Piggybacking on insider trades

UNIVtKb! I . OF
ILLINOIS LIBRARY
AT URBANA-CHA;
BOOKSTACKS

O A

/---

CENTRAL CIRCULATION BOOKSTACKS

The person charging this material is re-
sponsible for its renewal or its return to
the library from which it was borrowed
on or before the Latest Date stamped
below. You may be charged a minimum
fee of $75.00 for each lost book.

Theft, mutilation, and underlining of books are reasons
for disciplinary action and may result In dismissal from
the University.
TO RENEW CALL TELEPHONE CENTER, 333-8400

UNIVERSITY OF ILLINOIS LIBRARY AT URBANA CHAMPAIGN

APR 2 9 m
AUG 0 3 1998

When renewing by phone, write new due date below
previous due date. L162

STX

I COPY 2

BEBR

FACULTY WORKING
PAPER NO. 90-1631

Piggybacking on Insider Trades

'ary of the
MAR 2 4 1990

University cf Minds

of (Jrfc3rKi-Champa!gn

Asani Sarkar

College of Commerce and Business Administration
Bureau cf Economic and Business Research
University of Illinois Urbana-Champaign

Digitized by the Internet Archive

in 2011 with funding from

University of Illinois Urbana-Champaign

http://www.archive.org/details/piggybackingonin1631sark

BEBR

FACULTY WORKING PAPER NO. 90-1631

College of Commerce and Business Administration

University of Illinois at Urbana- Champaign

February 1990

Piggybacking on Insider Trades

Asani Sarkar

Department of Finance

University of Illinois

225G David Kinley Hall

1407 West Gregory

Urbana, IL 61801

(217) 333-9128

I am grateful to Franklin Allen, Gary Gorton, Richard Kihlstrom, George
Mailath, Andrew Postlewaite, Jean-Luc Vila and Asher Wolinsky for
valuable comments. I also thank seminar participants at the University
of Florida, Georgia Tech, McGill, Pennsylvania, Chicago Mercantile
Exchange, and the Federal Reserve Bank of Philadelphia.

ABSTRACT

The effect of piggybacking (copying of insider trades by other
traders) on insider trades is considered in the context of an asym-
metric information leader-follower model. When the piggybacker is
able to observe the insider's trades perfectly, insider information
is revealed exactly in the separating equilibrium. In order to signal
poor information, the insider restricts trading. A surprising result
is that these strategic effects may cause price inf orraativeness and
insider's trading volume to be negatively related to information
quality for some parameter values. The main policy implication is
that requiring trade intentions disclosure from insiders will reduce
trading volumes and market efficiency for some parameter values.

Insider trading spawns a host of people who try to piggyback on
these trades. Outsiders can monitor publicly available statistics
such as the SEC's "Official Summary of Insider Trading." Others, such
as bankers and brokers, may observe insider trades through their rela-
tionship with insiders. Rational insiders must condition their be-
havior upon the presence of piggybackers. The strategic interaction
that results between insiders and piggybackers affects insiders'
equilibrium behavior in ways which are important to our understanding
of how financial asset markets operate.

The ins ider-piggybacker interaction is modelled as a leader-
follower game between two traders who use trade size as the strategic
variable. It is based upon the one-shot version of Kyle's [1985]
trading model. In stage one noise traders and the insider (leader)
trade simultaneously. The former trade a fi>ced amount while the
leader's trading volume is based on his private piece of information
about the random value of a single risky asset. In stage two, the
piggybacker (follower) trades after observation of the leader's quan-
tity choice in stage one. When all the orders are in, the market-
maker sets the price of the asset to clear the market.

