ContentslistsavailableatScienceDirect
Journal of Financial Stability
journalhomepage:www.elsevier.com/locate/jfstabil
Litigation and mutual-fund runs
夽Meijun Qian
a, Bas¸ak Tanyeri
b,∗aAustralianNationalUniversity,WhartonFinancialInstitutionsCenter,Australia
bBilkentUniversity,TurkeyandMITGolubCenterforFinanceandPolicy,USA
a r t i c l e i n f o
Articlehistory:
Received19February2015
Receivedinrevisedform16May2017 Accepted19May2017
Availableonline8June2017
JELclassification:
G23 G14 Keywords:
Mutual-fundflows Litigation Returns
a b s t r a c t
Weinvestigatewhetheranticipationofadverseevents(litigationaboutmarkettimingandlatetrad- ing)maytriggermutual-fundruns.Wefindthatrunsstartasearlyasthreemonthspriortolitigation announcements.Pre-litigationrunsaccumulateto31basispointsofthetotalnetassetsoverathree- monthwindow;post-litigationrunsmaylastmorethansixmonthsandaccumulateto1.25percentover thefirstthree-monthwindow.Additionally,investorswhorunbeforelitigationannouncementsearn significantlyhigherrisk-adjustedandpeer-adjustedreturnsthanthosewhorunafterlitigation.Thedif- ferenceinreturnsisparticularlypronouncedforfundsholdingilliquidassets.Finally,securitiesheldby litigatedfundfamiliessignificantlyunderperformvis-á-visothersecuritiesintermsoflowerabnormal returnsandliquidity.Ouranalysissuggeststhatapro-rataownershipdesignisinsufficienttoprevent mutual-fundruns.
©2017ElsevierB.V.Allrightsreserved.
Introduction
Thefirst-come–first-servedprinciplegoverningdepositwith- drawalsatparmotivatesbankruns:depositorswanttowithdraw beforeothersbecausethoseatthebackofthelinemaynotrecover theirdeposits(DiamondandDybvig,1983;ChariandJagannathan, 1988;Zhu,2005;ChenandHasan,2006,2008;DwyerandSamartin, 2009;Aldasoroand Faia, 2016).Incontrast, mutual fundsallo- catetheproceedsfromassetsalesonapro-rata basis,a design thatshouldshieldthemfromruns.However,mutualfundsmay besusceptibletorunswhenadverseinformationaboutthequal- ityofmanagement oraboutunderlyingassets isrevealed,even thoughthereisnophysicalqueueofcustomerswaitingtowith- draw.Thispaperprovidesdirectevidenceofmutual-fundrunsboth beforeandafterrevelationofanadverseeventandinvestigatesthe motivationsbehindmutual-fundruns.
Wedefineafundrunasanabnormallyconcertedredemption of mutual-fundshares in anticipationof, or afterrevelation of,
夽 TheauthorswouldliketothanktheeditorIftekharHassan,theanonymous referees,EdwardKane,WayneFerson,PhilStrahan,ItayGoldstein,anddiscussants attheFRCG2016,FMA2012,EFA2010,EFMA2008meetingsfortheirvaluable comments.
∗ Correspondingauthorat:FacultyofBusinessAdministration,BilkentUniversity.
MITGolubCenterforFinanceandPolicy,Turkey.
E-mailaddresses:meijun.qian@anu.edu.au(M.Qian),basak@bilkent.edu.tr (B.Tanyeri).
anadverseevent.Theadverseeventswe focusonarethe2003 and 2004litigationsalleging that certainmutual fundspermit- tedsomeinvestorstoengagein latetradingormarkettiming,1 therebyallowingpreferentiallytreatedinvestorstoenjoyprofits attheexpenseofthoseinvestorswhodonotengageinsuchprac- tices.Whenshareholderssuspectorlearnthatfundmanagersdo notservetheinterestsofallinvestorsequally,thedisadvantaged investors may discipline the implicated funds by withdrawing existinginvestmentsand/orwithholdingnewinvestments.
Threereasonsmightmotivatefundrunsaroundlitigation.First, prior to litigation some investors may become aware of grey- area tradingpractices via the media, hence lose confidence in thequalityoffundmanagement andvotewiththeirfeet.These investors mayanticipate a possible futureindictment and thus decidetoexitbeforeitoccurs,creatingafirstwaveoffundruns.
Second,afterlitigationannouncements,investorspenalizeman- agement bywithdrawing theirinvestmentsand/orwithholding newinvestments,creatingasecondwaveofredemptions.Finally, post-litigation,funds may beforced tofire sale the underlying assetstomeettheconcertedwithdrawalsor,evenworse,tocom- pletelyliquidatetheportfolios.Theprospectofafiresalemotivates investorstowithdrawearlytoavoidhavingtoredeemsharesat
1Latetradingisthepurchaseorsaleofmutual-fundsharesafter4:00PM,thetime whenthenetassetvalue(NAV)isdetermined.Markettimingisshort-termtradingof mutual-fundsharestoexploitpriceinefficienciesbetweenthemutual-fundshares andtheunderlyingsecuritiesinthefundportfolios.
http://dx.doi.org/10.1016/j.jfs.2017.05.011 1572-3089/©2017ElsevierB.V.Allrightsreserved.
undesirabletimes.Suchastrategiccomplementarityamplifiesfund runs(Chenetal.,2010).
Inthispaper,wedocumentabnormaloutflowsbothbeforeand afterthelitigation,andinvestigatethemotivesforfundruns.To theextentthatinvestorsactoninformationleakedviathemedia, thereshouldbeabnormal fundoutflowswhen suchstories are published.Furthermore,ifinvestors runfundstopenalizeman- agement,thesizeoftherunsshouldbelargerforfundsthatsuffer frompoorreputation.Finally,ifinvestorsareconcernedaboutfire- salecosts,fundrunsshouldbemorecommonamongilliquidfunds thanamongliquidfunds.Moreover,ifthestrategiccomplemen- tarityhypothesisholds,investorswillbenefitbyredeemingshares beforesuchadverseinformationbecomespublic.Thatis,pre-event runsshouldearnrelativelyhigherreturnsthanpost-eventruns.
