74 views

Uploaded by kashif salman

save

- Ch03
- MEC-001 - Micronomic Analysis 2015
- Cbse Class 12 Eco
- Information Regulation (Including Regulation of Advertising)
- 4. Equilibrium Final(2)
- Posner.economic.analysis
- Learning Curve 3
- var paper
- A Future for Labour in the Global Economy (Bowles S., 2001)
- Nelson e Winter_an Evolutionary Economic Change_1982_livro
- ECO100Y1_Final_2009W
- Sol_Ch4
- Economics 101Chapter06
- macroeconomic Chapter 2 - DD & SS
- The Number That Killed Us - Presentation[1]
- 7 consumers_producers.ppt
- Pearl Essay
- Economic Solution
- cbse economics XII sample paper -4.doc
- 0. a New Economic Theory for Space Exploration
- mkw_chap_03
- Chap07
- APE_Unit_I_Prax_Test_Answers
- Chapter_7.ppt
- Analysis of Demand and Demand Forecasting
- NestedSimulationForPortfolioRiskMeasurement.pdf
- harald
- Homework Number 2
- Calvo, G (1988). Servicing the Public Debt- The Role of Expectations
- Ch03 Parkin Microeconomics
- imapct of selected hr parctices on percieved emplyee performance
- Children Attitude Toward TVC in Pak
- Introduction to Microsoft Office
- leadership of a culture change process
- gender and ownership in uk small firms
- women enterpreneurs- management skills and business problems
- leadership theory
- organisational support, employee development and commitment
- challenges and prospects of hrm practices in developing countries
- organisational knowledge leadership- grounded theory approach-2007
- sunflower
- leadership preferences in japan 2007
- uk university article
- organisational commitment among pakistani university teachers
- transformational leadership
- article4
- afshan 4
- holy places
- appraisal2
- article12
- markus
- alien definition
- advertising create a goodwill of producate
- advertising planning
- appraisal1
- appraisal employee development
- advertising under varticle producat differenate
- appraisal
- advertising bans
- max planck

You are on page 1of 30

Prolonged Mispricing

EITAN GOLDMANand STEVE L. SLEZAK

n

ABSTRACT

This paper examines how information becomes reflected in prices when in-

vestment decisions are delegated to fund managers whose tenure may be

shorter than the time it takes for their private information to become public.

We consider a sequence of managers, where each subsequent manager inherits

the portfolio of their predecessor.We showthat the inherited portfolio distorts

the subsequent manager’s incentive to trade on long-term information. This

allows erroneous past information to persist, causing mispricing similar to a

bubble. We investigate the magnitude of the mispricing. In addition, we exam-

ine endogenous information quality. In some cases, information quality in-

creases when the manager’s expected tenure decreases.

MORE THAN EVER BEFORE, investors delegate the management of their investment

portfolios to professional fund managers.

1

Due to mobility and turnover, the ex-

pected tenure of a manager may be shorter than the investment horizon of a typi-

cal investor.

2

Given this horizon mismatch, fund managers may have less

incentive to collect and trade on long-terminformationthan do the investors they

trade for. As a result, the prices in markets with more delegated investment will

reflect less long-term information than those markets where long-lived investors

manage their own portfolios. The amount of long-term information that is re-

flected in prices is likely also to change with market conditions. As a manager’s

THE JOURNAL OF FINANCE v VOL. LVIII, NO. 1 v FEB. 2003

n

Goldman is from the Kenan-Flagler Business School, University of North Carolina at Cha-

pel Hill and Slezak is from the College of Business Administration, University of Cincinnati.

We gratefully acknowledge the helpful comments of Douglas Breeden, Jennifer Conrad, Rick

Green (the editor), Harold Zhang, the anonymous referee, and the seminar participants at the

University of North Carolina at Chapel Hill. We are solely responsible for any errors that

might persist.

1

Allen and Gorton (1993) cite statistics that indicate that institutional ownership has in-

creased from 6.5 percent in 1965 to 45 percent in 1991. Since then, the percentage of institu-

tional ownership has increased further. The Conference Board provides information on

institutional ownership which shows that the percentage of institutional ownership peaked

at 59.9 percent by the end of 1997, falling only slightly to 57.6 percent by the third quarter of

1999 (see The Conference Board news release by Carolyn Kay Brancato at http://www.news-

wise.com/articles/2000/4/INVEST.TCB.html).

2

Chevalier and Ellison (1999) report that, in a sample of growth and growth and income

funds, 18 percent of the funds changed managers from 1993 to 1994. Thus, in their sample,

approximately one-fifth of the managers had tenures less than two years, far shorter than

the investment life of a typical investor.

283

expected tenure decreases due to increased mobility, or as the time it takes for

information to become public increases due to greater uncertainty, the incentive

to collect long-term information and hence the informational efficiency of stock

prices may decline.

We formalize the above argument byconsidering a model inwhichthe probabil-

ity that the manager’s private information will become public during her tenure

may be less than one. We refer to this probability as the revelation probability. We

examine a sequence of managers where each new manager inherits the portfolio

of the preceding manager. We assume that managers are compensated according

to the change in the value of the portfolio they manage during their tenure. The

model confirms that the incentive to trade on long-terminformation is increasing

in the revelation probability. However, the model also provides additional in-

sights. We show that the inherited position can create a lock-in effect whereby

the current manager may not sell on bad news or buy on good news.

3

This lock-

in effect cancause erroneous information fromthe past to have a prolongedeffect

on the stock price, even beyond the point in time at which the truth in past infor-

mation is publicly revealed.

The intuition for these results can be obtained by recognizing that the change

inthe value of the fund during the manager’s tenure is the trading profits plus the

change in the value of the inherited position. If the current manager inherits a

long position in a security for which she receives negative information, then she

faces a trade-off in determining howbest to trade. On the one hand, selling is the

best policy if her private information will be revealed during her tenure. This is

true because selling maximizes trade profits and has no adverse effect on the

value of her inherited position. The latter fact holds because the portfolio’s end-

of-tenure value fully reflects her information regardless of her trades. On the

other hand, selling will have a negative effect on the value of the fund if her pri-

vate informationwill not be revealed during her tenure. In this case, selling has a

neutral effect on her trade profits because the price she sells at prevails until the

end of her tenure. However, selling lowers the end-of-tenure price and causes the

value of the inherited position to decline. In summary, if her information is re-

vealed, selling increases trade profits and has a neutral effect on the inherited

portfolio, but when her information is not revealed, selling has a neutral effect

on her trade profits and a negative effect on the inherited portfolio. If it is suffi-

cientlylikely that her informationwill not be revealed, thenthe negative effect on

the value of the inherited portfolio will dominate the potential trade profits. As a

result, she will not sell on her negative information and the stock price will con-

tinue to reflect the positive information that prevailed when the prior long posi-

tion was established. If this past information contains errors, then the stock

price will persistently reflect these errors in a manner similar to a speculative

bubble.

3

This effect is similar to the lock-in effect associated with capital gains taxation, where

owners of appreciated securities who possess negative information fail to sell in order to

avoid taxes (see Constantinides (1983, 1984)).

The Journal of Finance 284

In the paper, we provide cross-sectional implications for stocks that differ in

the degree of institutional ownership. We show that firms with high initial insti-

tutional ownership are, on average, firms with overly optimistic initial signals.

These firms are the ones that are more likely to be persistently overvalued. We

also show that there will be more mispricing in stocks with a single institutional

investor relative to those with multiple institutional investors. This is consistent

with the literature on analyst following and market efficiency (Easley et al.

(1996)). Finally, we showthat the degree of mispricing is increasing in the quality

of the signals that prior managers receive. As the quality of the signal obtained

by the prior manager increases, the previous manager takes a more extreme posi-

tion in the stock. The more extreme is the prior position, the larger is its impact

on the trades of the current manager and the larger the mispricing. This effect

implies that the degree of mispricing will become more pronounced as institu-

tions collect higher quality information or as the time it takes for uncertainty

to be resolved and publicly revealed increases.

We also show that the inherited position and the revelation probability to-

gether affect the current manager’s incentive to collect higher quality informa-

tion. We identify two such effects. First, for a given signal-to-noise ratio, as the

revelation probability falls, the likelihood of realizing a positive trading profit

declines, thus reducing the incentive to trade and collect information. Second,

as the revelation probability falls, the inherited position has a greater impact

on the current manager’s trade. We show that this implies that current net order

flow contains more noise. This allows the current manager to hide her private

information better, increasing the incentive to collect information. Taking both

effects together, we show that a decline in the revelation probability can actually

lead to an improvement in private information qualityand price informativeness.

Finally, we identify several implications of the model that are consistent with a

variety of recently documented empirical regularities. First, the model generates

behavior that is consistent with the glamour-versus-value anomaly (Lakonishok,

Shleifer, andVishny (1994)) in which low-P/E stocks outperform high-P/E stocks.

Second, the model generates phenomena similar to speculative bubbles and the

recent run-up and decline of Internet stocks. Third, managers in the model en-

gage in positive-feedback trading as in De Long et al. (1990a) and Grinblatt, Tit-

man, and Wermers (1995). In contrast to the behavioral interpretations, in our

model positive-feedback trading is a rational consequence of delegated portfolio

management. Fourth, the model provides insight into the relationship between

past performance and subsequent risk taking recently documented in the empiri-

cal tournament literature (see Brown, Harlow, and Starks (1996), Chevalier and

Ellison (1997), Coval and Shumway (2000), and Busse (2001)).

Our paper contributes to the theoretical literature that examines when stock

prices solely reflect information about fundamental value. Our model belongs to

the branch of the literature that considers only rational agents.

4

This branch

4

A recent segment of the literature includes models with irrational agents (see, e.g.,

De Long et al. (1990a, 1990b), Barberis, Shleifer, and Vishny (1998), Daniel, Hirshleifer, and

Subrahmanyam (1998), Odean (1998), and Hong and Stein (1999)).

Delegated Portfolio Management and Rational Prolonged Mispricing 285

includes papers on speculative bubbles that establish that bubbles do not exist

under common knowledge and/or symmetric information (see Tirole (1982)) ex-

cept when a bubble completes the market (as in Samuelson (1958), Diamond

(1965), Gale (1973), Bewley (1980), and Santos and Woodford (1997)) or when its

growth is bounded by the growth of the real economy (as in Tirole (1985)).

