第七章效率市場證券投資學(xué)北大,楊云紅VIP

第七章效率市場主要內(nèi)容效率市場的定義效率市場的實證結(jié)果證券市場中的幾種反?，F(xiàn)象行為金融MauriceKendall,Theanalysisofeconomictimeseries,PartI:Prices,JournaloftheRoyalstatisticalsociety96,1953.Hecouldidentifynopredictablepatternsinstockprices.1.Whyarenotstockpricespredictable?例子：一種模型預(yù)測股票價格三天后將從20元/股漲至23元/股Theforecastofafuturepriceincreasewillleadinsteadtoanimmediatepriceincreasethestockpricewillimmediatelyreflectthegoodnewsimplicitinthemodel’sforecast.例子：2001年10月21日晚，證監(jiān)會宣布暫停國有股減持例子：2002年4月5日宣布，從5月1日起交易傭金使用0.3%上限向下浮動制。例子：中國人民銀行決定，從2004年10月29日起上調(diào)金融機構(gòu)存貸款基準(zhǔn)利率并放寬人民幣貸款利率浮動區(qū)間和允許人民幣存款利率下浮。金融機構(gòu)一年期存款基準(zhǔn)利率上調(diào)0.27個百分點，由現(xiàn)行的1.98%提高到2.25%，一年期貸款基準(zhǔn)利率上調(diào)0.27個百分點，由現(xiàn)行的5.31%提高到5.58%。由于存在聰明的投資者的原因，任何能夠用來對股票價格作預(yù)測的信息已經(jīng)反映在股票的價格中。任何新信息，如果是可以預(yù)測的，則已經(jīng)反映在價格中；如果是不可預(yù)測的，則導(dǎo)致的股票價格變動也是不可預(yù)測的，即隨機的信息導(dǎo)致隨機的股票價格變化。2.股票價格的隨機游走(randomwalk)股票價格的變動服從隨機游走的形式。隨機游走的形式并不是說明市場是非理性的，而恰恰表明這是投資者真相尋求相關(guān)信息，以使得自己在別的投資者獲得這種信息之前買或者賣股票而獲得利潤的結(jié)果。反之，如果股票價格不是隨機變動而是可以預(yù)測的，則表明所有的相關(guān)信息并沒有完全反映在價格中，這表明市場是非有效的。股票價格反映了所有可得信息稱為有效性市場假設(shè)。有效資本市場指的是現(xiàn)時市場價格能夠反映可得信息的資本市場，在這個市場中，不存在利用可得信息獲得超額利潤的機會。資本市場的有效性這一概念對于金融投資者而言具有非常重要的意義。因為市場的有效性消除了許多可以提高收益的策略。MarketefficiencymeansPricesarecorrectTheyfullyreflectallavailableinformationPeopleuseallavailableinformationinformingexpectationsaboutfuturecashflows.Thediscountrateisrightfortherisknessofthecashflows.PricesreacttonewinformationquicklyandtotherightextentThereisnofreelunchTheonlywayyoucangethigherreturnsisbytakingonmoreriskThereisnoinformationouttherethatcanbeusedtoconstructstrategiesthatearnreturnshigherthanrequiredfortheirrisk.Whenwesay‘pricesarecorrect’,weareimplicitlystatementwhat‘correct’is(i.e.,weareassuminganassetpricingmodel)3.有效資本市場的描述例子：F-stopCameraCorporationisattemptingtodevelopacamerathatwilldoublethespeedoftheauto--focusingsystemnowavailable.在特定的價格下，持有股票的原因著名的光學(xué)專家加盟股價是否上漲股價上漲的原因股價上漲的時間一個市場對于一個信息集來說稱為有效的，如果不存在利用該信息獲得超額利潤的機會。有效和非有效市場中價格對新信息的反應(yīng)股票價格

0宣布前（-）或者后（+）的天數(shù)過激反應(yīng)和回歸延遲反應(yīng)有效市場對新信息的反應(yīng)三種有效市場(efficientcapitalmarket)不同的信息集對證券價格產(chǎn)生影響的速度不一樣。為了處理不同的反應(yīng)速度，把信息集分成不同的類別。最常用的一種分類方法：過去價格的信息，可得的公共信息，所有信息。針對這三種信息集，有三種形式的有效市場的定義弱有效市場(theweakform)半強有效市場(thesemistrongform)強有效市場(strongform)弱有效市場定義：一個資本市場稱為弱有效的或者滿足弱有效形式，如果證券價格充分反應(yīng)了包含在歷史價格中的信息。弱形式有效通常表示成下面的數(shù)學(xué)形式

+Expectedreturn+Randomerror這里隨機誤差項的均值為0，且不同時間的隨機誤差項是不相關(guān)的。弱形式有效性是最弱類型的有效性。股票價格的歷史數(shù)據(jù)是可以免費得到的，如果這些數(shù)據(jù)里包含有用的數(shù)據(jù)，則所有的投資者都會利用它，導(dǎo)致價格調(diào)整，最后，這些數(shù)據(jù)失去價值?！畉echnical’analysisusingpastpricepatternswillnotproduceprofits.如果一個市場是弱有效的，考慮一個交易策略如果某股票的價格連續(xù)漲三天，就買進該股票；如果股票的價格連續(xù)降三天，就賣出股票。問題：這個策略能否賺錢。FairlyconvincedthatmarketsareweakformefficientBut,newevidence(i.e.,momentum)haschallengedthis.半強形式有效市場定義：一個市場是半強形式有效的，如果價格反應(yīng)了所有公共可得的信息。