• Aucun résultat trouvé

Copycat funds : information disclosure regulation and the returns to active management in the mutual fund industry

N/A
N/A
Protected

Academic year: 2021

Partager "Copycat funds : information disclosure regulation and the returns to active management in the mutual fund industry"

Copied!
46
0
0

Texte intégral

(1)
(2)
(3)
(4)

Digitized

by

the

Internet

Archive

in

2011

with

funding

from

Boston

Library

Consortium

IVIember

Libraries

(5)

'31

Dewey

[415 ;

DM

Massachusetts

Institute

of

Technology

Department

of

Economics

Worl<ing

Paper

Series

Copycat

Funds:

Information Disclosure Regulation

and

the

Returns

to

Active

Management

in

the

Mutual

Fund

Industry

Mary

Margaret

Myers

James

M.

Poterba

Douglas

A.

Shackelford

John

B.

Shoven

Working

Paper

C2-04

October

2001

Room

E52-251

50

Memorial

Drive

Cambridge,

MA

021

42

This

paper

can be

downloaded

without

charge from

the

Social

Science

Research Network Paper

Collection

at

(6)
(7)

f

Massachusetts

Institute

of

Technology

Department

of

Economics

Working Paper

Series

Copycat

Funds: Information Disclosure

Regulation

and

the

Returns

to

Active

Management

In

the

Mutual

Fund

Industry

Mary

Margaret

Myers

James

M.

Poterba

Douglas

A.

Shackelford

John

B.

Shoven

Working

Paper

02-04

October

2001

Room

E52-251

50

Mennorial

Drive

Cambridge,

MA

021

42

This

paper

can be

downloaded

without

charge

from

the

Social

Science

Research

Network

Paper

Collection

at

(8)

MASSACHUSEHS

INSTITUTE

OF

TECHWOLOGY

MAR

7

2002

LIBRARIES

(9)

Copycat

Funds:

Information

Disclosure

Regulation

and

the

Returns

to

Active

Management

in

the

Mutual

Fund

Industry

Mary

Margaret

Myers

Graduate School of Business, University

of

Chicago

1101 East 58th Street,

Chicago

IL

60637

James

M.

Poterba

Department of Economics,

MIT

50

Memorial

Drive,

Cambridge

MA

02142-1347

Douglas

A. Shackelford

Kenan-Flagler Business School, Universityof

North

Carolina

Campus

Box

3490,

McCoU

Building,

Chapel

Hill,

NC

27599-3490

John

B.

Shoven

Department of Economics,

Stanford University EncinaHall, Stanford

CA

94305

Revised October

2001

Contact Author:

James

Poterba,

Department of

Economics

E52-350,

MIT,

50

Memorial

Drive,

Cambridge

MA

02142-1347;tel (617)

253

6673;fax (617)

258

7804; email

poterba@mit.edu

ABSTRACT

Mutual

ftinds

must

disclosetheirportfolio holdingstoinvestorssemiannually.

The

costs

and

benefitsof suchdisclosures are along-standingsubjectofdebate.

For

actively

managed

funds, onecost ofdisclosure

isapotentialreductioninthe private benefits firomresearch

on

asset values. Disclosureprovidespublic accesstoinformation

on

the assets thatthefund

manager

views

asundervalued. Thispapertries to

quantifythispotentialcostofdisclosure

by

testing

whether

"copycat"mutual fijnds, fiands thatpurchase

the

same

assetsasactively-managedfiindsassoonasthoseassetholdingsare disclosed,can earnretums that aresimilartothose oftheactively-managedfiands.

Copycat

fiands

do

not incur the researchexpenses

associatedwiththeactively-managed fundsthattheyaremimicking,butthey

miss

the opportunityto

investinassets that

managers

identify as positivereturn opportunities

between

disclosuredates. Oior

resultsfora limited

sample

of

highexpensefiands inthe 1990s suggestthatwhile

retums

beforeexpenses

aresignificantlyhigherfortheunderlyingactively

managed

fundsrelativetothecopycatfunds,after expenses copycat funds

cam

statisticallyindistinguishable,

and

possiblyhigher, retums thanthe

underlyingactively

managed

fiands.

These

findings contributeto thepolicydebate

on

theoptimallevel

and

frequency of fimddisclosure.

We

aregrateful to

Dan

Bergstresser, Joel Dickson,JenniferWilson, seminarparticipants atthe University

ofIllinois-Chicago

and

IndianaUniversity,

and

especially

Mark

Wolfson

for helpfiildiscussions.

We

alsothankJennifer Blouin, V.J. Bustos,Rachel Ferguson, Michael Myers,

Angel Townsend,

and

Candice

Whitehurstforexcellentresearchassistance,theErnst

&

Young

Foundation,theNationalScience Foundation(Poterba)

and

theUniversityof

Chicago

(Myers)forresearch support,

and John

Rekenthaler

(10)
(11)

The

U.S. Securities

and

Exchange

Commission

is ciirrentlyconsidering

modifying

theregulations

thatgoverndisclosuresthat

mutual

funds

must

make

to theirshareholders.

Under

the

1940

Investment

Company

Act, investment

companies

must

disclose boththeir

performance

and

theircurrent portfolio

holdingsinsemi-annualreportstoshareholders.

These

reports

must be

senttothe shareholders

no

more

thansixty daysafterthereportingperiodends.

A

number

oforganizationsrepresentingfiand shareholders

have

calledfor

more

frequent disclosure

of

mutual fundholdings,

on

thegroundsthatthey

would

enable

investors to

more

accuratelyselect ftindsthat

match

theirinvestmentobjectives.

More

frequent

disclosure

would presumably

permit shareholderstodetect,sooner,

changes

infund investmentstrategy.

They would

consequently reducetheriskthatfund

managers

couldpursue investmentstrategies that

do

not coincide withtheirshareholders'wishes.

Previousresearch

on

financialdisclosurerecognizesthat

mandatory

disclosurehas bothcosts

and

benefits.

The

costsinclude thedirectexpensesassociatedwithproducing

and

disseminating information

on

investmentpositions,aswell as arange

of

potential coststhata discloser

may

face

when

private

information

becomes

publiclyavailable.

The

benefits ofdisclosure

emanate

from improved

monitoring

of flmd managers,

and

from

potential

improvements

in investorchoicethat result

from

detailed

information availability.

Both

the costs

and

benefits

of

disclosure are typicallyverydifficulttoquantify.

In themutual fundcontext,

Wermers

(2001) notesthatthere are

two

potentialcoststothe

investors inafiind

when

thefundisrequired to disclose itsholdings.First,

when

afiind disclosesits

holdings, it

becomes

easier forother investorstouse information

on

fiind inflowsto "front-run"thefiind's

trades,therebybidding

up

the prices

of

the securities thatthefund

manager

wishestobuy. Thiscostis

presumably

greater

when

disclosureis

more

frequent,

and

ittranslates intoalowerreturn

on

the fimd's investments. Second,disclosurereducesthetime period over

which

fundinvestors areabletoreap the

privaterewards oftheirmanager'ssecuritiesresearch.

The

managers

who

directinvestmentsin

actively-managed

open-end mutualfiindscarryoutresearchaboutvarious securities toidentify underpricedassets

that will generateabove-average, risk-adjustedreturns.

Because

disclosure reveals the identityofthe

(12)

Potentialcompetitors, as well as

fund

shareholders, learnaboutafund manager's investments

when

the

fiinddiscloses itsholdings.

The

fund manager's uniquereturntoinvestinginthesecuritiesthat his

researchsuggestsareunder-valued istherefore limitedtothetime

between

thecompletion ofthe research,

and

thenextdisclosuredate. This analysis

presumes

thatresearch

by

fund

managers

uncoverspositive

return opportunities;a largeempirical literature,reviewedfor

example

in

Gruber

(1996),suggeststhat

thisassumptionis

open

todebate.