Concertedredemptionsandthelackofnewsalesfollowinglitiga- tionannouncementswillforcefundstoquicklyliquidateassets,and thislargetradingvolumemaytemporarilydepresstheunderlying assetprices.Becauseshareholderswhoredeemsharesatthistime willincurlosses,investorswhocananticipatelitigationsandsub- sequentredemptionshaveanincentivetoredeemsharesearly.By exitingearly,informedinvestorsavoidthefire-salecostscreated bythesubsequentconcertedwithdrawals.Therefore,theincen- tivesforearlyrunswillbegreaterforfundsinwhichthereturn differencesfromthetimingofthewithdrawalsarelarger,either becausetheyholdilliquidassetsorbecausetheirpoorreputations willdrivelargeoutflows.
Ourpaperempiricallyaddressesthefollowingquestions:First, dorunsoccurbothbefore(pre-event)andafter(post-event)litiga- tion?Second,doinvestorswhorunfundspriortolitigationavoid thecoststhatinvestorswhorunpostsufferfrom?Third,arefund runslargerforfundsthathavepoorerreputation?Fourth,arethe fire-salecostslargerforfundswithilliquidassets?
Wefindthatfundrunstakeplacebothbeforeandafterthelit- igation.Pre-eventrunsbeginasearlyasthreemonthsbeforethe litigation.First,inthethreemonthsbeforethelitigation,abnormal monthlyoutflowsfromtheinvestigatedfundsare31basispoints ofthetotalnetassets(TNA),andinthethreemonthsfollowingthe litigation,abnormaloutflowsare1.25percent.Second,investors whorunbeforethelitigationearnsignificantlyhigherrisk-adjusted andpeer-adjustedreturns(asmuchas6basispoints)thanthose whorunafterthelitigation.Thisdifferenceinreturnsismorepro- nouncedforlitigatedfundsholdingilliquidassets.Third,fundsare notequallyvulnerabletoruns.Fundsholdingilliquidassetsand stand-alonefundsexperiencelargeroutflows.
Ourresultsindicatethatmutual-fundinvestorswhoanticipate outflowsfollowing litigationnewshave incentives towithdraw earlytoavoid fire-salecosts.Weexaminetheabnormalreturns andtheliquidityoftheunderlyingsecuritiesheldbytheinvesti- gatedfundsaroundthetimeoflitigation.Illiquidstocksheldbythe investigatedfundssignificantlyunderperform(intermsofcumu- lativeabnormalreturnsfollowinglitigation)vis-à-visotherstocks.
Furthermore,thebid-askspreadofilliquidstocksheldbytheinves- tigatedfundsalsoincreases.
Whenthetimingoftheaction(arun)mattersforthepayoff(the return),strategiccomplementarities(Bulowetal.,1985)comeinto play,whichamplifytheimpactofadverseeventsonthefundamen- talsandgeneratefinancialfragility.Nonetheless,mutual-fundruns maynotoccurunlessthereisasystemicliquidityshocktoallfund investors(Chenetal.,2010).Intheabsenceofsuchashock,other investorswillpurchasetheassetsatfiresalepricesandthuscor- rectthemispricing.2Consequently,althoughthereisafundrun,
2 Chenetal.(2008)showthathedgefundsthatpurchasetheunderlyingassetsof mutualfundsatdepressedpricesduringfiresalesgeneratearbitrageprofitswitha similarmagnitudetotheprofitsofshortsellers.
investorswhoarenotmotivatedtoexitwillholdontotheirshares andsurvivetopricerecovery.Asaresult,mostoftheindictedfunds survive.Thedatarevealasurvivalrateof80percentfortheinves- tigatedfundsduringthe2003–2007period,whichissimilartothe averagemutual-fundsurvivalrateduringthesameperiod.
Thefinancial fragility of themutual-fundindustry is under- scoredbytheU.S. Treasury’sdecision toinsuretheholdings of eligiblemoney-marketmutualfundsinthewakeoftheturmoil causedbytherunontheReservePrimaryFundinSeptember2008.3 Ourfindingsnotonlyexplainwhymutualfundrunsmayoccur,but alsoenlightenonhowtheeventsthatledtothedemiseofReserve PrimaryFundreflectthefragilityoftheindustry.4
Despite the significance of fund market fragility, there is a scarcityresearchonmutual-fundruns.StrahanandTanyeri(2015) examinewhetherthemoney-marketfundsthatwerehitwiththe largestoutflowsfollowingtheReservePrimaryFundbreakingthe buckchangedtheirportfolioriskprofiles.Chenetal.(2010)inves- tigatethepayoffcomplementarityby analysingtheflow-return sensitivityofilliquidmutualfunds.Ourpaperdirectlydocuments runsandsilentrunsinmutualfundsbystudyingtheSpitzerinves- tigationsof2003and2004.Examiningaperiodpriortothe2008 crisisisimportantbecauseitshowcasesthevulnerabilityofmutual fundspriortothecrisis.Moreimportantly,weextendthefind- ingsofChenetal.(2010)bylookingataspecificadverseevent andprovidedirectevidenceofpayoffcomplementarityinmutual funds.
Theremainder of this article proceedsasfollows: Section 2 developsthemethodology;Section3describesthedata;Section 4outlinestheempiricalresults;andSection5presentsourconclu- sions.
2. Methodology
We address four research questions. First, do investors run investigated funds both before and afterlitigation? Second, do investorswho runfunds beforelitigationrealize higherreturns thanthosewhorunfundsafterlitigation?Third,aresometypes offundsmoresusceptibletorunsthanothers?Finally,arestocks thatareintheportfoliosoflitigatedfundsaffectedbytheilliquidity ofthemutualfundsthatholdthem?
