Вы находитесь на странице: 1из 54

StockDurationandMisvaluation

MartijnCremers

AnkurPareek

ZachariasSautner

UniversityofNotreDame

RutgersUniversity

UniversityofAmsterdam

mcremers@nd.edu

apareek@business.rutgers.edu

z.sautner@uva.nl

December17,2012

Abstract
Westudywhetherthepresenceofshortterminvestorsisrelatedtoaspeculativecomponentinstock
prices using a new measure of holding duration. First, we characterize institutional investors holding
durations since 1985 and find that holding durations have been stable and, if anything, slightly
lengthened over time. Second, we document that the presence of shortterm investors is strongly
relatedtotemporarypricedistortions,consistentwithaspeculativestockcomponentinstockpricesas
modeled in Bolton, Scheinkman, and Xiong (2006). As shortterm investors move into (out of) stocks,
theirpricestendtogoup(down)relativetofundamentals.Asthepresenceofshortterminvestorsis
strongly meanreverting, this creates a predictable pattern in stock returns. We document such
predictabilityusingbothvaluationproxiesandassetpricingtests.

________________
WewouldliketothankWilliamGoetzmann,ZhiguoHe,FlorianPeters,DavidSraer,andRussWermersaswellasparticipantsat
the Roundtable on LongTerm Value in the Corporation at Harvard Law School (September 2012) for their comments.
Commentsareverywelcome.Allerrorsareourown.

Electronic copy available at: http://ssrn.com/abstract=2190437

1.Introduction

How shortterm are investors, and is the presence of shortterm investors related to a

speculativecomponentinstockmarkets,i.e.,toadeviationofstockpricesfromfundamentalvalueover
longer periods of time? Answering these questions is crucial to inform the current debate on the
efficiency of financial markets, on possibly inefficient shortterm firm investment policies, and on
managerial myopia driven by shortterm incentives in executive compensation.1 The extent and
consequences of shorttermism is likewise important for addressing calls for greater shareholder
engagement and power (see e.g. Bebchuk (2005)), or alternatively, for a system of corporate
governance where shareholders are kept at a greater distance from the board (e.g. the director
primacymodelofBainbridge(2006)andBlairandStout(1999)).
Froot, Perold, and Stein (1992) empirically link shortterm investor horizons with managerial
myopia,forwhichBolton,Scheinkman,andXiong(2006)provideaformalmodel.Theirtheorybuildson
the model of Scheinkman and Xiong (2003) and posits that thepresence ofshortterm investors with
heterogeneous beliefs that are caused by overconfidence can create temporary price distortions if
shortsalesconstrainsexist.Theresultingspeculativecomponentinstockpricescanaffectthebehavior
of firms managers (and vice versa), whose compensation has become largely based on stocks and
option grants with relatively short vesting periods.2 Daniel, Hirshleifer, and Subrahmanyam (1998)
incorporateinvestoroverconfidenceintheirtheoryaswell,showingthatstockpriceoverreaction(i.e.,
misvaluations) can be causedby investors overweighting of private information. Shortterm investors
may be more prone to such overconfidence, as suggested by DeBondt and Thaler (1995) or De Long,
Shleifer,Summers,andWaldmann(1991).

Inthispaper,weproposetostudythesequestionsusinganewmeasureoftheholdingduration

of institutional investors, called Stock Duration, which was introduced in Cremers and Pareek (2012).
StockDurationistheweightedaveragelengthoftimethatinstitutionalinvestorshavehelda stockin
their portfolios, based on their quarterly 13F holding reports and weighted by the dollar amount

See, for example, Narayanan (1985), Stein (1988, 1989), Bebchuk and Stole (1993) or Laverty (1996) on short
termismandmanagerialmyopia.Phelps(2010,p.17),forexample,isconcernedthatinestablishedbusinesses,
shorttermismhasbecomerampant.Similarly,ChancellorLeoStrine(2007)oftheDelaware'sCourtofChancery
raises concerns about the presence of shortterm investors for corporate America. Currently stalled, the SEC
drafted a proposal in 2010 allowing shareholders to nominate directors if they own at least 3% of a company's
sharesforatleastthepriorthreeyears.
2
Gopalan et al. (2012), for example, document with their new measure of executive pay duration that typical
compensationpackageshavestockandoptiongrantsvestingwithinoneortwoyears,whichseemsshortterm.

Electronic copy available at: http://ssrn.com/abstract=2190437

investedacrossallinstitutionscurrentlyholdingthestock.UsingthisnewStockDurationmeasure,our
paperprovidestwo maincontributions.First,we characterizetheinstitutionalholdingdurationssince
1985,andsurprisinglyfindthatholdingdurationshavebeenstableand,ifanything,slightlylengthened.
Second,wedocumentthatthepresenceofshortterminstitutionsisstronglyrelatedtotemporaryprice
distortions, consistent with a speculative stock component in stock prices as modeled in Bolton,
Scheinkman,andXiong(2006).Asshortterminvestorsmoveinto(outof)stocks,theirpricestendtogo
up(down)relativetofundamentals.Asthepresenceofshortterminvestorsisinherentlyinstable(i.e.,it
isstronglymeanreverting),thisseemstocreateapredictablepatterninstockreturns.Wedocument
suchpredictabilityusingbothvaluationproxies(i.e.,themarkettobookratioorthemisvaluationproxy
proposed by Pastor and Veronesi (2003)), and traditional asset pricing methodology (i.e., ex ante
portfoliosortsandcrosssectionalpredictivereturnregressions).3

TheaverageStockDurationequalsabout1.2yearsin1985andincreasestoaround1.5yearsin

2010,andthepercentageofinstitutionalstockownershipheldforover2yearsincreasedfrom19%in
1985to31%in2010.Suchlengthenedholdingdurationsmaybesomewhatsurprisingasitiswellknown
that share turnover has significantly increased over the same period (from an average annualized
turnover of 72% in 1985 to close to 300% in 2010). Indeed, business press articles routinely infer
average holding periods from share turnover, which our results indicate is incorrect (see, e.g., The
Economist (2012)). The most important reason that holding periods and share turnover are quite
distinctiveisthatturnoverisandinstitutionalholdingperiodsarenotstronglyaffectedbytherecent
increase in high frequency trading. As high frequency traders typically have little overnight exposure
(see, e.g.,Menkveld(2011)), they affectshareturnover butnotStock Duration, whichmoregenerally
ignoresanyroundtriptradesoccurringwithinaquarter.Ontheotherhand,lengtheningStockDuration
is consistent with the major recent shift towards indexing which is inherently longerterm (see, for
example, Cremers and Petajisto (2009) for evidence of closet indexing by actively managed mutual
funds,andCremersandPareek(2012)fortheincreaseinoverallinstitutionalindexing).Thecorrelation
ofStockDurationandshareturnoverequals41%,andcontrollingforshareturnoverdoesnotaffectany
ofourresults.4Acrossinstitutions,pensionfundsandendowmentshavethelongestmediandurations

OurresultsarerobusttousingthemisvaluationmeasureproposedinHobergandPhillips(2010).
AnotherdifferenceisthatourStockDurationmeasureonlyconsidersinstitutionalstockholdingsratherthanall
stockholdings,whichhasbecomelessofarestrictionovertime,asinstitutionalinvestorsasagrouphaveincreased
theiraverageownershipinstocksfrom33%in1985to75%attheendofoursamplein2010.Thedifferentresults
forturnoverare consistentwithHendershott,Jones,and Menkveld (2011),who showthat, rather than creating
temporarypricedistortions,highfrequencytradingimprovesliquidity.Litzenberger,Castura,andGorelick(2012)
4

(about 1.7 years), followed by banks and insurance companies (about 1.5 years), and by investment
companiesincludingmutualfunds(about1.25years).

Incontrasttoothermeasuressuchasshareturnoverorthefrequencyofinstitutionaltrading,

ourStockDurationproxycanbemeasuredatamoredeaggregatedlevel.Specifically,StockDuration
measureshowlongaspecificstockhasbeeninaninstitutionalinvestorsportfolio,ratherthanhowlong
allstocksonaveragehavebeenintheportfolio.Accountingforthisisimportantasonecanonlyobserve
institutional holdings at an aggregate institution level, e.g. a single quarterly holdings report for the
entire portfolio of Goldman Sachs, which includesboth client retirementportfolios with a more long
term focus and internal hedge fund positions with a perhaps more shortterm focus. Our proxy thus
allows any given institutional portfolio to be shortterm in some stocks and longterm in others. As a
result,weareabletodistinguishdurationsamongthetypicallylargenumberofdifferentstocksheldby
an institution at a particular time, which may be managed by different portfolio managers with
heterogeneous investment horizons. In contrast, existing measures such as the classification of
institutionsastransient(seeBushee(1998))andinstitutionalfundturnover(seeGaspar,Massa,and
Matos(2005)andGasparetal.(2012))donotaccountfordifferencesinholdingdurationsamongstocks
withinagiveninstitutionalportfolio.Forthesealternativemeasures,wedonotfindanyassociationwith
price deviations from fundamentals or persistent return predictability, and our results are robust to
controllingforthem.

Usingournewmeasure,wedocumentbothastrongnegativecontemporaneousandastrong

predictive positive association between Stock Duration and proxies for misvaluation in pooled panel
regressionswithfirmandyearfixedeffects.Ourmain(mis)valuationproxiesarethestandardmarket
tobookratioand theproxyproposedinPastor andVeronesi(2003).5 Thecontemporaneousnegative
association between Stock Duration and markettobook is economically quite meaningful, with a
standard deviation decrease in duration (of 0.54 years) associated with an increase in the marketto
book ratio of 14.6%. This indicates that valuations are generally going up (down) while shortterm
institutionalinvestorsarebuying(selling).Similarly,acontemporaneousstandarddeviationdecreasein
duration is associated with an increase in the PastorVeronesi misvaluation proxy of about 5.5%. The

conclude in their survey that automated and high frequency trading have improved market quality, including
narrowerbidaskspreads,increasedliquidityandloweredintradayvolatility.
5
The Pastor and Veronesi (2003) proxy is calculated as the residual from yearly firmlevel regressions of
log(markettobook ratio) on log(book value of total assets), leverage (longterm debt / total assets), return on
equity,1/(1+firmage),adummyforwhetherthefirmpaysdividends,andstockreturnvolatility.

contemporaneously negative association is followed by an equally meaningful predictive and positive


association. A standard deviation increase in duration this year is associated with a decrease in next
yearsmarkettobookratio of12%,andwithanotherdecreasein thefollowing yearof4%.The price
reversal pattern in the PastorVeronesi misvaluation proxy is analogous and both statistically and
economicallysignificant.6

We then show evidence suggesting that the price reversal pattern is consistent with limited

arbitrage allowing temporary price distortions (see, e.g., Shleifer and Vishny (1990, 1997)). Both the
contemporaneouslynegativeandthepredictivepositionassociationbetweenStockDurationandprice
levels isonly apparentinthesubsamples ofstocksthathave higheridiosyncratic volatilityorthatare
lessliquid.Likewise,thepricepatternisonlyapparentforstocksthathavelowerinstitutionalholdings
andthusaremorelikelytobesubjecttoshortsalesconstraints(e.g.,Nagel(2005)).Theseresultsare
againsimilarusingeitherthemarkettobookratioorthemisvaluationproxy.
While these findings suggest that the reversal pattern could at least partly be the result of
temporarypricepressureduetothepresenceofshortterminvestors,weacknowledgethatwedonot
have any instruments allowing us to only consider exogenous changes in Stock Duration that are not
relatedtofirmfundamentals.Indeed,thecontemporaneousassociation,ifconsideredinisolation,could
bedrivenbynewinformationthatshorttermfocusedinstitutionsarequickertotradeupon.However,
thedocumentedpredictabilitythatprovidesthereversalpatterncannoteasilybereconciledwithsuch
aninformationstory.Rather,ifinformationisdrivingchangesinthepresenceofshortterminstitutions,
thereversalsuggestsasystematicoverreactiontoanysuchinformationonpartofinstitutionsthathave
held these stocks for short period of time. Such overreaction is consistent with the theory in Daniel,
Hirshleifer,andSubrahmanyam(1998),andwiththeempiricalresultsinCremersandPareek(2012)who
documentthatthemomentum,reversalandshareissuanceanomaliesarestrongerforstocksheldby
shortterminvestors.

It is this price reversal pattern and particularly the predictability that is most suggestive of

temporarypricedistortion.Asaresult,weinvestigatethispredictabilityinmuchgreaterdetail,leaving

However, a price reversal pattern is not apparent using either share turnover, institutional turnover or the
presenceoftransientinvestors,whiletheresultsforStockDurationarerobusttotheirinclusion.Asourpaperis
closelyrelatedtothepricepressurevariableinCovalandStafford(2007),weverifythat,whileStockDurationis
correlatedwiththeirmeasure(correlationof32%),ourpricereversalresultsareindependentfromtheirvariable
whichisdrivenbyextrememutualfundflows.

thereasonsfortradingby(orfirmsattracting)shortterminstitutionsforfutureresearch.7Letusnote
herethatthepredictivepatterninstockpricesismirroredinthemeanrevertingpresenceofshortterm
institutions (e.g., those with holding durations of less than 6 months), whose presence is inherently
instable.ItcantakeseveralyearsbeforelargechangesinStockDurationarefullyreversed,andlikewise
wefindthatittakesabouttwoyearsbeforethepricereversalpatterniscompleted.

Our investigation of stock return predictability proceeds in three steps, each considering an

additional aspect of the three components that make up the reversal pattern: Stock Duration,
misvaluation, and longterm past stock return performance. Accordingly, we first start by forming
quintileportfoliosbasedonStockDurationattheendofthepreviousquarter,updatingportfolioseach
quarter.Alongshortequallyweightedportfoliothatbuys(sells)stockswithlong(short)durationshas
anannualized3factoralphaof4%(tstatisticof2.65).Thisunconditionalresultconfirmsthehypothesis
thatinvestorholdingdurationisrelatedtostockmisvaluation.

Thesecondstepistoconditiononmisvaluation.Theintuitionisthatifstocksheldbyshortterm

(longterm) investors are more likely to be overvalued (undervalued), we should expect the
predictability to be stronger conditional on misvaluation. We examine predictability conditional on
misvaluationbyusingindependent5x3doublesortsonStockDuration(intoquintiles)andmisvaluation
(intoterciles).UsingthePastorVeronesimisvaluationproxy,wefindforthetercilegroupofthemost
overvaluedstocksthatalongshortequallyweightedportfoliothatbuys(sells)stockswithlong(short)
durations has an annualized 3factor alpha of 5.7% (tstatistic of 3.38). The analogous 4factor alpha
remainseconomicallyandstatisticallysignificantat3.9%peryear(tstatisticof2.38).Resultsarevery
similar if we use the standard markettobook ratio and we cannot detect similar patterns of
misvaluationwhenusingeithershareturnoverorinstitutionalturnoverasproxiesforinvestorhorizons.
Consistent with the limits to arbitrage argument, we find that this predictability pattern is again

WhilethereversalpatterninstockpricesseemsquitesymmetricduringversusafterchangesinStockDuration,it
is quite possible that these changes in the presence of shortterm investors interact with managerial decision
making,whichlikewiseisleftforfutureresearch.Somesupportiveresearchinthisdirectionisalreadyprovidedby
Derrien,Thesmar,andKecskes(2012).Usinginstitutionalturnoverasameasureofinvestorhorizon,theyfindthat
mispriced firms cater in their corporate policies to the tastes of their shortterm investors. Polk and Sapienza
(2009) find that firms cater in their investment behavior to the tastes of shortterm investors (measured using
share turnover). Similarly, firms with shorter investor horizons, measured again using institutional turnover, do
worse in takeovers as targets or acquirers (Gaspar, Massa, and Matos (2005) and Chen, Harford, and Li (2007)).
Furtherevidenceconsistentwithaninteractionofinvestorhorizonandcorporatepoliciescomesfromsurveysin
which managers admit that they are sacrificing profitable longrun project to meet shortrun earnings targets
(Graham,Harvey,andRajgopal(2005)).

strongestinsubsamplesofstockswithhigherarbitragecosts,suchasforstockswithlowinstitutional
ownership,highidiosyncraticrisk,andhighlevelsofilliquidity.

Thethirdandfinalstepistofurthertakethelongertermpastperformanceintoaccount.Ifthe

reversalpatternindicatesthatstocksaretemporarilyovervalued,pastreturnscouldplayanimportant
roletocorroborateanyovervaluation.Aslargechangesinthepresenceofshortterminvestorscantake
twotothreeyearstomaterialize,weconsidertheeffectsofthestockreturnsoverthepastthreeyears.
Intuitively,ifthepastthreeyearstockreturnisrelativelyhighandastockhasahighvaluation,itseems
morelikelythatthestockisovervalued.Ifso,wewouldexpectourreversalpatterntobestrongerfor
suchstocks,particularlythoseheldbyshortdurationinvestors.Weconfirmthisintuitionusinga5x3x2
triple sort on Stock Duration, misvaluation, and the threeyear past return. We further study a direct
connection to past sixmonth momentum. Specifically, stocks with high (mis)valuations held by short
terminvestorsthatalsoexperiencednegativesixmonthmomentumseemgoodcandidatesforstocks
whoseovervaluationhasrecentlystartedtobecomerecognized.Forthesestocks,wefindthestrongest
reversal pattern (i.e., large negative alphas). In contrast, stocks with a similar valuation and a similar
presence of shortterm investors but with positive past sixmonth momentum do subsequently not
showeconomicallylargenegativealphas.Combinedwiththeevidenceconditioningonthepastthree
yearreturns,theseresultssupportthenotionthathighvaluationsindeedreflectmisvaluationsrather
than,forexample,superiorgrowthopportunities.
A closely related question is to what extent the presence of shortduration investors could
explain the value premium, i.e. that stocks with low markettobook ratios tendto outperform stocks
with high markettobook ratios. If shortduration traders distort prices, particularly allowing highly
valuedstockstobetemporarilyovervalued,thenwewouldexpectthatthevaluepremiumisstronger
for stocks held primarily by institutional investors with shorter durations. We find empirical results
consistentwiththis,namelythatthevaluepremiuminoursampleisdrivenbystockswithshorterStock
Durations,whilethereisnovaluepremiumamongstockswithlongStockDurations.
We complement our portfolio results with regressions using FamaMacBeth (1973)
methodology.Todothis,weestimatepredictivecrosssectionalregressionsofnextyearrawreturnsor
DGTW adjusted cumulative abnormal returns on Stock Duration, while controlling for other stock
characteristics(seeDanieletal.(1997)andWermers(2003)).Theregressionresultsareconsistentwith
theportfolioresultsandconfirmthatreturnpredictabilitybasedonStockDurationisstrongerformore

overvaluedstocks.Moreover,weagainfindthatthatpredictabilityisstrongestinsubsamplesofstocks
withhigherarbitragecosts.