These

two

costs

must be

balancedagainst thepotentialbenefits ofdisclosureboth

from

the

standpoint

of

anindividual ftind

manager

who

is considering increasing thefrequencyofdisclosure,

and

fi-omthestandpoint

of

aregulator tryingtodesign an optimalpolicy.

One

potential benefit

of

disclosure

may

be heightened

demand

forthe securities

owned

by

adisclosingfijnd. Ifother investorsdecideto

purchasethe securities thatafimd already

owns

because theybelieve thatthe disclosing

manager

has

private information, this

may

drive

up

theprice

of

these securities,therebyraising the returns

on

thefiind

that

makes

the disclosure.

Another

benefit isthat

some

investors

may

attachsubstantialvalue tofrequent disclosure,

and

therefore

be

preparedtoaccepthigherfeesorlowerreturnstoreceive frequentfunddisclosures. Frequent

disclosure

makes

itdifficultforftindstopursuestrategies thatare substantiallydifferent

from

theones thattheyadvertise.

The

gains

from

suchdisclosure

depend on

the likelihoodthat ftinds

change

their

strategieswithout informingtheirshareholders, and

on

the costtoshareholdersofdeviations

from

pre-announced

strategies.

Although

the current regulatory

environment

requiressemi-annualdisclosure,

some

managers

voluntarilydisclosetheir fiindpositions

more

frequentlythanthe

SEC

requires.

They

presumably

believethatinvestorsare

more

likely to invest ina fimdthatprovidestimely portfolio

information,

and

theyvaluethe associated increasein theirfiind'sassets

more

highlythan the potential

futureinflowsassociatedwithhighercurrentreturns.

The

possibility that

some

firms

may

voluntarilydisclose

more

than the regulatorsrequire, to

attractaparticular investorclienteleisnotuniquetothemutual fimd market. In discussing insider trading

(13)

firm'sshares,

some

firms

would

voluntarilychoosetoprohibitsuchtransactions, they

would

thereby

attractinvestors

who

were

preparedtoaccept

slow

incorporation

of

informationintoprices (ifinsiders

could nottrade) inreturn forexcludingbetter-informed traders

from

themarket.

A

thirdpotential benefit

of

disclosureisthatit

may

convey

information

on

a firm's successfiil

pastinvestmentsto prospective investors

and

therebyattract

them

tothefiind.Verrecchia (1983)notes

thatvoluntarydisclosures

of

product innovations or otherresearchresults

may

increasefirm value

by

persuading investors

of

thefirm'sresearch

acumen.

Thiscan occur

even though

disclosurefacilitates the

competitivestrategies

of

rivals

and

reducesthe

market

value

of

the proprietary returns

on

theresearch

beingdisclosed.

The

literature

on

"window

dressing"

by

mutual funds

and

otherinvestmentmanagers,

including Carhart, Kaniel,

Musto,

and Kadlec

(2000), Lakonishok,Shleifer, Thaler,

and Vishny

(1991),

and O'Neal

(2001), suggeststhat

managers

believethatinvestors will

judge

prospective

performance

on

thebasis

of

pastperformance.

The

problem of

deciding

how

toregulateinformationflows

between

a

mutual

fiind

and

itscurrent

and

prospective investorsisclosely relatedtoarange

of problems

in financialaccountingregulation.

For

example,regulators

and

managers have

long debatedthe extentto

which

requiringgeographic

segment

disclosures,

and

othertypes

of

detailed financialinformationrelease,

conveys

valuable informationtoa

firm'scompetitors withoutproviding

much

informationtoinvestors.

A

number

of

previousstudies

have

consideredthedesign

of

disclosure regulation forfinancialinformation. Foster(1980) notes thatthe

"extemalities" associatedwithfinancial reporting

may

leadfirmstounder-provide informationinan

unregulated market.

Admati and

Pfleiderer(2000) investigatethenature

of

voluntarydisclosure

equilibria, andthecircumstancesunder

which

disclosureregulationiswelfare-enhancing.

Both of

these

studiesnotethatthe optimalregulatory structure for disclosurewill

depend

on

the firm-specific costs of

disclosinginformation.

Optimal disclosure policy

depends

on

the costs

and

benefitsofdisclosure,yet thereisremarkably

little empiricalevidence

on

eitherofthese issues. This paper seekstoprovide

new

insight

on one

ofthe

(14)

reductionin potentialexcessreturnsearned

by

the

managers of

actively-managedequity funds.

We

emphasize

that thisisonly

one of

thepotentialcostsofdisclosure,

and

that an

improved measure of

this

costalonedoes notresolve thequestion

of

theoptimaldegree

of

disclosure,

which

must depend on

the

costs aswell asbenefits

of

potential disclosurerules.

To

investigate

how

disclosure affectsthe returnsearned

by

actively-managedmutual fiindsand

other investors,

we

create"copycat"fiindsthatallocate assetsto

match

thelatestpublicly-disclosed

holdings

of

actively-managed funds.

We

then

compare

the returnsof each copycatfiindwiththe returns

of

the

mutual

fundthatitmimics. Ifresearchisvaluable inuncoveringpositivereturn opportunities, the

copycatfund should earn lowerreturnsthanitsprimitiveactively-managedfiind.

The

active

manager

can

implement

theresults

of

new

researchimmediately,whilethe copycat

manager

can onlytrade

on

new

informationafteritispublicly disclosed.

The

copycatfiind'spotentialdisadvantage intimelyaccessto

research findings

may

be

offset,

however,

by

the fact thatthe copycatfundalsohasvirtually

no

research

expenses. Thus,itispossible thatacopycat fiind

and

an actively

managed

fiondcoulddeliver similar net

of expense

returns,

even

ifthe copycatfundearns alowerreturn before expenses.

We

recognize thatourtestsare ofgreatestinterest

when

research

by

active

managers

hasthe

potential togenerate positive returns beforefund expenses. Ifactive

managers

areunableto

add

value throughtheirresearch, copycatfiindsshould beableto

match

the returnsoftheirprimitivefundsbefore

expenses,

and

theyshouldoffersuperior returns net

of

expenses.

Our

research strivestoprovideinsight

on

theextentto

which

expense savings can

compensate

for

foregone assetallocation opportunities

on

the partof copycatfunds.

We

investigatetheviabilityofthe

copycat strategy

by

studyingthe returnstoaset

of

actively-managedfiindsandassociatedcopycat funds

from

1992-1999.

The

view

thatdisclosurerestrictsthecapacity

of

actively-managedfiindstoreapthe

potential benefits oftheirresearch findings

would

beconsistentwiththenet-of-expensereturnstocopycat

fiindsprovingindistinguishable

from

the returnsofprimitive fiinds,whilethebefore-expenseretums of

(15)

Our

analysis isdividedinto five sections. Section

one

summarizes

the disclosure regulationsthat

currentlyapplytomutual funds

and

provides

background

forunderstandingour copycat fundstrategy.

Section

two

describesour algorithmfor

managing

acopycatfluid, including thefrequency

of

portfolio

adjustments

and

the relationship betw^een the dates

when

actively-managedfiindsdisclose information

and

the dates

when

copycat funds rebalancetheirholdings. Sectionthree explains the selectionprocess

for the actively

managed

fiinds inourdata

sample

and

presents

summary

statistics

on

theexpenses

and

returns

on

these funds. Section four reportsourprincipal empiricalfindings. In

most

cases,

we

findthat

returns tothe primitivefunds

exceed

thoseofthecopycat funds beforeexpenses.

When we

compute

returns toboththe primitive

and

thecopycat fundsnet

of

expenses,however,returns

on

thecopycat fimds

often

exceed

those

on

theprimitivefiinds, although

we

typicallycannotrejectthe nullhypothesisthat

returns

on

thecopycatfiindsareequaltothoseofthe primitive funds.

A

briefconclusionoutlinesseveral

broad

issues relatedtoinformation disclosure

by

financial intermediariesthat

emerge

firomouranalysis.