2.1. Detectingpre-eventruns
Todocumentpre-eventruns,weneedbenchmarksofnormal flow,thefirstofwhichareflowstopeersnotnamedinthe2003 and2004lawsuits.Weconstructthreegroupsoffunds.Thefirst treatmentgroupcomprisesfundsnamedinthelitigations(inves- tigatedfunds);thesecondtreatmentgroupcomprisesfundsthat arenotnamedinthelitigationsperse,butaremanagedbycom- paniesthatarenamedforotherfunds(otherfundsinfamilieswith investigatedfunds);thecontrolgroupcomprisesfundsmanagedby
3TheU.S.Treasuryexpresseditsconcernsaboutuncertaintiesinthemutual- fundindustryandjustifieditsimplementationofaguaranteeprogramasfollows:
“Maintainingconfidenceinthemoney-marketfundindustryiscriticaltoprotecting theintegrityandstabilityoftheglobalfinancialsystem....Thisactionshouldenhance marketconfidenceandalleviateinvestors’concernsabouttheabilityformoney marketmutualfundstoabsorbaloss....”(U.S.TreasuryDepartmentPressRelease, September19,2008).
4SincethefailureofLehmanBrothers,therehasbeenagrowingbodyoflitera- turestudyingthecausesandconsequencesofrunsinthemoney-marketindustry (McCabe,2010;KacperczykandSchnabl,2010,2013;ChernenkoandSunderam, 2014;Duca,2013;StrahanandTanyeri,2015).
companiesthatarenotinvolvedinlitigation(fundsinfamilieswith noinvestigatedfunds).5Wethencomputethefundflowsasfollows:
Flowi,t = [TNAi,t−TNAi,t-1∗(1+ri,t)]/TNAi,t-1, (1) whereFlowi,tisthenetflowsoffundiinmontht,TNAi,t-1andTNAi,t arethetotalnetassetsoffundiinmontht-1andt,respectively, andri,tisthereturnoffundiinmontht.Todetectwhetherthe implicatedfundshavelowerflowsthanthenon-implicatedfunds, wecomparethenetflowsofthethreegroupsbeforeandafterthe litigationdates
Thesecondbenchmarkfornormalflowsistheestimatednet flowsfroma modeldesignedtocapturethemaindeterminants of fundflows. We develop a model that includes variables for fundcharacteristics,pastreturns,andindustry-levelandstyle-level flows.Previousstudiesshowthatpastreturnswillpredictfuture flows(Gruber,1996;ChevalierandEllison,1997;SirriandTufano, 1998;DelGuercioandTkac,2002,2008)andthatindustry-level andstyle-levelflowsalsoexplainindividualfund-levelflows(Qian, 2011).Todetectpre-litigation and post-litigationruns,wealso constructevent-windowindicatorsandtestthefollowingmodel:
Flowi,t = a+
bj∗fundcharacteristicsi,tj +
cj∗pastreturnsij+
dj∗aggregateflowstj+
j∗Event-windowindicatorsi,tj+εi,t, (2)
wherefundcharacteristicsaresize,thelogofTNA;age,thelog ofdays sincethefirstofferdate; expenseratio,a fund’soperat- ingexpensesasaratioofthetotalinvestment;andmanagement fee, the management fee as a ratio of the average TNA. Past returnsincludecompoundedreturnsduringthepastone(Ri,t-1), three (
s(1+ Ri,t-s)-1, s=1,2,3),and six months(
s(1+Ri,t-s)- 1,s=1,2,–-,6). Aggregateflowsare industry-leveland style-level flows:industry-levelflowsarethesumofflowsindollars(i(TNAi,t
−TNAi,t-1*(1+Ri,t)))toallfundsinthesampledividedbythesum ofthelaggedTNA(i(TNAi,t-1)),andstyle-levelflowsarethesumof flowsindollarstoallfundswiththesameinvestmentstyledivided bythesumofthelaggedTNA.Wedefineevent-windowindicators infourways.First,webenchmarkthefirstdatethatthefundliti- gationismentionedinthenewspaperstodefinethreeindicators:
pre-litigation(−3,−1)forthethreemonthsbeforethelitigation newsdate;newsmonth(0)forthemonthofthelitigationnews;
andpost-litigation(+1,+3)forthethreemonthsafterthelitiga- tionnews.Second,weidentifythedateofthelitigationfilingto definethethreeindicators:thepre-litigationfiling(−3,−1),the filingmonth(0),andthepost-litigationfiling(+1,+3).Third,we useboththelitigationnewsandthefilingdatestodefinethethree indicators: pre-litigationincludes themonthsbetweenthefirst newspaperarticleandthelitigationfilingmonth,thefilingmonth (0),andthepost-litigationfiling(+1,+3).Finally,thirteenindicators equal1ifitisthenthmonthfromthedateofthelitigationnews, otherwiseitis0(n=−6...−1,0,1...6).
2.2. Therationaleforpre-eventruns
Whatincentives existfor shareholderstorunamutual fund whenproceedsfromassetsalesaredeterminedbythepricesof theunderlyingassetsandaredistributedpro-rata?First,litigation
5WhenthefirstnewspaperarticleappearedonSeptember3,2003itwasnot clearwhichfundswouldbenamedinthelitigation.Hence,theremayhavealso beenoutflowsfromthecontrolgroupduetothepossibilityoffutureinvolvement.
Assuch,wemayunderestimatethemagnitudeoftheoutflowsobservedinthe investigatedfundsrelativetotheoutflowsofthecontrolgroupfunds.
mayindicatehowfaithfullyfundmanagersareservingtheinter- estsoftheinvestors.Hence,investorsmayredeemsharesassoon astheyareinformed,eitherpubliclyorprivately,thatthefundsare engaginginabusivepractices,suchasmarkettimingorlatetrading.
Whenasufficientnumberofinvestorslearnofthefund’sabusive behavior,arunmayensue.Shareholderswhoredeemsharesatthis pointconsequentlywillrealizenegativeabnormalreturnsbecause themutualfundsmustquicklyliquidateassetstosatisfytheshare redemptions,andthelargesellingvolumewilltemporarilydepress theunderlyingassetprices.