The remainder of this paper is organized as followed. In the next section, we explain the

constructionoftheholdingdurationmeasureandtheothermainvariablesusedanddescribetheirtime
seriesevolutionoveroursampleperiodfrom1985to2010.Insection3,werelatethemarkettobook
ratio and the PastorVeronesi misvaluation ratio to Stock Duration. Section 4 considers the cross
sectionalreturnpredictabilityusingportfoliosortsandFamaMacBethregressions.Section5concludes.
2.DataandSummaryStatistics
2.1DataandSummaryStatistics
WeuseinstitutionalinvestorholdingsdatafromtheThomsonFinancialCDA/Spectrumdatabase
ofSEC13FfilingstocreateourStockDurationmeasure.AllinstitutionalinvestorswithmorethanUSD
100 million of securitiesunder management are required to report theirholdings to the SEC on form
13F.Holdingsarereportedquarterlyandallcommonstockpositionsgreaterthan10,000sharesorUSD
200,000 must be disclosed. Stock return data is obtained from the monthly CRSP stock data files and
accountingdataisfromCOMPUSTAT.OuranalysisfocusesonUScommonstocksfromJanuary1985to
December 2010. Return forecasting and the stock selection analysis is performed from January 1980
onwards,asatleastfiveyearsofholdingsdata(whichstartin1980)isrequiredtocalculatetheholding
durationmeasure.WeeliminatestocksinthebottomNYSEsizedecileandstockswithpricesbelowUSD
1fromoursample.Further,werequirestockstobepresentinCRSPforatleasttwoyearsbeforethey
areincludedinthesampletomakesurethatIPOrelatedanomalies(e.g.,IPOunderpricing)donotaffect
ourresults.Toeliminateanysamplebias,werequireinstitutionalinvestorstobepresentfortwoyears
before being included in the sample. We do this as new institutions by construction have short past
holdingdurationsforstocksintheirportfolios.Table1providessummarystatisticsforthesampleofthe
firmsthatareincludedinourstudy.
2.2Methodology:CalculatingStockDuration
Using the methodology introduced in Cremers and Pareek (2012), we calculate the holding
durationofownershipofeachstockforeveryinstitutionalinvestorbycalculatingaweightedmeasure
ofbuysandsellsbyaninstitutionalinvestor,weightedbythedurationforwhichthestockwasheld.For
eachstockinagiveninstitutionsportfolio,theholdingdurationmeasureiscalculatedbylookingback
8

over the full time period since that particular stock has been held continuously in the portfolio.
Intuitively, this institutionalstock level duration variable measures how long a USD 1 investment in a
stockhasonaveragebeeninthatinstitutionsportfolioataparticularpointintime.Themostimportant
limitationofourmeasureisthatanyroundtriptradeswithinaquarterareignored,asweonlyobserve
institutionalholdingsattheendofeachquarter.
ThecalculationofthedurationforstockithatisincludedintheinstitutionalportfoliojattimeT
1,forallstocksi=1Iandallinstitutionalinvestorsj=1J,isgivenby:

Duration i , j ,T 1 d i , j ,T 1

(T t 1) i , j ,t

H B
t T W
i, j
i, j
T 1

(W 1) H i , j

H B
i, j
i, j

(1)

where
Bi,j=totalpercentageofsharesofstockiboughtbyinstitutionjbetweent=TWandt=T
1;t,Tareinquarters
Hi,j=percentageoftotalsharesoutstandingofstockiheldbyinstitutionjattimet=TW.
i,j,t = percentage of total shares outstanding of stock i bought or sold by institution j
betweentimet1andt,wherei,j,t>0forbuysand<0forsells.
Our duration measure takes into account tax selling and other temporary adjustments in the
portfoliosbecauseintermediatesellsarecancelledbyimmediatebuybacks,withonlyasmalleffecton
thedurationofcurrentholdings.Tocreatethevariable,wechooseW=20quarters,asveryfewstocks
are held continuously for longer than five years. If stock i is not included in institutional portfolio j at
timeT1,thenDurationi,j,T1=0.
Next, we compute at the individual stocklevel our Stock Duration proxy by averaging the
institutionalstock level Durationi,j,T1 over all institutions currently holding the stock, using as weights
eachinstitutionstotalcurrentholdingsinthestock.ThesummarystatisticsinTable1showthatStock
Durationhasamean(median)of1.41(1.39)acrossthesampleperiod,withchangesinStockDuration
beinghighlyinstableandmeanrevertingassuggestedbyitsnegativeautocorrelationof14%.
We will employ two other closely related measures of institutional investor horizon in our
empirical analysis. The first measure was introduced by Bushee (1998, 2001), whose methodology is
basedonfactorandclusteringanalysis to classifyinstitutionalinvestors intothreegroups:transient
institutions with high portfolio turnover and diversified portfolios, dedicated institutions with low
9

turnoverandmoreconcentratedportfolioholdings,andquasiindexerinstitutionswithlowturnover
and diversified portfolio holdings. To create the variable, we obtain the institutional investor
classification data from Brian Bushees website and include the percentage of a firms ownership by
transient institutional investors (Transient Investors) as an alternative measure for the level of
ownershipbyshorthorizoninvestors.Table1showsthattheaveragefirminoursamplehasabout13%
transientinstitutionalinvestors.
The second alternative measure is Institutional Turnover which is used, for example, by
Gaspar, Massa, and Matos (2005) and Gaspar et al. (2012). This variable does not account for
differences in holding durations among stocks within a given institutional portfolio. The variable is
defined as the weighted average turnover of the institutional investors holding a given stock.
InstitutionalTurnoveriscalculatedusingchangesinthequarterlyholdingsoverthepastfourquarters
and thestocklevel weightsare calculatedusing thecurrentholdingsinthestockin eachinstitutional
portfolio.Thevariablehasamean(median)valueof0.28(0.27)foroursamplefirms.Thecorrelation
between Stock Duration and the percentage of ownership by transient investors is 29%, and the
correlation between institutional turnover and Stock Duration equals 57%. Finally, the correlation
between Stock Duration and share turnover equals 41%. These figuressuggestthat these alternative
measures are clearly distinct from our new measure Stock Duration (see Appendix A2 for the full
correlationmatrix).
The major difference between Stock Duration and the percentage of transient investors and
institutional turnover is that these measures are calculated at the institutionlevel rather than the
institutionstocklevelbeforetheyareaggregatedacrossallinstitutionsholdingastock.Asaresult,they
donotallowforheterogeneityintheinvestmenthorizonacrossdifferentstocksinagiveninstitutional
portfolio. In contrast, Stock Duration is calculated by aggregating the institutionalstocklevel holding
durations, thus allowing the same institutional investor to be shortterm for some but longterm for
other stocks. Given that we only observe institutional portfolios at an aggregate level, with many
portfoliomanagerswithinthelargestinstitutionspotentiallyhavingdifferentinvestmenthorizons,thisis
animportantdistinction.
2.2EvolutionofStockDurationoverTime

Table 2 provides information on the evolution of Stock Duration, institutional turnover, share

turnover,andinstitutionalinvestorholdingsovertime.Theevolutionofthesevariablesisalsovisualized
10

inFigure1.Table2showsthattheaverageStockDurationequalsabout1.2yearsin1985,andincreased
by about 20% to 1.5 in 2010. The increase in duration is interesting as share turnover shows a very
differentpatternoverthesametimehorizon.Infact,whileStockDurationlengthened,shareturnover
hassubstantiallyincreasedfrom72%peryearin1985to276%peryearin2010,i.e.,atotalgrowthof
about300%.Theincreaseinshareturnoverhasbeenrelativelycontinuousoverthelastdecades,witha
slight drop in the years of the financial crisis. StockDuration has increased in the late 80s, and saw a
drop in the second half of the 90s. Since then, however, durations have become longer again, with a
dropsimilartothatofshareturnoverintheyearsofthecrisis.

Wecanillustratethesepatternsusingsomespecificexamples.Figure2Areportstheevolution

of Stock Duration over time for selected companies (Apple, Microsoft, P&G, Boeing, and Hewlett
Packard). For all of these firms Stock Duration has increased over the last decades. Figure 2B
complements the previous figure and shows the evolution of Stock Duration, share turnover, and
institutional investorholdingsforHewlettPackard. Thisgraphprovidesarepresentativeillustration of
anincreaseinshareturnoverandinstitutionalinvestorholdingsovertime,combinedwithalengthening
ofStockDuration.

Figure 3 complements the statistics in Table 2 and reports the percentage of institutional

holdings with a Stock Duration of lessthan six months (i.e., shortterm holdings), between six and 12
months,between12and24months,andabove24months(i.e.,longtermholdings).Theholdingsadd
upto100%,representingthetotalholdingsofallinstitutionalinvestorsatanypointintime.Thefigure
shows that the institutional holdings with Stock Durations above two years have increased over the
sampleperiod,fromabout19%in1985to31%in2010,anincreasebymorethan60%.Tothecontrary,
the holdings of shortterm investors with Stock Durations less than six months have not significantly
changedfrom1985(27%)to2010(26%).

Our data also allows us to measure Stock Duration across different types of institutional

investors.Weclassifyinstitutionalinvestorintofourcategories.ThecategoryBankincludesbanksand
insurance firms, Investment Company includes independent investment advisors and investment
companies (e.g., mutual funds), Pension Funds includes corporate (private) pension funds, public
pensionfunds,anduniversityandfoundationendowments,andOthersincludeallotherinstitutional
investors. As displayed in Table 1, we find that, across the entire sample period, pension funds and
endowments have the longest median durations (about 1.7 years), followed by banks and insurance
companies(about1.5years),andinvestmentcompaniesincludingmutualfunds(about1.25years).
11

Figure4AshowstheevolutionofmedianStockDurationbyinstitutionalinvestortype,withthe

dataunderlyingthisfigurebeingreportedinAppendixA3.Thefigureshowsthatthestockdurationof
banks has been relatively stable over time, moving in a range between about 1.2 and 1.8 years. The
patternforinvestmentcompaniesissimilar,thoughtheygenerallyseemtohavestockforshortertime
periodsintheirportfolios.ThestrongestincreaseinStockDurationcanbeobservedforpensionfunds,
withalengtheningofthemediandurationfrom0.85yearsin1985to2yearsin2010.Figure4Bprovides
asimilaranalysis,lookingattheweightedaverageStockDurationofallstocksheldintheportfolioofa
numberofselectedinstitutionalinvestors,namelyforBerkshireHathaway(averylongterminvestment
company), Fidelity (a major mutual fund), STRS Ohio (a public pension fund), and the University of
Chicago(auniversityendowment).

Appendix A4 complements Figure 4 and shows the evolution of institutional investor holdings

over time. It shows that investment companies such as mutual funds are by far the most important
investor type based on holdings. Their relative holdings as a fraction of all equities have substantially
increased over the past decades, from about 16% in 1985 to 54% in 2010. This is equivalent to an
increaseintheirdollarholdingsbymorethan200%.Tothecontrary,theholdingsofbanksandpension
fundshaveslightlydecreased overthesample periods (by 4%and 23%,respectively).Theholdingsof
otherinvestorsalsoincreasedovertime,thoughfromamuchlowerbasisandwiththetotalholdings
stillbeingbelow5%ofalloutstandingequities.

AppendixA5providesananalysissimilartothatinAppendixA4,butonlyreportsholdingswith

stockdurationsthatarelessthansixmonths.Theappendixshowsthatmutualfundsareagainthemost
importantshortterminstitutionalinvestors,makingupmorethan80%ofallshortterminvestmentsby
institutionalinvestors.Whileotherinvestorssuchasbanksalsohaveshorttermholdings,mutualfunds
playadominantrolesimplybecauseasagrouptheyhavethelargestholdings.

How can we reconcile the simultaneous increase of both Stock Duration and share turnover?

First and foremost, it indicates that institutional investors (particularly the most important, namely
banks,insurancecompanies,investmentcompanies,pensionfundsandendowments)asagrouparenot
thosethathavesignificantlyincreasedtrading.Asa group, institutionalinvestorshave increasedtheir
average ownership in stocks from 33% in 1985 to 75% at the end of our sample period in 2010.
However,averageinstitutionalturnoverhasnotincreasedovertime,suggestingagainthatinstitutions
generally have not shortened their investment horizons. The main caveat here is that we can only
observe quarterly holdings and are thus do not observe changes in holdings due to roundtrip trades
12

within a quarter. However, it seems rather unlikely that institutional investors could significantly
increaseshorttermtradingthatwouldnotresultinanyincreaseintheirquarterlyportfolioturnover.In
effect,thatwouldrequirethatalltheirincreasedtradingwouldhavetobeexclusivelyduetoroundtrip
tradesthatfallneatlywithincalendarquartersratherthanbeingmorerandomlyspreadoutoverone
tothree month periods across quarterends. This requirement seems quite restrictive, as we are not
aware of any special institutional incentives to carefully time round trip trades this way. Window
dressingissuesaretypicallyrelatedtoreturnsandpricesratherthanthelengthoftimestocksareheld.
Rather, we conjecture that the increase in both share turnover and Stock Duration is consistent,
respectively, with the increase in high frequency trading and the major shift towards indexed
investments,whichisinherentlylongerterm.
AsshowninHendershott,JonesandMenkveld(2011),theincreaseinhighfrequencytradingis
partlyduetoasignificantlowering oftrading costs during ourtimeperiod. Ashigh frequencytraders
typicallyhavelittleovernightexposure(seeMenkveld(2011)),theirtradingincreasesshareturnoverbut
doesnotaffectStockDuration,whichisbasedoninstitutionalportfoliosattheendofeachquarterand
thus ignores any roundtrip trades occurring within a quarter. We will later show that controlling for
shareturnoverdoesnotaffectourresults.FindingthatStockDurationbutnotshareturnoverisrelated
totemporarypricedistortionsisconsistentwiththemicrostructureliterature.Hendershott,Jones,and
Menkveld (2011) and Litzenberger, Castura, and Gorelick (2012), for example, document that
automatedandhighfrequencytradinggenerallyseemstoimprovemarketquality.
Theincreaseinholdingdurationoveroursampleperiodcanpartlybeexplainedbytheincrease
in(closet)indexingoverourtimeperiod.CremersandPetajisto(2009)documentthisprevalencefor
activelymanagedmutualfundsandinCremersandPareek(2012)forallinstitutionalinvestors.Cremers,
Ferreira,Matos,andStarks(2012)documentthesignificantincreaseinexplicitindexingthroughmutual
fundsandETFS,andfindthatveryactivemanagers(i.e.,managerswhoseportfoliosareverydifferent
fromtheirbenchmarksorwithahighActiveShare)aremoresuccessfulinmarketswithmoreindexing,
whichmaymakemarketlessinformationallyefficient.
3.MisvaluationandStockDuration
3.1MisvaluationandStockDuration:ResultsfromPanelRegressions

In this section, we provide evidence suggesting that the presence of shortterm investors is

related to temporary price distortions. We estimate pooled panel regressions relating a set of
13

(mis)valuationproxiesandStockDuration.Weconsiderboththerawstandardmarkettobookratio
asavaluationproxythatcouldbepartlypickingupmisvaluationoncewecontrolforalargenumberof
controls,andthemisvaluationproxyproposedinPastorandVeronesi(2003).Theirmisvaluationproxyis
definedastheresidualfromregressingmarkettobookratiosonasetoffundamentalvariablessuchas
assets,leverageorprofitability.Asanadditionalrobustnesscheckwewillalsoreportresultsusingthe
relatedmisvaluationproxyproposedinHobergandPhillips(2010).

Westudyboth thecontemporaneousandthepredictiverelationbetween StockDurationand

our (mis)valuation proxies. We include firmfixed effects throughout all regressions, such that we
effectively test how changes in Stock Duration are related to changes in (mis)valuation. We further
controlforyearfixedeffectsaswellasasetofothervariablesthatmayberelatedtomisvaluations.For
example, we control for the percentage of holdings by institutional investors, share turnover, asset
horizon,firmcharacteristics,illiquidityoftheshares,andidiosyncraticvolatility.Wewilllatertesthow
someofthesevariablesinteractwithStockDuration.Forallpooledpanelregressions,weemployrobust
standarderrorsclusteredatthefirmlevel.

The corresponding regression estimates are reported in Table 3, containing results for both

standard markettobook ratio and the PastorVeronesi misvaluation proxy. Results using the Hoberg
and Phillips (2010) misvaluation proxy are reported in columns 1 to 4 of Appendix A6. The results
indicatethatStockDurationand(mis)valuationhaveasignificantnegativecontemporaneousaswellas
asignificantpositivepredictiveassociation,whichcombineintoareversalpatternthatisconsistentwith
temporarypricedistortionsthatarerelatedtothepresenceofshortdurationinvestors.
First, we find that there is a strong negative contemporaneous association between Stock
Durationandthemisvaluationproxies.Thisindicatesthatstockpricesaregoingup(down)whileshort
terminstitutionalinvestorsarebuying(selling).Intermsofeconomicmagnitudes,usingtheregression
estimatesfromcolumn5,astandarddeviationdecreaseinStockDurationisassociatedwithanincrease
in the markettobook ratio of (0.27*0.54=) 14.6%. Similarly, such a standard deviation decrease in
durationisassociatedwithanincreaseinthePastorVeronesimisvaluationproxyofabout5.4%.Both
seemeconomicallymeaningful.8

WeperformasetofrobustnessteststoensurethatourStockDurationvariableisnotsimplypickinguptheeffect
ofinstitutionalpricepressuredrivenbyextrememutualfundflowsasinCovalandStafford(2007),Asexpected,
bothvariablesarecorrelatedwitheachother,withacorrelationof32%.Wefindthatourpricereversalpatterns
arerobusttoincludingtheCovalandStafford(2007)measureforinstitutionalpricepressure.

14

Second,thecontemporaneouslynegativeassociationisfollowedbyanequallystrongpredictive

and positive association between the misvaluation proxies and Stock Duration. This effect seems
likewise uniform across the different misvaluation proxies. A standard deviation increase in Stock
Duration this year is associated with a decrease in next years markettobook ratio of 12%, and with
anotherdecreaseinthefollowingyearof4%(thelatterofwhichisinsignificant).Thereversalpatternin
thePastorVeronesimisvaluationproxyisanalogousandbothstatisticallyandeconomicallysignificant.
AppendixA6showsthatourresultsarerobusttousingthemisvaluationproxyproposedinHobergand
Phillips(2010),withagainboththenegativecontemporaneousandthepositivepredicateeffectsbeing
statisticallysignificant.

AppendixA6furthershowsthatourresultsarerobusttousingthepercentageofstocksthatis

ownedbyshortterminstitutionsratherthanStockDuration.Wecreatethisvariablebymeasuringthe
percentage that is owned by institutional investors (relative to all holdings by institutional investors)
withaStockDurationthatisinthebottom50%.Theresultsincolumns5to8showthatfirmsthathave
moreshortterminvestorshavehighercontemporaneousandlowerfuturevaluations.

As mentioned above, the main difference between our Stock Duration measure and related,

existing measuresof investor horizon is that our measure can be calculatedat a more disaggregated
level.Specifically,quarterlyholdingreportsallowustodistinguishdurationsamongthetypicallylarge
numberofdifferentstocksheldinthesameportfolioofaparticularinstitutionalinvestorataparticular
point in time. For example, some stocks may have been inthatinstitutions portfolio for a long time,
whileotherstocksmayappearforthefirsttimethatquarter.Tothecontrary,theothertwomeasures
based on institutional holdings classify institutions (i.e., all stocks in their portfolios) as short or long
term,andthusdonotdistinguishbetweenpotentiallyverydifferentholdingdurationsacrossthestocks
inagiveninstitutionsportfolio.