1. DisclosureRegulations

and

the

Mutual

Fund

Industry

Mutual

fundsarerequiredtodisclosesemiannuallytheirbalancesheets, including alist

of

the

securities thattheyholdandthevalue

of

thesesecurities. Section 30(e)oftheInvestment

Company

Act

of

1

940

stipulatesthe relevant disclosurerequirements,

and

specifiesthatthelistofsecuritiesheld

must

be

fora "reasonablycurrent date." Securities

and

Exchange

Commission

Rule

30bl-l is

more

specificin

outlining theprocessofdisclosure. Registeredinvestment

companies

must

file

form

N-SAR

not

more

than

60

calendardaysafterthe closeoftheirfiscalyear,

and

again afterthecloseofthe

second

quarterof

their fiscalyear.

Fund

familiesvaryin theirdisclosurepolicies.

Some,

suchasfiinds in the Fidelity family,

do

not

make

voluntarydisclosures.

Most

fiinds,however,disclosetheirholdingsbeforethe

end

ofthe

two-month

graceperiod, and

many

fiindsdisclose portfolioholdings

on

a quarterlybasis.

For

example,the

Vanguard

Group

disclosesitsfunds' holdingseveryquarterwitha

one-month

lag. Inaddition,

monthly

(16)

reportsthat

30

percentof mutual funds,includingtheJanus

and

USAA

Investment

Management

fiind

families,releasecompleteportfolioholdings

monthly

to

Momingstar,

a firm that tracks returns

and

providesinvestorswith informationthatthey

may

findusefiilinevaluating

mutual

funds.

The

factthat

investors

pay

substantial

sums

to

Momingstar

toobtain these datasuggeststhat at least

some

market

participantsregardfunddisclosuresasvaluableinformation, perhapsbecauseit offersaguidetothe

fiiturebehaviorof fimd managers.

Some

firms

even

disclose

more

fi-equentlythanmonthly.

Laderman

(1999)reportsthat the

Open Fund

postsall

of

itstrades

and

reportsitsentireportfolioinrealtime

on

its

web

site.

Fund

familiesthat

do

not voluntarily disclosetheirholdingstypicallycitedistributioncostsasthe

major

impediment

to

more

frequent disclosures.

For

instance,the

Omni

hivestment

Fund

historically

mailed

monthly

statements

of

fund holdings toitssmall

group

ofshareholders. Fairley(1997)reportsthat

after

Berger

Associatesfiind familyacquired

Omni,

disclosures

were

reducedtosemiannualreports

because

of

distributioncosts.

However, even

when

ftinds

do

notmaildetailed disclosurestoall investors,

some

ftindswillprovideacompletelist

of

theirinvestments, or apartial listingwiththeir

most

significant

holdings, to investors. Investors also

may

obtaininformationthatisnotmailedtoall shareholders

by

contacting theftmd

manager

directly.

Fund

web

sitesincreasinglydisseminate additional portfolio

information.

2. Primitive

and

Copycat

Funds

To

evaluate

one

ofthe costs

of

informationdisclosureforactively-managedequityfiinds,

we

design "copycatfiands" that spend nothing

on

researchbutselectassets

by

followingan actively-managed

fiind.

The

fiind that carriesout research

on

asset selection istheprimitivefund. Letthe return

on

this

fundequalRprimitive, preexpense,

sud

Ictthcftind'sexpenses equal eper period.

The

net-of-expensepretax

retum

toan investorinthisfundis

(17)

When

detailedinformation

on

the primitivefund'sportfolioholdings

become

available, the

copycatfiind will alignitsportfolioexactlywith thereportedholdings oftheprimitive fiand. Consistent

withthe

SEC

mandatory

disclosurerules,theprimitive flmdis

assumed

todiscloseexactlysixtydays

afterthe close

of

the primitive fimd's

second and

fourth quartersofthefiscal year. Thisassumption

implies thatthecopycatfundis alwaysatleast

two months

"outofdate"intracking the primitive fimd. It

can

be

as

much

aseight

months

behindtheprimitive fimd, inthedaysimmediatelyprior toa

new

semiannualdisclosure.

The manager

of

the actively-managedfiindchanges assetallocation

between

the datesofrequired

disclosure,whilethecopycatfiindonlyadjustsitsportfolio

when

the

manager

oftheactively-managed

fiinddiscloses

new

holdings. Ifidentification

of

new

stocks

by

actively-managed fimdsgenerates a

uniform

distribution

of

trades

between

disclosure dates,the

manager

oftheprimitivefundwill holda

securityfor fivemonths,

on

average,beforethe copycat fundwillpurchaseit. Thislag

may

be longerif

the active

manager

pursuesstrategies designedtopreventimitation, suchasdelayingthepurchase(sale)

of

some

securities that

may

have

positive (negative) returnpossibilitiesuntiljust afterthedisclosuredate.

Ifthe active

manager

iscompletelysuccessful incamouflaging hisorherfund's

tme

portfolioholdings,

thenthecopycat fundwillnot

even

be

abletoholda laggedversionoftheactively-managedfund's

portfolio. Inthiscase,thecopycatfundwillbe holdinga portfoliothat correspondsto

whatever

assetsthe

actively

managed

fund

found

itattractivetopurchaseas partofthe

camouflage

program.

The

extentto

which managers

atactively

managed

fundstradetodisguisetheirholdingsatthe

time

of

disclosure isan

open

issue.

While

masking

strategiescan avoid informing competitors

and

investorsaboutcurrent portfolio positions,they

may

also

impose

potentially substantialtransaction costs

on

thefund pursuing them.

Musto

(1999)discusses

more

generallythepotential gains

from

short-term

tradingthatisdesignedto affecttheinformationtransmittedto investors,

and

O'Neal (2000)presents

some

evidence offiindreturnabnormalities aroundfund disclosuredates,suggestingthat

some

unusual

(18)

We

denotethebefore-expensereturn

on

thecopycatfundasRcopycat.preexpense- Ifthestock

selectionassociatedwithactive

management

generates positiveretums,

we

would

expectthat

\^) -^Vropycat,pre-expense -^^primitive,

pre-expense-The

copycat fundisalwaysrelying

on

dated informationin

making

portfolio choices, soitsretums should

be lower

thanthose ofthe actively

managed

fimdthattakesfull advantage of

new

informationasit

arrives.

However,

thecriticalquestionforinvestorsis

whether

thecopycat fund'sreturn, net

of

expenses,

exceedsthe

comparable

return

on

the primitivefiind. Ifafraction

X

of

theactively-managedfund's

expensesisassociatedwith research

and

other costsofactive

management,

suchasbrokeragefees, then

thecopycat fundcangenerateanafter-expense returnof

\^) -^Vropycat, net ^M:opycat, pre-expense "

vA'^^J^-The

parameter

X

islikely tovaryacrossfiinds

of

differenttypes.

Our

analysisfocuses

on

pretaxretums,

but

we

notethatfortaxable investors, thecapital gains taxliability associatedwith investmentsincopycat

funds

might

be lowerthan those for primitive funds, since thecopycatfiinds will

presumably

tradeless

thanthe primitive fund.

Because

the copycatfund onlytradestwiceeachyear,torealignitsportfolio

and

that

of

theprimitive fiond,itis

somewhat

lesslikelythan the primitive ftindtorealize capital gains.

Our

empirical

work

computes

thedifferentialreturnofthe primitive

and

copycatfiinds

on

a

pre-expense

and

apost-expensebasis,

and

testsforstatisticallysignificantvalues

of

\

V

^pre-expense ^S^r'nii^i^^'pre-expense ~-^Nropycal,pre-expense

and

\p} ^net l^primitive, net~ ^N:opycat,net •

To

compute

Rprimitive,net,

wc

uscthe

monthly

return asreported

by Momingstar.

Thisisthe

change

inthe fund's net assetvalue

(NAV)

duringthe

month

divided

by

the net assetvalueatthe beginningofthe

month, assuming

reinvestment

of

dividends

and

capital gains distributions.

