-
Electronic copy available at:
http://ssrn.com/abstract=2498743
PatientCapitalOutperformance:TheInvestmentSkillofHighActiveShareManagers
WhoTradeInfrequently
MartijnCremersAnkurPareekUniversityofNotreDameRutgersBusinessSchool
Firstdraft:December2013Thisdraft:September2014
This paper documents that among highActive Share portfolios
whose holdings differ substantiallyfrom theholdingsof
theirbenchmarkonly thosewithpatient investment strategies
(i.e.,with longstock holding durations of at least 2 years)
outperform their benchmarks on average. Funds
tradingfrequentlygenerallyunderperform,regardlessofActiveShare.Amongfundsthatinfrequentlytrade,itiscrucialtoseparateclosetindexfundswhoseholdingslargelyoverlapwiththebenchmarkfromtrulyactivefunds.Theaverageoutperformanceofthemostpatientanddistinctportfoliosequals2.30%peryearnetofcosts
forretailmutual funds. Stocksheldbypatientandactive institutions
ingeneraloutperformby2.22%peryearandbyhedgefundsinparticularby3.64%peryear,bothgrossofcosts.
JELClassifications:G12,G24Contact info:Martijn Cremers: 264
Mendoza College of Business, University of Notre Dame,Notre Dame,
IN46556,Phone:5746314476,
Email:[email protected]:RutgersBusiness
School,1WashingtonPark,Newark,NJ07102,Phone:9733531646,Email:ankur.pareek@business.rutgers.edu.WethanktheQGroupforfinancialsupport.
-
Electronic copy available at:
http://ssrn.com/abstract=2498743
Introduction
Which, if any, activelymanaged portfolios can outperform passive
benchmarks? Theprevious literature has documented that, on average,
the longterm net performance
ofactivelymanagedmutualfundsissimilartotheperformanceoftheirbenchmark(withactivelymanaged
funds generally underperforming their benchmarks but without strong
statisticalsignificance on average).However, some papers argue that
some smaller subset of activelymanagedmutual funds that is
identifiableexante isable
toconsistentlyoutperform,onaverage,overfairlylongperiodsoftime(see,e.g.,Cohen,CovalandPastor,2005;Kacperczyk,Sialm,andZheng,2005;Mamaysky,SpiegelandZhang,2008;KacperczykandSeru,2007;andCremersandPetajisto,2009).Anecessaryconditionfor
longtermoutperformance
isthattheactivelymanagedportfolioissubstantiallydifferentthanthebenchmark,whichisconsideredinKacperczyk,Sialm,andZheng
(2005)andCremersandPetajisto (2009).Both
findthatmutualfundswhoseholdingsaremoredifferent
fromtheirbenchmarks,onaverageand inthe
longterm,outperformtheirbenchmarksnetoffees.Kacperczyk,Sialm,andZheng(2005)consideronly
industrybetsusing the IndustryConcentration
Index,andCremersandPetajisto (2009)consider all stock positions
using Active Share, i.e., the proportion of fund holdings that
isdifferentfromthebenchmark.
FundswithhighActiveSharehave
littleoverlapwiththebenchmarkholdingsandthusaretrulyactivelymanaged.MutualfundswithlowActiveShareshavesimilarholdingsastheirbenchmark.CremersandPetajisto(2009)foundthatsuch
closet
indexfundshavetendedtosignificantlyunderperformtheirbenchmarks(afterfees)inthefuturebyabout1%peryear.Atthe
same time, theyalso found that
fundswithabove90%ActiveShareoutperformed
theirbenchmarksbyabout1%ayear,afterfeesweretakenout.Cremers,Ferreira,MatosandStarks(2014)analyzealargeinternationalsampleofmutualfundsdomiciledin32differentcountries,andlikewisefindthathighActiveSharefundsonaverageoutperformtheirbenchmarksinthefuturewhilelowActiveSharefundsunderperform.
Inthispaper,weexaminethesourceofthisapparentinvestmentskillbythemostactivemanagers,
focusingonhow fundoutperformance isrelatedto
fundholdingdurationsorhow
-
frequentlythefundmanagertrades.Exante, it
isnotclearwhetherfundswouldgenerallybemore successful
throughholding stocks for longperiodsor through frequently changing
theportfolio.Ontheonehand,ifmarketsarefairlyinformationefficient,thenanyinformationthatafundmanageristradingonwillbeincorporatedinthemarketquickly.Asaresult,managersmayneedtofrequentlytradeinordertobenefitfromtheirtemporarysuperiorinformation.
Ontheotherhand, fundmanagersmaybeabletospotmarketmispricingthat
isonlyreversedoverlongerperiods,requiringstrongmanagerconvictionandinvestorpatience.Inthiscase,managerswouldonlyoutperformbyholdingstocksover
longerperiods.Thisrequiresastrong conviction on the part of
themanager, as stock pricesmay initiallymove
adverselybeforereversinganymispricing.Similarly, itrequiresthat
investorsare fairlypatient ingivingthe manager time to see the
strategy through, rather than evaluate the performance
arerelatively brief periods of time. An investment approach with
long holding durations maybenefit from opportunities to buy
relatively illiquid or deepvalue stocks on the cheap (seeAmihud,
Mendelson and Pedersen, 2005), at the detriment of potentially
allowingoverconfidentfundmanagerstopersistinwrongconvictions.
Finally,itseemspossibleexantethatbothshorttermandlongtermopportunitiesmaycoexist,
and that differentmanagerswith different skills set, convictions
and opportunitiesfocusonwhere they think theyhave the
strongestadvantage.Whether
somemanagersareabletobenefitfromtheirshorttermtradesthatcouldpotentiallyexploittemporarymispricingor
shortlived information advantages, and whether other managers can
exploit longtermundervaluationorbenefit froman
illiquiditypremiumarethusseparatebutrelatedempiricalquestions.
Theextantempiricalliteraturehasnotdocumentedastronglinkbetweenfundholdingdurations(ortradingfrequency)andsubsequentfundperformance.Wearguethatthereasonforthis,andtheexplanationthatweareabletofindastrongassociation,isthatwedistinguishbetween
truly actively managed funds and those whose holdings are fairly
similar to
thebenchmarks,i.e.,thatwedistinguishbetweenfundswithlowversushighActiveShare.Amongfundswith
longholdingdurations,onewillnaturally findboth (closet) index funds
and very
-
activemanagerswhopatientlywait for themarket to reward their
longtermbetsagainst
it.Similarly,amongfundswithshortaverageholdingperiods,onemayexpecttofindfundswhichareaggressivelytradingwiththebulkoftheirportfolioaswellasfundswhichholdsubstantialpositions
overlapping the benchmark butwho churn or very frequently turn over
a smallerfractionoftheirportfolio.
Our basic setup is straightforward. We consider both a large
sample of activelymanaged allequity U.S. retail mutual funds from
the CRSP Survivorshipfree mutual
funddatabase,andthesampleofallaggregatedinstitutionalinvestorportfoliosfromtheirquarterly13F
statements. For each sample, we perform 5x5 double sorts of
portfolios into
holdingduration(orturnover)quintilesandActiveSharequintiles,andthencomparetheperformancealongbothdimensions.Foroursampleofmutual
funds,weconsiderthe futureperformanceover the 19year period from
1995 2013, and evaluate the performance of both the netreturns to
investorsafterall trading costsand fees (except frontand rearend
loads)andoftheir holdingsbased gross returns. For our sample of
aggregate institutional portfolios,weanalyzethe
longerperiodof19862013,butonlyconsiderholdingsbased
grossreturnsaswedonotobservetheiractualafterfeenetreturns.WeevaluateperformanceofthenetaftercosttradingreturnsafterbenchmarkadjustingasmotivatedbyCremers,PetajistoandZitzewitz(2013)
generally using the mutual funds selfdeclared benchmarks and
evaluate
theperformanceofthegrossholdingsbasedreturnsaftercontrollingforsizeandbooktomarketcharacteristicsaccordingtoDaniel,Grinblatt,TitmanandWermers(1997).
Weemploy threeproxies forhow long fundshold stocks in
theirportfolios.The firstproxy istheFundDurationasrecently
introduced
inCremersandPareek(2013)andCremers,PareekandSautner(2013)attheportfoliolevel.FundDurationmeasurestheweightedaveragelengthoftimethatthefundhasheld$1ofequities
intheportfoliooverthe
last5years.Thesecondproxyistheholdingsbasedfundturnover,whichwecalculateourselvesforthemutualfund
and institutionalquarterlyportfolios as inGaspar,Massa, andMatos
(2005). The
thirdproxyisonlyavailableforthemutualfundsample,andistheselfdeclaredFundTurnoverRatio,
-
i.e., the ratio of all sales / buys in the calendar year over
the number of fund sharesoutstanding.
ThemedianActiveSharei.e.,theproportionofportfolioweightsthatdoesnotoverlapwithbenchmarkweightsofmutualfundsequals79%inoursample,withahighof86%atthebeginningofoursample
in1994anda lowof73% in2002.Funds
inthebottomActiveSharequintilegenerallyhaveanActiveSharebelow60%and
can thusbe considered closet
indexfunds.FundsinthetopActiveSharequintileportfoliohaveanActiveShareofatleast90%,andthushaveholdingsthatarequitedistinctfromtheirbenchmarks.Themedianholdingdurationisourmutualfundsampleequals14months,whilemutualfundsinthebottomFundDurationquintile
portfolio hold stocks generally for less than 7 months and those in
the top
FundDurationquintileformorethan24months.ThecrosssectionofFundDurationshasbeenfairlystableovertime.1
Ourevidenceshowsthat,amonghighActiveSharefunds,patientlymanagedportfolioshave
been most likely to outperform. Patient funds are those which trade
relativelyinfrequently, i.e., fundswith longholdingdurationsor
lowportfolio turnover.The longtermoutperformance of the highActive
Share andpatient funds is economically remarkable. Forexample,
letsconsidertheequallyweightedportfolioofmutualfundswhereboththeActiveShareandtheFundDurationareinthetopquintile.Thesepatientfundswereabletobeattheirbenchmarksinthe12monthsfollowingportfolioconstruction,aftercosts,by1.92%peryear(tstatof2.08).Adjustingthebenchmarkadjustednetreturnsforexposuretothestandardfivefactors(market,size,booktomarket,momentumandliquidity)increasestheoutperformanceto
2.30% per year (tstat of 3.14). Compounding this over the 19year
period (19952013)producesacumulativeoutperformanceof54%.
1ThemedianActiveShareforourinstitutioninvestorsampleissimilartothemutualfundsampleat73%,HedgefundsarethemostactivegroupofinstitutionswithamedianActiveShareof84%andpublicpensionfundsaretheleastactivewithamedianActiveShareof21%.HedgeFundsarealsothemostshorttermwithamedianFundDurationof0.75
yearsandpublicpension fundsare themost longtermwithamedian
FundDurationof2.41years.
-
Similarly, the stocks held by high Active Share and patient
institutional
portfoliosoutperformby2.22%peryearonaverage(tstatisticof4.10),whichcompoundedoverthe27year
period (19862012) accumulates to an outperformance of 80%. Finally,
among patientinstitutions with high Active Share we find that hedge
funds have particularly strongoutperformance,with an annualized
alpha of 3.64% (tstatistic of 3.04),which compoundedover
the19yearperiod (19942012)aggregates to97%outperformance.
