Liquidity premium in stock returns, the case of vietnam

PROBLEMSTATEMENT

This article explores the vital role of stock markets in countries, highlighting their numerous benefits for investors, corporations, and the overall economy It will discuss the establishment of stock markets globally and the growing interest in stock returns among various participants The subsequent sections will focus on research into the determinants of stock returns, particularly the emergence of liquidity as a key factor Additionally, it will address the current trend of liquidity research, which has predominantly focused on developed markets, emphasizing the need for comprehensive studies on the impact of liquidity on stock returns in frontier markets like Vietnam.

Stockmarketisundeniablyanintegralpartofeachcountryeconomy.Stockmarketproviden u m b e r o f benefitsf o r i n d i v i d u a l i n v e s t o r s , corporationsandeconomy.Forcorporation,s t o c k marketal lowcompanytogainaccesstohugecapitalmarket.Oncethecompanyislisted,itcane x p a n d itsca pitalthroughshareissuance.Inaddition,mergeandacquisitioncanbefacilitatedbysharepurchaseinth estockmarket.Forinvestors,stockmarketprovideinvestorsachannelfori n v e s t i n g theirmo ney.Therearemanydifferenttypesof companieswhichshouldsuitthetasteofdifferentinvestors.Abouteconomy,thekeybenefitofstockmar ketisthatitpromoteseconomicgrowthbyencouraginginvestorstoputtheirsavingintolistedcompani es.Asaresult,itencouragesthe companies’developmentand promoteseconomicgrowth.

The stock market has developed significantly in many countries, particularly in developed nations where it has been established for centuries The involvement of various social parties has made stock returns a key interest for many market participants, especially investors who prioritize stock returns when deciding where to allocate their funds For corporations, stock returns serve as a reliable measure of performance, influencing investor decisions on buying and selling stocks based on fundamental and prospective evaluations Additionally, stock returns act as a dependable barometer for assessing a country's economic condition, reflecting macroeconomic trends and major changes The fluctuations in stock prices often align with a country's economic cycles, allowing governments to monitor the health of their stock markets and implement appropriate policies to foster economic development.

Fromt h i s realisticneed,m a n y researchershavebeend e v o t i n g their ef fo rt s i n studyingab outdeterminantsofstockreturnsanditsmechanism.Theliteratureonstockreturnsnowadaysi s incredi blyenormousandmanyscholarsstillkeepsearchingforunknowndeterminantsandnewmethodologies.

Inreality,onefactorthatattractedattentionofstockmarketparticipantsandithasbeennotic edforalongtime,thisisliquidity.Liquidityiscommonlydefinedastheabilitytopurchaseo r sellalarge quantityofstocksquicklyatlowcostwithoutaffectingthepricesignificantly(Choe

&Yang,2008).Throughmanyyears,investorsobservedthatliquidstockcanbeeasilyconvertedi n t o ca shwithoutmuchdifficulties.Incontrast,illiquidstockcausedsomelevelofdifficultiesforinvestorswhent heywanttoconvertthestocksintocash.Itoftenrequiresinvestorstosellatalowerpriceorendurea greatertransactioncostforthesale.Withthisinmind,investorswellawareo f t he l i q u i d i t y premium( t h e pricespreadbetweenl i q u i d andi l l i q u i d stocks).However,th er e wasn’tanyremarkablestudyabo utliquidityuntilthewell-knownpaperofAmihudandMendelson(1986).

Amihud and Mendelson (1986) pioneered liquidity research with their transaction cost theory, revealing a significant liquidity premium in asset returns This foundational paper sparked extensive scholarly interest, leading to the development of various liquidity proxies to measure liquidity across different stock markets Over the years, many researchers have concurred that investors demand higher stock returns as compensation for illiquidity (Amihud, 2002; Brennan & Subrahmanyam, 1996; Brennan, Chordia, & Subrahmanyam, 1998; Chordia, Roll, & Subrahmanyam, 2000), indicating a negative impact of liquidity on stock returns However, some researchers have challenged this perspective, presenting evidence of a positive relationship between liquidity and returns, supported by their own arguments and empirical studies (Bali, Peng, Shen, & Tang, 2014; Abzari, Fathi, & Kabiripour, 2013).

Most research on liquidity has primarily focused on developed financial markets, particularly in the United States, where data is abundant However, Bekaert, Harvey, and Lundblad (2007) argue that the liquidity effect is more pronounced in emerging markets They provide several reasons for this conclusion First, poor liquidity has historically deterred foreign institutional investors from committing capital to various stocks in emerging markets, which exacerbates the liquidity premium, especially in frontier markets that face even greater liquidity gaps Second, many emerging markets experienced market liberalization during the study period, which significantly increased liquidity compared to the pre-liberalization phase Lastly, developed markets often feature a diversified ownership structure that includes both long-term and short-term investors, further influencing liquidity dynamics.

(2007)suggesttheclienteleeffectsinselectinginvestmentportfolioshouldreducethepricingofliquidity.I ncaseofemergingmarkets,t h e diversificationinnumberofsecuritiesandownershipislackedof whichoftenintensifythel i q u i d i t y effects.

The need for research on liquidity premium in frontier markets, particularly in Vietnam, is evident due to differing opinions among researchers regarding the impact of liquidity on stock returns The liquidity gap between liquid and illiquid stocks in Vietnam is substantial, highlighting a significant area of study Current research on liquidity in frontier markets, especially in Vietnam, is scarce, limiting a comprehensive understanding of its role in the local stock market The only notable study by Xuân Vinh and Hồng Thu (2013) found a positive correlation between liquidity measures and stock returns, while noting a negative relationship for illiquid stocks Their findings suggest that frequent trading by small investors may drive demand for liquid stocks, resulting in higher returns Additionally, the dynamic nature of the Vietnamese stock market, with numerous newly listed stocks each year, contributes to increased liquidity and stock prices during their Initial Public Offerings.

Ibelievedt h e i r rationalei s q u i t e vagueandt h e i r m e t h o d f o r constructingt h e l i q u i d i t y measuresisdifferentfrom thewell–knownmethodofFama andFrench(2013).

Allabovereasonshavemotivatedm e t o c o n d u c t a thoroughresearchaboutl i q u i d i t y pre miuminVietnamfinancialmarketusingawell- knownmethodfromFamaandFrench(2013).T h i s researchhopefullywillprovideanadditionalpi eceinliteratureaboutliquiditypremiuminVietnamstockmarket,afrontiermarket.

RESEARCHOBJECTIVE

 ToempiricallyexaminetheinfluenceofliquidityonstockreturninVietnam(afrontiermarket)byu singrenownedFama andMacBeth(1973)method.

RESEARCHQUESTION

RESEARCHSCOPE

C h i M i n h S t o c k Exchange(HOSE)from January2008 toDecember2013.

(3)thetotalmarketcapitalization ofHOSEiss i g n i f i c a n t l y largerthanHNX(asof16April20 15,themarketcapitalizationofHOSEis1.051.445b i l l i o n VND,w h i l e HNXmarketcapitalizationi s

1 3 9 9 8 0 billionVND),and( 4 ) t h e marketl i q u i d i t y ofHOSEisalsoremarkablygreaterthan HNX(asof16April2015,thedaily tradedamountinHOSEis1.825billionVND,whileHNX’sdai lytradedamountis729billionVND).W i t h allofthesereasons,selecting

HOSEasprimarytradingexchangeforcollectingdatashouldensurethe qualityandprevent possiblebiases for thisresearch.

RESEARCHMETHODOLOGY

MacBeth(1973)regressionmethod.Thismethod allowustoresolveissuesinpaneldata,especiallydesigned fordealingwithfinancialdata.Inaddition,allfactorsandportfoliosinthisthesisareconstructedbyusingpr emiummethodofFamaandFrench( 1 9 9 3 ) T h i s m e t h o d i s well- acceptedasa s t a n d a r d i n b u i l d i n g e x p l a n a t o r y factorsforexplaining stockreturns.Thiswillhelpimprove thevalidityinfinalconclusionofthist h e s i s

Thebasicprocedureforconstructingpremiumfactorincludesthreemainphasesforeachvariab le.Allvariablesarederivedfromthisprocedure,exceptforSIZE,HML,RMWandCMAvariableswher etheirdailyvaluesarenotavailable.Thefirstphaseisprimarydatacalculationinwhichrequiredd a i l y d a t a f o r eachvariablei s collectedfromVietstockandfinancialauditedstatement.The dailydataforeachvariablewillcontainallnon- financialfirmstradedinHOSEina specificyear.Theseindividualfirmdataisthencalculatedbasedon adistinctformulaofeachvariabletoderive individualfirmvalueforeachvariable.Thesedailyfirmvalueofeachvariablei s thenaveragedout to obtain theaveragedailyvalue ofeachfirmduringayear.

Inthesecondphase,allyearlyaveragefirmvalueofeachvariablewillusedtocalculatedbreakpoi ntforpremiumportfolios.Theseportfoliosareconstructedonannualbasisinordertoupdateaboutfu ndamentalchangeineachvariableforeachyear.Thebreakpointsaredecidedincorrespondingtoeachv ariablewhichare oftenselectedat30and70percentileofyearly averagevalueo f allfirms.Subsequently,eachfirmi s allocatedi n t o o n e appropriateportfoliocorresp ondingto its ownyearlyaveragevalue.

Int h e t h i r d p h a s e , a f t e r eachfirmi s allocatedi n t o o n e d i s t i n c t p o r t f o l i o basedo n i t s i n d i v i d u a l variablevalue,averagemonthlystockreturnofallfirmswithinaportfolioiscombined t o calculatetheaveragemonthlyreturnofawholeportfolio.Thepremiumfactorisfinallyderivedbysubtra ctingtheaveragemonthlyreturnoffirstandthelastportfoliowithineachvariable.Theunderlyingideaf orthiswholelengthyprocedureistocomputetheexcessstockreturnbysimplyemployingapassiveinvest mentstrategywitheachpremiumfactor.Forexample,withSizefactor,t h e sizepremiumistheexcessstoc kreturnaninvestorcanearnbysimplybuyingstockofsmallcompaniesandsellingtheircurrentstockofl argecompanies.

THESTRUCTUREOFTHISTHESIS

Chapter1providesanoverviewofthesiswithproblemstatementsforselectingthissubject,research q u e s t i o n , researchs c o p e , b a s i c i n f o r m a t i o n aboutdataandmethodology.C h a p t e r 2 fo cusesliteraturereview.Inthischapter,somefundamentaltheoriesinfinanceisfirstreviewed.Then,rela tedliteratureaboutdependentandindependentvariableswillalsobeexplained.Notably,t h e theoreticala ndempiricalstudiesaboutliquidityiscarefullypresentedinaseparatesectionofchapter2 T h e finals e c t i o n o f thischapterw i l l b e dedicatedf o r t h e m a i n hypothesisandi t s rationales.

Chapter3 concentrateso n dataandmethodology.Firstly,t w o regressionm o d e l s andresear chframeworkwillbediscussedtoprovideanoverviewtoreaders.Followingthissection,dataandreg ressionmethodsectionwillbepresented.Afterthat,variablessectionwillconciselyintroduceaboutd e p e n d e n t andindependentv a r i a b l e s i n t h i s t h e s i s T h e subsequents e c t i o n i n chapter3willco ncentrateondataprocessingandconstructionofeachvariables.Thelastsectiono f thischapterwillbede dicatedtopotentialeconometricissuesandhowtheycanbesolvedinthist h e s i s

Chapter4w i l l focuso n statisticalandempiricalresulto f t h i s study.A t t h e endofth e chapter,s o m e d i s c u s s i o n s a b o u t t h e finalresultarea l s o p u t forwardw h i c h s h o w noticeableobser vationsaswellascommentsaboutthisresearch.Chapter5willbedevotedfortheconclusion,p o s s i b l e impl icationsandsomedrawbacksofthisthesis.Italsosuggestsadirectionforfurtherresearchaboutstockre turns infrontier market,especiallyinVietnam.

Thecapitalmarkettheorieslaidthefoundationforthesubsequentformationoffinancialassetpr icingmodels T h e currentandp r o m i n e n t v i e w o ffinances c h o l a r s s u p p o r t t h e v i e w o fp e r f e c t l y efficientcapitalmarketinwhichfinancialassetpricesisquicklyandaccuratelyadjusttonewinform ationasitisavailabletothemarket.Inaperfectefficientcapitalmarkets,investorsareconsideredto berisk– averseandrationalin theirdecisionmaking.

Undert h e assumptiono f perfectefficientcapitalmarket,EfficientMarketHypothesis(EMH)w asdevelopedandpublishedbyFama(1970)whichisconsideredasafundamentaltheoryo f modernfinanc e.However,theefficientmarkethypothesisdidn’tgiveinvestorsanyguidancei n assigningtheirasset s.Asaresult,othertheoriesareintroducedtoassistrisk– averseinvestorsi n allocatingtheirassetsinanefficient capitalmarket.ThemostrenownedtheoryamongthemareModernPortfolioTheory(MPT)

Withinthestreamofmarketportfolioframework,risk– averseinvestorsarebelievedtohavehomogenousexpectationsaboutthecovariance,varianceandme anoftheassetreturns,andt h e y alwaysstrivetoobtainmaximumexpectedutilityintheirinvestme ntdecisions.Inessence,M P T theorysuggestsamethodtodiversifyandachieveoptimalriskportfoliofori nvestment.Theo n l y sourceofrisktotheportfolioisderivedfromitsco- movementcorrespondingwiththemarketportfolio.CapitalAssetPricingModel(CAPM)issubsequentl ydevelopedasanextensionoftheM P T theory,this theoryprovidesinvestorswithatool topriceassetsinanefficientmarket.

Thisliteraturewillfirstreviewsomefundamentaltheoriesofcontemporaryfinancewhichareeffic ientmarkethypothesis,efficientcapitalmarket,andcapitalassetpricingtheory.Th es e theoriesformt h e f o u n d a t i o n f o r t h e currentresearchi n assetreturns.Afterreviewingt h e s e theories,theliter atureandempiricalresearchofrelatedvariablesinthisstudywillbediscussed.Baseon theliteratureandempiricalreviewofallvariables, main hypothesis forthis studywill beconstructed.

FUNDAMENTALTHEORIES

TheEfficientMarketHypothesis

Theconceptofefficientmarketisusedtodefineamarketwhereinvestorscannotc o n s i s t e n t l y outperformotherinvestorsbymakingabnormalreturnafteradjustingforrisk.Fama(1970)definedpe rfectefficientcapitalmarketas:“amarketinwhichfirmcanmakeproductioninvestmentdecisionsandin vestorscanchooseamongthesecuritiesthatrepresentsownershipof firms’activitiesundertheassumptionthatthesecuritypricesatanytimefullyreflectallavailableinformatio n”.

Intradingactivities,investorsemployalltradinginformationthatavailabletothemastoolsi n ordert o m a k e p r o f i t s i n t h e market.Therea r e numberso f informationwhichavailablet o i n v e s t o r s Themostbasicinstrumentsarethehistoricalpricepatterns,tradedvolumedata.Otherinformationi s c o m p a n y relatedp u b l i c a n n o u n c e m e n t Besides,t h e r e i s anexistenceo f i n s i d e informatio nwhichusuallydid not completelycomprehendedoraccessibletoallinvestors.

In his 1970 review, Fama analyzed prior research on asset prices in capital markets and proposed his theory of efficient capital markets, categorizing them based on the types of information incorporated into asset prices He identified three levels of market efficiency: strong form, semi-strong form, and weak form Each level excludes certain investor groups from consistently outperforming the market using specific information as trading tools In a weak form efficient market, all historical price patterns are reflected in asset prices, making it impossible for technical analysts to achieve abnormal returns The semi-strong form asserts that asset prices incorporate all current and publicly available information, preventing fundamental analysts from outperforming the market In a strong form efficient market, insider information is accessible not only to company insiders but also to outsiders, meaning that investors using such information do not outperform the market.