Bubbles, however, can exist when common knowledge fails (see Brunnermeier

(2001) for a detailed survey) or when asymmetric information or moral

hazard exists (see Allen and Gorton (1993), Allen, Morris, and Postlewaite

(1993), and Dow and Gorton (1994)). Our paper belongs to this latter

group.

Our paper is most closely related to Dowand Gorton (1994). In their model, the

agents have investment horizons that are shorter than the time it takes for their

information to become public.While both Dowand Gorton’s and our paper exam-

ine the pricing impact of agency and time-horizon, our paper differs in the direc-

tion of the informational externality. In Dow and Gorton, the possibility that

future speculators will not have a sufficient incentive to trade on information

leads current speculators not to trade on long-term, value-relevant information.

Thus, future speculators generate an informational externality on current

speculators. In contrast, in our model, the previous portfolio position creates

an informational externality that distorts the current manager’s incentive to

trade.

Our paper is also related to the literature on trade-based price manipulation

(see, e.g., Allen and Gale (1992), Jarrow (1992), Kumar and Seppi (1992), and

Fishman and Hagerty (1995)). Profitable trade-based manipulation is possible

when an agent can, by trading, cause a price change that does not fully reverse

when he unwinds his position. This is possible when the relationship between

trade size and price impact is asymmetric around zero. The above papers provide

structural models in which this type of asymmetry occurs in equilibrium. In our

model, the fund manager also manipulates the stock price. However, unlike in

the manipulation literature, the agent in our model does not unwind her position

in order to profit. Rather, she receives compensation based on the interim ma-

nipulated price, passing her position (and the consequences associated with un-

winding it) on to the next manager. Thus, as in the manipulation literature, our

manager instigates a price impact without having to bear the full cost of its re-

versal.

The paper is organized as follows. In Section I, we describe the model.

In Section II, we solve for, analyze, and provide intuition on the equilibrium

when private information quality is exogenous. In Section III, we examine

the resulting properties of the equilibriumprice process under fixed information

quality. In Section IV, we endogenize information quality and examine the

impact of the revelation probability and the fund’s prior position on the in-

centive for current managers to collect information. In Section V, we discuss

the robustness of our results when there are multiple informed fund

managers competing in the same security. In SectionVI, we relate the behavior

generated by the model to existing empirical regularities. Section VII

concludes.

The Journal of Finance 286

I. The Model

There are three types of risk-neutral agents: (1) fund managers, (2) liquidity

traders, and (3) a competitive market maker. Over multiple discrete dates, the

agents trade two securities: a risk-free bond and a risky stock. The following sec-

tions discuss each element of the model in detail.

A. Managers

A.1. Timing

We consider a succession of managers, with each manager staying with the

fund for one unit of time. At t 50, the fund is formed with an initial endowment

of cash and zero shares of the risky asset. At the same time, an initial manager,

A0, is installed. This manager collects/receives private information about the fu-

ture value of the risky security and, based on this information, trades to create

the fund’s initial position in the risky asset, X

I

0

. Following that trade (i.e., at

t 50

1

), the manager leaves the fund. Immediately following this point at t 51 a

new fund manager, A1, is installed. This new manager collects her own private

information regarding the future value of the risky securityand decides howbest

to trade. As with the first manager, this manager also leaves the firm after her

trades clear (i.e., at t 51

1

).

A.2. Information about FutureValue

We assume that the value of the risky security at some future date beyond

t 51

1

is V =

VV ÷ Z

0

÷ Z

1

, where

VV is a constant and Z

0

and Z

1

are independent

mean zero normal random variables with variances s

2

Z

0

and s

2

Z

1

, respectively. The

variables

VV, s

2

Z

0

, and s

2

Z

1

are common knowledge. The first manager, A0, receives

information about Z

0

, while the second manager, A1, receives information about

Z

1

. ManagerA0 receives a noisy signal y

0

5Z

0

1e

0

, where e

0

~ N(0; s

2

e

0

) is indepen-

dent of Z

0

, while manager A1 receives the noisy signal y

1

5Z

1

1e

1

, where e

1

~

N(0; s

2

e

1

) is independent of Z

0

, Z

1

, and e

0

.

We make the following assumption regarding the resolution of uncertainty

about Z

0

.

ASSUMPTION 1: At t 50

1

(i.e., after the first manager trades and before the new man-

agerarrives) Z

0

becomes publicknowledge. However, neither y

0

nor e

0

are ever revealed.

This assumption is made for two reasons. First, it specifies that past information

about the true value of the firm is revealed over time while the specific noisy in-

formation possessed by the manager is not. Second, this assumption allows us to

focus on the incentives of the second manager in the most innocuous situation in

which the inherited portfolio does not start out mispriced. If Z

0

is publicly re-

vealed at t 50

1

, by the time A1 is installed, the market price of the risky asset is

VV ÷ Z

0

, which is the expected per-share terminal value given all available infor-

mation at that time. Thus, when A1 joins the fund, she inherits a portfolio that is

Delegated Portfolio Management and Rational Prolonged Mispricing 287

appropriately valued. As a consequence, the results we obtain on prolonged mis-

pricing do not stem from initial mispricing.

To capture the feature that a manager’s tenure may be shorter than the

time it takes for information to become public, we assume that there is some

probability that the truth in the second manager’s information will not become

public before she leaves the fund at t 51

1

. In particular, we adopt the following

assumption.

ASSUMPTION 2: The stock price at t 51

1

is given by

P

1

÷ =

V with probability d

P

1

with probability 1 ÷ d

;

where P

1

is the market price that results given the realized net order flowat t 51.

Thus, with probability d, the true value of the security is revealed while, with

probability 1 ÷d, the price at t 51

1

reflects only the information already con-

tained in P

1

.

Companies that rely heavily on research and development (e.g., Internet-re-

lated stocks or pharmaceuticals and biotechnology) are low d stocks. Stocks in

fairly stable industries for which the future business environment is

fairly predictable and whose assets are more tangible are high d stocks. Although

d is a characteristic of a security, d can also be a feature of the fund manager

and/or of prevailing general economic conditions. For example, d declines as a

fund manager ages and moves closer to retirement. It also decreases during

periods of exuberant economic activity characterized by increased managerial

turnover.

Assumptions 1 and 2 create an asymmetry between the first and second man-

ager in that the truth in the first manager’s information is always revealed prior

to her departure while the second manager’s information is only revealed with

some probability. Rather than consider a history of many prior managers, we

use a single‘‘first’’ manager to represent all prior managers. Given this interpre-

tation, it is reasonable to assume that some of the truth in the old information

possessed by a long-since departed manager will have been revealed by the time

a new manager is installed. If, under these assumptions, we can show that pro-

longed mispricing obtains, then it will also obtain for a longer sequence of man-

agers who each face do1.

A.3. Incentive Compensation

Managers have an incentive to collect and appropriately trade on information

due to performance-based compensation. We make the following assumption re-

garding the manager’s objective.

ASSUMPTION 3: Each manager trades to maximize the expected difference between

the value of the fund’s portfolio immediately before her first trade (but after the prior

The Journal of Finance 288

manager’s last trade) and the value of her portfolio when she leaves (prior to the next

manager’s trading).

This assumption has two important features. First, the manager’s performance

is measured by the change in the value of the fund (or, equivalently, the fund’s

return) only over the interval in which she manages the fund.

5

Second, the man-

ager’s objective function is linear in performance. This assumption is consistent

with the manager receiving all or part of any performance-based fee received by

the fund (as distinct from the manager), where the performance fee satisfies the

restrictions imposed by the Investment Advisor Act of 1940 and its amendments.

This act specifies that a mutual fund can assess its investors a ‘‘fulcrum’’-type

performance fee. Fulcrum fees are those that are linear and symmetric about a

prespecified benchmark. Our assumed objective function corresponds to a ful-

crum fee with a nonstochastic benchmark. Because our managers are risk neu-

tral, this objective function is also consistent with performance measured

relative to a stochastic benchmark (e.g., the return on the S&P 500) if the man-

ager’s trade has a negligible impact on that benchmark.

Some papers suggest that even though fees are restricted to be linear, a fund’s

net income may be nonlinear in performance. Chevalier and Ellison (1997) show

that fund size is typically nonlinear in the fund’s past performance. Thus, if the

net income of the fund increases with fund size (e.g., due to increasing returns to

scale), the objective of the fund and its manager may be nonlinear in perfor-

mance. This is important because nonlinear contracts may introduce secondary

motives for trade, such as variance manipulation. Here we focus onthe first-order

effect of manipulation of the first moment. Nonlinear contracts may potentially

impact the magnitude of the pricing bias we identify, but not its existence.

B. LiquidityTraders and Market Makers

Liquidity traders are assumed to be nonstrategic where the liquidity demand

at date t is a randomvariable: X

L

t

~ N(0; s

2

L

t

). As in Kyle (1985), the orders of the

manager and the liquidity traders are combined, with the market maker obser-

ving only the aggregate net order flow, w

t

5X

t

I

1X

t

L

.The market maker then satis-

fies the individual orders at the prespecified net-order-flow-contingent price.

Since we assume that the market maker is competitive, the price function is sim-

ply the expectationof the future value of the asset giventhe net order flowand all

other relevant information available to the market maker. Here we consider line-

ar price functions of the following form:

P

t

= E(V[O

t

) ÷ l

t

(w

t

÷ E(w

t

[O

t

)); (1)

5

An alternative contract could specify that her compensation be based on the change in the

portfolio value between when she arrives and after the truth in her information (Z

0

) is fully

revealed. While this type of contract will eliminate prolonged mispricing, this type of contract

is never observed in practice since it exposes the current manager to risks not under her con-

trol, namely the trades of predecessors. In addition, this type of contract is not feasible since

the date at which the truth in the manager’s information is fully revealed is not observable

and therefore not contractible.

Delegated Portfolio Management and Rational Prolonged Mispricing 289

where l

t

is the inverse of the market depth and O

t

is public information at t ex-

cluding w

t

.

We further assume that the market maker does not observe the initial position

of the fund or the information used to choose it. However, the market maker has

rational expectations about the distribution of the initial position. As will be

shown in Section II.B below, if the market maker knows the initial position, then

he will be able to‘‘undo’’any trade biases it creates.

Figure 1 provides a summary timeline for all of the events in the model.