If the leader's trade is observed exactly (perfect piggybacking)
then the leader has an incentive to restrict trades in order to signal
poor information. In a separating equilibrium, however, the follower
correctly infers the leader's information. In this setting the use
of valuable information is costly to the leader because the follower
successfully piggybacks on his trades and drives up the asset price.
1 show, in fact, that the leader may throw away valuable information

-2-

by reducing his trades in response Co better information. This is
consistent with empirical evidence that the size of insider trans-
actions is unrelated to the value of information. As a consequence,
when information precision increases, prices may become less informa-
tive (unlike Grossman and Stiglitz [1981]) and markets more liquid
(unlike Kyle) .

The perfect piggybacking model is comparable to the limit pricing
paper of Milgrora and Roberts [1981], who solve a signalling model in
which prices convey information about an incumbeat firm's cost func-
tion. Gal-Or [1987] constructs a leader-follower model with asymmetric
information in the context of a product market. The difference with
this paper is that here the pricing rule is endogenously determined
and non-linear decision rules are also worked out. The idea of trade
size conveying information is present in Easley and O'hara [1987],
Golsten and Milgrom [1985], Madhavan [1988], and numerous empirical
papers (see footnote 1 for some references).

Under noisy piggybacking, the follower observes the sura of noise
trade and the leader's trade. Therefore observation of the leader's
trade only conveys statistical information to the follower. A striking
result is that arbitrarily small amounts of noise trading is enough to
reverse the comparative static effects of piggybacking. The leader's

See, for example, Scholes [1972], Jaffe [1974], Seyhun [1986],
and Givoly and Palraon [1985].

Markets are more liquid in the sense of "depth" not volume (see
Kyle). In liquid markets prices are relatively insensitive to move-
ments in trades.

-3-

trade is now positively related to both its information precision and
the amount of noise trading and inversely related to the follower's
information quality. These comparative static results are consistent
with those in Mathews and Mirraan [1983] who consider a limit pricing
model where the entrant has a noisy observation of the incumbent's
price.

Other extensions to the perfect piggybacking model show that
(1) information sharing between the two insiders makes the leader
strictly better off and (2) as the number of follows go to infinity,
each follower's trades approach zero but total trades of all followers
increase without bound.

An interesting application of the models is in analyzing pre-trade
disclosure laws. The noisy piggybacking model is taken as the bench-
mark case of no-regulation. A policy requiring trade intentions
(Tl) disclosure can be interpreted as the leader having to reveal his
trades exactly, which is the perfect piggybacking model. The result-
ing analysis suggests that the Tl policy reduces trading volume and
may lower market efficiency. 1 propose instead a policy of de-
regulating insider trading along with a lump-sura grant to outsider
traders financed by a tax on insiders' profits. An additional ad-
vantage of this policy is that it maintains the benefit of insider
trading, which is to increase the inf orraat i veness of the pricing
system through the trading process.

The paper is organized as follows. Section 1 surveys the rele-
vant literature. Sections 2 and 3 deal with the perfect piggybacking
model. Section 4 describes the model with noisy piggybacking.

-4-

Section 5 considers other extensions to the perfect piggybacking
model. Section 2 suggests policy implications and Section 7 con-
cludes .

Section 1. Review of Literature
Markets communicate information between rational market partici-
pants in different ways. In the context of securities markets, the
focus has long been on the role of equilibrium prices as a reflector
of information possessed by various traders. Grossman and Stiglitz,
Diamond and Verrecchia [1981], Hellwig [1980], and others have used
the concept of a competitive rational expectations model to study the
information content of equilibrium prices. Information is incorpor-
ated into prices by the trades of insider traders who have private
information about the uncertain value of the asset. These traders are
rational in the sense that, although they do not know actual outcomes,
they are aware of the underlying stochastic model. Traders therefore
understand that prices convey information and use this knowledge to
form expectations which are correct in equilibrium. The link between
information and prices via trades provides an explicit mechanism for
information transmission between informed and uninformed traders.
Also Grossman and Stiglitz, in particular, show how changes in the
precision of private information affect price inf orraat iveness and
other variables of interest.