NAV

equals the fiind's total

assets, lessfees

and

expenses,divided

by

the

number

ofsharesoutstanding. Thisreturnisnetof expenses paid

from

fundassets,suchas 12b-1

and

management

and

administrative fees.

(19)

Our

primitivefundreturnisnotreducedfor loads,brokeragecosts,

and

other coststhat

do

not

easilyconvertto

monthly

returns.

To

theextentthatactively-managedprimitivefundsare

burdened

more

than passivecopycats withthese additionalcosts, Rprimitive,netunderstates the potential net return

advantagesof copycatftmds. In theempiricaltests, Rprimitive, pre-expenseequalsRprimitive,netplusanestimate

of

the

monthly

expenses paidfirom tundassets.

The

estimated

monthly

expenseis 1/12ofthepercentage

of

thefund'sassetsdeductedeachfiscalyearforfund expenses.

For

example,

assume

that

Momingstar

reportsareturn

of

2.0percent for afund ina given

month, and

thatthe fund'sannual

expense

ratiois 1.2

percent. In thiscase,theestimated

monthly

expenseratiois0.1 percent, 1.2percent/12,

and

theadjusted

monthly

returnis2.1 percent (2.0

+

0.1).

We

compute

thecopycat'spre-expensereturn, Rcopycat,prc-expense,as the

sum

ofthevalue-weighted

monthly

returns foreach stockheld

by

the primitivefiind. Portfolioholdings

and

theirweightsare

collected

from Momingstar.

The

monthly

returnsforstocks are

computed by

compounding

dailyreturns

inthe daily

CRSP

files. Ifthestockislisted

on

CRSP,

we

useitsdaily

retums

includingdistributions. If

a

common

stock isnotlisted

on

CRSP,

for

example

ifitisa closely-held or foreign-controlled

company,

we

assume

itsdaily returnequalsthe distribution-inclusive return

on

thevalue-weighted

market

portfolio

as reportedinthe

CRSP

files. Ifa

common

stockheld

by

thecopycatfimd dropsfi-om

CRSP

duringthe

six-month

buy

and

hold period

between

information disclosures

we

assume

thattheassetspreviouslyheld

inthatstockearn the value-weighted

market

returnuntilthenextdisclosureof information

from

the

primitive fund. Thisistantamountto

assuming

thatthecopycat

fund

manager

reinvests, inabroad

market

index, theproceedsfiromsellingsharesthatstoptrading.

Many

of

theprimitivefiinds inour

sample

hold

some

of

theirassets in securities otherthan

corporate stocksthatare included

on

the

CRSP

tape.

Computing

theretums

on

theseotherassetsis

problematicbecause

we

oftenlack detailed information

on

theidentity

of

theasset, the return

on

theasset,

or both.

To

overcome

theseproblems,

we make

a range

of

assumptions withrespecttocopycatfimd

retums. First,

we

assume

that

bonds

eam

theIbbotson Associates

monthly

return

on

long-termcorporate

bonds. Second,

we

assume

thatcashearns theTreasurybill

monthly

rate. Third,

we

assume

thatthe

(20)

returns

on

small equity

and

otherassetholdingsare proportionaltothe return

on

the fund's otherassets.

Momingstar

roundsthe portfoliopercentage weighttozeroifthefund holdslessthan 0.006 percent

of

its

portfolioinaspecific security.

Only

aboutone percent ofthe equitiesheld

by

fundsinour

sample

have

weights

below

thisthreshold,

and

the

median

fundinbothsamples

had

no

holdings

below

thisthreshold.

'

We

assume

thattheassets in thisunreported categoryareinvestedintheodierassets inthecopycat fund

(i.e.,equities,

bonds and

cash),withweightsequal to the share

of

theotherassets inthecopycat fiind's

portfolio. Finally,

we

assume

thatpreferredstockthatisnotlisted

on

the

CRSP

filesalsoearns the

average

retum

of

thecopycatfund'sotherassets.

We

do

thisbecausepreferred stockhas bothequity

and

bond

features.

To

illustrateourprocedureforconstructingcopycatfundreturns,suppose afiind'sassetsare

invested

40

percentin

Common

Stock

A,

which

isincludedinthe

CRSP

files,25 percentin

Common

Stock B,

which

isnotlistedin

CRSP,

30

percentin

Bond

C,

and

4.98percentincash.

The

remaining

0.02percentofthe portfolioisinvested infour stocks,eachcomprising 0.005 percentoftheportfolio.

First,

we

dropthefourstocksthatcomprise only 0.005 percent

of

the fund,because

Momingstar

reports

theirportfolioweightsas percent,

and

reweighttheremaininginvestments. Stock

A's

weightis

now

assumed

tobe 40.008 {40/(100-0.02)} percent, B's weightis25.005 percent,C'sweightis 30.006

percent,andthecash weightis

now

4.981 percent. IfStock

A

has a

4

percent

retum

forthe period, the

market

retum

is3 percent, theIbbotson Associatesretuin

on

long-termcorporate

bonds

is2percent,

and

theTreasurybillrateis 1 percent,thenthecopycat

retum

is 3 percent(.03

=

(40.008*0.04

+

25.005*0.03

+

30.006*0.02

+

4.981*0.01)/100).

At

least

one

possible

measurement problem

ariseswiththis computationmethod.

Momingstar

does notnecessarilysimultaneouslyreport equity portfolioweights

and

the overallassetallocation

from

which

we

set

bonds and

cashweights. Thus, ifinformation isreleased

on

different dates,ourcomputation

oftotalweights

may

notequal 100percent.

We

therefore reweighttheholdingstoachievea consistent

'The

maximum

percentage ofassetsinholdingsbelowthisthresholdis 16.4percentinSample 1,and 21.9 percent

inSample2.

(21)

outcome.

For

example,

suppose

Momingstar

releases equity portfolioweightsthatdisaggregate95

percentofaftind'sequityholdingsatyear-end, whilealsoreportingthatthefund's overall asset

allocation is

97

percent equity

and

3 percent

bonds

forthe

month

priortothe year-end.

The

portfolio

weights

would

be

adjusted

by

theratio 100/(95

+

3), leaving thetotal equityholdingsat 96.9 percent

(95/98)

and

the

bond

holdingsat 3.1 percent(3/98)

of

thefund. Inpractice,thereare

no

more

than

sixteen days,

on

average,

between

thedates

of

portfolio

and

asset allocation disclosure, so

we

suspectthat

the inconsistencies associatedwiththedifferential dating arelimited. Further, theaveragedifference

between

the

sum

of

the portfolioequityweights

and

thereported allocationtoequitiesisonly

one

percent.

The

SEC

allowsfimdssixtydays followingthe

end

oftheperiodtodisclosetheirholdings.

We

thereforebeginestimating returnstothecopycatfund

two months

afterthe

end

ofthereporting period.

For

example, a

mutual

fund withacalendaryear-end

must

discloseitsyear-endportfolioholdings

by

the

end of

February. It

must

make

a similar report

on

itsholdingslateinthe

second

quarter

by

the

end

of

August. Therefore, thecopycatftind

retums

forthe

March

-

August

periodusetheprimitiveftind's

portfolioholdings reportedatthe

end

of

February,

and

copycat

fund retums

for

September

-February use

theprimitive fund's disclosure

from

lateAugust.

The

estimatedcopycatfund

retums

foreach

of

these

periodsarethen

compared

withtheprimitivefund'sretumsforthe

same

period.

Our

assumptionthat the

copycatfiindcan onlytracktheprimitiveftind'sportfolio

from

thesemiannual

mandatory

disclosure is

conservativebecauseifthe primitiveflmd

makes

voluntary disclosures,thecopycatftindcantrackits

assetholdings

more

closely.

We

assume

thattheexpensesincurred

by

copycat funds equaltheexpenses

ofthe

Vanguard

TotalStockIndexfund. Thisisan index ftmdthatinvestsinbothlarge

and

small

capitalization stocks,

and

we

view

its expensesasillustrative

of

thecostsa

passively-managed

copycat

ftind

might

incur.