ModernPortfolioTheory

Onthegroundoftheefficientmarkethypothesisandprincipleofdiversification, Marko witz(1952)developedafirsttheoryofportfoliomanagementusingtheconceptofrisk.Alli n v e s t o r s areassumedtob e r i s k – a v e r s e b a s e o n t h e expectedu t i l i t y t h e o r y ( a conventionalexpectedutilitycurveisplotted inFigure1 below). j j

Figure1presentsthatthewealthispositiverelatedtoitsutilityasagreaterwealthorassetp o s i t i o n p r o v i d e a h i g h e r u t i l i t y t o i n v e s t o r s However,t h e curvealsoe x h i b i t s diminis hingp r o p e r t y whichindicatesthemarginalutilityderivedfromtheincreaseinassetpositionwillri seataslowerpacethantheincreaseinassetposition.Theinterpretationisthatinvestorswillnotacc eptforriskyinvestmentwithoutadequatecompensationforits risk.

Theinvestmentdecisionbasedontheutilityfunctionaboveisrationalanddonotexposet o anyp sychologicalbiasesasthedecisionissolelydependonthecalculationdifferentpossiblecombination ofassetportfolio inEquation1:

With x 1, x 2 ,…x n arethe possibleasset position in acombination

Employingtheconceptofriskaversionintoportfolioselectionprocess,rationalinvestorsw o u l d prefertoselectinvestmentthatofferhigherexpectedlevelofreturnforthesamelevelofr i s k orlow erriskforthesamelevelofexpectedreturnintotheirportfolio.Theexpectedvalueandriskforaportfolioisca lculatedinamannerofEquation2and3belowwhichisamathematicalequationformeanandvarianceof2 twoassets:

Fromtheaboveequation,portfolioreturniscalculatedasweightedaverageofthereturno f its components,whereastheriskoftheportfolioiscomputedfromthestandarddeviationofhistoricalret urnofeachcomponentwhichislessthanweightedaverageofthestandarddeviationo f itscomponents.It isduetothe returnofeachcomponentintheportfolioisrarelyperfectlycorrelated,therefore,thesp ecificriskoftheportfolioissuccessfullydiversified.AsEquation3s h o w n , thelowertheco rrelationefficientbetweeneachpairofcomponentintheportfolio,theloweristhestandarddeviati onandvarianceoftheportfolio.Therefore,theportfolioriskdoesnotr i s e asthesameratioastheincrease inexpectedreturnofportfoliowhennewassetsiscombinedi n t o t h e portfolio.

Incorporatingthediscussedtheory,Markowitz(1952)derivedtheefficientfrontierofriskyassetfro moptimizationofmeanvariance(theobjectiveistomaximizetheexpectedreturnateverylevelofvariancefro mpossiblecombinationofassets).Theassetcombinationlaidontheefficientfrontierrepresentstheefficie ntmean– variancecombinationofassets,whichmeansthecombinationprovidehighestexpectedreturnforaspe cificlevelofriskandispreferredbyrisk–averseinvestors.

However,t h e Markowitzefficientfrontieronlycompriseso f r i s k y a s s e t s Ino r d e r t o m anagetheriskmoreeffective,investorscanallocateafractionoftheircapitalintoassetsthatp r o v i d e certaintyreturn(riskfreeassets).CommonriskfreeassetsareTreasurybillandbondwithv i r t u a l l y zero probabilityofdefault.

Considertheefficientmean– varianceportfolioAandBinFigure2,allthecombinationbetweenriskfreeassetandportfolioAcanbede pictedbyCapitalAllocationLineforA(CALA)whichisconnectedfromRiskFreepointtoAontheEfficient

(CALB)iscreatedinthesamemanner.CALAoutperformsCALBs i n c eanycombinationbetweenportfolioAa ndriskfreeassetsprovideahigherexpectedreturnthananycombinationbetweenportfolioBandriskfr eeassetsgiventhesamelevelofrisk.Therefore,CALAw o u l dbepreferredbyriskaverseinvestorsthanCALB

.Usingthesameapproachandkeepmovingupwardalongt h e E f f i c i e n t Frontier,t h e o p t i m u m Capi talAllocationLinec a n b e reachedbytangentlinebetweenCALofMandEfficientFrontierCurve.Th isoptimumCALiscalledtheCapitalMarketLine(CML)whichdeliversthehighestexpectedreturnf oragivenlevelofriskandlowestrisk foranygivenlevelofexpectedreturn.

Equation4showsthemathematicalcalculationofCapitalMarketLinethatexpectedreturno n an EfficientFrontierPortfolioisequalthereturnonriskfreeasset(Rf)plustheMarketRiskPremium(E(R

TheCapitalAssetPricingModel

MarketPortfolioTheoryprovideriskaverseinvestorsafoundationforunderstandingaboutassetallocatio ninanEfficientMarket.However,thesetheorydidnotprovideamechanismforpricingassetsorportf olioinanefficientmarket.Therefore,CapitalAssetPricingModel(CAPM), asanextension of MPT, provides risk averseinvestors asinglefactorlinearmodelforcomputingt h e equilibriumrateo f assetsreturni n anefficientmarket.T h e m o d e l i s developedbyt h r e e researchers:Sharpe(1964),Lintner(1965),Mossin(1966),eachresear chercontributei n d e p e n d e n t l y to CAPMmodel.

Thebasicassumptiono f C A P M i s t h a t unsystematicr i s k (firms p e c i f i c r i s k ) canb e diversifiedaway,therefore,t h e o n l y systematicr i s k (marketr i s k ) i s requiredt o endurebyi n v e s t o r s Furthermore,thesystematicriskforagivenassetcanbemeasuredbyitscovariance w i t h marketp o r t f o l i o (i,M).

Substituting(i,M)intoEquation4,arelationshipbetweenriskandexpectedreturncanb e deri vedasEquation 5 below:

Bydefiningβii(Betaofasseti)=i,M/M 2,Equation5canberewrittenas:

Equation6istermedtheSecurityMarketLine(SML)equation.Equation6basicallysaidt h a t , inequilibriumcondition,theexpectedreturnonportfolioi(E(Ri))isequaltoreturnofRiskfreeasset(Rf) plusthemarketriskpremium(E(RM)–Rf)inproportionto itssystematicrisk(βii).A s aresult,itcanbeinterpretedthatassetswithhigherβii(highersystematicrisk)mus tprovideahigherreturntoinvestorsforenduring ahigherriskinvestment.A systematic risk – expectedreturnrelationshipof SMLisdescribedinFigure3:

Whenthemarketisinequilibrium,anefficientportfolioshouldbegraphedonSMLlinewhic hdeliversareturnappropriatetotheirsystematicrisk.Anundervaluedassetisdrawnabovet h e SMLli neasitdeliversahigherreturnthanexpectedfromthemarketincorrespondingtoitssystematicrisk.On contrary,anovervaluedassetissketchedbelowtheSMLlineasitdeliversalowerreturnthan expected incorrespondingto itssystematicrisk.

AlthoughtheCAPMmodelwasbuiltonasolidfoundationwithitsfundamentaleconomictheory,t h e r e a r e n u m b e r o f drawbacksfromC A P M m o d e l Firstly,R o l l (1977)critiquedt h a t marketin CAPMtheorydoesnotonlyconsistonEquityMarket,wheremostofEmpiricalTestonC A P M iscarriedo ut,butaPortfolioofallwealth.AmarketPortfolio,accordingtoRoll(1977),s h o u l d comprisesbon ds,stocks,properties,humancapitalandanyotherassetsthatcontributetot h e wealthofmankind.From thisviewpoint,hefurtherassertedthattheCAPMcanonlybetestedwhenthetrue marketportfolio is knownwithcertainty.AndtestsofCAPMis onlyatest ofmean

– varianceefficiencyofthePortfoliothatisconsideredasmarketproxy.Due tothis reason,af i n d i n g ofamean– varianceefficiency Portfoliowithin anysamplecannot informresearcherswhetherCAPMmodeli scorrector not.

Moreover,apartfromitsweakness,thereareincreasingevidencesthatBetaisnottheonlyr i s k fac toraffectingstockreturns.Therearemanyotherfactorswhichcanbenamed,suchasbookt o marketratio(F ama&French,1992),companysize(Banz,1981),price/ earningsratio(Basu,1 9 7 7 )anda n u m b e r o f o t h e r systematicfactors.Thesefactorsw i l l b e d i s c u s s e d i n detailsi n followingsections.

ASSETRETURNSLITERATURE

TheoreticalvsDataMiningResearch

Following the identification of potential flaws in the Capital Asset Pricing Model (CAPM), researchers have sought additional factors to explain asset returns more effectively These models refine CAPM by incorporating various explanatory factors that better capture asset price movements Notably, factors influencing stock returns, aside from the Beta factor of CAPM, are typically referred to as anomalous variables or anomalies While some models, such as the Fama and French (1993) Three-Factor Model and the Carhart (1997) model, have demonstrated enhanced explanatory power for observed asset returns, they have often been criticized as merely fitting data rather than being grounded in theoretical foundations Black (1998) notably evaluated the Fama and French Three-Factor Model in this context.

1993)as“datamining”becauseconductingaresearchwithsizable ofp o t e n t i a l explanatoriesva riablesshouldproducesomepositive result.

Furthermore,Subrahmanyam(2010)expressedthat:“thetendencyofscholarstouseonem e t h o d o l o g y o r t h e o t h e r s r a i s e t h e q u e s t i o n ofwhethert h e resultsa r e r o b u s t t o d i f f e r e n t methodology”.

Asaresult,aquestionthatmanypeoplemightconcerniswhyacademicinFinanceweres o com mittedto“datamining”.Therearesome convincingexplanationsforthisquestion.Firstly,D e m p s e y (2013)revealedthat:“agooddealoffina nceisnowaneconometricexerciseinminingdataeitherfor confirmationofaparticularfactormodelorfor theconfirmationofdeviationfromt h e model’spredictionasanomalies”.FromtheargumentofDemp sey(2013),thepersistentuseo f “datamining” originatedfromthedesireto support thescholar’spriorbelief.

26November2010)gaveusanotherpersuasivereasonf o r highl e v e l d a t a m i n i n g t h a t s t e m f r o m t h e e s t a b l i s h m e n t o f C R S P database.Iti s estimatedthatmorethanone– thirdofpublishedresearcheshavebeenconductedbyusingdatafromC R S P database.T h e databaseh asgivenfinancialscholarsane n o r m o u s o p p o r t u n i t y toperformtheireconometric and“data mining”research.

Fama–FrenchThreeFactorsModel

Fromaboveliterature,wewouldacceptedexplanatoryfactorsarenotalwaysderivedfromtheoreti calbackground,but itmightbederivedfromdifferentresearcheson aparticulardatabaseo r “datamining” activities.Therefore,itishardtoclaimwhatAssetPricingMod elisthebestm o d e l AmoresuitablequestionwouldbetoaskwhatModelandExplanatoryVariab lesshouldb e themostappropriateforaparticularresearch.ThreerenownedAssetPricingModel willbediscussedbelowwhichwereconstructedd u r i n g a searchbymanyscholarst o exploreo t h e r significantexplanatoryfactorsforassetreturn.

Twofamousresearchersi n Finance,EugeneFamaandKenFrenchh a v e performede x t e n s i v e researchesaboutAssetPricing.In1993,theycameupwithaconclusionthatapartfromt h e market riskpremium(Beta),“value”factorand“size”factorscanexplainasignificantportiono f assetreturns.Ino rdertoaccountforthesefactors,SMBisconstructedtorepresent“size”factorandHMLisconstructedtoadd ress“value”factor.

The SMB variable, short for Small Minus Big, represents the excess return investors can earn by investing in small-cap companies instead of large-cap firms Similarly, the HML variable, or High Minus Low, captures excess returns from investing in companies with high book-to-market ratios compared to those with low ratios Although these factors lack a robust theoretical foundation, they offer valuable insights into their significance The SMB factor suggests that small companies are more susceptible to risks due to their size and lack of diversification, leading investors to seek higher returns as compensation for this additional risk Conversely, the HML factor indicates that companies with high book-to-market ratios are perceived to have diminished value in the market, making them more vulnerable to risks and yielding lower returns compared to those with lower ratios Consequently, investors demand higher returns as a premium for the extra risk associated with high B/M ratio stocks.

R it –R Ft =a i +b i (R Mt –R Ft )+s i SMB t +h i HML t +e it (7) b iis a coefficientof relationshipbetweenasset returnandmarketriskpremium. s Ai s a coefficientofrelationshipbetweenassetreturnandsizefactor h Ai s a coefficientofrelationshipbetweenassetreturnandvaluefactor.

TheModelcanbeconsideredasane x p a n s i o n ofC A P M m o d e l s i n c e i t combinest h e trad itionalmarketriskfactorwithtwonewfactors.Thismodelexplainedthereturnofassetsincorresp ondencewith threeriskfactors:marketriskpremium,sizeriskandvaluerisk.

Theeffectivenesso f F a m a andFrenchT h r e e factorsm o d e l h a s b e e n t e s t e d bymanye mpiricalstudies indifferentmarket.Most of thesestudiesprovide positivefeedbackstoward them o d e l

In 1997, a study was conducted on the Hong Kong Stock Exchange to examine the relationship between stock returns and various factors such as beta (βi), size, leverage, book-to-market equity ratio, and earnings/price ratio, following the Fama and French approach The findings indicated that beta did not significantly capture the average monthly returns of stocks listed from July 1984 to June 1997 However, size, book-to-market ratio, and earnings/price ratios were found to explain a substantial portion of the cross-sectional variation in average monthly returns Additionally, the study concluded that while book leverage and market factors have explanatory power over cross-sectional return variation, they become redundant when the aforementioned three factors are included in the model These results remained consistent across different size groups and months, indicating that extreme observations and abnormal return behaviors did not affect the findings.

Ajili(2003)conductedatestbetweenFamaandFrenchthreefactorsandtheCharacteristicm o d e l ofDanielandTitman (1997)o nFrenchstock marketin1976 –

2002 period.Stock aresortedintodifferentportfoliosofsizeandbook/ marketratiofollowingFamaandFrenchmethodology.S o m e noticeableinferencesf r o m t h i s research a r e th at 9 0 5 % on averageof th e expectedreturni s explainedbyt h e p o w e r o f 6 s i z e andB/ Mratioportfolios.Moreover,t h e prediction,thattheinterceptsinthe regressionofcharacteristicbala ncedportfolioofFFThreefactorsmodelshouldbezero,issupportbythestudyovertheCharacteristicmo delofDanielandTitman(1997) whichpredicted that theseintercepts should benegative.

InastudybyDrew(2003),heconductedatestbetweenCAPMmodelandFFThreefactorm o d e l f orHongKong,Korea,MalaysiaandPhilippineduringthe1990–

1999period.Hefoundt h a t CAPMmodelalonewerenotcapableofexplainingcrosssectionalofstoc kreturnsinthesecountriesandt h e a b s o l u t e p r i c i n g errorso f t h e C A P M i s largeri n c o m p a r i s o n w i t h FFT h r e e factorsmodel.Overall,theresearchconcludedthatsizeandbooktomarketratiofactor sdidplaya significantrolein capturingthe movement ofcrosssectionalstockreturns

O'Brien, Brailsford, and Gaunt (2008) conducted a comprehensive study on the value premium in the Australian stock market, analyzing data from 98% of Australian listed firms over a 25-year period Prior to their research, the "size effect" was acknowledged in Australia, but limited data hindered the analysis of the "value effect." In a subsequent study, Brailsford, Gaunt, and O'Brien (2012) highlighted that, for the first time in Australia, their constructed factors revealed a significantly positive price premium using both time series and cross-sectional tests They also noted that the Fama-French three-factor model had superior explanatory power compared to the Capital Asset Pricing Model (CAPM), accounting for nearly 70% of the movement in asset returns.

CarhartFourFactorsModel

ThefourfactorsmodelwasputforwardbyCarhart(1997)asaresultfrommomentumeffectpr esentedbyJegadeeshandTitman(1993).Themomentumeffectcouldnotbeexplainedbythefamous ThreeFactorModelofFamaandFrench.