Incontrast, there
islittleevidencethattheholdingsofhedgefundsgenerallyoutperform(seealsoGriffinandXu,2009).ThisagainhighlightstheimportanceofdistinguishingbetweenhighandlowActiveSharehedge
funds, and among highActive Share hedge funds between shortterm and
longterminvestors.2
We findnoevidence thateven themostactive
(i.e.,highActiveShare)mutual
fundswithshortdurationsorfrequenttradingwereabletooutperformtheirbenchmarksonaverage.Rather,
we find that frequently trading mutual funds systematically
underperformed theirbenchmarks, regardless of how different their
holdings are relative to their benchmark. Forexample, the
equallyweighted portfolio ofmutual fundswith short FundDuration and
lowActiveShare(both
inthefirstquintile)underperformedconsiderablywithanabnormalreturnof2.46%peryear(tstatof5.52).ThemutualfundportfoliowithshortFundDurationandhighActiveSharefundssimilarlyunderperformedwithanannualabnormalreturnof1.94%(tstatof
2.45). This suggests that only active bets that were also patient
(or longerterm) wererewarded
inthemarkets,whileweactiveshorttermmutualfundbetswereonaveragequiteunprofitable.3
2Our resultsare robust to removing fromoursample thegroupof
institutionswhoseportfoliosaredominated(withportfolioweightsabove90%)byonesinglestockthatisheldlongterm.ExamplesaretheinstitutionalfilingsofAmericanExpressCompany,variousESOPsholdingsalmostexclusively
thestockof
thesponsoringcompany,andfoundationsandendowmentswhose13Ffilingsalmostexclusivelycontainasinglestock(suchastheGannettFoundation,theGeneralElectricFoundation,theWilliam&FloraHewlettFoundation,theRobertWoodJohnsonFoundation,theKresgeFoundation,theLillyEndowmentandtheUniversityofDelaware).3Forour
institutional sample, the stocksheldby institutions (andhedge
funds)with shortFundDurationshadinsignificant alphas
andDGTWadjusted returns.Naturally, that does notmean that these
institutions did
notoutperformorunderperform,asweonlyobservetheirholdingsandnot,forexample,theirwithinquarterroundtrip
trades. Themore often an institution trades, the less informative
about investment skill the
quarterendholdingswillbe.Therefore,wecannotconcludeanythingabouttheinvestmentskilloftheshortterminstitutions
-
The lack of profitability from active shortterm bets is
consistentwith the literaturearguing that shortterm trading by
institutional investors is associated with greater
pricingefficiency (see,e.g.,Collins,Gong, andHribar,2003;Ke
andRamalingegowda,2005;Bartov,RadhakrishnanandKrinsky,2000;andBoehmerandKelley,2009).This
literatureargues
thatthestocksinwhichshortterminstitutionstradetendtobeefficientlypricedandthusunlikelyto
allow profitable shortterm trades.On the other hand, Puckett and
Yan (2011) find
thatinstitutionalroundtriptradeswithinthequarteraregenerallyprofitable,estimatingthattheseshortterm
trades add about 26 basis points per year on average. However,
Chakrabarty,MoultonandTrzcinka
(2014),examiningthesamedataasPuckettandYan (2011),documentthat
shortterm roundtrip trades ingeneral (i.e.,not limited toonly those
thatarewithinacalendarquarter)arenotprofitable.Onecontributionofourpaper
istoconsiderthisdebatewithinthecontextofthelargeliteratureonmutualfundperformance.
Holding stocks for a relatively long period is in itself on
average not associatedwithgreater investment skill. Rather, Active
Share matters most among funds with long FundDurations,while among
fundswith short FundDurations the difference in the
performancebetweencloset indexfundsandhighActiveSharefunds
iseconomicallysmallandstatisticallyinsignificant. For example and
using equalweighting, the longshort portfolio that is long
infundswithhighActiveShareand longFundDurationandshorts fundswith
lowActiveShareand long Fund Duration has an annualized alpha of
3.47% (tstatistic of 4.78), while
theanalogouslongshorthighlowActiveShareportfolioamongshortFundDurationportfolioshasanalphaof0.44%peryear(tstatisticof0.57).
We try to explain the managerial skill of the high Active Share
and patient
fundmanagersbyexaminingtheirexposureto7factorssuggestedbytheexistingliterature:market,size,
booktomarket, momentum, systematic liquidity, low versus high beta
and earningsquality. For the last two factors,we use the recently
proposed Betting against Beta (BaB)factor (see Frazzini and
Pedersen, 2013) and Qualityminus Junk (QmJ) factor
(seeAsness,Frazzini,andPedersen,2013), respectively.We find that
these latter two factorscanexplainper se, and only observe that
their quarterend holdings do not exhibit any investment skill. See
further ourdiscussioninthenextparagraph.
-
mostof theoutperformanceof
thepatientandhighActiveSharemanagersofmutual funds,butonlya fairly
small fractionof thegrossoutperformanceof the stocks in the
institutionalportfolioswithhighActiveShareandlongFundDuration.Specifically,theclearmajorityoftheoutperformanceof
thepatientandactivemutual fundmanagers seemsdue to theirpickingsafe
(low beta), value (high booktomarket) and quality (profitable,
growing, less valuationuncertainty,higherpayout)stocks,and
thenstickingwith theirconvictionsandholdingon tothoseover
relatively longperiods.Our results thus suggest thatWarrenBuffetts
investmentskill (see Frazzini, Kabiller and Pedersen, 2013) seems
generally shared by mutual
fundmanagersinthetopActiveShareandFundDurationquintiles.
In conclusion, our results indicate that Active Share and Fund
Duration are bothimportantdimensionsofactivemanagement thatstrongly
interact inpredicting theabilityoffundmanagers tobeat
theirbenchmarksand inpredicting theoutperformanceof
thestocksheldininstitutionalportfolios.Whilelongtermfundmanagersingeneralwereunabletobeattheirbenchmark,onlyhighActiveSharemanagerspursuingpatientstrategiesweresuccessful.Thisunderscoresthe
importanceofdistinguishingbetweentrulyactivefundsandthosefundswhose
holdings have significant overlapping holdings with their
benchmarks. Similarly, ourresults highlight the importance of
distinguishing among the truly active funds betweenmanagers
pursuing shortterm mispricing (generally unsuccessfully in our
sample) versusmanagerswho stickwith the courage of their
convictions and patiently (and in our
samplesuccessfully)followabuyandholdstrategywithadistinctportfolio.
2.Data,methodologyanddescriptivestatistics
2.1Dataandmethodology
Weconsidertwodifferentsamplesoffunds.ThefirstsampleincludesactivelymanagedallequityU.S.retailmutualfundsfromtheCRSPSurvivorshipfreemutualfunddatabase,andthe
second sample all aggregate institutional investor portfolios from
their quarterly 13Fstatements.
-
FromtheCRSPSurvivorshipfreemutualfunddatabasethatincludesdead,mergedanddelistedfunds,weusethenetfundreturns(afterfees,tradingcosts,otherexpensesincludingbrokerage
commissions, but before taking out any rear or frontend loads),
total net
assets(TNA)undermanagement,theannualturnoverratioandtheannualexpenseratio.For
fundswithmultipleshareclasses,wesumthetotalnetassetsineachshareclasstoarriveatthetotalnet
assets in the fund. For theexpense ratio, turnover, and
thepercentageof stocks in
theportfolio,weaveragetheseacrossshareclassesweightingbythevalueoftheassets.
We focusonactivelymanaged funds investingalmostexclusively
inU.S.equities.Asaresult,weusethefollowingsampleselectioncriteria,whicharestandard
inthe literatureandwere also used in Cremers and Petajisto (2009).
First, we require the Lipper Prospectusobjectivecode, theStrategic
Insightobjectivecode,and theWeisenbergerObjective code
inCRSPtoindicatethatthefirmispursuinganactiveU.S.equitystrategythatisnotfocusingonone
ormore particular industries or sectors.4 Second,we exclude index
funds and ETFs asindicated by CRSP. Third, we verify that the fund
is primarily focusing on U.S. equities byrequiring
thepercentageofstocks in theportfolioasreportedbyCRSPtobeat
least80%ormissing.For fundswherethisvariable
ismissing,wecalculatethevalueofthestockholdingsfrom
theThomsonholdingsdatabaseandrequire that
thevalueofU.S.commonshares
isatleast80%ofthefundTNA.Fourthandfinally,werequireeachfundtohaveatleast$10millionundermanagement,whichalsoservestomitigateanyincubationorreportingbias.Wemergetheremainingfunds
inCRSPwiththemutualfundholdingsdatabasemaintainedbyThomsonFinancialasavailablethroughWRDSusingthemflinkslinkingfilesonWRDS.
TheinstitutionalinvestorholdingsdatainthisstudycomesfromtheThomsonFinancialCDA/Spectrumdatabaseof
SEC 13F filings.All institutional investorswith greater than
$100millionof securitiesundermanagement are required to report
theirholdings to the SECon
4Specifically,werequiretheLipperProspectusobjectivecodetobeequaltoEI,EIEI,ELCC,G,GI,LCCE,LCGE,LCVE,LSE,MC,MCCE,MCGE,MCVE,MLCE,MLGE,MLVE,MR,S,SCCE,SCGE,SCVE,SESE,SGormissing;werequiretheStrategic
Insight objective code to be equal to AGG, GMC, GRI, GRO, ING, SCG
or missing; we require
theWeisenbergerobjectivecodetobeequaltoGCI,IEQ,IFL,LTG,MCG,SCG,G,GI,GIS,GS,GSI,GS,I,IG,IGS,IS,ISG,S,SGI,SI,SIGormissing;andwerequiretheCDA/Spectrumcodetobe2,3,4ormissing.Wefinallyrequireatleastoneofthesefourcodestobenonmissing.
-
form 13F.Holdings are reported quarterly; all common stock
positions greater than 10,000shares or $200,000must be disclosed
starting in 1980.We obtain the institutional
investorclassificationdatafromBrianBusheeswebsiteandcombinethosewiththeinformationintheThomsonFinancialdatabasetoidentifydifferenttypesofinstitutions:banks(i.e.,equitiesheldinbanktrustdepartments),insurancecompanies,investmentadvisors(includingmutualfundsforretailandinstitutionalclients),hedgefunds,pensionfundsandother(mostlyendowmentsandunclassifiedinstitutions).
Stock returns data are obtained from monthly stock data files
from the Center forResearch inSecuritiesPrices
(CRSP),andaccountingdataare
fromCOMPUSTAT.Tocalculatetrackingerror,weusedailymutual fund
returns fromStandard&Poorsand
theWallStreetWebuntilSeptember1998,andafterwardsfromtheCRSPdailymutualfundreturnsdatabase.The
samedata sourceswereused inCremers andPetajisto (2009) to
calculatemutual
fundtrackingerrors,andseethatpaperfordetails.Thefundtrackingerroristhestandarddeviationofthedifferencebetweenthedailyfundreturnandthedailybenchmarkreturnoverthepast12months.
Weusemutual fund selfdeclaredbenchmarkswhereveravailable.The
sourceof theselfdeclared benchmarks
isMorningstarDirect,whichwemerge toour other data by fundticker
and/or cusip. As Morningstar Direct makes available only the
current selfdeclaredbenchmark and these benchmarks could change
over time for a particular fund, we
usehistoricalMorningstarDirectdataobtainedin2009,2011and2013,andassumethattheselfdeclared
benchmarks in 2009 were applicable also before that year. If the
selfdeclaredbenchmark
isnotavailable,weassignabenchmarkourselvesbasedon thebenchmark
indexthat has the closet fit in terms of holdings (i.e., has the
lowest Active Share across allbenchmarks considered). The main
advantage of this is that if we assign an incorrectbenchmark in
case the selfdeclared benchmark is not available, then that only
happensbecause the funds holdings actually resemble that benchmarks
holdings more than theholdings of any other benchmark. Another
advantage of assigning benchmarks using
theminimumActiveShareisthatthisrequiresnoreturnhistoryandcantakeaccountofchangesin
-
the funds style. Finally,we verify that our results are robust
to not using the selfdeclaredbenchmarks at all but only using the
minimum Active Share benchmarks that we
assignourselvesbasedonthemostcurrentholdings.
The set of benchmarks includes all selfdeclared benchmarks
chosen by funds in
oursampleasavailableinourMorningstarDirectdata.Thebenchmarksarefromthesebenchmarkfamilies:
(1) Calvert Social (1 benchmark), Down Jones (6 benchmarks,
including the DJIndustrialAverage, theDJUSSelectDividend,
theDJWilshire4500and theDJWilshire5000benchmarks),FTSE
(4benchmarks, including
theFTSEHighDividendYield,FTSERAFIUS100andMid Small 1500
benchmarks),Mergent (1 benchmark),MSCI (15 benchmarks,
includingsmall,mid and large capbenchmarksplus their value and
growth components),NASDAQ (2benchmarks,namely theNASDAQ100and
theNASDAQCompositebenchmarks),Russell (13benchmarks, including
small, mid and large cap benchmarks plus their value and
growthcomponents), Standard and Poors (14 benchmarks, including
small, mid and large capbenchmarksplus
theirvalueandgrowthcomponents),andSchwab (2benchmarks,
includingtheSchwab1000andSchwabSmallCapbenchmarks),foratotalof58benchmarks.Weverifythatour
resultsare robust tousinga smaller setof19benchmarksasused
inCremersandPetajisto(2009),whoonlyincludebenchmarksfromtheRussell,S&PandWilshirefamilies.
Theholdingsofthevariousbenchmarkscomefromavarietyofsources.FortheRussellandS&Pbenchmarkfamilieswhicharethemostfrequentlyusedasselfdeclaredbenchmarks
we have the official benchmark constituent weights, from Russell
directly and fromCompustat for S&P. For all other benchmarks,
we approximate the benchmark constituentweights by using the
weights in passive ETFs and passive mutual funds with the
samebenchmarks,averagedoverallavailablepassive
fundswithcompleteholdings information inThomson or CRSP, analogous
to the methodology in Cremers, Ferreira, Matos and
Starks(2014).DailyandmonthlybenchmarkreturnsarefromBloomberg.