We

will demonstrate

below

that ourqualitative findingswithrespect to the

performance

differentials

between

primitive

and

copycat fundsarerelatively insensitiveto

modest

changesinourassumptions abouttheexpensesof copycatfunds.

We

report

monthly

returndifferentials, Apre-expenseandAnd, foreachofthe six

months between

the

disclosure dates ofthe primitive fund.

Because

thecopycat ftind'sportfolioholdingsshouldstray

more

(22)

from

theprimitive fund'sholdingsas thetime since thelastdisclosure increases, there

may

be

some

informationinthe patternsof

retum

differentialsfordifferentmonths.

We

alsoreportcumulativereturn

differentials forthe

one

tosix

month

intervals

between

thesedisclosures.

3.

Data

Sample and

Fund

Selection

The

potential net

retum

advantage

of

a copycatfiindisgreatest fora primitive fitnd thathasa

high expenseratio.

Such

fimds

might

be,butarenotnecessarily,

engaged

in

more

researchthanother

funds. If

one

were

goingtointroduce acopycatftind intothemutual fund marketplace,it

would

be

natural touseahighexpense fundas a primitive, sinceitshighexpenses

would

offer thegreatestpromise

forthecopycat,through it'slow-coststrategy,tooutperform.

For

thisreason,

we

begin ourempirical analysis

by examining

a

sample

of

large equity funds with highexpenses.

To

assess therobustnessof our

findings,

we

repeat the analysis

on

a broader

sample

offunds.

We

focus

on

equityfundsbecauseitis

relativelyeasy,usingthe

CRSP

tapes, totrackthe

retum

on

thesefluids' investments. Inpractice,thereis

no

reasona copycatfund needsto

be concerned

aboutthe availabilityof

CRSP

data.

The

copycat

strategycouldeasily

be

appliedtofundsthathold

more

exoticassets,providedtheinformationdisclosed

by

theprimitivefund

made

itpossibleforthecopycat

manager

toidentifytheunderlyingassets inthe primitive fund'sportfolio.

We

draw

our

sample

from

theequity

mutual

ftindsincluded

on Momingstar's

July

1992

Principia

database.

Because

CRSP

onlyreports returns forequitysecurities

on

domestic stock exchanges,

we

eliminate internationalequity funds.

We

alsoexcludesmallcapitalization hands

and

specialtyfunds

from

our sample; thesecouldbethe subject

of

separate,follow-on studies.

Our

sample

restrictions limitour

sample

universeto

812

funds.

We

draw two samples

offtinds

from

thisuniverse.^

The

High-Expense

Fund

Sample

.

The

first

sample

comprisesthe

20

fundsthat appearto

meet

most

closelythe definitionoflarge,diversified equityfunds with largeinvestorfees.

The

sample

^We

supplementedourinitialsample,whichwascollectedby hand from 1992-93

Momingstar

reports,witha

databasesuppliedby Momingstar,Inc. for1993-1999.

(23)

excludes fundsthatinvestin assetsotherthan equities

and

cash. Furthermore,thecash allocation

must

be

lessthan 1 percentofallinvestments.

We

exclude funds with saleschargesequaltozero

and

expense ratios lessthan 1 percent.

The

sample

alsoexcludes indexfunds,funds withassetsoflessthan

$200

million,

and

flindswith

more

thanhalftheirassetsallocatedtoequitiesin

one

industry.

For

these

20

mutual

funds,

we

collecteddata

from

Momingstar

forall

of

the

SEC-mandated

semiannualreporting

periods

between

1992 and 1999.' Since ourdatasetspansjustoversixyears,disclosures occur twice

eachyear,

and

we

have

twentyfiinds,

we

would have

a

maximum

of

roughly

240 (=20*6*2)

observations

on fund

disclosure. Infact,

we

have

a

somewhat

smaller

sample

188disclosures.

We

have

not

addressedissuesconcerning survivorship biases for theflindsinour sample, since

we

arenot

comparing

returnsforthesefiindswithotherfunds orthebroad market, butratherwith a setofhypotheticalcopycat

fundsthataretracking thefundsinour sample.

The

Broader

Equity

Fund

Sample.

Our

second

sample

islarger

and

is

drawn

withfewer

restrictions. Itincludes the largest 100flinds(by netassetvalue)thatallocate lessthanfortypercent

of

theirassets tobonds,preferredstock or convertiblesecurities.

As

withtheprevious sample, index funds

areexcluded,

and

thedata

come

fi^om the

SEC-mandated

semiannualreports for1992 -1999, available

through

Momingstar.

Appendix

A

liststhesetof fundsforboththehigh-expense fiind

sample and

the

broaderequityfiindsample.

We

define an observationforthepurpose of our

sample

sizeas afund

disclosure.

Table 1 providesdescriptivestatisticsfor the

two

fund samples.

On

average, equitiestracked

by

CRSP

comprise91 (81)percentofthehigh-expense(broader) sample'sportfolios. Forthe

median

flinds

ineach sample, the analogousstatisticsare

92

percentand

86

percent, respectively.

Some

funds ineach

sample

have

more

than 100 percent

of

theirassets inequity; this reflectslevered equitypositions.

'Theearliestpossible reportingperiod-endfor thissampleisJanuary31, 1992,so thesix-monthbuyand hold

copycat calculations

commence

April 1, 1992. Thelatestpossiblereportingperiod-endfor thissampleisApril30,

1999,so thesix-monthbuyand hold copycatcalculationsendon

December

31,1 999. Thissamplecutoffimplies

that 1999only has3 fundsforsample 1 and 23 fundsforsample2.

The

final sampledoesnot necessarilyreflectall

oftherestrictivescreensineveryperiod. Ifafundmettherestrictions inJuly 1992,itwasincludedinthesample

evenifatothertimesduring1992- 1999itdid notmeettherestrictions.

(24)

Average

turnoveris similar,at 85

and

81 percent, respectively,inthe

two

samples.

Funds

inthehigh

expense

sample have

annual expensesthataverage 1.4percent

of

NAV,

versus0.9percentfor thebroader fiandsample.

On

average, thehigher

expense

ratio fiindsaresmallerthan thoseinthebroader sample,

holdslightlysmallercompanies,

and

employ

newer

portfoliomanagers.

The

funds inthehigh

expense

ratio

sample have

a slighdylongerperiod

between

disclosuresofportfolioholdings

and

asset allocation.

The

average

number

ofdisclosures to

Momingstar

overtheprevious twelve months,

and

theaverage

difference

between

the

sum

of

the portfolio equity weights

and

thereportedallocationto equities, are

similarforthe

two

samples.

4. Empirical Findings

In thissection,

we

reportourresults

on

therelative returns

on

copycat funds

and

on

their

primitiveactively-managedfunds.

We

begin

by

reportingourfindings for the

sample

of high-expense

funds,

and

we

then

move

on

tothebroader

sample

ofequityflinds.

We

conclude

by

reporting theresults

of

aregression analysis

of

the factorsthatexplainthedifferential

between

copycat

and

primitivereturns.

4.1 Resultsforthe

Sample

of

High Expense

Ratio

Funds

Table 2

summarizes

ourfindings

on

relativereturns

on

fundsinthehighexpenseratio

group

and

theircopycats. Panel

A

reports

mean

and

median

returnsineachofthesix

months between one

disclosuredate

and

the next,while panel

B

presents cumulativeresultsforperiodsof

between one and

six

months.

The

first

and

second

rows

present

summary

statistics

on

returns forbothprimitive

and

copycat

funds,while thethird

and

fourth

rows

present returndifferentials

between

the

two

setsoffunds.

Row

three considers returndifferentialsbefore expenses, while

row

fourconsidersthedifferential net

of

expenses.

The

mean

primitive returnisgreaterthanthe

mean

copycatfundreturninall but

one

ofthe six

months.

However,

only in

month

6isthe return differentialsignificant atthe.05 levelusinga two-tailed

test.