We benchmarknormal returnsusingfivereturnmodelsand introduceindicators(asdefinedintheflowmodels)forthepre- andpost-eventmonthstoidentifythereturndifferencesbetween investors who withdraw beforeand after litigation. These five return models are the market model (Sharpe, 1964; Lintner, 1965), the market model withlagged market returns (Scholes andWilliams,1977),the3-factorFama-Frenchmodel(Famaand French,1992,1993),theFama-Frenchmodelwithafourthfactor thatcapturesmomentum(JegadeeshandTitman,1993;Carhart, 1997),andthemarketmodelwithafactorthatcapturesliquidity (PástorandStambaugh,2003):
ri,t =˛+ˇ∗rm.t +
˛n∗event-windowindicatorn+εi,t, (3)
ri,t =˛+ˇ1∗rm.t+ˇ2∗rm.t–1 +
˛n∗event-windowindicatorn+εi,t, (4)
ri,t =˛+
ˇj∗FFtj+
˛n∗event-windowindicatorn+εi,t, (5)
ri,t =˛+
ˇj∗FFtj+1∗MOMt +
˛n∗event-windowindicatorn+εi,t, (6)
ri,t =˛+ˇ∗rm.t+2∗LIQt
+
˛n∗event-windowindicatorn+εi,t. (7) whereri,tistheexcessreturns(netoftherisk-freerate)offundiin montht,andrm.tistheexcessmarketreturninmontht.FFjincludes marketreturns,size(SMB),andvalue(HML)factors;MOMisthe momentumfactor;andLIQistheliquidityfactor.
WeestimatetheflowandreturnofEqs.(2)through(7)intwo ways. First, werun pooledregressions usingthefull sampleof boththeinvestigatedfundsandthenon-investigatedfunds.Pooled regressionsincludefundandyear-monthfixedeffects.Second,we runfund-levelregressionsusingonlythesubsampleoflitigated funds.Thepooledregressionsareefficientinthattheypoolinfor- mationfromallofthefunds;however,theymayalsobeinefficient due to thefact that all fundcoefficients must be thesame. In fund-by-fundestimations,fundcoefficientsmayvarybuttheyare restrictedtoinformationonsinglefunds.
Wealsousedailyreturnstoexaminethereturnimpactofthe withdrawal.Werunpooledregressionsusingthefiveassetpricing models,Eqs.(3)–(7),atthedailyfundreturnlevelandincluding bothlitigatedfundsandnon-litigatedfunds.ri,tistheexcessreturns
(netoftherisk-freerate)offundiondayt,andrm.tistheexcessmar- ketreturnondayt.Thepre-eventindicatorindicates[−20,−1]days beforethelitigationannouncement,andthepost-eventindicator indicates[1,20]daysaftertheactuallitigation.Theeventindicator indicatesthedaysbetweentheannouncementandthelitigation.
Observationsofthelitigatedfundscoverthesethreeeventwin- dows.Observationsofthenon-litigatedfundscoverthemaximum numberofcalendardaysofobservationsofalllitigatedfunds.
2.3. Theimpactoffundcharacteristicsandliquidityonfundflows andreturns
Pre-eventrunsaremotivatedbyworriesaboutpotentialliti- gationandanticipationoftheliquidationcoststhatwillariseto satisfythepost-litigationredemptions.Hence,investordecisions torunareinfluencedbytheirbeliefaboutorawarenessofabusive behaviourand byfactorsthatwillincrease fire-salecosts.Wor- riesaboutpoorbehaviourareaffectedbythereputationofthe fundmanagement companyaswell asby theinvestors’ ability tocollectandprocessinformation.Wemeasurethefundreputa- tionusingtheownershipstructureanditshistoryofSECcharges.
Investorsmayassumethatfundsinconglomeratefamilieswillbe lesslikelytoengageindishonestbehaviourbecausealossofrepu- tationwillhurtboththefundinquestionaswellasotherbusinesses intheconglomerate.Hence,illicitbehaviourmayhavemoreseri- ousconsequencesforconglomerates thanforfundfamilies that onlymanagemutualfunds.Moreover,conglomerates,especially thosewithcommercialbanksubsidiaries,havealargercapacityto copewithliquidityshocksthanstand-alonefunds.Therefore,they willbelesssusceptibletorunsmotivatedbyattemptstoavoidfire sales.Likewise,pastactionsmaypredictfuturedecisions.Investors mayassumethatfundswithnohistoryofaberrantbehaviourinthe pastwillbelesslikelytobehavepoorlyinthefuture.
Toinvestigatewhetherthecharacteristicsoffundsandinvestors influencethesusceptibilityoffundstopre-eventruns,wegenerate indicatorvariablesforthefollowingcharacteristics.Theconglom- erateandchargehistoryindicatorsequal1ifthefundispartofa conglomerateand ifit hadbeensubjecttoanSECinvestigation withinthepasteightyears,respectively;otherwisetheyareequal to0.
Theeconomicrationaleforpre-eventrunsistheliquidationcost (thepricedepression)thatfundswillbearwhentheyareforcedto sellassetsuponrevelationofanadverseevent.Thisliquiditycost increaseswiththeilliquidityoftheunderlyingassetsaswellaswith thevolumeoftheredemptions.Henceinvestorsinfundswithilliq- uidassetshavestrongerincentivestorunbecausetheymayreap greaterbenefits.Wethereforeinvestigatetheimpactofunderlying assetliquidityontheincentivestorunandthebenefitsofrun- ningearlybygeneratinganindicatorvariable(illiquid)forilliquid funds.Wecategorizefundsasilliquidbasedontheassetsinwhich theyinvestusingLipperobjectivecodes. Illiquidfundsinvestin small-capstocks,internationalequityandbonds,andasset-backed securities(Lipperobjectivecodes:corporatebondslow,derivative mortgages,growthorsmallormidcapequity,internationalbonds, andinternationalequity),whereasliquidfundsinvestinlarge-cap stocksandtreasurybills.6Asecondliquidityclassificationusesthe cashholdingsofthefunds.Whenwelimitthesampletodomestic equityfunds,wecanalsoapplyathirdliquidityclassificationthat computestheweightedaverageofthebid-askspreadoftheunder- lyingstocksasofJune30,2003,(threemonthspriortothefirst
6 Lanetal.(2015),inaclassificationofmutualfundsaccordingtotheirinvest- menthorizons,findthatfundswithlong-terminvestmenthorizonsinvestinilliquid stocks.Shareredemptionsforfundswithlong-terminvestmenthorizonsandilliquid stockholdingsmayhavelargerimpactsonprice.
litigation).Thesetwoalternativeliquiditymeasuresclassifyfunds withabovethesamplemedianweightedaveragebid-askspreador belowthesamplemediancash-holdingratioasilliquidfunds.