Active Share measures the proportion of the funds holdings
(considering only U.S.equitypositions) that isdifferent from
theholdingsof the fundsbenchmark at
aparticularpointintime.Itiscalculatedasfollows:
-
, , (1)
wherewfund,i istheweightofstock i inthe fundandwbenchmark,i
istheweightofstock i
inthebenchmark.ActiveSharethussumsuptheabsolutedifferenceinweightsacrossallstocksthatareineitherthefundorthebenchmarkanddividesthatsumbytwo,treatingoverweightsandunderweight
identically.Afundwithnooverlappingholdings
inthebenchmarkshasanActiveShareof100%,anda fundwithholdingsthatare
identicaltothebenchmarkholdingshasanActiveShareof0%. Ifafunddoesnot
leveruporshort,theActiveSharewillbebetween0%and100%, andmeasures
the sizeof the active (i.e.,different)positions as a fractionof
theentireportfolio.
FundDuration, introduced inCremersandPareek (2013),
iscalculatedas
follows.Wefirstcalculatethedurationofownershipofeachstock
ineveryfundbycalculatingaweightedmeasure of buys and sells by a
fund (either a mutual fund or an institutional
portfolio),weightedbythelengthoftimethestockwasheld.Foreachstockinagivenfund,theholdingdurationmeasureiscalculatedbylookingbacktodeterminehowlongthatparticularstockhasbeenheldcontinuouslyinthatfundsportfolio.
Wecalculate theduration for stock i that is included in the
institutionalportfolio jattimeT1,forallstocks i=1 Iandall
institutional investors j=1J,byusingthefollowingequation:
jiji
jiT
WTt jiji
tjiTjiTji BH
HWBH
tTdDuration
,,
,1
,,
,,1,,1,,
)1()1(
,(2)
where
Bi,j=totalpercentageofsharesofstock iboughtby institution
jbetweent=TWandt=T1;t,Tareinquarters.
Hi,j=percentageoftotalsharesoutstandingofstockiheldbyinstitutionjattimet=TW.
-
i,j,t=percentageoftotalsharesoutstandingofstockiboughtorsoldbyinstitutionjbetweentimet1andt,wherei,j,t>0forbuysand
-
the FundHoldingsTurnover,which is calculatedas
thepercentageofholdings that
changedfromtheendofthisquartersholdingsreporttothepreviousquarterendholdingsreport(seeGaspar,Massa,andMatos,2005).
To evaluate the performance of the fund accruing to the
endinvestor, we use
thebenchmarkadjustedreturnformutualfunds.AsexplainedbyCremers,PetajistoandZitzewitz(2013),
subtracting the benchmark return from the netmutual fund return is
a simple androbustmethod to adjust for the funds exposure to a
particular style or to particular stockcharacteristics.We calculate
the fivefactoralphasof thebenchmarkadjustednet returns tocontrol
for any remaining exposure to the market, size, booktomarket,
momentum andliquidity factors,using thedata for those factors
fromKen Frenchswebsite and the
tradedliquidity(PastorStambaugh)factorfromWRDS.
Forbothmutualfundsandinstitutional13Fportfolioswecalculatetheholdingsbasedreturn.For
this,weassume thatall tradesbetweenquarterlyholdings reportsaremade
justbeforetheholdingsreport
ismadepublic.Weupdateweightsalsoatthetwomonthends
inbetweenquarterlyportfolioreportsbyadjustingtheweightsatthebeginningofthemonthforthe
stock returns during themonth, in order to approximatemore closely
the return of
anactualportfoliowithidenticalholdingsatthebeginningofthequarter.Wecontrolforafundsexposure
to aparticular styleor toparticular stock characteristicsby
calculating
theDGTWadjustedreturnsoftheholdingsbasedreturns.TheDGTWadjustedreturnofeachstockinthefundsportfolioiscalculatedasthedifferenceoftheCRSPmonthlystockreturnandanequallyweightedportfoliowith
similar size, value andmomentum as the stock in the portfolio
(fordetailsseeDaniel,Grinblatt,Titman,andWermers,1997).Finally,
theholdingsbasedDGTWadjustedmonthly return of the fund is
theweightedaverage return of theDGTWadjustedstock returnsusing the
funds (adjusted)portfolioweights.Weagaincalculate the
fivefactoralphas of the holdingsbased DGTWadjusted monthly returns
to control for any
remainingexposuretothemarket,size,booktomarket,momentumandliquidityfactors.
Givenourdemandingdatarequirements(includingtherequirementofat
least2yearsof holdings reports to calculate Fund Duration), we
start our performance analysis for the
-
mutualfundsampleattheendof1994,andfortheinstitutionalportfoliosampleattheendof1985.Atthebeginningofeachyear,wesortfundsinto5quintileportfoliosdependingontheirFundDuration,andwithineachFundDuration(orfundturnover)quintilewesortfundsinto5quintile
portfolios depending on their Active Share. We show results using
both
equallyweightingandvalueweightingtheperformanceofeachofthefundswithineachportfolio.Forrobustness,
we also employ FamaMacBeth crosssectional as well as pooled
panelperformanceregressions.
2.2DescriptiveStatisticsoftheMutualFundSample
Table1providesannualdescriptivestatisticsofthemutualfundsample.Thenumberoffundsstartsat239at
theendof1994andclimbs to1,212 in2002,andremains
fairlystablearound1,200thereafter.Themedianfundsizeequals$180millionin1994and$418millionattheendofoursamplein2013.Expenseratioshavebeenfairlystablewithamedianof1.1%peryear.Thetypicalnumberofstocksintheportfolioisabout73.ThemedianActiveShareinoursampleequals86%in1994,dropsto73%in2002,afterwhichitclimbsbackto80%attheendofthesample.About66%oftheActiveSharesarebasedontheselfdeclaredbenchmarkfromMorningstarDirect,abouthalfthesampleinthebeginningandabout90%attheend.MedianFundDurationswere1.1yearsatthebeginningofthesample,shortenedabitto0.92yearsin2001,afterwhich
theyconsiderably lengthened to1.7yearsat theendof the
sample.FundTurnoverRatioandFundHoldingsTurnoverfollowasimilarpattern,thoughwithlessevidenceoflongerholdingsperiodsattheendofthesamplerelativetothebeginning.Forexample,FundTurnoverRatiohasamedianof38%in1994,77%in2001anddroppingto49%in2013.
PanelAofFigure1showsthepercentageofallTNA
inoursampleattheendofeachyearbyfiverangesofActiveShare:(i)closetindexfundswithActiveShare(below60%),(ii)lowActiveShares(between60%and70%),(iii)moderateActiveSharebetween(70%and80%),(iv)highActive
Share (between80%and90%),and (v) veryhighActive Share
(above90%).ThepercentageofTNA incloset index fundsequals2%
in1994,climbs tooverhalf theassets
in2002andsteadilydropsafterthattoe.g.31%in2008andonly14%in2013.Thepercentageof
-
assetsinveryhighActiveSharefundshasbeenmorestable,withasamplehighof27%atthebeginningofoursamplein1995,alow8.5%in2001andequaling12%in2013.
PanelBofFigure1showsthepercentageoffundsinoursampleattheendofeachyearwithanActiveShare
in the same five rangesasused inPanelAofFigure1.Comparing
thisfigureof thepercentageof funds to the figurebasedon TNA shows
thatmost closet
indexfundsintheearly2000swerelargefunds,thoughthatismuchlessthecaseattheendofoursample.A
largegroupof
fundshasahighActiveSharebetween80%and90%,whichgroupcomprises 28% of
funds in 1994, 18% of funds in 2001 and 29% of funds in 2014.
ThepercentageofveryhighActiveSharefundshassteadilydeclinedovertime,standingat38%in1994,29%in1999and22%in2013.
Panel A of Table 2 gives basic descriptive statistics mean,
standard
deviation,minimumandmaximumforthemainvariablesinourpaperacrossthefullsample.PanelBofTable2presentsthetimeseriesaverageoftheannualquintilebreakpointsforsortsonActiveShareandthethreeproxiesfortradingfrequency.TheActiveSharebreakpointsshowthatthebottomActiveSharequintileportfolioconsistsofcloset
indexfunds,andthetopActiveSharequintile portfolio of very high
Active Share funds. On average, funds in the shortest
FundDurationquintileportfolioholdstocksforatmost0.65years(lessthan8months),whilefundsinthelongestFundDurationquintileportfolioholdstocksgenerallylongerthan2years.FundinthebottomFundTurnoverRatioquintileportfoliohaveaturnoverratiobelow27%peryear,andthose
inthetopFundTurnoverRatioquintileportfoliohaveaturnoverratioabove119%peryear.
PanelCofTable2providestheSpearmanrankcorrelationmatrixofthemainvariables.ActiveShare
isnothighlycorrelatedwithanyofthethreetradingfrequencyproxies,withthehighestrankcorrelationequalto
16%forFundDuration.Thetwoholdingsbasedproxiesaremosthighlycorrelated,withaSpearmanrankcorrelationof82%forFundDurationandFundHoldingsTurnover.Finally,FundTurnoverRatiohasaSpearmanrankcorrelationof70%withFundDurationandof78%withFundHoldingsTurnover.Othernotable
rankcorrelationsarethatfundswithhighActiveSharetendtobesmaller(rankcorrelationof17%withTNA),more
-
expensive(rankcorrelationof29%withthetotalexpenseratio)andcontainfewerstocks(rankcorrelationof29%withthenumberofstocks).
TableDofTable2showstheaveragepercentageofTNAinoursampleineachofthe25portfolios
in the5x5 sortonFundDurationandActiveShare,witheachof
the25portfolioscontainingaboutthesamenumberoffunds.FundsintheshortestFundDurationquintiletendto
be small, representing on average only 9.7% of TNA, while funds in
the longest FundDurationquintile
representonaverage41.6%ofTNA.Abouta thirdof that represents
fundsthat also are in the bottom Active Share quintile, i.e.,
showing that many funds with
longholdingdurationsarenotveryactive.This furtherunderscores the
importanceof separatingfunds in both Active Share and FundDuration
(or fund turnover) dimensions. In the
ActiveSharedimension,closetindexfundsinthebottomActiveSharequintileonaverageholdabout32%oftheassets
inthesample,andtheveryhighActiveSharefunds
inthetopquintileholdabout10%oftheassets.
Panel A of Figure 3 provides the historical Active Share and
Fund Duration for 8generally largemutual funds.These8
fundsarechosen for theirdistinctiveActiveShareandFundDurationat
theendofour sample in2013,andonly serve for
illustration.ExamplesoffundsthatattheendofthesamplehaveahighActiveShareandlongFundDuration(bothtopquintile),includedinthisfigure,arethetheBaronGrowthFundafter2009,theRoycePremierFundafter2005andtheArielFundalsoafter2005.AllofthesefundsgenerallyhadconsistentlyhighActiveShares,thoughexhibitedconsiderablyshorterFundDurationspreviously.ExamplesoffundswithlowActiveShareandhighDurationwouldbetheBlackrockEquityDividendFundafter2005and
theFidelityMagellanFundaround2001.Both funds
canalsobeclassifiedasclosetindexfundsattheendofoursamplein2013,wheretheFidelityMagellanFundhasfairlyshortFundDurationandtheBlackrockEquityDividendFundfairlylongFundDuration.
2.3DescriptiveStatisticsoftheInstitutionalInvestorSample
PanelAofTable3providesdescriptivestatisticsofActiveShareandFundDurationforthe
institutional investorsample
for19862012.ThemeanActiveShareandFundDurationacross all
institutions in thepooled sample is73% and17months, respectively.As
a group,
-
these institutions own 36% of public tradedU.S. equities at the
beginning of our sample in1986,45%in2000and48%in2012.
Wealso report the summary statistics separately for the
followingeight typesof
institutionsbasedontheclassificationschemegiven
inBushee(2001):Banks(BNK),InsuranceCompanies(INS), Investment
Companies (INV), Independent Investment Advisers (IIA), Public
PensionFunds (PPS), Corporate Pension Funds (CPS),University and
Foundations Endowments
(UFE)andother.IndependentInvestmentAdvisers(IIA)arethemostactivewithameanactiveshareof77%andfurtherwithintheIIAgroup,HedgeFunds5aremostactivewithmeanActiveShareof
84%. Panel A of Figure 2 shows that the proportion of assetsmanaged
by institutionalinvestorswithActiveShare lessthan60%hasgoneup
from49.4% in1985to73.5%
in2012.TheproportionofassetsmanagedbytheactivefundswithActiveSharegreaterthan80%hasbeenrelativelystableataround10%of
totalassets (less thanhalfofwhich is in
institutionalportfolioswithActiveShareabove90%).PanelBofFigure2showsthatthenumberoffundsisfairlyevenlydistributedacrossthevariousActiveShareranges,implyingthatlowActiveShareinstitutionalportfoliostendtobemuchlargerinsizethantheinstitutionalportfolioswithhighActiveShare.