(We

usetwo-tailedteststhroughoutthe paper.) Neithertheabsolutenorthe

median monthly

return

differenceever exceeds

20

basispoints.

Mean

differencesremainfairly constantoverthesixmonths.

(25)

Consistentwith deteriorationinthe copycats'abilitytotracktheirunderlyingprimitive funds, the

difference

between

the copycat

and

theprimitivefiindreturns increasesovertime.

The

standard

deviationofthe

monthly

differences

between

the primitive

and

thecopycatreturn,while not

shown

inthe

table,averages

80

basispointsinthefirstthree

months

aftertheinformationdisclosure

and

rises to 110

basispointsinthelastthreemonths.

The

averageabsolute

monthly

return differences(alsonotreported)

alsoincrease

from 50

basispointsin

month

1 to80 basispointsin

month

6. Similarly, the correlation

between

actual

and

copycatreturnsdeclinesslightlyduringthesixmonths.

The

correlation coefficient

forthe

two

returnsaverages 0.98 duringthefirstthree

months

aftertheinformationdisclosure

and

0.96

duringthenextthreemonths.

When

boththe primitive

and

copycatfundreturns are adjustedfor theirestimated

monthly

expenses, the copycatfiindsoutperform theprimitivefundsinfourofthesixmonths.

However,

the

spread remainsinsignificantly different

from

zeroexceptin

month

4.

The

absolute valuesofthe

mean

return differences

remain

ator

below 20

basispoints.^

Panel

A

inTable 2presentsretumsfor individual

months between

portfoliodisclosures

by

primitive ftmds. Panel

B

translatesthese

monthly retums

intocumulative buy-and-hold retums.

Descriptive statisticsareprovided for

retums

over one- through six-month holdingperiods.

The

key

questionthattheseresultscan address concernsthe statisticalsignificanceofthecumulativereturn

differential

between

theprimitive

and

copycatflmds.

The

table

shows

thatthe difference

between

the

primitive

and

copycatfiind

retums

atthe

end of

thesix-month holdingperiod, net

of

expenses,isnot

significantlydifferent

from

zero.

The

mean

differenceatthe

end

ofsix

months on

thebefore-expense

returnis

42

basis points,

and

thisdifferential isstatisticallysignificantlydifferent

from

zero.

Thus

the

primitive funds

eam

ahigherreturnthanthe copycats.

On

a net-of-expensebasis,

however,

thecopycat

In futurework

we

plantodevelop

more

sophisticatedmeasuresoftherelativeriskinessoftheprimitiveand copycatfunds, includingboththevarianceofretums andthecovariancebetweentheretums andthemarket

portfolio.

The

results arerobusttotheassumptions abouttheretumsgeneratedbynon-equityinvestments.

We

computed

retumstoprimitiveand copycat funds underdifferentassumptions abouttheretumstonon-equityassets,suchas

(26)

fundreturnexceedstheprimitivefundreturn

by

an averageof

24

basispoints,

and

we

cannotrejectthe

nullhypothesisthatthe

two

cumulativereturnsare equal.

The

differences

between

the primitive

and

the

copycat fundreturns

grow

largerasthe return horizonislengthened.

The

onlystatisticallysignificant

differenceincumulativereturnsbefore expenses occurssix

months

afferthe informationdisclosure.^

The

results

from

thehigh

expense

ratio

sample

indicatethat acopycatfund cantracka primitive

fund

closely for

up

tosix

months

followingportfolio disclosures. Returndifferences

between

thecopycat

and

theassociatedprimitivefundarestatisticallyinsignificantregardless of whether

we

deduct

none of

theprimitiveftind'sexpenses,orall

of

theseexpenses,

from

itsreturns. Thisindicatesthatourfindings

areunaffected

by

choices

of

X,the fraction

of

the fund'sexpensesthat areattributabletoresearch

expenses,inequation(3)above. If

we

includedthe tax coststhattaxable investors face asaresult

of

turnover

by

actively

managed

fiinds,thenet-of-expensereturnadvantageforcopycatfunds

would

become

even

larger.

4.2 Results forthe

Broader

Sample of

Equity

Funds

Table3 presentsbothindividual

month

returns

and

cumulative returnsforthe

847

observationsin

our broader

sample

of

actively-managedfunds.

Our

findings

from

thehighexpenseratio

sample

carry

overtothebroader

sample

as well. Table3

shows

that

when

we

ignore expenses, primitivefiinds

outperformtheirassociatedcopycat fundsin five

of

thesix

months between

disclosureepisodes.

(The

return reversal occursinthethird

month

afterthe disclosure,

when

we

estimate the averagereturn

on

the

copycat fimdsto

be

greaterthanthat

on

theprimitive fiinds.)

The

differencein

monthly

returns,

however,

isneverstatisticallysignificantly different

from

zero,

and

it isneversubstantivelyverylarge.

When

fund expensesarededucted

from

the returnsof boththe primitive

and

copycat funds, the

mean

differencein

zeroreturnandassumingthatallnon-equityassetsearnedtheequityindexreturn. Theseassumptionsdid notaffect

ourbasic findingthatcopycatreturnsandprimitivefundreturns are notstatisticallysignificantlydifferent.

*Toaddressthe potentialproblemsassociatedwithasynchronousreportingofportfolioweights andthe overall allocationtoequities,

we

repeatedouranalysisexcludingtheone percent of funds withthe largest differences betweentheirportfolioweightsforequitiesandtheiroverallassetallocationto equities. Fortheremaining 179 observationsinthehighexpense sample,theabsolute differencewaslessthanor equaltoeightpercentagepoints, whichgives us greaterconfidencethatthecopycatfund'sassetallocationmimicstheunderlyingmutualfund's. The

conclusionsfromPanels

A

and

B

of Table2 arenot affectedbythissamplerestriction.

(27)

the

monthly

returnsisnegativefor five

of

sixmonths, butthedifferenceisstatisticallysignificantonlyin

month

three.

The

copycatfiindsgeneratehigher net-of-expensereturnsinallmonths, exceptthesixth

month,

between

disclosuredates.

The

cumulativereturnsinthelower panel of Table3

show

thatbefore expenses,theretumsto

holdingprimitive ratherthan copycat fundsareverysimilar. Aftersix

months,

thereisonlyathirteen

basispoint difference,

on

average,

between

the

two

sets

of

retums,

and

thisdifferenceisnotstatistically

significantly different

from

zero.

When

we

compute

the difference in

retums

net

of

expenses,however,

the averagereturn

on

the copycatfiindsishigher thanthe averagereturn

on

theprimitivefiinds,

and

we

canrejectthe nullhypothesisthat this differentialiszero.

Most

of

thedifferentialreturn infavor

of

the

copycat fimds

emerges

inthefirstfour

months

aftertheinformation disclosure. Aftersixmonths,the

cumulative

retum

differentialis 25 basispointsinfavorofthecopycatfiinds.

The

evidenceforthebroader

sample of

funds providesstrongersupportforthe

view

that copycat

fundscan outperformtheirprimitive funds, netof expenses, thanthe

comparable

resultsforthesample of

high

expense

ratiofiinds. This appearsto

be due

to thegreater

sample

size,

and

correspondinglysmaller

standarderrors,inthebroaderfiindsample.

The

cumulative

retum

differential

between

the primitive

and

thecopycatfiinds,netof expenses,is similarinthe

two

samples.

4.3 Investigating theSource

of

Retum

Differences

The

summary

statisticsinTables2

and

3 offerinsight

on

theviability

of

copycatfiindsasa

competitive altemativetotheirprimitivefiinds,butthey

do

notprovide

any

insight

on

the factorsthat

contributeto largerorsmaller

retum

differentials.

To

explorethis issue,

we

relatethecumulative

retum

differential beforeexpensestoa smallset

of

characteristicsoftheprimitivefiind.

These

characteristics,

while chosenina

somewhat

arbitraryfashion,aredesigned tocaptureboth factorsthat

might

mechanically leadtodifferences

between

thecopycat

and

primitive fiindretum, aswell as factors that

might

make

it

more

difficultfor thecopycattotracktheprimitive fiind.

