Rochester Institute of TechnologyRIT Scholar Works

Articles

2004

Price-volume relationships and stock returnsSteven Gold

Follow this and additional works at: http://scholarworks.rit.edu/article

This Article is brought to you for free and open access by RIT Scholar Works. It has been accepted for inclusion in Articles by an authorizedadministrator of RIT Scholar Works. For more information, please contact ritscholarworks@rit.edu.

Recommended CitationGold, Steven, "Price-volume relationships and stock returns" (2004). Journal of Accounting and Finance Research, 85-94. Accessed fromhttp://scholarworks.rit.edu/article/295

PRICE-VOLUME RELATIONSHIPS AND STOCK RETURNS

Steven C. Gold, Rochester Institute (IfTeehnology

ABSTRACT

It is popular among technical analyststo use high trading volume as a positivesele<:lioll or filter criteria. Yellhe findings inthe finance literature are no! clear on thepredictive validity or even the diroction ofthe impact of tratling volume on stockreturns. One stream of finance researchfinds that high changes in lrading volumeare associmcd with information asymmetryor differences in beliefs between traders,suggesting stock price reversals and returnvariances are higher with high tradingvolume. A secoml stream of research findsIhal high trading volume is attributed tomfomled trading, suggesting stock pricereversals and return variances are lower wilhhigh trading VOllilllC, A third stream ofresearch, modem portfolio theory, rejects thepredictive validity of using past infonnalion.In this study, an alternative hypothesis isdeveloped using an intuitive market demandand supply model. supporting the hypothesisthat large price reactions coupled withnonnal tmding volume are less likely to bereversed and arc more stable lhan in the caseof high trading volume These hypothesizesarc tested empirically and have importantimplications for investment analysts, and thecontroversies surrounding the meaning oftrading volume,

INTRODUCTION

Finallcial economists have beenstudytng the relationship between stockreturns and lrading volume for mallY years.Of particular interesl to investors is whetherinfonnation about trading volume is usefulin he1pmg forecast stock returns, It iscommon to see slock charts studied by

inveslors displaying stock prices in the topportion of the chart and volume al thebollorn, Many technical analysts use hightrading volume as a eritcrion to filter andselect stocks \\'ith promising returns. Yet,there is much controversy concerning thesignificance and predictive validity oftrading volume. According to theefficient market hypothesis, past price orvolume changes in a competitively tradedfinancial market do not help predict futurepnces. However, recent studies havequestioned the efficient market hypothesisand have supported the notion that stockmarket excess returns can bc predicted bypublicly available infonnation (eg., seeFama & French, 1995; Pesaran &Timmennan, 1995; Ferson & Harvey, 1993).With respect to trading volume, Lee andSwalUllinalhan (2000) stated: "Thefacllhata market staliSlic widely U5ed in technicalanalysis can provide informalion aboulrelalive under- or OVer-l'(J!ualio1l issurprising and is difficult 10 reconcile withexiSling Iheorelical work." (p.2019) Yet.their study finds that past trading volumecan be used to prcdiet future stock pricemomentum. Chen, et al. (2001) also findsvolume to be useful 1Il predicting changes instock prices and volatility. Gervais, Kaniel,and Mingelgrin (2002) test and eonfinn aninvestment strategy using trading volumethat realizes positive economic profits.

The objective of this study is 10further examine and test the validity of usingtrading volume to forecast stock returns.This paper extf.>nds the work in this area Inseveral ways, First. different schools ofthought are categorized in teons of the wayin which volume arfeets and predicts stockreturns, Second, the impact of low, normal,and high volume associated with high,

-

stable. or low relUrns are ellamined. High(low) volume changes associated willi high(low) mums are bypothesized to havedifferent afftets than "illi nonnal returns.Third. an imuiti\'e demand and supply modelof !he market is used to explain price­\olume relalionships_

The p~ proceeds as follOW$. Fim!he relevant lito:rarure is reviev..cd. Second. •set of hypothesizes iIIl: formulated and !hetcsling melhodology described. Third, thecmplncal results iIIl: presented and analyzed.The final section discusses the implicationsand am1S of futUTC research.