PanelsB andCof Table 3 reports thebreakdownofnumberof funds and
assetsbyinstitutionaltypeandfurtherbyvariousActiveSharerangesattheendofoursampleperiod.Independent
Investment Advisers is the largest group with 1,944 institutions
with a totalportfolio valueof$3.47 trillionoutof totalof 2,491
institutionswith an aggregateportfoliovalue of $6.64 trillion.
Hedge Funds comprised a major portion of the investment
advisersgroupwith 707 fundswith totalportfolio valueof$1.28
trillion.Assets forhedge funds areconcentrated in high Active Share
groups (except for Blackrock with $404 billion investedmostly
inpassive fundswithanoverallActiveShareof14%, though is
stillmisclassifiedasahedgefund).Othermajorgroupsofinstitutionsareinvestmentfirmswithtotalassetsof$1.05
5We thankVikasAgarwal forkindlyprovidingus the listofHedgeFunds
inThomson13Fdatabase.This listofhedge funds is based on 13F filings
and ismanually classified by Agarwal, Jiang, Tang, and Yang ( 2013)
andAgarwal,Fos,andJiang(2013).
-
trillionwhicharemostlycomprisedofthemutualfundcompaniesandBankswhicharelargelyinactivewith$1.1trillionundermanagementwithanaverageActiveShareof55%.
Similarly,PanelsAandBofAppendix1presentsthebreakdownofassetsandnumberoffundsby
institutional typesandFundDuration ranges.HedgeFundshave the
lowestholdingdurationwithmore thanhalfof the
fundswithholdingduration less
thanoneyearwhereasindependentinvestmentadvisersareevenlydistributedwithFundDurationsofupto3.5years.
Attheendofeachyear,wefirstsorttheinstitutionsintoquintilesbyFundDurationandthenwithineachFundDurationquintile,wefurthersorttheinstitutionsintoquintilesbasedontheir
yearend Active Share. Panel D of Table 3 reports the average
percentage of
totalinstitutionalinvestorstockholdingsineachofthese5*5portfoliossortedbyHoldingDurationandActiveShare,averagedacrosstime.Eachoftheseportfoliosincludes4%oftotalnumberofinstitutions
in each year but may differ in the amount of assets depending on
the size
ofequitiesundermanagementattheinstitutionsineachoftheseportfolios.ThelowActiveShareandhighFundDurationportfolio(bottomandtopquintile,respectively),i.e.,withpassivelongterminstitutions,comprisesof20.5%oftotalinstitutionalinvestorportfoliovalue.InstitutionswithhighActiveShareand
longFundDuration (both
topquintile)hold2.1%ofallassetsonaverage,jointlyholdingequitiesworth$93.2billionintheirstockportfolioattheendof2011.
Panel B of Figure 3 gives some examples across time of large
institutional investorswhoseActive Share and FundDurations are
distinctive. These examples are only chosen
asillustrative.Theportfolioofpubliclytradedstocksasreportedinthequarterend13FholdingsofBerkshireHathawayshowsageneral
loweringoftheActiveShareovertime,while itsFundDuration
isabout3years from19871992andagainafter2007,but
reachedover4yearsaround19972002.Other large institutions
thatcombineagenerallyhighActiveShareandlong Fund Duration are
Gardner Russo & Gardner, SPO Partners & Co., Ruane Cunniff
&Goldfarb,EarnestPartnersandColumbiaWangerAssetManagement.BlackrocksActiveShareisgenerallylowandhasparticularlydeclinedafter2007,
indicatingthegrowing
importanceofpassiveproducts.Thequarterend13FportfoliosreportedbyBankofAmerica,GoldmanSachsandJPMorgancanallbeclassifiedasclosetindexfunds.
-
Panel C of Figure 3 shows the Active Share and FundDuration of
nine hedge
fundsacrosstime,againsolelyforillustrativepurposes.HedgeFundswhosequarterend13Fholdingsexhibit
high Active Share and long Fund Duration only MHR Fund Management,
WayzataInvestmentPartners,FortressInvestmentGroupandTPGCapital.Hedgefundswhosequarterend
13F portfolios indicate fairly low Active Shares and short Fund
Durations include
AQRCapitalManagement(especiallyafter2008),D.E.ShawandCitadelInvestmentGroup.TheirlowActiveSharesdonotmeanthatthesehedge
fundsdonottendtotakemarketpositionsthataresubstantiallydifferentfromsomebenchmark
index.Aspreviouslymentioned,quarterendportfolioswithfairlyshortFundDurationsare
inherently lessreflectiveofthe
longsideofthetradingpositions,astheshortFundDurationshowsthataconsiderablypartoftheinstitutionstrading
is likelyround tripwithin thequarter.Therefore,
theirquarterendholdingsare likelymuch less informativeabout the
funds tradingstrategyrelative to
institutionswithgenerallylongFundDurations.
3.PatienceProxies,ActiveShareandFundPerformance
A.MutualFundPerformance
Sortingallmutual funds
into5x5portfoliosbasedonFundDurationandActiveShareresultsin25portfolios.UsingmonthlyperformancedatafromJanuary1994toDecember2013,wefirstconsidertheequallyandvalueweighted(bytotalnetassets)performanceofthefundsin
theseportfolios.PanelA (B)ofTable4provides the fivefactoralphasof
thebenchmarkadjustednetfundreturnsusingequally(value)weightedportfolios,andPanelC(D)ofTable4provides
the benchmarkadjusted returns themselves again using equally
(value) weightedportfolios.
Amongallportfoliosof funds,onlythemostactiveandpatientlymanaged
fundshaveonaveragebeenabletooutperformtheirbenchmarksinoursample.Activeandpatientfundsaredefinedasthosewhoseholdingsaresubstantiallydifferentthanthebenchmarks(i.e.,havehighActiveShare)andwhotraderelativelyinfrequently(i.e.,theyhavelongholdingdurationsor
low portfolio turnover). The outperformance of themost active and
patient funds theportfolioofmutual fundswhereboth theActiveShareand
theFundDurationare in the top
-
quintile
iseconomicallyconsiderableandstatisticallystrong.Asshown
inPanelCofTable4usingequallyweightedportfolios,thesepatientfundsbeattheirbenchmarksinthe12monthsfollowingportfolioconstructionby1.76%peryear(tstatof1.98)aftercosts.
Adjustingthebenchmarkadjustednetreturnsforexposuretothefivefactors(market,size,
booktomarket,momentum and liquidity) increases the outperformance
to 2.30% peryear(tstatof3.14),asshown
inPanelAofTable4.Valueweightingtheportfoliosofmutualfundsgivessimilarresults,albeitwithweakerstatisticalsignificance(outperformanceof1.86%peryearandatstatof1.66afterbenchmarkadjusting,seePanelDofTable3,andof1.85%peryearandatstatof1.88afteralsoadjustingforexposuretothefivefactors,seePanelBofTable
4). Theweaker statistical significancemay be due to a few large
funds dominating aportfoliooffunds,giventhelargeskewinfundsize.
The most active mutual funds with short durations or who
frequently trade weregenerally unable to outperform their
benchmarks. Rather, we find that mutual funds thatfrequently trade
on average underperformed their benchmarks by a considerable
margin,acrossall levelsofActiveShare.Forexample,
theequallyweightedportfolioofmutual
fundwithshortFundDurationandlowActiveShare(bothinthefirstquintile)underperformedtheirbenchmarksconsiderablywithanabnormalreturnof2.46%peryear(tstatof5.57seePanelAofTable4).Theunderperformanceof
theanalogousportfoliowith
shortFundDurationandhighActiveShare(inthetopquintile)similarlyunderperformedwithanannualabnormalreturnof1.94%(tstatof2.45,againseePanelAofTable4).Wethusfindthatonlyactivebetsthatwerealsopatient
(or longerterm)were rewarded in themarkets,but findnoevidence
thatactiveshorttermbetswereprofitable.
Further,theresultsforoursampleconfirmthatActiveShareisusefulforpickingactivemanagers,
but find less evidence that highActive Share alone is generally
sufficient to findmanagerswho outperform their benchmarks.We
replicate themain result in Cremers
andPetajisto(2009)thathighActiveSharefundsoutperformlowActiveSharefunds.However,wedo
not find that high Active Share funds as a group have economically
and
statisticallymeaningfuloutperformance,butonlyhighActiveSharefundsthatalsohavelongerinvestment
-
durations.Unconditionally,thefundsinthetopActiveSharequintiletendtooutperformtheirbenchmarks
on average for equallyweighted portfolios, but this outperformance
ismodesteconomically (e.g.,25basispointsperyearalpha
inPanelAofTable4)andnot
statisticallysignificant(tstatisticof0.45).Furthermore,theoutperformanceofthehighActiveSharefundsis
on average negative using valueweighted portfolios, showing that
large funds with highActiveShare(whoarenaturallyweightedmoreheavily
inthevalueweightedportfolios)werelesslikelytooutperform,asalsofoundinCremersandPetajisto(2009).6
LongerFundDuration
isassociatedwithbetterperformance,andespecially for largerfunds.
Across all Active Share quintiles, funds with long Fund Duration
(top quintile)considerable and statistically significantly
outperformed funds with short Fund Duration(bottom quintile) for
valueweighted portfolios. The stronger positive association
betweenperformanceandFundDuration for larger funds isconsistentwith
those larger fundshavingmorepriceimpactwhentheytrade.
Holding stocks for a relatively long period is in itself on
average not
associatedwithmanagerialskillorsignificantoutorunderperformance,andtheonlyportfolioswithlongFundDuration
that outperform are those that also contain high Active Share
funds. Theunconditionalportfolioof funds in the
topFundDurationquintilehasaneconomically
smallandstatisticallyinsignificantabnormalreturn.ThissuggeststhatthepositiveabnormalreturnsdocumentedbyCremers,SautnerandPareek(2014)forportfoliosofstockssortedbyduration(i.e.,averagingFundDurationoverallfundsowningthestock,usingthedollaramountinvestedbyeachfundinthestockastheweights)isduetothestockselectionskillsofthemostactivefundsratherthantosomestockrelatedanomaly.
ActiveSharemattersmostamongfundswith
longFundDurations,whileamongfundswithshortFundDurationsthedifference
intheperformancebetweencloset index fundsandhighActive Share funds
iseconomically small and statistically insignificant. Forexample
and6WecanreconcileourresultsfromthoseinCremersandPetajisto(2009),whoreportedafourfactoralphaforthehighestActiveSharequintileportfolioof1.15%peryearwitha
tstatisticof1.86, from thedifferences in
timeperiod.CremersandPetajisto (2009) show results
for19902003,whileour time seriesextends from19952013. Inour sample,
thecumulativeabnormal returnof the
topActiveSharequintileportfolioequals4%
from1995totheendof2003,butisnegative2%from2004totheendof2013.
-
usingequalweighting,thelongshortportfoliothatislonginfundswithhighActiveShareandlongFundDuration
(bothtopquintile)andshorts fundswith lowActiveShareand
longFundDurationhasanannualizedalphaof4.32%(tstatof5.18).Incomparison,theanalogouslongshorthighlowActiveShareportfolioconditionalonshortFundDuration(bottomquintile)hasafivefactoralphaof0.44%peryear(tstatof0.57).
PanelAofFigure4showsthecumulativeabnormalnetreturnsofa$1investmentinthefourmutual
fund portfolios defined by the extremes in the 5x5 sort on
FundDuration
andActiveShare.Asshowninthefigure,the$1investmentintheportfolioofmutualfundsattheend
of 1994 would have outperformed a $1 investment in the portfolio of
respectivebenchmarks over 40% over our time period,while the other
portfolios underperformed thebenchmarks.
One possible interpretation of these results is that
fundmanagerswith very
distinctportfolios(i.e.,withhighActiveShare)whoalsotradealotmaybeoverconfident.Alternatively,fundsinthetopActiveSharequintilemaybefocusingonrelativelyilliquidstocks(i.e.,relativeto
the stocks in thebenchmark), and frequent trading in those
stocksmay significantlyhurtperformance. This would mean that the
impatient and very active fund managers aredemandingcostly
liquidityprovision.Insubsequentanalysis,wetrytoshedfurther
lightuponthesealternativeexplanations.
Forrobustness,wealsoconsidertwoalternativeproxies
forhowpatientthe
investingstrategyofthefundis,andtheseresultsconfirmthattheoutperformancetendstobestrongestforfundswithhighActiveSharesthattrademorepatiently.First,PanelAofTable5containsthefivefactoralphasbasedonnetbenchmarkadjustedreturnsforequallyweightedportfoliosinadoublesortofFundTurnoverRatioandActiveShare,andPanelBofTable5presentstheanalogousresultsforthedoublesortofFundHoldingsTurnoverandActiveShare.