InastudybyJegadeeshandTitman(1993),portfolioswerecreatedbyusinganassumptionthatthesto ckshasmadegain(orloss)inthepastwillcontinuetomakegain(orloss)inthefuture.T h e timeframeformo mentumeffectfluctuatefrom3–12months.Thefourfactormodelwast h e n testedbyCarhartin 1997.ThismodelwascreatedbyaddingMomentumfactorofJegadeeshandTitman(1993)totheThreefa ctorsModelofFama–French.TheformationofModelisasbelow:

TheonlyfactorwhichwasaddedisPR1YR(representMomentum)calculatedbytakingt h e differencebetweenhighest30%returnofprior12monthsandlowest30%returnofprior12m o n t h s ThetestwascarriedoutbyCarhartandsuccessfullyconfirmedthevalidityofMomentumfactorin explainingstockreturns.

FourfactorsmodelofCarhartwastestedinmanycountriesandalsoyieldedpositiveresultstowardth e m o d e l L’Her,Masmoudi,andSuret(2004)conducteda Fourfactorm o d e l t e s t o n Canadianstockm arketover1960–

The 2001 study utilized the FF methodology to identify risk factors in the Canadian market, revealing that the "size" factor is significantly higher in January compared to other months Additionally, the "momentum" factor consistently demonstrates meaningful results throughout the year, with January being the only exception The book-to-market ratio shows significant value in downtrend markets, while its explanatory power diminishes in uptrend scenarios The research also takes into account the macroeconomic environment, particularly the impact of monetary policy, finding that during periods of expansive monetary policy, both the "size" and "value" premiums are highly significant.

BennaceurandChaibi(2007)performedatestonTunisianstockexchangewithdifferentm o d e l fromC A P M t o FFT h r e e factorm o d e l andCarhart4 factorsm o d e l i n ordert o teste x p l a n a t o r y poweroftheseassetpricingmodels.Theperformanceofeachmodelisevaluatedbyu s i n g Sc hwartzandAkaikeindicators.Theresultfromthemodelsuggestedthatthemomentumpremiumisth ehighestpremium(8.8%)inTunisianStockExchange(TSE).Thevaluepremium

 is2.88%andrelativelysmallerthanotherpremium.Surprisingly,thesizepremiuminthispaperhasa n egativev a l u e (-3.4%)w h i c h disagreeswith otherliteratureswherethes m a l l firmc o m m o n l y deliverhigherreturn.Overall,thepaperclaime dthatthebestAssetPricingModelforTunisianStockMarketistheCarhartfourfactormodelwhenallth eselectingcriteriaguidedtot h e sameconclusion.

K.S.Lam,Li,andSo(2010)providedanotherstudysupportingthe4factorsmodel.Thes t u d y wasconductedinHongKongstockmarketduring1981–

2001period.Aftertestingthefourfactorsm o d e l w i t h datafromH o n g Kongstock market,allf o u r fact orswerefoundedt o havesignificante x p l a n a t o r y power.Moreover,t h e interceptalsowasu n v e i l e d t o b e insignificantdifferentfromzero.AfurthersupportiveevidencefortheCarhartfourfactormodel whenvalueofadjustedR 2i s relativelyhighandt h e r e s i d u a l standarddeviationofadditionale x p l a n a t o r y variablesisinsignificantwhichmeanalargeportioninstockreturnsvariationhadbeens u c c e s s f u l l y explainedbyCarhartmodel.

Fama–FrenchFiveFactorsModel

Manyresearchers,whocompletedtheirempiricalresearchesindifferentmarket,claimedt h a t FamaFrenchThreeFactorsModel(TFM)indeedprovideexplanatorypowerofaveragestockreturns.How ever,theexplanatorypoweroftheTFMisstillweaksinceaportionofaveragereturnsleftunexplained.Asaresu lt,FamaandFrenchpersistentlyputtheireffortindevelopingabetterassetp r i c i n g m o d e l D u r i n g t h e i r research,p r o f i t a b i l i t y andinvestmentfactorsaroseasnewe x p l a n a t o r y factorsw ithsupportingtheoreticalframeworkfromdividenddiscountmodel Additionally,therearenum berofempiricalstudieswhichhave beenconductedandgavepositiveresultofthesetwo factors.

Initially,theDividendDiscountModelclaimedthatthemarketvalueofastockisasumo f disco untedvaluestream ofexpecteddividendsforeachshare.

M tdenoted sharepriceattime t ris thelongtermexpectedstockreturnorinternalrateofreturn ofexpecteddividendsE(d t+)istheexpecteddividendpersharefortime(t+).

Fromequation(9),wecaninferthatiftwodifferentstockswhichofferinvestorsthesameexpected dividendsbuttheirsharepricesaredifferent,thelowerprice’sstockwouldberequired

 ahigherexpectedreturnbyinvestors.Orwecaninferthatthestockwithlowerpricewouldhavehigherrisk oruncertaintyinpayingouttheirexpecteddividendthantheotherstockdespitehavinga sameamountofexpec teddividend.

MillerandModigliani(1961)furtherdevelopedequation(9)andpresentedequation(10)whic hshows the totalmarketvalueofafirm’sstockataparticulartime t:

With Yt+istotalequityearningsforperiodt+ dBt+=Bt+-Bt+-1isthechangeintotalbookvaluebetweenperiodt+andt+-

Mt)andexpectedstockreturn(r).WecaninferthatalowervalueofM t stock(orahigherb o o k tomarkete quity B/Mratio stock)impliesahigherexpectedreturn.

Secondly,i f eachterminequation( 1 1 ) i s keptunchangedexceptf o r e x p e c t e d futureearni ngs(Yt+)andexpectedstockreturn(r).Wewouldinterpretthatahigherexpectedearnings(profitabil ity)impliesa higherexpectedreturn.

(Bt,MtandYt+)exceptforexpectedgrowthi n bookequity(dBt+)andexpectedreturn(r).Thiswillprovideu swithaninterpretationthatah i g h e r growth inexpectedbookequity(investment)impliesalowerexpectedreturn.

Throughabovec o n c l u s i o n s fromequation(11),w e cano b s e r v e t h e r e aretheoretical relationshipsbetweenbookstomarketequityratio,expectedearnings(profitability),growthinb o o k e q u i t y (investment)ands t o c k s ’ expectedreturn.A s a result,t h e s e conclusionsa r e t h e pre misesfromwhich theFamaandFrench(2013)FiveFactorsModel (FFM) isdeveloped.

R it –R Ft =a i +b i (R Mt –R Ft )+s i SMB t +h i HML t +r i RMW t +c i CMA+e it (12)

Ritisthereturnon stockorportfolioiforperiodt.RFti s t h eriskfree return

RMtis th ereturnonvalueweightmarketportfolio s ii s a coefficientofrelationshipbetweenasset returnandsizefactor. h ii s a coefficientofrelationshipbetweenasset returnandvaluefactor r ii s a coefficientofrelationshipbetweenasset returnandp r o f i t a b i l i t y factor. c ii s a coefficientofrelationshipbetweenasset returnandinvestment.

FromtheinterpretationofMillerModiglianivaluationequation,FamaandFrenche x p l a i n e d t h e o r e t i c a l l y therea r e relationshipsbetweenf u t u r e s t o c k r e t u r n s , currentB/

M,firmlevel,expectedprofitabilityandfirmlevelinvestment.Theyalsoperformedanempiricaltesto nt h e s e relationships.Aharoni,Grundy,andZeng(2013)studiedanempiricaltestforthedatadrawnfro mCenterf or ResearchinSecurity Prices(CRSP) forperiodof1 96 3 - 2 0 0 9 Aharonietal. (2013)claimedt h a t , w h e n variablesa r e m e a s u r e d att h e firml e v e l , t h e expectedr e l a t i o n s h i p betweenthesevariablesholdtrue.Therationalewasthatsharelevelmeasurementwillbeaffectedbyissu anceandrepurchasewhichwillalterthecalculationofsharegrowth.Asaresult,thefirm

– levelgrowthandpersharegrowthcanbesignificantlydifferent.Atthefirmlevel,Aharonietal.(2013) confirmedfollowingrelationshipsbetweenvariables:apositiverelationshipbetweenB/

Mandstockreturns,apositiverelationshipbetweenexpectedprofitabilityandreturns,andanotablynegati ve relationshipbetweenexpectedinvestment andreturn.

In a study by Titman, Wei, and Xie (2004), it was found that there is a negative relationship between abnormal capital investment and stock returns The researchers argued that while increased investment can signal better opportunities, it may also indicate that management is prone to overinvestment They noted that investors often underestimate the negative implications of high investment levels, leading to a decline in stock returns Furthermore, firms with high cash flows or low debt levels are particularly susceptible to the issues of overinvestment and low returns.

Marx(2013)),h e alsoc o n f i r m e d therei s a p o s i t i v e relationshipbetweenprofitabilityandcross sectionofaveragereturns.Heconcludedthatprofitability,calculatedbygrossprofitstoitsassets,hassi milarexplanatorypowertoB/

Mratioinpredictingt h e abnormalaveragereturns.Healsorealizedthathighprofitablefirmshadmuc hhigherstock returnsthanunprofitablefirmsdespitethesestockshavearemarkablyhigherB/

Mandprofitability,investorscangainadvantagesofcapturingtheprofitabilitypremiumwithout withstandinganyadditionalr i s k T h e mechanism,throughwhichp r o f i t a b i l i t y influencea veragereturns,representsconsiderationofinvestorsaboutfirmvalue.While“valuestrategy”investorsg aintheirabnormalreturnsthroughpurchasinginexpensiveassets(highB/

Mratiostocks),“profitabilitystrategy”investorsreceivetheirabnormalreturnsbybuyingproductivea s s e t (highp r o f i t a b i l i t y s t o c k s ) a n d sellingu n p r o d u c t i v e a s s e t s ( l o w p r o f i t a b i l i t y stocks).

LIQUIDITYLITERATURE

TransactionCost Theory

AmihudandMendelson(1986)w e r e t h e firstt o introducet hi s m o d e l i n whicht r a d i n g diff erentassetsincludenumbersofelements,suchastransactioncostsandtradingfrequencies.T h e modellaterwasfurtherdevelopedbyKane(1994).

Assumetherewasamarketwiththreerisk- freeassetsandeachassetonlyhastwooutstandingshares.EachassethasitsownexpectedreturnR i ,i 1,2,3.ThetransactioncostsfortradingtheseassetsareS i ,i=1,2,3(assumingS 3 -S 2=S2 -

S 1= S>0andS 1=0) Themarkethast w o typesofinvestors,eachtypeofinvestorshavethreepeopleandeachty peofinvestorshasdifferenttradingintervalorfrequency.

Inthisframework,liquiditypremium(r i )isdefinedasthedifferencebetweenreturnsofassets withpositivetrading expensesandthereturnoftheassetwithzerotradingcost,whichis, r i

R 1,i=1,2,3 (asweassumedthatthetradingexpenseofasset1=0).Hence,fortype1investors,byholdingassetI,inve storswillexpecttoreceive(R 1 +r i )-

Anotherassumptionisthatinvestorsarerisk- neutralandselecttoretaintheassetwiththehighestnetexpectedreturn.Theequilibriuminthiseconomywi lldisplayaclienteleeffect,whichmean,investorswillnotselecttokeepallassets,buttheywillonlyretaint heassetswhichmatcht h e i r preferences.Iftwoassets1and2arekeptbytype1investors,itmustsi gnifythatthenetreturnsonasset1and2areequalandtheirexpectedreturnsarehigherthannetreturns onotherassets.Hence,

Switchinginequality(13.1) and(13.3),wehaver 1 -r 2= T 1(S1 -S 2 )andT 2(S1 -S 2 )>r 1 - r 2 C o m b i n i n g above inequalities,wehave (T 1 -T 2 )(S 1 -S 2 ) T 2(S2 -S 1 )andT 2(S3 -S 2 )=r 3 - r 2 D i v i d i n g theinequalitiesbythedifferenceintradingvolumeorcost,inequalitiesintermsofthe firstderivativearederiveasbelow: r 2 r 1 TURN 2 TUR

Accordingly,ast u r n o v e r increases,t h e l i q u i d i t y premiumdecreasesconcavely;i t i s alsoanin creasing andconcavefunction ofthetrading cost Therelationshipcanbeseen in thefollowingfigure:

( 2 ) t h e expectedl i q u i d i t y p r e m i u m r ii s a concaveandincreasingfunctionofitstradingexpe nses,and(3)theexpectedliquiditypremiumisaconcaveanddecreasingfunction ofitstradingturnover.

Someresearchershavebeencriticizedthetradingfrequencyhypothesisinacircumstancewhere aneconomyhasonlyonetypeofinvestorwithtradingfrequencyT,investorswillkeepallassetswith thesamenetreturns.Hence, r1–TS1=r2–TS2=r3–TS3=0 (16); and,ri=TSi, forallI=1,2,3

Withthisassumption,thetradingfrequencyisthesameforallassets,therewillbenocrosssectionalre lationshipbetweent u r n o v e r andt r a d i n g frequency.T h e c r o s s sectionall i q u i d i t y pre miumwillincreaselinearlywith thetransactioncost.

Two theories explain the relationship between turnover and liquidity premium through transaction costs The first, the information-based trading hypothesis (Easley, Kiefer, O'Hara, & Paperman, 1996), suggests that stocks with higher turnover have a lower probability of information-based trading, which in turn decreases transaction costs The second theory, the order-processing cost hypothesis, posits that in a fixed cost environment, higher turnover stocks lower the dealer's average costs, subsequently reducing investor costs As a result, lower costs lead to a decreased liquidity premium required by investors for their investments.

SubsequentDevelopmentinLiquidityresearch

Divergence between the effect of liquidity on stock returns:

AfterthetheorydevelopedbyAmihudandMendelson(1986),aneweraforresearchinl i q u i d i t y and assetreturnhasbegun.SomeresearchessubsequentlyfoundtobeinagreementwithresultfromAmihudand Mendelson(1986)’spaperthatliquiditynegativelyrelatedtoassetreturns.Belowa r e s o m e prominentres earcherst h a t p o s e a negative(positive)relationshipbetweenl i q u i d i t y (illiquidity)measure s andexcessstockreturns:

BrennanandSubrahmanyam(1996),usingpricechangesandorderflowsformeasuringilliq uidity,claimedt h a t t h e i r i l l i q u i d i t y measurements p r o v i d e a p o s i t i v e relationshipbetwe eni l l i q u i d i t y ands t o c k r e t u r n Theirr e s u l t hasb e e n verifiedandp r o v e d significant, evenaftercontrollingforFamaandFrench(1993)’sthreeriskfactors.InanotherpaperbyBrennaneta l.(1998),shareturnover is utilizedasaproxyforliquidityandthismeasuresuggestedthatliquidityandassetreturnshasanegativerelat ionship.

BakerandStein(2004)builtamodelthatpredictedanincreaseliquiditywillsubsequentlyleadt o a l owers t o c k r e t u r n s atb o t h firmlevelandaggregated a t a Int h i s m o d e l , n u m b e r o f l i q u i d i t y proxieswereusedsuchasbid- askspread,priceimpactoftradeorturnover.They allpointedtothesameconclusionofanegativerelati onshipbetweenliquidityandstockreturns.Ani n n o v a t i v e featureofthismodelisthatitincludesa classofirrationalinvestors,whounder-reactt o informationinorder- flow.Andhighliquidityisanindicationthatthemarketisoverwhelmedbyirrationalinvestors,thus it isover-valued.

InapaperofAmihud(2002),heemployedanilliquiditymeasure(thedailyratioofabsolutes t o c k return t o dollarvolume).T h i s paperrevealedt h a t , overa l o n g timeframe,t h e expectedmarketi l l i q u i d i t y p o s i t i v e l y influencest h e s t o c k excessr e t u r n T h i s suggestst h a t t h e r e i s a negativ einfluenceofliquidityonstockexcessreturnoveralongtimeframe.However,hefurtherstressedthat,sto ckreturnsarenegativelycorrelatedwithtemporaryunexpectedilliquidity( l i q u i d i t y shock).

Incontrarytoaboveresearchpaper,thereareexistingempiricalstudiesin stockmarketsw h i c h showedacontraryrelationship.Inarecentstudy byBalietal.