TheresultsinTable4,usingthemarkettobookratioinPanelAandthemisvaluationproxyof

PastorVeronesi(2003)inPanelB,underscorethatresultsareverydifferentwhenanyofthealternative
investorhorizonmeasuresareused.Specifically,noneofthealternativemeasuresshowanyevidence
for a reversal pattern, while our baseline results are generally robust to adding both lagged and
contemporaneousvaluesofthealternativemeasures.Forexample,incolumns1to3,wefindthatshare
turnoverhasapositiveandsignificantcontemporaneousassociationwithvaluations,butnopredictive
association.Incolumns4to6,institutionalfundturnoverhasbothapositivecontemporaneousanda
positive predictive association with valuations, thus providing again no evidence for any reversal
15

pattern.9 Finally, the percentage of transient investors in columns 7 to 9 has a positive


contemporaneousassociationbutnopredictiveassociation.Therefore,wecannotdetectforanyofthe
otherthreeproxiesthepricereversalpatternthatwasapparentinTable3.Atthesametime,theprice
reversal pattern associated with Stock Duration is largely robust to adding any of the other proxies,
especiallywhenconsideringthemisvaluationproxyinPanelB.
Figures5Aand5Bcomplementthepreviousregressiontables,providingananalysissimilarto
aneventstudyandplottingmarkettobookratiosaroundlargedecreases(PanelA)andincreases(Panel
B)inStockDuration.Tocreatethesefigures,wedefinelargedecreases(increases)inStockDurationas
calendaryearendchangesinStockDurationthatareinthebottom(top)decile.Forthiseventstudy,we
define as t = 0 the year in which a large change in Stock Duration occurs, and report mean values of
markettobook ratios and Stock Duration in the five years before and after t = 0. For the marketto
bookratios,wealsoreportconfidenceintervals(twostandarddeviationsaroundthemean).
Figure5Ashowsthatthemarkettobookratioislargest(atabout3.0)intheyearofthedropin
StockDuration,andissmallerbothintheyearsbeforeandafterthedecreaseinStockDuration.Itthus
seemstotakeabouttwotothreeyearsbeforethepricereversalpatterniscompleted.Thefigurealso
documentsthatStockDurationshowssubstantialmeanreversionafterthelargedrops,goinggradually
back until year +5 to its level in the years before the large increase in shortterm investors. Again, it
seems that it takes rather long, about three to five years, until the large drop in Stock Duration is
reversed.WithrespecttoincreasesinStockDuration,Figure5Bshowsthatmarkettobookratiosare
lowest(ataround2.6)int=0,i.e.,theyearinwhichStockDurationgoesup.StockDurationdecreases
substantiallyintheyearsbeforethisincrease,comingdownfromaround3.2fiveyearsbeforethedrop,
anditincreasesagainthereafter.
3.2MisvaluationandStockDuration:ResultsfromSubsamples

The negative contemporaneous and positive predictive associations of Stock Duration with

misvaluationcreateapricereversalpatternthatissuggestiveofatemporarydistortioninstockprices.If
so, this is evidence of limits to arbitrage as in Shleifer and Vishny (1990), such that it is important to

We note as an exception the results in column 6 of Panel A where the predictive and thus most crucial
associationbetweenStockDurationandthemarkettobookratioremainspositiveandstatisticallysignificant,but
statistical significance of the contemporaneous association between Stock Duration and markettobook is
subsumed by institutional turnover (though the economic effect is still positive). Both variables are statistically
significantinPanelBwhereweusethePastorVeronesimisvaluationproxy.

16

uncover what frictions prevent arbitrageurs from trading these away. We consider three stock
characteristics that the literature has identified as associated with more limitations to arbitrage: (i)
lowerinstitutionalinvestorholdings;(ii)higherilliquidityasproxiedbytheAmihud(2002)measure;and
(iii) higher idiosyncratic volatility. The first characteristic relates to stocks that are more likely to be
subjecttoshortsalesconstraints(e.g.,Nagel(2005)),andthelattertwomoregenerallytostocksthat
aremoredifficulttotrade,implyingalsogreaterlimitstoarbitrage(e.g.,ShleiferandVishny(1997)).For
eachofthesethreecharacteristics,wecreatesubsamplesbasedonwhetherafirmscharacteristicfalls
beloworabovethesamplemedianeachyear.

TheresultsarereportedinTable5,withmarkettobookratioregressionsprovidedinPanelA,

andthoseusingthePastorVeronesiproxyreportedinPanelB.Forbothproxies,weshowregressions
for all subsamplesseparately, and provide pvalues of Chow tests comparing the coefficients of Stock
DurationandlaggedStockDurationacrosssubsamples.10Consistentlyacrossallthreemeasures,wefind
that both the contemporaneously negative and the predictive position association between Stock
Durationandpricelevelsisonlyapparentinthesubsampleofstockswithgreaterlimitstoarbitrage,i.e.
stocks that have lower institutional holdings, are less liquid, and have higher idiosyncratic volatility.
These results are again similar using either the markettobook ratio or the misvaluation proxy.
AppendixA7alsoshowsthatourresultsarerobusttousingtheHobergandPhillips(2010)misvaluation
measure.
4.ReturnPredictability
4.1ResultsfromPortfolioSorts

Inthissection,weuseassetpricingteststoexaminereturnpredictabilityconditionalonStock

Duration and the (mis)valuation measures. We proceeds in three steps, each step considering an
additional aspect of the three components that make up the reversal pattern: Stock Duration,
misvaluation,andlongtermpastperformance.Inthefirststep,wesortportfoliosbasedonlyonStock
Duration,inthesecondstepwedodoublesortsonbothStockDurationandmisvaluation,andfinallyin
thethirdstepweconsidertriplesortsonStockDuration,misvaluation,andlongtermpastperformance.

10

The Chow tests are performed based on pooled regressions that include a dummy variable that indicates
whetherafirmbelongstothehighorlowgroup,andaninteractionofthisvariablewithStockDurationandlagged
StockDuration.

17

Theresultsfrombasicportfolio sortsonStockDuration arereported inTable6. InPanelA of


Table6wesortthesampleintoquintilesbasedonStockDurationattheendofthequarter,calculating
returnsforthenextfourquartersandupdatingtheportfolioeveryquarter.Allofourportfoliosortsare
intended to capture any reversal of longerterm mispricing, with longerterm meaning over the next
four quarters. We account for overlapping portfolios by following the methodology in Jegadeesh and
Titman(1993),suchthatstocksrankedineachofthelastfourquartersformonefourthoftheportfolio.
Returns from each of the four subportfolios are equally weighted to calculate the monthly portfolio
returns.WereportmonthlyreturnsfortheCAPM,3factorand4factormodel,andthe4factorloadings
foreachofthefivequintileportfoliosaswellasthelongshortportfoliothatbuysstockswithlongStock
Durations(topquintile)andsellsstockswithshortStockDurations(bottomquintile),

Unconditionally, we find some but not strong evidence for a positive relation between Stock

Durationandfuturestockreturns.Thisfirststepthusprovidesweakinitialsupportforthehypothesis
thatthestocksheldbyshortterminvestorsaremorelikelytobeovervalued.Forexample,asshownin
the second column of Table 6 Panel A, a longshort equalweighted portfolio with long positions in
stocksinthehighestStockDurationquintileandshortpositionsinstocksinthelowestStockDuration
quintileearnsa3factoralphaof(12*0.34%=)4.1%peryear(tstatisticof2.65).Thenegativealphaof
stocks held by shortduration investors is generally not significant though, and disappears after
controllingformomentum.Thepositivealphaofstocksheldbylongdurationinvestorsissignificantand
robusttocontrollingformomentum.
Constituting our second step, a stronger test of the hypothesis that the reversal pattern that
wasasdocumentedinthemarkettobookregressionsactuallyconstitutestemporarypricedistortionsis
to consider the joint importance of Stock Duration and misvaluation for stock return predictability.
Intuitively, if stocks held by the shortterm (longterm) investors are more likely to be overvalued
(undervalued), then the above return reversal/predictability results should be stronger conditional on
stocks that may be more misvalued. We examine this by independently double sorting stocks into
quintilesbasedontheirstockdurationattheendofeachquarter,andintothreegroupsbasedonour
(mis)valuationmeasures.TheresultsarereportedinTable6usinginPanelBthemarkettobookratio
and in Panel C the proxy from Pastor and Veronesi (2003). We again report monthly alphas for the
CAPM,3factorand4factormodels.

18

We find that stocks with high valuations have negative future alphas and thus appear
overvaluedonlywhenheldbyshortterminvestors.Tothecontrary,stockswithhighvaluationstendto
havefuturepositivealphasifthesestocksareheldbymorelongterminvestors.Forexample,inTable6
PanelBtheCAPMalphaforstocksinthehighmarkettobookratioquintileandthelowStockDuration
quintile equals 0.41% per month (tstatistic of 1.92), whereas the CAPM alpha for stocks in the high
markettobookratiogroupandthehighStockDurationgroupispositive(0.18%permonth),withat
statistic of 2.20. Thedifference of 0.59% per month betweenthe two Stock Duration groups is highly
significant (tstatistic = 2.91). This difference in returns remains statistically significant for Fama and
French 3factor alphas (0.45% monthly, tstatistic = 3.47) and for Carhart 4factor alphas (0.31%, t
statistic=2.49).AsreportedinPanelC,theresultsaresimilarwhenusingthealternativemisvaluation
measure from Pastor and Veronesi (2003). For example, the CAPM alpha for stocks that appear most
overvalued equals 0.40% per month (tstatistic of 1.78) for stocks that are also in the shortest Stock
Duration quintile, but it equals 0.23% per month (tstatistic of 2.63) for stocks that are also in the
longestStockDurationquintile.Thedifferenceof0.62%isagainhighlysignificant.Ourresultsaresimilar
ifweusetheHobergPhillipsmisvaluationmeasure(seeAppendixA8PanelA).
To complement this analysis, Figure 6 provides cumulative stock returns of different Stock
Durationportfolios. The figure shows thatunconditionally, one dollar invested in a portfolio of stocks
with long Stock Duration (top quintile) outperforms a portfolio of stocks with short Stock Duration
(bottom quintile), with a longshort portfolio that is long (short) in stocks with long (short) Stock
Durationprovidingacumulativereturnofabout170%overtheperiod1984to2011.Mostimportant,
thesedifferencesinstockreturnsarestrongestforaportfoliothatalsoconditionsonmarkettobook
ratios, with a longshort portfolio in stocks in the highest markettobook ratio group (top tercile)
yieldingareturnofabout280%,versusonly37%forstocksinthelowestmarkettobookratiogroup
(bottomtercile).Takentogether,alltheseresultsprovideevidencein supportofthe maintheoretical
prediction in Bolton, Scheinkman and Xiong (2008), namely that stocks withhigh valuations are more
likelytobeovervaluedinthepresenceofshortterminvestors.Moreover,thisovervaluationdoesnot
existorislesssevereinthepresenceoflongterminvestors.
Interestingly, we find that the documented overvaluation pattern is not present when using
shareturnover(Table6PanelD),institutionalturnover(PanelE),ortransientinvestors(PanelF).Panel
D, for example, shows for stocks that are more likely to be overvalued that a portfolio with long
positionsinstocksinthelowestshareturnoverquintileandshortpositionsinstocksinthehighestshare
19

turnoverquintiledoesnotproduceeconomicallyorstatisticallysignificantalphas.Thisisconsistentwith
ourfindingsinTable4,wherewedocumentedthatthereversalpatternsresultsareverydifferentwhen
anyofthealternativeinvestmenthorizonmeasuresareused.
The analysis in Table 6 also addresses a closely related question, namely to what extent the
presence of investors with shortdurations could explain the value premium, i.e. that stocks with low
markettobookratios(valuestocks)tendtooutperformstockswithhighmarkettobookratios(growth
stocks). If shortduration traders distort prices, particularly allowing highly valued stocks to be
temporarily overvalued, then we would expect that the value premium is stronger for stocks held
primarilybyinstitutionalinvestorswithshorterdurations.OurempiricalresultsinPanelBandCofTable
6areconsistentwiththisintuition.Theyshowthatthevaluepremiuminoursampleisdrivenbystocks
withshorterStockDurations,whilethereisnovaluepremiumamongstockswithlongStockDurations.
Forexample,usingthereturnsreportedinTable6PanelB,wefindthatifweonlyconsiderstocksinthe
shortestdurationquintile,thenalongshortequallyweightedportfoliothatbuys(sells)stockswithlow
(high)markettobookratioshasanannualized4factoralphaof(12*0.55%=)6.6%(tstatisticof3.82),
andaCAPMalphaof(12*0.68=)8.2%peryear(tstatisticof2.95).Ontheotherhand,thecorresponding
longshort annualized 4factor alpha for stocks in the highest duration quintile is only 1.6%, which is
statistically insignificant (tstatistic of 1.34). The difference between the longshort portfolios for long
andshortdurationinvestorsisalsohighlyeconomicallyandstatisticallysignificant(withatstatisticof
2.55incaseofthe4factoralphas).ResultsaresimilarusingthePastorVeronesimisvaluationmeasure
inTable6PanelC.Asaresult,thesefindingssuggestthatthevaluepremiumisdrivenbystockswith
moreshortterminstitutions.

Ourthirdandfinalstepincorporatesthenotionthatifreversalpatternsindicatethatstocksare

temporarily overvalued, past stock returns should play an important role to corroborate any
overvaluation.Aslargechangesinthepresenceofshortterminvestorscantaketwotothreeyearsto
materialize,weconsiderthestockreturninthepast36months.Ifthatpastthreeyearstockreturnis
relatively high and the stock has a high valuation and the stock has a short Stock Duration, it seems
more likely that that stock is overvalued. In other words, we would expect previously documented
predictabilitytobestrongerforstockswheresuchpredictabilitywouldindeedcreateareversalpattern.
This would imply that negative future alphas of stocks held largely by shortterm investors should be
more prevalent for stocks with relatively high returns in the past three years and high current
valuations.
20

We formally study this by conducting 5x3x2 independent triple sorts on Stock Duration,
misvaluation,andthreeyearpastreturns.ThecorrespondingresultsarereportedinTable7,withPanel
A reporting 3factor alphas and Panel B reporting 4factor alphas. We only report monthly alphas for
stocks in the highest and lowest Stock Duration quintiles. Panel A shows that for growth stocks with
short Stock Durations and with low past stock returns that there is no evidence for overvaluation or
predictability,withanannualized3factoralphaofonly0.51%(tstatisticof0.27).Incontrast,growth
stockswithshortStockDurationsandhighpaststockreturnsseemovervalued,withanannualized3
factoralphaof3.7%(tstatisticof2.61).
TheresultsinPanelCandDofTable7showthatthereisalsoadirectconnectiontopastsix
monthmomentum.Specifically,stockswithhigh(mis)valuationsheldbyshortterminvestorsthatalso
experiencednegativesixmonthmomentumseemgoodcandidatesforstockswhoseovervaluationhas
recently started to become recognized. For these stocks, we find in Table 7 Panel C the strongest
reversal, with an annualized 3factor alpha of 7.08% (tstatistic of 3.56). In contrast, stocks with a
similar valuation and a similar presence of shortterm investors but with positive past sixmonth
momentumhavenosubsequenteconomicallylargenegativealphas(annualized3factoralphaofonly
0.94%,withatstatisticof0.69).Combinedwiththeresultswereweconditionedonthepastthreeyear
returns,thesefindingsthussupportthenotionthatthesehighvaluations(thoughonlyslowlyreversed)
areindeedreflectingmisvaluationsratherthan,forexample,superiorgrowthopportunitiesoffirms.
Ourresultsareconsistentwithandexpandtheexistingliteratureconsideringtheinteractionof
valuegrowth strategies with either momentum or the presence of shortterm investors, see, for
example,Asness(1997)andLeeandSwaminathan(2000).Asness(1997)alsofindsthatgrowth(value)
strategiesworkbestforhigh(low)momentumstocks,andLeeandSwaminathan(2000)notethatstocks
withlow(high)shareturnovertendtohavevalue(growth)characteristics.
4.2PortfolioResultsforSubsamples:TheRoleofLimitstoArbitrage
Havingshownthatthestocksheldbyshortterminvestorsaremorelikelytobeovervalued,we
next examine the effects of limits to arbitrage on this overvaluation. The corresponding results are
presentedinTable8. Tocreatethistable, weform5x3x2 independenttriplesorts on StockDuration,
markettobook, and proxies for arbitrage costs. Consistent with the analysis in Table 5, we use
institutional ownership (Panel A), the Amihud illiquidity measure (Panel B), and idiosyncratic volatility
(PanelC)asproxiesforarbitragecosts.Inallthreepanel,wereportboth3factorand4factoralphas.
21

WefindthatthepatternsdocumentedinTable6,namelythatstockswithhighvaluationshave
negativefuturealphasonlyifheldbyshortterminvestors,aredrivenbythesubsetofstockswithhigh
arbitragecosts.Forexample,wefindinPanelAofTable8thatthemonthly3factoralphasforstocks
withhighvaluationsthatareheldbyshortterminvestorsequal0.49%(tstatisticof2.81)ifinstitutional
investor holdings are low. However, these 3factor alphas are not statistically and economically
significant if institutional investor holdings are large (0.11% per month, tstatistic of 0.61). Arbitrage
costsdonotseemtomatterifstocksareheldbyinvestorwithlongtermdurations.
The results are similar for the other two proxies of arbitrage costs. In Table 8 Panel C, for
example,thedifferencein4factoralphasbetweenhighandlowStockDurationquintilestockswithhigh
markettobookratiosequals6.4%peryear(tstatisticsof2.39)ifidiosyncraticvolatilityishigh,butonly
0.6% per year (tstatistic of 0.42) if idiosyncratic volatility is low. Again, this confirms the notion that
predictabilityisstrongestandmisvaluationmostlikelyifshortterminvestorsarepresentandiflimitsto
arbitragearehigh.OurresultsaresimilarifweusethePastorVeronesiorHobergPhillipsmisvaluation
measures(seeAppendixA8PanelBtoD).
4.3ResultsfromFamaMacBethRegressions
Wefinallyexaminetheeffectofinvestorhorizononstockreturnpredictabilityinamultivariate
regression setting. We use the FamaMacBeth (1973) methodology and estimate predictive cross
sectional regressions of next oneyear raw returns or next oneyear DGTW adjusted cumulative
abnormal return on stock duration, while controlling for other stock characteristics (see Daniel et al.
(1997)andWermers(2003)).ResultsarepresentedinTable9.
TheregressionestimatesaregenerallyconsistentwiththeportfolioresultsreportedinTables6
and 8. For example, the coefficient of Stock Duration is positive and highly significant in most of the
specifications,forexample,incolumn3thevalueofthecoefficientis0.023,withatstatisticof3.77.
ThecoefficientforStockDurationisrobusttousingDGTWadjustedreturnsasthedependentvariable,
forexampleincolumn9ofthetable.Incolumns4and10,wefindthatthecoefficientoftheinteraction
between Stock Duration and a high markettobook ratio dummy (top tercile) is positive and highly
statistically significant. This confirms that return predictability based on stock duration is stronger for
moreovervaluedstocks.NotethatourresultsoninstitutionalturnoverareconsistentwiththoseinYan
andZhang(2009).Likeintheirwork,wealsofindthatfirmsheldbyinstitutionsthattendtotrademore

22

have higher future stock returns (see column 3). We find that this result does not appear to be very
robustinoursamplethough(seecolumns2and9).