II. Equilibrium

In this section, we solve for the equilibrium of the multiperiod trade game.

Since each manager maximizes only the change in the value of the portfolio dur-

ing her tenure, the first manager’s problemcan be solved separately fromthe sec-

ond manager’s problem. However, the second manager must take as given the

trade of the first manager. Thus, we solve the first-period problem first and then

use the result to solve the second-period problem.

A. Equilibriumin the First Period

The initial manager chooses an order X

I

0

to solve the following optimization

problem:

max

X

I

0

E[X

I

0

(

VV ÷ Z

0

÷ P

0

)[y

0

[; (2)

Figure1. Sequence of events. This figure shows the sequence of events in the model. The

first manager is denoted by A0 and the second manager is denoted by A1. The trades of a

fund manager at time t are denoted X

I

t

. The trades of the liquidity traders at time t are

denoted X

L

t

. The net order flowat time t is denoted w

t

and is equal to the sum of the man-

ager’s trade and the liquidity trades at time t (i.e., w

t

= X

I

t

÷ X

L

t

). The terminal value of the

asset is V =

VV ÷ Z

0

÷ Z

1

. The first fund manager receives a signal y

0

5Z

0

+e

0

while the sec-

ond fund manager receives a signal y

1

5Z

1

+e

1

.The price of the asset at time t is denoted P

t

.

Finally, E([) denotes a conditional expectation and d is the revelation probability.

The Journal of Finance 290

subject to

P

0

=

VV ÷ l

0

(w

0

÷ E(w

0

[O

0

));

w

0

= X

I

0

÷ X

L

0

:

Here the objective function for A0 is simply the expected value of her portfolio,

net of its initial cost. The constraints reflect her rational expectations about the

expected effect her trade will have on the transaction price. Because the fund

initially holds zero shares of the risky asset and Z

0

is publicly announced right

before A0 leaves the fund, this problem is essentially the same as in a static Kyle

(1985) model. The manager’s optimal demand is

X

I

0

=

1

2l

0

g

0

y

0

; (3)

where g

0

= s

2

Z

0

=s

2

y

0

and s

2

y

0

= s

2

Z

0

÷ s

2

e

0

. Knowing the formof this demand, the mar-

ket maker sets market depth l

0

such that

P

0

= E[V[w

0

= X

I

0

÷ X

L

0

[ =

VV ÷ l

0

(w

0

÷ E(w

0

[O

0

)): (4)

Theorem1 specifies the equilibrium l

0

.

THEOREM1: In the rational expectation equilibriumof the first period, the equilibrium

price function parameter is

l

0

=

1

2

s

2

Z

0

s

L

0

s

y

0

: (5)

Proof: The proof is a straightforward extension of Kyle (1985) and is, therefore,

omitted.

B. Equilibriumin the Second Period

We begin by analyzing the optimization problem of the second fund manager,

A1. Based on her signal, y

1

, and the portfolio she inherits, X

I

0

, she chooses her

trade, X

I

1

, so as to maximize the expected change in the value of her portfolio

from the time she was hired to the time she leaves. Because Z

0

is revealed prior

to her joining the fund, the value of the portfolio she inherits is W

1

= X

I

0

(

VV ÷ Z

0

).

Given the trade X

I

1

and the realized prices at t 51 and t 51

1

, the value of her

portfolio at t 51

1

when she leaves is simply W

1

÷ = X

I

0

P

1

÷ ÷ X

I

1

(P

1

÷ P

1

). Hence,

the realized change in the value of her portfolio is

DW = W

1

÷ ÷ W

1

= X

I

0

[P

1

÷ ÷ (

VV ÷ Z

0

)[ ÷ X

I

1

(P

1

÷ ÷ P

1

): (6)

As shown intheAppendix, substituting the price function P

1

=

VV ÷ Z

0

÷ l

1

(w

1

÷

E[w

1

[Z

0

; w

0

[) into equation (6) and taking expectations shows that the

Delegated Portfolio Management and Rational Prolonged Mispricing 291

manager’s objective depends on her trade X

I

1

and her inherited position X

I

0

as

follows:

E(DW[X

I

1

; y

1

; X

I

0

; Z

0

) =X

I

0

dg

1

y

1

÷ (1 ÷ d)l

1

[X

I

1

÷ E(w

1

[Z

0

; w

0

)[

÷ X

I

1

d g

1

y

1

÷ l

1

[X

I

1

÷ E(w

1

[Z

0

; w

0

)[

;

(7)

where g

1

= s

2

Z

1

=s

2

y

1

.

Equations (6) and (7) decompose the change in the portfolio value into two

parts: The first term on the right-hand side is the change in the value of her

inherited position, and the second term on the right-hand side is her trade

profit. There exists a trade that maximizes the trading profits as in Kyle (1985).

However, this trade may adversely affect the value of the inherited position. For

example, if the current manager has negative information and has inherited a

positive position, the trade that maximizes her trade profit will lower the value

of her inherited position. In the absence of an inherited position, the manager

would sell. However, in the event that her private information does not

become public (thus P

1

÷ = P

1

), her selling will lower the end-of-tenure price

P

1

÷, leading to a decrease in the value of her inherited long position. Thus, the

optimal trade that maximizes the sumof the two components will be less negative

than if she had inherited a zero position. Unless the information is sufficiently

negative, she may not sell since the trade profits from selling may be too small

relative to the loss in the value of the inherited position. On the other hand,

if she inherits a short position, then a lower end-of-tenure price will improve

the value of the inherited position and her optimal demand will be even more

negative.

The first-order conditionwith respect to X

I

1

implies that the manager’s optimal

trade is

X

I

1

=

g

1

2l

1

y

1

÷

1 ÷ d

2d

X

I

0

÷

E[w

1

[Z

0

; w

0

[

2

:

Given this trade function,

E[w

1

[Z

0

; w

0

[ =

1 ÷ d

d

E[X

I

0

[Z

0

; w

0

[:

and the optimal trade function becomes

X

I

1

=

g

1

2l

1

y

1

÷

(1 ÷ d)

2d

X

I

0

÷ E(X

I

0

[Z

0

; w

0

)

: (8)

Since the market maker extracts E[w

1

[Z

0

; w

0

[ from w

1

when setting the stock

price, the inherited position impacts the price only to the extent that it differs

fromits expectation givenpublic information, E[X

I

0

[Z

0

; w

0

[.That is, themanager’s

trade impacts the unexpected net order flow by X

I

1

÷ E(w

1

[Z

0

; w

0

) = X

I

1

÷

1÷d

d

E[X

I

0

[Z

0

; w

0

[ =

g

1

2l

1

y

1

÷

(1÷d)

2d

D

0

, where D

0

= X

I

0

÷ E[X

I

0

[w

0

; Z

0

[ is the‘‘initial-po-

sition deviation,’’defined as the deviationof the actual inherited position fromits

conditional expectation. The net demand then consists of two parts: The first

part reflects relevant private information, and the second part contains noise

The Journal of Finance 292

from the inherited position D

0

.

6

In addition, if the initial position X

I

0

exceeds its

expectation, then D

0

is positive and the signal has to be sufficiently negative in

order for the manager’s trade to lower the price on average. If it is not sufficiently

negative, the manager will forgo trade profits and buy in order to increase the

value of her inherited position. Thus, the inherited position creates a lock-in ef-

fect whereby negative information possessed by the manager does not get ex-

pressed to the market via selling. A similar argument implies that positive

information will not get expressed if the manager inherits a short position.

The equilibriumprice functionunder competitive market making is simply the

conditional expectation of the terminal value given the information content of

the realized unexpected net order flow, w

1

÷ E[w

1

[Z

0

; w

0

[. Specifically, we have

the following theorem.

THEOREM 2: The equilibriumprice function at t 51 is

P

1

=

VV ÷ Z

0

÷ l

1

[w

1

÷ E(w

1

[w

0

; Z

0

)[; (9)

where l

1

is as follows:

l

1

=

1

2

s

2

Z

1

s

y

1

1

[s

2

L

1

÷ (

1÷d

2d

)

2

Var(e

0

[Z

0

; o

0

)[

0:5

; (10)

with Var(e

0

[Z

0

; o

0

) = (s

2

L

0

s

2

e

0

)=(s

2

y

0

÷ s

2

e

0

).

Proof: See the Appendix.

The features of the equilibrium are discussed in detail in the next section.

III. Price Properties and Mispricing

A. The Effect of the Inherited Position and Prolonged Mispricing

To highlight the dependence of the current price on the inherited position, the

equilibrium t 51 price function can be written as follows:

P

1

=

VV ÷ Z

0

÷ 0:5

s

2

Z

1

s

2

y

1

y

1

÷ l

1

X

L

1

÷

1 ÷ d

2d

l

1

D

0

: (11)

As in a standard Kyle-type model, half of the pertinent information possessed by

the informed agent (i.e., (s

2

Z

1

=s

2

y

1

)y

1

) is reflected in the price while the current

liquidity shock X

L

1

creates noise. In addition to these standard components,

6

If the initial position is observable by the market, then there is no noise and D

0

50. How-

ever, an alternative formulation is obtained if the inherited position is observable but d is

random. In this case, from the perspective of the market maker, the optimal demand of the

current manager remains a function of two unknown variables (now y

1

and d rather than y

1

and X

I

1

). If the market maker knows the inherited position but not d, then the second term in

the net demand remains unknown and the qualitative results obtained herein will be sus-

tained.

Delegated Portfolio Management and Rational Prolonged Mispricing 293

the inherited position creates a bias in the price. For a given initial-position

deviation D

0

, the expected difference between the terminal value,V, and the cur-

rent price, P

1

, is simply

E(V ÷ P

1

[D

0

) = ÷l

1

1 ÷ d

2d

D

0

: (12)

Thus, the price of a security that is held bya manager who inherits a larger-than-

expected position (i.e., one for which D

0

40) will be biased upward on average.

Similarly, when the current manager holds a smaller-than-expected position,

the security will be undervalued on average.