LITERATURE REVIEW; VOLUMEAND PRlCE REACTIONS

Early research has found price andvolume reactions to be similar, andconsequently, have considered price andvolume as substitute measures of markelreaction. Several studies dating back 10 asearly as 1970 demonstrate thaI there is apositive correlation between stock returnsand daily trading volume (see Crouch(I970), Clark (1973). and Wood, Mclnish,and Ord (1985)). Positive correlalJOns werealso found betv.-een the ,-ariance of stockmums and tradmg \'olume (set Epps andEpps( 1916). Morgan (1916), Westerfield(1911). Taugo:n and Pins (1983)), ThemtphcauDn 15 that \'olume docs I10t pro\'ideaddiuonal infonnation from stock pnces.Also, the suggestion that eitlln hIgh or lowtradmg volume would have any predictive\"IlidllY IS inconsislCJlt WIth the cfficlmtmarket hypothesis.

More rccmtty. sludies hay!; :>/lU"lltradmg volume to behave quite diffCTmtlythan stock pnee mo'-cmeTl1S and doesprovide different information_ ConceptuallyKim and Vem:cchia (1991) bave arguedprice changes are associated with themarkel's average beliefs. while lrading

volume is the sum of all individual trades.Hicmslnl and Jones (1994) provideempirical suppan that more can 1M: learnedabout the stock market by examining bothstock pnces and \1)lume than byconccnlnlting only on price dynamics.EvidaJCe of non-hnear causality fromvolume to stock retums is found. Volumeserves as a proxy for information flow. andthere is a positive autOCOl'relation betweenuading "olume and absolute stock rerums.Bamber and Cheon (1995) dtmOfl5tr.1te thalearmngs announctmmts that cause hightrading volume with small price changes arcfollowed by price increases. Stickel andVerr«chia (1994) present e"idence thatprice changes are more likcly to be re,·erscdfollowing low tradc volume than highvolume. They argue high InIding volumeindicates that the increase in demand comesfrom informed investors. High volume isclaimed to cause information fueling anddiffusion (Daniel et al. (1998); Hong andStein (1999). Informed trading implies thatprice changes arc less likely to be reversed.In contrast, low volume is associaled withuninformed or liquidity motivated uading.Therefore, price changes with low volumeare more likely to be reversed. b«ause theyresult from some temporary effect wluch isnoI accurately relaled to the mformation_Consistent willi this study. Gervais, Kamel.and Mingelgrin (2002) find that stocks withhigh (low) tradmg volume OV£I'" a day orweek tend to appreciate (depreciate) over thenext month. It is hypothesized that the: high­volume premium is a result of the: stock'sincreased visibility. (Referred to asmformation fueling or diffusion in Lee andSwaminathan (2000»

These COnclllSions iIIl: mconsistell1with several studies showing that highvolume is associaled with differences ofopinion among investors. Bamber (1995)found price and volume reactions to be very

-

different for 20-24 percent of the sample of8.180 earnmgs announcernnus betw«n1986 to 1989, Tradmg \'olume reaction wasshown to be relau\'ely high (compared to

price reaction) when an announcementcreated a greater dIvergence of opinionbetween individual investors. The studyconcluded eamillgs announccmelliS thatcause high volumc relativc to pncemO"ements arc related to: (I) moredlvergem earnings forecasts by analysts; (2)large number of analysts; and (3) higherrandom-walk unexpected eammgs relati\'e10 analysts-based unexpected earnings.Harris and Ravi ... (1991) suggest that even ifall tr3den are homogeneously informed,differences in opinion are the only f..tor­explairung high trading volume. Thesefindings are eorunstent with many paststudies indicatmg volume reatlion is directlyrelated to the degree of asynunetricinformation or diffcrenccs in beliefs. ThisImplies that trading volume would berelatively low if there were no dIfferences inbeliefs among traders (see Kim &Verrccchta, 1991; Holth.usen &Verrrecchai, 1990). Campbell, Grossman,and Wang (1993) find that price changes...ith high nading volume will more likely bere\'ersed than with low InJdmg volume.Kramer (1999) argues that high tradingvolume is a source of risk because itincreases the traden' marginal transactioncost. Lec and Swaminathan (2000) find thatfirms with high (low) volume expcricnccsignificarttly lower (higher) future returns.The reason given is based on investormispercepllOns about future earnings.Analysts gi\'e lower (hIgher) long-termeilJf1Ings gro...th forecasts for low (high)\-olwne stocks, but firms with lou' (high)trading volume experienee slgruficantlybener (worse) future operat,"g performance.