In the regressions in columns 5 to 7, we include as additional independent variables the

interaction terms between Stock Duration and dummy variables for three proxies of arbitrage costs,
namelyinstitutionalownership,theAmihudilliquiditymeasure,andidiosyncraticvolatility.Toconstruct
thedummyvariables,wedividethestockseachquarterintothreetercilegroupsbasedononeofthese
threeproxiesoflimitstoarbitrage.Therefore,thelowinstitutionalinvestorholdingsdummyequalsone
for stocks in the lowest institutional investor holdings tercile. Similarly, the high Amihud illiquidity
dummyequalsoneforstocksinthehighestAmihudilliquiditymeasuretercile,andthehighidiosyncratic
volatilitydummyequalsoneifstocksbelongtothehighestidiosyncraticvolatilitytercile.Theresultsin
columns5to7ofTable9showthatthecoefficientscorrespondingtotheinteractionsbetweenthese
high arbitrage costs dummies and stock duration are positive and highly significant in all three
regression specifications. This suggests again that predictability is stronger in the subsampleof stocks
with higher arbitrage costs. For example, one standard deviation decrease in stock duration leads to
103 bp lower returns over the next one year for stocks in the bottom institutional ownership tercile
comparedtothoseinthetopinstitutionalownershiptercile.
5.Conclusions

This paper empirically studies whether the presence of shortterm investors is related to a

speculativecomponentinstockprices,i.e.,todeviationsofstockpricesfromfundamentalsoverlonger
periods of time, as predicted, for example, by Bolton, Scheinkman, and Xiong (2006). Their paper
providesaformaltheorywhichpositsthatshortterminvestorswithheterogeneousinvestorcombined
with limits to arbitrage can create temporary price distortions and a speculative component in stock
prices.ArelatedexplanationisgiveninDaniel,Hirshleifer,andSubrahmanyam(1998),whointroducea
theorythatincorporatesinvestoroverconfidence,whichresultsininvestoroverreactiontopublicnews.
Shortterm investors may be more likely to be prone to such overconfidence, as suggested by, for
example,byDeBondtandThaler(1995)andDeLong,Shleifer,SummersandWaldmann(1991).

Westudythisquestionusinganewmeasureoftheholdingdurationofinstitutionalinvestors,

called Stock Duration, which is defined as the weightedaverage length of time that institutional
investors have held a stock in their portfolios, and was introduced in Cremers and Pareek (2012). In
contrast to other measures such as share turnover or institutional turnover, our new proxy can be
23

measuredatadeaggregatedlevel,witheachinstitutionalinvestorbeingpotentiallyshortterminsome
stocksandlongterminothers.

Usingthisnewmeasure,ourpaperprovidestwomaincontributions.First,wecharacterizethe

evolutionofStockDurationovertime,andsurprisinglyfindthatholdingdurationshavebeenstableand,
ifanything,slightlylengthenedsince1985.TheaverageStockDurationequalsabout1.2yearsin1985
andincreasestoaround1.5yearsin2010.ThisincreaseinStockDurationmaybesomewhatsurprising
asshareturnoverhassignificantlyincreasedoverthesameperiod(fromanaverageannualizedturnover
of72%in1985tocloseto300%in2010),whileinstitutionalinvestorsasagrouphaveincreasedtheir
averageownershipinstocksfrom33%in1985to75%attheendofoursamplein2010.However,the
simultaneous increase in both share turnover and Stock Durations can be explained by two separate
phenomena occurring during our time period: the increase in automated and high frequency trading,
andthemajorshifttowards(closet)indexingwhichisinherentlylongerterm.

Second, we document that the presence of shortterm institutions is strongly related to

temporarypricedistortions.Whenshortterminvestorsmoveinto(outof)stocks,theirpricestendtogo
up(down)relativetofundamentals.Wedocumentsuchpredictabilityusingbothvaluationproxiesand
traditional asset pricing methodology. We cannot detect a similar a price reversal pattern for either
share turnover, institutional turnover or the presence of transient investors, while results for Stock
Durationarerobusttotheirinclusion.Finally,predictabilityisstrongerinthesubsampleofstockswith
higher arbitrage costs, such as for stocks with low institutional ownership, high idiosyncratic risk, and
highlevelsofilliquidity.

Overall, our findings are consistent with a speculative stock component in stock prices as

modeled in Bolton, Scheinkman, and Xiong (2006). Our findings raise interesting questions for future
research,forexample,whatthereasonsfortradingbyshortterminstitutionsareorhowchangesinthe
presenceofshortterminvestorinteractwithmanagerialdecisionmakingandcompensation.

24

References
Amihud, Yakov, 2002, Illiquidity and stock returns: crosssection and timeseries effects, Journal of
FinancialMarkets5,3156.
Asness,CliffordS.,1997,Theinteractionofvalueandmomentumstrategies,FinancialAnalystsJournal,
March/April,2936.
Bainbridge,Stephen,2006,Directoryprimacy andshareholder disempowerment,HarvardLawReview
119,17351749.
Bebchuk,Lucian,2005,Thecaseforincreasingshareholderpower,HarvardLawReview118,833861.
Blair, Margaret M. and Lynn A. Stout, 1999, A team production theory of corporate law, Virginia Law
Review85,247286.
Bushee, Brian, J., 1998, The influence of institutional investors on myopic R&D investment behavior,
AccountingReview73,305333.
Bushee, Brian, J., 2001, Do institutional investors prefer nearterm earnings over longrun value?,
ContemporaryAccountingResearch18,20746.
Bolton, Patrick, Jose Scheinkman, and Wei Xiong, 2006, Executive compensation and shortterm
behaviorinspeculativemarkets,ReviewofEconomicStudies73,557610.
Coval, Joshua, and Erik Stafford, 2007, Asset fire sales (and purchases) in equity markets, Journal of
FinancialEconomics86,479512.
Cremers,K.J.Martijn,andAnttiPetajisto,2009,Howactiveisyourfundmanager?Anewmeasurethat
predictsperformance,ReviewofFinancialStudies22,33293365.
Cremers,K.J. Martijn, Miguel A.Ferreira,PedroP. Matos andLaura T. Starks, 2012, Themutualfund
industryworldwide:Explicitandclosetindexing,feesandperformance,Workingpaper.
Cremers, K. J. Martijn, and Ankur Pareek, 2012, Shortterm trading and Stock Return Anomalies:
Momentum,Reversal,Accruals,ShareIssuanceandR&DIncreases,WorkingPaperUniversityofNotre
DameandRutgersUniversity.

25

Daniel, Kent, Mark Grinblatt, Sheridan Titman, and Russ Wermers, 1997, Measuring Mutual Fund
PerformancewithCharacteristicBasedBenchmarks,JournalofFinance52,10351058.
Daniel, Kent, David Hirshleifer, and Avanidhar Subrahmanyam, 1998, Investor psychology and security
marketunderandoverreactions,JournalofFinance53,18391885.
De Bondt, Werner F.M., and Richard Thaler, 1995, Financial decisionmaking in markets and firms: A
behaviouralperspective.In:Jarrow,R.etal.(Ed.),HandbooksinOperationsResearchandManagement,
Vol.9,ElsevierScience,Amsterdam,385410.
DeLong,J.Bradford,AndreiShleifer,LawrenceH.Summers,andRobertJ.Waldmann,1991,Thesurvival
ofnoisetradersinfinancialmarkets,JournalofBusiness64,119.
Derrien,Francois,AmbrusKecskes,andDavidThesmar,2012,Investorhorizonsandcorporatepolicies,
JournalofFinancialandQuantitativeAnalysis,forthcoming.
Froot, Kenneth A., Andre F. Perold, and Jeremy C. Stein, 1992, Shareholder trading practices and
corporateinvestmenthorizons,JournalofAppliedCorporateFinanceSummer,4258.
Gaspar,JosMiguel,MassimoMassa,andPedroMatos,2005,Shareholderinvestmenthorizonsandthe
marketforcorporatecontrol,JournalofFinancialEconomics76,135165.
Gaspar, JosMiguel,MassimoMassa,PedroMatos, RajdeepPatgiri, and ZahidRehman,2012,Payout
policychoicesandshareholderinvestmenthorizons,ReviewofFinanceforthcoming.
Gopalan, Radhakrishnan, Todd Milbourn, Fenghua Song, and Anjan V. Thakor, 2012, Duration of
executivecompensation,WorkingPaper,WashingtonUniversityinSt.Louis.
Graham,JohnR.,CampbellR.Harvey,andShivaRajgopal,2005,Theeconomicimplicationsofcorporate
financialreporting,JournalofAccountingandEconomics40,373.
Harford,Jarrad,DirkJenter,andKaiLi,2011,Institutionalcrossholdingsandtheireffectonacquisition
decisions,JournalofFinancialEconomics99,2739.
Hendershott, Terrence, Charles M. Jones and Albert J. Menkveld, 2011, Does algorithmic trading
improveliquidity?,JournalofFinance66,133.

26

Hoberg, Gerald, and Gordon Phillips, 2010, Real and financial industry booms and busts, Journal of
Finance65,4586.
Jegadeesh, Narasimhan, and Sheridan Titman, 1993, Returns to buying winners and selling losers:
Implicationsforstockmarketefficiency,JournalofFinance48,6591.
Laverty, Kevin J., 1996, Economic shorttermism: The debate, the unresolved issues, and the
implicationsformanagementpracticeandresearch,AcademyofManagementReview21,825860.
Lee,CharlesM.C.,andBhaskaranSwaminathan,2000,Pricemomentumandtradingvolume,Journalof
Finance55,20172069.
Litzenberger, Robert J., Jeff Castura and Richard Gorelick, 2012, The impacts of automation and high
frequencytradingonmarketquality,AnnualReviewofFinancialEconomics4,5998.
Menkveld, Albert J., 2011, Electronic trading and market structure, UK Government Foresight Driver
Review16.
Nagel,Stefan,2005,Shortsales,institutionalinvestorsandthecrosssectionofstockreturns,Journalof
FinancialEconomics78,277309.
NarayananMP.,1985,Managerialincentivesforshorttermresults,JournalofFinance40,14691484.
Newey,WhitneyK.,andKennethD.West,1987,Asimple,positivesemidefinite,heteroskedasticityand
autocorrelationconsistentcovariancematrix,Econometrica55,703708.
Pastor, Lubos, and Pietro Veronesi, 2003, Stock valuation and learning about profitability, Journal of
Finance58,17491789.
Phelps,EdmundS.,2010,ShorttermismisunderminingAmerica,NewPerspectivesQuarterly27,1719.
Polk, Christopher, and Paola Sapienza, 2009, The stock market and corporate investment: A test of
cateringtheory,ReviewofFinancialStudies22,187217.
Scheinkman, Jose, and Wei Xiong, 2003, Overconfidence and speculative bubbles, Journal of Political
Economy,111,11831219.

27

Stohs,MarkHoven,andDavidC.Mauer,1996,Thedeterminantsofcorporatedebtmaturitystructure,
JournalofBusiness69,279312.
Shleifer, Andrei, and Robert W. Vishny, 1990, Equilibrium short horizons of investors and firms,
AmericanEconomicReviewP&P80,148153.
Shleifer,Andrei,andRobertW.Vishny,1997,Thelimitsofarbitrage,JournalofFinance52,3555.
Stein,JeremyC.,1988,Takeoverthreatsandmanagerialmyopia,JournalofPoliticalEconomy96,6180.
Stein, Jeremy C., Efficient capital markets, inefficient firms: A model of myopic corporate behavior,
QuarterlyJournalofEconomics104,655669.
Strine,LeoE.,2007,Towardcommonsenseandcommonground?Reflectionsonthesharedinterestsof
managersandlaborinamorerationalsystemofcorporategovernance,JournalofCorporationLaw33,
120.
TheEconomist,2012,Takingthelongview.Thepursuitofshareholdervalueisattractingcriticismnot
allofitfoolish,TheEconomist,IssueofNovember24.
Yan,Xuemin,andZheZhang,2009,Institutionalinvestorsandequityreturns:Areshortterminstitutions
betterinformed?ReviewofFinancialStudies22,893924.
Wermers, Russ, 2003, Is money really smart? New evidence on the relation between mutual fund
flows,managerbehavior,andperformancepersistence,WorkingPaper,UniversityofMaryland.

28

Table1:SummaryStatistics
Thistableprovidessummarystatisticsofthevariablesusedintheempiricalanalysis.ThesampleconsistsofUSfirmsfromCompustat.
WeexcludefirmsthatareinthebottomNYSEsizedecileandfirmswithstockpricesbelowUSD1attheendofafiscalyear.Thesample
periodis1985to2010.WereportthesesummarystatisticsforallobservationsusedintheregressionsreportedinColumn1ofTable3.
Allvariablesarewinsorizedat1%.VariablesaredefinedinAppendixA1.

Variable

Mean

Median

STD

25%

75%

StockDuration
ChangeStockDuration(tminust1)
InstitutionalTurnover
TransientInvestors
ShortTermOwnership
ChangeShortTermOwnership(tminust1)
ShareTurnover
InstitutionalInvestorHoldings
MBRatio
MBResidualPV
MBResidualHP
MarketCap(inmUSD)
AssetHorizon
R&D/Assets
Capex/Assets
PPE/Assets
Debt/Assets
Cash/Assets
SalesGrowth
Log(Assets)
Assets
AmihudIlliquidity
IdiosyncraticVolatility
BankStockDuration
InvestmentCompanyStockDuration
PensionFundsStockDuration
OthersStockDuration
BankHoldings
InvestmentCompanyHoldings
PensionFundsHoldings
OthersHoldings

1.41
0.05
0.28
12.87
0.42
0.02
0.01
53.12
2.94
0.31
0.22
3304
10.65
0.04
0.07
0.59
0.19
0.14
0.16
6.49
2635
0.12
0.02
1.53
1.29
1.73
1.22
14.01
32.50
3.06
1.54

1.39
0.07
0.27
10.07
0.39
0.01
0.00
54.40
2.16
0.25
0.17
564
7.05
0.00
0.05
0.51
0.18
0.07
0.09
6.35
570
0.01
0.02
1.52
1.25
1.73
1.01
13.33
31.28
2.65
0.51

0.54
0.33
0.08
10.79
0.22
0.18
0.01
23.33
2.53
0.65
0.57
13269
10.46
0.06
0.06
0.37
0.16
0.18
0.40
1.60
6491
0.33
0.01
0.62
0.54
0.92
0.93
7.94
17.50
2.57
2.49

1.02
0.10
0.23
4.62
0.26
0.12
0.00
35.25
1.46
0.13
0.14
213
3.62
0.00
0.03
0.28
0.04
0.02
0.01
5.33
206
0.00
0.01
1.08
0.90
1.02
0.52
7.82
18.38
1.07
0.07

1.76
0.23
0.32
18.32
0.56
0.08
0.01
71.52
3.43
0.69
0.55
1851
13.62
0.04
0.08
0.84
0.31
0.20
0.22
7.54
1890
0.07
0.03
1.94
1.63
2.36
1.67
19.24
45.74
4.06
1.95

Autocorr.
(t,t1)
0.80
0.14
0.53
0.68
0.59
0.31
0.84

Obs.
29418
25554
29413
29413
28977
25079
29418
29418
29418
29418
29418
29418
29418
29418
29418
29418
29418
29418
29418
29418
29418
29418
29418
29407
29405
28313
11501
29407
29405
28313
11501

Table2:StockDurationandRelatedVariablesoverTime
This table provides information of the evolution of Stock Duration, institutional turnover, institutional investor holdings, and share
turnoverovertheperiod1985to2010.Thetablereportsmeanvaluesofthesevariablesforeachyear.ThesampleconsistsofUSfirms
fromCompustat.WeexcludefirmsthatareinthebottomNYSEsizedecileandfirmswithstockpricesbelowUSD1attheendofafiscal
year.Allvariablesarewinsorizedat1%.VariablesaredefinedinAppendixA1.

Year

1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Change(19852010)

StockDuration
Mean
1.23
1.23
1.28
1.40
1.45
1.45
1.44
1.47
1.44
1.44
1.44
1.30
1.30
1.32
1.31
1.29
1.32
1.37
1.41
1.46
1.51
1.52
1.51
1.54
1.38
1.47
20%

Obs.
865
1103
1196
1211
1268
1353
1403
1416
1451
1568
1666
1716
1625
1661
1711
1615
1495
1369
1315
1215
1135
1051
956
974
966
402

InstitutionalTurnover
Mean
0.31
0.31
0.32
0.27
0.26
0.25
0.29
0.28
0.29
0.29
0.30
0.32
0.31
0.29
0.33
0.32
0.28
0.25
0.25
0.26
0.27
0.29
0.30
0.27
0.34
0.32
0.5%

Obs.
1137
1377
1436
1407
1408
1522
1627
1707
1854
1913
1997
2116
1965
1920
1991
1883
1643
1435
1359
1271
1212
1128
1032
1035
1006
409

InstitutionalInvestor
Holdings
Mean
33%
35%
36%
37%
39%
39%
40%
42%
44%
45%
46%
46%
50%
51%
49%
51%
57%
63%
66%
70%
72%
73%
75%
74%
73%
75%
126%

Obs.
1109
1341
1415
1385
1379
1477
1572
1664
1787
1864
1934
2047
1942
1913
1930
1866
1637
1434
1354
1275
1222
1137
1065
1086
1178
430

ShareTurnover
(annualized)
Mean
72%
92%
103%
87%
94%
93%
120%
126%
147%
133%
162%
180%
169%
171%
223%
228%
207%
207%
215%
233%
233%
251%
282%
318%
316%
276%
283%

Obs.
1296
1515
1607
1563
1591
1739
1896
2046
2251
2361
2527
2774
2616
2615
2828
2754
2410
2096
2057
2034
2032
2037
2128
2154
2202
597

Table3:MisvaluationandStockDuration:MainSpecification
ThistableprovidespanelregressionslinkingvaluationproxiesandStockDuration.Weusethemarkettobookratio(MBRatio)andthemisvaluationmeasureproposedbyPastor
andVeronesi(2003)(MBResidualPV)asvaluationproxies.ThesampleconsistsofUSfirmsfromCompustat.WeexcludefirmsthatareinthebottomNYSEsizedecileandfirms
withstockpricesbelowUSD1attheendofafiscalyear.Allvariablesarewinsorizedat1%.VariablesaredefinedinAppendixA1.tstatistics,calculatedbasedonrobuststandard
errorsclusteredatthefirmlevel,arereportedinparentheses.***,**,*indicatesignificancelevelsof1%,5%,and10%,respectively.

StockDuration

(1)
0.192***
(4.39)

StockDuration(t1)

(2)

0.085**
(2.00)

StockDuration(t2)
InstitutionalInvestorHoldings
ShareTurnover(coeff./100)
AssetHorizon
Cash/Assets
SalesGrowth
PPE/Assets
Log(Assets)
Debt/Assets
Capex/Assets
R&D/Assets
R&DMissing
AmihudIlliquidity
IdiosyncraticVolatility
FirmFixedEffects
YearFixedEffects
Obs.
adj.Rsq.