Werunfund-by-fundflowsandreturnregressionsinthesub- samplesofthefundsgroupedperSEClitigationhistory,whetherthe parentisaconglomerate,andtheilliquidityofthefundportfolio.7 Thefirststepestimatestheflowmodelsandthereturnmodels foreachfundusingtime-seriesobservations.Thecontrolvariables fortheflowmodelsincludetheaccumulatedreturnsinthepast one,three,andsixmonths,andtheindustry-levelandstyle-level flows,managementfees,expenseratio,size,age,andyearindi- cators.The control variables for the return modelsinclude the yearindicatorsandtherelevantreturnfactorsinthefivereturn models.Theexplanatoryvariablesincludethethreeindicatorsfor theseven months surroundingthe litigation (i.e.,three indica- torscoveringthelitigationmonth,threemonthsbeforeandthree monthsafter).Thesecondstepcomparestheestimatedflowsand therisk-adjustedreturnsinthecrosssection.Weinvestigatethe cross-sectionaldifferencesaccordingtofunds’SECchargehistory, ownershipstructure,andtheliquidityoftheirunderlyingassets.
2.4. Returnsandliquidityoftheunderlyingassets
Weanalysethereturnstotheunderlyingstocksintheportfolios oflitigatedfundstodirectlytestwhethermutualfundsbearthe costsassociatedwithliquidatingportfoliopositions.Wecompile theholdingsofthelitigatedandthenon-litigatedfundssurround- ingalltheeventmonths,fromSeptember2003toMarch2004,and AugustandNovember2004.Foreachlitigationevent,wecompute thecumulativeabnormalreturns(CARs)inthe[−3,3],[−3,10], [−3,20],[−3,40],and[−3,60]daysaroundthelitigationfilingfor allstockslistedontheCRSP.
The WRDS event study module estimates the market model—whichusestheCRSPequallyweightedportfolioasthemar- ketportfolio—usingthedailyreturnsfrom282daysto30daysprior tothelitigationannouncement.Foreachstock,wealsoconstruct,% heldbylitigatedfunds,whichequalsthesharesofstockheldbythe litigatedfunds(inthatlitigationevent)overthetotaloutstanding shares.Theilliquidindicatorequals1ifthestocks’bid-askspreadis abovethemedianbid-askspreadofallthesamplestocks.Finally, weexaminehowtheinteractionbetweenlitigatedfundholdingsof thestockandilliquidity(interactionequallingthepercentageheld bythelitigatedfundsandtheilliquidityindicator)affectCARs.
Ifthefire-salepropositionholds,theliquidityoftheunderly- ingassetsmayalsobeaffected.Weuseadifference-in-difference approachtoinvestigatethiseffect.Wecomputetheaveragebid- askspreadofeachstockforeacheventmonthaswellasforJune 2003.Wethenregressthedifferenceofthebid-askspreadbetween theeventmonthandJune2003on%heldbylitigatedfunds,illiquid indicator,andinteraction.
3. Data
Toidentifythefundsandthefundfamiliesnamedinthemarket- timingandlate-tradinglitigations,weconductakeywordsearchin theFinancialTimesandtheWallStreetJournal.8Wealsosearchthe SEClitigationfilingsonEDGARandintheStanfordLawSchoolSecu- ritiesClassActionClearinghouse.9Table1summarizestheresults
7Theresultsofthepooledpanelregressionsofflowsandreturns,availableupon request,arequalitativelythesame.
8Weusethreekeywords—investigation,mutualfund,andSpitzer—tosearchthe FinancialTimesandtheWallStreetJournalbetweenSeptember3,2003andDecember 31,2005.
9TheStanfordLawSchoolSecuritiesClassActionClearinghouse(availableonline athttp://securities.stanford.edu/index.html)compilesdetailedinformationonthe
Table1
Listoffundfamiliesinvolvedintradingscandals.Table1liststhefundfamiliesnamedinlitigationsonmarkettimingandlatetrading.Hedgefunds,brokeragefirms,and investmentbankingservicesareexcluded.Litigationdataishand-collectedfromtheStanfordLawSchoolSecuritiesClearinghouseandSEClitigationnews.MTstandsfor markettimingandLTstandsforlatetrading.
Fundfamily Newspaperdate Litigationdate Settlement(inmillion$) Practiceunderinvestigation
NationsFundsTrustFamily 3-Sep-03 5-Sep-03 535 MT<
OneGroupFamily 3-Sep-03 9-Sep-03 90 MT<
JanusFamily 3-Sep-03 5-Sep-03 226 MT<
StrongFamily 3-Sep-03 5-Sep-03 175 MT
INVESCOFamily 3-Sep-03 31-Oct-03 415 MT<
PutnamFamily 3-Sep-03 21-Oct-03 194 MT
MFSFamily 8-Sep-03 11-Dec-03 350 MT<
AllianceBernsteinFamily 8-Sep-03 2-Oct-03 250 MT<
FederatedFamily 9-Sep-03 24-Oct-03 100 MT<
FranklinFamily 8-Oct-03 6-Feb-04 49 MT
AlgerFundsFamily 16-Oct-03 31-Oct-03 45 MT
SalomonSmithBarneyFamily 22-Oct-03 9-Aug-04 MT
ScudderFamily 5-Nov-03 22-Jan-04 208 MT
PBHGFamily 13-Nov-03 14-Nov-03 90 MT
ExcelsiorFamily 14-Nov-03 20-Nov-03 1 MT
ColumbiaFamily 14-Nov-03 13-Feb-04 460 MT
FremontFamily 21-Nov-03 12-Mar-04 4 MT
PIMCOFamily 13-Feb-04 20-Feb-04 90 MT
RSGrowthandValue 3-Mar-04 12-Nov-04 30 MT
AmericanFamily 24-Mar-04 24-Mar-04 MT
Sources:MoneyManagementExecutiveCompilation,January31,2004.