Low trading volwne has also beenslwwn to occur with asymmetric

information. George, Kaul, and Nimalendran(1994) dernonstnlle that trading "olume cartbe negam'dy related 10 the degree ofmfonnation asymmetry in a speciaJistmartel with endogenous transaction costs.This is suppor1td by an eacher study byBlack (1986). More recently. a study byChae (2002) shows that decreases in tradingvolume occurs before earningsannouncemertlS duc to infonnationasymmetry. This would imply that pncechanges associated WIth low trnding volumewill more likely be fC'\'ersed owing 10

information U)mmeuyOf low visibility,

DEMAND "'....D SUPPLY MODEL

Tlte different schools of thought andtheir price-volume relationships can beexplained intuitively with a simple demandand supply analysis. The demand andsupply model is also applied to devclop anallernative view that is referred to as the"S}mmetrie information" hypolhesls. Withsymmetric information it is assumed there ISconsistency in beliefs among investors.

AS)'mmetrie Illformalioll

Demand and supply analysis will fimbe used to illustrate gr;Iphically price andvolume relationships given differentialbeliefs between traders, and lIlert contrastedwith symmetric beliefs. Suppose aneconomic evem causes an optimistic marketdcmand reaction for a stock, as shown inFigure I by a shift in demand from OJ to02. Assuming no changc in supply, bothpric.e lU](\ Irading "olumc incrcQC frompoint A to B. But if stock ownen Interpretthe evem pesslmlsllcally. and sell stock,IIJCrcasclng supply from 51 10 S2, pornt Cbecomes the new equilibrium. In thiS case, adivergence of opinion between buyers andsellers causes a relatIVely large increase in

-

trnding volume. The extent of the pncelJlCrUSe depends on the relath·e shIfts indemand and supply.

FIGURE IDemand & Supply with As)mmetric Beliefs

,,

VI V2 V3

Trodlng Vot""",

An intriguing question is why different~l1d opposing strategies occur betweenbuyers and sellers? Andreassen (1988)argues that different people pay attention todifferent aspects of the same information,and that the same investment role causesdifferent results if applied to one aspect asopposed to anolher. For example, whenstock price rises. some people may belie,'ethat il tS roow abo,'e the a,·erage and II willsoon fall. bUI olhers focus on price "elocityand fann posItive expectations about IheSIOC)e'S future. It has also been argued thaio....ners of stock may be more opllmistic thannon--owners.

S,-mmtlrie Information

Now consider price and volumereaclloos .",~"mm& the same beliefs belweentnders. Suppose an economic evenl cauSC!lan optImIstic market demand and supplyreaction for a stock, as shown in Figure 2 bya simullaneous upward shift in demand tothe right, DI to D2. and supply to the left.Sl to 52. The new equilibrium moves frompoint A to point C. In this situation,

relath'ely high price changes occur with nochange in volume. i.e. volume remains atnormal levels. This point of view suggeststhat a large incruse in price coupled withnorma1 tnding volwne Sterns fromagreemenl of opinion. Conslslent wllh thelilenlwe on asymmetnc information,agm:rnrnl of OJllnion suggests there is lessuncertainty in the informalion provided.

FIGURE 2Demand & Supply with Symmetric Beliefs

D'

"

Tradina Volu",<

S)mmetric beliefs imply that pricechanges (high or low) with normal tradingvolume are less likely to be reversed owingto greater consistency (less uncertainty)surrounding the informanon.

IlHPLlCATlOJl;"S OF PAST RESEARCHA!\'"D HYPOTlIESIZES

lbe litenture review provides a basis10 categories the pasl researt.h findings intothree schools of thought with COlTeSpOndingimplications in termS of the impact oflnIding ,-olume on fu'ut<l returns. The threeschools of thought are referred to as: fullinformation. asymmetric information, andefficient markets. The asswnption ofs)mmetric beliefs, evaluated with a demandand supply model in this paper provides afourth testable point of view.