MBRatio
(3)

(4)
0.307***
(6.57)
0.242***
(5.99)

0.011***
(5.39)
0.302***
(5.56)
0.011
(1.36)
1.684***
(5.65)
0.637***
(9.58)
0.403
(1.57)
0.850***
(10.51)
2.570***
(9.59)
3.201***
(6.59)
6.406***
(5.41)
0.187
(1.56)
0.114
(1.61)
15.051***
(7.28)

(5)
0.269***
(5.43)
0.216***
(5.74)
0.071
(1.56)
0.010***
(4.78)
0.350***
(5.92)
0.012
(1.33)
1.750***
(5.46)
0.676***
(8.24)
0.389
(1.41)
0.859***
(9.88)
2.511***
(8.72)
3.165***
(6.13)
6.490***
(4.92)
0.196
(1.49)
0.152**
(1.98)
13.646***
(6.21)

0.009***
(5.07)
0.286***
(5.66)
0.010
(1.48)
1.369***
(5.02)
0.531***
(9.53)
0.391*
(1.71)
0.874***
(11.85)
2.531***
(10.11)
3.177***
(7.13)
6.178***
(5.89)
0.163
(1.46)
0.088
(1.51)
14.331***
(7.47)

0.010***
(5.39)
0.367***
(7.02)
0.011
(1.26)
1.701***
(5.79)
0.642***
(9.62)
0.421*
(1.65)
0.870***
(10.87)
2.610***
(9.87)
3.169***
(6.56)
6.136***
(5.24)
0.173
(1.46)
0.140*
(1.95)
14.749***
(7.21)

0.121**
(2.54)
0.010***
(4.85)
0.375***
(6.91)
0.012
(1.38)
1.766***
(5.67)
0.688***
(8.55)
0.425
(1.57)
0.877***
(10.32)
2.473***
(8.83)
3.139***
(6.15)
6.278***
(4.85)
0.183
(1.42)
0.159**
(2.03)
13.455***
(6.28)

Yes
Yes

Yes
Yes

Yes
Yes

Yes
Yes

Yes
Yes

29418
0.131

25841
0.133

22836
0.134

25554
0.135

22362
0.133

(6)
0.069***
(5.94)

(7)

MBResidualPV
(8)

0.019*
(1.68)

(9)
0.104***
(8.58)
0.073***
(6.86)

0.003***
(5.42)
0.097***
(7.63)
0.002
(1.17)
0.461***
(6.52)
0.197***
(12.99)
0.176***
(3.13)
0.193***
(10.28)
0.649***
(10.94)
1.135***
(9.41)
0.846***
(4.10)
0.030
(0.84)
0.036*
(1.81)
7.240***
(13.14)

(10)
0.099***
(7.64)
0.076***
(7.76)
0.009
(0.79)
0.003***
(4.83)
0.107***
(7.52)
0.002
(1.41)
0.482***
(6.38)
0.212***
(11.49)
0.167***
(2.71)
0.191***
(9.52)
0.639***
(10.08)
1.087***
(8.46)
0.868***
(3.86)
0.041
(1.07)
0.060**
(2.57)
7.148***
(11.99)

0.002***
(4.90)
0.091***
(7.60)
0.002
(1.17)
0.397***
(6.04)
0.166***
(13.30)
0.183***
(3.49)
0.205***
(11.70)
0.674***
(11.90)
1.095***
(9.89)
0.805***
(4.34)
0.026
(0.78)
0.002
(0.15)
7.028***
(13.87)

0.003***
(5.37)
0.115***
(9.51)
0.002
(1.14)
0.470***
(6.73)
0.200***
(13.12)
0.186***
(3.33)
0.198***
(10.65)
0.658***
(11.21)
1.136***
(9.45)
0.789***
(3.89)
0.027
(0.77)
0.045**
(2.24)
7.162***
(13.09)

0.026**
(2.19)
0.003***
(4.75)
0.113***
(8.56)
0.002
(1.47)
0.498***
(6.73)
0.215***
(11.80)
0.176***
(2.92)
0.194***
(9.89)
0.629***
(10.10)
1.086***
(8.50)
0.860***
(3.94)
0.039
(1.05)
0.062**
(2.56)
7.184***
(12.19)

Yes
Yes

Yes
Yes

Yes
Yes

Yes
Yes

Yes
Yes

29418
0.133

25841
0.135

22836
0.133

25554
0.139

22362
0.138

Table4:MarkettoBookRatiosandStockDuration:ControllingforOtherMeasuresofInvestorHorizon
This table provides panel regressions linking valuation proxies with Stock Duration, share turnover, institutional turnover, and a
measure for transient investors. Panel A provides regressions where we use the markettobook ratio (MB Ratio) as a proxy for
misvaluation,whilePanelBprovidesregressionswhereweusethemisvaluationmeasureproposedbyPastorandVeronesi(2003)(MB
ResidualPV).ThesampleconsistsofUSfirmsfromCompustat.WeexcludefirmsthatareinthebottomNYSEsizedecileandfirmswith
stockpricesbelowUSD1attheendofafiscalyear.Allvariablesarewinsorizedat1%.VariablesaredefinedinAppendixA1.tstatistics,
calculatedbasedonrobuststandarderrorsclusteredatthefirmlevel,arereportedinparentheses.***,**,*indicatesignificancelevels
of1%,5%,and10%,respectively.

PanelA:MBRatio

StockDuration

(1)
0.192***
(4.39)

StockDuration(t1)
ShareTurnover(coeff./100)

(2)

0.061
(1.39)
0.286***
(5.66)

ShareTurnover(t1)(coeff./100)

0.130**
(2.43)

(3)
0.302***
(6.51)
0.230***
(5.59)
0.327***
(5.92)
0.052
(0.98)

InstitutionalTurnover

(4)
0.007
(0.15)

(7)
0.070*
(1.65)

(8)

0.125***
(2.88)

(6)
0.055
(1.20)
0.178***
(4.40)

0.076*
(1.80)

(9)
0.165***
(3.75)
0.200***
(5.07)

1.675***
(6.15)

3.990***
(13.93)
1.091***
(4.22)

0.011***
(5.24)

0.039***
(17.79)
0.003
(1.64)

MBRatio
(5)

4.233***
(15.82)

InstitutionalTurnover(t1)
TransientInvestors

0.043***
(19.69)

TransientInvestors(t1)
ControlsasinTable3
FirmFixedEffects
YearFixedEffects
Obs.
adj.Rsq.

Yes
Yes
Yes

Yes
Yes
Yes

Yes
Yes
Yes

Yes
Yes
Yes

Yes
Yes
Yes

Yes
Yes
Yes

Yes
Yes
Yes

Yes
Yes
Yes

Yes
Yes
Yes

29418
0.131

25851
0.128

25547
0.134

31141
0.143

27349
0.131

27049
0.146

31141
0.149

27349
0.131

27049
0.151

(7)
0.032***
(2.90)

(8)

0.034***
(2.93)

(6)
0.029**
(2.44)
0.057***
(5.19)

0.014
(1.30)

(9)
0.062***
(5.37)
0.059***
(5.67)

0.569***
(8.22)

1.222***
(17.62)
0.389***
(6.03)

0.003***
(5.62)

0.012***
(22.36)
0.001
(1.25)

PanelB:MBResidualPV

StockDuration

(1)
0.069***
(5.94)

StockDuration(t1)
ShareTurnover(coeff./100)

(2)

0.009
(0.77)
0.091***
(7.60)

ShareTurnover(t1)(coeff./100)

0.031**
(2.35)

(3)
0.102***
(8.47)
0.067***
(6.20)
0.108***
(8.77)
0.025*
(1.95)

InstitutionalTurnover

(4)
0.014
(1.27)

MBResidualPV
(5)

1.272***
(19.47)

InstitutionalTurnover(t1)
TransientInvestors

0.013***
(24.27)

TransientInvestors(t1)
ControlsasinTable3
FirmFixedEffects
YearFixedEffects
Obs.
adj.Rsq.

Yes
Yes
Yes

Yes
Yes
Yes

Yes
Yes
Yes

Yes
Yes
Yes

Yes
Yes
Yes

Yes
Yes
Yes

Yes
Yes
Yes

Yes
Yes
Yes

Yes
Yes
Yes

29418
0.133

25851
0.127

25547
0.139

31141
0.148

27349
0.131

27049
0.155

31141
0.158

27349
0.128

27049
0.162

Table5:MarkettoBookRatiosandStockDuration:ResultsfromSubsamples
This table provides panel regressions linking valuation proxies and Stock Duration. Panel A provides regressions where we use the
markettobookratio(MBRatio)asaproxyformisvaluation,whilePanelBprovidesregressionswhereweusethemisvaluationmeasure
proposed by Pastorand Veronesi (2003)(MBResidual PV).Weprovidethese regressionsbased onsubsamples using:(i)institutional
investor holdings; (ii) the Amihud illiquidity measure; and (iii) idiosyncratic volatility. We create these subsamples by separating the
sample firms each year based on median values of the respective variables. We also report pvalues of Chowtests comparing the
coefficients of Stock Duration and lagged Stock Duration across subsamples. The sample consists of US firms from Compustat. We
excludefirmsthatareinthebottomNYSEsizedecileandfirmswithstockpricesbelowUSD1attheendofafiscalyear.Allvariablesare
winsorized at 1%. Variables are defined in Appendix A1. tstatistics, calculated based on robust standard errors clustered at the firm
level,arereportedinparentheses.***,**,*indicatesignificancelevelsof1%,5%,and10%,respectively.

PanelA:MBRatio

StockDuration
StockDuration(t1)
ShareTurnover(coeff./100)
AmihudIlliquidity
IdiosyncraticVolatility

Inst.InvestorHoldings
Low
High
(1)
(2)
0.419***
(7.25)
0.384***
(7.22)
0.412***
(4.05)
0.187**
(2.20)
7.050**
(2.36)

pvalueChowtest(StockDuration)
pvalueChowtest(StockDuration(t1))
ControlsasinTable3
FirmFixedEffects
YearFixedEffects
Obs.
adj.Rsq.

0.097
(1.29)
0.049
(0.80)
0.212***
(3.17)
0.231**
(2.06)
19.549***
(7.32)

MBRatio
AmihudIlliquidity
Low
High
(3)
(4)
0.024
(0.26)
0.130
(1.54)
0.084
(1.15)
21.041***
(5.19)
21.796***
(6.82)

0.0310
0.0000

0.370***
(7.59)
0.386***
(9.02)
0.393***
(3.88)
0.275***
(4.39)
5.127**
(2.05)

IdiosyncraticVolatility
Low
High
(5)
(6)
0.087
(1.64)
0.057
(1.25)
0.296***
(3.00)
0.532***
(4.53)
10.111***
(2.66)

0.0000
0.1010

Yes
Yes
Yes

Yes
Yes
Yes

11750
0.142

13804
0.142

0.554***
(6.99)
0.356***
(5.63)
0.233***
(3.45)
0.024
(0.29)
20.780***
(6.06)

0.0000
0.0000

Yes
Yes
Yes

Yes
Yes
Yes

14165
0.165

11389
0.183

Yes
Yes
Yes

Yes
Yes
Yes

15543
0.113

10011
0.182

PanelB:MBResidualPV

StockDuration
StockDuration(t1)
ShareTurnover(coeff./100)
AmihudIlliquidity
IdiosyncraticVolatility

Inst.InvestorHoldings
Low
High
(1)
(2)
0.113***
(7.65)
0.103***
(7.52)
0.111***
(5.24)
0.054**
(2.43)
4.619***
(6.03)

pvalueChowtest(StockDuration)
pvalueChowtest(StockDuration(t1))
ControlsasinTable3
FirmFixedEffects
YearFixedEffects
Obs.
adj.Rsq.

0.068***
(3.33)
0.018
(1.07)
0.091***
(5.56)
0.117**
(2.41)
9.176***
(12.73)

MBResidualPV
AmihudIlliquidity
Low
High
(3)
(4)
0.014
(0.67)
0.032*
(1.77)
0.026
(1.43)
4.707***
(4.68)
9.047***
(11.43)

0.912
0.000

0.114***
(8.20)
0.117***
(9.67)
0.144***
(6.95)
0.089***
(4.77)
3.838***
(6.07)

IdiosyncraticVolatility
Low
High
(5)
(6)
0.036***
(2.60)
0.015
(1.16)
0.081***
(3.37)
0.248***
(5.46)
5.134***
(5.00)

0.000
0.030

Yes
Yes
Yes

Yes
Yes
Yes

11750
0.135

13804
0.164

0.182***
(9.26)
0.116***
(6.89)
0.094***
(6.32)
0.018
(0.83)
7.370***
(9.31)

0.000
0.000

Yes
Yes
Yes

Yes
Yes
Yes

14165
0.185

11389
0.176

Yes
Yes
Yes

Yes
Yes
Yes

15543
0.146

10011
0.160

Table6:ReturnPredictability:PortfolioResults
This table presents monthly equalweighted CAPM alphas, 3factor FamaFrench alphas, and Carhart 4factor alphas for portfolio strategies from unconditional sorts based on
averageStockDuration(PanelA),stockturnover(PanelD),institutionalturnover(PanelE),andownershipoftransientinvestorsasapercentageoftotalinstitutionalownership
(PanelF).PanelsBandCreportequalweightedalphasforconditionalsortsbasedonStockDurationandamisvaluationmeasure(markettobookratioinPanelBandthePastor
andVeronesi(2003)misvaluationproxyinPanelC).Atthebeginningofeachquarter,stocksarefirstdividedintofivegroupsbasedonStockDuration.InPanelsBandCtheyare
thenindependentlydividedintothreegroupsbasedonamisvaluationmeasure.Wethenreportreturnsforthese15portfolioswhicharecalculatedovernextfourquarters.To
accountforoverlappingportfolios,wefollowthemethodologyinJegadeeshandTitman(1993)suchthatstocksrankedineachofthelastfourquartersformonefourthofeach
portfolio.Allthereportedreturnsareinmonthlypercentages.5%significancelevelsaredenotedinboldandtstatisticsarereportedinparentheses.

PanelA:SortsonStockDuration

StockDuration
1(Short)

5(Long)

51

CAPMAlpha
0.15
(0.82)
0.14
(0.95)
0.26
(2.16)
0.33
(2.93)
0.35
(3.08)
0.50
(2.69)

AverageStockDuration
3FactorAlpha
4FactorAlpha
0.15
0.06
(1.40)
(0.74)
0.05
0.20
(0.56)
(2.85)
0.12
0.22
(1.67)
(3.33)
0.17
0.24
(2.24)
(3.23)
0.19
0.23
(2.40)
(2.86)
0.34
0.16
(2.65)
(1.40)

MKTRF
1.19
(63.04)
1.09
(68.41)
1.01
(67.59)
0.95
(57.68)
0.87
(48.95)
0.32
(12.18)

Carhart4FactorLoadings
HML
SMB
0.10
0.87
(3.44)
(32.57)
0.19
0.67
(7.57)
(29.81)
0.34
0.50
(14.79)
(23.81)
0.43
0.36
(16.95)
(15.25)
0.45
0.34
(16.57)
(13.46)
0.55
0.53
(13.64)
(14.27)

UMD
0.25
(14.42)
0.18
(12.48)
0.12
(8.62)
0.08
(5.34)
0.04
(2.73)
0.21
(8.50)

PanelB:DoubleSortsonStockDurationandMBRatio

StockDuration
Uncond.

1(Short)

5(Long)

51

CAPMAlpha
MBRatio
2
3 (High)
0.26
0.10
(2.19)
(0.79)
0.06
0.41
(0.35)
(1.92)
0.23
0.10
(1.61)
(0.64)
0.28
0.03
(2.31)
(0.26)
0.31
0.18
(2.66)
(2.12)
0.34
0.18
(2.77)
(2.20)
0.28
0.59
(1.66)
(2.91)

1 (Low)
0.41
(2.43)
0.27
(1.19)
0.33
(1.68)
0.43
(2.58)
0.44
(2.60)
0.47
(2.79)
0.20
(1.05)

31
0.51
(2.88)
0.68
(2.95)
0.43
(2.25)
0.47
(2.72)
0.26
(1.66)
0.28
(1.80)
0.39
(2.14)

1 (Low)
0.15
(1.62)
0.05
(0.33)
0.09
(0.70)
0.18
(1.81)
0.17
(1.68)
0.20
(1.97)
0.14
(0.82)

FF3FactorAlpha
MBRatio
2
3(High)
0.12
0.03
(1.55)
(0.48)
0.03
0.27
(0.24)
(2.49)
0.12
0.03
(1.32)
(0.33)
0.13
0.01
(1.57)
(0.07)
0.15
0.17
(1.78)
(2.09)
0.17
0.17
(1.86)
(2.19)
0.20
0.45
(1.51)
(3.47)

31
0.18
(1.99)
0.33
(2.07)
0.12
(0.94)
0.17
(1.65)
0.00
(0.01)
0.03
(0.24)
0.30
(1.83)

1 (Low)
0.35
(4.98)
0.41
(3.39)
0.36
(3.77)
0.35
(4.13)
0.33
(3.82)
0.30
(3.07)
0.11
(0.73)

FF4FactorAlpha
MBRatio
2
3 (High)
0.19
0.04
(2.54)
(0.59)
0.16
0.14
(1.59)
(1.36)
0.21
0.06
(2.49)
(0.66)
0.20
0.07
(2.40)
(0.86)
0.18
0.18
(2.08)
(2.23)
0.19
0.17
(1.98)
(2.06)
0.02
0.31
(0.19)
(2.49)

31
0.31
(3.59)
0.55
(3.82)
0.30
(2.55)
0.28
(2.77)
0.15
(1.69)
0.13
(1.34)
0.42
(2.55)

Table6(cont.)
PanelC:DoubleSortsonStockDurationandMBResidualPV

StockDuration
Uncond.

1(Short)

5(Long)

51

1 (Low)
0.42
(2.40)
0.29
(1.09)
0.32
(1.50)
0.46
(2.69)
0.45
(2.66)
0.47
(2.82)
0.19
(0.83)

CAPMAlpha
MBResidualPV
2
3 (High)
0.25
0.08
(2.11)
(0.63)
0.05
0.40
(0.25)
(1.78)
0.23
0.10
(1.44)
(0.60)
0.26
0.06
(2.03)
(0.43)
0.31
0.16
(2.61)
(1.78)
0.38
0.23
(3.30)
(2.63)
0.32
0.62
(1.60)
(2.91)

31
0.50
(3.08)
0.68
(3.06)
0.42
(2.18)
0.52
(3.36)
0.29
(1.85)
0.25
(1.55)
0.44
(2.25)

1 (Low)
0.18
(1.68)
0.12
(0.62)
0.10
(0.68)
0.24
(2.08)
0.19
(1.73)
0.20
(1.96)
0.08
(0.39)

FF3FactorAlpha
MBResidualPV
2
3(High)
0.14
0.02
(1.71)
(0.22)
0.01
0.26
(0.05)
(2.16)
0.15
0.02
(1.43)
(0.22)
0.13
0.03
(1.38)
(0.32)
0.16
0.16
(1.76)
(2.04)
0.22
0.22
(2.51)
(2.68)
0.21
0.48
(1.32)
(3.38)

31
0.20
(2.02)
0.38
(2.07)
0.12
(0.84)
0.27
(2.39)
0.03
(0.26)
0.02
(0.16)
0.40
(2.12)

1 (Low)
0.41
(5.06)
0.55
(3.67)
0.42
(3.74)
0.46
(4.67)
0.37
(3.83)
0.32
(3.27)
0.24
(1.33)

FF4FactorAlpha
MBResidualPV
2
3 (High)
0.24
0.05
(3.18)
(0.77)
0.28
0.13
(2.25)
(1.14)
0.28
0.06
(2.88)
(0.62)
0.24
0.04
(2.71)
(0.47)
0.20
0.17
(2.21)
(2.08)
0.23
0.19
(2.58)
(2.31)
0.05
0.32
(0.35)
(2.38)

PanelD:SortsonShareTurnover

ShareTurnover
1 (Low)

5 (High)

51

Uncond.
0.18
(1.94)
0.09
(0.68)
0.27
(1.52)

FF3FactorAlpha
MBRatio
1(Low)
2
0.26
0.13
(2.59)
(1.25)
0.00
0.04
(0.02)
(0.28)
0.26
0.08
(1.10)
(0.42)

3(High)
0.01
(0.07)
0.15
(1.20)
0.14
(0.86)

31
0.27
(2.60)
0.16
(0.82)
0.11
(0.56)

31
0.20
(1.82)
0.26
(1.73)
0.05
(0.34)

Uncond.
0.20
(2.22)
0.17
(1.55)
0.04
(0.22)

FF4FactorAlpha
MBRatio
1(Low)
2
0.34
0.12
(3.37)
(1.20)
0.46
0.31
(2.77)
(2.37)
0.12
0.19
(0.60)
(1.06)

Uncond.
0.16
(1.95)
0.14
(1.63)
0.02
(0.18)

FF4FactorAlpha
MBRatio
1 (Low)
2
0.27
0.14
(2.86)
(1.43)
0.43
0.20
(3.95)
(2.06)
0.16
0.07
(1.23)
(0.55)

3 (High)
0.02
(0.21)
0.02
(0.19)
0.00
(0.01)