(2014) f o r allcommons t o c k s tradedonNYSE,Nasdaqfrom1963to2010,theyconcludedther eisanegativerelationbetweenfirmlevelilliquidityandstockreturns.Theirdecileportfoliosthatbuynegative illiquiditys t o c k s andsellpositiveilliquiditystocks provide arisk-adjustedreturn largerthan1%per month.T h e resultissignificantevenaftertestingwithdifferentmeasuresofliquidityandcontrollingfo rdifferentr i s k factorsaswellasfirmcharacteristics.Theirexplanationf o r t h e resultsi s thati n v e s t o r s underreacttoilliquidityleveloffirms.Thisunder- reactiontendstobestrongerforstockshavinglessattentionfromgeneralpublic.

Exchangeatm o n t h l y levelfromMarch2002toNovember2011.Theyshowedthatilliquidityisan importantfactorinexplainingtheassetreturnsanditnegativelyimpactsonassetreturns.Theauthors alsosuggestedthat illiquidi ty measurescanpartly coverthemarketfactorseffect.Inaddition,the ystatedthatfour factormodels (market,illiquidity,sizeandvaluefactors)isthe bestmodelsto e x p l a i n assetreturnsandthe momentumfactorisnotpricedinIranstockmarket.

Withb o t h p o s i t i v e andn e g a t i v e evidenceso n t h e r o l e o f l i q u i d i t y i n e x p l a i n i n g s t o c k returns,thereseemtobeanambiguousconclusionabouttheeffectofliquidityonstocke xcessreturns. b)

Liquidity effect in emerging market:

AresearchbyHarvey(1995),herevealedthatthecorrelationbetweenemergingmarketequitie sandothermarketsislow.Asaresult,itsignificantlyreducesunconditionalportfoliorisko f aworldinv estors Additionally,t h e standard globalassetpricingmodel, whichadopta fu ll integrationofgl obalcapitalmarket,cannotexplaintheaveragereturns inemergingcountries.Healsoconcludedthattheequityreturnsin emergingmarkettendto be controlbylocalinformation.

Witht h e datafrom2 7 emergingmarketsfromJ a n u a r y 1 9 9 2 t o December1 9 9 9 , Jun,Mar athe,andShawky(2003)indicatedaggregatemarketliquidityin emergingmarketspositivelyaffectedonstockreturns.Turnoverratio,tradingvalueandturnover– volatilitymultipleare usedasliquiditymeasuresin theirstudy.Importantly,theyalsoconcluded thatemergingmarketshavelowerintegrationlevel withglobalcapitalmarket.

In 2007, researchers identified emerging markets as ideal environments to study the impact of liquidity on stock excess returns They noted that these markets typically exhibit relatively poor liquidity compared to developed markets, which has deterred many institutional investors from allocating funds to various stocks This situation exacerbates the disparity between liquid and illiquid stocks Additionally, many emerging markets experienced market liberalization during the research period, providing a unique opportunity to analyze the significance of liquidity on expected returns, especially since liquidity often improves significantly post-liberalization In contrast, developed markets usually feature a diversified ownership structure that includes both long-term and short-term investors.

(2007)suggesttheclienteleeffectsinselectinginvestmentportfolioshouldreducethepricingofliquidit y.Inrespectofemergingmarkets,thediversificationinnumberofsecuritiesandownershipi s lackedofwhic hoftenintensifythe liquidityeffects.

Withallaboveresearchesandarguments,w e c a n expectt h a t thel i q u i d i t y researchi n fronti ermarketshouldyieldaveryconvincingresult.Thefactthatliquiditygapbetweenliquidandilliqu idstocks,whichisvery significantinemergingmarket(evenmoresevereinfrontiermarket), givesresearchersarealopportunitytodiscovertheeffectofliquidityonstockreturns.Inaddition,therea r e v e r y l i m i t e d n u m b e r o f liquidityresearchesconductedi n frontiermarkets(pre- emergingmarkets).Mostliquidityresearcheswereconductedindevelopedmarket(Bekaertetal.,2007). Therefore,thereisaneedforfurtherliquidityresearchinfrontiermarketenvironmenti n ordertoclearlyun derstandabouttheliquidityeffectindifferentenvironmentsandeconomicconditions.

Vietnamisaninterestingcaseforliquidityresearchforsomereasons.Firstly,liquidityinVietna mmarkethasincreasegraduallyyearafteryearduetomoreliberalizingpoliciesfromthegovernment.T hisprovidesopportunitytostudytheeffectofimprovedliquidityonstockreturns.Moreover,t h e l i q u i d i t y gapbetweenl i q u i d andi l l i q u i d s t o c k i n Vietnami s quitelarged ue to foreigninvestorsofteninve stinlargeandreputablestocks.Thus,observedliquidityinotherstocks areoftennoticeablylessthanthoseliquidstocks.Lastbutnotleast,Vietnamequitymarketadoptedmatchingorder system,whichdonotallowcommonliquiditymeasuressuchasbidaskspread,z e r o dailyreturn,etc

… tobeemployedinVietnamequitymarket.Therefore,liquidityresearchinVietnamequitymarketwillinv olvetheuseofdifferentliquiditymeasurestoyieldfinalconclusion.T h i s s h o u l d enableresearchers t o t e s t whetherd i f f e r e n t l i q u i d i t y measuresgivea convincing conclusionabout liquidityeffectin thecaseofVietnam.

Fromabovearguments,therei s ano b v i o u s needt o s t u d y aboutt h e liquidityeffecti n Vietn ammarketforsomereasons.T h e firstreasonisthatthereisalackofliteratureaboutliquidityeffectinVietnam,afrontiermarket.Moreover,thereisarealneed fromvarioussocialparties(investors,p o l i c y makers,c ompanies,etc…)t o thoroughlyunderstand h o w l i q u i d i t y influence assetreturns.Sothat,each partycanhavetherightdecisionsfortheirparticularinterestswhichw i l l benefitthesocietyatawho le.

MAINHYPOTHESIS

Froma b o v e literaturea b o u t l i q u i d i t y ands t o c k return,m o s t researchersagreedt h a t l i q u i d i t y playsarolein explainingstockreturn.However,thereis still adebateaboutwhethertheeffectis positive ornegativesign.

Inthecaseofthisresearch,theresearchwascarriedoutinafrontiermarket,particularlyi n Ho ChiMinhCitystockexchange.Someconsiderationswasputforwardinordertoestablisht h e hypothes isforthisstudy.ThefirstconsiderationwasHo ChiMinhCitystockexchange wasanewandrelativesmallcapitalmarket.Thus,itpreventedlargeforeigninstitutionalinvestorstoi n v e s t theirfundinmostofcompanystocks(Bekaertetal.,2007).Asaresult,highliquiditystocks,whichoftens electedbyinstitutionalinvestorsforinvestment,commonlyhavegoodfundamentalandrepresentlowris kinvestment.

Secondly,illiquidstocksusuallyholdbyonlyalimitedinvestors.Hence,theyareriskert h a n l i q u i d stocksastheriskfromilliquid stocksarenotsharedbymanyinvestors.

ThethirdconsiderationwasamajorityofinvestorsinVietnamis smallandinexperiencedi n v e s t o r s Andinexperiencedinvestorsaremoreinclined toparticipateintoover-valuestocksandtrend- chasingbehavior(Greenwood&Nagel,2009).Thismeanstheytendtoinvestintostockswhichquick lyincreaseinpriceandmaycreatebubble.Andanimportantconditionforstockstoq u i c k l y riseisthat itsliquidityshouldbelimitedandadduponlysmallportionsoftotaloutstandingstocks.Incaseofhighliquidityst ocks,theyrarelycansoarrapidlyinpriceorcreatebubbledue tovoluminous awaitingsellersinthemarket(include biginvestmentfirms) whenthesesellers observeabubbleinassetprice.

Fromthese considerations,wecaninfer that illiquidstocks oftenh a v e higherr is k than l i q u i d stocks.Andinvestorswilldemandapremiumforbuyingilliquidstocksasacompensationf o r t akingextrarisk.Additionally,illiquidstocks’pricescanshiftupanddownquicklyduetotheiri l l i q u i d i t y na ture.Thismeansilliquidstock canremaina significantriseinpricewhichhardlyoccurin liquid stocks.

As aresultof abovearguments, themainhypothesisforthis thesis is:

Thischapterwill bededicatedtodataandresearchmethodologyof thisthesis.Regressionm o d e l s andresearchframeworkofthisthesisare firstlypresentedtogiv ereadersanoverviewaboutvariablesandtheirlinkage.Subsequently,Datasectionispresentedtoshowdat asourceandi t s reliability.T h e u p c o m i n g sectionw i l l discussaboutt h e selectionofFamaand MacBethregression.Afterregressionm e t h o d i s discussed,allvariablesandt h e i r calculationw i l l b e m e n t i o n e d toprovideaclearpictureabouttheprocedureofconstructingpremiumfact ors.Thischapterwillendwiththepresentationofpotentialeconometricissuesandhowtheyaresolvedin t h e thesis.

REGRESSIONMODELANDRESEARCHFRAMEWORK

RegressionModels

AlthoughFamaandMacBethregressionisacomplexmethod,thisthesiswillemployFamaandMacB eth(1973)regressiont o t e s t t h e empiricalrelationshipbetweenl i q u i d i t y ands t o c k returns.Thisr egressionmethodisconsideredasoneofthemosteffectiveapproachforempiricalresearchaboutstockre turns.ThedetailaboutthisapproachispresentedinAppendixsectionforinterestedreader.

Inliteraturereviewpart,someofthemostprominentpremiumfactorsintheassetreturnshavebee nintroduced(FamaFrenchfivefactors,momentumfactor).Thesefactorsfrequentlyusedbygovernmentala gencies,institutionalinvestorstocalculatetherequiredreturnfortheirinvestment.Theyoften provedeffectiveandapplied world-wideforassetvaluationpurpose.

Themainobjectiveofthisstudyistoverifytheroleofliquidityinexplainingassetreturns.Therefore,all well- knownpremiumfactorsinassetreturns’literaturearecombinedintoregressionm o d e l s Thesefactor swillrepresentcontrolvariablesinthisresearch.Sothat,anyliquidityeffectt h a t canbemeasuredisthetru eeffectof liquidityonassetreturns.

Thisthesiswill involve theuseoftwo proxies ofliquidity (turnoverratio measurea ndAmihudilliquiditym e a s u r e ) fortwoseparatedmodels.Therearemanyavailableliquidityme asuresintheliterature.Nonetheless,thesetwomeasuresshowedtheyarethebestcandidatesf o r l i q u i d i t y researchi n t h e c a s e o f Vietnammarket.A l l t h e controlvariablesw i l l b e keptunchan gedbetweenthetwo models.Insummary,two modelsaredefinedas follow:

RETURNi,t =α0+α1TURNOVERi,t+α2RISKi,t+α3SMBi,t+α4HMLi,t+α5CMAi,t+ α6RMWi,t+α7RET23i,t+α8RET46i,t+α9RET712i,t+εi,t

RETURNi,t =α0+α1AMIHUDi,t+α2RISKi,t+α3SMBi,t+α4HMLi,t+α5CMAi,t+ α6RMWi,t+α7RET23i,t+α8RET46i,t+α9RET712i,t+εi,ti : cross –listedfirms,t:time periodfrom2008 to2013.

RETURNi,t :Excessstockreturnsonportfolioiformontht,itiscalculatedbyusingaveraged ailyreturn on portfolio i minus riskfreeratereturnforthesameperiod(R it–

TURNOVER :Turnovermeasure,it iscalculatedasthedifference betweenaverages t o c k returnofhighandlowTurnoverportfolios. AMIHUD : Amihud illiquidityratiomeasure,it iscalculatedasthedifferencebetweenaveragereturn ofhighand low Amihud illiquidityportfolios.

RISK :Marketriskpremium,it iscalculatedbyusingaveragedailyreturn onmarketportfolio (VNindex) minus riskfree ratereturnforthesameperiod(R Mt– R Ft).

SMB :SizePremiumfactorwhichisthedifferenceinaveragereturnbetweens m a l l a n d b i g capitalizationcompanies HML :ValuePremiumfactorwhichis thedifferenceinaveragereturnbetweenhighandlowBook toMarketratio(B/Mratio)companies.

CMA :InvestmentPremiumfactorwhichis thedifferenceinaveragereturnbetweenlowandhighgrowth in book equitycompanies.

7and12,previously.Itiscalculatedast h e differencebetweenhighandlo wcumulativereturnportfolios.

Risk Small Minus Big (SMB) High Minus Low (HML) Conservative Minus Aggressive (CMA) Robust Minus Weak (RMW) Cumulative 2-3 (RET23)

ResearchFramework

DATA

Thisresearchinvolvedifferentsortsofdata,rangingfromdaytodaytradingdatatofirmindicator s.Forthedaytodaytradingdata,thesevaluesareobtainedfromVietstockwebsite(oneo f themosttruste dwebsiteforfinancialinformationinVietnam).Thefirmvaluesandimportantindicatorsarecollectedfr omfirms’auditedf i n a n c i a l reportw h i c h s h o u l d b e t h e m o s t t r u s t e d sourceoffirminformatio n).

REGRESSIONMETHOD

Thetraditionalwayoft e s t i n g a capitalassetp r i c i n g m o d e l i s t o r u n crosssectionalre gressionbetweenaverageexcessreturnofstocksorportfoliosandinfluencedriskfactors.Iftheresults h o w s a z e r o interceptterm,t h e n , w e ha ve at ru em od e l w i t h allnecessary riskfactors.However,n oneofthesevariablesareknown,sothat,estimationoftheseriskfactorsareutilizedt o testthemodel.By usingtheestimatorsoftheserisk,weacceptedthattherewillbemeasurementerrorsandtheestimatorswillbebi ased.Apossiblewayofreducingthisproblemistogroupsimilars t o c k s intocommongroups.Nonetheless

Jensen,Black,andScholes(1972)wereamongthefirstgroupofresearcherswhoproposeda metho dtoovercometheissueofselectionbias.Previousperiodestimatedbetawasutilizedasaninstrumentalva riabletogroupthesecuritiesforupcomingyear.Thebasicargumentforusingt h i s instrumentalvariab leisthatiftheassetsmodelistrue,noneofgroupingmethodscangenerateaninterceptdifferentfromzero,sta tistically.ThisideawaslaterappliedintoaresearchofFamaandFrench(1993)in which theyformedportfoliosbasedon sizeandbook tomarketequityratio.Anotherissueappearsduringourtestofamodelisthattheresidualwillbecrosssection aldependencewhenweaggregatethedataonlargenumberofsecurities.Thereasonisabnormalr eturnamongdifferentassetsusuallyvarysignificantlytogether.Forfixingthisproblem,Famaand

VARIABLES

DependentVariables

Inthisthesis,theonlydependentvariablewillbeexcessportfolios’return.Thisvariablewasus uallydefinedas:“Differencebetweenanasset ‘sreturnand riskfreerate”(Nasdaq,2015).

Theexcessreturndefinition is sometimesmisinterpretedwiththeconceptofabnormalreturnsinceabnormalreturn is commonlyusedtoindicatethereturninexcessof thoserequiredreturnbysomeassetpricingmodels(Nasdaq,2015).

Inthispaper,theeachportfolio’sexcessreturnarecalculatedbysubtractingtheportfolioreturna ndtheVNindex’sreturnfor thesameduration.Thisisthecommonapproachforcalculatingtheportfolio’sexcessreturninfinancialstu dies(Fama&French,1993).

ExplanatoryVariables

Untilrecently,withthedevelopmentinliquidityliterature,thereareexistingaconsiderablen u m b e r ofliquiditymeasures.Itisagoodsignasresearchershavevariouschoicesofliquiditymeasure fortheirresearch.However,thisalsocreatetroublesomeforresearchersastheyneedtoselectthebestpo ssiblemeasurefortheirstudy.Inthissection,someof themost popularliquiditymeasureswillbeintroducedinaccordingtotheircategories.Moreover,themostfeasibleli quiditymeasureswillalsobeselectedfor thisresearchduetotheirsuitability.

