Momentum 交易策略

2020/3/151Momentum交易策略2020/3/152ReturnstoBuyingWinnersandSellingLosers：ImplicationsforStockMarketEfficiency《JournalofFinance》,1993,65-126.2020/3/153文章结构IntroductionTradingStrategiesTheReturnsofRelativeStrengthPortfoliosSourcesofRelativeStrengthProfitsProfitabilityofRSSWithinSize-andBeta-basedSubsamples2020/3/154文章结构（续）SubperiodAnalysisPerformanceofRelativeStrengthPortfoliosinEventTimeBack-TestingtheStrategyStockReturnsAroundEarningsAnnouncementDatesConclusions2020/3/155Introduction:ContrarianStrategies(1)Apopularview:individualstendtooverreacttoinformation.Adirectextension:DeBondtandThaler(1985、1987)Stockpricesalsooverreacttoinformation：contrarianstrategies(buyingpastlosersandsellingpastwinners)achieveabnormalreturns.(反向投资策略)Debate:unclearwhethertheirresultscanbeattributedtooverreaction.1.Canbeexplainedbythesystematicriskoftheircontrarianportfoliosandthesizeeffect.2.Thelong-termlosersoutperformandthelong-termwinnersonlyinJanuaries.2020/3/156Introduction:ContrarianStrategies(2)Morerecentpapers:Jegadeesh(1990)andLehmann(1990)shorter-termreturnreversals(aweekoramonthbased).However,theirapparentsuccessmayreflect1.thepresenceofshort-termpricepressure,or2.alackofliquidityinthemarket,ratherthan3.overreaction.(JegadeeshandTitman(1991))Inaddition,alargepartoftheabnormalreturnsisattributabletoadelayedstockpricereactiontocommonfactorsratherthanoverreaction.(LoandMacKinlay(1990))2020/3/157Introduction:RelativeStrengthStrategies(1)Earlyliteratureonmarketefficiencyfocusedon:RelativeStrengthStrategies(RSS)tobuypastwinnersandsellpastlosersMostnotably,Levy(1967):atradingrulethatbuysstockswithcurrentpricesthatsubstantiallyhigherthantheiraveragepricesoverthepast27weeksrealizessignificantabnormalreturns.JensenandBennington(1970),however,attributeLevy’sresulttoaselectionbias.(outsideLevy’soriginalsampleperiod)2020/3/158Introduction:RelativeStrengthStrategies(2)Althoughthecurrentacademicdebatehasfocusedoncontrarianratherthanrelativestrengthtradingrules,anumberofpractitionersstilluserelativestrengthasoneoftheirstockselectioncriteria.GrinblattandTitmann(1989,1991):amajorityofmutualfundsshowatendencytobuystocksthathaveincreasedinpriceoverthepreviousquarter.Inaddition,ValueLinerankingsareknowntobasedinlargepartonpastrelativestrength.CopelandandMayers(1982)andStickel(1985)2020/3/159Introduction:CSvs.RSS(Reconciliation)Onepossibility:theabnormalreturnsrealizedbythesepractitionersareeitherspuriousorareunrelatedtotheirtendenciestobuypastwinners.Asecondpossibility:thedifferencebetweentimehorizonsusedintradingrulesexaminedintherecentacademicpapersandthoseusedinpractice.CS:eitherveryshort-term(1weekor1month)returnreversals,orverylong-term(3to5years)returnreversals.RSS:overthepast3to12months.2020/3/1510Introduction:aboutthispaper(1)AnalysisofRSSover3-to12-monthhorizonsNYSEandAMEXstocksSampleperiod:1965-1989Result:theanalysisdocumentssignificantprofitsforeachoftherelativestrengthstrategiesexamined.DecompositionandrelativeimportanceTests:theprofits:1.Notduetothesystematicriskofthetradingstrategies.2.Cannotbeattributedtoalead-lageffectresultingfromdelayedstockpricereactionstoinformationaboutacommonfactor.3.Consistentwithdelayedpricereactionstofirm-specificinformation.2020/3/1511Introduction:aboutthispaper(2)Furthertestssuggestthatpartofthepredictablepricechangesthatoccurduringthese3-to12-monthholdingperiodsmaynotbepermanent.1.negativeabnormalreturns:12months,uptothe31month2.9.5%overthenext12months,butlosesmorethanhalfinthefollowing24monthsAnalysisofstockreturnsaroundearningsannouncementdatessuggestsasimilarbiasinthemarketexpectations.Structureofthispaper.2020/3/1512Ⅰ.TradingStrategies:RSS(1)Formationperiod:1,2,3,or4quarters.Holdingperiod:1to4quarters.16=4×4strategiesAsecondsetof16strategies:skippingaweekbetweentheportfolioformationperiodandtheholdingperiodtoavoidsomeofthe:1.Bid-askspread2.Pricepressure3.Laggedreactioneffects.(Jegadeesh(1990)andLehmann(1990))2020/3/1513Ⅰ.TradingStrategies:RSS(2)J-month/K-monthstrategy（J,K)strategyInanygivenmontht,thestrategiesholdaseriesofportfoliosthatareselectedinhecurrentmonthaswellasinthepreviousK-1months.Inanygivenmonth,werevisetheweighton1/Kofthesecuritiesintheentireportfolioandcarryovertherestfromthepreviousmonth.Wepresentonlytherebalancedreturns2020/3/1514Ⅱ.TheReturnsofRSS.Sampleperiod:1965—1989TheCRSPdailyreturnsfile:allstockswithavailablereturnsdataintheJmonthsprecedingtheportfolioformationdate.TableⅠ:averagereturnsofthedifferentbuyandsellportfoliosaswellasthezero-cost,winnersminuslosersportfolios,forthe32strategiesdescribedabove.2020/3/1515TableⅠMostsuccessfulZero-coststrategy略高3.1%permonth1.t值小，不够显著2.MostsuccessfulZero-coststrategy2020/3/1516Ⅲ.SourcesofRelativeStrengthProfits-----twosimplereturn-generatingmodeltodecomposetheexcessreturnsandidentifytheimportantsourcesofrelativestrengthprofits.Thefirstmodel:允许因素模拟组合序列相关。但要求股价对共同因素实时反映。超额回报被分解为三部分：其中两部分与系统风险有关，存在于有效市场。第三部分与特定的公司相关，只有当市场无效时才会对RSS的回报有贡献。Thesecondmodel:放松了股价对共同因素实时反映的假设。用以评估由于股票价格的lead-l