We

alsorepeatedthisanalysisonasubsampleof fundsthatmeasureassetallocation

more

precisely,and againour conclusionsdidnotchange.

(28)

Table4presents the resultsofordinaryleastsquares regressions in thisspirit.

While

the analysis

isnotmotivated

by

a tightly-specified

model of

returndifferentials,thefindingsshould provide

some

guidancefor future

model

development.

The

dependentvariableforthe regressionsinthefirst

column

of

Table

4

isthepre-expensesix-month cumulativereturndifferential.

The

estimatesinthe

upper

panel

correspond tothehigh-expensefund sample, whilethoseinthelowerpanel,applyto thebroader sample.

The

explanatoryvariables include thepercentage

of

stocksthat

we

were

able to find

on

the

CRSP

return

tapes,turnover, theprimitivefund'sexpenseratio,itstotal assetvalue, the capitalizationofthe

companies

thatthefiindholds,

dummy

variables foreachyear,

and

a variablecapturingthemanager'stenure.

Inboththehigh-expense

and

broaderfimd samples,

we

find thatthedifferenceinpre-expense

returns

between

the primitivefund

and

thecopycatisdecreasinginthepercentage

of

fiind assetsinvested

in equities covered

by

CRSP.

Thisresultindicatesthat

when

the copycatflind isabletopositively

identifya higherfi-actionoftheprimitive fund's assets,thecopycatfund's return iscloserto thatofthe

primitive fund. Greateridentifiabilitydecreasestheability

of

aprimitive fund

manager

toearn superior

returns.

Most

oftheother variables that

we

includeintheregression

model have

astatistically

insignificanteffect

on

the returndifferential

between

primitiveand copycatfiinds.

The

second

column

of Table

4

presents theresultsofestimating

models

in

which

the dependent

variableistheabsolutevalue

of

thepre-expensereturndifferential

between

theprimitive

and

thecopycat

fiind. In this case,

we

again find thatthe share

of

the portfoliothatconsisted

of

CRSP-identifiable

equitiesisnegatively associatedwiththis

measure

of

the variation

between

theprimitive

and

thecopycat

fiind.

The

effectisstatisticallysignificant inboth thehigh-expense

sample

and

inthebroader sample, but

itis

much

largerinthehighexpense sample.

There

are also otherstatisticallysignificantcovariatesin

thiscase.

We

find, forexample,thattheabsolutevalue

of

thereturn differential isincreasinginthe

primitivefiind'sturnoverrate.

Copycat

fimds

on

averagetracktheprimitivefluidreturnslesswell

when

the primitivefiindsexhibithighturnover. This

makes

sense; thecopycatfimd

manager

ischasinga

"faster

moving

target"

when

theprimitivefiindhasahigher

tumover

rate.

We

alsofind,inthebroader

(29)

sample,thathigherexpenseratiosfortheprimitivefundsare associatedwithlargerabsolute return

differentials

between

the primitive

and

thecopycatfiinds, beforeexpenses.

The

resultsin thissection arestrictly descriptive. Nevertheless,they provide

some

guidance

on

the potential efficacyof copycat fundsintracking the returnsofactively

managed

equity funds.

We

do

notfind

any

largedifferences

between

the returns

of

the primitivefunds

and

thecopycat fimds inour

sample,

and

we

uncover

plausible patternsinthetype

of

actively-managed fundsthatcopycatsarelikely

to

have

the

most

successintracking.

5. Conclusions

Inthispaper,

we

constructhypothetical "copycat" fundsthat will

mimic

the portfolioofthe

associated "primitive" fundeach timetheprimitivefunddiscloses itsportfolioholdings.

We

findthatfor

a

broad sample

ofdiversifiedU.S. equity

mutual

funds overthe

1992-1999

period, theaveragereturns

before expenses

on

thecopycatfimdsare

lower

thanthecorrespondingreturns

on

the primitivefiands.

Whether

we

canrejectthenullhypothesisthatthe

two

setsof retums

have

the

same

mean

issensitiveto

our choiceof

sample

withrespecttoprimitive funds.

We

rejectthe equality

of

before-expenseretums in

a

sample of

high expenseratiofunds,while

we

do

notreject thisequalityforabroader

sample

of

funds.

However,

theretumsnetof expenses,

which

are the

retums

availabletoinvestors in

mutual

fiinds,are

higher

on

thecopycat funds than

on

the primitivefiinds. Thisistrue forbothsamples

of

funds,although

we

onlyrejectthenullhypothesis

of

equal

retums

forthebroad

sample

offunds.

The

disparity

between

thenet-of-expense retums

on

thecopycat

and

primitivefundsclearly

depends

on

theassumptionthat

we

make

abouttheexpensesassociatedwith

managing

acopycatfimd.

We

assume

thatthese expenses

would

be comparable

totheexpenses

of

currentindex fimds;iftheexpenses

were

actuallylarger, the

corresponding

retum

differential

would

besmaller.

Our

findings suggest thatcopycat fimds

have

the potentialtogenerate

retums

thatare roughly

comparable

totheretums

on

theprimitivefundsthattheyaredesignedtomimic. This suggeststhat

one

ofthecostsassociatedwith financialdisclosures,

namely

theprospect

of

competitorstrading

on

the

(30)

informationthatsuch disclosure revealsabouttheprimitive fund'sportfolio,

may

be substantial.

We

have

nottriedtoexplain

why

investorsholdactively

managed

equity funds,

and

we

have

not contributedtothe

debate

on whether

actively-managedfiinds, orcopycat funds based

on

thesefiinds,can outperform broad

market

benchmarks.

These

are large issuesthat

go

well

beyond

the current paper.

However,

we

have

shown

that

whatever

benefitinvestors inactively

managed

fundsthinkthattheyreceive

from

these

investments, intermsof subsequent retums, can beimitatedtoa substantialdegree

by

copycatfiinds.

Our

analysis considersonly

one

ofthe potential costs offinancialdisclosure, anditdoes not

attempttoquantify

any

of

the potential benefits associatedwithdisclosure.

As

such,itcannotbe

construedasprovidingultimateguidance

on

thedesign ofdisclosure regulations,although it

might

be an

inputtotheregulatory process.

Two

featuresof our study design

may

leadustooverstate the return

on

thecopycatftind,relative

tothe primitivefiind.

Both

reflectouranalysis ofhypotheticalcopycatfiinds,ratherthancopycatfiinds

thatactuallyoperateinthe

market and

thatprimitivefundsrecognize

and

respondto.

First,ifthe securitypurchases

by

the copycat funddrive

up

the prices ofsecuritiesalreadyheld

by

theprimitivefiind,thentheprimitivefund isina sense "frontrunning" purchases

by

thecopycatfiind.

This couldincrease the

retums

on

theprimitivefiind,particularly in thefirst

month

afterportfolio

holdingsare disclosed.

There

is

some

evidencethatfront-mnningofstockpurchases

by

mutualfiindscan

generate substantialretums.

Gasparino

(1997)reportsthatthe

Vanguard

Group

stoppedreporting

informationaboutthenetcash flowsinto itsfundsbecausethirdparties

were

apparentlyusingthis

informationto "frontrun"

Vanguard

funds. Ifapotentialinvestor

knew

thelargestholdings

of

agiven

Vanguard

fund,

and

healso

knew

thatthefund

had

experienceda largecash inflow,

he

might beableto

identifysecuritiesfor

which

there

would

be substantial

demand

inthenearfiiture. Fidelity,

which once

released dailyinformation

on

the size

of

some

of

itssectorfiinds,hassimilarlystoppedreportingthis

informationbecauseit

may

be

of usetoinvestors

who

aretryingto profit

from

thefund's prospective purchases.

We

are not

aware of any

evidencethatquantifies thepotentialretumstofront-running.