A fundamental problem with the competitive paradigm is that in-
formed traders are assumed to be price-takers even though they
influence prices through trades. This "schizophrenic" behavior, to

-5-

use Hellwig's terra, leads to certain problems — chiefly, the paradox
discussed by Grossman and Stiglitz. If traders are price-takers and
prices are fully-revealing, no trader will want to be informed. But
then, there will be no information for prices to reflect and informa-
tion will have value.

Different approaches have been taken to get over this problem.
One way is to add noise to the system, keeping the price-taking assump-
tion intact. This can be done through noisy aggregate supply of the
asset, adding noise traders, etc. Another is to work with a "large
market," as in Hellwig, where each informed trader is made "small"
relative to the market in a specified manner.

A last method involves dropping the price-taking assumption and,
instead, allow insiders to act strategically. There are now several
papers in this category. Kihlstrora and Postlewaite [1987] look at a
well-informed dominant trader who sets prices in a futures market.
Realizing that other traders may be able to infer his information from
market prices, the monopolist uses a randomized pricing strategy to
optimally determine how much of his information to use. In Grinblatt
and Ross [1985], uninformed price-taking investors try to infer the
monopolist insider's information from the market-clearing price of a
risky asset. The monopolist insider is modelled as a Stackelberg
leader. Aware of his information leakage, the insider does not use
his information to the extent a price-taker would. Gould and
Verrecchia [1985] look at a price-setting specialist who has private
information about a risky asset. A trader, upon observing prices,
can infer information as before. The specialist is assumed to

-6-

exogenously add noise to his pricing rule. Grinblatt and Ross allow
the insider to add noise to demand and find that adding noise is not
an optimal strategy.

Allen [1987] asks: what is the social value of asymmetric informa-
tion? He shows that cheaper information is Pareto-inf erior because
(1) risk-sharing opportunities are reduced as prices become better
signals and (2) uninformed agents have to trade with more informed
traders. An important implication of the model is that insider
trading may improve the incentives of current and future managers,
as originally suggested by Manne [1966]. Dye [1984] has also demon-
strated this result in the context of a principal-agent model.

Another strand of the literature looks at equilibrium trades as
information signals. Empirical research provides considerable support
to the contention that insider trades have informative content.
Scholes [1972] finds that secondary offerings (many of which are issued
by insiders) may act as a signal to other investors that the seller
has adverse information about the firm. Jaffe [1974] studies the
information content of the SEC's "Official Summary of Insider Trading"
and finds that uninformed outsiders can get significant abnormal
returns from replicating these trades. And, finally, Givoly and
Palmon [1985] content that "significant abnormal returns are generated
in the wake of these trades themselves . . . outside investors follow
the footsteps of insiders."

The theoretical literature primarily focuses on market-makers and
their ability to infer insiders' information by observing their

-7-

market -orders . The seminal article in this literature was by Kyle,
where a monopolist insider behaves strategically by explicitly taking
into account the effect of his trades on the price established. Unin-
formed liquidity traders camouflage the insider from competitive
market-makers who infer information through observation of aggregate
trades in the market. Easley and O'hara presents a model with two
exogenously fixed order size levels — large and small. Since informed
traders wish to trade larger amounts at any given price, trade size
conveys information to the market-maker. In Glosten and Milgrom trade
size is fixed but the act of trading (whether the trader buys or sells,
for example) is informative to a specialist who sets bid-ask spreads.
Madhavan distinguishes between continuous and periodic trading mech-
anisms. In the continuous dealer market, dealers learn from the se-
quence of traders' market-orders, while traders learn from dealers'
bid-ask quotes. In the periodic batch system, traders submit price-
dependent orders so that prices act as information signals.

Section 2. The Model With Perfect Piggybacking
There are three kinds of traders who exchange among themselves a

risky asset for a riskless asset: uninformed noise traders who trade

3 -

randomly; two insider traders with private informations s, and s0

3
Noise-traders as a group lose money in equilibrium. Their

presence is often justified by considering them as life-cycle or

liquidity constrained traders. An alternative assumption is noisy

total asset supply. Allen uses the assumption of a stochastic birth

rate and fixed total supply to obtain random per capita asset supply.

about the liquidation value v of the risky asset; and market-makers
who set prices efficiently conditional on the aggregate quantities
traded in the market.

