Prediction of US Export to China.docx
《Prediction of US Export to China.docx》由会员分享,可在线阅读,更多相关《Prediction of US Export to China.docx(11页珍藏版)》请在冰豆网上搜索。
PredictionofUSExporttoChina
西南财经大学
SouthwesternUniversityofFinanceandEconomics
商科研究方法II
论文题目:
PredictionofU.SExporttoChina
学生
姓名:
郭宏宇
所在学院:
国际商学院
专业:
国际经济与贸易
学号:
41110024
2014年6月
PredictionofU.SExporttoChina
GuoHong-yu
(TheSouthwestUniversityofFinanceandEconomics,Internationaleconomicsandtrade)
Abstract:
ChinaandAmericaisthelargesttradingpartner.SinceChina’saccessiontoWTO,thevolumeoftradebetweenChinaandAmericahaverisensharply.However,therearestillsometradebarriersandfrictionbetweentwocountries.TheresearchbasedontradebetweenAmericaandChinaisimportanttothetwocountrieseventheworldeconomy.IselecttheU.SexporttoChinaasdependentvariableandusedistributedlagmodeltoforecasttheU.Sexport.
Keywords:
U.SexporttoChina;Q-test;
1.Introduction
1.1Motivation(BackgroundandSignificanceofTopics)
AmericaandChinaarethefirsttwolargesteconomicentitiesintheworld.Atthesametime,chinaandAmericaisalsothelargesttradingpartner.Forecastingtheexportandimportvolumeininternationaltradeistheprerequisiteofagovernment’spolicy-makingandguidanceforahealthierinternationaltradedevelopment.SinceChina’saccessiontoWTO,thevolumeoftradebetweenChinaandAmericahaverisensharply.However,therearestillsometradebarriersandfrictionbetweentwocountries.TheresearchbasedontradebetweenAmericaandChinaisimportanttothetwocountrieseventheworldeconomy.Here,IselecttheU.SexporttoChinaasdependentvariableandusedistributedlagmodeltoforecasttheU.Sexport.
1.2Literaturereview
MostofliteraturessetupARMAmodelstopredictfutureexports.AndmanyprofessorspointoutthatARMAmodelisusefulinshort-runforecastratherthanlong-runwhichmayhavefiercevolatility.RonaldL.Coccariinvestigatedtheresultsofafewdifferentmodelsin“AlternativeModelsforForecastingU.S.Exports”(RonaldL.Coccari,1998).HepointedoutthatOneobviousconclusionfromthisanalysisisthatforecastevaluationshouldnotberegardedasaprocedureforacceptingoneforecastingmodeltotheexclusionofothers.XiaoZ&GongKinvestigatedacombinedforecastingapproachbasedonfuzzysoftsets(XiaoZ&GongK,2007),whichpointedoutthatacombinedforecastingapproachbasedonfuzzysoftsetsisapromisingforecastingapproach.DiamantopoulosAandWinklhoferH.researchedthetechniqueutilizationanditsimpactonforecastaccuracy,thepaperpointedthat“inordertoimproveforecastaccuracy,attentionneedstobefocusedbeyondthequestionoftechniqueselection”.(DiamantopoulosA&WinklhoferH.,2003).Eachalternativemethodusuallycontainsavaluablepieceofinformationthatmaybeusedbycombiningallavailableforecastsintoacomposite.OnemightsimplywanttousethetrendandseasonalmodelsincetherewasfoundtobesuchastrongtrendcomponentinU.S.exports.Italsorevealstheobviousdependencyofexportsuponpreviouslaggedobservations(especiallyathreequarterlag),andthusasimpleautoregressiveschemeseemsappropriate.
2.Data
2.1Datasource
IfindmonthlydataofexportfromtheUStoChinaandforeignexchangeratebetweentheUSandChinainthewebsiteFederalReserveEconomicData(FRED).TheyareseasonallyadjusteddatafromJanuary1985toAugust2013.ButChinaisofficiallynotaWTO(WorldTradeOrganization)memberuntilDecember2001.SoIdropthedatawhichispreviousthanJanuary2002andfocusonthelaterones.Andthedataisseasonallyadjusted.
2.2Datadescription
Iplotthepointsofexportandexchangerateandgetthefollowinggraph.Andwecanseethattherearetrendbothintheexportandexchangerate.
ThenImakeanAugmentedDickey-Fuller(ADF)testtoseeifthereisunitrootinthedataoritisstationary.First,Itestthedataofexport.
Theabovestataresultsshowthatexportisstationary,butthereisunitrootinexchangerateasshownbelow.
Soitneedtomakedifferenceofexchangeratetoget
Zt=ert-ert-1
Andthentestifit’sstationary.Theresultisasbelowandit’sstationary.
3.Model
3.1Modelselection
Exportisaffectedbychangeinexchangerate.Asintheeconomictheory,whenacountry’scurrencyappreciates(risesinvaluerelativetoothercurrencies),thecountry’sgoodsabroadbecomemoreexpensiveandforeigngoodsinthatcountrybecomecheaper(holdingdomesticpricesconstantinthetwocountries).Sothecountryexportsless.From2002to2005,thedecreaseofexchangeratebetweentheUSdollarandothercurrencieshelpedUSindustriesexportmoreandsellmoregoods.Theimpactofexchangerateonexportshouldbeaccountedfor.
Forexport,firstlyItakethelag,itisstillstationary.ThenIremovethelineartrendfromlogexportandgettheresidual.Nextisidentifycycleintheresidual.AutocorrelationgraphhasatrailPartialautocorrelationgraphhasatruncation.SoitisanARmodel.