Usingeitherproxy,wefindthatoutofthe25portfolios,theonlyportfolioswithpositivealphasarethosewithfundsinthetopActiveSharequintileandthebottomthreequintilesforthealternativepatienceproxy.Forexample,
theportfoliowithhighActiveShare
(quintile5)andlowFundHoldingsTurnover(quintile1)hasanannualizedalphaof1.44%withatstatistic
-
of1.91.TheportfoliowithhighActiveShare(quintile5)andlowFundTurnoverRatio(quintile1)hasaneconomicallymeaningfulalphaof0.90%peryear
that is insignificant (tstatisticof1.10),while
theportfoliowithhighActive Share (quintile5) andnextlowest Fund
TurnoverRatio(quintile2)hasanannualizedalphaof1.07%that
isstatisticallysignificant(tstatisticof1.82).
Usingeither turnoverproxy,we find that funds that trademore tend
tohavegreaterunderperformance,andespeciallysoforfundsthatalsohaveahighActiveShare.First,
inthesinglesortonFundTurnoverRatio, theportfolioof funds trading
themost
(quintile5)hasafivefactoralphaof1.58%peryear(tstatisticof2.81),whilethequintile1portfolioexhibitsnosignificant
underperformance (alpha of 0.38% per yearwith a tstatistic of
1.05).
Similarly,usingFundHoldingsTurnoversinglesorts,thequintile5portfoliohasanannualizedalphaof1.97%
(tstatistic of 4.29) and the quintile 1 portfolio has an annualized
alpha of 0.14% (tstatistic of 0.39). In both cases, the difference
in performance across the quintile 1 and
5portfoliosisstatisticallysignificantaswell.
The final robustness check for the main result in this section
using net
benchmarkadjustedreturnsistorun19annualFamaMacBethpredictivecrosssectionalregressionsoftheyearly
net benchmarkadjusted fund return on various fund
characteristics,with theirmeancoefficientspresented
inTable6.Theseregressionsallowustocontrolforseveralotherfundcharacteristicssuchastrackingerror,the
logoffundsizeandtheexpenseratio,andtoverifywhether the
outperformance of the funds with high Active Share and relatively
infrequenttrading is robust across years.We enable an easy
comparison to the portfolio approach
byfocusingontheHigh(Low)ActiveSharedummy,whichequals1ifthefundsActiveShareisin
the top (bottom)quintile that year,plus the interactionsof these
twodummieswith thethreepatienceproxies.Aspreviouslyalsoshown
inthedoublesorts,fundperformance inoursample isnot
linearlyrelatedtoActiveShareandtheproxiesfortradingfrequency,suchthatusing
Active Share itself rather than the dummies does not produce
statistically significantresults.
-
Column (1)ofTable6 shows that fundswithhighActiveShare tended
tooutperformtheirbenchmarksbyabout79basispoints,thoughwithoutstatisticalsignificance(tstatisticof1.45).Likewise,noneofthe3patienceproxiesaresignificantlypredictiveoffundperformance.InColumn
(1),theonly fundcharacteristicstronglyrelatedto future
fundperformance istheexpenseratio,
indicatingthatmoreexpensivefundsperformedworseonaverage.Column(2)showsapositiveandstatisticallysignificantcoefficientfortheinteractionofHighActiveShareand
FundDuration, Column (3) for the interaction ofHigh Active Share
and Fund
TurnoverRatio,andColumn(4)fortheinteractionofHighActiveShareandFundHoldingsTurnover.Thisconfirms
thatonlymutual fundswithhighActiveShare thatalsopursuedpatient
investmentstrategieswereabletooutperformtheirbenchmarksinoursample.
Next,weevaluatethefundperformanceusingtheholdingsbasedgrossreturns,whichis
lessrelevanttoend investorsbut is
informativeaboutmanagerialskillbeforetradingcostsandotherexpensesare
incurred.Asdescribed inSection2,wecalculate
theDGTWadjustedholdingsbasedreturns,andcontrol
foranyremainingexposure to
thestandardmarket,size,booktomarketandmomentumfactors.Table7showstheresultsfortheequallyweighted5x5double
sort on Fund Duration and Active Share. One noticeable change from
the
previousresultsisthatonceweuseDGTWadjustedgrossreturns,wenolongerseemanyportfoliosinthe
double sortwith economically and statistically significant negative
alphas. This indicatesthatmanymanagershaveon averagemanagerial
skill,even if (againon average)notmanyinvestors benefit from those
in the form of net performance that is superior to
theperformanceofthebenchmark.
TheonlyportfoliosexhibitingapositiveandstatisticallysignificantalphafortheDGTWadjusted
gross returns is the portfolio with high Active Share (quintile 5)
and high FundDuration (quintile 5),with an annualized fivefactor
alpha of 1.26% (tstatistic of 1.94).
Thisconfirmsthemainresultthatactiveandpatient
investorsoutperform.BothActiveShareandFundDurationmatter.Similar to
theprevious resultsusingnetbenchmarkadjusted returns,fundswith
longFundDurationtendtooutperformfundswithshortFundDuration,especiallywhenActiveShare
ishigh.Forexample,the longshortportfoliobuying (stocks
in)fundswith
-
longFundDurationandselling
fundswithshortFundDurationunconditionallyhasa4factoralphaof1.21%peryear
(tstatisticof2.31) in the single sort,declining to0.60%peryear
(tstatisticof1.62)forfundsthatarealso
inthebottomActiveSharequintileand,alternatively,rising to 2.38% per
year (tstatistic of 3.28) for funds that are also in the top Active
Sharequintile.
Previously,wehad raised twoalternative interpretationsof
theunderperformanceofvery active managers with short investment
horizons: overconfidence versus a focus onrelatively illiquid
stocks.TheDGTWadjusted gross returns areuseful
todistinguishbetweenthesealternative interpretations,asgross
returnsarenotaffectedby trading costsor costlyliquiditydemands from
impatient trading.Comparing the alphasusingDGTWadjusted
grossreturnsinTable7totheanalogousalphasusingbenchmarkadjustednetreturnsinPanelAofTable3suggeststhatlowActiveSharefundslosemoreintradingcostsandotherexpensesthanhighActiveSharefunds.Thiscastsdoubtontheliquidityexplanation.Wewillempiricallyrevisitboth
interpretations
inSection4,byexploringwhatstrategiesthepatientmanagerswithhighActiveShareusedtosubstantiallyoutperform.
B.InstitutionalPerformance
This subsection uses the institutional 13F stock holding reports
to evaluate theperformanceofaggregated institutional
fundsbasedonDGTWadjustedgrossholdingsbasedreturns.Asfarasweknow,wearethefirsttoexploretheabilityofActiveShareanddifferentproxies
for the trading frequency to predict the performance of the stocks
held by
theseinstitutions.PanelAofTable8reportstheequallyweightedfivefactoralphasina5x5sortsonActiveShareandFundDuration,andPanelBofTable8reports
theequallyweightedDGTWadjustedreturnsthemselveswithoutadjustingtheseforanyremainingexposuretothemarket,size,booktomarket,momentumandliquidityfactors.
Theunconditional sortsonActive Share indicate
thatactivemanagershad
significantmanagerialskill,withthestocksheldbythe institutions
inthetopquintileportfolioofActiveShare outperforming
considerably,with an annualized 5factor alpha of 2.10% (tstatistic
of2.75)andaDGTWadjustedreturnof1.15%peryear(tstatisticof2.36).Astheseinstitutional
-
holdingsreportsareatanaggregatedlevelandincludingsubstantiallylargerassetsthanthosecovered
in the retailmutual fund database considered in the previous
section aswell as
inCremersandPetajisto(2009),theseresultsaretoasignificantextentanoutofsampletestforthe
ability of Active Share to predict outperformance. The other five
Active Share
quintileportfoliosshowstatisticallyinsignificantperformance.
TheoutperformanceofthehighActiveSharemanagersislargelyduetoinstitutionswiththelongestinvestmenthorizons.ForexampleandasshowninPanelAofTable8,theportfolioof
institutionswithhighActiveShare (quintile5)andshortFundDuration
(quintile1)hasanannualized 5factor alpha of 0.40% (tstatistic of
0.50),while the portfoliowith high
ActiveShareandlongFundDuration(quintile5)hasanannualalphaof2.22(tstatisticof4.10).Weillustrate
the substantialoutperformanceof theholdings in
institutionalportfolioswithhighActiveShareandlongFundDurationinPanelBofFigure4.
Stocks inhighActive Share (topquintile) institutionsoutperform
stocks in lowActiveShare (bottom quintile) institutions, but the
difference in performance is only
economicallylargeandstronglystatisticallysignificantforinstitutionsthatalsohavelongFundDuration.Forexample,
among short FundDuration institutions (bottom quintile), the
longshort portfoliothat is long (short)
institutionalportfolioswithhigh(low)ActiveSharehasa5factoralphaof0.65%peryear(tstatisticof0.84),whiletheanalogous
longshortportfolioamong longFundDuration institutions
(topquintile)hasa5factoralphaof2.07%peryear
(tstatisticof3.85).Similarly,thedifferencebetweentheperformanceoftheshortandlongFundDurationquintileportfolios
for thehighActiveShare (topquintile) institutions
iseconomicallymeaningfulandstatistically significant, but is
insignificant for the other Active Share quintiles. Results
usingDGTWadjusted returns in Panel B of Table 8 are similar, ifwith
generally (slightly)weakerstatisticalsignificance.
PanelCofTable8 shows robustness resultsusingadouble sortof all
institutionsonFundHoldingsTurnoverandActiveShare.Weagainfindthatthebestperformanceisachievedby
theportfolioswithhighActive Share (topquintile) and low
FundHoldings Turnover.Thethree portfolios with top quintile Active
Share and bottom three quintiles Fund Holdings
-
Turnoveraretheonlythreeportfolioswithpositiveandstatisticallysignificantperformanceinthe
5x5 sort (though the performance among these 3 portfolios is not
statistically differentfromeachotheror from theperformanceof
theportfoliowith
topquintileActiveShareandtopquintileFundHoldingsTurnover).
Table9focusesonthesampleofhedgefunds.AsdiscussedinSection2.3,hedgefundsarean
importantsubsetof institutions thatat theendofour
timeseriesconstitutesaboutaquarter of the observations in the
institutional sample and about a fifth of the assets.
Asdiscussedpreviously,westartourhedgefundsample
in1995.Quarterendholdingsofhedgefunds tend tohavehighActive Share
(medianof89%),andoftenwithoutmuchpersistenceovertime(medianFundDurationofaround9months).Twomajorlimitationsinevaluatingtheperformanceofhedgefundsarethatweonlyobservethelongpositions(whileshortpositionsmaybeevenmoreimportantformanyhedgefunds)andonlythequarterendpositions(whileroundtrip
trades within the quarter may make up the bulk of their trading
activity).Nonetheless,ashedge fundsareamong
themostactiveandunconstrained institutions,andfastgrowing
insizeandnumber, it
isworthconsideringhowtheirActiveSharesandpatiencearerelatedtotheirholdingsbasedperformance,evenifbasedonquitelimiteddata.InTable8,we
use a 5x3 sort on Fund Duration and Active Share (rather than a 5x5
sort) due to theconsiderably lower number of institutions and
reduce noise. We report 5factor alphas
ofDGTWadjustedreturns,thoughresultsusingDGTWadjustedreturnsthemselvesaresimilar.
Hedge fundquarterend longholdings significantlyoutperform if
theirportfolioshavehigherActiveShare.IntheunconditionalsortofhedgefundsinActiveSharetercileportfolios,thelowActiveSharetercileportfoliohasa5factoralphaisbasicallyzero,whilethehighActiveShare
tercile portfolio has a 5factor alpha of 1.52% per yearwith a
tstatistic of 2.08.
ThedoublesortindicatesthatthisismorelikelytobeduetohedgefundswithlongFundDuration.Whilealphasarenotmonotonicacrossthe
fiveFundDurationquintileswithinthetopActiveSharequintile,amongthosetheonlyportfoliowithapositiveandstatisticallysignificantalphaistheportfoliowithlongestFundDuration,withaneconomicallyremarkable3.64%peryear(tstatisticof3.04).
We illustrate the substantialoutperformanceof the
longonlyquarterend
-
holdings inhedgefundportfolioswithhighActiveShareand
longFundDuration inPanelCofFigure4.
A final robustness check is given in Table 10 using FamaMacBeth
predictive
crosssectionalregressionsofDGTWadjustedholdingsbasedannualreturnson
lagged
institutionalcharacteristics.Columns(1)(3)usethesampleofallinstitutionalportfoliosandcolumns(4)(6)usethesampleofhedgefundsonly.Incolumn(1),wefindthathavingahighActiveSharepredictsanoutperformanceof0.79%peryear
(tstatisticof2.31)and lowActiveShare fundspredicts an
underperformance of 0.28% per year (tstatistic of 1.68, and thus
marginallyinsignificant).Column(2)addstheinteractionsoftheHighandLowActiveSharedummieswithFund
Duration, and finds strong evidence that the outperformance of high
Active Shareinstitutions is due to those that also have longer Fund
Duration. The coefficient of
theinteractionofFundDurationandHighActiveShareequals0.64%(tstatisticof3.16),andonceitis
included thecoefficienton theHighActiveSharedummy
itselfbecomesveryclose
tozeroandinsignificant.Economically,astandarddeviationincreaseininstitutionalholdingdurationisassociatedwitha55basispoints(=0.86x0.64%)higherperformanceamonghighActiveShareinstitutions.