(31)

Second,ifcopycat funds

were

an importantfeature

of

themutual fundlandscape,

actively-managed

funds

might

take actionsthat

would

reducetheinformationcontent

of

theirsemiannual

disclosures. Thisisanalogoustotheproblem, described in

Lemer

(1995),thatfacesa firmthatplansto

patenta

new

technology,

when

thedetailsrevealedinthe patent applicationwillprovide valuableinsights

topotential competitors.

Ifthe

managers

ofprimitivefundscould

camouflage

theiractual portfolioholdings,this

would

potentiallyincrease the returndifferential

between

theprimitive

and

thecopycatfunds. If actively

managed

flinds

do

earnpositivereturns asaresultoftheirinformation, it

would presumably

raisethe

return

on

the primitivefimdsrelativeto thatofthecopycats.

There

is

no

consensusatpresent

on

the

extentto

which

mutual fluid

managers engage

in

"window

dressing,"orchangingthe composition oftheir

portfoliosnearthe

end of

a reporting period.

The

attractiveness

of

window

dressing

depends

on

the

transaction costsassociatedwith

moving

in

and

outofa particularsecurity,including

any

costs

of

executionassociatedwiththebid-askspread.

Such

a trading strategy

might

make

senseforlarge

companies

whose

shares are actively tradedinliquidmarkets.

For

smaller

and

less liquid securities,

we

suspectthatthetransaction costs

would

outweigh

thegain

from

dissembling

from

competitors.

Musto

(1999)findsevidence of

window

dressing

among money

market

fiindmanagers.

The

potential benefits

from

securityselectionintheequitymarkets probably exceeds the

comparable

benefitsinthe short-term

money

market; thissuggeststhat

window

dressing

may

alsoexistinthe equityfund market,as

O'Neal

(2001)suggests.

While

these

two

considerations

may

leadustounderstate the returnadvantage ofthe primitive

fiind,

one

other factoris likely tooperateinthereversedirection. Thisisourassumptionthatthecopycat

onlyobtainsinformationaboutthe primitive fund's holdingseverysixmonths. Inpractice,

we

know

that

many

actively

managed

fundsrevealinformation

more

oftenthanthis, and

presumably

a copycatfiind

tryingto

mimic

such afiind

would

beable to track the primitive's

retum

performance

more

closely.

The

issuesthat

we

have

discussedwithrespecttomutual fiindsalsoarise ina varietyofother

contexts

where

imitation can reducethevalue

of

initialresearch investments.

Tax

sheltersprovide

one

(32)

example. Sullivan(1999)writesthat "taxshelterscannot becopyrighted. Eventually, the

word

getsout

tootherclients, tocompetitors,

and even

totheIRS. It ishardto justify large feesfortaxshelters that

many

firms

market

..."

One way

that

law and

accounting firmsthatdesigntaxsheltersattempttoreduce

this diffusionof informationis

by

askingpotential clients to sign confidentialityagreements before they

leam

aboutthedetails

of

a potential shelter.

These

agreements

presumably

slow,but

do

not prevent, the

ultimate diffusionofinformation.

Our

analysis

of one

potentialcost

of

information disclosure,

and

ourdiscussionofother potential

costs,bears

on

only

one

side ofthebalancethat

must

ultimatelybeusedtodetermine optimal regultaory

policywithrespecttoinformationdisclosure.

The

most

important

need

forpolicydesignin thisareais

information

on

the potential benefitsthatinvestorsreceive

from

information disclosure. This

presumably

requiresinformation

on

thelikelihoodthatfund

managers

will

change

theirinvestmentobjectiveswithout

informingshareholders,

and on

the levelofdisclosurethat

would

takeplaceifthere

were no

regulatory

requirementsfor disclosure.

(33)

REFERENCES

Admati,

Anat and

Paul Pfleiderer (2000), "Forcing

Firms

toTalk: Financial DisclosureRegulation

and

Externalities,"

Review

ofFinancial Studies 13,479-519.

Carhart,

Mark,

Ron

Kaniel,

David

Musto,

and

Adam

Reed

(2001), "LeaningfortheTape:

Evidence

of

Gaming

Behavior

inEquity

Mutual

Funds," Journalof Finance(forthcoming).

Fairley, Juliette(1997),

"Keeping

More Under

TheirHats,"

New

York

Times

(June 15),p.F7.

Foster,

George

(1980), "Externalities

and

FinancialReporting," Journal of Finance35, 521-533.

Freeman, John

P.

and

StewartL.

Brown

(2001).

"Mutual

Fund

Advisory

Fees:

The

Cost

of

Conflicts

of

Interest,"Journal of Corporation

Law

26

(3),610-673.

Gasparino, Charles(1997). "Vanguard's

Cutback of

Fund

Data

May

Mean

ItFears a

Market

Drop,"

Wall

StreetJournal(October 14).

Gigler,

Frank

(1994), "Self-enforcingVoluntaryDisclosures,"Journal

of Accounting Research

32:2,

224-240.

Gmber,

Martin J. (1996), "AnotherPuzzle:

The Growth

of

Actively-Managed

Mutual

Funds,"

Joumal

of

Finance

51, 783-810.

Laderman,

Jeffrey

M.

(1999),

"A

Mutual

Fund

That

LetsItAll

Hang

Out,"Business

Week

(September

27), p. 126.

Lakonishok,Josef, AndreiShleifer,RichardThaler,

and

Robert

Vishny

(1991),

"Window

Dressing

by

Pension

Fund

Managers,"

American

Economic Review

81 (May), 227-231.

Lemer,

Josh(1995), "Patenting inthe

Shadow

of Competitors,"

Joumal

of

Law

and

Economics

38 (2),

463-495.

Manne, Henry

G. (1966). InsiderTrading

and

the Stock

Market

(New

York:

The

FreePress).

Musto, David

K. (1999),"Investment Decisions

Depend

on

PortfolioDisclosures,"

Joumal

of Finance 54 (June),935-952.

O'Neal,

Edward

(2001),

"Window

Dressing

and

Equity

Mutual

Funds,"

mimeo, Babcock

Graduate School

of

Management,

Wake

Forest University.

Sullivan,Martin (1999),

"One

Shelterata

Time?,"

Tax

Notes

(December

6), 1226-1229.

Verrecchia, Robert. (1983), "Discretionary Disclosure,"

Joumal

of

Accounting and

Economics

, 5, 179-194.

Wermers, Russ

(2001),

"The

Potential Effects

of

More

FrequentPortfolioDisclosure

on Mutual

Fund

Performance," Investment

Company

InstitutePerspective 7 (June).

Figure

Table 1: Summary Statistics for Samples Using Information from Mandatory Disclosure Periods (1/92^/99)
Table 1 (continued): Summary Statistics for Samples Using Information from Mandatory
Table 4: Regression Results Explaining Difference in 6-Month Buy and Hold Returns Between Primitive and Copycat Funds

Références

Documents relatifs

%20in%20Europe%202010.pdf (analyzing a database of European activist interventions and finding mean abnormal returns of 4.4% around the disclosure date); see also Bebchuk et al.,

In the context of financial markets and the indirect regulation of hedge funds, regulatory competition induced by entrusting a relatively large num- ber of prime brokers

The delimitation of these conflict of interest areas has been done assuming that invested amounts are lower than $50,000. Nevertheless, depending on the importance of the

Takes note that no supplementary voluntary contributions to the Intangible Cultural Heritage Fund have yet been received in line with the two funding priorities approved

Mis-alignment of objectives of the fund manager and shareholders results in discount/premium, and (2) for a given risk aversion parameters, the diversification benefit to

The model shows that the fund share trades at a discount or a premium, depending on the risk aversion parameters of investor 1 and the manager.. On open-ending the fund, fund

In an empirical study conducted on a sample of 12 hedge funds, Lo (2002) noticed that the serial correlation in monthly hedge fund returns can overestimate the Sharpe ratio by up

The copycat strategy is perceptually effective for solving the correspondence problem of associating a shadow with its caster; copycatting can be more effective than other