31
0.29
(2.66)
0.45
(3.31)
0.16
(1.02)

FF4FactorAlpha
MBRatio
1(Low)
2
0.29
0.17
(3.32)
(1.95)
0.38
0.17
(3.38)
(1.66)
0.10
0.00
(0.79)
(0.02)

3(High)
0.10
(1.06)
0.00
(0.04)
0.10
(0.77)

3 1
0.39
(3.44)
0.38
(2.81)
0.01
(0.06)

3(High)
0.01
(0.07)
0.01
(0.05)
0.01
(0.08)

31
0.33
(3.18)
0.47
(2.83)
0.14
(0.71)

PanelE:SortsonInstitutionalTurnover

Institutional Turnover
1 (Low)

5 (High)

51

Uncond.
0.08
(0.95)
0.02
(0.27)
0.05
(0.49)

FF3FactorAlpha
MBRatio
1 (Low)
2
0.13
0.11
(1.27)
(1.10)
0.18
0.13
(1.38)
(1.32)
0.05
0.03
(0.39)
(0.23)

3 (High)
0.07
(0.75)
0.07
(0.68)
0.00
(0.01)

PanelF:SortsonTransientInvestors

Transient Investors
1(Low)
5(High)

5 1

Uncond.
0.08
(1.08)
0.01
(0.08)
0.07
(0.68)

FF3FactorAlpha
MBRatio
1(Low)
2
0.15
0.13
(1.50)
(1.42)
0.14
0.07
(1.06)
(0.62)
0.00
0.06
(0.01)
(0.51)

3(High)
0.14
(1.52)
0.07
(0.63)
0.07
(0.53)

3 1
0.29
(2.50)
0.21
(1.51)
0.07
(0.50)

Uncond.
0.17
(2.31)
0.14
(1.50)
0.03
(0.27)

31
0.35
(4.16)
0.68
(4.26)
0.36
(2.77)
0.42
(3.91)
0.20
(2.11)
0.13
(1.29)
0.56
(3.01)

Table7:ReturnPredictability:PortfolioResultsforStockswithHighversusLowPastReturns
Thistablepresentsmonthlyequalweighted3factorFamaFrenchalphas(PanelAandC)andCarhart4factoralphas(PanelBandD)for
portfoliostrategiesbasedon5x3x2independenttriplesortsbasedonStockDuration,markettobook,andpastreturns.PanelAandB
(Panel C and D) report these portfolios separately for stocks with low and high stock returns over the past 36 (six) months. At the
beginningofeachquarter,stocksarefirstdividedintofivegroupsbasedonStockDuration.Theyarethenindependentlydividedinto
threegroupsbasedontheirmarkettobookratio.Wethenreportreturnsforthese15portfolioswhicharecalculatedovernextfour
quarters.We onlyreport results for thetopand bottomStock Durationgroups. Toaccount foroverlapping portfolios,wefollow the
methodologyinJegadeeshandTitman(1993)suchthatstocksrankedineachofthelastfourquartersformonefourthofeachportfolio.
All the reported returns are in monthly percentages. 5% significance levels are denoted in bold and tstatistics are reported in
parentheses.

PanelA:FF3FactorAlphasforStockswithHighversusLow36MonthsReturns

Past36MonthsStockReturns=Low
MBRatio

StockDuration
1(Short)

5(Long)

51

1(Low)
0.01
(0.07)
0.12
(1.02)
0.13
(0.69)

2
0.03
(0.21)
0.12
(1.03)
0.09
(0.58)

3(High)
0.04
(0.27)
0.16
(1.13)
0.21
(1.07)

31
0.03
(0.16)
0.05
(0.28)
0.08
(0.36)

FF3FactorAlpha

Past36MonthsStockReturns=High
MBRatio

1(Low)
0.21
(1.43)
0.42
(3.77)
0.22
(1.27)

2
0.01
(0.07)
0.26
(2.58)
0.25
(1.87)

3(High)
0.31
(2.61)
0.20
(2.16)
0.51
(3.75)

31
0.52
(3.25)
0.22
(1.89)
0.30
(1.60)

HighLow
Past36
MonthsRet.
0.49
(2.29)
0.27
(1.52)
0.22
(0.88)

PanelB:FF4FactorAlphasforStockswithHighversusLow36MonthsReturns

Past36MonthsStockReturns=Low
MBRatio

StockDuration
1(Short)

5(Long)

51

1(Low)
0.35
(2.50)
0.25
(2.30)
0.11
(0.60)

2
0.23
(1.83)
0.16
(1.42)
0.06
(0.41)

3(High)
0.16
(1.06)
0.16
(1.08)
0.00
(0.01)

31
0.19
(1.05)
0.09
(0.58)
0.10
(0.48)

FF4FactorAlpha

Past36MonthsStockReturns=High
MBRatio

1(Low)
0.41
(3.07)
0.40
(3.53)
0.01
(0.04)

2
0.14
(1.22)
0.22
(2.17)
0.08
(0.64)

3(High)
0.21
(1.77)
0.17
(1.82)
0.38
(2.85)

31
0.62
(3.86)
0.23
(1.94)
0.38
(2.05)

HighLow
Past36
MonthsRet.
0.42
(1.95)
0.14
(0.81)
0.28
(1.10)

PanelC:FF3FactorAlphasforStockswithHighversusLow6MonthsReturns

FF3FactorAlpha

Past6MonthsStockReturns=High
MBRatio

Past6MonthsStockReturns=Low
MBRatio

StockDuration
1(Short)

5(Long)

51

1(Low)
0.07
(0.34)
0.14
(1.20)
0.21
(1.06)

2
0.12
(0.72)
0.12
(1.11)
0.24
(1.43)

3(High)
0.59
(3.56)
0.11
(1.04)
0.70
(4.18)

31
0.52
(3.20)
0.03
(0.24)
0.49
(2.70)

1(Low)
0.19
(1.40)
0.30
(2.97)
0.11
(0.71)

2
0.11
(0.98)
0.24
(2.38)
0.13
(1.12)

3(High)
0.07
(0.69)
0.23
(2.59)
0.30
(2.31)

31
0.26
(1.66)
0.07
(0.65)
0.19
(1.09)

HighLow
Past6
MonthsRet.
0.26
(1.51)
0.04
(0.34)
0.30
(1.57)

PanelD:FF4FactorAlphasforStockswithHighversusLow6MonthsReturns

FF4FactorAlpha

Past6MonthsStockReturns=High

MBRatio

Past6MonthsStockReturns=Low
MBRatio
StockDuration
1(Short)

5(Long)

51

1(Low)
0.40
(2.90)
0.31
(2.99)
0.09
(0.51)

2
0.25
(1.97)
0.24
(2.38)
0.01
(0.05)

3(High)
0.23
(1.89)
0.20
(1.96)
0.43
(2.94)

31
0.63
(3.87)
0.11
(0.91)
0.52
(2.79)

1(Low)
0.28
(2.15)
0.27
(2.68)
0.01
(0.07)

2
0.10
(0.88)
0.17
(1.69)
0.07
(0.59)

3(High)
0.08
(0.75)
0.14
(1.69)
0.23
(1.74)

31
0.36
(2.34)
0.13
(1.19)
0.24
(1.34)

HighLow
Past6
MonthsRet.
0.26
(1.50)
0.02
(0.14)
0.28
(1.45)

Table8:ReturnPredictability:PortfolioResultsforStockswithHighversusLowLimitstoArbitrage
This table presents monthly equalweighted 3factor FamaFrench alphas and Carhart 4factor alphas for portfolio strategies based on 5x3x2 independent triple sorts on Stock
Duration,markettobook,andproxiesforlimitstoarbitrage.Wereporttheseportfoliosseparatelyforstockswithloworhighinstitutionalinvestorholdings(PanelA),loworhigh
Amihudilliquidity(PanelB),andloworhighidiosyncraticvolatility(PanelC).Atthebeginningofeachquarter,stocksarefirstdividedintofivegroupsbasedonStockDuration.They
arethenindependentlydividedintothreegroupsbasedontheirmarkettobookratio.Wethenreportreturnsforthese15portfolioswhicharecalculatedovernextfourquarters.
WeonlyreportresultsforthetopandbottomStockDurationgroups.Toaccountforoverlappingportfolios,wefollowthemethodologyinJegadeeshandTitman(1993)suchthat
stocksrankedineachofthelastfourquartersformonefourthofeachportfolio.Allthereportedreturnsareinmonthlypercentages.5%significancelevelsaredenotedinboldand
tstatisticsarereportedinparentheses.

PanelA:TripleSortsonStockDuration,MarkettoBookRatio,andInstitutionalInvestorHoldings

Stock
Duration
1(Short)

5(Long)

51

InstitutionalInvestorHoldings=Low
MBRatio
1(Low)
2
3(High)
31

0.06
(0.32)
0.24
(1.93)
0.18
(0.84)

0.07
(0.52)
0.21
(2.08)
0.28
(1.81)

0.43
(3.30)
0.16
(1.37)
0.59
(3.86)

0.49
(2.81)
0.08
(0.56)
0.41
(2.05)

FF3FactorAlpha

InstitutionalInvestorHoldings=High

MBRatio
1(Low)
2
3(High)
3 1

0.04
(0.22)
0.12
(1.08)
0.08
(0.48)

0.08
(0.60)
0.12
(1.10)
0.04
(0.25)

0.07
(0.54)
0.24
(2.52)
0.32
(1.98)

0.11
(0.61)
0.12
(1.08)
0.24
(1.26)

High
LowInst.
Inv.
Holdings
0.38
(2.00)
0.20
(1.27)
0.17
(0.78)

Stock
Duration
1(Short)
5(Long)
51

FF4FactorAlpha
InstitutionalInvestorHoldings=Low
InstitutionalInvestorHoldings=High
MBRatio

MBRatio
1(Low)
2
3(High)
3 1
1(Low)
2
3(High)
3 1

0.39
(2.64)
0.30
(2.40)
0.10
(0.49)

0.13
(1.08)
0.18
(1.79)
0.06
(0.40)

0.26
(2.12)
0.15
(1.31)
0.41
(2.84)

0.65
(3.81)
0.14
(1.01)
0.50
(2.50)

0.40
(2.99)
0.27
(2.72)
0.13
(0.82)

0.24
(1.84)
0.17
(1.53)
0.08
(0.52)

0.01
(0.04)
0.22
(2.26)
0.22
(1.35)

0.39
(2.40)
0.05
(0.50)
0.34
(1.83)

High
Low
Inst.Inv.
Holdings
0.25
(1.35)
0.09
(0.58)
0.16
(0.71)

PanelB:TripleSortsonStockDuration,MarkettoBookRatio,andAmihudIlliquidity

Stock
Duration

AmihudIlliquidity=Low
MBRatio
1(Low)
2
3(High)
31

1(Short)

5(Long)

51

0.25
(1.01)
0.22
(1.92)
0.03
(0.14)

0.08
(0.47)
0.11
(0.98)
0.03
(0.17)

0.06
(0.38)
0.24
(2.61)
0.29
(1.61)

0.31
(1.38)
0.02
(0.16)
0.33
(1.47)

FF3FactorAlpha

AmihudIlliquidity=High

MBRatio
1(Low)
2
3(High)
31

0.08
(0.45)
0.18
(1.55)
0.10
(0.54)

0.04
(0.29)
0.24
(2.24)
0.20
(1.16)

0.36
(2.67)
0.33
(2.25)
0.69
(4.16)

0.44
(2.21)
0.15
(0.85)
0.59
(2.71)

High
Low
Amihu.
Illiq.
0.13
(0.50)
0.13
(0.64)
0.27
(0.96)

Stock
Duration
1
5

51

FF4FactorAlpha
AmihudIlliquidity=Low
AmihudIlliquidity=High
MBRatio

MBRatio
1(Low)
2
3(High)
31
1(Low)
2
3(High)
31

0.77
(4.11)
0.35
(3.34)
0.41
(2.03)

0.34
(2.35)
0.17
(1.53)
0.17
(1.08)

0.12
(0.88)
0.21
(2.30)
0.09
(0.51)

0.64
(3.24)
0.14
(1.19)
0.50
(2.30)

0.30
(1.84)
0.27
(2.33)
0.03
(0.17)

0.23
(1.66)
0.22
(2.02)
0.01
(0.09)

0.24
(1.84)
0.29
(1.96)
0.54
(3.32)

0.54
(2.71)
0.03
(0.16)
0.57
(2.56)

High
Low
Amihu.
Illiq.
0.10
(0.40)
0.17
(0.80)
0.07
(0.24)

PanelC:TripleSortsonStockDuration,MarkettoBookRatio,andIdiosyncraticVolatility

Stock
Duration
1(Short)

5(Long)

51

IdiosyncraticVolatility=Low
MBRatio
1(Low)
2
3(High)
31
0.15
(1.59)
0.22
(2.14)
0.07
(0.68)

0.11
(1.14)
0.23
(2.26)
0.12
(1.24)

0.10
(0.99)
0.21
(2.34)
0.11
(1.15)

0.05
(0.43)
0.01
(0.11)
0.03
(0.27)

FF3FactorAlpha

IdiosyncraticVolatility=High

MBRatio
1(Low)
2
3(High)
31

0.02
(0.07)
0.12
(0.95)
0.13
(0.64)

0.05
(0.32)
0.24
(2.43)
0.18
(1.03)

0.50
(3.76)
0.01
(0.07)
0.49
(3.21)

0.48
(2.27)
0.13
(0.73)
0.35
(1.56)

High
LowIdio.
Vola.
0.44
(2.15)
0.12
(0.65)
0.32
(1.44)

Stock
Duration
1
5
51

FF4FactorAlpha
IdiosyncraticVolatility=Low
IdiosyncraticVolatility=High
MBRatio

MBRatio
1(Low)
2
3(High)
31
1(Low)
2
3(High)
31
0.26
(2.93)
0.29
(2.77)
0.03
(0.26)

0.14
(1.43)
0.21
(2.02)
0.07
(0.71)

0.10
(0.93)
0.18
(1.97)
0.08
(0.85)

0.16
(1.62)
0.11
(1.12)
0.05
(0.42)

0.42
(2.48)
0.31
(2.78)
0.12
(0.59)

0.36
(2.55)
0.27
(2.81)
0.08
(0.53)

0.35
(2.78)
0.07
(0.44)
0.42
(2.73)

0.77
(3.96)
0.24
(1.37)
0.53
(2.39)

High
LowIdio.
Vola.
0.61
(3.07)
0.13
(0.72)
0.48
(2.18)

Table9:ReturnPredictability:FamaMacBethRegressions
This table provides quarterly FamaMacBeth predictive regressions linking next twelve month stock returns (columns 1 to 7) or next
twelvemonthDGTWadjustedcumulativeabnormalreturns(columns8to10)withlaggedStockDuration,institutionalturnover,share
turnover, markettobook ratio, pasttwelvemonthreturns(Momentum12 Months), institutional investor holdings, andidiosyncratic
volatility.WefurtherincludeinteractionsofStockDurationwiththemarkettobookratio,anddummyvariablesforhighidiosyncratic
volatility, low institutional investor holdings, and high Amihud Illiquidity. To construct the dummy variables, we divide stocks each
quarter into three groups. Low institutional holdings equals one for stocks in the lowest institutional investor holdings tercile, high
AmihudilliquidityequalsoneforstocksinthehighestAmihudilliquiditymeasuretercile,andhighidiosyncraticvolatilityequalsoneif
stocks belong to the highest idiosyncratic volatility tercile. Variables are defined in Appendix A1. and NeweyWest (1987) adjusted t
statistics, calculated based on 2lags, are reported in parentheses. . ***, **, * indicate significance levels of 1%, 5%, and 10%,
respectively.

Intercept

Log(StockDuration)

Log(InstitutionalTurnover)

Log(MarketCap.)

Log(MBRatio)

Momentum12Months

Log(ShareTurnover)

Log(InstitutionalInvestorHoldings)

Log(IdiosyncraticVolatility)

HighMBRatio*Log(Stock
Duration)

HighMBRatio

LowInstitutionalInv.Holdings*
Log(StockDuration)

LowInstitutionalInv.Holdings

HighIdiosyncraticVolatility*
Log(StockDuration)

HighIdiosyncraticVolatility

HighAmihudIlliquidity*Log(Stock
Duration)

HighAmihudIlliquidity

AverageRsq.(in%)
Obs.
NumberofQuarters

Return12Months
(1)

0.117
(0.79)
0.015**
(2.35)

0.004
(1.06)
0.022**
(2.28)
0.020
(0.97)
0.005
(0.56)
0.014*
(1.69)
0.005
(0.26)

(2)

0.098
(0.66)

0.020
(1.53)
0.002
(0.49)
0.024**
(2.49)
0.018
(0.85)
0.002
(0.18)
0.016*
(1.95)
0.006
(0.34)

(3)

(4)

(5)

(6)

0.123
(0.83)
0.023***
(3.77)
0.033***
(2.58)
0.0029
(0.79)
0.023**
(2.34)
0.019
(0.88)
0.002
(0.18)
0.013
(1.62)
0.004
(0.25)

0.152
(1.01)
0.005
(0.82)

0.183
(1.37)
0.008
(1.22)

0.138
(1.26)
0.004
(0.53)

0.007*
(1.71)

0.003
(0.92)
0.023**
(2.30)
0.021
(1.00)
0.007
(0.91)

0.003
(0.76)
0.023**
(2.33)
0.020
(0.93)
0.003
(0.36)
0.014*
(1.76)

0.017
(0.77)
0.004
(0.43)
0.017**
(2.02)
0.003
(0.14)

0.023**
(2.15)
0.038
(1.61)

7.6
161230
104

7.1
161230
104

7.4
161230
104

7.4
161230
104

7.4
161230
104

7.4
161230
104

(7)

0.128
(0.85)
0.003
(0.40)

0.004
(0.89)
0.021**
(2.20)
0.021
(1.03)
0.005
(0.55)
0.014*
(1.68)
0.006
(0.34)

0.034***
(4.55)
0.081***
(4.24)

0.007
(0.40)

DGTWAdjusted
Return12Months
(8)
(9)
(10)

0.013
0.016
0.042
(0.22)
(0.28)
(0.71)
0.017*** 0.022***
0.009
(2.68)
(3.48)
(1.49)

0.017

(1.52)

0.003
0.002
0.004
(1.10)
(0.89)
(1.62)
0.004
0.005
(0.71)
(0.77)

0.017
0.017
0.015
(1.27)
(1.29)
(1.07)
0.002
0.001
0.002
(0.35)
(0.11)
(0.31)
0.013*
0.013*
0.014**
(1.94)
(1.91)
(2.10)
0.009
0.009
0.011
(0.76)
(0.75)
(0.89)

0.024***
(3.42)
0.044***
(3.25)

0.021**
(2.38)
0.045**
(2.29)

0.037***
(3.92)
0.065***
(4.26)

7.7
126730
104

4.2
161230
100

3.9
161230
100

4.0
161230
100

Figure1:EvolutionofStockDurationandRelatedVariablesoverTime
This figure shows the evolution of Stock Duration, institutional turnover, institutional investor holdings, and share turnover over the
period1985to2010.Thetablereportsmeanvaluesofthesevariablesforeachyear.ThesampleconsistsofUSfirmsfromCompustat.
WeexcludefirmsthatareinthebottomNYSEsizedecileandfirmswithstockpricesbelowUSD1attheendofafiscalyear.Allvariables
arewinsorizedat1%.VariablesaredefinedinAppendixA1.