~ "Full Information~ schoolsuggests that high trading \<olume isassociated with higher Ie-'els of ."'wencss.visibility. and informahon dIffusIon andfueling. Based on this school Qfthought it ishypothesized that price changt5 llSS()Ciatedwith high volume will be maintained and theprobability of a price reversal and variancesill relUms will be less than in the case ofnormal or IQW volume.

The "Asynunetric Information cauo;inghigh \'ol~- school suggeslS that high\'olume is due to differences Qf QPinion ordisagreement in beliefs. Many studies arguethat diffeTmC:e5 in beliefs creatc uncauinly,

causing trading volume to increase. Basedon lIIis school of IOOUght it is hrpothesiudthat price changes associated high volume",ould more likely be re\erscd and ha\'ehigher variances in returns than in the caseof normal or low volume_

The "Asymmetric Information Cllusinglow volume" SChool sU&gesls lIIat lowtrading \'olume is created by differences ofopmlOn Qr disagreement In beliefs.Accoromg to these studies. the un<:ertaintyc;luset! by dlffermees of opllllOll wooldreduce tradmg, not increase il. Sued 011 thisschool Qf thought it is h)pothesiud tha1price changes associated low \'Qlume wouldmore likely be reversed and have highervariances in returns than in the case ofnonnal or high volume.

The "Symmetric Infonnation"hypothesis developed using a demand andsupply model in this paper argues thatconsistent beliefs betwccn traders wouldcause trading \<olume to remam at nQnnallevels. Consistem willi the llter1.lUre 011

asYJl1lIldrie Informauon, agreement ofIlpmlon suggests there is less uncenamty int~ mfonnalion provided Less W1Cenamtywould lead to the hypothesis that a pricechange (hIgh or low) wlIh IIOrmaJ tnldingvolume is less likely to be reversed and have

!ov>'CT variances in returns (relative tQ tughor m' volume) Qwing to symmetric beliefssurrounding the mformauon.

The -Efficient Market~ school ormodem ponfolio theory suggeslS tbat neitherhigh nor low trading volume would have anypredictive validity, Modem ponfolio theorycontends that daily stock price changes arerandom and cannOt be predicted by pastinfonnation. The future is unknown andstock prices change "ery quickly to companydisclosutCS, pUblic news releases, and othereconomic events.

Table I summarizes the four testableh)plHbesize:s ronc:eming the impact oflrading volume.

DATA AND RESEARCHMETIIODOLOGY

The database of the Center forRescarch in Security Prices (CRSP) is used,consisting of listed stocks on the r-.rySE,AMEX, and NASDAQ exchanges over thepast ten years. For twing and controlpwposc:s. !he database is limited to theNYSE. including about 2.600 li5led stocksin the database. Tl\c: 1D\'estment models an:tested for two $eleclN years: 1989 and 1995.This lime period encompasses bo!h a sablemarket and bull market.

In each of the chosen ~:us. 12 tradingdays are selected at random. one from eachmonth, On each random day selected, a sctof stocks is chosen that best fit the followingthree categories: (l) high positive ornegative price perfonnance with high'<olume; (2) high posit;"e or negatl\·e priceperformance with low volume; and (3) highpositi\'e or negall\'C price paforman<:e withncmnal volume

Price performance is measured as the Iday total return In stock price. Comistenlwi!h Campbell, Grossman, and Wang(2001), volume is measured by turnover,

which is the ralio of tho: number of shan::traded to the number of oUlstanding shares.

High (low)

Fullinfonnauon Hypothesis'

TABLE ITestable- II ypothtsi:a:es

LArge pn« changQ (~/-) Q1'f' Ins (morr) likely 10 ~

'"'t"ed alld have lowrr (AigMr) vonallCC! III I'f'funu withhI!!" (low) lradlng vofruu..

Asymmetric Information Hypothesis: Large pri« chQJIgrs (+/.) are t!JllU. ilk.ely to ~ rrversedand haw: ",gher (lo ..-er) vorlan« 111 rerurnJ WIth rilnrrfuKl!j+/-) Iradlng 1"Olumr.