FundScandalScorecard,WallStreetJournal,April27,2004.
SECPressReleases,September2003–December2004.
FinancialTimes,2003–2005.
StanfordLawSchoolLibrarySecuritiesClassActionClearinghouse.
ofthissearch—includingthenamesofthefundfamilies;theactiv- itiesforwhichtheyareinvestigated;thelitigationannouncement datesandthenewspaperannouncementdates;andthesettlement inmilliondollars.WealsousetheStanfordClearinghousedatabase toidentifythefundswithineachfundfamilythatareexplicitly namedinthelitigation.
Aformalinvestigationintothetradingpracticesofmutual-fund companiesbeganinSeptember2003,whenNew YorkAttorney GeneralEliotSpitzerfiledacomplaintintheNewYorkSupreme Courtallegingthatthemutual-fundcompaniesofBankofAmer- ica,BankOne,JanusCapitalGroup,andStrongCapitalManagement hadallowedcertainhedgefundmanagerstoillegallytradeintheir fundunits.Subsequently,betweenSeptember2003and August 2004,theSEC,theNewYorkStateAttorneyGeneral,andotherreg- ulatoryauthoritiesfiledlitigationinvolvingfundsintwenty-five mutual-fundfamilies.10
Newsarticlesonabusivetradingpracticesbymutualfundspre- datethefirstlitigationannouncementinSeptember2003.Infact, theSECwasawareoffairpricingproblemsinmutualfundsasfar backas1997,andaprobeintohedgefundtradesthattookadvan- tageofsuchproblemshadbeenunderwaysince 2002.Thefirst articleindicatingpossibleactiveinvolvementbymutual-fundman- agementisdatedMarch5,2003,andonMarch26,2003Congress beganconsideringoptionstostrengthenmutual-fundregulation.It ishighlyprobablethatbyMarch2003investorshadalreadybegun tosuspectabusivebehaviourandthepossibilityofacriminalinves- tigation.Thepre-eventindicatorsintheflowmodelsmaycapture whetherinvestors whosuspectedaberrantbehaviourredeemed theirsharespriortothelitigationannouncements.
Fortheuniverseofmutualfunds,werelyontheCRSPmutual- funddatabase(from WRDS),which providesmonthlyanddaily
prosecution,defense,andsettlementoffederalclass-actionsecuritiesfraudlitiga- tion.
10Wearenotabletoidentifythenamesoftheindividualfundsinvestigatedinfive ofthemutual-fundfamilies.Therefore,Table1andthesampleonlyincludetwenty mutual-fundfamilies.
observationsofthetotalnetassets(TNA)andreturns(R)offunds.11 ThesamplecoversthemonthsfromJanuary1999toDecember 2007.WemergethelistoflitigatedfundswiththeCRSPuniverseof fundstoproduceasampleinwhichtheinvestigatedfundsaredif- ferentiatedfromthenon-investigatedfunds.Weexcludeallfunds withmissingtickerandmanagementcodesymbols,fundsintheir incubationperiod,fundswithlessthansixmonthsofobservations aroundthelitigationdate,andfundswithaTNAoflessthan5mil- lionUSD.WealsodropfundswithoutflowsgreaterthantheTNA andwithinflowsgreaterthanfivetimestheTNA.
PanelsA,B,andCofTable2provideasnapshotofthesam- plefundsthree monthsbefore, duringthemonth of,and three monthsafterthe firstlitigationannouncement. Specifically,the panelsshowthenumberoffunds,themean,andtheaggregateTNA inthesubsampleoftheinvestigatedfunds,ofotherfundsinfamilies withinvestigatedfunds,andoffundsinfamilieswithnoinvesti- gatedfunds.First,theaverageflowofinvestigatedfundsdecreased from1.1percentinJune2003to−0.2percentinSeptember2003.
Similarly,theaverageflowofotherfundsintheinvestigatedfam- iliesdecreasedfrom0.5percentto−0.5percent.Furthermore,the negativeflowspersistedinDecember2003forbothgroups.The descriptivestatisticsindicatethattheinvestigatedfundssuffered outflowsintheeventmonthandbeyond.Theaverageinflowfor thecontrolgroup(fundsinfamilieswithnoinvestigatedfunds) droppedfrom1percentinJune2003to0.7percentinSeptember 2003,butitrecoveredto1.1percentinDecember2003.
The CRSP also providesinformation onfund characteristics, including the expense structure (management fees and expense ratio),investmentstyle,age(age),andthecashholdingsofeach fundrelativetotheTNA.Wealsohand-collectdataonthefund
11Samplingdailyflowswouldbeidealtodocumentfundruns.However,dailyflow dataareinferredfromthedailyfundsurveyoftheTNA.Thedailynetinflowondate tisthedifferencebetweenthefundTNAattheendofdaytandthefundTNAat theendofdayt–1,adjustedbythereturnondayt.Inthesurvey,somefundsreport theTNAoftandothersreporttheTNAoft–1onsurveydayt.Thesamefundmight reportaTNAoftonsomedaysandaTNAoft–1onotherdays.Thereisnowayto tracewhatandwhenfundsreport.Assuch,thedailyflowdataarenotreliable.
Table2
Summarystatistics:Overviewofsamplefunds.PanelsA,B,andCprovidesnapshotsofthreegroupsoffunds:(1)investigatedfunds,(2)otherfundsinfamilieswith investigatedfunds,and(3)fundsinfamiliesthatarenotinvestigatedasofJune,September,andDecember2003.Thetableliststhetotalnumberoffunds,theaverageTNA andtheflowofeachfund,andtheaggregateTNAandtheflowofallfundsinthethreegroups.
