马克维茨投资组合选择.docx
《马克维茨投资组合选择.docx》由会员分享,可在线阅读,更多相关《马克维茨投资组合选择.docx(21页珍藏版)》请在冰豆网上搜索。
马克维茨投资组合选择
马克维茨投资组合选择
PortfolioSelection
HarryMarkowitz
TheJournalofFinance,Vol.7,No.1.(Mar.,1952),pp.77-91.
StableURL:
http:
//links.jstor.org/sici?
sici=0022-1082%28195203%297%3A1%3C77%3APS%3E2.0.CO%3B2-1
TheJournalofFinanceiscurrentlypublishedbyAmericanFinanceAssociation.
YouruseoftheJSTORarchiveindicatesyouracceptanceofJSTOR'sTermsandConditionsofUse,availableat
http:
//www.jstor.org/about/terms.html.JSTOR'sTermsandConditionsofUseprovides,inpart,thatunlessyouhaveobtained
priorpermission,youmaynotdownloadanentireissueofajournalormultiplecopiesofarticles,andyoumayusecontentin
theJSTORarchiveonlyforyourpersonal,non-commercialuse.
Pleasecontactthepublisherregardinganyfurtheruseofthiswork.Publishercontactinformationmaybeobtainedat
http:
//www.jstor.org/journals/afina.html.
EachcopyofanypartofaJSTORtransmissionmustcontainthesamecopyrightnoticethatappearsonthescreenorprinted
pageofsuchtransmission.
TheJSTORArchiveisatrusteddigitalrepositoryprovidingforlong-termpreservationandaccesstoleadingacademic
journalsandscholarlyliteraturefromaroundtheworld.TheArchiveissupportedbylibraries,scholarlysocieties,publishers,
andfoundations.ItisaninitiativeofJSTOR,anot-for-profitorganizationwithamissiontohelpthescholarlycommunitytake
advantageofadvancesintechnology.FormoreinformationregardingJSTOR,pleasecontactsupport@jstor.org.
http:
//www.jstor.org
SunOct2107:
53:
252007
PORTFOLIOSELECTION*
HARRYMARKOWITZ
TheRandCorporation
THEPROCESSOFSELECTINGaportfoliomaybedividedintotwostages.
Thefirststagestartswithobservationandexperienceandendswith
beliefsaboutthefutureperformancesofavailablesecurities.The
secondstagestartswiththerelevantbeliefsaboutfutureperformances
andendswiththechoiceofportfolio.Thispaperisconcernedwiththe
secondstage.Wefirstconsidertherulethattheinvestordoes(orshould)
maximizediscountedexpected,oranticipated,returns.Thisruleisrejected
bothasahypothesistoexplain,andasamaximumtoguideinvestment
behavior.Wenextconsidertherulethattheinvestordoes(or
should)considerexpectedreturnadesirablethingandvarianceofreturn
anundesirablething.Thisrulehasmanysoundpoints,bothasa
maximfor,andhypothesisabout,investmentbehavior.Weillustrate
geometricallyrelationsbetweenbeliefsandchoiceofportfolioaccording
tothe"expectedreturns-varianceofreturns"rule.
Onetypeofruleconcerningchoiceofportfolioisthattheinvestor
does(orshould)maximizethediscounted(orcapitalized)valueof
futurereturns.lSincethefutureisnotknownwithcertainty,itmust
be"expected"or"anticipatded7'returnswhichwediscount.Variations
ofthistypeofrulecanbesuggested.FollowingHicks,wecouldlet
"anticipated"returnsincludeanallowanceforrisk.2Or,wecouldlet
therateatwhichwecapitalizethereturnsfromparticularsecurities
varywithrisk.
Thehypothesis(ormaxim)thattheinvestordoes(orshould)
maximizediscountedreturnmustberejected.Ifweignoremarketimperfections
theforegoingruleneverimpliesthatthereisadiversified
portfoliowhichispreferabletoallnon-diversifiedportfolios.Diversification
isbothobservedandsensible;aruleofbehaviorwhichdoes
notimplythesuperiorityofdiversificationmustberejectedbothasa
hypothesisandasamaxim.
*ThispaperisbasedonworkdonebytheauthorwhileattheCowlesCommissionfor
ResearchinEconomicsandwiththefinancialassistanceoftheSocialScienceResearch
Council.ItwillbereprintedasCowlesCommissionPaper,NewSeries,No.60.
1.See,forexample,J.B.Williams,TheTheoryofInvestmentValue(Cambridge,Mass.:
HarvardUniversityPress,1938),pp.55-75.
2.J.R.Hicks,Val~eandCapital(NewYork:
OxfordUniversityPress,1939),p.126.
Hicksappliestheruletoafirmratherthanaportfolio.
78TheJournalofFinance
Theforegoingrulefailstoimplydiversificationnomatterhowthe
anticipatedreturnsareformed;whetherthesameordifferentdiscount
ratesareusedfordifferentsecurities;nomatterhowthesediscount
ratesaredecideduponorhowtheyvaryovertime.3Thehypothesis
impliesthattheinvestorplacesallhisfundsinthesecuritywiththe
greatestdiscountedvalue.Iftwoormoresecuritieshavethesamevalue,
thenanyoftheseoranycombinationoftheseisasgoodasany
other.