Liquidity measures are primarily classified based on the frequency of data used in their construction, which includes high-frequency data (intraday) and low-frequency data (daily or weekly) High-frequency data offers greater accuracy but is limited in research timeframe due to the vast amount of data collected over a short period, making it ideal for microstructure studies Conversely, low-frequency data, which involves a smaller volume of information over a longer duration, allows for extensive research due to its greater availability Each type of data has its own set of advantages and disadvantages, influencing the choice of liquidity measure in financial analysis.

The second important classification of liquidity measures is based on their underlying concepts, which researchers categorize into four basic groups The first group focuses on trading quantity, measuring liquidity by the ability to trade large volumes, including metrics such as the number of trades and turnover The second group addresses price impact, assessing liquidity by how trades influence asset prices, utilizing measures like Amihud's illiquidity measure and Kyle's lambda The third group examines trading costs, quantifying liquidity through transaction costs, such as the proportional bid-ask spread and Roll's spread Lastly, the fourth group emphasizes trading speed, evaluating liquidity by the speed of trade execution, with measures developed by Liu and Lesmond et al.

Afurtherdetailintoeachofwell- knownliquiditymeasurewillbediscussedbelowtogivereadersabroaderunderstandaboutthesemeasures Thesemeasureswillbedividedintofourmaincategoriesofliquidityconcept:tradingquantity,priceimpa ct,tradingcost,tradingspeed.

Volume:tradingvolumeusuallydefinedastotalstocksornumberofassetstradedinap redefinedperiodo f t i m e BrennanandSubrahmanyam( 1 9 9 5 ) wereconsideredpioneersw h o f o u n d theimportanceoftradingvolume.InBrennanetal.

(1998),theyfurtherusedollartradingv o l u m e measureintheir assetpricingtestandclaimedthat volume negativelyrelatedtostock returns.Volumealsofoundtobehighlycorrelatedwithothermeasur esofliquidityby(Chordiaetal.,2000).

Turnover:i sdefinedast h e ratiobetweens t o c k tradingv o l u m e i n a fixperiod/ totaln u m b e r o f outstandingstocks.Thismeasuresuggesthowmuchstockistradedinaknownperiod t t

V i td oftimeandtherefore,arelativeunderstandaboutthespeedofstockturnoverinafixduration.Data r,N a i k , andRadcliffe( 1 9 9 8 ) firstappliedturnoveri n t o h i s researchf o r crosssectionalrelationsh ipbetweenstockreturnsandliquidity.Inastudyfor20emergingmarkets,Rouwenhorst(1999)foundturnov er is asignificantexplanatoryfactorofstockreturns.

Amivestliquidityratioistheratiobetweentotalstocktradedvolume/absolutereturnforpre- definedtimeframe.Themeasureis calculatedasfollowed:

Thenotionbehindthismeasureisthatahighlyliquidstockwillonlyfluctuateslightlyinpricew henalargetradingvolumewasexecuted.Cooper,Groth,andAvera(1985)andAmihud,Mendelson,and Lauterbach(1997)wereamongthefirstresearcherswhoinvolvethismeasureintot h e i r liquiditypapers.

Amihudproposedthis illiquiditymeasurein his paper in 2002:

Theintuitionbehindthismeasureisthatifahighvolumetradeisexecutedandthestock’spricemov eslightly,thestockhaslowvalueofAmihud(asAmihudisanilliquiditymeasure,sol o w Amihudval uemeansthestockisliquid).Incontrast,ifahighvolumetradeisexecutedandstock’spricemovesignifi cantly,thestockhashighAmihudvalue(illiquidstock).Inhispaper,A m i h u d (2002)demonstrate dthatstocks’illiquidityhasapositiverelationshipwithexcessstockreturn.

Kyle(1985)’sLamda(λ)):Thismeasurecomposedbytwoseparatecomponentswhichareperman entandtemporaryeffects.Thepermanenteffectcapturestheinformationcontentoftheorderflows( Kyle,1985).Whereas,thetemporaryeffectiscapturedbytransientliquidityeffects,price discreteness… etc.Theestimatingregressionforthismeasureis asfollow:

Pt=λVVt+(Qt–Qt-1)+et (19) λVis theadverse selectioncostdueto thepermanenteffectsoftrades.

Qti sthedirectionalvariable whichshowtrade direction(+1forbuy- initiatedtrade,-1forsell– initiatedtrade)

The institutionbehindKyle(1985)’smeasure isthatitmeasuresthepriceimpactforeachshare.Thesmaller is Kylemeasure,themoreliquid is thestock.

BidAskSpread:isthedifferenceinbuyingandsellingpriceatthesamemomentintime.A t t h i s pri ce,i n v e s t o r c a n i n s t a n t l y buyorsellt h e i r preferredassets.M a n y studiesh a v e b ee n carriedoutb yusingthismeasureofliquidity.SomeofthepioneersandprominentresearchesareA m i h u d a n d Mendelson(1986),EleswarapuandReinganum(1993),Eleswarapu(1997).

Therea r e t h r e e b a s i c t y p e s o f spread.T h e firsttypei s q u o t e d spread,i t i s s i m p l y t h e difference betweenthebid andaskprice.

(20) t  1(Ask i,t Bid i,t )/2 Thethirdtypeiseffectivespread,thisspreadisthedifferencebetweentherealexecutedpriceand middlepointofbid- askpriceatthetimethetradewascarriedout.Thereasonissomei n v e s t o r s canorderatabetterpriceth antheindicatedbid- askspreadinthemarketduetodifferentunderlyingreasons.Therefore,effectivespreadwasutilizedin somestudiesinadditiontotraditionalbid-askspreadandproportionalspread.

Roll(1984)’sspread:inhis 1984paper,Rollproposesaspreadmeasure asfollow:

Thisformulacalculateseffectivespreadbyu s i n g t h e b i d - a s k b o u n c e i n d u c e d n e g a t i v e a u t o - covarianceindailyreturns.Inthisformula,COVii s t h eauto-covarianceofreturnsforstocki.

Theconceptofzerodailyreturnwasthatinthedaywithzeroreturn,theexecutingorderfrominv estorsweredelayedorhaddifficulties.Therefore,thezerodailyreturniscorrelatedwith tradingspeedandutilizedasameasureofliquidity.Bekaertetal.

Thismeasurei s a revolutionaryv e r s i o n o f z e r o d a i l y returnmeasure.T h i s measurei s con sideredasastandardizedturnoveradjustedfornumberofzerodailytradingvolumesoveradefinedp eriodoftime.Theformulais inequation22below:

  xmonthturnover=sum of dailyturnover overpriorxmonths

Throughoutt h e analysiso f allabovemeasures,s o m e importantinferencef r o m t h o s e l i q u i d i t y measurescan bedrawnout:

Therearetwoimportant aspectsofliquidity whicharedatafrequency andt h e buildin gconceptbehindthemeasure.Forthisthesis,duetoalongtimeframefrom2007to2013,onlyt h o s e measureswithlowfrequencydatacanbeobtainedforthestudy.Moreover,asVietnamstockmarketis imp lementingthematchedpriceorder mechanism,sothatonly certainmeasuresarep o s s i b l e t o c alculatef o r t h e researchp u r p o s e T h e s e measuresc a n b e l i s t e d suchasvolume,turnover,amiv estliquidityratio,Amihud(2002)’silliquidityratio…etc.

Despitet h e s e constrains,t h e writero f t h i s t h e s i s s t r i v e d t o achievet h e bestp o s s i b l e measurewhichwillnotcompromisetheaccuracyinmeasuringtheinfluenceofliquidityonassetretu rn.InathoroughresearchbyChoeandYang(2008),mostofwell- knownliquiditymeasuresarecomparedagainstothersbyusingtwomethods.Thefirstmethodwascorrelatio ncheckinordert o exposetheinternalrelationbetweendifferentmeasures.Thesecondmethodwasempi ricaltesti n whichtherelationshipbetweenstockreturnsanddifferentliquiditymeasuresareexamined.The empiricaltestwascarriedout on stock listedfirmsinKoreanandUS stockmarketfor1993-

2004period.Afterconductingtwo mentioned methods, ChoeandYang(2008)concludedthat:

 Inthecorrelationcheck,theAmihud(2002)andproportionalbid- askspreadshowedt h e y arebestmeasuresinboth KoreaandUS market.Therefore,th e concernaboutconflictingofliquiditymeasuresaresubdued.

 Intheempiricaltest,Amihud(2002)illiquidityratioandturnovermeasureprovedtob e th ebestmeasureswhentheyareallsignificantlyrelatedtostockreturnsinKoreanandUSmark et.

Fromaboveconclusionandargument,itseemthat A m i h u d (2002)i l l i q u i d i t y ratioandTu rnovermeasurearethebestcandidatesasliquiditymeasuresforthisresearch.Therefore,thewriterd ecidedtoemployallthesetwomeasuresintothisthesisinordertoverifywhethertheresultfromthesetwomeasur essupporteachother.Inaddition,itwillalsoimproveaccuracyaswellasv a l i d i t y of theempiricaltestaboutinfluenceofliquidityonassetreturns.

ControlVariables

The asset pricing literature has rapidly evolved since the introduction of key theories such as the Efficient Capital Market and the Capital Asset Pricing Model Extensive research has identified numerous factors that can explain asset returns A significant contribution to this field was made by Fama and French (1993), who identified two important explanatory variables in addition to the market risk premium (βi): SMB (small minus big) and HML (high minus low) These factors have demonstrated their effectiveness in explaining a substantial portion of asset returns across various studies and are now utilized by many governments and organizations to assess the value of their assets.

Nonetheless,thereisstillaportionofunexplainedreturn,evenafterapplyingthreewell- k n o w n factors(marketriskpremium(βi),SMB,HML).Thisleftroomforfurtherresearchintoasset returns.O t h e r pr om in en t e x p l a n a t o r y variablesf o r assetr e t u r n s areR M W (robustm i n u s we ak)andC M A (conservativem i n u s a g g r e s s i v e )

(Chordia,Subrahmanyam,& Anshuman, 2001).Thesevariableswillbecalculatedandaddedintom odeltocontrolfor theirpotentialexplanatorypowero n assetreturns.

DATA PROCESSING

PrimaryDataCalculation

Int h i s section,t h e d i s c u s s i o n w i l l o n l y b e f o c u s e d o n p r i m a r y datap r o c e s s i n g f o r allvariables.Thecalculationprocesswasmainlybasedonoriginalpaperswherethesefactors arefirstdeveloped.Itshouldbebettertogivereadersageneralunderstandabouttheprocedureofc reatingthefactorsandhowdataisanalyzed.

ThedailytradingpriceforlistedstocksonHOSEareobtainedfromVietstockwebsite. Fromthesedailyprice,daily stockreturnwillbecalculated.Theaveragedaily returnforeachs t o c k inamonthisaresultoftotalstock returnduring amonth / totaltradingdayduringthisparticularmonth.

3.5.1.2For SMB,HML, CMA,RMW:

SMB(SmallMinusBig):themarketcapitalizationforeachlistedfirmiscalculatedbyu s i n g thestockpriceattheendoftheyearmultiplyforitsaverageoutstandingstockduringtheyear.Th ismarketcapvaluewillthenbeutilizedtorankthemarketcapofeachfirmanddividedt h e s e stocks into different portfolios Theseportfolios will beusedtoperformfactorcalculation.

HML(HighMinusLow):the book tomarketvalueofeachfirmiscalculatedastheratioo f thefirm’sbookvalueattheendofthefiscalyearan dmarketcapitalizationofthefirm(bookvalue/ marketvalue) Thisratioisthen involvedin theprocessofsplitting stockintodifferentportfolio sdependingonB/

1accountingdata.T h e appliedformulai s = (totalfirmrevenue– costo f goodss o l d – s e l l i n g – generalandadministrative expenses–interestexpense)/bookequity.

2f o r i t s derivation.T h e appliedformula=(totalasset inyeart-1–totalasset inyeart-2)/totalasset inyeart-1.

Tosumup,thefinancial datasize(SMB),book/marketvalue(B/M),Profitability(RMW)andInvesment(CMA)isfirstcollectedatt heendofyeart-

1.Theseinformationisthengatheredu p andcalculatedtoderivethebreakpointsforeachfactors(30 th ,50 t hor70thpercentile).Afterthebreakpointis finalized,stocksareallocatedintoportfoliosbaseon thebreakpointsofeachfactors t t

Afterallocatedi n t o a particularportfolioi n yeart , d a i l y s t o c k r e t u r n o f eachs t o c k i s cal culatedbasedonthedailystockprice.Later,thisdailyreturnforeachstockisusedtoobtaint h e avera gedailyreturnforaparticularstockduringamonth.Thisaveragedailystockreturnisonceagaingat hereduptocalculatetheaveragedailyreturnofallstocksinoneportfolio.Thisist h e processforachievi ngaveragedailyreturnforeachportfolioforaspecificmonth.Thisprocessrepeatedon monthlybasis foreachyear.

Afterthat,basedonthenextyearfinancialdatasize(SMB),book/marketvalue(B/

M),P r o f i t a b i l i t y (RMW)andInvesment(CMA),newportfoliosfornextyearwillbeformedan dthew h o l e processrepeata g a i n T h e researchp e r i o d f o r t h i s t h e s i s i s 2 0 0 7 to2 0 1 3 (year

1,hence,theirfinancialinformationisonlyusedtocreateportfolios.Therew i l l n o t b e a n y factorsfor thisyear).

VNindexreturn)iscomputedondailybasis.Then,theaveragedailyreturnofVnIndexforeachmonthiscomp utedusingdailyreturnofeverytradingdaysinamonth.Thebetavalueisaresultofsubtractionofaverag edailyVNIndexreturnforeachmonthandaveragedailyriskfreeratereturn(discountratefromStateBan kofVietnam) ofthesamemonth.

Therearetwomeasuresofliquidityforthisthesis:Amihud(2002)’silliquidityratioandtur nover.Alltwomeasuresarecalculatedon dailybasis.Theyarethenaveragedoutto obtaintheaveragedailyvalueforeachmeasure.Theformulafortwomeasureswerealreadydiscu ssedinl a s t section.However,for theconvenienceofreaders,theyarecopieddownfromthelastsection.

Turnover: is definedastheratiobetweenstocktradingvolumeinafixperiod/ totalnumbero f outstandingstocks.

Theinitialprocessofcumulative returndatawasadoptedfrompaperofChordiaetal. (2001).A l l t h e returnsf o r eachi n d i v i d u a l s t o c k ared a i l y averager e t u r n o f o n e m o n t h T h e threemeasuresofcumulativereturnarecalculatedasfollow:

3:thecumulativereturnoftwomonths(month2 ndand 3rdpreviously,ignoredfort h e closedpreviousmo nth t-1).

RET4-6:thecumulativereturnof3months(month4 tht o 6thpreviously,ignoredforthel a s t 3 m o n t h s return).

RET7-12:thecumulativereturnof6months(month7 thto 12thpreviously, ignoredforthel a s t 6 m o n t h s return).

FactorConstruction

AccordingtoFamaandFrench(2013),therearetwopossibleapproachesforformingtheportfoli osofdependencevariables(theaveragemonthlyexcessreturn– thereturninexcessofthebenchmarkreturn).

Thefirstpossibleapproach is to sortallstocksbyusingtwoindependentvariables:Size–B/M,Size–Investment,Size–

YorkStockExchange(NYSE)marketcapbreakpointswhichweredividedintoequalportionbetweeneach group(20 th ,40 th ,60 th ,8 0 thpercentiles).

Mgroupswhichalsoe q u a l l y divided bybreakpoints Subsequently,bymerging s t o c k s wh ichhavethesameSizeandB/

Mpercentilesintooneportfolio,wehave25portfoliosw i t h eachstockbelongtoonlyonespeci ficportfolio.Thesementioned25portfoliosarecalledaveragem o n t h l y excessreturnsportfoliosfo rmedo n S i z e andB/

M(itiscalledexcessreturnportfoliosbecauseaveragereturnfromtheseportfolioswerealreadysubtracted forriskfreerate).Similarly,t h e 2 5 S i z e – OperatingProfitabilitya n d 2 5 S i z e –InvestmentPortfolioscanb e constructedin thesamemanner.