Substituting the closed-form expression for the initial-position deviation into

the price function allows us to rewrite equation (11) as a function of the noise in

A0’s signal, e

0

:

P

1

=

VV ÷ Z

0

÷ 0:5

s

2

Z

1

s

2

y

1

y

1

÷ l

1

X

L

1

÷

1 ÷ d

2d

l

1

s

2

y

0

s

2

y

0

÷ s

2

e

0

e

0

÷

s

y

0

s

L

0

s

2

e

0

s

2

y

0

÷ s

2

e

0

X

L

0

¸ ¸

: (13)

This expression reveals that the signal error of the previous manager, e

0

, has a

long-term impact on the stock price even after Z

0

is revealed. Hereafter we refer

to this feature as prolonged mispricing. Prolonged mispricing occurs because X

I

1

depends upon the inherited position, which, in turn, depends upon the previous

manager’s noisy signal.

There are four situations in which there is no prolonged mispricing. First,

when there is no noise in the prior manager’s signal (i.e., s

2

e

0

= 0 and e

0

50) the

market maker can use Z

0

to perfectly infer the inherited position. In that case,

D

0

= X

I

0

÷ E[X

I

0

[w

0

; Z

0

[ = 0 for all realizations of Z

0

and w

0

and no bias is caused

by the inherited position.

7

Second, when there is no liquidity noise in the first

period, the past net order flow perfectly reveals the inherited position and D

0

=

X

I

0

÷ E[X

I

0

[w

0

; Z

0

[ = 0 for all realizations of the prior signal and net order flow.

Third, when the revelation probability d is one, the inherited position has no

impact on the current manager’s trade. As a result, the current net order flow is

unaffected by the past signal error.The fourthcase is trivial and occurs whenthe

realized value of D

0

50.

Because D

0

is mean zero, equation (12) implies that the unconditional expected

deviation between price and terminal value is zero.This does not imply that there

is no bias for each individual security, however. Rather, it implies that in a cross

section of securities, the pool will not be biased on average. That is, the uncondi-

tional expected bias corresponds to the mean of the deviations that occur in a

7

Mutual funds are required to periodically disclose portfolio holdings. If these disclosure

times coincide with management turnover, then there will be no prolonged mispricing or bias.

However, if position disclosures do not coincide with management turnover, then the position

inherited by the current manager will not be perfectly predictable, and the market will not

know what portion of the manager’s current trade is based on new information. In addition,

even if the starting position is disclosed when there is manager turnover, as long as there is

some other parameter uncertainty (perhaps about d), the market will not be able to fully infer

the extent of the trade bias (see footnote 6).

The Journal of Finance 294

cross section of securities, with each security traded by a different manager who

inherits a different initial position. Thus, on average, securities are priced with-

out bias, but only because there are as manyoverpriced securities as underpriced

securities. The existence of a bias for an individual security is analogous to the

standard omitted variable bias in econometrics.

8

The following three propositions characterize how the degree of the bias for a

single security depends upon the parameters of the model.

PROPOSITION 1: For do1, s

2

e

0

40, and a given initial-position deviation D

0

, the average

absolute deviation of price fromterminal value is

a. increasing in s

2

Z

1

,

b. decreasing in s

2

L

1

, and

c. decreasing in s

2

e

1

.

Proof: It is easy to verify the above by inspecting

@

2

E([V ÷ P

1

[[D

0

)

@s

2

Z

1

@D

0

;

@

2

E([V ÷ P

1

[[D

0

)

@s

2

L

1

@D

0

;

and

@

2

E([V ÷ P

1

[[D

0

)

@s

2

e

1

@D

0

;

respectively.

The above results derive solely from the manner in which each of these variables

affects market depth. For consequence (a), the manager trades relatively more

aggressively onthe basis of her signal and relativelyless aggressively onthe basis

of her initial position as the variance of the news increases. Hence, the informa-

tion content of the unexpected net order flow increases, causing market depth to

increase. As market depth increases, the initial position deviation D

1

has a larger

impact on the realized price. Consequence (b) derives from the following argu-

ment. The larger the liquidity variance, the greater is the noise in the net order

flow, and hence the lower is l

1

. As with consequence (a), a lower l

1

implies

8

To integrate over all possible values of X

I

0

and claim there is no bias is similar to conclud-

ing that there is no omitted variable bias just because the correlation between the omitted,

included, and dependent variables is unknown. Even if the correlation is unknown, as long as

it is nonzero, we know that the estimated coefficient is not correct on average. Similarly, there

is a bias in our model as long as the initial position is nonzero.

Delegated Portfolio Management and Rational Prolonged Mispricing 295

that d has a smaller impact on the price. A similar argument justifies conse-

quence (c).

PROPOSITION 2: The impact of the past noise shock, e

0

, on the deviation of current price

fromfundamental value is decreasing in the revelation probability, d. That is,

@

2

[V ÷ P

1

[

@d@e

0

o0:

Proof: This is easily verified given the derivative of the price function in equa-

tion (13) and the reduced-form expression for the equilibrium l

1

in equation (10).

As d rises, the probability that the current manager’s informationwill be publicly

revealed during her tenure increases. The current manager trades less aggres-

sively based on her inherited position ((1 ÷d)/2d goes down), thus trading rela-

tively more aggressively based on her current signal. This has two effects. First,

the current net order flowdepends less on the inherited position and, as a result,

less on the prior manager’s signal error. This results in the signal error having a

smaller impact onthe current price.The second effect, however, is inthe opposite

direction. Because the current manager trades more aggressively based on her

current signal, her trades become more informative and the market becomes less

deep (i.e., l

1

increases). As a result, the component of the manager’s trade that

depends on the inherited position has a greater impact on the price. The above

proposition indicates that the first effect dominates the second effect, implying

that the bias is decreasing in d.

PROPOSITION 3: The impact of past noise shock, e

0

, on the deviation of current price

fromfundamental value is increasing in the quality of A0’s signal (i.e., 1=s

2

e

0

). That is,

@

2

[V ÷ P

1

[

@s

2

e

0

@e

0

o0:

Proof: This can be easily verified given the derivative of the price function

in equation (13) and the reduced-form expression for the equilibrium l

1

in

equation (10).

This result seems counterintuitive initially; higher quality information implies a

more biased price. The intuition for this result, however, can be obtained by rea-

lizing that the lower the variance of past noise, the more reliable is A0’s informa-

tion. Thus, she takes a more extreme position on the basis of that information.

The more extreme the position passed on to A1, the more A1’s trade is based on

this position and the less it is based on new information. As a consequence, the

initial position has a greater impact on the current stock price than the current

information, resulting in a greater deviation between price and fundamental

value.

The Journal of Finance 296

The above result implies that, conditional on a particular error, a firm in an

industry with high-quality information will have a greater bias than a firm in

an industry where information is less reliable. However, because extreme errors

are less likely to occur with higher quality signals, it is a priori unclear how the

degree of mispricing depends upon the quality of the prior signal. In the next

section, we integrate over all possible errors and ascertain the efficiency of

prices as measured by the conditional variance of the terminal value conditional

on the market price of the stock.

B. Informational Efficiency

The following lemma specifies the variance of the terminal value conditional

on the market price (i.e., Var(V[P

1

)).

LEMMA 1: For all d and X

I

0

, the conditional variance is a constant:

Var(V[P

1

) =

s

2

Z

1

2

1 ÷

s

2

e

1

s

2

y

1

¸ ¸

: (14)

Proof: See the Appendix.

Thus, the conditional variance is independent of d and X

I

0

. Furthermore, it is also

independent of the prior signal error variance, s

2

e

0

. This result is similar to the

result in a standard Kyle model that the degree of informational efficiency is in-

dependent of the level of the liquidity-shock variance. As the liquidity-shock var-

iance increases in a standard Kyle model, the aggressiveness with which the

informed agent trades on his information will also increase until the signal-to-

noise ratio in the net order flow returns to the original level, independent of the

level of the liquidity-shock variance. In our model, d falling from one introduces

an extra source of noise, namely that fromthe initial position. As the total noise

from liquidity shocks and the initial position increases, the fund manager in-

creases the aggressiveness with which she trades on her private information un-

til a fixed portion of it is revealed.

The conditional variance in Lemma 1 has two possible interpretations. One in-

terpretation is that the conditional variance represents the uncertainty the mar-

ket maker faces regarding the terminal value given that the prior signal error is

unobservable. Integrating over all the possible values of signal error results in

Var(V[P

1

). This variance considers both the magnitude of the effect of a given

noise shock as well as the probability that that noise shock will be realized. The

second interpretation is that Var(V[P

1

) represents the variance of the market as

a whole, with the inherited positions varying across the securities in the market.

Given the above interpretations, Lemma 1 indicates that the revelation prob-

ability has no impact on (1) the efficiencyof the market as awhole, and (2) the risk

faced by the market maker. However, this result only holds if information quality

Delegated Portfolio Management and Rational Prolonged Mispricing 297

is exogenously fixed. We consider endogenous information quality in the follow-

ing section.

IV. Endogenous Information Quality

If d and s

2

e

0

affect the equilibrium quality of information collected, then the

term in brackets in equation (14) is not independent of d and s

2

e

0

, and price effi-

ciency varies with these parameters. In this section, we investigate how these

parameters affect the fund manager’s incentive to collect information. We also

examine whether differences in the manner inwhichuncertainty is resolvedhave

different effects on price efficiency. In particular, we compare situations inwhich

longer revelation horizons are manifested by d falling relative to situations in

which longer revelation horizons simply imply greater intrinsic uncertainty

(i.e., s

2

Z

1

increasing).We show that, in contrast to the effect of increasing intrinsic

uncertainty which always reduces price efficiency, price efficiency may increase

as the revelation probability d falls.

A. The Incentive to Collect Information

To model information acquisition, we assume that prior to trading, the man-

ager collects information and incurs a personal cost. The more precise her infor-

mation (i.e., the lower the variance of the error, s

2

e

1

), the greater is her personal

cost. Specifically, we assume that the personal cost is c

2

=s

2

e

1

, where c is a constant

parameter. Prior to knowing the signal she will receive, the expectedvalue of her

compensation (i.e., the expected value of equation (7)) is as follows:

E DW ÷

c

2

s

2

e

1

¸ ¸

= H ÷

d

4l

1

s

2

Z

1

2

s

2

Z

1

÷ s

2

e

1

÷

c

2

s

2

e

1

; (15)

where H is independent of s

2

e

1

. The first-order condition implies that

s

2

+

e

1

=

cs

2

Z

1

d

4l

1

1=2

s

2

Z

1

÷ c

; (16)

for co(d/4l

1

)

1/2

s

2

Z

1

.