Column (3) robustly confirms that institutionswith high Active
Share
aremorelikelytooutperformiftheytradelessusingthealternativeproxyofFundHoldingsTurnover.
TheFamaMacBethregressionsforthehedgefundsamplelikewiseshowrobustnessofour
main result that hedge funds with higher Active Shares tend to have
better futureperformance, and especially if they also have longer
holding durations or less
quarterendturnover.Asthesampleissmallerandstartsonlyin1995,statisticalsignificanceisabitweakercompared
to the full sample of all institutions, especially using Fund
Holdings Turnover.Economically, the results are strong. For
example, the interaction coefficient in column (5)indicates that a
standarddeviation increase in FundDuration is associatedwith a
138basispoints(=0.86x1.6%)higherperformanceamonghighActiveSharehedgefunds.FundDurationdoes
not seem to predict performance for low Active Share hedge funds,
which tend tounderperformbyabout90basispoints.
-
C.WhathelpsexplaintheoutperformanceofthehighActiveShareandpatientfunds?
ThissubsectionconsidershowtoexplainthemanagerialskillshownbythefundswithhighActiveSharethattradeinfrequently.Weaimtodosobyregressingthemonthlyreturnsofthe
top quintile Active Share and Fund Duration portfolios on seven
different factors
tomeasuretheportfoliosexposuretodifferentgroupsofstocks.Westartwiththebenchmarkadjustedreturnsfortheportfolioofmutualfundswiththemostpatientanddistinctportfolios,andpresentresultsinPanelAofTable11.
Column(1)doesnot
includeanyfactorandreportsanannualizedbenchmarkadjustedreturnof1.92%(tstatisticof2.08),correspondingtotheresultinPanelCofTable4.Addingthemarketfactorincolumn(2)increasesthealphasubstantiallyto2.63%(tstatisticof3.01).Thisindicates
that thepatient and activemanagers tend tobuy stockswith
lowermarketbetas.Column(3)showsanegativeloadingonthesizefactor,suggestingthattheseportfoliostendtobuylargerstocksrelativetothebenchmark.Thepositiveloadingonthebooktomarketfactorin
column (4) indicates that active and patient managers tend to be
value managers.Economically, exposure to the booktomarket factor
seems important, explaining about
54basispointsayear(=2.92%1.36%).Next,theportfoliohasanegativeloadingonmomentumin
column (5) that increases the alpha to 2.22% a year (tstatistic of
2.46), suggesting
acontrariantendencythatalsomoderatelymattersforthealpha.ThetradedsystematicliquidityfactorproposedbyPastorandStambaugh(2003)doesnotplayasignificantroleincolumn(6).
Column(7)confirmsthatactiveandpatientfundstendtobuylowbetastocks,andthatthisexposureisbettercapturedbytheBettingagainstBeta(BaB)factorintroducedbyFrazziniandPedersen(2013)thanbythemarketfactor.RegressingthebenchmarkadjustednetreturnsonaconstantandtheBaBfactorgivesastatisticallyinsignificantalphaof0.92%peryearwithatstatisticof1.07.Column(8)showsthatamajorpartofthemanagerialskillisfurthercapturedbythe
QualityminusJunk (QmJ) factor
introducedbyAsness,Frazzini,andPedersen
(2013).TheregressiononaconstantandtheQmJfactorresultsinanannualizedalphaof1.18%withatstatisticof1.38.
-
Column(9)showsthe5factorresultscorrespondingtoPanelAofTable4,withanalphaof2.30%peryearwithatstatisticof3.14.Thenext3columnsshowthatbothBaBandQmJfactorshelptoreduceandthusexplainthisalpha,withcolumn(11)suggestingthatmostofthereduction
fromthe5factoralpha incolumn (9)comes fromcontrolling
forexposuretohighqualitystocksasmeasuredbytheQmJfactor.Inparticular,the7factormodelsthatincludesallfactors
consideredhereproducesanannualizedalphaof0.59%peryearwitha
tstatisticof0.83,whichconstitutesreductionof74percentagepointsofthe5factoralpha.
We thus conclude that themanagerial skill from themost active
andpatientmutualfund managers is quite similar to that manifested
by Warren Buffett, as documented byFrazzini,KabillerandPedersen
(2013).Theydocument thatBerkshireHathaway,managedbyWarren Buffett,
as well as Berkshires public equity holdings identified from
quarterly 13Fholdingshavesignificantlypositivealphasthatbecome
insignificantonlywhencontrolling fortheir exposure to BaB
andQmJ.Our results thus indicate thatWarren Buffetts skill
seemsgenerallysharedbymutualfundmanagersinthetopActiveShareandFundDurationquintiles:the
clearmajorityof theiroutperformance seemsdue to theirpicking safe
(lowbeta),value(highbooktomarket)andquality(profitable,growing,
lessuncertainty,higherpayout)stocksandholdingontothoseoverrelativelylongperiods.
Next,weconsiderwhethertheperformanceofthemostactiveandpatientinstitutionalportfoliosislikewiseexplainedbytheirbuyingofsafe,valueandqualitystocks.Asindicatedbycolumns
(1) (3) inPanelBofTable11,wedonot findevidence for
this.Startingwith
theaverageDGTWadjustedreturnof1.88%peryear(withatstatisticof3.49)incolumn(1)seealsoPanelBofTable8addingallsevenfactorsonlyreducesthealphato1.73%peryear(withatstatisticof3.06)incolumn(3).WethusconcludethatWarrenBuffettsstyleascapturedbytheBaBandQmJfactorsasinFrazzini,KabillerandPedersen(2013)isnotgenerallysharedamongotherinstitutionswhoinvestactivelyandpatiently.
Finally,columns(4)(6)ofTable11,PanelBfocusesontheperformanceofthestocksinthequarterendportfoliosofthehedgefundswithhighActiveShareandlongFundDuration,whohaveanaverageDGTWadjustedoutperformanceof3.53%peryear(seecolumn(4)).We
-
find that about a thirdof thisoutperformanceor1.2%per year, see
column (6)
canbeexplainedbybuyingvaluestocks,beingcontrarianwithalargenegativeloadingonmomentum,andbuyingqualitystockswithalargepositiveloadingonQmJ.LiketheoutperformanceofthehighActiveShareandpatient
institutions,we leave theexplanationof the remainderof
thisoutperformancebythesehedgefundsforfutureresearch.
4.Conclusion
This paper documents that both Active Share and trading
frequently and theirinteractionare
importantdimensionstoexplainandpredicttheabilityofstockportfoliostooutperform.
For both retail mutual funds and general institutional portfolios,
we find thatamonghighActiveShareportfoliosonly thosewithpatient
investmentstrategiesareable tooutperform their benchmarks on
average. Funds which trade frequently
generallyunderperformtheirbenchmarks,regardlessofActiveShare.
Thismeans that among funds that infrequently trade, it is
crucial to separate closetindex fundswhoseholdings
largelyoverlapwiththebenchmark fromtrulyactive
fundswhoseholdingsarealmostcompletelydifferentfromtheirbenchmarkswithhighActiveShares.Economically,ourresults
indicatesubstantialbenefitsfromthe
investmentskillofhighActiveSharemanagersthattradeinfrequently.Theaverageoutperformanceofthemostpatientanddistinctportfolios(usingquintilesorts)equals2.30%peryearnetofcostsforretailmutualfunds(tstatisticof3.14).
Compoundingthis2.30%annualoutperformanceoverthe19yearsoverwhichitisestimated(19952013)givesacumulativeoutperformanceof54%.Stocksheldbypatientandactive
institutions
ingeneraloutperformby2.22%peryear(tstatisticof4.10)andbyhedgefundsinparticularby3.64%peryear(tstatisticof3.04),bothgrossofcosts.
The average underperformance of retail mutual funds that
frequently trade
andgenerallyholdstocksforlessthan7monthsarelikewiseeconomicallysubstantial.Forexample,theportfolioofmutualfundswithshortFundDurationand
lowActiveShare(both inthefirstquintile)hadanabnormalnet returnof
2.46%peryear (tstatof5.52),which compounded
-
overour19yearperiodfrom19952013wouldaccumulatetoatotallossof38%oftheinitialamountinvestedrelativetoapassiveinvestmentinthemarketatnocost.
OurresultssuggestthatU.S.stockmarketsmaybetoo
informationefficient,withtoomuchcompetitionamongactive
investorsortoomuchprice impactof informationtrades,forinvestors to
systematically benefit from frequently trading mutual fund managers
pursuingshortterm predictability. At the same time, our results
also suggest that one can ex
anteidentifyadistinctivegroupoffundmanagersthatseemtoexhibitsubstantial
investmentskillwhenmeasuredoversufficientlongperiodsoftime.ThisisthegroupofpatientfundmanagersthatsticktotheirconvictionsandcombineahighActiveSharewith
infrequenttrading.Inour5x5sortof retailmutual funds,
theoutperformance isstrongest in the2portfolioswithhighActiveShare
(topquintile)andtop40%ofFundDuration
(top2quintiles),constituting8%offundsandonaverageabout6%ofassetsinoursampleofretailmutualfunds.
How did these patient and active mutual fund managers
outperform? Perhapsunsurprisingly, their outperformance can largely
be explained by their focus on stocks
thatotherinvestorsshunorfindlessattractivefortheirimpatientstrategies:pickingsafe(lowbeta),value
(high booktomarket) and high quality (profitable, with growing
profit margins,
lessuncertainty,higherpayout)stocksandthenstickingwiththoseoverrelativelylongperiods.OurfindingsthussuggestthatWarrenBuffettsinvestmentskill(seeFrazzini,KabillerandPedersen,2013)seemsgenerallysharedbytheseretailmutualfundmanagersinthetopActiveShareandFundDurationquintiles.
We have less success in explaining the large gross
outperformance of institutions especiallyhedge
fundswithhighActiveShareand longFundDuration.Thisgroup
includesBerkshireHathaway,probably thebest knowpatient investor,
formuchofour timeperiod.However,we find
thatBerkshireHathawaypublicequityholdings in the
last10yearsofoursamplearenolongerverydistinctfromthemarketportfolio(withanActiveSharearound80%since2007andaround70%attheendofoursample
in2012).BerkshireHathawayremainsapatient investor in public
equities, even though the holding duration of public stocks in
thequarterend13FreportsofBerkshireHathawaysportfoliohasshortenedfromover4years
in
-
2002 to a bit over 3 years. Frazzini, Kabiller and Pedersen
(2013) show that BerkshireHathaways remarkable performance can
largely be explained by exposure to safe and
highqualitystocks,whichwe
likewisedocumentforpatientandactivemutualfunds.However,wefindthatexposuretolowbeta,booktomarketandqualityfactorsexplainsonlyasmallfractionoftheoutperformanceofthestocksinthequarterendportfoliosofhighActiveShareandlongFundDurationinstitutionsandhedgefundsgenerally.
We thus leave the the investment skillofpatient
institutionswithverydistinctpublicequityportfolioslikeRuaneCunniff&Goldfarb,GardnerRusso&Gardner,SPOPartners&Co,for
example, as largely unexplained. The investment skill of the fund
managers at
theseinstitutionsmaynotbeeasilycapturedinquantitativefactorsormaychangesubstantiallyovertimedependingonthecircumstances.Perhapstheirsuccessdependsonbeingtrulypatientandcontrarian,willing
towait yearsbeforedeploying cash in aworldwhere typical
institutionalportfolioshave100%turnoverbasedonquarterendholdings(andthusignoringanyroundtriptradingwithinquarters).Relatedly,perhapstheseinstitutionsaretooidiosyncraticanduniquetosummarizetheircollective
investmentskillaswecanforthemostactiveandpatientretailmutualfunds.
At the same time, the outperformance we document is based on
publicly
availablequarterendportfolioholdingsthatanyonecaneasilyaccessontheSECwebsite.Thissuggeststhat
other market participants may be either unaware of the persistent
success of theseinstitutions, or, if they are aware, other market
participants may underestimate theirinvestmentskill.
Or,alternatively,evenifothermarketparticipantsseeandappreciatethetrackrecord,theymaysimplybediscouragedbecauseoftheirown
lackofpatience.Asaresult,conditionssuchashavingastableinvestorbaseandarelativelackofpersistentcopycatsmayallowtheseinstitutionstobeuniqueandpatient(and,itturnsout,outperform).