這些信息包括：歷史價格數(shù)據(jù)、與公司生產(chǎn)有關(guān)的基本數(shù)據(jù)、管理的質(zhì)量、資產(chǎn)負債表、專利情況、收益預(yù)測、會計處理‘fundamental’analysis(e.g.sortingthroughincomestatements)willnotproduceprofits.Ex.Formingportfoliosonaccountingratios,balancesheet,orincomestatementinformationwillnotgenerateabnormalprofits.NoevidenceofabnormalreturnsafterapublicannouncementProfessionalmoneymanagersdonotoutperformthemarketMarketseemtobesemi-strongefficientBut,B/MandE/Pstrategiesstillchallengethis.(mayberisk,maybenot)強形式有效市場定義：一個市場是強有效的，如果價格反應(yīng)了所有的信息，不管是公共的還是私有的。InsidertradingwillnotproduceprofitsEx.Knowingamergerisgoingtotakeplacebeforeitisannouncedpublicly.Althoughillegal,thereisevidencethatpricesmovebeforeannouncements,suggestinginsidertradestakeplaceInsidertradingappearsprofitable,indicatingmarketsarenotstrongformefficient.But,theseprofitsareshortlived,suggestingthemarketmaybeclosetoefficient.強形式有效性半強形式有效性弱形式有效性說明三種有效性的例子：總是在股價上漲后賣出股票，能賺到錢投資者在一家公司宣布增加收益后買該公司股票，能賺到錢知道采礦公司是否開采到了金子的內(nèi)部消息后買該公司股票，能賺到錢4.Howcanwetellifmarketsareinefficient?Lookforstock-pickingstrategiesbasedonsomepastinformationwhichhaveearnedhighreturnswithlittlerisk.Unfortunately,wecanneverbesureofinefficiency.Itisalwayspossiblethatwearenotmeasuringriskproperly,i.e.,wedonotknowwhattherightdiscountrateisThisisthe‘Jointhypothesisproblem’5.Whywouldweexpectmarketstobeefficient?theforcesofarbitragesmartinvestorsexploitthemispricinginsecuritiesuntilitdisappearsToshowthatmarketsareineffcient,needtoshowthatpeoplemakeerrorsinsettingpricesthatarbitragefailstoeliminatetheseerrors一些例子GrossmanandStiglitz的結(jié)果不同市場的有效性不同：中國和美國的股市小股票和大股票的有效性不同RandomWalksandEfficientMarketsitisoftenthoughtthatefficientmarketspricesmoverandomlythisisnotnecessarilytrueStrictlyspeaking,weshouldcharacterizestockpriceasfollowingasubmartingale,meainngthattheexpectedchangeinthepricecanbepositive,presumablyascompensationforthetimevalueofthemoneyandsystematicrisk.Moreover,theexpectedreturnmaychangeovertimeasriskfactorschange.returnsaremean-revertingifthediscountrateforanassetdoesnotchangeovertime,thenitistruethatefficientmarketsrandomwalke.g.overshorttimeframes,returnsshouldlookrandom6.Evidenceformarketefficiencystockpricesappeartomoverandomlynewinformationappearstobequicklyincorporatedintopricese.g.announcementofatakeoverdoan‘eventstudy’tolookatthestockpricereactiontothenewsaverageovermanycompaniesProfessionalmoneymanagersdonotclearlybeatthemarketonaverage.7.EvidenceagainstmarketefficiencydidthevalueoftheU.S.economyreallydrop20%inOctober1987?thevolumeoftradingonstockexchangesistoohightobeconsistentwithrationalinvestorsthevolatilityofthemarketistoohigh(Shiller,1982).why?whyaretheresomanymutualfunds?thereareinvestmentstrategieswhichappeartohaveearnedhigheraveragereturnsthanisconsistentwiththeirriskThesearesocalled‘marketanomalies’suggeststhatnewinformationisnotalwaysimmediatelyincorporatedintopricesThisevidencereferstoweak/semi-strongversionsofmarketefficiencyHowaboutstrong-formefficiency?canyoumakemoneyusinginside,non-publicinformation?YES!stockpricesoftenmoveinadvanceofimportantcompanyannouncementsfromrecordsofinsidertrades,wefindthattheymakemoneyonaverageBUT:it'sillegalforcompanyinsiders(orpeopletippedbythem)totransactbasedonmaterialnon-publicinformationHowaboutinstitutionalinvestors?Wementionedmoneymanagersdonotexhibitabnormalperformanceonaverage,butmaybesomelocaladvantage.8.ThemarketanomaliesAnomaliescanbethoughtofasinvestmentstrategieswhichseemtoearnhighreturnswithoutbeingveryriskyThestrategiesarenormallybasedonsomefirmcharacteristic:sizeofthefirmitsprice-earningsratioRecipe:formaportfoliobasedonobservablecharacteristics,andmeasureitsreturnsovertimedoesthestrategygivehighreturnsonaverage?Whylookataveragereturns?IFYES:thestrategymayberiskyandthehighaveragereturnsarejustfaircompensationforthatriskhowdowemeasurerisk?(seebelow)ifriskdoesnotexplainthehighreturns,isitevidenceofmarketinefficiency?itmaybespurious,theresultofdata-miningifyoutrymanystrategies,someofthemwilldogreatinhistoricaldatadoesn'ttellyouanythingaboutfutureperformancepoorriskmeasurementfrictionstotradingandexploitingtheanomaly(bid-askspread,transactionscosts,liquidity,taxes,etc.)Measuringtheriskofastrategystandarddeviationdownsideriskdoesthestrategysometimesperformverypoorly?betalooksatstrategy'spayoffrelativetothemarket'spayoffahighcovarianceisunattractive(risky)lookatcovarianceofstrategy'spayoffwithvariableslikeGNPgrowth(factormodels)astrategywhichdoeswellingoodstatesoftheworldandpoorlyinbadstatesisriskyNote:evenifnoneoftheaboveturnupameasureofrisk,efficientmarketsenthusiastswillsay:theriskispresent,butjusthasn'tsurfacedwithinthesampleanalyzedOR:we'rejustnotlookingattherightmeasureofriskSmallvs.LargestocksSmallstockshaveoutperformedlargestocksbyabout12%ayearover1929-1979timeperiodShouldwebuylotsofsmallstocks?DependsonwhethertheyareriskiersmallstockshavehigherstandarddeviationsandbetasbutnothighenoughtoexplaintheirreturnsSo,smallstocksmaybemispricedBUT:althoughsmallstockreturnsarehigh,thismayonlybebasedonafewextraordinaryyearsWemaybemissingsomedimensionofriskTheJanuaryEffectMuchofthesmallfirmeffectseemstooccurinJanuaryMuchofsmallfirmpremiumoccursinfirstfivedaysofJanuaryoftenexplainedbytax-losssellinghowever,effectiswidespreadininternationalmarketsalso,evenwhenthere'snocapitalgainstaxand,therestillseemstobeasizeeffectaftercontrollingforthis.IfthepositiveJanuaryeffectisamanifestationofbuyingpressure,itshouldbematchedbyasymmetricnegativeDecembereffect.maynotaccordwithefficientmarketswhydon'tpeoplebuyinDecemberinanticipation?andthepredictableJanuaryeffectfliesinthefaceofefficientmarkettheory.‘Windowdressing'byinstitutionalinvestorsinfusionofcapitalatbeginningofyearOverreactionstudiessomestudiessuggestthatthereareinefficienciesduetopeopleoverreactingtoinformation（1）LosersandWinners----Longtermreversal