Andit’sknownthatpreviousvolumeofexporthaveimpactonlattervolumeofexport.SoIcombinetheAR(p)modelofexportanddistributedlagmodelofexportondifferenceofexchangerate.Andthenumberofdistributedlagsisq.Thespecificnumberofpandqshouldbeidentifiedtofindthetruemodel.
3.2Modelspecification
ThentheauthorusetheinformationcriteriaAICandBICtogetthespecificnumberofpintheARmodel.Finally,theresultshowedthatAICandBICisthesmallestwhenpequals13
Asforthedifferenceofexchangerate(z),thetrendofexchangeratehasbeenremovedwhenmakethedifference.SotheauthorjustmaketheautoregressiveofZonitslagsandthenuseAICandBICtofindthenumberoflags.Finally,Igetqequals1thatiswiththesmallestAICandBIC.Zt=α+βZt-1.Thisisthegraphofdifferenceofexchangerate(z).
Throughthepreviousprocess,thetruemodelIfoundisbelow:
lnexportt=α0+α1t+γZt-1+β1lnexportt-1+β2lnexportt-2+β3lnexportt-3+β4lnexportt-4+β5lnexportt-5+β6lnexportt-6+β7nexportt-7+β8lnexportt-8+β9lnexportt-9+β10lnexportt-10+β11lnexportt-11+β12lnexportt-12+β13lnexportt-13
Thegraphaboveistheregressionresultusingclassicalstandarderror.TheR2isprettyhigh,whichiscloseto1.Asfortheeffectofexchangerateonexport,ImakeaGrangerCausalitytestandderivethefollowingresult.P-valueissmall(p=0.0210),sowecanrejecthypothesisofnon-causalityandindicatethatdifferenceofexchangeratedoespredictivelycauseln(export)andhelptopredictit.
3.3Modeltest
First,theauthortestsiftheerrortermofthewholemodeliswhitenoise.Therearetwomethods.ThefirstoneisQ-test.
Theresult0.7205showsitiswhitenoise.
Forsecondmethod,theauthorderivestheautocorrelationofresidualsfromthestata.Andwecanmakeaconclusionthattheerrortermiswhitenoise.efromthegraph.
Second,theauthorusestheWhiteTesttoseeifthereisheterokedasticityinthemodel.Thenderivethefollowingresult.TheP-valueis0.4583,sowecannotrejectthehypothesisofhomoskedasticity.
4.Resultsandforecasts
4.1Evaluatingtheresults
Aftersettingupthemodel,wecanuseittogeneratethefittedvaluesandcomparetheresultwiththeactualvaluestomakesurethatthemodelreallyfitthehistoricaldatawell.Basically,theforecastmodelisagoodoneknownfromthefollowinggraph.
4.2Makingforecast
Theauthorusesthemodelderivedtoestimateone-step-aheadtotwelve-step-aheadpointandintervalforecast.
one-step-aheadforecast
Yt+1=α0+α1t+γZt+β1yt+.....β13yt-12
fromtwototwelvestep-aheadforecast,theforecasterrorisnotWN,itisMA(h-1).sinceitiscorrelated,theauthoruseanewwaytogetbetterresults.
Yt+2=α0+α1t+γ1Zt+γ2Zt-1+β1yt+.....β13yt-12
Yt+12=α0+α1t+γ1Zt+....γ12Zt-12+β1yt+.....β13yt-12
Theresultsareasfollows.
5.Conclusion
Thispaperanalyzetheexportdatafrom2002m1andestablishAR(13)modelwithtrendanddistributedlagsofexchangerate.Thensuccessfullypredicttheexportinthefutureyear.
Accordingtoourprediction,theexportincreasesteadilyinthenextyear.Thepredictionconformstotherealityforthefollowingreasons.
Forthefirstplace,afterthefinancialcrisisin2008,theglobaleconomystepintoanewbusinesscycle.Incurrentstage,theeconomyofU.S.isrecoveringanddevelopingatafastspeed.Thegoodeconomicconditionwilldefinitelypromoteexportasawhole.
Secondly,withthetrendofglobalization,Chinaismakingmorepoliciestofacilitateinternationaltrade.IthaslaunchedShanghaiPilotFree-TradeZoneonSeptember29,2013.What’smore,itwillcontinuetoestablishTianjinPilotFree-TradeZoneduringnextyear.Theseopenreformationsdefinitelylowerthebarriersoftrade.U.ScanexporttoChinamoreeasily.AndtheconsumptiondemandinChinawillbecomelargerandlarger.
Thirdly,thelastingdepreciationofdollarstoyuanmakeU.SgoodscheaperinChina.Thistrendwillcontinue,forthesimplereasonthatChina,asanew-borninternationaleconomyafterattendingWTO,isplayinganincreasinglyimportantrolearoundtheworld.Thedemandforcurrencyyuanissurgingrapidly.
Finally,themaingoodsthatU.SexportstoChinaarehigh-techproductssuchaselectronicproductsandfinancialservices.Nowadays,thesekindsofindustriesarefacingunprecedentedopportunitiestoboom.EspeciallyinU.S,whereitisabundantinhigh-techhumancapital,theseindustriescangrowfasterandcontributealargeportiontoexport.
REFERENCE
ClementsMP,SmithJ.Evaluatingtheforecastdensitiesoflinearandnon‐linearmodels:
applicationstooutputgrowthandunemployment[J].JournalofForecasting,2000,19(4):
255-276.
CoccariRL.Alternative