9

Equation (16) specifies the optimal information quality col-

lected by the current fund manager as a function of the price function parameter

l

1

. Recall, however, that the price function parameter l

1

is a function of s

2

e

1

. Plug-

ging the above expression for s

2

+

e

1

into equation (10) yields a single equation in the

unknown l

1

:

l

2

1

÷ l

1=2

1

c

2

1

Gd

1=2

÷

s

2

Z

1

4G

= 0; (17)

9

For, c _ (d=4l

1

)

1=2

s

2

Z

1

the manager collects no information and s

2

e

1

= ·.

The Journal of Finance 298

where

G = s

2

L

1

÷

1 ÷ d

2d

2

Var(e

0

[Z

0

; o

0

):

Let l

+

1

denote the value of l

1

that solves equation (17). The equilibrium informa-

tion quality is the right-hand side of equation (16) evaluated at l

+

1

.

B. Price Efficiency under Endogenous Information Quality

Because analytical solutions to equation(17) do not exist in general, in Figure 2

we provide numerical solutions to illustrate howequilibriuminformation quality

and price informativeness depend upon the revelation probability and the noise

in the inherited position. Panel Ashows howthe equilibriuminformation quality

varies with d and Var(e

0

[Z

0

; o

0

) and Panel B shows the corresponding values of

the equilibrium market depth parameter l

+

1

. Panel C shows the equilibrium price

efficiency and illustrates two important features. The first important feature is

0.875

0.661

0.448

0.234

s

i

z

e

1.90

1.36

0.81

0.27

l

a

m

b

d

a

1

r

e

l

c

o

n

v

0.733

0.687

0.641

0.595

1.000

0.767

0.533

0.300

delta

1.000

0.767

0.533

0.300

delta

1.000

0.767

0.533

0.300

delta

0.00

0.67

1.33

2.00

cndvare0

0.00

0.67

1.33

2.00

cndvare0

0.00

0.67

1.33

2.00

cndvare0

Panel A

Equillibrium Information Quality

Panel B

Equillibrium Market Depth

Panel C

Equillibrium Price Efficiency

Figure 2.Various aspects of equilibriumunder endogenous informationquality. Pa-

nel A shows the equilibrium amount of signal noise s

2

e

1

as a function of the revelation

probability d and the conditional variance of the first-period signal error Var(e

0

[Z

0

; w

0

).

Panel B shows how the inverse of the market depth, l

1

depends upon d and Var(e

0

[Z

0

; w

0

).

Panel C shows the equilibrium values of Var(V[P

1

)=s

2

Z

1

, where Var(V[P

1

) is the uncer-

tainty remaining after observing P

1

, and s

2

Z

1

is the uncertainty prior to observing P

1

. For

all of the above solutions c 50.1, s

2

L

t

= 0:05, and s

2

Z

t

= 1 for t 50,1.

Delegated Portfolio Management and Rational Prolonged Mispricing 299

that, for Var(e

0

[Z

0

; w

0

) = 0, as d falls, the informational efficiency of the price

falls (i.e., the conditional variance increases). This is intuitive, since the lower d

is, the less likely it is that the fund manager will profit on her information. As a

result, in equilibrium, she spends less effort collecting information (i.e., she col-

lects signals with a high value for s

2

e

1

). Since the informativeness of the price is

decreasing in the quality of the fund manager’s information, prices become less

informative as d falls. Thus, if there is no uncertainty regarding the inherited

position (Var(e

0

[Z

0

; w

0

) = 0 shorter managerial investment horizons reduce

price efficiency.

The second important feature is that for a fixed level of d, price efficiency

improves as Var(e

0

[Z

0

; w

0

) increases. When d is close to one, the improvement is

very slight, while for a low d, the improvement is striking. This happens

because, as Var(e

0

[Z

0

; w

0

) increases, the amount of noise in the net order

flow that emanates from the inherited position X

I

0

increases, causing the

market depth parameter l

+

1

to fall as shown in Figure 2, Panel B. The decline in

l

+

1

allows the fund manager to trade more aggressively on her information

without affecting the price as much. This increases the return from information,

leading to higher equilibrium information quality and more informative

prices.

This second result implies that price efficiency may actually improve as, ce-

teris paribus, either (1) managerial investment horizons become shorter due to

increased turnover (shorter expected tenure length), or (2) the length of time it

takes for private informationto become public increases. Bothof these situations

correspond to decreases in d. As Figure 2, Panel C, shows, for sufficiently high

levels of Var(e

0

[Z

0

; w

0

), a decrease in d leads to a decrease in the conditional var-

iance of the terminal value (i.e., an increase in price efficiency). In contrast, if

there is greater intrinsic uncertainty due to increases in s

2

Z

1

, then price efficiency

unambiguously decreases.

10

Thus, an increase in uncertainty caused by an in-

crease in the time it takes for private information to become public may have a

different effect on price efficiency than a common increase in intrinsic uncer-

tainty (s

2

Z

1

) faced by all agents.

V. Empirical Implications

In this section, we show that the model generates behavior that is consistent

with some empirical regularities that have been documented in the literature. In

10

As can be seen by inspection of equation (14), a sufficient condition for

@Var(V[P

1

)=@s

2

Z

1

40 is that @(s

2

e

1

=s

2

Z

1

)=@s

2

Z

1

o0. By equation (16) @(s

2

e

1

=s

2

Z

1

)=@s

2

Z

1

=

c[

d

4l

1

1=2

s

2

Z

1

÷ c[

÷2

d

4l

1

1=2

s

2

Z

1

2l

1

@l

1

@s

2

Z

1

÷ 1

. Thus the condition holds if the term in square brackets

is negative. Implicitly differentiating equation (17) with respect to s

2

Z

1

implies that

@l

1

@s

2

Z

1

=

[4G(2l

1

÷ l

÷1=2 c

4Gd

1=2

)[

÷1

. Using this result, we have

s

2

Z

1

2l

1

@l

1

@s

2

Z

1

÷ 1o0 = l

2

1

4

1

4

s

2

Z

1

4G

÷ l

1=2

1

c

2Gd

1=2

.

Equation (17) also implies l

2

1

=

s

2

Z

1

4G

÷ l

1=2

1

c

2Gd

1=2

. Therefore the right-hand side of the last inequal-

ity equals l

2

1

and the required condition is satisfied.

The Journal of Finance 300

particular, the pricing and trade bias we identifyabove is consistent with the gla-

mour versus value anomaly (see Lakonishok et al. (1994)) and positive feedback

trading among institutions (see Grinblatt et al. (1995)). In addition, the results on

endogenous information quality also generate some implications with respect to

the relationship between past profitability and subsequent risk taking documen-

ted in the empirical mutual fund tournament literature (see Brown et al. (1996),

Chevalier and Ellison (1997), Coval and Shumway (2000), and Busse (2001)).

A. Glamour versusValue

Lakonishok et al. (1994) consider the subsequent returns on securities that de-

viate from their historical average price-to-earnings ratio. Those securities with

a current price-to-earnings ratio higher than their expected ratio are defined as

‘‘glamour’’ stocks, while those that are below their expected ratio are ‘‘value’’

stocks. The glamour versus value anomaly is that low-P/E (i.e, value) stocks out-

perform high-P/E (i.e., glamour) stocks. In the context of our model, we define a

glamour (value) stock as one that has a higher- (lower-) than-expected price rela-

tive to fundamental value, where fundamental value is defined given past public

information. Specifically, we define the fundamental value as

PP =

VV ÷ Z

0

. Given

the price function at t 51 and this definition of fundamental value, we have

P

1

÷

PP =

g

1

2

(Z

1

÷ e

1

) ÷ l

1

X

L

1

÷

(1 ÷ d)

2d

l

1

D

0

: (18)

A glamour stock is one in which P

1

÷

PP40 and a value stock is one in which

P

1

÷

PPo0. Recall that the long-run price level of the stock will be

P

T

=

VV ÷ Z

0

÷ Z

1

.

To address the existence of any differential expected price movements between

glamour and value stocks, we focus on glamour stocks and examine

E[P

T

÷ P

1

[P

1

÷

PP40[.

PROPOSITION 4: For glamour stocks: E[P

T

÷ P

1

[P

1

÷

PP40[o0.

Proof: See the Appendix.

Thus, the expected return on a glamour stock is negative. A glamour stock (i.e.,

one whose realized price is above its expected value) is one where the prior man-

ager had amassed an above-average positionbased on anoverlyoptimistic signal.

This large position prevents current selling for all but the most extreme negative

signals. Hence, once the truth eventually is revealed (possibly after the current

manager’s tenure), the price sharply declines, generating, on average, a negative

return. A similar argument can be made for value stocks with realized prices

below the mean. In that case, overly pessimistic past signals persist and when

the truth is revealed, the price rises. Furthermore, since

@[E[P

T

÷P

1

[P

1

÷

PP40[[

@d

o0,

the greater the revelation probability, the smaller is the extent of a glamour

Delegated Portfolio Management and Rational Prolonged Mispricing 301

reversion to the mean. Thus, those securities with little long-term information

are less likely to display the glamour-versus-value differential.

B. Positive-FeedbackTrading

De Long et al. (1990a) theoretically examine the effect of positive-feedback tra-

ders on prices. Positive-feedback traders are agents who trade based on past per-

formance, typically buying (selling) securities after they have gone up (down).

They show that if such traders exist, then the trades of rational speculators can

cause volatility to increase. Grinblatt et al. (1995) empirically examine the quar-

terly holdings of 155 mutual funds from1974 to1984 and document that 77 percent

of those funds displayed positive feedback, with most funds buying past winners.

We measure the extent of positive-feedback trading by the covariance between

the prior price change and the subsequent trade: Cov(X

I

1

; P

0

÷ P

÷1

), where

P

÷1

=

VV is the price prior to the first manager’s trade. Apositive correlation be-

tween past returns and subsequent trades corresponds to momentumor positive-

feedback trading, while a negative value of Cov(X

I

1

; P

0

÷ P

÷1

) represents contra-

rian trading. Nonzero correlation occurs because the initial position affects both

the realized return in the first period and the trade of the subsequent manager.

For example, if A0 receives a positive signal, she establishes a long initial posi-

tion. In doing so, she (1) bids up the price (i.e., P

0

÷P

÷1

is positive on average),

and (2) shifts up A1’s trade function (equation (8)). The two effects together result

in the following proposition.