takeathreeyearperiodandrankstocksonthebasisoftheirperformanceoverthatperiodforma‘winner"portfolioofthetop10%best-performingstocksforma‘loser"portfolioofthebottom10%worst-performingstocksthisiscalledacontrarianstrategylookattheirreturnsoverthenextfewyearstheloserportfolioseemstooutperformthewinnerportfolio(DeBondtandThaler(1985))Twoexplanations:overreaction:thewinnersarefirmsthatpeoplehavebecometooexcitedaboutsubsequently,theyrealizethattheyweretoooptimisticpricefallsandreturnsarelowrisk:thelosersareriskierfirmstheirhigherreturnsarejustcompensationforriskbutlosersdonotappearriskieronstandardmeasuresofriskHowdoweaccountforrisk?VarianceUseregressionanalysisandapricingmodelCAPMFF3-factormodelExaminewhenthestrategyexhibitsthehighestandlowestpayoffs(i.e.,doesthestrategydowellwhenthemarketdoes?whenweareinthepeakofabusinesscycle?arecession?,etc.)Catastropherisk（2）ValueandGrowthformportfoliosofvaluestocksavaluestockisonewithlowpricerelativetosomemeasureoffundamentals,i.e.highbooktomarket(B/M)ratioscashflowtoprice(C/P)ratiosearningstoprice(E/P)ratiosalsoformportfoliosofgrowthstocks,i.e.withlowvaluesoftheseratiosWhygrowthvs.value?findthatvaluestocksdramaticallyoutperformgrowthstocksWhy?rational:Representsadistressfactorintheeconomy.Valuestocksaremorepronetothissourceofriskthangrowthstocks.higheraveragereturns.Valuestocksaretypically`fallenangels'irrational:Growthstocksare`glamorous'.Peopletendtowanttobuytheseandstampedetowardsthem,pushinguptheprice,anddepressingfuturereturns.Valuestockshavebeenneglected,causingtheirpricetofall,andexpectedreturnstorise.Thisisanoverreactionstory.Also,valuestocksdonotappearriskier,however.theydon'thavehighervariancetheydon'thavehighdownsiderisk(i.e.,donotunderperformoftenorbythatmuch)theydon'thavehigherbetastheydon'tunderperforminbadstatesoftheworldLakonishok,Shleifer,andVishny(1994)somaybeit'saninefficiency,drivenagainbyoverreactionPsychologicalfoundationsforoverreaction:representativenessheuristicsmallsamplebiasbutofcourse,itcouldstillberisk!TheFama-FrenchdebatesmallstocksandhighB/Mstocksearnreturnsthatarehigherthanisrequiredfortheirrisk,accordingtotheCAPMmeasureofrisktwopossibilities:smallstocksandhighB/MstocksaremispricedOR:theCAPMisn'tmeasuringriskproperlyFamaandFrench(1993)constructanewasset-pricingmodel,the‘’3-factormodel"whichmakessmallandhighB/Mstockslookriskierstartwiththemarketfactor,andaddtwonewfactors,F(small)(=SMB)andF(B/M)(=HML)theysupposedlytrackgoodandbadstatesoftheworldsmallstockshavehighcovariancewithF(small)highB/MstockshavehighcovariancewithF(B/M)theyareriskier!EvidenceinfavorofFF:Adjustingthelong-runcontrarianprofitsfoundbyDeBondtandThaler(1985)usingthe3-factormodel,theprofitsdisappear(i.e.,alpha=0.).pastlong-termlosersloadhigheronSMBandHMLthanpastwinners,eventhoughtheyhavethesamemarketbeta.long-termlosersmoreriskythanlong-termwinners.The3-factormodelcapturesahostofotheranomaliesaswell!EvidenceagainstFF:LSV(1994):HMLandotherbook-to-priceratiosperformwellinpoortimes.DanielandTitman(1997):characteristicsratherthanfactorloadingspriceassetsbettercontrolforsizeandB/Mcharacteristics,SMBandHMLnolongerexplainaveragereturnsex:twostockswithsamesizebutdifferentbetasonSMBhavethesameaveragereturn.twostockswithsamebetaonSMBbutdifferentsizeshavedifferentaveragereturns.DoesthisnecessarilyimplysizeandBE/MEareirrationalanomalies?No,couldproxyforsometrueunknownfactor,bettercapturedbythecharacteristic.Measurementofbetasarepronetoestimationerror,whichincreasesnoiseintherelationbetweenthebetasandaveragereturns.Underreactionstudiesotherstudiessuggestthatthereareinefficienciesduetopeopleunderreactingtoinformation（1）Momentumformportfoliosofstocksthatperformedverywellinrecentpast(`winners')i.e.overthepast3monthstooneyearsimilarly,formaportfolioof`loser'stocksthewinnersoutperformtheloserse.g.buythewinnersandselltheloserse.g.azero-costportfoliowhichbuyswinnersandsellslosersfromthepast6monthsearns12%(annually)overthenext6months!contrasttothemean-reversionresultsourceofinefficiency:underreactiontoinformation?Or,isitrisk?（2）EarningsAnnouncementsslowpriceresponsetoearningsannouncements(post-earningsannouncementdrift)eachquarter,rankstocksonthesizeofthesurpriseintheirearningsannouncement(surprise=actual-expected)formaportfolioofstockswithlargestpositivesurprises(portfolioA)andaportfolioofstockswiththelargestnegativesurprises(portfolioB)AoutperformsBisitrisk?beta,factormodeltypechecksdon'tfindanyover50quartersfrom1974-1986,strategyearnedpositiveabnormalreturns46timesManyexamplesofunderreactiontoinformation:tochangesindividendpolicystrategy:buycompaniesthathavejustannouncedadividendincrease;sellthosethathavejustannouncedacutindividendstorepurchasesofsharesstrategy:buycompaniesthathavejustannouncedasharerepurchasePsychologicalFoundationsforUnderreaction:conservatism,pessimismOtheranomaliesThenewissuespuzzleHotoffering

Iftheseanomaliesareinefficiencies,whydoesn'tarbitrageeliminatethem?inpractice,arbitrageislimiteddonothaveinfinitenumberofstocks,thereforethereissomerisksometimesthestrategycandoverypoorly,andyoulosealotofmoneymoneymanagersandindividualsmayhaveshorthorizons(duetoregularevaluationsorpsychologicalpreferences)amispricingcantakeawhiletocloseinfactitmaynotclosewithintheinvestor'shorizoninvestorwillrestrictthesizeofpositiontakentherecanbehightransactionsandtradingcostsduetoturnoverinthesestrategiesliquiditycanbelowattimeswhenyouneeditmostmonitoringcanbehighthereareshortsalesconstraintsMeasurementissues:Theseanomaliesaremarket-basedmeasurese.g.,theyincludepricesize,BE/ME,pastreturns(contrarianandmomentum),E/P,etc.,allcontainpriceinthem.soanythingmissedbythepricingmodelwillshowupinoneofthesevariablese.g.,appearsonbothsidesoftheregressionequationthiswasnotedbyBall(1978)andformalizedinBerk(1995)wemayneverknowifpickingupmispricingorinadequacyofpricingmodel.Dataminingtechniquesoffindinganomaliesareoftensubjecttothedataminingcritiqueifyoutryenoughvariables,somethingwilleventuallyappeartopredictreturnsbut,theforecastpowerofthisvariable

willbecompletelyspurious(i.e.,itwon'tworkoutofsample)e.g.,generate100differentdataseriesofcompletelyrandomnumbersrunaregressionofactualstockreturnsoneachofthe100randomdataseries.someoftheregressionswillproducesignificantresults:doesthismeanyoucanmakemoney?NO!thedataminingcritiqueisverypowerful:bearitinmindwhenpeopletrytoimpressyouwithstrategiesthatworkedgreatinthepast.theyprobablywon'tworkinthefuture!ResponsetoDataMiningCritique:Manyoftheseanomaliesappearinother(international)markets(seeHawa