Symmetnc Information Hypothesis: lArge prier chunges (+/.) are IDJ. likely to be reversedund have~ (higher) \'Qrianer in relurns wilh nQrmal(rading ,'Olumr comparrd 10 high lrading vol,m,e.

Efficient Markel Hypothesis: Past (railing .'Oilime has no offrci 011 flltrlre slock pricere~'ersuls or mrionce.l i'l relUfllJ.

stock price performance is defined as oncstandard dcviation above (below) Ihe meandaily return. The high (low) performers arethen separated by volume lumo,er Ic'cls:high, low. or normal. High (low) "olume isdefined as one slandard deviation abo"e(below) the mean daily volume measured bytumo,·cr.

After the three ponfolios are selecled,tho: lolal return of each S10ck m the neXItradmg day IS used to detcrmine lhefrequency of pnce ~ersals and the v:lnaneeof the rcturns in each portfolio. In Ihe caseof Ihe high positive (ntgalivc) pnceperformancc portfolio, a rf:'l"t'rsal IS definedto occur if the rerum changes direcuon frompositive to negative. or Vice-vena.

RESULTS AND DISCUSSION

Tables 2 and 3 summarize the resultsof the study for Ihe years 1989 and 1995.The year 1995 was a growth year for the

market, while 1989 was a stable year withliule growth in prices. The data issegmented in 1\0,'0 pans: high stock returns(·1-) with normal volume on day I versushigh slock .-elums with high \'olume on dayI. The focus is to compare the reversals andreturn variances on the second trading day(day 2) of the high return normal volumegroup with tho: high return Irigh volumegroup. The eornprehensive New York Stockhehange (!\'YSE) results ser....e as abenchmarl<.

TIle first lest compares the next dayreversals and return varianeClil or the nQmlu/volume group 10 the NYSE. In Table 2(1995), out ortho: 12 selected tradilli days, 5next day re~"tTSa15 were stahSl,callysignificant, but 3 were greater and 2 werelower than the NYSE. There was noconsistent direelJon, and aggregate reversalsfor 1995 were not sigmficanl. In Table 3(1985), 4 next day reversals wcre significantwuh 3 be,ng greater than the NYSE. But

-

again, the aggregate reversals for 1985 werenot sigJllficanl. However, next da)' WJrlDllCQ

lIere highly slgnificam and greater !han theNYSE ...ariances in almost all cases for both1989 and 1995. Aggrcganng atl tradingd.ys. the return variances in the tugh returnnormal volume group were significantlyhIgher !han the NYSE for both yean 1989and 1995.

The sceond lcst compares the ne;l:t dayl'('\'c:rsal.s and return variances of the lughvol~ 10 the llQrllWI volume group. InTable 2 (1995) only 2 out of the 12 reversalswere statistically significant at the 0.05level. In Table 3 (1989), only lout of the12 re,·ersals were significant at the 0.05level. Howe...er. nC;l:t day rcturn variances ofthe high volume compared to the normalvolume group were again statisticallysignificant in most cases. In 1995, 10 out of12 variances were statistically significanland greater lhan the variances in the nomlalvolume group. Similarly, in 1989, 8 out of12 variances wcre statistically significant.Of the signifiClllt variances in the highvolume group, 6 oUi of 8 were greater thanthe variances 111 the normal \·olume group.Aggregating all trading days, the retumvariances in the high volume group weresignificantly higher than the normal \"Olumegroup in both 1989 and 1995.

BNDI:'\'GS AND CONCtUSIOiliS

These findin~ bring into quesuonthe cOlTlmonly held beliefby many tcchnicalanalysts that tugh trading '·olume IS •positIve mdicator of fu~ rerums. The fullInfOnnatlon h)'JlOthcsls suggests that hIghstock pnee returns associated WIth !J.Jz.b.trading volume should be more stable andexhIbit less freqUC11t reversals. The data isnot consistent lI'ith this hypotheSIS. TherelIere no slglllficanl differences in thefrequency of I'e\'crsals and, to thc contrary,

the vanances in stock returns with hightrading \'olume were significantly greaterthan with normal volume

The asymmetric informationh)JlOlhesis is consistem WIth the sampledata, supporting the notion thai. hIgh volumeis a signal of disagreement In the market,causing greater uneertainty. Although stockprioe re\'ersa!s sbowed no significantdifferences, the variance in returns ....oeresignificantl)' higheT when slOCk pricesincreased with high volume as opposed to

normal volume. This also supports thes)mrnetric information argument that highreturns with nonnal trading volume IS asignal of agreemenl in the markel, andshould cause greater stability and lessvariance in stock returns.