1.60

350%

1.50

300%
250%

1.40

200%
1.30
150%
1.20

100%

1.10

50%

1.00

0%
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010

StockDuration(inyears)

StockDuration,InstitutionalTurnover,ShareTurnover,andInstitutional
InvestorHoldingsfrom19852010

StockDuration(leftaxis)

InstitutionalInvestorHoldings(rightaxis)

ShareTurnover(annualized)(rightaxis)

InstitutionalTurnover(rightaxis)

Figure2:EvolutionofStockDurationforSelectedCompanies
Figure2AshowstheevolutionofStockDurationforselectedcompaniesovertheperiod1985to2010.Wereportthisvariableforfive
companies:Apple,Microsoft,P&G,Boeing,andHewlettPackard.Figure2BshowstheevolutionofStockDuration,shareturnover,and
institutionalinvestorholdingsforHewlettPackardovertheperiod1985to2010.

Figure2A:StockDurationofSelectedCompanies

StockDurationforSelectedCompanies
3

StockDuration(inyears)

2.5
2
1.5
1
0.5

1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010

Apple

Microsoft

P&G

Boeing

HewlettPackard

Figure2B:StockDuration,ShareTurnover,andInstitutionalInvestorHoldingsforHewlettPackard

300%

StockDuration(inyears)

2.6
2.4

250%

2.2
200%

2
1.8

150%

1.6

100%

1.4
50%

1.2
1
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010

0%

StockDuration(leftaxis)

ShareTurnoverandInstitutionalInvestor
Holdings(in%)

StockDuration,ShareTurnover,andInstitutionalInvestorHoldingsfor
HewlettPackard

InstitutionalInvestorHoldings(rightaxis)

ShareTurnover(annualized)(rightaxis)

Figure3:EvolutionofInstitutionalInvestorHoldingswithDifferentStockDurations
Thisfigureshowstheevolutionofinstitutionalinvestorholdingsovertheperiod1985to2010.Wereporttheholdingsofinstitutional
investorsthathaveaStockDurationoflessthansixmonths(i.e.,shorttermholdings),betweensixand12months,between12and24
months,andabove24months(i.e.,longtermholdings).Theholdingsaddupto100%,representingthetotalholdingsofallinstitutional
investors.The sampleconsistsofUSfirmsfrom Compustat.Weexclude firmsthatareinthe bottomNYSEsize decileand firmswith
stockpricesbelowUSD1attheendofafiscalyear.Allvariablesarewinsorizedat1%.VariablesaredefinedinAppendixA1.

InstitutionalInvestorHoldings:InvestorswithDifferentStockDurations
90%
80%
70%
60%
50%
40%
30%
20%
10%

StockDuration6Months

6Months<StockDuration12Months

12Months<StockDuration24Months

StockDuration>24Months

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

0%
1985

InstitutionalInvestorHoldings(in%)

100%

Figure4:StockDurationoverTimebyInstitutionalInvestorType
Figure4AshowstheevolutionofStockDurationovertheperiod1985to2010byinstitutionalinvestortype.Weclassifyinstitutional
investor into four categories. The category Bank includes banks and insurance firms. Investment Company includes independent
investmentadvisorsandinvestmentcompanies.PensionFundsincludescorporate(private)pensionfunds,publicpensionfunds,and
university and foundation endowments. Others include all other institutional investors. The table reports median values of these
variablesforeachsampleyear.Figure4BshowstheevolutionoftheaverageStockDurationforselectedinstitutionalinvestorsoverthe
period1985to2010.WecalculatethisvariableastheweightedStockDurationofallholdingsheldbyaninstitutionalinvestor,usingas
weights the current portfolio holdings of an institutional investor. We report this variable for four institutional investors: Berkshire
Hathaway (investment company), Fidelity (mutual fund), STRS Ohio (public pension fund), and the University of Chicago (university
endowment).VariablesaredefinedinAppendixA1.

Figure4A:StockDurationbyInstitutionalInvestorType

2.50
2.00
1.50
1.00
0.50
0.00
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010

StockDuration(medians,inyears)

StockDurationbyInstitutionalInvestorTypefrom19852010

BankStockDuration

PensionFunds

InvestmentCompany

Others

Figure4B:StockDurationofSelectedInstitutionalInvestors

5
4.5
4
3.5
3
2.5
2
1.5
1
0.5
0
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010

InstitutionalDuration(inyears)

StockDurationforSelectedInvestors
(WeightedAverageStockDurationAcrossallStocksofanInvestor)

BerkshireHathaway

Fidelity

STRSOhio

UChicago

Figure5:MisvaluationandStockDurationinEventTime:LargeChangesinStockDuration
Thesefiguresshowhowmarkettobookratioschangearoundlargedecreases(Figure5A)andincreases(Figure5B)inStockDuration.
Wedefinealargedecrease(increase)inStockDurationasachangeinStockDurationthatisinthebottom(top)decile.Wedefineast=0
theyearinwhichthelargechangeinStockDurationoccurs,andreportvariablesinthefiveyearsbeforeandaftert=0.Forallvariables,
we report mean values. We also report standard error bounds around the means of the markettobook ratio (+/ 2 standard errors
aroundthemean).Allvariablesarewinsorizedat1%.VariablesaredefinedinAppendixA1.

Figure5A:LargeDecreasesinStockDuration

3.20

1.80

3.00

1.70
1.60

2.80

StockDuration

MBRatio

MBRatioandStockDurationifStockDurationGoesDown
(changeinbottom10%)

1.50

2.60

1.40

2.40

1.30

2.20

1.20
1.10

2.00
t5

t4

t3

t2

t1

t0

t+1 t+2 t+3

t+4

t+5

EventTimeinYears
(t=0isdefinedasyearinwhichChangeinStockDurationfromt1tot0isin
Bottom10%)
MBRatio(leftaxis)

+2SE

2SE

StockDuration(inyears)(rightaxis)

Figure5B:LargeIncreasesinStockDuration

1.70

3.40
3.30
3.20
3.10
3.00
2.90
2.80
2.70
2.60
2.50
2.40

1.60
1.50
1.40
1.30
1.20
1.10

StockDuration

MBRatio

MBRatioandStockDurationifStockDurationGoesUp
(changeintop10%)

1.00
0.90
0.80
t5

t4

t3

t2

t1

t0

t+1 t+2 t+3

t+4

t+5

EventTimeinYears
(t=0isdefinedasyearinwhichChangeinStockDurationfromt1tot0isinTop
10%)
MBRatio(leftaxis)

+2SE

2SE

StockDuration(inyears)(rightaxis)

Figure6:CumulativePortfolioReturnsinCalendarTime
This figure shows cumulative stock returns over the period 1984 to 2011 for different Stock Duration portfolios that invest USD 1 in
1984. We calculate these cumulative returns based on monthly stock returns, constructed using different portfolios. The first two
portfolio(ShortStockDurationandLongSockDuration)consistofstockswithshort(bottomquintile)andlong(topquintile)Stock
Duration.Thenextportfolio(Long/ShortStockDuration)consistsoflongpositionsinstockswithlongStockDuration(bottomquintile)
andshortpositions(topquintile)instockwithshortStockDuration.Thenexttwoportfoliosconsistagainoflongpositionsinstockswith
long Stock Duration and short positions in stock with short Stock Duration, but now for stocks that have low (Long/Short Stock
Duration(MBRatio=Low)orhigh(Long/ShortStockDuration(MBRatio=High)markettobookratios.Stockshavelow(high)market
tobookgroupiftheyareinthebottom(top)tercile.Tocalculatethemonthlystockreturns,atthebeginningofeachquarter,stocksare
firstdividedintoquintilesbasedonStockDuration.Forthereturnswherewealsoconditiononmarkettobookratios,theyarethen
independently divided into three groups based on the markettobook ratio. The cumulative returns are calculated based on these
monthlystockreturns.VariablesaredefinedinAppendixA1.

CumulativePortfolioReturnsinCalendarTime
4
3.5

DollarValue

3
2.5
2
1.5
1
0.5
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011

ShortStockDuration

LongStockDuration

Long/ShortStockDuration

Long/ShortStockDuration(MBRatio=Low)

Long/ShortStockDuration(MBRatio=High)

AppendixA1:DefinitionsofVariables
Thistableprovidesdefinitionsofthevariablesusedintheempiricalanalysis.VariablesaredefinedinAppendixA1.
Variable
StockDuration

Definition
This variable is defined as the weighted average duration a stock has been in the portfolios of institutional
investors.Thisvariableiscalculatedastheholdingdurationofownershipofeachstockforeveryinstitutional
investor by calculating a weightedmeasure of buys and sells by an institutional investor, weighted by the
duration for which the stock was held. For each stock in a given fund managers portfolio, the holding
durationmeasureisthuscalculatedbylookingbackoverthefullthetimeperiodsincethatparticularstock
hasbeenheldcontinuouslyinthatfundsportfolio.Thecalculationofthedurationforstockithatisincluded
intheinstitutionalportfoliojattimeT1,forallstocksi=1Iandallinstitutionalinvestorsj=1J,isgiven
by:

Durationi , j ,T 1 d i , j ,T 1

InstitutionalTurnover

TransientInvestors(in%)

ShareTurnover
InstitutionalInvestor
Holdings
ShortTermOwnership

MBRatio
MBResidualPV

MBResidualHP

MarketCap(inmUSD)
AssetHorizon

R&D/Assets
R&DMissing
Capex/Assets
PPE/Assets
Debt/Assets
Cash/Assets
SalesGrowth
Assets
MarketCapitalization
AmihudIlliquidity

(T t 1) i , j ,t

H B
t T W
i, j
i, j
T 1

(W 1) H i , j

H B
i
,
j
i
,
j

whereBi,j=totalpercentageofsharesofstockiboughtbyinstitutionjbetweent=TWandt=T1;t,Tarein
quarters;Hi,j=percentageoftotalsharesoutstandingofstockiheldbyinstitutionjattimet=TW;i,j,t=
percentageoftotalsharesoutstandingofstockiboughtorsoldbyinstitutionjbetweentimet1andt,where
i,j,t > 0 for buys and <0 for sells. We choose W = 20 quarters, as very few stock positions are held
continuously for longer than 5 years. If stock i is not included in institutional portfolio j at time T1, then
Durationi,j,T1=0.Next,wecomputeattheindividualstocklevelourStockDurationproxybyaveragingthe
institutionalstocklevelDurationi,j,T1overallinstitutionscurrentlyholdingthestock,usingasweightsthe
totalcurrentholdingsinthestockofeachinstitution.
Thevariableisdefinedastheweightedaverageoftheturnoverofinstitutionalinvestorsholdingagivenstock.
InstitutionalTurnoveriscalculatedusingchangesinthequarterlyholdingsoverthepast4quartersandthe
stocklevelweightsarecalculatedusingthecurrentholdingsinthestockineachinstitutionalportfolio.
This variable measures the percentage ownership of transient institutional investors. The measure was
introducedbyBushee(1998,2001),whosemethodologyisbasedonfactorandclusteringanalysistoclassify
institutional investors into three groups: transient investors with high portfolio turnover and diversified
portfolios,dedicatedinstitutionswithlowturnoverandmoreconcentratedportfolioholdings,andquasi
indexerinstitutionswithlowturnoveranddiversifiedportfolioholdings.Weobtaintheinstitutionalinvestor
classification data from Brian Bushees website and calculate the percentage of a firms ownership by
transientinstitutionalinvestors.
This variable is defined as the number of a firms shares that are traded divided by the number of shares
outstanding.
Thisvariableisdefinedasthepercentageownershipofinstitutionalinvestors.
This variable measures the percentage that is owned by institutional investors (relative to all holdings by
institutionalinvestors)withaStockDurationthatisinthebottom50%(i.e.,thathaveheldthestockforthe
shorttermonly).
Thisvariableisdefinedasthemarketvalueofequityoverthebookvalueofequity.
This variable is based on Pastor and Veronesi (2003) to proxy for misvaluation. It is defined as the residual
from yearly regression of log(MB Ratio) on log(total assets), leverage (longterm debt/total assets), ROE,
1/(1+firmage),adividendpayerdummy,andstockreturnvolatility.
ThisvariableisbasedonHobergandPhillips(202010)toproxyformisvaluation.Itisdefinedastheresidual
fromregressingforeachyearandindustrylog(MBRatio)onlog(assets),leverage,ROE,1/(1+firmage),stock
returnvolatility,adividendpayerdummy.Industriesaredefinedby2digitsiccodes.
Thisvariableisdefinedasthemarketcapitalizationoftheequityofafirm.
This variableisbased on Stohsand Mauer(1996) anddefinedasthe(book)valueweighted averageofthe
maturitiesofcurrentassetsandnetproperty,plant,andequipment.Thematurityofcurrentassetsisdefined
ascurrentassetsdividedbythecostofgoodssold.Thematurityofnetproperty,plant,andequipmentisthat
amountdividedbyannualdepreciationexpense.
ThisvariableisdefinedasR&Dexpendituresovertotalassets.
ThisvariabletakesthevalueoneifdataonR&Dismissing,andzerootherwise.
Thisvariableisdefinedascapitalexpendituresovertotalassets.
Thisvariableisdefinedasnetproperty,plant,andequipment(PPE)overtotalassets.
Thisvariableisdefinedasdebtovertotalassets.
Thisvariableisdefinedascashandcashequivalentsovertotalassets.
Thisvariableisdefinedastheyearbyyearchangeinsales.
Thisvariableisdefinedasthetotalassetsofafirm.
Thisvariableisdefinedasthemarketvalueoftheequityofafirm.
This variableisbased on Amihud (2002)and isdefined as the annualaverageof the ratioof dailyabsolute
stockreturnstodailydollartradingvolume.

AppendixA1(continued)
Variable
IdiosyncraticVolatility
BankStockDuration
InvestmentCompanyStock
Duration
PensionFundsStock
Duration
OthersStockDuration
BankHoldings
InvestmentCompany
Holdings
PensionFundsHoldings
OthersHoldings
Return12Months
DGTWAdjustedReturns12
Months
Momentum12Months

Definition
This variable is defined as the residual that is obtained from a 3factor Fama and French model of stock
returns.Itisestimatedusingdailyreturnsoverthequarterbeforethefiscalyearend.
This variable measures the stock duration of banks. The category bank includes banks and insurance
firms.
This variable measuresthestockdurationof investment companies.The category investment company
includesindependentinvestmentadvisorsandinvestmentcompanies(e.g.,mutualfunds).
This variable measures the stock duration of pension funds. The category pension funds includes
corporate(private)pensionfunds,publicpensionfunds,anduniversityandfoundationendowments.
This variable measuresthestockduration of others. The category others includes all otherinstitutional
investorsthatarenotcapturedinthecategoriesBank,InvestmentCompany,andPensionFunds.
This variable is defined as the percentage ownership of banks. The category bank includes banks and
insurancefirms.
Thisvariableisdefinedasthepercentageownershipofinvestmentcompanies.Thecategoryinvestment
companyincludesindependentinvestmentadvisorsandinvestmentcompanies(e.g.,mutualfunds).
This variable is defined as the percentage ownership of pension funds. The category pension funds
includescorporate(private)pensionfunds,publicpensionfunds,anduniversityandfoundationendowments.
This variable is defined as the percentage ownership of others. The category others includes all other
institutionalinvestorsthatarenotcapturedinthecategoriesBank,InvestmentCompany,andPensionFunds.
Thisvariableisdefinedasthetwelvemonthsrawstockreturn.
ThisvariableisdefinedasthetwelvemonthsDGTWadjustedcumulativeabnormalstockreturn.
Thisvariableisdefinedasthepasttwelvemonthsrawstockreturn.

AppendixA2:CorrelationsofVariables
Thistablereportspairwisecorrelationsofsomeofthekeyvariablesofthepaper.*indicatessignificanceat1%.

StockDuration
InstitutionalTurnover
TransientInvestors
ShareTurnover
InstitutionalInvestorHoldings
MBRatio
MBResidualPV
MBResidualHP
MarketCap(inmUSD)
AssetHorizon
PPE/Assets
Amihudilliquidity
IdiosyncraticVolatility

[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]

[1]
1
0.5734*
0.2863*
0.4118*
0.0512*
0.0922*
0.0507*
0.0304*
0.1858*
0.1871*
0.2151*
0.0280*
0.3609*

[2]

[3]

[4]

[5]

[6]

[7]

[8]

[9]

[10]

[11]

[12]

[13]

1
0.3321*
0.3378*
0.0212*
0.1389*
0.0932*
0.0796*
0.1103*
0.1254*
0.1486*
0.0371*
0.2615*

1
0.3951*
0.5792*
0.1021*
0.1785*
0.1438*
0.0022
0.1503*
0.1484*
0.2173*
0.0786*

1
0.3215*
0.1737*
0.1614*
0.1194*
0.0123
0.1910*
0.2504*
0.1840*
0.3604*

1
0.0066
0.0699*
0.0420*
0.0956*
0.1571*
0.1055*
0.3569*
0.1583*

1
0.8490*
0.6958*
0.2000*
0.1609*
0.1996*
0.0329*
0.1329*

1
0.8211*
0.2441*
0.1471*
0.1624*
0.0810*
0.0387*

1
0.2185*
0.0786*
0.0792*
0.0558*
0.0278*

1
0.0168*
0.0197*
0.0857*
0.1320*

1
0.7830*
0.0002
0.2312*

1
0.0149
0.2323*

1
0.2099*

AppendixA3:StockDurationbyInstitutionalInvestorTypeoverTime
ThistableshowstheevolutionofStockDurationovertheperiod1985to2010byinstitutionalinvestortype.Weclassifyinstitutional
investor into four categories. The category Bank includes banks and insurance firms. Investment Company includes independent
investmentadvisorsandinvestmentcompanies(e.g.,mutualfunds).PensionFundsincludescorporate(private)pensionfunds,public
pension funds, anduniversityand foundation endowments. Others include all otherinstitutionalinvestors. The tablereports mean
valuesofthesevariableforeachsampleyear.ThesampleconsistsofUSfirmsfromCompustat.Weexcludefirmsthatareinthebottom
NYSEsizedecileandwithstockpricesbelowUSD1attheendofafiscalyear.Allvariablesarewinsorizedat1%.Variablesaredefinedin
AppendixA1.

Year

1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Total
Change(19852010)

Median
1.23
1.23
1.36
1.57
1.71
1.67
1.65
1.64
1.53
1.42
1.41
1.19
1.30
1.30
1.39
1.37
1.37
1.41
1.49
1.64
1.73
1.76
1.78
1.80
1.60
1.62
1.50
32%

BankStock
Duration

Obs.
865
1103
1196
1211
1267
1349
1402
1414
1450
1567
1664
1714
1625
1661
1711
1615
1494
1369
1315
1215
1135
1051
953
1068
1124
423
33961

InvestmentCompany
StockDuration

PensionFundsStock
Duration

Median
1.11
1.05
1.10
1.20
1.19
1.26
1.25
1.25
1.27
1.30
1.29
1.15
1.13
1.16
1.15
1.15
1.20
1.24
1.27
1.33
1.36
1.36
1.39
1.43
1.34
1.41
1.23
27%

Median
0.85
1.06
0.75
0.92
1.34
1.68
1.90
2.09
2.15
1.97
1.80
1.90
1.63
1.68
1.50
1.48
1.50
1.58
1.74
1.97
2.17
2.28
2.31
2.10
1.96
2.00
1.72
134%

Obs.
863
1102
1196
1210
1267
1352
1403
1413
1450
1567
1664
1714
1625
1659
1707
1615
1494
1369
1315
1215
1135
1051
953
1068
1125
424
33956

Obs.
779
921
1080
1199
1256
1334
1385
1400
1431
1553
1658
1620
1529
1534
1499
1435
1414
1352
1312
1214
1134
1049
951
1067
1123
422
32651

Others
StockDuration
Median
1.75
1.60
1.53
1.98
1.52
2.18
1.52
1.44
1.81
1.63
1.71
2.10
1.70
1.69
1.15
0.82
0.67
0.82
0.86
0.65
0.79
0.93
1.01
1.18
1.22
1.23
1.00
30%

Obs.
197
211
246
117
150
109
187
273
262
303
298
286
322
291
295
290
577
1043
1069
1160
1130
1049
950
1066
1122
422
13425

AppendixA4:InstitutionalInvestorHoldingsbyInvestorTypeoverTime
This figure shows the evolution of the holdings by different institutional investor types. We classify institutional investor into four
categories.ThecategoryBankincludesbanksandinsurancefirms.InvestmentCompanyincludesindependentinvestmentadvisors
andinvestmentcompanies(e.g.,mutualfunds).PensionFundsincludescorporate(private)pensionfunds,publicpensionfunds,and
universityandfoundationendowments.Othersincludeallotherinstitutionalinvestors.Thesumofallholdingsaddsupto100%,i.e.,
tothetotalholdingsofallinstitutionalinvestorstogether.