Wecanseethisanalytically:
supposethereareNsecurities;letritbe
theanticipatedreturn(howeverdecidedupon)attimetperdollarinvested
insecurityi;letdjtbetherateatwhichthereturnontheilk
securityattimetisdiscountedbacktothepresent;letXibetherelative
amountinvestedinsecurityi.Weexcludeshortsales,thusXi20
foralli.Thenthediscountedanticipatedreturnoftheportfoliois
Ri=xm
di,Titisthediscountedreturnoftheithsecurity,therefore
t-1
R=ZXiRiwhereRiisindependentofXi.SinceXi20foralli
andZXi=1,RisaweightedaverageofRiwiththeXiasnon-negative
weights.TomaximizeR,weletXi=1foriwithmaximumRi.
IfseveralRa,,a=1,...,Karemaximumthenanyallocationwith
maximizesR.Innocaseisadiversifiedportfoliopreferredtoallnondiversified
poitfolios.
Itwillbeconvenientatthispointtoconsiderastaticmodel.Instead
ofspeakingofthetimeseriesofreturnsfromtheithsecurity
(ril,ri2)...,rit,...)wewillspeakof"theflowofreturns"(ri)from
theithsecurity.Theflowofreturnsfromtheportfolioasawholeis
3.Theresultsdependontheassumptionthattheanticipatedreturnsanddiscount
ratesareindependentoftheparticularinvestor'sportfolio.
4.Ifshortsaleswereallowed,aninfiniteamountofmoneywouldbeplacedinthe
securitywithhighestr.
PortfolioSelection79
R=ZX,r,.Asinthedynamiccaseiftheinvestorwishedtomaximize
"anticipated"returnfromtheportfoliohewouldplaceallhisfundsin
thatsecuritywithmaximumanticipatedreturns.
Thereisarulewhichimpliesboththattheinvestorshoulddiversify
andthatheshouldmaximizeexpectedreturn.Therulestatesthatthe
investordoes(orshould)diversifyhisfundsamongallthosesecurities
whichgivemaximumexpectedreturn.Thelawoflargenumberswill
insurethattheactualyieldoftheportfoliowillbealmostthesameas
theexpectedyield.5Thisruleisaspecialcaseoftheexpectedreturnsvariance
ofreturnsrule(tobepresentedbelow).Itassumesthatthere
isaportfoliowhichgivesbothmaximumexpectedreturnandminimum
variance,anditcommendsthisportfoliototheinvestor.
Thispresumption,thatthelawoflargenumbersappliestoaportfolio
ofsecurities,cannotbeaccepted.Thereturnsfromsecuritiesare
toointercorrelated.Diversificationcannoteliminateallvariance.
Theportfoliowithmaximumexpectedreturnisnotnecessarilythe
onewithminimumvariance.Thereisarateatwhichtheinvestorcan
gainexpectedreturnbytakingonvariance,orreducevariancebygiving
upexpectedreturn.
Wesawthattheexpectedreturnsoranticipatedreturnsruleisinadequate.
Letusnowconsidertheexpectedreturns-varianceofreturns
(E-V)rule.Itwillbenecessarytofirstpresentafewelementary
conceptsandresultsofmathematicalstatistics.Wewillthenshow
someimplicationsoftheE-Vrule.Afterthiswewilldiscussitsplausibility.
Inourpresentationwetrytoavoidcomplicatedmathematicalstatements
andproofs.Asaconsequenceapriceispaidintermsofrigorand
generality.Thechieflimitationsfromthissourceare
(1)wedonot
deriveourresultsanalyticallyforthen-securitycase;instead,we
presentthemgeometricallyforthe3and4securitycases;
(2)weassume
staticprobabilitybeliefs.Inageneralpresentationwemustrecognize
thattheprobabilitydistributionofyieldsofthevarioussecuritiesisa
functionoftime.Thewriterintendstopresent,inthefuture,thegeneral,
mathematicaltreatmentwhichremovestheselimitations.
Wewillneedthefollowingelementaryconceptsandresultsof
mathematicalstatistics:
LetYbearandomvariable,i.e.,avariablewhosevalueisdecidedby
chance.Suppose,forsimplicityofexposition,thatYcantakeona
finitenumberofvaluesyl,yz,...,y,~L.ettheprobabilitythatY=
5.U'illiams,op.cit.,pp.68,69.
80TheJournalofFinance
yl,bepl;thatY=y2bepzetc.Theexpectedvalue(ormean)ofYis
definedtobe
ThevarianceofYisdefinedtobe
VistheaveragesquareddeviationofYfromitsexpectedvalue.Visa
commonlyusedmeasureofdispersion.Othermeasuresofdispersion,
closelyrelatedtoVarethestandarddeviation,u=.\/Vandthecoefficient
ofvariation,a/E.
Supposewehaveanumberofrandomvariables:
R1,...,R,.IfRis
aweightedsum(linearcombination)oftheRi
thenRisalsoarandomvariable.(ForexampleR1,maybethenumber
whichturnsupononedie;R2,thatofanotherdie,andRthesumof
thesenumbers.Inthiscasen=2,a1=a2=1).
Itwillbeimportantforustoknowhowtheexpectedvalueand
varianceoftheweightedsum(R)arerelatedtotheprobabilitydistribution
oftheR1,...,R,.Westatetheserelationsbelow;werefer
thereadertoanystandardtextforproof.6
Theexpectedvalueofaweightedsumistheweightedsumofthe
expectedvalues.I.e.,E(R)=alE(R1)+aZE(R2)+...+a,E(R,)
Thevarianceofaweightedsumisnotassimple.Toexpressitwemust
define"covariance."ThecovarianceofR1andRzis
i.e.,theexpectedvalueof[(thedeviationofR1fromitsmean)times
(thedeviationofR2fromits