Initially, noise traders and the insider who receives information
first (leader) trade simultaneously. Noise traders trade a fixed
quantity u while the leader trades x. based on his private information
s.. In the second stage, the other insider (follower) trades after
observing x. and his private information s~. Finally, market-makers
set a price and trade the quantity that clears the market.

Denoting I. as the leader's information set and I? as the fol-
lower's information set, we have:

(2.1) Ij = {Sl} l? = {s2, Xj}

Further, the insiders' signals are of the form:

(2.2) sk = v + ek, k = 1,2

where e, is a random noise terra uncorrelated with v and independent
k

of e., j * k. All random variables are assumed to be normally dis-
tributed with zero mean and constant variance.

(2.3) v ~ N(0, ZQ)

e ~ N(), Z ), k = 1,2
K ek

U ~ N(0, Eu).

These distributional assumptions, though restrictive, allow me
to find a unique equilibrium under linear trading rules and derive

-9-

coraparative static results. Also, 1 will only consider pure strate-
gies. Non-linear trading rules are discussed in Section 5.

Note that the market described has the character of an auctions
market because prices are determined only at the final stage. The
insiders place market orders while the market-makers choose a pricing
rule P as a function of total trade (insider trade plus noise trade).
Insiders' market orders and noise trade, along with the pricing rule
P, determines equilibrium trading price at the final stage. Figure 1
summarizes the sequence of moves.

Modelling the trading protocol in this manner allows the follower's
piggybacking to affect the leader's trading strategy and the exploita-
tion of his private information. So long as insiders are not able to
exploit all their information instantaneously, piggybacking will have
real effects on insiders' behavior. The model is able to capture
these effects in a relatively simple and stylized setting.

The equilibrium concept followed is that of sequential equilibrium,
introduced by Kreps and Wilson [1982].

2 . 1 Sequential Equilibrium

A sequential equilibrium is a strategy triple (X. , X~,P) and a set
of beliefs on the follower's part such that:

(1) The leader's trading strategy X, and the follower's strategy
X_ are best responses to each other.

(2) For any x. , X„(x.) maximizes the follower's expected profits
where these expectations are taken with respect to some be-
liefs over the leader's information.

-IO-

CS) Given X. and X„ , P satisfies the following efficiency con-
dition.

(2.4) p = E(v|y = Xl+x2+y) + Ty.

Condition (2.4) is derived by assuming that market-makers earn

4
zero expected profits conditional on y, the total trade in the market.

(It is required that market-makers take their expectations with re-
spect to the same beliefs as the follower.) Thus equilibrium prices
satisfy serai-strong efficiency. The linear relation between p and y
follows from the assumption of normality, which implies that r is the
regression coefficient of the linear project of v and y and is given
by the normal equation:

(2.5) r=Co^vlJ2

y

where Cov(v,y) is the covariance between v and y and £ is the var-
iance of y. Kyle has interpreted l/T as market liquidity or "depth."

2 . 2 Market Liquidity

Market liquidity is defined as the volume of trading required to
change prices by one dollar and is measured as 1/T.

Intuitively, liquid markets are those that allow investors to
trade large volumes of stocks in a short period of time without chang-
ing prices by large amounts.

4
This assumption can be justified by interpreting (2.4) as the

equilibrium outcome of a Bertrand game between at least two market-
makers who only observe y, as noted by Kyle.

-11-

The sequential equilibrium concept defined in 2.1 requires in-
siders' and market-makers' strategies to be optimal with respect to
given beliefs over s. . Condition two requires that these beliefs be
given by Bayes ' rule along the equilibrium path but places no restric-
tion on beliefs of f-the-equilibrium path. In general, I will look
for a separating equilibrium which will be defined rigorously below.
In these equilibria, the leader's information is revealed perfectly
to the follower and so the impact of piggybacking on the leader's
trading strategy is maximized.