60%
50%
40%
30%
20%
10%

BankHoldings

InvestmentCompanyHoldings

PensionFundsHoldings

OthersHoldings

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

0%
1985

InstitutionalInvestorHoldingsbyType(in%)

InstitutionalInvestorHoldingsbyInvestorType19852010

AppendixA5:InstitutionalInvestorHoldingsbyInvestorTypeifStockDurationislessthan6Months
This figure shows the evolution of the holdings of different investor types. Hereby, we only report those holdings of institutional
investorswithastockdurationoflessthansixmonths(i.e.,shorttermholdings).Weclassifyinstitutionalinvestorintofourcategories.
The category Bank includes banks and insurance firms. Investment Company includes independent investment advisors and
investment companies (e.g., mutual funds, hedge funds). Pension Funds includes corporate (private) pension funds, public pension
funds,anduniversityandfoundationendowments.Othersincludeallotherinstitutionalinvestors.

90%
80%
70%
60%
50%
40%
30%
20%
10%

BankHoldings

InvestmentCompanyHoldings

PensionFundsHoldings

OthersHoldings

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

0%
1985

InstitutionalInvestorHoldingsbyTypein%

InstitutionalInvestorHoldingsbyInvestorType:
OnlyifStockDurationis6months

AppendixA6:MisvaluationandStockDuration:RobustnessChecks
Thistableprovidesincolumns1to2panelregressionslinkingamisvaluationproxyandStockDuration.Asvaluationproxyweusethe
misvaluation measure proposed by Hoberg and Phillips (2010), MB Residual HP. Columns 3 to 4 provide panel regressions linking
valuation proxies and the percentage that is owned for the shortterm. This variable measures the percentage that is owned by
institutionalinvestors(relativetoallholdingsbyinstitutionalinvestors)withaStockDurationthatisinthebottom50%(i.e.,thathave
held the stock for the shortterm only). As valuation proxies we use in columns 3 and 4 the markettobook ratio (MB ratio) and in
columns5and6themisvaluationmeasureproposedbyPastorandVeronesi(2003),MBResidualPV.ThesampleconsistsofUSfirms
fromCompustat.WeexcludefirmsthatareinthebottomNYSEsizedecileandwithstockpricesbelowUSD1attheendofafiscalyear.
All variables are winsorized at 1%. Variables are defined in Appendix A1. tstatistics, calculated based on robust standard errors
clusteredatthefirmlevel,arereportedinparentheses.***,**,*indicatesignificancelevelsof1%,5%,10%,respectively.

MBResidualHP
(1)
(2)
0.076*** 0.070***
(6.48)
(5.54)
0.074***
0.074***
(7.18)
(7.31)
0.016
(1.45)

StockDuration
StockDuration(t1)
StockDuration(t2)

ShortTermOwnership

MBRatio
(3)

(4)

0.412***
(4.30)
0.166**
(2.05)

0.003***
(5.58)
0.088***
(6.89)
0.002
(1.60)
0.477***
(6.68)
0.197***
(12.97)
0.175***
(3.09)
0.184***
(9.77)
0.648***
(10.87)
1.105***
(9.01)
0.848***
(4.18)
0.029
(0.80)
0.062***
(2.72)
7.500***
(13.45)

0.168***
(6.38)
0.046**
(2.14)
0.008
(0.36)
0.003***
(5.29)
0.096***
(6.80)
0.003*
(1.79)
0.505***
(6.56)
0.212***
(11.56)
0.158**
(2.51)
0.181***
(8.90)
0.621***
(9.65)
1.049***
(7.97)
0.862***
(3.84)
0.037
(0.97)
0.067**
(2.46)
7.459***
(12.29)

0.002***
(5.07)
0.055***
(4.31)
0.004**
(2.50)
0.332***
(5.24)
0.158***
(11.45)
0.197***
(3.70)
0.188***
(11.58)
0.238***
(4.19)
0.845***
(6.71)
0.753***
(3.83)
0.034
(1.10)
0.022
(1.15)
5.110***
(10.16)

0.002***
(4.33)
0.067***
(4.61)
0.005***
(2.78)
0.351***
(5.15)
0.168***
(10.04)
0.202***
(3.52)
0.190***
(10.97)
0.212***
(3.46)
0.814***
(6.02)
0.775***
(3.47)
0.046
(1.36)
0.038*
(1.66)
4.876***
(8.99)

0.011***
(5.46)
0.297***
(5.43)
0.013
(1.60)
1.687***
(5.67)
0.640***
(9.84)
0.406
(1.58)
0.840***
(10.40)
2.586***
(9.59)
3.199***
(6.45)
6.312***
(5.31)
0.174
(1.43)
0.210***
(2.84)
15.269***
(7.23)

Yes
Yes

Yes
Yes

Yes
Yes

Yes
Yes

Yes
Yes

Yes
Yes

25554
0.087

22362
0.086

25219
0.135

21906
0.132

25219
0.140

21906
0.137

ShortTermOwnership(t2)

ShareTurnover
AssetHorizon
Cash/Assets
SalesGrowth
PPE/Assets
Log(Assets)
Debt/Assets
Capex/Assets
R&D/Assets
R&DMissing
AmihudIlliquidity
IdiosyncraticVolatility
FirmFixedEffects
YearFixedEffects
Obs.
adj.Rsq.

MBResidualPV
(5)
(6)

0.342***
(3.22)
0.266***
(3.06)
0.015
(0.17)
0.011***
(5.14)
0.344***
(5.77)
0.015*
(1.71)
1.761***
(5.43)
0.666***
(8.69)
0.407
(1.45)
0.847***
(9.69)
2.494***
(8.53)
3.094***
(5.82)
6.228***
(4.62)
0.184
(1.36)
0.268***
(2.99)
13.873***
(6.14)

ShortTermOwnership(t1)

InstitutionalInvestorHoldings

0.177***
(7.54)
0.025
(1.19)

AppendixA7:MisvaluationandStockDurationforSubsamples:RobustnessChecks
ThistableprovidespanelregressionslinkingamisvaluationproxyandStockDuration.Weusethemisvaluationmeasureproposedby
HobergandPhillips(2010)(MBResidualHP).Weprovidetheregressionsbasedonsubsamplesusing:(i)institutionalinvestorholdings;
(ii) the Amihud illiquidity measure; and (iii) idiosyncratic volatility; We create these subsamples by separating the sample firms each
year based on median values of the respective variables. We also report pvalues of Chowtests comparing the coefficients of Stock
DurationandlaggedStockDurationacrosssubsamples.ThesampleconsistsofUSfirmsfromCompustat.Weexcludefirmsthatarein
thebottomNYSEsizedecileandwithstockpricesbelowUSD1attheendofafiscalyear.Allvariablesarewinsorizedat1%.Variables
are defined in Appendix A1. tstatistics, calculated based on robust standard errors clustered at the firm level, are reported in
parentheses.***,**,*indicatesignificancelevelsof1%,5%,10%,respectively.

Inst.InvestorHoldings
Low
High
(1)
(2)

StockDuration
StockDuration(t1)
ShareTurnover(coeff./100)
AmihudIlliquidity
IdiosyncraticVolatility

0.087***
(6.11)
0.094***
(7.08)
0.062***
(2.93)
0.025
(1.17)
3.408***
(4.80)

pvalueChowtest(StockDuration)
pvalueChowtest(StockDuration(t1))
ControlsasinTable3
FirmFixedEffects
YearFixedEffects
Obs.
adj.Rsq.

0.039**
(1.98)
0.032*
(1.86)
0.053***
(3.16)
0.112**
(2.38)
6.227***
(9.01)

0.008
(0.39)
0.044**
(2.37)
0.000
(0.01)
2.736***
(2.79)
5.890***
(7.77)

0.7730
0.0000

0.087***
(6.46)
0.112***
(9.46)
0.102***
(4.82)
0.060***
(3.15)
3.111***
(5.05)

Yes
Yes
Yes

Yes
Yes
Yes

11750
0.085

13804
0.097

IdiosyncraticVolatility
Low
High
(5)
(6)
0.020
(1.43)
0.033**
(2.57)
0.033
(1.28)
0.184***
(4.36)
2.465**
(2.43)

0.0000
0.0850

MBResidualHP
AmihudIlliquidity
Low
High
(3)
(4)

0.134***
(7.05)
0.100***
(6.16)
0.063***
(4.41)
0.015
(0.65)
5.451***
(7.36)

0.0000
0.0020

Yes
Yes
Yes

Yes
Yes
Yes

14165
0.115

11389
0.113

Yes
Yes
Yes

Yes
Yes
Yes

15543
0.076

10011
0.117

AppendixA8:ReturnPredictability:RobustnessChecks
ThistablepresentsinPanelAmonthlyequalweightedCAPMalphas,3factorFamaFrenchalphasandCarhart4factoralphasforportfoliostrategiesfromconditionalsortsbased
onStockDurationandthemisvaluationmeasuresfromHobergandPhillips(2010).Atthebeginningofeachquarter,stocksarefirstdividedintofivegroupsbasedontheStock
Duration. Theyare then independentlydividedintothree groupsbased on the misvaluation measure andreturnsfor these 15 portfoliosare calculatedovernext fourquarters.
Panel B to D present monthly equalweighted Carhart 4factor alphas for portfolio strategies based on 5x3x2 independent triple sorts on Stock Duration, markettobook, and
proxiesforlimitstoarbitrage.Wereporttheseportfoliosseparatelyforstockswithloworhighinstitutionalinvestorholdings(PanelB),loworhighAmihudilliquidity(PanelC),and
loworhighidiosyncraticvolatility(PanelD).Atthebeginningofeachquarter,stocksarefirstdividedintofivegroupsbasedontheStockDurationandtheyarethenindependently
divided into three groups based on the PastorVeronesi or HobergPhillips misvaluation measure. We then calculate returns for these 15 portfolios over next four quarters. We
reportresultsforthetopandbottomStockDurationgroups.Toaccountforoverlappingportfolios,wefollowinallpanelsthemethodologyinJegadeeshandTitman(1993)such
thatthestocksrankedineachofthelastfourquartersformonefourthoftheportfolio.Allthereturnsareinmonthlypercentage.5%significancelevelisdenotedinboldandt
statisticsaregiveninparentheses.

PanelA:DoubleSortsonStockDurationandMBResidualHP

StockDuration
Uncond.

1(Short)

5(Long)

5 1

1(Low)
0.36
(2.25)
0.19
(0.75)
0.29
(1.48)
0.40
(2.55)
0.40
(2.58)
0.44
(2.93)
0.25
(1.13)

CAPMAlpha
MBResidualHP
2
3(High)
0.25
0.03
(2.13)
(0.23)
0.02
0.32
(0.11)
(1.42)
0.15
0.03
(0.97)
(0.16)
0.23
0.06
(1.79)
(0.48)
0.38
0.15
(3.28)
(1.55)
0.42
0.23
(3.61)
(2.60)
0.44
0.54
(2.25)
(2.57)

3 1
0.39
(3.13)
0.51
(2.63)
0.32
(2.05)
0.34
(2.93)
0.25
(2.07)
0.22
(1.73)
0.29
(1.67)

FF3FactorAlpha
MBResidualHP
2
3(High)
0.12
0.01
(1.57)
(0.08)
0.08
0.19
(0.58)
(1.50)
0.04
0.02
(0.41)
(0.19)
0.09
0.05
(0.98)
(0.57)
0.22
0.11
(2.58)
(1.33)
0.25
0.18
(2.90)
(2.38)
0.33
0.37
(2.19)
(2.62)

1(Low)
0.17
(1.73)
0.07
(0.37)
0.13
(1.02)
0.22
(2.06)
0.18
(1.70)
0.21
(2.17)
0.14
(0.74)

3 1
0.16
(2.08)
0.26
(1.59)
0.11
(0.87)
0.17
(1.86)
0.06
(0.70)
0.03
(0.29)
0.23
(1.35)

1(Low)
0.374
(4.92)
0.46
(3.19)
0.40
(3.91)
0.41
(4.55)
0.33
(3.52)
0.31
(3.24)
0.16
(0.91)

FF4FactorAlpha
MBResidualHP
2
3(High)
0.224
0.10
(3.09)
(1.40)
0.15
0.03
(1.31)
(0.24)
0.19
0.13
(1.92)
(1.30)
0.21
0.13
(2.52)
(1.50)
0.27
0.15
(3.18)
(1.72)
0.27
0.18
(3.07)
(2.24)
0.12
0.20
(0.85)
(1.52)

31
0.27
(3.81)
0.49
(3.32)
0.27
(2.20)
0.28
(3.28)
0.18
(2.18)
0.13
(1.46)
0.36
(2.13)

PanelB:TripleSortsonStockDuration,Misvaluation,andInstitutionalInvestorHoldings

Stock
Duration
1(Short)

5(Long)

51

FF4FactorAlpha
InstitutionalInvestorHoldings=Low InstitutionalInvestorHoldings=High
MBResidualPV

MBResidualPV
1(Low)
2
3(High)
31
1(Low)
2
3(High)
3 1

0.58
(2.93)
0.29
(2.50)
0.29
(1.25)

0.35
(2.36)
0.25
(2.66)
0.11
(0.64)

Stock
Duration

0.25
(1.86)
0.18
(1.51)
0.42
(2.68)

High
Low
Inst.Inv.
Holdings
0.43
(1.86)
0.09
(0.51)
0.34
(1.23)

1(Short)

5(Long)

51

0.83
(3.98)
0.11
(0.79)
0.71
(2.96)

0.39
(2.61)
0.31
(2.53)
0.09
(0.51)

0.28
(1.87)
0.20
(1.69)
0.08
(0.49)

0.00
(0.00)
0.28
(2.78)
0.28
(1.64)

0.39
(2.20)
0.03
(0.21)
0.37
(1.72)

FF4FactorAlpha
InstitutionalInvestorHoldings=Low
InstitutionalInvestorHoldings=High
MBResidualHP

MBResidualHP
1(Low)
2
3(High)
3 1
1(Low)
2
3(High)
3 1

0.51
(2.72)
0.30
(2.63)
0.21
(0.96)

0.14
(0.94)
0.27
(2.90)
0.13
(0.73)

0.10
(0.74)
0.17
(1.49)
0.27
(1.66)

0.61
(3.18)
0.13
(0.97)
0.48
(2.11)

High
Low
Inst.Inv.
Holdings
0.37
(1.63)
0.07
(0.39)
0.30
(1.09)

0.32
(2.13)
0.28
(2.33)
0.04
(0.21)

0.20
(1.47)
0.28
(2.58)
0.08
(0.51)

0.08
(0.49)
0.22
(2.07)
0.14
(0.82)

0.24
(1.41)
0.06
(0.57)
0.18
(0.89)

AppendixA8(cont.)

PanelC:TripleSortsonStockDuration,Misvaluation,andAmihudIlliquidity

Stock
Duration
1

51

FF4FactorAlpha
AmihudIlliquidity=Low

AmihudIlliquidity=High
MBResidualPV

MBResidualPV
1(Low)
2
3(High)
31
1(Low)
2
3(High)
31

0.76
(4.07)
0.37
(3.36)
0.39
(1.87)

0.42
(2.83)
0.13
(1.23)
0.29
(1.78)

0.09
(0.62)
0.20
(2.17)
0.11
(0.62)

0.67
(3.37)
0.17
(1.42)
0.50
(2.27)

0.33
(2.07)
0.27
(2.29)
0.05
(0.28)

0.15
(1.02)
0.27
(2.49)
0.12
(0.69)

0.24
(1.76)
0.26
(1.81)
0.50
(2.98)

0.56
(2.99)
0.01
(0.08)
0.55
(2.38)

High
Low
Amihu.
Illiq.
0.11
(0.46)
0.16
(0.76)
0.05
(0.17)

Stock
Duration
1

51

FF4FactorAlpha
AmihudIlliquidity=Low

AmihudIlliquidity=High
MBResidualHP
MBResidualHP
1(Low)
2
3(High)
31
1(Low)
2
3(High)
31

0.75
(4.10)
0.26
(2.29)
0.49
(2.36)

0.35
(2.59)
0.22
(2.33)
0.13
(0.83)

0.16
(1.08)
0.18
(2.04)
0.02
(0.14)

0.60
(3.14)
0.08
(0.74)
0.51
(2.47)

0.17
(1.21)
0.30
(2.72)
0.13
(0.78)

0.02
(0.12)
0.27
(2.49)
0.29
(1.78)

0.11
(0.80)
0.21
(1.61)
0.32
(1.91)

0.28
(1.74)
0.09
(0.59)
0.19
(0.96)

High
Low
Amihu.
Illiq.
0.32
(1.41)
0.01
(0.05)
0.33
(1.29)

PanelD:TripleSortsonStockDuration,Misvaluation,andIdiosyncraticVolatility

Stock
Duration
1

51

FF4FactorAlpha
IdiosyncraticVolatility=Low

IdiosyncraticVolatility=High
MBResidualPV

MBResidualPV
1(Low)
2
3(High)
31
1(Low)
2
3(High)
31

0.22
(2.09)
0.35
(3.36)
0.13
(1.05)

0.25
(2.29)
0.24
(2.40)
0.01
(0.12)

0.09
(0.77)
0.17
(1.85)
0.09
(0.79)

High
Low
Idio.
Vola.
0.91
(4.15)
0.11
(0.60)
0.80
(3.06)

0.13
(1.16)
0.18
(1.73)
0.05
(0.33)

0.74
(3.63)
0.36
(2.87)
0.38
(1.73)

0.40
(2.51)
0.34
(3.24)
0.06
(0.34)

0.30
(2.22)
0.07
(0.49)
0.37
(2.36)

1.04
(4.77)
0.29
(1.62)
0.75
(2.92)

Stock
Duration

5 1

FF4FactorAlpha
IdiosyncraticVolatility=Low
IdiosyncraticVolatility=High
MBResidualHP

MBResidualHP
1(Low)
2
3(High)
31
1(Low)
2
3(High)
31

0.25
(2.50)
0.30
(2.77)
0.04
(0.37)

0.14
(1.25)
0.28
(2.94)
0.14
(1.29)

0.09
(0.83)
0.19
(2.09)
0.10
(0.93)

0.16
(1.63)
0.11
(1.14)
0.06
(0.44)

0.58
(2.90)
0.39
(3.20)
0.19
(0.91)

0.17
(1.14)
0.31
(3.16)
0.14
(0.85)

0.13
(0.94)
0.11
(0.76)
0.24
(1.48)

0.72
(3.47)
0.29
(1.76)
0.43
(1.83)

High
Low
Idio.
Vola.
0.55
(2.62)
0.18
(1.04)
0.37
(1.49)