Man)' financial analysts use tradingvolume activity as a positive screening orfiltering tool 10 sclect OJld predict stockrcturns. The common opinion amongpractitioners in the field is thal high lTadingvolume confirms that high pnceperformance is associated with informeddecision-making that is widespread in themarket and that this is a good climate forinvestmenl. Yet. the findings in this study donot suppan this belief academics in the fieldof fmance disagree on the meaning oftnding v"Olume. as evidmced by thediversity of opinions in the rese3fcllliterature.

The findIngs In lhis study arcImportant 10 tWQ ......ys For practicinginvestment analysIS, a warning flag is raisedin terms ohiewmg tugh tradm8 volume: as apositiv·e signal. Second, for academics, thestudy further suppans the notion that hightnoding volume is a signal of disagreementwith respect (0 the interpretauon of newinfonnation. Consistent with this view, highstock price returns, coupled with normalvolume, implies greater agreement and lessullCcnainty (or more stability) in the market.

TABLE 2Rt\trsal$ and Return VariancC!I of PonroliO$ wilb High Rtlurns (+/-) 00 Day I

u Statistical Slglllficallce at the 0.05 level or greater, Slg111ficance at 0.10 Ie' el.

Stllmtnlcd b\' Volume in Year 1995Normal Volume- Oa I iii h Volume - Da I ~'VSE - all slocks

Nelli Day NellI Day Nexl Day Nexi DayDates No, Reversals Variances N,. Reversals Variances Reversals Variances1/12-13 ll8 41.62% 0,0012u 47 36.17% 0.0043-- 41.93% 0,00064

1-0.12) (I.S9) 1·0.76) 13.57)122-23 '" 42.27% 0,0012u

" 37.93% 0.0027-- 41.11% 0.000511044\ 12.41\ 1-0.6 {2.23\

} 4-5 2S8 43.41% 0.0013-- 41 43.90% 0.0021-- 40.23% 0.00051/L~\ /2.45\ /0.06\ 1l.68'

4.1J.19 367 41.23% 0.0013u 32 50.82% 0.0023u 38.94% 0.0005411.29\ /2.49\ /0.89\ li.7i\

5110-11 '" 41.61% 0.0003-- 61 49.18% 0.0023u 4O.7rh 0.00051(Om '1.49) '1.20) '2.98'6113_14 '68 42.54%" 0.0011" 71 42.27% O.OOOS-- 44.81% 0.00040

(-0.75) (2.87) (-0.05) (US)712fr17 '" 49.77%-' 0.0014u 24 ~!,67~ 0::"::;' 43.98% 0.00052

".", - '12. 721 -0.79 2.60&18·9 33' 38.02%- 0.0007-- 74 44.59% 0.0019-- 42.86% 0.00048

(-1.79' - 11.561 fU61' ".so,9'5-6 422 4455%" O~;;' 41 24.39%-- 0.0017-- 39.43% 0.00039

(2.15-\ 1.82 -,:,.0, r2.'"10124_25 247 47.7:~· Ooo13u

" ~.62~- O;':Jl,;;' 42_15% 0.00052{1.79 -1',4, -1.90 1.43

II 21·22 '27 37.24·,_··O.~~~~· 16

3(~5~~ ~i~~~42.2S-h 0.00050

-(~2.lll 2.34 002 1.2212.6-7 "0 4f~·2~~ O~;;' 81 ~.5~~-

O.OOllu 42_38% 0.001040.78 1.40 -1.75 -iJ .45)

AUrtz·' 4,319 4;i9~3~· O.~Oll~~· 59' 3{~.8~~. O·~IOI:;~. 41.76% 0.00055• 1.53 1.95 -1.53 \.82., .