Thesecondapproachistoconstructportfoliosonthreeindependentvariablesinsteadoft w o asabove.Initially,allstocksareseparatedintotwoSizegroups(onlysmallandbiggroups)byu s i n g N

Y S E median( 5 0 % breakpoint)marketcapast h e separationp o i n t Therefore,alluniverseofstock saredividedintoeithersmallorbigSizegroup.Subsequently,stocksineachS i z e groupareindepe ndentlyassignedtofourB/

MgroupsandfourOperatingProfitabilitygroupsu s i n g theequallydividedbreakpoints.Asaresult,inea chSizegroup(eithersmallorbiggroup),w e have16sub-portfoliosofB/

MandOperatingProfitability.AndfortwoSizegroups,wehavetotal3 2 portfoliosf o r t h i s approach.In a similarmanner,averagem o n t h l y excessreturnsf o r portfoliosconstructedonSize,B/

MandInvestmentorportfoliosconstructedonSize,OperatingP r o f i t a b i l i t y andInvestm entarecreatedfollowingabove methodology.

FamaandFrench(2013)proposedthreemethodsforbuildingsetofexplanatoryfactors.O n l y Size,B/M,profitability,andinvestmentareincluded inportfolios’constructingprocess,themarketrisk factoriscalculatedd if fe re nt ly byusingthedail yaveragereturnof marketin eachm o n t h minusfordailyreturnofGovernmentbond(riskfree)r ateinthesamemonth.Therearethreesortswhichare:

- 2 x3 sorts onSize andB/M,SizeandOperatingProfitability,SizeandInvestment.

- 2 x2 sorts onSize andB/M,SizeandOperatingProfitability,SizeandInvestment.

- 2 x2x2x2 sorts onSize,B/M,OperatingProfitability,andInvestment. a) 2 x 3 sorts on Size andB/M,Size andOperatingProfitability, Size andInvestment:

Inthe2x3sorts,FamaandFrench(2013)assignedallstocksintotwoSizegroupsusingt h e NYS Emedianasitsbreakpoint.Allstocksalsoindependentlyassignedintothreegroupsofo t h e r facto rs(eitherB/M,operatingprofitability(OP)orinvestment(Inv))using the30 thand 70thpercentilesasthebreakpointineachsort.Theportfoliosareestablishedbytheintersecti onsoft h e s e groups.Theseportfoliosarenamedbyusing thefollowingletters:

- ThesecondletterdescribestheB/Mgroup,with high (H) orneutral(N)orlow(L)o r labelstheOPgroup,withrobust(R) orneutral(N)orweak(W) orrepresentstheInvgroup, withconservative(C)orneutral(N) oraggressive(A)

Table 1:2x3factorsbuildingprocess Breakpoints Factorsand theircomponents

SMB B/M =( SH +SN +SL)/3–(BH+BN+BL)/

3SMB OP =( SR +SN +SW)/3–(BR+BN +BW)/

3SMB Inv =( SC +SN +SA)/3–(BC+BN+BA)/

HML=(SH +BH)/2–(SL+BL)/

2RMW=(SR +BR)/2–(SW+BW)/

2CMA=(SC +BC)/2–(SA +BA)/2 b) 2 x 2 sorts on Size andB/M,Size andOperatingProfitability, SizeandInvestment:

Procedureforbuildingupfactorsin2x2sortissimilarto2x3sort.Theonlyexceptionist h a t thereis noneutralportfoliosforanygroupsinthe2x2sortsandthebreakpointsaretheNYSEmedian.

Table 2:2x2factorsbuildingprocess Breakpoints Factorsand theircomponents

SMB=( SH +SL+SR +SW+SC +SA)/6–(BH+

BL+BR+BW+BC+BA)/6

HML=( SH +BH)/2–(SL+BL)/

2RMW=( SR +BR)/2–( SW+BW)/

2CMA=( SC +BC)/2–(SA +BA)/2 c) 2 x 2 x 2 x 2 sortsonSize,B/M,OperatingProfitability,and Investment:

(B)andsmall(S).However,therewillbeacombinationofother3charactersasthissortingmethodinvolve allgrou psforbuildinguptheexplanatoryfactors.ThesecondletterisB/

Mgroup,withhigh( H ) andl o w (L).T he th ir dl e t t er is O P group,w i t h r o b u s t ( R ) andweak( W ) , t he f o u r t h characterisInvgroup,withconservative(C)andaggressive(A).Thederivationoffactorsfollo wst h e procedurein below table:

Table 3:2x2x2x2factorsbuildingprocess Breakpoints Factorsand theircomponents

SMB=(SHRC+SHRA+SHWC+SHWA+SLRC+SLRA+SLWC+SLWA)/8

–(BHRC+BHRA+BHWC+BHWA+BLRC+BLRA+BLWC+BLWA)/8

HML=(SHRC+SHRA+SHWC+SHWA+BHRC+BHRA+BHWC+BHWA)/8

–(SLRC+SLRA+SLWC+SLWA+BLRC+BLRA+BLWC+BLWA)/8

RMW=(SHRC+SHRA+SLRC+SLRA+BHRC+BHRA+BLRC+BLRA)/8

–(SHWC+SHWA+SLWC+SLWA+BHWC+BHWA+BLWC+BLWA)/8

CMA=(SHRC+SHWC+SLRC+SLWC+BHRC+BHWC+BLRC+BLWC)/8

–(SHRA+SHWA+SLRA+SLWA+BHRA+BHWA+BLRA+BLWA)/8

Asmentioned above,therearethreepossibleapproachesforformingthesetoffactors.H owever,anempiricalresearchbyFamaandFrench(2013)suggestedthatthesethreeapproachesf o r for ming thefactorsmight providesimilardescriptions ofaveragereturns forthe Stock Returnportfolios(dependentvariable).

Atfirst,the2x2sortofSize,HML,RMWandCMAisbelievedtobemoreeffectivethant h e 2 x 3 sortssincethe 2x2 sortsincludesallstocks tocalculatethefactorswhilethe 2 x3 sortso n l y involves60%ofstocksinitsformation.However,anempiricaltestfromfivefactorsmodelo f FamaandFrench(2013)indicatedthat2x2and2x3sortsprovideasimilarexplanatorypowero n theaverag ereturnoftheexcessreturnportfolios.Asaresult,theyconcludedthatselectionbetween2 x2and2 x3 sortsseeminconsequential.

Inaddition,regardingthe2x2x2x2sort,thissorttheoreticallyprovidesanefficientmethodf o r isolatin gtheeffectoffourfactors.However,italsocarriesnumberofflaws.Thefirstdrawbacki s thatthemultiplere gressionwasdesignedtomeasurethemarginaleffectsofeachfactors,soiti s unclearwhetherthefactorsc onstructedfrom2x2x2x2sortcanbetterdisentangletheexposurest o variationinaveragereturnsrelatedt oSize,B/

M,profitability,andinvestment.Moreover,tryingt o controlformorefactorscanalsobeproblematic.T hecorrelationbetweenexplanatoryfactorscanresultininadequatediversificationofsomeportfolios whicharethenutilizedinformingthefactors.Forthisreason,anoveruseofcontrollingvariablesinformingth eportfolioscanpotentiallyreduceitseffectiveness.

Tosumup,FamaandFrench(2013)suggestedthatthesortsof2x3wouldbethebestouto f threeap proachesasitprovidesbothflexibilityandeffectivenessincomparisonwith2x2and2 x 2 x 2 x 2 s o r t s

Thekeypointisthatthefirminformationisupdatedeachyearbyusingthenewfirm’sfinanciali nformation.Allthefirmchangeswillthenbeupdatedintoeachofthefactors.Therefore,firmswillbelocatedi ntheappropriateportfoliosinaccordancewithitsfundamentalforeachyear. b) Forexcessreturns portfolios:

According to research by Fama and French (2013), there are two methods for forming portfolios to calculate average stock returns: the 5x5 sorts based on two variables and the 2x4x4 sorts based on three variables Their findings indicate that the 2x4x4 sorts explain a smaller portion of excess returns for average return portfolios compared to the 5x5 sorts Specifically, the average R² values for the 5x5 formation range from 0.91 to 0.93 in five-factor regressions, while the 2x4x4 formation yields average R² values between 0.85 and 0.89.

Thisincidencewasclaimedtobearesultofpoordiversificationonthe2x4x4sortsthant h e 5 x 5 s o r t s Firstly,d e s p i t e t h e 2 x 4 x 4 s o r t s produce3 2 portfolios,t h e breakpointsf o r eachvaria blesismedianand25 thwhich islessdiversifiedincomparisonto5x5sorts(eachvariable’sbreakpointis20 th).Furthermore,thecorrelationbetweenvariablesinthesamesortreducesthediversificationofexce ssreturnportfolios(forexample,thenegativecorrelationbetweenOPandB/Msuggeststhereshould befewbigstocks in thetopquartilesofB/MandOP).

Intheperiod2007to2013,thefirstfewyearsdidn’thavesufficientfirmstofillintoall25returnportf olios.Moreover,2 5 portfoliosw i l l n o t p r o v i d e sufficientv a l i d i t y f o r FamaandMacBeth( 1973)’sregressionprocedureasthisprocessrunsintwostageswhichinvolvesboth cross- sectionalandt i m e seriesregression.D u e t o abovereasons,t h i s t h e s i s w i l l e m p l o y 1 6 portf oliosforeachsorts(sortonSize-B/M,sortonSize–Investment,sortonsize–

Profitability).Totally,w e w i l l have16x 3 = 4 8 portfolioso n excesss t o c k returnswhichp r o v i d e suf ficientv a l i d i t y forFama andMacBeth(1973)’sregression.Thebreakpointwill be25 th ,50 th ,and75 thf o r allthesortsinsteadof 20 th ,40 th ,60 thand 80 thas intheoriginalpaper. c)Liquiditymeasure:

M,investmentandprofitability.L i q u i d i t y p o r t f o l i o s areconstructedo n annualbasisi n ordert o p r o v i d e acomprehensiveview aboutfundamentalchangeinliquidi ty foreachyear.Allofth ei n d i v i d u a l stocksareallocatedinto3differentportfoliosin accordancetotheir averageann uall i q u i d i t y breakpoints( t h e breakpointsare3 0 thand 7 0 th ).Thereareonly3 portfoliosf o r this measure(high,neutralandlow)ofliquidity.Thefactorderivationisobtainedbysubtractingtheaverager eturnofhighliquidity portfoliofortheaveragereturnoflow l i q u i d i t y portfolio.Thedifferenceb e t w e e n t h e s e t w o portfolioswill bet h e i n p u t f o r FamaandMacBeth(1973)’sregression.ThisprocessalsosimilarforcalculationofAmihud’sill iquidityratioinwhichtherew i l l b e asubtractionofhigh illiquidityportfolio for thelow illiquidityportfolio. d)Cumulative return:

Threecumulativer e t u r n measuresaref o r m e d i n t h e samemanner.A s a r e s u l t , t h e procedurewillbeintroducedonlyonetime.Firstly,stockreturnsar ecumulatedinaccordancewitht h e i r measure(RET 2-3iscumulatedofprevious2and3 monthsreturn).Next, cumulativereturno f stocksarearrangedintothreedifferentportfoliosforeachyearfrom2008to201 3(RET7-

12used2007dataasbaseyearforitscalculation).Therearethreeportfoliosineachyearrepresent030 t hpercentile,3 0 th  70thpercentile,70 th 100thpercentileofc u m u l a t i v e return.T h e allocationprocedurewillberepe atedonyearlybasis.Subsequently,thecumulativereturnintheseportfoliosareaveragedtoobtainmonthlya veragecumulativereturnforwholeportfolio.Thefinali n p u t t o FamaandMacBeth( 1 9 7 3 ) ’si s att ainedbys u b t r a c t i n g t h e h i g h e s t cumulativer e t u r n portfolioforthelowestone.

SOLVINGPOTENTIALECONOMETRICISSUES

Dealingwith HeteroscedasticityandAutocorrelation

Financial research often involves time series and panel data, which frequently exhibit issues such as serial correlation and heteroscedasticity Savoiu (2013) highlighted that these problems significantly affect model construction for financial datasets, violating the fundamental assumptions of Best Linear Unbiased Estimators (BLUE) in Ordinary Least Squares (OLS) regression Petersen (2009) noted that a substantial 45% of recent research papers using panel data failed to adjust for standard errors due to potential residual dependence Among the studies that did adjust, methodologies varied: 31% utilized dummy variables for each cluster, while 34% employed the Fama and MacBeth (1973) procedure for estimating coefficients and standard errors The remaining studies adopted one of these two methods for standard error adjustments.

WestproceduremodifiedforuseinpaneldatasetorRoger‘sstandarderrorswhichi s t h e adjustedW h i t e st andarderrorst o a c c o u n t f o r pos si bl e correlationw i t h i n a cluster.”

According to Petersen (2009), financial data commonly exhibits two types of correlations: cross-sectional correlation, which refers to the correlation of residuals among different firms in a given year (termed a time effect), and time series correlation, which involves the correlation of residuals within a specific firm over different months or years (referred to as a firm effect) While the Fama and MacBeth (1973) procedure effectively addresses time effects and cross-sectional correlation in residuals, the data for this thesis still faces firm effects in the residuals To resolve this issue and obtain more reliable standard errors, this thesis will utilize the Heteroskedasticity and Autocorrelation Consistent (HAC) Standard Errors method, which accounts for heteroskedasticity and autocorrelation of unknown form in the data Basic knowledge about this method will be discussed further.

HACstandarderrorswasfirstintroducedbyNeweyandWest(1986)asanapproachtooverco meh e t e r o s k e d a s t i c i t y andautocorrelationi n t h e errortermsoft i m e seriesd a t a Int h i s m o d e l , a covariancematrixforaconsistentestimatoris:

LR is theestimatorfor longruncovariancematrixof theerrorterm:

Essentially,HACstandarderrorsistakingaverageofserialcorrelatederrorsoverapre- definedtimehorizon Sothat,the mostimportanttaskistodefinehow manytimeperiodsfor t a k i n g theaverageandindividualweightofeachresidualinthatprocedure.Therearedifferentm e t h o d s toassignweightforeacherrorterm(thesemethodsarenamedkernel).Inaddition,then u m b e r oftimeperiodsfortakingaverageisnamedbandwidth.Somecommonweightedaveragem e t h o d s a reBarlettkernelandParzenkernel.Bandwidthusuallyderivedfromaspecifiedformulawhichiscalculated basedonsamplesize(nw1=0.75N 1/3 ,nw2=4(N/

100) 2/9 ,Nisthesamplesize).Int h i s t h e s i s , H A C s t a n d a r d errorsw i l l u s e t h e commonBarlettkern elw i t h bandwidthn w 2 = 4(N/100) 2/9f o r fixingpotential issueof heteroskedasticityandautocorrelationin theerrorterms.

DealingwithMulticollinearity

Multicollinearityisalsoacommoneconometricissue.Therefore,thisissueisalsoconsideredin thestudy.Multicollinearitycantriggeramisleadinginterpretationofeconometricresultduetoc orrelationsbetweenexplanatoryvariables.Somecommonpracticalconsequencesofmulticollinearityincl ude:largevarianceandstandarderrorsofOLSestimators,wilderconfidentintervals,insignificanttrati os,ahighR 2value butfewsignificanttratios…etc.

Page| 48 LIQUIDITY PREMIUMIN STOCKRETURNS,THECASE OFVIETNAM

TURNOVER RISK SMB HML CMA RMW RET23 RET46 RET712

Forinterpretingthecorrelationmatrixtable,somekeynumbersworthtobenoticed.Firstly,theTURNOVERdatashowedtwosignificantc orrelationwith HMLandRMWvariables(thecorrelationis-0.406and-

0.465,respectively).Secondly,thereisnoticeablecorrelationbetweenHML,CMAandRMW(correlationbetweenHMLandCMA=-

Page| 48LIQUIDITY PREMIUMIN STOCKRETURNS,THECASE OFVIETNAM

Page| 49 LIQUIDITY PREMIUMIN STOCKRETURNS,THECASE OFVIETNAM

AMIHUD RISK SMB HML CMA RMW RET23 RET46 RET712

Thecorrelationmatrixt a b l e p r o v i d e s o m e i n t e r e s t i n g correlationf i g u r e s First,therearehighcorrelationsbetweenAmihu di l l i q u i d i t y ratioandSMB,HMLvariables(thecorrelationis0.601and0.481,respectively).Inaddition,correlationsbetweenHML,CMAandRMWarealsosignificant(correlationbetween HMLandCMA=-0.683, HMLandRMW=0.840, CMAandRMW=-0.580).Thesecorrelationvaluesindicatepossible multicollinearityissuein thesampledata.