PROPOSITION 5: For all do1,

Cov(X

I

1

; P

0

÷ P

÷1

) =

1 ÷ d

2d

g

0

2l

0

2

s

2

Z

0

÷

g

0

gg

2l

0

s

2

e

0

÷

2l

0

gg

g

0

s

2

L

0

¸

40; (19)

where

gg =

b

0

s

2

e

0

(b

0

)

2

s

2

e

0

÷ s

2

L

0

and b

0

=

g

0

2l

0

:

That is, mutual fund managers engage in positive-feedback trading.

Proof: See the Appendix.

Badrinath and Wahal (1999) investigate conditional positive feedback. The

authors divide their sample into two subsamples: (1) securities that are already

present in a fund, and (2) extreme-change securities for which either a prior posi-

tion is completely reversed or a nonexisting position is established for the first

time. They show that funds tend to add securities for the first time following po-

sitive returns (i.e., momentum) but complete liquidations do not tend to occur

after negative returns. Rather, for complete liquidations, there is a negative cor-

The Journal of Finance 302

relation between the past return and trade, indicating contrarian behavior. Our

model implies these results as well.

PROPOSITION 6: The following conditional covariances hold:

a. Conditional on a zero prior position (i.e., X

I

0

= 0):

Cov(X

I

1

; P

0

÷ P

÷1

[X

I

0

= 0) =

1 ÷ d

2d

s

2

e

0

s

2

e

0

÷ s

2

y

0

s

2

L

0

: (20)

b. Conditional on complete liquidation of an existing position (i.e., X

I

1

= ÷X

I

0

):

Cov(X

I

1

; P

0

÷ P

÷1

[X

I

1

= ÷X

I

0

) = ÷s

L

0

s

y

0

: (21)

Proof: See the Appendix.

Badrinath andWahal (1999) also document that the magnitude of the momentum

effect for a zero prior position is stronger than the contrarian effect for comple-

tely liquidated existing positions. Acomparison of equations (20) and (21) shows

that this results holds if

1 ÷ d

2d

s

L

0

s

y

0

÷ 2

¸

s

2

e

0

4s

2

Z

0

: (22)

Note that if the revelation probability is one, then this condition is never satis-

fied. However, when do1, this conditioncanbe satisfied as longas the ratio of the

liquidity noise to the signal variance and the signal noise are sufficiently large.

Moreover, the lower the revelation probability, the lower the ratio of liquidity

noise to signal variance must be for the condition to hold.

C. Performance and Subsequent ProfitVariability

Our model also provides insight into recently documented empirical relation-

ships between past performance and subsequent risk taking (see, e.g., Brown

et al. (1996), Chevalier and Ellison (1997), Coval and Shumway (2000), and Busse

(2001)). In the model, the liquidity variance in the first period creates a link be-

tween past performance and future volatility. When liquidity variance is high,

profits in the first period are also high. High liquidity variance in the first period

also makes it more difficult for the market to separate the part of A1’s demand

that is based on current information fromthe part that is based on her inherited

position. GivenA1’s greater ability to hide her information, she collects more pre-

cise information in equilibrium. More precise information implies that she takes

more extreme portfolio positions, which results in a higher variance of trade prof-

its. We are unable to sign the effect on the variance of the value of the existing

Delegated Portfolio Management and Rational Prolonged Mispricing 303

portfolio. However, if the total return on the portfolio is due mostly to the trade

profits, then the correlation between past performance and current portfolio re-

turn volatility is positive.

VI. Discussion of Robustness

In this section, we focus on the sensitivity of the results to (1) the number of

informed traders, (2) the form of the compensation contract, and (3) the short

horizons of the managers.

A. Multiple InformedTraders

In this section, we examine the extent of the price bias whenthere are multiple

informed fund managers who trade in a particular security. Specifically, we con-

sider the case inwhichthere are Kcompeting initial fund managers, each receiv-

ing a signal y

j

0

= Z

0

÷ e

j

0

, where j 51, 2, y, K and e

0

j

are i.i.d. We further assume

that subsequent fund managers only know their own inherited positions. Thus,

every informed fund manager in the second period must predict the trades of the

other informed managers given their current signal (y

j

1

), the truth that is re-

vealed (Z

0

), and the prior net order flow (w

0

). The equilibrium with multiple

agents is given in the following lemma.

LEMMA 2: The equilibriumtrade of the ith informed manager is

X

1i

=

g

1

(K ÷ 1)l

1

y

1

÷ m

X

0

X

0i

÷ m

X

0

b

0K

[K(^gg ÷ b

0K

g

+

) ÷ (^gg ÷ 1)[

Z

0

÷ m

X

0

[b

0K

g

+

÷^gg[

w

0

; (23)

where m

X

0

=

(1÷d)

(2÷^gg)d

, g

+

=

b

0K

Ks

2

e

0

b

2

0K

Ks

2

e

0

÷s

2

L

0

, ^gg =

b

2

0K

(K÷1)s

2

e

0

b

2

0K

(K÷1)s

2

e

0

÷s

2

L

0

, and b

0K

= g

0

=(K ÷ 1)l

0

.

The equilibriumprice function is

P

1

=

VV ÷ Z

0

÷

K

K ÷ 1

g

1

(Z

1

÷ e

1

) ÷ m

X

0

b

0K

l

1

(1 ÷ g

+

b

0K

)

¸

K

j=1

e

i

0

÷ g

+

X

L

0

¸

÷ l

1

X

L

1

; (24)

where

l

1

=

K

1=2

K ÷ 1

g

1=2

1

s

Z

1

(m

X

0

b

0K

)

2

[(1 ÷ b

0K

g

+

)

2

Ks

2

e

0

÷ (g

+

)

2

s

2

L

0

[ ÷ s

2

L

1

1=2

: (25)

Proof: Available fromthe authors upon request.

As in a Kyle model with multiple agents, the fraction K/(K11) of the pertinent

private information is revealed; the market becomes deeper with competition

The Journal of Finance 304

among the informed agents. Now, rather than a single prior signal error having

an impact on the price, the sum of the prior signal errors affects the current

price. When there are a finite number of fund managers, the distribution of the

sumof the prior signal errors minus its conditional expectation does not collapse

on zero. Moreover, the market has finite depth (i.e., l

1

40). As a consequence, past

signal errors persist. However, as the number of informed agents increases to in-

finity, the variance of the fourth term in the price function (equation (24)) con-

verges to zero.

11

Since the mean of this term is also zero, past errors have no

impact on the current price and no prolonged mispricing exists in the limit as

the number of informed fund managers goes to infinity.

B. Long-lived Managers and the Incentive to Maintain Managerial Continuity

A natural question arises as to how the equilibrium would differ if managers

were long-lived but still faced the possibility that some of their information may

not get publicly revealed until after theyleave. Arelated question is howthe equi-

librium might change if funds were managed by longer-lived teams of multiple

agents, whereby management continuity could be maintained even though indi-

vidual team members might come and go. On these questions, we want to make

the following points.

First, even if the first manager is retained as the manager in the second period

and she chooses her first position taking into consideration its effect on her sec-

ond-period trade, a bias will still exist as long as the revelation probability is less

than one.

12

Second, if the fund is managed by a team, then the problem becomes

far more complex. Presumably trade decisions will be made jointly by the team,

with each member making a recommendation on the basis of private information

not observed by the other team members. In this case, the expected tenure of an

individual will likely affect the recommendation she makes. To the extent that

the joint decision depends upon the collection of recommendations, then the in-

centives of a single manager will affect even long-lived teams. Third, funds may

have no incentive to adopt policies or organizational forms that would minimize

prolonged mispricing. It is unclear how the fund would sufficiently privately ben-

efit fromadopting a rule that promotes a public good like reduced prolonged mis-

pricing. Even if there is a private benefit, to maintain management continuity,

the fundwill have to incur some costs. For example, it might have to have preemp-

tive salary increases to reduce turnover that decrease the employer’s surplus for

the fund. Thus, unless management continuity is costless to maintain, the issues

discussed in the previous sections will prevail. Nonetheless, these issues point

out that funds will be interested in managing these costs and benefits. We leave

this analysis for future research.

11

The proof is available from the authors upon request.

12

No closed-form solutions for equilibrium are available when the manager is long-lived.

Thus, we analyze the issues in this paper by considering a model with a sequence of short-

lived managers. However, in an appendix available upon request, we characterize the equili-

brium with a long-lived manager and show that the bias will have a smaller magnitude than

that which exists with a series of short-lived managers.

Delegated Portfolio Management and Rational Prolonged Mispricing 305

VII. Conclusion

A stock market with delegated portfolio management in which there is

a mismatch between the manager’s investment horizon and the time it takes for

her information to become public can result in the mispricing of securities.

We focus on the interaction of the trade strategies of a sequence of fund

mangers and show that early fund managers impose an informational external-

ity on subsequent fund managers. This causes obsolete or erroneous past infor-

mation to have a prolonged effect on prices. The magnitude of the resulting

pricing bias displays cross-sectional variation that depends on some economic-

ally important variables. Stocks with higher institutional holdings, for

example, are more likely to be mispriced. Stocks about which managers get high

quality information are also likely to have larger mispricings. We demonstrate

how price efficiency is affected by the parameters of the model, notably the

revelation probability and the variance of the size of the initial equity

position. We also show that prices can actually become more efficient as the re-

velation probability falls, which occurs when the manager’s tenure becomes

shorter.