TABLEJRe"cnals and Reluni Vlrilnus or portrolios ....ilh High Rl'Iurns (+/.) on Dar I

Stal\$'Uca! slgmflCance all:he 0.05 level or greater; Slgmfteance at 0.10 ICl'el.

S_menltd b,· Volume in Yur 1989Normal Volume - DI 1 f1j b Volume Do 1 IIo'YSE all SIOC;ks

Nc:x1 Diy Ncxl Day NUl Diy Nal DiyDates No. Re>'ersa!s Variances No. Revasals Variances Reversals VariancesI '24_2~ 227

, 47.58%·· 0.0015· 2 0.00% 0.0027 40.54% 0.0006(2.1 (2.62\ (·1.17\ (1.85\

2:2.) 226 45.13% 0.0014·· 14 35.71% 0.0014 43.40% 0.0005(0.5)\ (2.6\1 (·0.71) (1.02\

'M 173 40.46% 0.0020·· 8 62.50% 0.01100· 41,1:Z-;' 0.0005(".17) (l.70) 0.2 (lAS)

412·1) 215 45.12% 0.0013·· 17 58.S2% 0.01 IS·· 45.95% 0.0008(·024) 0.70) 0.14) (9.26)

~ 17.18 204 43.14% 0.0025·· 12 50.00'l'. 0.0018 39.5",. 0.0010(104' (2.4(1 IOAS) (136)

6120-21 223 46.64% 0.OO14 u 1 0.00% 00000 42.26% 0.0005(1.32) (2.89) (-0.93) (n,a.)

7126·27 214 3~i4~~. 0;~28\· J7 1~:92~~" 0.0008·· 30.61% 0.00161.85 1.38 -2.22 . 12.6.)

8/IS.16 "3 44.26%"" O.~3~~· 4 5(~.0?;;. O·~51~~· 59.930/. 0.0001r~321 4.10 0.23 5.29

9127·28 2.7 ~~.O~:; O,~~:~· IS 5(~.3~~ O.~I,~~. 43.70"/0 0.0009-0.48 4.18 0.89 2.94

10110-11 209 40.19% O.~I~~· 34 5~.9~~ O.~~~;· 36.36% 0.0007(1.1'5)- 2.11 1.52 3.50

11'28·29 23' 4~;0~~" O.~~~" 26 3~.77~. 'O.~;~· 41.00% 0.00071.86 4.27 -1.66 1.42

l1i19·20 239~~.~~

0,0024"· 19 ~=.II~ O.~~~~. 40.46% 0.0009-;2.7 -0.12 1.26"urtt't ,~.. 4J.5~;' O.:~OI:~· ". 40.74~ I O.~~~· 42.07% 0.11008• 11.54 2.14 1-0.79 2.60•• . .

AL,.HORS' ,,"OTE REFERENCES

RllUI SlIlghvi. a graduate student In

Ihe MBA program at the Rochester Inslituteor Technology. must be acknowledged rorthe Slgnifiealll conlributions he has made tothiS study wilh respect 10 lhe analysis or theCRSI' database.

Andreassen, P. 1988. Explaming Ihe price­volume relationship: The dilTcrcncc:between price changes and changingprices. OrgafllzaliQllal Behavior andHuman Decision PrOCeSSI!$ 41: 371·389.

Bamber, S, and Y. Chcon. 1995DiffeTtnual price and \'Olume reactionsto accountmg eanungs anl\OW'K:errtmlS.

AlXOIln/lng Re"iew, 70 (3): 411-441

Black, F. 1986. NOise, Jotl,na! of Finance,41: 529-543.

Campbell, J, S. Grossman, and J. Wang1993. Trading volume and serialcom:llltion in stock returns, QuanerlyJournal ofEconomIcs lOB: 905-939.

Chat. J. 2002. Timing information.infonnation 3S}'IlllIleU)', and mKling\'orume. Working Papa 4380-<11, MITSloan School of Managemenl: 1-39.

Chen, G.. M, Firth. and O. Rul. 2001. Thed)nllmlC relation belween slockreturns, trading volume, and volatihty.The FillQncial Rniew, 38: 153-174.