TreatmentGroup1:
InvestigatedFunds
TreatmentGroup2:Other FundsinFamilieswith InvestigatedFunds
ControlGroup:Fundsin FamilieswithNo InvestigatedFunds PanelA:SnapshotonJune2003
Total#offunds 1560 1408 4769
AverageTNAofeachfund(million$) 738 313 485
TotalTNA(million$) 1,151,025 440,262 2,315,065
Averageflowofeachfund(%) 1.1 0.5 1.0
Totalflow(million$) 5846 1454 16,735
PanelB:SnapshotonSeptember2003
Total#offunds 1560 1420 4825
AverageTNAofeachfund(million$) 761 317 500
TotalTNA(million$) 1,187,109 449,813 2,412,347
Averageflowofeachfund(%) −0.2 −0.5 0.7
Totalflow(million$) −715 −945 12,347
PanelC:SnapshotonDecember2003
Total#offunds 1561 1460 4882
AverageTNAofeachfund(million$) 823 331 552
TotalTNA(million$) 1,285,480 483,042 2,696,624
Averageflowofeachfund(%) −0.9 −0.2 1.1
Totalflow(million$) −4355 −4003 14,776
characteristics. We use SEC EDGAR filings and firm Web sites todeterminewhethertheparentcompanyisaconglomerateor anassetmanagementcompany,andtheSEClitigationfilingsto checkwhetherthefundshada priorhistory ofSECcharges.To estimatefundperformance,wecompilemonthlydataonmarket returns(rm);therisk-freerate(rf);andthevalue(SMB),size(HML), momentum(MOM),andliquidity(LIQ)factorsusingWRDS’sFama- French,momentum,andliquiditydatabases.
Wecreateacross-sectionaldatasetattheindividualstocklevel coveringallthestockslistedontheCRSP.WeusetheWRDSevent studyapplicationtocalculate7-day(−3,+3),14-day(−3,+10),24- day(−3,+20),44-day(−3,+40),and64-day(−3,+60)cumulative abnormalreturnsaroundeachlitigationfilingforeachstock.We compilethebid-askspread(calculatedatthedailylevelastheask minusthebiddividedbytheaverageofthebidandtheask)for allCRSPstocks.Wecalculatetheaveragedailybid-askspreadin themonthofeachlitigationfiling(September2003throughMarch 2004,August 2004,andNovember2004) andtheaveragedaily bid-askspreadinthemonthofJune 2003(threemonthsbefore thefirstlitigationannouncement).Wedefinetherelativebid-ask spreadasthebid-askspreadofeachstockineacheventmonthrel- ativetoitsbid-askspreadinJune2003(theevent-monthbid-ask spreadminustheJune2003bid-askspread).Weclassifythestocks asilliquidifthebid-askspreadofthestockwaslargerthanthe medianbid-askspreadinJune3003.Wethenmaptheholdingsof themutualfundsontotheCRSPstocks.Wecalculate%heldbylit- igatedfundsasthenumberofsharesheldineacheventmonthby thefundslitigatedinthatmonthdividedbythetotalnumberof sharesoutstanding.
4. Empiricalresults
First,weinvestigatefundflowsbothbeforeandafterthelit- igationnews and the filing dates. Second, we analysewhether investors who run prior to the litigation announcements earn higherrisk-adjustedreturnsthaninvestorswhorunpostthelitiga- tionannouncements.Third,weexaminecross-sectionaldifferences intheflowsoflitigatedfundsaccordingtotheliquidityofthesecu- ritiesin which theyinvest.Fourth, we investigatewhetherthe liquidityandabnormalreturnsofstocksthatthelitigatedfunds holdareaffectedbytheliquiditysqueezeoftheirinvestors.
Fig.1. Time-seriestrendofaveragefundflows.
ThefigureplotsaveragemonthlyflowsfromSeptember2001toSeptember2005in threefundsubsamples:treatmentgrouponeconsistsofinvestigatedfunds,treat- mentgrouptwoisconsistsoffundsinfamilieswithinvestigatedfunds,andthe controlgroupconsistsoffundsinfamilieswithnoinvestigatedfunds.Flowi,tis calculatedas[TNAi,t−TNAi,t-1*(1+Ri,t)]/TNAi,t-1.
4.1. Detectingpre-eventruns
Wedetectpre-eventrunsusingtwobenchmarks:aunivariate analysistobenchmarktheflowsofinvestigatedfundsandother fundsinthesamefamilyagainsttheflowsoffundsinfamilieswith noinvestigatedfunds,andamultivariateanalysistobenchmarkthe flowsofinvestigatedfundsagainsttheflowsestimatedusingthe normal-flowmodel.Theaveragemonthlyflowsofthethreesub- samplefundsfromSeptember2002toSeptember2004areplotted inFig.1.Asthefigureshows,beforeJune2003theflowsofinves- tigatedfundsareeitherhigherthanornotdifferentfromtheflows offundsinfamilieswithnoinvestigatedfunds,butthereafterthey areconsistentlylower.Thatis,flowsshiftedthreemonthsbefore thefirstlitigationfiling,suggestingthatinvestorsranfundsboth beforeandafterthefirstlitigationannouncement.
Table3showstheaveragemonthlyflowsforthetwotreatment groupsandthecontrolgroupfromSeptember2002toSeptember
Table3
Acomparisonofflowsamongfundgroupsovertime.Thistableshowstheaveragemonthlyflowsforthethreefundgroups:investigatedfunds,otherfundsinfamilieswith investigatedfunds,andfundsinfamilieswithnoinvestigatedfundsfromSeptember2002toSeptember2004.Theeventmonthisthefirstmonthinwhichthelitigationwas announced(i.e.,September2003).Fundsinfamilieswithnoinvestigatedfundsareusedasbenchmarkstotestforflowdifferencesagainsttheinvestigatedfundsandother fundsinfamilieswithinvestigatedfunds.**and*denotesignificanceatthe1%and5%levels,respectively.