Appendix

Derivation of Optimal Demand at t 51: Given the price function, the expected

change in the portfolio’s wealth is given by

E[DW[X

I

1

; y

1

; X

I

0

[

= X

I

0

d(

VV ÷ Z

0

÷ g

1

y

1

) ÷ (1 ÷ d)(

VV ÷ Z

0

÷ l

1

[X

I

1

÷ E(X

L

1

)

÷ E(w

1

[Z

0

; w

0

)[) ÷ (

VV ÷ Z

0

)

÷ X

I

1

d(

VV ÷ Z

0

÷ g

1

y

1

) ÷ (1 ÷ d)[

VV ÷ Z

0

÷ l

1

(X

I

1

÷ E(X

L

1

) ÷ E(w

1

[Z

0

; w

0

))[

÷ [

VV ÷ Z

0

÷ l

1

(X

I

1

÷ E(X

L

1

) ÷ E(w

1

[Z

0

; w

0

))[

= X

I

0

dg

1

y

1

÷ (1 ÷ d)l

1

[X

I

1

÷ E(w

1

[Z

0

; w

0

)[

÷ X

I

1

d g

1

y

1

÷ l

1

[X

I

1

÷ E(w

1

[Z

0

; w

0

)[

:

(A1)

The first-order condition for the optimal trade X

I

1

is

@E(DW[X

I

1

; y

1

; X

I

0

)

@X

I

1

= X

I

0

(1 ÷ d)l

1

÷ d[g

1

y

1

÷ l

1

(X

I

1

÷ E[w

1

[Z

0

; w

0

[)[ ÷ X

I

1

dl

1

= 0:

(A2)

This implies that the optimal trade is

X

I

1

=

g

1

y

1

2l

1

÷

(1 ÷ d)

2d

X

I

0

÷

E[w

1

[Z

0

; w

0

[

2

; (A3)

The Journal of Finance 306

where X

I

0

= b

0

y

0

with b

0

5g

0

/(2l

0

). Before preceding, we need to calculate

E[w

1

[Z

0

; w

0

[:

E[w

1

[Z

0

; w

0

[ = E[X

I

1

÷ X

L

1

[Z

0

; w

0

[

= E

g

1

y

1

2l

1

÷

(1 ÷ d)

2d

b

0

(Z

0

÷ e

0

) ÷

E(w

1

[Z

0

; w

0

)

2

[Z

0

; w

0

¸

=

(1 ÷ d)

2d

b

0

[Z

0

÷ E(e

0

[Z

0

; w

0

)[ ÷

E(w

1

[Z

0

; w

0

)

2

(A4)

=

E(w

1

[Z

0

; w

0

)

2

=

(1 ÷ d)

2d

b

0

[Z

0

÷ E(e

0

[Z

0

; w

0

)[ (A5)

= E(w

1

[Z

0

; w

0

) =

(1 ÷ d)

d

b

0

[Z

0

÷ E(e

0

[Z

0

; w

0

)[ =

(1 ÷ d)

d

E[X

I

0

[Z

0

; w

0

[: (A6)

That is, the expected net order flow from the first period is proportional to the

conditional expectation of the initial position of the informed manager.Thus, the

optimal trade at t 51 is

X

I

1

=

g

1

y

1

2l

1

÷

(1 ÷ d)

d

X

I

0

÷ E(X

I

0

[Z

0

; w

0

)

2

¸

: (A7)

Proof of Theorem 2: Given the optimal trade function, the price is given by

P

1

=

VV ÷ Z

0

÷ l

1

X

I

1

÷ X

L

1

÷ E[X

I

1

÷ X

L

1

[Z

0

; w

0

[

=

VV ÷ Z

0

÷ l

1

¦X

I

1

÷ E[X

I

1

[Z

0

; o

0

[¦ ÷ X

L

1

= P

1

=

VV ÷ Z

0

÷ l

1

g

1

y

1

2l

1

÷

(1 ÷ d)

2d

(X

I

0

÷ E[X

I

0

[Z

0

; w

0

[) ÷ X

L

1

¸

;

(A8)

since

X

I

1

÷ E[X

I

1

[Z

0

; w

0

[ =

g

1

y

1

2l

1

÷

(1 ÷ d)

2d

[X

I

0

÷ E(X

I

0

[Z

0

; w

0

)[ ÷

(1 ÷ d)

2d

[2E(X

I

0

[Z

0

; w

0

)[

= X

I

1

÷ E[X

I

1

[Z

0

; w

0

[ =

g

1

y

1

2l

1

÷

(1 ÷ d)

2d

[X

I

0

÷ E(X

I

0

[Z

0

; w

0

)[:

(A9)

Delegated Portfolio Management and Rational Prolonged Mispricing 307

Given the information content of the unexpected net order floww

1

÷E(w

1

/Z

0

, w

0

),

l

1

solves

l

1

=

Cov(

g

1

(Z

1

÷e

1

)

2l

1

÷

(1÷d)

2d

(X

I

0

÷ E[X

I

0

[Z

0

; w

0

[) ÷ X

L

1

; Z

1

)

Var(

g

1

(Z

1

÷e

1

)

2l

1

÷

(1÷d)

2d

(X

I

0

÷ E[X

I

0

[Z

0

; w

0

[) ÷ X

L

1

)

=

g

1

2l

1

s

2

Z

1

g

1

2l

1

2

(s

2

Z

1

÷ s

2

e

1

) ÷ s

2

N

; (A10)

where

s

2

N

= Var

(1 ÷ d)

2d

[X

I

0

÷ E(X

I

0

[Z

0

; w

0

)[ ÷ X

L

1

¸

=

(1 ÷ d)

2d

¸

2

b

2

0

s

2

e

0

b

2

0

s

2

e

0

÷ s

2

L

0

¸ ¸

s

2

L

0

÷ s

2

L

1

: (A11)

Solving for l

1

and simplifying yields equation (10) in the text. Q.E.D.

Proof of Lemma 1: Because P

1

= E(V[w

1

; O

1

), then Var(V[P

1

) = Var(V ÷ P

1

)

and the price function can be written as

P

1

=

VV ÷ Z

0

÷

1

2

s

2

Z

1

s

2

y

1

(Z

1

÷ e

1

) ÷ l

1

X

L

1

÷ l

1

1 ÷ d

2d

g

0

2l

0

[e

0

÷ E(e

0

[w

0

; Z

0

)[: (A12)

Thus,

V ÷ P

1

= 1 ÷

g

1

2

Z

1

÷

g

1

2

e

1

÷ l

1

¸

X

L

1

÷

1 ÷ d

2d

g

0

2l

0

(e

0

÷ E e

0

[w

0

; Z

0

[) [

; (A13)

and

Var(V ÷ P

1

) = s

2

Z

1

1 ÷

g

1

2

2

÷

g

1

2

2

s

2

e

1

÷ l

2

1

s

2

L

1

÷ A

; (A14)

where

A =

1 ÷ d

2d

2

Var(e

0

[Z

0

; w

0

) =

1 ÷ d

2d

2

g

0

2l

0

2

s

2

L

0

s

2

e

0

g

0

2l

0

2

s

2

e

0

÷ s

2

L

0

:

The Journal of Finance 308

Given that l

1

2

satisfies equation (A10), further simplification of (A14) yields

Var(V ÷ P

1

) = s

2

Z

1

1 ÷

g

1

2

2

÷

g

1

2

2

s

2

e

1

÷

1

4

g

1

s

2

Z

1

(s

2

L

1

÷ A)

(s

2

L

1

÷ A)

= s

2

Z

1

1 ÷ g

1

÷

g

2

1

4

¸

÷

g

2

1

4

s

2

e

1

÷

g

1

4

s

2

Z

1

= s

2

Z

1

1 ÷ g

1

÷

g

1

4

÷

g

2

1

4

(s

2

Z

1

÷ s

2

e

1

)

= s

2

Z

1

1 ÷ g

1

÷

g

1

4

÷

g

1

4

s

2

Z

1

= s

2

Z

1

1 ÷ g

1

÷

g

1

4

÷

g

1

4

= s

2

Z

1

1 ÷

g

1

2

=

s

2

Z

1

2

1÷

s

2

e

1

s

2

y

1

¸ ¸

:

(A15)

Q:E:D

Proof of Proposition 4:

E[P

T

÷ P

1

[P

1

÷

PP40[

= E

VV ÷ Z

0

÷ Z

1

) ÷ (

VV ÷ Z

0

÷

g

1

2

(Z

1

÷ e

1

) ÷ l

1

X

L

1

÷

(1 ÷ d)

2d

l

1

D

0

[P

1

÷

PP40

¸

= E Z

1

÷

g

1

2

(Z

1

÷ e

1

) ÷ l

1

X

L

1

÷

(1 ÷ d)

2d

l

1

D

0

[P

1

÷

PP40

¸

= E[Z

1

÷ (P

1

÷

PP)[P

1

÷

PP40[ = E[Z

1

[P

1

÷

PP40[ ÷ E[P

1

÷

PP[P

1

÷

PP40[

= E E[Z

1

[P

1

÷

PP[[P

1

÷

PP40

÷ E[P

1

÷

PP[P

1

÷

PP40[

= E

g

1

2

s

2

Z

1

Var(P

1

÷

PP)

(P

1

÷

PP)[P

1

÷

PP40

¸ ¸

÷ E[P

1

÷

PP[P

1

÷

PP40[

=

g

1

2

s

2

Z

1

Var(P

1

÷

PP)

E P

1

÷

PP[P

1

÷

PP40

÷ E[P

1

÷

PP[P

1

÷

PP40[

= ÷ 1 ÷

g

1

2

s

2

Z

1

Var(P

1

÷

PP)

E P

1

÷

PP[P

1

÷

PP40

:

(A16)

Standard results on conditional expectations under normality (see, e.g., Green

(1997, pp. 951^952)) yields

E[P

1

÷

PP[P

1

÷

PP40[ = E[P

1

÷

PP[ ÷ Var(P

1

÷

PP)[f(0)=:5[; (A17)

where f(z) is the p.d.f. of the standard normal distribution. That is,

f(0) =

1

2p

exp(

÷(0)

2

2

) =

1

2p

exp(0) =

1

2p

: (A18)

Because E[P

1

÷

PP[ = 0, equation (A17) simplifies to

E[P

1

÷

PP[P

1

÷

PP40[ = Var(P

1

÷

PP)=p; (A19)

where

Var(P

1

÷

PP) =

g

1

2

2

(s

2

Z

1

÷ s

2

e

1

) ÷ s

2

N

: (A20)

Delegated Portfolio Management and Rational Prolonged Mispricing 309

Combining (A19), (A20), and (A16) yields

E[P

T

÷ P

1

[P

1

÷

PP40[ = ÷

1

p

g

1

2

2

(s

2

Z

1

÷ s

2

e

1

) ÷ s

2

N

÷

g

1

2

s

2

Z

1

o0: (A21)

Q.E.D.

Proof of Proposition 5: Given that P

÷1

=

VV, P

0

=

VV ÷ l

0

w

0

, and the optimal de-

mand at t 51, the correlation is

Cov(X

I

1

; P

0

÷ P

÷1

)

= E

g

1

2l

1

y

1

÷

1 ÷ d

2d

g

0

2l

0

y

0

÷ggX

L

0

÷ 1 ÷

g

0

gg

2l

0

e

0

¸

g

0

2l

0

y

0

÷ X

L

0

¸

; (A22)

which simplifies to the expression in the statement of the proposition. Q.E.D.