Page| 49LIQUIDITY PREMIUMIN STOCKRETURNS,THECASE OFVIETNAM

Asafurtherexaminationonthe correlationbetweene x p l a n a t o r y variables.Auxiliary regres sionsofeachexplanatoryvariableagainstothervariableswillbecarriedout.

Duetolimitedpresentingarea,onlytheauxiliaryregressionofTurnoverandAmihudi l l i q u i d i t y ratioagainstotherexplanatoryvariableswillbedisplayedbelow.Otherauxiliary regres sionswill onlybereportedwiththeiradjustedR 2

Variable Coefficient Std.Error t-Statistic Prob

Table 7:AdjustedR–squaredofeachauxiliaryregressionin first model

TURNOVER RISK,SMB,HML,CMA,RMW,RET23,RET46,RET712 0.277

RISK TURNOVER,SMB,HML,CMA,RMW,RET23,RET46,RET712

SMB TURNOVER,RISK,HML,CMA,RMW,RET23,RET46,RET712

HML TURNOVER,RISK,SMB,CMA,RMW,RET23,RET46,RET712

CMA TURNOVER,RISK,SMB,HML,RMW,RET23,RET46,RET712

RMW TURNOVER,RISK,SMB,HML,CMA,RET23,RET46,RET712

0.715RET23 TURNOVER,RISK,SMB,HML,CMA,RMW,RET46,RET712 -0.017

RET46 TURNOVER,RISK,SMB,HML,CMA,RMW,RET23,RET712 0.018 RET712 TURNOVER,RISK,SMB,HML,CMA,RMW,RET23,RET46 0.132

SimilarwithTurnovermeasure,auxiliary regressionresultofAmihud’sIlliiquidityRatioisdisp layedbelow:

Variable Coefficient Std.Error t-Stat Prob

The analysis of various financial factors reveals significant correlations among them The AMIHUD risk metric shows a value of 0.494, indicating a moderate positive relationship with other variables The SMB factor also demonstrates a notable correlation at 0.394, while HML exhibits the strongest connection at 0.763 CMA and RMW display positive values of 0.473 and 0.685, respectively, suggesting their relevance in risk assessment Interestingly, RET23 has a minimal impact with a value of 0.021, and RET46 shows no correlation at all These insights are essential for understanding the dynamics of risk and return in financial modeling.

RET712 AMIHUD,RISK,SMB,HML,CMA,RMW,RET23,RET46 0.106

Aftertestingpossibleauxiliaryregressionsforallexplanatoryvariables,wecaneasilyobser venoticeablyhighR 2f o r auxiliaryregressionofTURNOVER,AMIHUD,HML,CMAandRMW.T hisresultsandthecorrelationmatrixsuggestthattheremightbemulticollinearityproblemin thesample.

Tohandlethisproblem,anapproachoftenusedbyresearchersistoseparatepotentialmulticol linearityvariables.Theyarethenestimatedindifferentregressionstogetherwithotherindependentvaria bles.Thismethodwillassistresearchersinmaneuveringaroundthemulticollinearityproblem andestimate bettertheeffectsofeachexplanatoryvariables.

Inthisthesis,theforemostpurposeistoestimatetheeffectofliquidityonstockreturn.Therefore ,t h e p r i o r i t y ist o alleviatea n y p os s i b l e multicollinearityissue betweenl i q u i d i t y measure andothercontrolvariables.Aftercloselyinspectauxiliaryregressionsintable6andt a b l e 8,itcan beobservedthatCMA,RMWhavesignificanteffectonTurnovermeasureint h e firstmodel.Int hecaseofAmihudilliquidityratiointhesecondmodel,SMB,HMLweref o u n d tohavesignifica ntexplanatorypoweronAmihudilliquidityratio(theirt- statisticaresignificantat10%level).

Inordertoalleviatepossiblemulticollinearityeffects,thesehighlycorrelatedvariablesw i l l bedroppedonebyoneindifferentregressions.Inaddition,infinalregressionofeachl i q u i d i t y measures,allofth es eh ig hl y correlatedvariablesw i l l b e d ro pp ed Ultimately,t h e underlyi ngpurposeofdroppingthesevariablesisto examinehowmuchthesevariablesaffectt h e e x p l a n a t o r y p o w e r ofl i q u i d i t y o n s t o c k r e t u r n s Detailsabouttheseregressionsarepresentedinempiricalresultsection.

This chapter focuses on the descriptive statistics of data and the empirical results of the two models presented in Chapter Three The first section provides an overview of the general statistics for each variable, including mean, median, maximum, and minimum values The second section delves into the empirical results of the study, which is further divided into two subsections to explain the results of the different models Finally, the last section offers a thorough interpretation and discussion of the empirical results for each liquidity proxy, as well as other control variables, including the Fama-French five factors and cumulative returns.

DESCRIPTIVE STATISTIC

TurnoverMeasure

Int h i s part,t h e d e s c r i p t i v e statisticsw i l l b e dividedi n t o 2 sectionsf o r 2 differentl i q u i d i t y variables.Thefirstpartwillcoverdescriptivestatisticforturnovermeasure.

Table 10:Turnoverregressiondescriptive statistics Mean Median Maximum Minimum Std.Dev Skewness Kurtosis TURNOVER 0.007 0.013 0.883 -0.472 0.272 0.407 3.140

Inthistable,somebasicstatisticsaboutvariablesarepresented.Thesestatisticsincludedmean ,median,maximum,minimum,standarddeviation,skewnessandkurtosis.Then u m b e r o f obse rvationsf o r eachvariablesare72correspondedw i t h 7 2 m o n t h s i n 6 years(2008-

2013).Anoteisthatallofthevariablesarecalculatedasthedifferencebetweenthe1 stand the3rdportfolios,e xceptforriskvariableswhereriskisthemarketriskpremium(βi)andcalculatedasdifferencebetwee naveragedailyVNindexreturnandaveragedailyriskfreeratereturn.

Thereasonf o r s m a l l turnovervaluei n t h e r e p o r t e d t a b l e i s becausei t s h o w e d t h e averaged a i l y t u r n o v e r f o r s t o c k s d u r i n g a m o n t h Furthermore,eachobservationi s t h e differencebetweenaveragedailyreturnofhighturnoverportfolioandaveragedailyreturnofl o w turnoverportfolio.

Anothernoticeableinterpretationfromthestatistictableisthattheturnoverserieshasskew nessof0.407andkurtosisof3.14.Thesenumbersmeantthatdistributionofturnoverdatahasnearsymmetric alandnormalGaussiandistribution.Thisconfirmedthatturnoverdataiswellcollectedandp r o v i d e s reliabler e s u l t A l l o t h e r variablesalsoh a v e decentvaluesofskewnessandkurtosis.The onlyexceptionisthatcumulative7-

12(RET712)whichhashighskewnessandkurtosis value.Overall,thenumber ofobservations mightbe quitesmall (72observations)whichcreatetroublesomeforgettingnormaldistribut ionsample.Nonetheless,t h e sample hasbeenprocessedcarefullyandprovidesanacceptablerangeofdistribution.

AmihudIlliquidityRatio

Table 11:Turnoverregressiondescriptive statistics Mean Median Maximum Minimum Std.Dev Skewness Kurtosis AMIHUD -0.010 -0.022 0.470 -0.916 0.233 -0.646 5.381

ThedescriptivestatisticforAmihud’silliquiditytableissimilartoTurnovervariableinasenset h a t allothe rindependentvariablesarethesame,exceptforturnovermeasureischangedbyt h e u s e o f A m i h u d ’ s i l l i q u i d i t y ratio T h e skewnessindicatesa goodp r o p o r t i o n i n sampled i s t r i b u t i o n wherethereisanearequaldistributionbetweenleftandrightside.However,theKurtosisvalue o f 5 3 8 1 indicatesa Leptokerticd is tr ib ut io n wherem o s t o f t he sampledataconcentratecloset o themeanvalue.

ECONOMETRICRESULTS

EmpiricalResultsforTurnoverMeasure

Alltheeconometricresulttablesinthispartwillcontainfourdifferentregressions.Thefirstregre ssionwillincludeallexplanatory variables.Thesecondandthirdregressionswillseparatet w o p o t e n t i a l multicollinearityv a r i a b l e s i n t o t w o differentregressions.T h e finalregres sionforeachliquidity measurewilldropallpossible multicollinearityvariables.T h i s w i l l alleviatethemulticollinearityissueandhelpestimatingtheeffectofliquiditymeasureons t o c k returnprecisely.

REGRESSIONRESULTOFTURNOVER Regression1 Regression2 Regression3 Regression4

*, **, ***representsignificant at 1%level,5%leveland 10%level,respectively

Baseonsomecommonindicatorsofeachregression,someobservationscanbedrawno u t AllregressionsshowedasimilarvalueofadjustedR- squared,S.Eofregressionandsumsquareresidual.

Inanotheraspect,DurbinWatsonstat,whichisanindicatorofautocorrelation,suggestedtha t thirdregressionsufferlessfromautocorrelationissuethanotherregressions.

Tosumup,thethirdandfourthregressionsprovidemostreliableresultswhentheybothhavehighestR -squaredandDurbinWatsonstat.Nonetheless,thereis atrade-offin selectingeithert w o m o d e l s whichdependon the priorityof users.

Fromtheregressionresultoftable12,itcanbeobservedthatalltheregressionsstronglys u p p o r t th eroleofturnoverinexplainingstockreturns.Firstly,allthesemodelsindicatethatt h e empiricalre sultsofturnoverareatleastsignificantat5%levelwhichisahighsignificantlevel.Secondly,thelast r egression,whereCMAandRMWaren’tincluded,showedaveryhighl e v e l o f c o n f i d e n c e t h a t T u r n o v e r i s a p r e m i u m factorf o r e x p l a i n i n g s t o c k return.Furthermore,DurbinWatsonstatsarecloseto2whichmeanstheseregressionsonlysuffers l i g h t l y fro mautocorrelationissue.

EmpiricalResultsforAmihud’sIlliquidityMeasure

Table 13:Empiricalresult ofAmihud’sIlliquidityRatio Independent

*, **, ***representsignificant at1%level,5%leveland 10% level,respectively

InthiscaseofAmihudilliquidityratio,itcanbeobservedthatadjustedR- squared,S.Eo f regressionandSum ofsquareresiduals arequitesimilarbetweenallfourregressions.

TheregressionsofAmihudilliquidityratioshowedsomediscrepanciesintheirresults.T h e fi rstmodel,whichincludedallcontrolvariables,showedtheAmihudilliquidityratioiso n l y sig nificantat5%level.However,thesecond,thirdandfourthregressionsshowedthat

Amihudilliquidityratioishighlysignificantat1%level.JudgingbyR- squaredandDurbinW a t s o n stat,thefirstandthesecondregressionsshouldprovidethemostrelia bleresultsinallf o u r regressions.Overall,alloftheseregressionsdosupporttheroleofAmihudilliquidit yratioi n e x p l a i n stockreturns.

DISCUSSIONS

Inthisstudy,themainpurpose istoexamine theeffectofliquidity onstockreturn.

This study examines two key measures of liquidity: turnover and the Amihud liquidity ratio The findings reveal a strong negative relationship between turnover and stock returns, supported by all four regression analyses, which are significant at or below the 5% level This aligns with the main hypothesis of the study and is consistent with established liquidity research, particularly the work of Amihud and Mendelson (1986), who introduced the transaction cost theory This theory posits that investors demand a premium for illiquid stocks due to higher transaction costs compared to liquid stocks Additionally, investors tend to hold illiquid stocks for longer periods, which increases their risk relative to liquid stocks, leading to a higher premium for illiquid assets in their portfolios.

AnotherliquiditymeasureisAmihudilliquidityratio.Allofthefourempiricalregressions pointedtoasameconclusionthatAmihudilliquidityratiohasapositiverelationshipwithst ockreturns.Alloftheresultsarereliableastheyarealready treatedforcommonissuesi n p a n e l data(heteroscedasticity,autocorellation,multicollinearity… etc.).Despitethefirstregressionisonlysignificantat5%level,allotherregressionsaresignificantat1

%levelwhichisverystrongindicationofthisrelationship.BecauseAmihudratioisani l l i q u i d i t y measure,sothat theirexpected signwillreversewiththesignofTurnovermeasure.T h e rationalebehind this positiverelationshipissimilartoturnovermeasure.

Withverypersuasiveresultsfromtwoproxiesofliquidity,thisthesisconcludethattherei s a negative influenceofliquidity onstockreturnsinVietnammarket.

ThefindingsofnegativerelationshipbetweenliquidityandstockreturnsinVietnamisi n c o n t r a r y withfindingsinpaperofVoandBatten(2011),XuânVinhandHồngThu(2013).Int h e s e p a p e r s , theyclaimeda p o s i t i v e r e l a t i o n s h i p betweent w o v a r i a b l e s i n t h e c a s e o f

Vietnam.However,theresultfromthisthesisshouldbemorereliablewhenallthedataarecollec tedfromtrustedsource.Inaddition,alldataistreatedandfollowedstrictlytheprocedureo f FamaandFr enchinconstructingportfoliosandfactors.Furthermore,theempiricalresultfromthisthesisshowar easonableeffectoftwoliquiditymeasuresonstockreturnswhilethesepapersshowedextremelylargeors malleffectsofdifferentliquiditymeasuresonstockreturns.Lastbutnotleast, the result fromthisth esisis inagreementwith basictheoriesaswellasviewpointofwell- knownresearchers.Thisdemonstratedaveryimportantroleofselectingarightmethodologyfore achresearchastheoutcomecouldbeverydifferentdespitethesamet i m e frameandresearchlocati on.

Althoughtheprimarypurpose ofthisresearchwasn’ttoinvestigatetheeffectofFamaandFrenchfactorso n s t o c k returns,t h i s rese archalsodemonstratest h a t FamaandFrench(2013)’sfactorswereveryeffectiveinexplainingstoc kreturns.Particularly,SMBfactorhas1 % significantlevelinbothsetofregressions(regressionso nTurnoverandAmihudratio).T h e positivesignofSMBfactorwasalsoinagreementwithth eoryofFamaandFrenchinwhichtheyclaimedthatsmallfirmsshouldbemore vulnerabletovariousrisksthanbigfirmsd u e totheirrelativesmallsizeandundiversifiednature.

About HMLfactor,theresultwasveryencouragingwhen theirtstatisticin bothsetofregressionswasaround8.TheirpositivesignwasalsoinagreementwithFamaandFrenc hexpectationashighB/

Mratiomeansthevalueofcompanyinpublicmarkethasdecreasedduet o currentbusinessconditionan dlowexpectedfutureearnings.Therefore,investorsrequireane x t r a compensationabovenormal returnforinvestinginsuchcompanies.

Interestingly,t h e regressionresultfromt h i s thesisshoweda negativerelationshipbetwee nC M A factora n d s t o c k r e t u r n s T h i s meanst h a t s t o c k returnf r o m l o w investmentcompa nies(conservativecompanies)willactuallylessthanthestockreturnofhighinvestmentcompanies(agg ressivec o m p a n i e s ) T h i s r e s u l t divergesf r o m expectedsigno f Famaa n d Frenchwhent heyexpectedhighgrowthinbookequitycompanies(highinvestment)shouldi m p l y lowerex pectedreturn.The specificexplanationforthisresultmightbehighinvestmentfirmsinVietnamareexpectedtoprovid ehigherfutureearningsandmarketevaluatestheir s t o c k atahigherpricefortheirpotential.

ThelastFamaandFrenchfactorisRMWalsoprovedtobeeffectiveinexplainingstockreturnwithi tst- statisticaround10.ThepositivesignfromthisresearchisinalignmentwithexpectedsignfromFam aandFrenchtheory.InFamaandFrenchtheory,stockreturnsfromhigherexpectedearningfirmswo uldbemorethanreturnsfromlowexpectedearningfirms.A n d thismakesenseasinvestors wouldbe willingtopayhigherpriceforpurchasinghigh expectedearningstocks,thustheywillpushupthestockpricewhichresultinhigherstockret urnsforitsholders.

