatorsproducebiasedestimates.TheaimofthisworkistostudythebiaseffectsofpooledandindividualeffectsOLSestimators

6、.Threeendogeneityeffectsarediscussed:i)anovertimeendogeneityeffect“inside”thesinglepanel,ii)aglobal“overtime”effectoftheaveragedindividualdata,andiii)an“acrossindividual”effect.Anendogeneitycorrectionprocedureisexploredandsomeasymptoticresultsareproposed.Theplan

7、ofthepaperisasfollows.ThenextsectiondescribesthepoolingandindividualeffectsOLSasymptoticpropertiesfornear-cointegratingnonstationarypaneldata.ThethirdsectionshowsthemainresultsofanendogeneitycorrectionprocedureforpoolingandindividualeffectsOLSestimators.2.D.G.P.

8、forpooleddata.Letusconsiderthefollowingregressionmodelforpooleddata:Yit=a’Xit+uit(uit=f1iuit-1+eit)(1)WhereisYitanT´1vector,

9、f1

10、<1.eitisazero-meanIIDprocess.(1;2;…;i;