Clad.. P. 1973. A subordinated stochasticprocess model with finlle \'anance forspe<:ulall"e prices. Eamomet"ro, 41:135-155,

Crouch. R. 1970. The "olume oflnl1$IlClions and price changes on !heNew Yort Stock Exchange, Fi/lo./IciaJAnyalystsJourno.l26: 104-109.

Cutler, D., J. Poterba" and L. Summers.1991. Speculative dynamics, Ren'ew ofEconomic Swdies 58: 529-546.

Kent. D, D. Hirshliefer. and A.Subnhmanyam. 1998 A Theory ofo,eroonfidence. 5Clf-anribuuon. andsecunly madet under· ando,·~uons. Journal of Finanu 53:1839·1886,

Epps, T W, and M. Epps. 1976. The

stochastic depend= of security pncechangc:5 and lraIlSaChon "OhUDes:Implicalioos for the mixture--of­distributIOns hypolhe:sIS. Econometrico­44: 305·321

Filma, E. 1991. Efficient eapilal markets: II.JOl1rno.l ofFmunce 46: 1575-1617.

Fama, E., and K. French 1992. The cross­seclion of expeclcd SIOCk. returns. TheJOUl7fal of Fmance 47 (June): 427­465.

Fama, E., and K. Frmch. 1995. Size andbook-la-market racton in earnmgs andrerums. The Journal ofFillan~ 50 (I):131-155.

Ferson, W., and C. Harvey. 1993, The riskand predictabilny of illlemationalequity returns, Review of FllllmcialStudies 6: 521-566.

Gffvais, 5., R. Kamel, and D. Mingelgrin.2001. The high-volume returnpremium. The./ounrol ofFifUJnce LVI(3): 811-919.

George, T.• G. Kaul, and M. Nimalendran_1994. Trading volume and lr1lnUctionCOSlS in specialist mad.C\s. Th~

Journal of Fmw,ce XLIX (4); 1489­1505.

Hiemstra, C., and J. Jones. 1994. Tesling forlinear and nonlinear granger causalityin the StoCk. pnce-volume relauQn. TheJournal of Finance XLIX (5): 1639-

'''''.Hollhauscn. R., and R. Ven-ttChia. 1990.

Thc Effecl of informedness andconsertsu-s on price and volumebehavior The Accountmg ReVl~w 65:

•

191-208.

Hong, H., and J,Slcin. 1999. A unifiedtheory of undcfTeaClion, momentumtrading and overreaction In asselmarkets, JOlmlGl ofFinance 54: 2143­2184,

Klm,O" and R. Verrecchia. 1991. Tradingvolume and price reactions [0 publicannouncements. Journal ofAecGan/ingResearch 29 (Autumn): 302-32L

Kramer, C. 1999. Noise trading, transactioncosts, and the relationship of stockreturns and trading volume.fnlernmional Review of Economicsand Finance 8: 343-362.

Kothari 5., J. Shanken, and R, Sloan. \995,Another look at the cross section ofexpected slock returns. Journal ofFinance, SO (1): 185-224

Lee, c., and B. Swaminathan. 2000. Pricemomentum ami trading volume. TheJournal ofFinance LV (5): 2017-2069.

La, A" and j, Wang. 2000, Trading volume:definitions, data analysis, andimplications of poJ1folio theory.Review of Financial Swdies 13. 257-

300.

Morgan, L 1976. Stock pnces andhetCTOskedasticity. JOlmwl ofBusiness49: 1069-1088.

Pesaran, H., and A. Timmermann, 1995.PrediClability of slock returns:robustness and economic significance.Journal ofFillWlce L (4): 1201-1228.

Slickel, S, and Vcrrecchia, R, 1994.Evidcnce lhat trading volume sustainsstock pnce changes, FinancialAnalYSiS Journal, 50 (6): 57-67.

Taugcn, G., and M, Pills 1983, Thc pricevariability-volume relationship onspeculative markets. Ecanamelrica_51:485-505,

Westerfield, R. 1977. Thc distribution ofcommon stock price changes: Anapplication of transactions time andsubordinated stochastic models.Journal af Financial and Q,WlllilaliveAnalysis 12: 743-765.

Wood, R., T. Mdnish, and J. Ord. 1985. Aninvesligation of transactions data forNYSE stocks. JOlmral of Fimmce 60:723-739.