Accordingtotheempiricalresultofthisthesis,thecumulativereturnmeasuredidn’tp l a y a significantr o l e ine x p l a i n i n g s t o c k r e t u r n s asallt h r e e c u m u l a t i v e variables h a v e insignificantlevelofconfidence.Thissimplymeansthestockthatdeliversahighreturninpastw i l l notnecessarilydeliverahighreturnincurrentmonth.Theeffectivenessofthiscumulativereturnsme asureisquitecontroversialasempiricalresultsofthismeasureindifferentstockmarketyielded diverseconclusionaboutitseffectiveness.

OVERVIEW

ThemainfocusofthispaperistostudytheinfluenceofliquidityonstockreturnsinViet nammarket,afrontiermarket.Thesampledataforthisthesiswascollectedfromnon- financialfirmsinHoChiMinhCitystockexchange(HOSE)forperiod2007-

2013.Thedatawast h e n processedf o l l o w i n g FamaandFrenchmethodt o p r o v i d e i n p u t varia blesf o r t h i s t h e s i s TwomainproxiesforliquidityareTurnoverratioandAmihudilliquidity ratio.Thesep r o x i e s showedagreatconsistencyastheypointedtothesameconclusionwithhig hlevelofconfidence.Inordertoobtainareliableempiricalresult,FamaMacBethregressionmetho d,HeteroskedasticityandAutocorrelationConsistent(HAC)StandardErrorswereemployedtos o l v e commonissuesinfinancialresearchdata.

EMPIRICALRESULTS

The empirical test conducted in this thesis reveals significant findings regarding the relationship between liquidity and stock returns in Vietnam, a frontier market The regression results indicate a negative influence of liquidity on stock returns, which is supported by transaction cost theory (Amihud & Mendelson, 1986) This theory suggests that investors demand a premium for illiquid stocks due to their higher transaction costs compared to liquid stocks Furthermore, the negative liquidity effect is reinforced by Bekaert et al (2007), who argue that this effect is more pronounced in emerging markets The primary factor hindering foreign institutional investors from allocating funds to various stocks in these markets is poor liquidity, which exacerbates the liquidity premium Additionally, developed markets typically feature a diversified ownership structure with both long-term and short-term investors, further highlighting the challenges faced in frontier markets.

(2007)suggestt h e clienteleeffectsi n selectingi n v e s t m e n t portfolioshouldreducethepricingo fliquidity.Incaseofemergingmarketsandfrontiermarkets,thediversificationinnumberofsecuri tiesandownershipislackedofwhichoftenintensifythel i q u i d i t y effect.

Table 14:MainFindingof theThesisResearchQuestion Doesliquidityinfluenceonstock returns? YES

Secondly,althoughFamaandFrench’sfactorsaren’tthemainfocusofthisresearch,t h e re gressionresultsgreatlysupporttheroleofFamaand

French’sfactorsinexplainingstockreturns.Particularly,therearepositiverelationshipsbetweenS MB,HML,RMWfactorsands t o c k returns.CMAistheonlyFamaandFrench’sfactorwhichn egatively relatedtostockreturns.

Thirdly,thethesisalsotestthepossibleinfluenceofcumulativepastreturnsoncurrentreturn.H owever,theresultisnotverysupportivewhentwooutofthreecumulativevariablesshowedinsignif icantlevelofconfidence.RET46istheonlycumulativereturnvariablethathassignificantlevelat10%.N onetheless,thisresultdidnotkeepconsistentthroughallregressions.

CONTRIBUTIONS

This thesis contributes to the literature on asset returns by demonstrating a negative relationship between liquidity and stock returns in Vietnam, indicating the presence of a liquidity premium Empirical tests reveal a significant confidence level for liquidity proxies, with a liquidity premium of approximately 0.1% for every 1% of excess return This finding contradicts previous research by Vo & Batten (2011) and Xuân Vinh & Hồng Thư (2013), which suggested a positive relationship between liquidity and stock returns in Vietnam Additionally, the negative impact of liquidity on stock returns supports the claims made by Bekaert et al.

IMPLICATIONS

The research findings indicate that various social parties can benefit from understanding liquidity in stock markets Investors should be aware that illiquid stocks carry higher risks compared to liquid ones, and thus, investments in illiquid stocks should yield higher returns to compensate for the additional liquidity risk Additionally, companies with illiquid stocks tend to be either small in size or excessively controlled by large shareholders, resulting in a limited availability of shares in the market This situation increases the risk for non-insider investors To mitigate this issue, companies should impose limits on large shareholders to reduce excessive control Furthermore, the government and the State Securities Commission of Vietnam should implement new policies to enhance liquidity in the Vietnamese stock market, thereby attracting more prospective investors Suggested policies include shortening the transaction timeframe to T+2 and establishing upper limits on the percentage of company shares that can be held.

LIMITATIONSANDFUTURERESEARCHSUGGESTION

This study provides valuable insights into the relationship between liquidity, Fama and French factors, and stock returns in the Vietnamese market However, it faces limitations, including a narrow research timeframe of just seven years (2007-2013), which results in a limited number of observations and may affect the reliability of the findings Additionally, the presence of multicollinearity among explanatory variables inflates the variance and standard errors of OLS estimators, complicating the interpretation of results Furthermore, the study's control variables are restricted to common factors in stock return literature, and the intercept from empirical regressions is not equal to zero, indicating that the models cannot fully account for stock returns This highlights the need for further research to explore additional influential factors affecting stock returns in the developing Vietnamese market.

Ajili, S.(2003).Explainingthe Cross-SectionReturnsinFrance:Characteristics orCovariances?

Amihud, Y.(2002).Illiquidityandstockreturns: cross-sectionandtime-serieseffects.

Amihud, Y., &Mendelson, H.(1986).Assetpricingandthebid- askspread.Journaloffinancialeconomics,17(2),223-249.

Basu,S.(1977).Investmentperformanceofcommon stocks inrelationto theirprice‐ earningsratios:Atestof theefficientmarkethypothesis.Thejournal ofFinance,32(3),663- 682.

Bennaceur,S., &Chaibi,H.(2007).Thebest assetpricingmodel forestimatingcostofequity:evidencefrom thestockexchangeof Tunisia.SSRNPapers. Black,F.(1998).Betaandreturn.Streetwise:TheBestof theJournalof PortfolioManagement,74. Brailsford,T.,Gaunt,C.,&O'Brien,M A.(2012)

Theinvestmentvalueofthevaluepremium.Pacific-BasinFinance Journal,20(3),416-437. Brennan,M J.,Chordia,T., &Subrahmanyam,A.(1998).Alternative factorspecifications,s e c u r i t y characteristics,and thecross- sectionofexpectedstockreturns.Journal offinancialeconomics,49(3),345-373.

Brennan,M J., &Subrahmanyam,A.(1996).Marketmicrostructure andassetpricing:Ont h e compensationforilliquidityinstockreturns.Journal of financialeconomics,41(3),441-464.

Carhart,M M.(1997).Onpersistencein mutual fundperformance.Thejournal ofFinance,52(1),57-82.

Choe,H., &Yang,C.-W.(2008).ComparisonsofLiquidityMeasures in theStockMarkets.

Chordia,T., Roll, R., &Subrahmanyam,A (2000) Commonalityin liquidity.Journaloffinancialeconomics,56(1),3-28.

(1985).Liquidity,exchangelisting,andcommons t o c k performance.Journal ofEconomicsandBusiness,37(1),19-33.

Daniel,K.,&Titman, S.(1997).Evidenceon thecharacteristicsofcrosssectionalvariationi n stockreturns.Thejournal ofFinance,52(1),1-33.

Drew,M (2003).Beta,firm size,book-to-marketequityandstockreturns.Journalof theA s i a PacificEconomy,8(3),354-379.

(1996).Liquidity,information,andi n f r e q u e n t l y tradedstocks.Thejournal ofFinance,51(4),1405-1436.

Eleswarapu,V R (1997) Cost oftransacting andexpectedreturnsin theNasdaqmarket.Thej o u r n a l o f Finance,52(5),2113-2127.

Eleswarapu,V R., &Reinganum,M R.(1993).Theseasonalbehavior oftheliquiditypremiuminassetpricing.Journalof financialeconomics,34(3),373- 386.

Fama,E.F.(1965).Thebehaviorofstock-marketprices.Journalof business,34-105.

Fama,E.F.(1970).Efficientcapitalmarkets:A reviewof theoryandempiricalwork*.Thej o u r n a l o f Finance,25(2),383-417.

Fama,E.F.,&French,K R.(1992).Thecross‐section ofexpectedstockreturns.Thejournal ofFinance,47(2),427-465.

Fama,E.F.,&French,K R.(1993).Common riskfactorsin thereturnsonstocksandbonds.

Fama,E.F.,&French,K R (2013) Afive-factorassetpricingmodel.Fama-

Millerworkingpaper,Universityof Chicago, DartmouthCollegeand NBER.

Fama,E.F.,&MacBeth,J.D.(1973).Risk,return,andequilibrium:Empiricaltests.TheJournalof

Jensen, M C.,Black,F.,&Scholes,M S.(1972).Thecapitalasset pricing model:

Jun,S.-G.,Marathe,A.,&Shawky,H.A.(2003).Liquidityandstockreturns inemerginge q u i t y markets.EmergingMarketsReview,4(1),1-24.

Kyle,A S.(1985).Continuousauctionsandinsidertrading.Econometrica:Journalof theeconometricsociety, 1315-1335.

L’Her,J.-F.,Masmoudi,T., &Suret,J.-M (2004).Evidenceto support thefour- factorpricingmodelfromtheCanadian stockmarket.Journal of

Lam,K.(2005).Is theFama-Frenchthreefactormodelbetter than theCAPM?,Departmento f Economics-SimonFraserUniversity.

Lam,K S (2002).Therelationshipbetweensize,book-to-marketequityratio,earnings– priceratio,andreturnfor theHongKongstockmarket.GlobalFinanceJournal,13(2),163- 179.

Lam,K S.,Li,F.K.,& So, S M.(2010).On thevalidityof theaugmentedFamaandFrench’s(1993)model:evidencefrom the

Lintner,J.(1965).Thevaluation ofriskassetsand theselectionofriskyinvestments instockportfoliosandcapitalbudgets.Thereview ofeconomicsand statistics,13-37.

Liu,W.(2006).Aliquidity-augmentedcapitalassetpricingmodel.Journalof financialeconomics,82(3), 631-671.

Miller,M H., &Modigliani,F.(1961).Dividendpolicy,growth,andthe valuation ofshares. theJournalofBusiness,34(4),411-433.

Mossin,J.(1966).Equilibriumin acapitalassetmarket.Econometrica:Journal of theeconometricsociety, 768-783.

Newey,W K.,&West, K D.(1986).A simple,positivesemi-definite, heteroskedasticityandautocorrelationconsistentcovariancematrix:NationalBureauofEc onomicResearchCambridge, Mass., USA.

Novy-Marx,R.(2013).Theotherside ofvalue: Thegross profitabilitypremium.Journaloffinancialeconomics,108(1),1-28.

O'Brien,M.A.,Brailsford,T., &Gaunt, C (2008).Size andbook-to- marketfactorsinA u s t r a l i a Paperpresentedatthe21stAustralasianFinance andBankingConference.

Petersen,M A (2009).Estimatingstandarderrorsinfinancepanel datasets:Comparingapproaches.ReviewofFinancialStudies,22(1),435-480.

Pujari,S.(2010).Someimportantfunctionofstockexchange market.Retrieved25May,2 0 1 5 Roll, R.(1977).Acritiqueof theassetpricingtheory'stestsPartI:On pastandpotentialt e s t a b i l i t y ofthetheory.Journalof financialeconomics,4(2),129- 176.

Roll, R.(1984).Asimple implicitmeasureof theeffectivebid‐askspreadin anefficient market.Thejournal ofFinance,39(4),1127-1139.

Sharpe,W.F.(1964).Capitalassetprices:A theoryofmarketequilibriumunderconditionso f risk*.Thejournal ofFinance,19(3),425- Subrahmanyam,A (2010).The Cross‐442.SectionofExpectedStockReturns:WhatHaveWeLearntfrom thePastTwenty‐

XuânVinh, V.,&HồngThu,B.(2013).Liquidity,LiquidityRisk and Stock Returns–

EvidencefromVietnam.Bùi,Liquidity,LiquidityRisk andStock Returns–Evidencef r o m

Forbetterunderstandoftheprocedure,wefirstassumedthatwehavenassetsorexcessreturnportf oliosoveraperiodoftime(Ri,t).T h e p o t e n t i a l explanatoryfactorswillbedenotedasF1,t,F2,t,

….,Fm,t InthefirststageofFamaMacBethregression,eachassetorportfolioisregressedagainstthepr oposedexplanatoryfactors.Therefore,anequalnumber oftimeseriesregressiontothenumberofassetsorportfolioswillberun.Anillustrationoftheseequat ionsaredepictedbelow:

Afterregressingthesetimeseries,theinfluenceofeachexplanatoryfactorontheexcessreturnport foliosisrecognizedindividually.Thisisduetothefactthateachtimeserieswasperformeds e p a r a t e l y w i t h t h e samen u m b e r o f e x p l a n a t o r y factors.However,therewardpremiumt oeachfactorexposureisstillnotrealizeduntilasecondstageofFamaMacBethRegressionisim plemented.

 regression.  i,Fkis anestimateofβisforeachassetorportfolioforeachFriskfactor.The differencebetweenthesetwovariablesisthatisatrueunobservablefactorloadwhile  is anestimationofthetrueβiandderivedfromtimeseriesregressioninfirststageo f FamaMacBethr egression.Thefactorpremiumforeachfactorisfoundedbyregressingthef o l l o w i n g setofcross– sectionalregressions:

The iskeptunchangedforeachportfolioinallthecrosssectionalregressionsfrom thefirsttothefinalcrosssectionalregressions.Theonlydifferenceisthechangeofeachasseto r portfo lioreturni n eachcrosssectionalregressionoverdifferentt i m e period.T h e r i s k premiumforeach factoristhencomputedbyaveragingalltheγj,ttermintoasingleγj.Thestandarderrorforγjiscalculate dbyconsideringeachγj,tasanindependentcomponentandcomputeitst–statistic(thet– statisticofwhetherγjisstatisticallydifferentfromzero)withbelowformula:

WHY FAMAMACBETH TWO STAGE REGRESSION IS EMPLOYED:

Theγkcoefficientrepresentsfactorpremiumforanaverageexposureto i,Fkof  1.The questionaroseisthatwhywehavetousesuchalengthyanddemandprocedure insteadofu s i n g onlyonesinglecross– sectionalregressionthatinvolvetheaveragereturnofallassetso r portfoliosovert im e Ifonlyone singlecrosssectionalregressionwasi n use,w e wou ld expecttohave belowcomparativeequation:

E(Ri,t) =E(ɑi)+E(βii,F1F1,t) +E(βii,F2F2,t)+… +E(βii,FmFm,t)

Ri=ɑ+βii,F1E(F1,t)+βii,F2E(F2,t)+…+βii,FmE(Fm,t)

Ri=ɑ+γ1βii,F1+γ2βii,F2+…+γmβii,Fm

Noticethat ɑisequaltoE(ɑi),γkisequalto E(Fk,t),Rii sequalto E(Ri,t).

Accordingtothederivationofaboveequation,itcanberecognizedthattheγkisequalt o E(Fk,t) (themeanorexpectedvalueoftradablefactorisequaltofactorpremium)whichist h e wholepurpo seofourlengthyanddemandtask.Thereasonforusingsuchaprocedureist h a t asthesamplesize approachestheinfinity,thetruemean(ortrueexpectedvalue)ofthefactorw i l l b e a pp ro xi ma t e ly thesamplemeanoft h i s factor.Nonetheless,a sufficientt i m e periodtoderivethetrueexpected valueofthefactorisunknown,therefore,Fama-