Proof of Proposition 6:

(a)

Cov(X

I

1

; P

0

÷ P

÷1

[X

I

0

= 0)

= E

g

1

2l

1

y

1

÷

1 ÷ d

2d

g

0

2l

0

y

0

÷ (ggX

L

0

÷ 1 ÷

g

0

gg

2l

0

e

0

¸

g

0

2l

0

y

0

÷ X

L

0

[y

0

= 0

¸

=

1 ÷ d

2d

s

L

0

s

y

0

ggs

2

L

0

=

1 ÷ d

2d

s

2

e

0

s

2

e

0

÷ s

2

y

0

s

2

L

0

40: (A23)

(b)

Cov(X

I

1

; P

0

÷ P

÷1

[X

I

1

= ÷X

I

0

)

= E ÷

g

0

2l

0

y

0

g

0

2l

0

y

0

÷ X

L

0

¸

= ÷

g

0

2l

0

2

s

2

y

0

= ÷s

L

0

s

y

0

o0:

(A24)

REFERENCES

Allen, Franklin, and Douglas Gale, 1992, Stock-price manipulation, Reviewof Financial Studies 5, 503^

529.

Allen, Franklin, and Gary Gorton, 1993, Churning bubbles, Review of Economic Studies 60, 813^836.

Allen, Franklin, Stephen Morris, and Andrew Postlewaite, 1993, Finite bubbles with short-sale con-

straints and asymmetric information, Journal of EconomicTheory 61, 206^229.

Badrinath, S. G., and Sunil Wahal, 1999, Momentum trading by institutions, Working paper, Emory

University, Goizueta Business School.

Barberis, Nicholas, Andrei Shleifer, and RobertVishny, 1998, Amodel of investor sentiment, Journal of

Financial Economics 40, 307^343.

Bewley,Truman, 1980,The optimumquantityof money, inJohnKareken and Neil Wallace, eds.: Models

of Monetary Economics (Federal Reserve Bank, Minneapolis, MN).

Brown, Keith C.,W.V. Harlow, and LauraT. Starks, 1996, Of tournaments and temptations: An analysis

of managerial incentives in the mutual fund industry, Journal of Finance 51, 85^110.

The Journal of Finance 310

Brunnermeier, Markus K., 2001, Asset Pricing underAsymmetric Information; Bubbles, Crashes,Techni-

cal Analysis and Herding (Oxford University Press, NewYork).

Busse, JeffreyA., 2001, Another look at mutual fund tournaments, Journal of Financial and Quantita-

tive Analysis 36, 53^73.

Chevalier, Judith, and Glenn Ellison, 1997, Risk taking by mutual funds as a response to incentives,

Journal of Political Economy 105, 1167^1200.

Chevalier, Judith, and GlennEllison, 1999, Are some mutual fund managers better thanothers? Cross-

sectional patterns in behavior and performance, Journal of Finance 54, 875^899.

Constantinides, George, 1983, Capital market equilibrium with personal taxes, Econometrica 51, 611^

636.

Constantinides, George, 1984, Optimal stock trading with personal taxes: Implications for prices and

the abnormal January returns, Journal of Financial Economics 13, 65^89.

Coval, Joshua D., and Tyler Shumway, 2000, Do behavioral biases affect prices? Working paper, Uni-

versity of Michigan Business School.

Daniel, Kent, David Hirshleifer, and Avanidhar Subrahmanyam, 1998, Investor psychology and secur-

ity market under and overreactions, Journal of Finance 53, 1839^1885.

De Long, BradfordJ., Andrei Shleifer, Lawrence H. Summers, and Robert J.Waldman, 1990a, Positive-

feedback investment strategies and destabilizing rational speculation, Journal of Finance 45,

379^396.

De Long, Bradford J., Andrei Shleifer, Lawrence H. Summers, and Robert J. Waldmann, 1990b, Noise

trader risk in financial markets, Journal of Political Economy 98, 703^738.

Diamond, Peter, 1965, National debt in a neoclassical growth model, American Economic Review 55,

1126^1150.

Dow, James, and Gary Gorton, 1994, Arbitrage chains, Journal of Finance 49, 819^849.

Easley, David, Nicholas M. Kiefer, Maureen O’Hara, and Joseph B. Paperman, 1996, Liquidity, infor-

mation, and infrequently traded stocks, Journal of Finance 51, 1405^1436.

Fishman, Michael J., and Kathleen M. Hagerty, 1995, The mandatory disclosure of trades and market

liquidity, Reviewof Financial Studies 8, 637^676.

Gale, David, 1973, Pure exchange equilibrium of dynamic economic models, Journal of Economic

Theory 6, 12^36.

Green, William H., 1997, Econometric Analysis, Third Edition (Prentice Hall Publishers, Englewood

Cliffs, NJ).

Grinblatt, Mark, Sheridan Titman, and Russ Wermers, 1995, Momentum investment strategies, port-

folio performance, and herding: A study of mutual fund behavior, American Economic Review 85,

1088^1105.

Hong, Harrison, and Jeremy Stein, 1999, Aunified theory of underreaction, momentum trading, and

overreaction in asset markets, Journal of Finance 54, 2143^2184.

Jarrow, Robert A., 1992, Market manipulation, bubbles, corners, and short squeezes, Journal of Finan-

cial and Quantitative Analysis 27, 311^336.

Kumar, Praveen, and Duane J. Seppi, 1992, Futures manipulation with ‘‘cash settlement,’’ Journal of

Finance 47, 1485^1502.

Kyle, Albert, 1985, Continuous auction and insider trade, Econometrica 53, 1315^1335.

Lakonishok, Josef, Andrei Shleifer, and RobertW.Vishny, 1994, Contrarian investment, extrapolation,

and risk, Journal of Finance 49, 1541^1578.

Odean, Terrance, 1998,Volume, volatility, price and profit when all traders are above average, Journal

of Finance 53, 1887^1934.

Samuelson, Paul A., 1958, An exact consumption-loan model of interest with or without the social

contrivance of money, Journal of Political Economy 66, 467^482.

Santos, Manuel S., and Michael Woodford, 1997, Rational asset pricingbubbles, Econometrica 65, 19^57.

Tirole, Jean, 1982, On the possibility of speculation under rational expectations, Econometrica 50,

1163^1182.

Tirole, Jean, 1985, Asset bubbles and overlapping generations, Econometrica 53, 1499^1528.

Delegated Portfolio Management and Rational Prolonged Mispricing 311

- Ch03Uploaded bySyifa034
- MEC-001 - Micronomic Analysis 2015Uploaded byAshwin R John
- Cbse Class 12 EcoUploaded byVed Gazta
- Information Regulation (Including Regulation of Advertising)Uploaded bylegalmatters
- 4. Equilibrium Final(2)Uploaded byVaibu Mohan
- Posner.economic.analysisUploaded bydido29
- Learning Curve 3Uploaded byRavi Kiran Navuduri
- var paperUploaded byAdityaGarg
- A Future for Labour in the Global Economy (Bowles S., 2001)Uploaded byDiego Quartulli
- Nelson e Winter_an Evolutionary Economic Change_1982_livroUploaded byvigillatto
- ECO100Y1_Final_2009WUploaded byexamkiller
- Sol_Ch4Uploaded byElizabethBuana
- Economics 101Chapter06Uploaded byArkitel to
- macroeconomic Chapter 2 - DD & SSUploaded bytoushiga
- The Number That Killed Us - Presentation[1]Uploaded byGustavo Barbeito Lacerda
- 7 consumers_producers.pptUploaded byKaran Gandhi
- Pearl EssayUploaded byFreaker Lily
- Economic SolutionUploaded byPeterBenj
- cbse economics XII sample paper -4.docUploaded byFirdosh Khan
- 0. a New Economic Theory for Space ExplorationUploaded byBayron Flores
- mkw_chap_03Uploaded byPatrickLnandu
- Chap07Uploaded byAftab Hashmi
- APE_Unit_I_Prax_Test_AnswersUploaded byMohib Hasnain
- Chapter_7.pptUploaded byWoei Chen
- Analysis of Demand and Demand ForecastingUploaded byyashvimehta93
- NestedSimulationForPortfolioRiskMeasurement.pdfUploaded byPradeep Srivatsava Manikonda
- haraldUploaded bymaxi2366
- Homework Number 2Uploaded byGrace Christanti
- Calvo, G (1988). Servicing the Public Debt- The Role of ExpectationsUploaded byDaniela Sanabria
- Ch03 Parkin MicroeconomicsUploaded byMarlene William

- imapct of selected hr parctices on percieved emplyee performanceUploaded bykashif salman
- Children Attitude Toward TVC in PakUploaded bykashif salman
- Introduction to Microsoft OfficeUploaded bykashif salman
- leadership of a culture change processUploaded bykashif salman
- gender and ownership in uk small firmsUploaded bykashif salman
- women enterpreneurs- management skills and business problemsUploaded bykashif salman
- leadership theoryUploaded bykashif salman
- organisational support, employee development and commitmentUploaded bykashif salman
- challenges and prospects of hrm practices in developing countriesUploaded bykashif salman
- organisational knowledge leadership- grounded theory approach-2007Uploaded bykashif salman
- sunflowerUploaded bykashif salman
- leadership preferences in japan 2007Uploaded bykashif salman
- uk university articleUploaded bykashif salman
- organisational commitment among pakistani university teachersUploaded bykashif salman
- transformational leadershipUploaded bykashif salman
- article4Uploaded bykashif salman
- afshan 4Uploaded bykashif salman
- holy placesUploaded bykashif salman
- appraisal2Uploaded bykashif salman
- article12Uploaded bykashif salman
- markusUploaded bykashif salman
- alien definitionUploaded bykashif salman
- advertising create a goodwill of producateUploaded bykashif salman
- advertising planningUploaded bykashif salman
- appraisal1Uploaded bykashif salman
- appraisal employee developmentUploaded bykashif salman
- advertising under varticle producat differenateUploaded bykashif salman
- appraisalUploaded bykashif salman
- advertising bansUploaded bykashif salman
- max planckUploaded bykashif salman