How patient does one need to be? In Panel A of Figure 4,
plotting the cumulativeabnormaloutperformance forportfoliosofmutual
funds shows that theportfolioofmutualfundswithhighActiveShareand
longFundDurationoutperformedonlyslightly inthe first5
-
years(19951999),thensubstantialoutperformancewasachievedin2000,withagainlittlegainin
the 8 year period of 20012008, after which again substantial
outperformance followed.PanelCofFigure4showssimilarlythat forhedge
fundsoneneededtobepatience fromthemiddleof2000 to
theendof2008before thequarterendholdingsof thepatient andhighActive
Sharehedge funds substantiallyoutperformed.PanelBof Figure4 for all
institutionsindicates that the portfolio of high Active Share and
long Fund Duration institutionsoutperformed quite consistently over
time. However, thismasksmuch dispersion and
timevariationwithintheportfolio,asshown
inPanelDofFigure4,whichpresentsthecumulativeabnormalreturnsforsomeofthe
largest institutionswith longFundDurationandhighActiveShare. These
figures underscore that both fundmanagers and investors sometimes
need towaitseveralyearsbeforegettingrewardedfortheirpatience.
A good example of an institutionwith a very active and patient
approach is
Ruane,Cunniff&GoldfarbInc.,managerofe.g.theSequoiafundthattypicallyinvestsinahandfulofpublicly
traded stocks. The large idiosyncratic element in their investment
strategy couldillustratewellwhy trying toexplain
theiroutperformance ishard ifnot impossible, i.e.,highActive Share,
patient institutions may generally be very distinct from each other
andidiosyncratictothemarket.Thisdoesnotmeanthatthereisnoparticularinvestmentstrategy.Forexampleforthe
investmentstrategyoftheSequoiafund,apreferenceforstocksthatcanbe
held for the longterm rather than for stocks whose performance may
depend on thebusiness cycle (potentially warranting trading at
business cycle turning points) seemsfoundational: We're really
onebalancesheetata time, onecompanyatatime kind
ofinvestors,Mr.Poppe [comanagerof theSequoia fund]said inMay2008at
theshareholdermeeting.Wedon'tspendalotoftimeonmacro.Today,thefocusisstillonbalancesheets,but
the Sequoia managers try to minimize the impact of economic ups and
downs on theportfolio. There'saheavierburdenonacyclicalcompany to
find itsway into theportfolio,saysMr.Goldfarb [comanager of the
Sequoia fund], as quoted in theWall Street Journal(2010).
-
ReferencesAgarwal, V., Jiang, W., Tang, Y. and Yang, B., 2013,
Uncovering hedge fund skill from the
portfoliostheyhide,JournalofFinance68,739783.Agarwal, V., Fos,
V. and Jiang, W., 2013, Inferring reportingrelated biases in hedge
fund
databasesfromhedgefundequityholdings,ManagementScience,59,12711289.Amihud,Y.,Mendelson,H.andPedersen,L.H.,2005,LiquidityandAssetPrices,inFoundations
andTrendsinFinance14,269364.
Asness, C., Frazzini, A. and Pedersen L. H., 2013, Quality Minus
Junk, working paper, AQRCapitalManagement,NewYorkUniversity.
Bartov,E.,Radhakrishnan,S.andKrinsky,I.,2000,Investorsophisticationandpatternsinstockreturnsafterearningsannouncements,AccountingReview75,4363.
Boehmer, E. and Kelley, E., 2009, Institutional investors and
the informational efficiency
ofprices,ReviewofFinancialStudies22,35633594.
Bushee, B. (2001) Do institutional investors prefer near-term
earnings over long-run value?,
Contemporary Accounting Research 18, 207246.
Chakrabarty,B.,Moulton,P.C.,andTrzcinka,C.,2014,ShorttermInstitutionalTrades,WorkingPaper.
Cohen,R.B.,Coval, J.D.,andPastor,L.,2005, JudgingFundManagersby
theCompanyTheyKeep,JournalofFinance60(3),105796.
Collins,D.,Gong,G.andHribar,P.,2003,Investorsophisticationandthemispricingofaccruals,ReviewofAccountingStudies8,251276.
Cremers, K. J. M., Ferreira, M., Matos, P., and Starks, L.,
2014, The Mutual Fund
IndustryWorldwide:ExplicitandClosetIndexing,Fees,andPerformance,WorkingPaper.
Cremers,K. J.M.,andPetajisto,A.,2009,Howactive isyour
fundmanager?Anewmeasurethatpredictsperformance,ReviewofFinancialStudies22(9),33293365.
-
Cremers,K. J.M., andPetajisto,A., and Zitzewitz, E., 2013,
ShouldBenchmark
IndicesHaveAlpha?RevisitingPerformanceEvaluation,CriticalFinanceReview2,148.
Cremers, K. J. M., and Pareek, A., 2013, Shortterm trading and
Stock Return
Anomalies:Momentum,ReversalandShareIssuance,WorkingPaper.
Cremers,K.J.M.,Pareek,A.,andSautner,Z.,2013,StockDurationandMisvaluation,WorkingPaper.
Daniel, K., Grinblatt, M., Titman, S. and Wermers, R., 1997,
Measuring mutual
fundperformancewithcharacteristicbasedbenchmarks,JournalofFinance52,10351058.
Frazzini,A.,Kabiller,D.,andPederseon,L.,2013,BuffettsAlpha,NBERworkingpaper19691.
Frazzini, A. and Pedersen, L.H., 2013, Betting Against Beta,
Journal of Financial Economics,Forthcoming.
Gaspar,JM.,Massa,M.andMatos,P.,2005,Shareholderinvestmenthorizonsandthemarketforcorporatecontrol,JournalofFinancialEconomics76,135165.
Griffin,J.,Xu,J.,2009,HowSmartaretheSmartGuys?AUniqueViewfromHedgeFundStockHoldings,ReviewofFinancialStudies22(7),25312570.
Kacperczyk, M., Sialm, C., and Zheng, L., 2005, On the industry
concentration of
activelymanagedequitymutualfunds,JournalofFinance60(4),19832011.
Kacperczyk,M.,andSeru,A.,2007,Fundmanageruseofpublicinformation:Newevidenceonmanagerialskills,JournalofFinance62,485528.
Ke, B. and Ramalingegowda, S., 2005, Do institutional investors
exploit the
postearningsannouncementdrift?,JournalofAccountingandEconomics39,2553.
Mamaysky,H.,Spiegel,M.,andZhang,H.,2008,Estimatingthedynamicsofmutualfundalphasandbetas,ReviewofFinancialStudies21,233264.
-
Pastor, L. and Stambaugh, R., 2003, Liquidity risk and expected
stock returns, Journal ofPoliticalEconomy1113,642685.
Puckett,A.,andYan,X.,2011,The InterimTradingSkillsof
Institutional Investors, JournalofFinance66(2),601633.
WallStreetJournal,2010,StickingtoWhatWorks,articlebyTomLauricella,July10,accessedathttp://online.wsj.com
-
Table1.MutualFundSample:MediansbyYearThistableprovidesannualdescriptivestatisticsforthemutualfundsample.Medianvaluesforthevariablesarereportedforeachsampleyear.
Year #ofFunds TNA ExpenseRatio #Stocks ActiveShareFund
DurationFund
TurnoverRatio
FundHoldingTurnover
1994 239 180 1.11% 69 86% 1.08 38% 1.291995 382 181 1.15% 72 85%
0.91 44% 1.581996 500 227 1.14% 75 84% 0.87 47% 1.581997 583 259
1.15% 72 84% 0.88 50% 1.511998 727 290 1.16% 71 82% 0.91 63%
1.591999 881 290 1.19% 70 79% 0.93 65% 1.662000 958 261 1.18% 71
76% 0.93 72% 1.642001 1,036 242 1.23% 73 74% 0.92 77% 1.442002
1,166 172 1.25% 74 73% 1.02 68% 1.362003 1,180 197 1.30% 77 75%
1.09 69% 1.472004 1,167 219 1.27% 76 78% 1.14 67% 1.422005 1,185
239 1.25% 75 79% 1.19 66% 1.422006 1,187 240 1.21% 74 80% 1.21 64%
1.432007 1,141 248 1.18% 72 80% 1.26 64% 1.372008 1,136 165 1.16%
73 78% 1.22 72% 1.402009 1,196 197 1.16% 80 79% 1.18 75% 1.472010
1,263 225 1.14% 76 78% 1.22 60% 1.342011 1,255 269 1.12% 73 79%
1.40 54% 1.232012 1,182 298 1.11% 71 79% 1.55 50% 1.152013 1,066
418 1.10% 74 80% 1.69 49% 1.15
-
Table2.MutualFundSample:DescriptiveStatisticsPanelAreportsthedescriptivestatisticsforthemutualfundsampleusedinthispaper.PanelBreportsthequintilecutoffsforActiveShareandthreevariablescapturingmutualfundtradingfrequency:FundDuration,FundTurnoverRatioandFundHoldingTurnover.PanelCpresents
thecorrelationbetweenthemutual fundvariables.At theendofeachyear,
fundsare firstsorted intoquintilesby
theirFundDurationvalueandthenwithineachfunddurationquintilefurthersorted
intoquintilesbasedontheirActiveShare. PanelD reports
thepercentageofassets ineachof
the25FundDuration/ActiveSharegroupsobtainedby sorting funds
intoquintiles firstby theirFundDurationvalueand
thenbyActiveSharevaluewithineachFundDurationquintile.PanelA
Mean StDev. Min MaxActiveShare 77% 16% 20% 100%FundDuration 1.36
1.75 0.00 4.74FundTurnoverRatio 85% 80% 0% 2579%FundHoldingTurnover
162% 49% 0% 747%Log(TNA) 5.54 1.63 2.30 11.62ExpenseRatio 1.22%
0.42% 0.03% 5.32%#Stocks 117 195 10 3,450
PanelB
Q20 Q40 Q60 Q80ActiveShare 62% 74% 84% 92%FundDuration 0.65 0.99
1.38 2.00FundTurnoverRatio 27% 50% 77% 119%FundHoldingTurnover 83%
122% 164% 227%
PanelC
ActiveShare
FundDuration
FundTurnoverRatio
FundHoldingTurnover Log(TNA)
ExpenseRatio
#Stocks
ActiveShare 100% FundDuration 16% 100% FundTurnoverRatio 3% 70%
100% FundHoldingTurnover 7% 82% 78% 100% Log(TNA) 17% 20% 11% 12%
100% ExpenseRatio 29% 22% 16% 16% 32% 100% #Stocks 29% 8% 17% 11%
25% 16% 100%
-
PanelD
ActiveShare Q1 Q2 Q3 Q4 Q5 Sum
Fund
Duratio
n Q1 3.2% 2.4% 1.6% 1.4% 1.1% 9.7%Q2 3.7% 3.5% 2.6% 1.7% 1.2%
12.7%Q3 4.3% 3.9% 2.9% 2.4% 1.7% 15.2%Q4 6.0% 5.5% 3.8% 3.2% 2.4%
20.8%Q5 14.6% 9.8% 7.5% 5.8% 3.9% 41.6%
Sum 31.8% 25.1% 18.4% 14.4% 10.3%
-
Table3.InstitutionalInvestors:SummaryStatisticsPanelAreportsthesummarystatisticsforActiveShareandFundDurationvariablesfortheinstitutionalinvestorsample.Thesampleperiodforinstitution
investors is from1986 to2012except forHedgeFundswhere
thesampleperiod is from1995 to2012due to fewer funds in
theearlieryears.Ndenotesthenumberoffundyearobservations.FirstandSecondrowspresentthestatisticsforallinstitutionsandhedgefundsrespectively,remainingrowspresentthesummarystatisticsfordifferent
institutional investortypesasdefined
inBushee(2001).PanelBandPanelCpresents thenumberof institutionsand
institutional investorassets ($billion)
indifferentActiveSharegroups.PanelD reports
thepercentageofinstitutionalinvestorassetsineachofthe25FundDuration/ActiveSharegroupsobtainedbysortingfundsintoquintilesfirstbytheirFundDurationvalueandthenbyActiveSharevaluewithineachFundDurationquintile.PanelA
ActiveShare(%) FundDuration N Mean Stdev P20 Median P80 Mean
Stdev P20 Median P80
AllInstitutions 38769 72.60 19.48 56.29 74.80 91.61 1.42 0.85
0.64 1.34 2.16HedgeFunds 7713 83.90 15.38 72.33 89.20 96.61 0.92
0.74 0.27 0.74 1.50
BNK 4828 55.1 13.9 45.0 54.7 65.7 1.97 0.69 1.40 1.98 2.56INS
1673 62.7 23.5 41.4 66.7 83.2 1.75 0.89 1.00 1.59 2.58INV 1353 70.5
19.6 55.4 73.5 88.4 1.45 0.70 0.88 1.34 2.04IIA 26265 77.0 16.6
62.7 79.3 92.9 1.28 0.78 0.57 1.19 1.94CPS 865 58.5 27.8 32.7 61.8
85.1 1.79 0.98 0.91 1.64 2.64PPS 410 33.5 23.9 12.4 23.8 60.9 2.21
0.79 1.53 2.30 2.88UFE 325 72.2 19.6 57.0 75.4 90.9 1.89 0.96 1.08
1.73 2.76
OTHER 3050 78.4 18.7 60.6 83.3 96.4 1.25 1.11 0.14 1.02 2.26
-
PanelB
NumberofInstitutionsActiveShare All HedgeFunds BNK INS INV IIA
CPS PPS UFE MSC
010 11 0 1 2 3 2 2 11020 26 1 5 4 8 8 12030 31 4 3 4 1 16 3 3
13040 43 7 10 7 3 14 3 4 24050 118 9 39 3 1 59 3 3 1 95060 281 33
52 4 7 189 5 2 226070 436 56 29 10 4 355 7 1 307080 462 112 19 7 6
388 4 2 368090 467 164 8 4 8 398 1 3 4590100 616 321 3 4 514 6 2
87Total 2491 707 168 44 36 1944 34 25 6 234
-
PanelC
TotalAssets($billion)ActiveShare All HedgeFunds BNK INS INV IIA
CPS PPS UFE MSC
010 522.0 0.0 0.0 6.4 392.1 21.3 7.0 39.9 55.41020 1438.7 406.4
691.7 105.0 0.0 472.5 0.0 104.7 64.82030 137.4 29.5 7.8 23.1 10.0
57.5 3.5 20.7 14.83040 552.1 52.8 165.0 90.3 31.7 236.7 1.4 26.1
0.94050 1080.8 155.8 155.2 7.8 260.9 536.2 1.0 13.9 3.1 102.85060
1037.3 94.2 50.6 39.6 212.5 692.9 19.0 1.9 20.76070 555.6 106.9
27.6 21.2 51.9 429.2 5.8 5.0 14.97080 670.0 158.9 4.6 6.3 64.2
513.7 1.6 7.2 72.58090 396.9 127.8 1.9 7.5 18.8 319.8 0.2 0.0 11.4
37.490100 253.3 143.9 1.4 0.0 6.8 192.0 8.5 0.0 1.0 43.6Total
6644.3 1276.2 1105.9 307.2 1048.8 3471.7 48.0 219.4 15.5 427.7
PanelD
ActiveShare Q1 Q2 Q3 Q4 Q5
F
u
n
d
D
u
r
a
t
i
o
n
Q1 3.7% 1.5% 1.0% 0.7% 0.3%Q2 6.5% 3.3% 1.8% 1.1% 0.5%Q3 12.9%
4.4% 2.9% 1.7% 0.8%Q4 16.0% 4.7% 2.4% 2.5% 1.2%Q5 20.5% 2.6% 2.6%
2.4% 2.1%
-
Table4.ActiveShareandFundDuration:MutualFundBenchmarkAdjustedNetReturnsThis
table reports abnormal returns for mutual funds conditional on
their Active Share and
FundDuration.Fundsareassignedtooneofthe25portfoliosattheendofeachyearbysortingintoquintilesfirst
based on Fund Duration and then further by Active Share within each
Fund Duration quintile.Returnsare
thencalculatedovernextoneyear.Panel A reports
theequallyweighted5factoralphaobtainedastheinterceptfromtheregressionoftheaveragemonthlynetfundreturnsinexcessoftheirselfdeclaredbenchmarkonFamaandFrench3factors,theCarhartmomentumfactorandPastorandStambaughtradedliquidityfactor.Iftheselfdeclaredbenchmarkismissing,thenthebenchmarkindexwithminimumdeviation
isassignedas that fundsbenchmark.The first row reports the returns
for5portfoliosobtainedbysortingfundsunconditionally
intoquintilesbytheirActiveShare.