Monthsfrom September2003
Date TreatmentGroup1:
InvestigatedFunds
TreatmentGroup2:Other FundsinFamilieswith InvestigatedFunds
ControlGroup:Fundsin FamilieswithnoInvestigated Funds
DifferenceinMeansTest
T-test
(1) (2) (3) (1)=(3) (2)=(3)
−12 Sep-02 1.02 0.56 0.38 [2.00]** [0.58]
−11 Oct-02 0.39 −0.08 0.58 [−0.65] [−2.15]*
−10 Nov-02 1.28 0.01 0.71 [1.81]* [−2.54]**
−9 Dec-02 0.31 −0.53 0.16 [0.49] [2.43]**
−8 Jan-03 0.53 −0.15 0.96 [−1.52] [−3.84]**
−7 Feb-03 0.72 0.44 0.37 [1.99]* [0.29]
−6 Mar-03 0.71 0.13 0.91 [−0.61] [−2.29]**
−5 Apr-03 1.46 1.01 1.30 [0.51] [−0.91]
−4 May-03 1.14 0.33 0.94 [0.72] [−2.19]**
−3 Jun-03 1.07 0.50 1.05 [0.07] [−1.52]
−2 Jul-03 0.63 0.10 0.86 [−0.72] [−2.38]**
−1 Aug-03 0.03 0.23 0.89 [−3.07]** [−2.26]**
0 Sep-03 −0.22 −0.53 0.71 [−4.03]** [−5.20]**
1 Oct-03 0.37 0.56 1.31 [−2.66]** [−2.22]**
2 Nov-03 −0.76 0.90 0.95 [−8.39]** [−0.19]
3 Dec-03 −0.86 −0.22 1.11 [−6.56]** [−4.01]**
4 Jan-04 −0.10 0.54 1.75 [−5.97]** [−3.80]**
5 Feb-04 −0.37 0.66 1.32 [−7.09]** [−2.20]**
6 Mar-04 −0.53 0.32 1.01 [−4.35]** [−1.85]*
7 Apr-04 −0.67 0.14 0.72 [−5.70]** [−2.08]*
8 May-04 −1.52 −0.95 0.10 [−7.12]** [−4.42]**
9 Jun-04 −1.11 0.56 0.72 [−6.04]** [−0.44]
10 Jul-04 −0.90 −0.14 0.72 [−5.18]** [−2.54]**
11 Aug-04 −0.57 0.25 0.43 [−3.67]** [−0.61]
12 Sep-04 −0.82 0.40 0.37 [−6.16]** [0.10]
2004.TheresultsinTable3confirmthevisualtrendinFig.1.Up untiltwomonthspriortothefirstlitigationannouncement,the flowstotheinvestigatedfundsarelarger(orsmallerbutinsignif- icant)thantheflowstothefundsinfamilieswithnoinvestigated funds.However,thistrendisreversedpriortoSeptember2003.
Investigatedfundsthatenjoyedlargeflowsuptooneyearpriorto theonsetoflitigationbegantoexperiencerunsbeforeSeptember 2003andcontinuedtodosoafterSeptember2003.12Inmostofthe twelvemonthsfollowingSeptember2003,theflowsofthesecond treatmentgroup(otherfundsinfamilieswithinvestigatedfunds) arealsosignificantly lowerthantheflows ofthecontrolgroup (fundsinfamilieswithnoinvestigatedfunds).Theresultssuggest thatinvestorsmayseeinvolvementinlawsuitsasanindicatorthat fundfamilymanagershavefailedtoservetheinvestors’interests.
Asa result,theywillpunishallfundsintheimplicatedfamilies regardlessofwhetherornotthefundinquestionallowedabusive practices.
Weinvestigatewhetherpre-eventoutflowsaredrivenbyout- flowspriortothefirstlitigationinSeptember2003orbyoutflows fromfundsthatarelitigatedafterthefirstlitigation.First,wedisag- gregatethelitigatedfundsintotwogroups.Thefirstgroupcovers litigatedfundsinthemonthsbeforethemonthoftheirlitigation filing(inthefirstspecification)andthenewsdate(inthesecond specification).Thesecondgroupcoverslitigatedfundsinthemonth ormonthsaftertheirlitigationfilingandthenewsdate.Second, wedifferentiatebetweeninstitutionalfundsversusretailfunds.
12 Thispatternoflowerflowspersistsmorethantwoyearsaftertheeventmonthin non-tabulatedresults.Thelongertime-seriesofflowcomparisonsinnon-tabulated resultsalsoshowthatthesignificantflowdifferencesbetweentreatmentgroup 2andthecontrolgroupduringOctober2002andJanuary2003israndomrather thansystematic;therefore,thereisnoalternativehypothesistoexplainthelater systematicflowpatternsaroundthelitigation.
AppendicesA1andA2presenttheresults.Outflowsfollowingthe September2003litigationsaredrivennotonlybyfundslitigated inthatmonthbutalsobyfundsthatarelikelytobeimplicatedbut havenotyetfacedlitigation.Thetablealsoshowsthattheflowpat- ternsremainconsistentregardlessofthechoiceoftheeventdate.
Retailinvestorsseemtobemoreresponsivetolitigation.
Table4reportstheresultsofthepooledregressionestimates fortheflowmodeldescribedinEquation(2).Monthlyflowsare regressedonindicatorsfortheevent-window(pre-event,theevent month, and post-event)and four setsof controls—fund charac- teristics, past returns, fee structures, and aggregate flows. The regressionsincludefundandyear-monthfixedeffects.Thethree specificationsdefinetheeventtimeusingthelitigationnewsdate, thefilingdate,andthewindowbetweenthenewsdateandtheliti- gationdate.Observationsareatthemonth-fundlevelandcoverthe monthsfromJanuary1999toDecember2007.Theregressionsuse cluster-robustvariance/covarianceestimatorsinwhichtheclusters arethefunds.
TheresultsinTable4confirmtheoutflowsduringandafterthe eventmonth,aspreviouslyindicatedinFig.1,Table3,AppendixA andB.Therearesignificantoutflowsfrominvestigatedfundsduring theeventmonth(−56basispointsoftheTNAinthesecondspecifi- cation)andinthepost-eventperiod(−107basispointsoftheTNA inthesecondspecification).Funds,pastreturns,andaggregateflow controlsarealsosignificant.First,youngerandlargerfirmsenjoy significantlyhigherflowsthandotheirolderandsmallercounter- parts.Second,investorschasepastreturns.Third,fund-levelflows increase(decrease) significantlywithstyle-level(industry-level) flows.
Averagingapre-event periodmayglossover monthlyvaria- tionsinflowsduringeachmonth.Therefore,werunthepooled regressions with13 event-month indicators for the12 months surroundingthelitigationnews.AppendixBpresentstheresults.
Therearefoursetsofcontrols—fundcharacteristics,pastreturns,