Similarly,thefirstcolumnreportsalphasforportfoliosobtainedbysortingunconditionallyonFundDurationmeasure.RestofthePanelpresentreturnsforportfoliosobtainedbysortingbothonFundDurationandthenonActiveShare.PanelBpresentsthevalueweighted5factoralphas.PanelC,PanelDreporttheaveragevalueweightedandequalweightednetfundreturnsinexcessofthebenchmarkindex.PanelA
EqualWeighted5FactorAlphaofBenchmarkAdjustedNetReturns
ActiveShare
FundDuration Uncond. 1 2 3 4 5 51Uncond. 0.98 1.45 1.44 0.84
0.25 1.22
(4.49) (3.97) (3.64) (1.74) (0.45) (2.47) 1 1.96 2.46 2.00 1.64
1.75 1.94 0.44 (3.87) (5.52) (4.01) (2.57) (2.13) (2.45) (0.57)2
1.20 1.46 1.84 1.70 0.67 0.34 1.09 (3.20) (4.39) (3.84) (3.37)
(1.20) (0.50) (1.64)3 0.96 1.52 1.14 1.54 0.31 0.27 1.20 (2.39)
(5.39) (2.47) (3.33) (0.49) (0.37) (1.74)4 0.67 0.91 1.48 1.06 0.55
0.65 1.69 (1.96) (3.17) (3.94) (1.20) (1.07) (1.08) (2.89)5 0.37
1.06 1.11 1.54 0.44 2.30 3.47 (0.98) (4.44) (2.84) (3.09) (0.75)
(3.14) (4.78)
51 1.68 1.28 0.89 0.39 1.52 4.32 (3.46) (2.99) (1.83) (0.57)
(1.80) (5.18)
-
PanelB
ValueWeighted5FactorAlphaofBenchmarkAdjustedNetReturns
ActiveShare
FundDuration Uncond. 1 2 3 4 5 51Uncond. 0.98 1.45 1.44 0.84
0.25 1.22
(4.49) (3.97) (3.64) (1.74) (0.45) (2.47)1 2.52 1.64 2.39 3.85
3.59 3.35 1.55 (3.31) (2.46) (3.39) (4.47) (2.67) (2.11) (1.03)2
1.49 1.36 1.75 2.14 1.31 1.66 0.21 (2.79) (2.39) (2.36) (3.13)
(1.58) (1.41) (0.17)3 0.78 2.10 0.68 1.10 0.58 0.40 1.50 (1.26)
(4.29) (0.80) (1.49) (0.67) (0.42) (1.52)4 0.68 0.48 0.78 1.48 1.63
0.71 0.41 (1.25) (0.95) (1.36) (1.58) (2.30) (0.48) (0.27)5 0.04
0.27 1.34 0.30 0.73 1.85 2.57 (0.09) (0.60) (2.30) (0.38) (0.83)
(1.88) (2.57)
51 2.30 1.10 0.97 3.46 2.76 5.22 (3.15) (1.52) (1.22) (3.64)
(2.20) (3.15)
PanelC
EqualWeightedBenchmarkAdjustedNetReturn
ActiveShare
FundDuration Uncond. 1 2 3 4 5 51Uncond. 0.92 0.97 1.06 0.49
0.05 0.96
(3.30) (2.40) (2.26) (1.01) (0.08) (1.83) 1 1.15 1.60 0.97 0.92
1.09 1.22 0.37 (1.67) (2.76) (1.53) (1.02) (1.12) (1.31) (0.45)2
0.81 0.83 1.34 1.22 0.42 0.26 0.58 (1.87) (2.22) (2.66) (2.09)
(0.73) (0.35) (0.84)3 0.57 0.90 0.61 1.01 0.27 0.60 0.29 (1.40)
(3.19) (1.24) (2.06) (0.46) (0.81) (0.40)4 0.48 0.36 1.19 0.84 0.58
0.56 0.93 (1.39) (1.39) (2.89) (0.73) (1.07) (0.90) (1.42)5 0.38
0.90 0.75 1.33 0.65 1.92 2.66 (0.79) (4.40) (1.85) (2.21) (0.90)
(2.08) (3.09)
51 0.78 0.69 0.22 0.42 0.44 2.98 (0.92) (1.28) (0.35) (0.38)
(0.35) (2.31)
-
PanelD
ValueWeightedBenchmarkAdjustedNetReturn
ActiveShare
FundDuration Uncond. 1 2 3 4 5 51Uncond. 0.16 0.43 0.70 0.44
0.54 0.38 (0.37) (0.74) (1.19) (0.61) (0.63) (0.42)
1 1.40 0.60 1.74 2.30 1.98 1.70 1.11 (1.60) (0.75) (2.18) (2.13)
(1.36) (1.10) (0.76)2 0.82 0.59 0.66 1.68 0.88 1.57 0.98 (1.18)
(1.00) (0.71) (2.13) (0.99) (1.21) (0.76)3 0.24 0.88 0.53 0.84 0.56
0.76 0.12 (0.38) (1.56) (0.64) (1.09) (0.71) (0.63) (0.10)4 0.58
0.58 0.40 1.32 1.07 0.52 0.07 (0.93) (1.18) (0.56) (0.99) (1.61)
(0.31) (0.04)5 0.25 0.02 0.38 0.15 0.89 1.66 1.67 (0.42) (0.04)
(0.62) (0.18) (1.00) (1.53) (1.53)
51 1.66 0.58 1.36 2.46 1.08 3.36 (1.79) (0.71) (1.51) (2.03)
(0.70) (1.91)
-
Table5.FundTurnoverandActiveShare:5FactorAlphaofBenchmarkAdjustedNetReturnsThistablereportstheaverageabnormalmutualfundreturnsconditionalonFundTurnoverandActiveShare.
Average equally weighted returns for quintile portfolios obtained
by sorting funds either
onActiveShareoronTurnoverarepresentedinfirstrowandcolumnrespectively.Averagereturnsfor25portfoliosobtainedbysorting
funds intoquintiles
firstonTurnoverandthenonActiveSharearealsoreported.PanelAreportstheabnormalreturnsconditionalon
fundsselfreportedturnoverratioandActiveShare.PanelBpresentsreturnsconditionalonfundsaveragequarterlyturnoverratiocalculatedfromchangesinfundsholdingsoverlast4quartersusingthemethodologyinGaspar,MassaandMatos(2005).Allthereturnsareinyearlypercentages.Significanceatthe5%levelisdenotedinbold,andtstatisticsaregiveninparentheses.PanelA
EqualWeighted5FactorAlphaofBenchmarkAdjustedNetReturns
ActiveShare
FundTurnoverRatio Uncond. 1 2 3 4 5 51
Uncond. 0.98 1.45 1.44 0.84 0.25 1.22 (4.49) (3.97) (3.64)
(1.74) (0.45) (2.47)
1 0.38 0.42 0.98 0.91 0.47 0.90 1.32 (1.05) (1.77) (2.82) (1.95)
(0.91) (1.15) (1.63)2 0.47 0.61 1.31 0.86 0.62 1.07 1.68 (1.40)
(2.44) (3.22) (2.05) (1.15) (1.82) (3.10)3 0.61 0.91 1.25 1.50 0.01
0.62 1.52 (1.62) (3.10) (2.86) (2.90) (0.02) (0.94) (2.50)4 1.43
1.85 1.57 2.20 1.17 0.35 1.50 (3.86) (6.05) (3.58) (1.15) (1.99)
(0.54) (2.24)5 1.58 1.48 1.40 2.08 1.28 1.66 0.17 (2.81) (3.41)
(2.18) (3.30) (1.60) (1.92) (0.22)
51 1.20 1.06 0.42 1.17 0.81 2.56 (2.34) (2.44) (0.70) (1.73)
(1.00) (2.75)
-
PanelB
EqualWeighted5FactorAlphaofBenchmarkAdjustedNetReturns
ActiveShare
FundHoldingsTurnover Uncond. 1 2 3 4 5 51Uncond. 0.98 1.45 1.44
0.84 0.25 1.22
(4.49) (3.97) (3.64) (1.74) (0.45) (2.47) 1 0.14 0.24 0.73 1.01
0.14 1.44 1.68 (0.39) (1.10) (2.19) (2.14) (0.27) (1.91) (2.28)2
0.51 0.82 0.97 0.98 0.63 0.83 1.66 (1.49) (2.88) (2.46) (2.04)
(1.33) (1.42) (2.77)3 0.64 1.28 1.37 1.44 0.12 1.02 2.30 (1.61)
(4.18) (3.00) (3.00) (0.20) (1.45) (3.55)4 1.21 1.38 1.57 1.82 1.08
0.20 1.18 (2.89) (4.32) (3.23) (1.33) (1.89) (0.25) (1.50)5 1.97
1.74 1.86 2.21 2.05 1.99 0.25 (4.29) (4.77) (3.63) (3.85) (2.64)
(2.68) (0.35)
51 1.84 1.50 1.14 1.19 1.91 3.43 (3.82) (4.28) (2.34) (1.94)
(2.34) (3.52)
-
Table6.MutualFundActiveShareandFundDuration:RegressionEvidenceThis