Let z = X.(s,) be the leader's trade and s* the follower's beliefs
about the leader's information s. upon observing z. Then s* = X. (z)
if X. is one-to-one. Define the follower's problem as:

(2.6) max E [ {( v-Ty )x2 } | s* , s2 ] .
X2
The follower maximizes his expected profits conditional on his

information set I~. Condition (2.6) incorporates the fact that
p = Ty (see equation 2.4). The first-order condition for this problem
yields x? = [E(v|s*,s~) - Tz]/ir. The second-order condition is sat-
isfied by T > ). It is shown in the appendix that:

(2.7) E(v|s*,s2) = tjO-Ds* + Ts2

t (1-t ) £

(2.8) T = -fVT- where t. = -™— ,

l-tlt2 I EQ+E

i

i = 1,2

L/E

1/E + 1/E + 1/E

-12-

Note that t. e [0 1] and is a measure of the unconditional precision

of s.. For example, t. = 1 implies that s. is a perfect signal.

Second, since 1/2 is the precision of s. (conditional on v), T is
e . l

l

the proportion of total precision contributed by s~. Similarly,
t.(l-T) is the proportion explained by s, . These, then, are the
weights placed by the follower on his signal s~ and his inference s*
in learning about r. Re-writing the follower's first-order condition:

(2.9) x2 = [tjCl-Ds* + Ts2 - Vz]/2T.

Denote V(s.,s*,z) as the leader's expected profits when his in-
formation is s., the follower's inference is s* and the leader chooses
z. V(s ,,s*,z) = E[{(v - Tz - Tx - Tu)x }|s ]. Substituting (2.9)
for x- and using the facts that (i) E(v|s ) = t.s, and (ii) E(s9|s.) =
t.s, yields the following form for V:

t (l-T)s* Tt s _

(2.10) V(Sl,s*,z) - (tjSj 2— ~-~-)z.

2 . 3 Separating Equilibrium Strategy

X (s,) is a separating equilibrium strategy if it is one-to-one
and satisfies the following incentive compatibility (IC) condition:

(2.11) K^Sj) = argmax V(Sl,X~l(z),z)

z

X (s ) is a linear separating equilibrium strategy if X (s ) is linear

in s .

This exposition follow Mailath [1987].

-13-

Section 3. Effect of Piggybacking on Insider Trades
A unique separating equilibrium will be shown to exist when X.(s,)
is linear in s. leaving non-linear trading rules for Section 5. How-
ever, comparative static results will only be derived for the linear
separating equilibrium.

Suppose X. is linear and satisfies X, = A.s,. Then X. (z) = z/A.
and the IC condition (2.11) requires

(3.1) t1(l-T)s1/2 - t1(l-T)z/A1 - Tz = 0.

The following proposition characterizes the equilibrium.

Proposition 3.1. If t. e (0 1), t- > 0, and £ > 0, then there is a
unique separating equilibrium under linear trading strategies for the
model described in Section 2. The equilibrium X,, X9 , and P are given
by:

(3.2) X^Sj) = A!sr X2(x1,s9) = BjXj + B?s2,
P(y) = Ty, y = x1 + x + y

where A., B, , B? , and T are defined as

(3.3) Aj = t1T/2T, B2 = T/2T, B = (1/T) - 1.5,

r = (QZ /E )1/2/2, Q = t.(l-T) + T - t,T2/4,
0 u 1 1

and t ., t?, T are defined in (2.8).
Proof: The solution to (3.1) gives the equilibrium A.. B and B» are

obtained from (2.9) by substituting s* = z/A. and the equilibri

urn

-14-

value for A.. Finally, T is derived from the pricing rule (details
are given in the appendix).

Figure 2 illustrates the linear separating equilibrium described

1/2
in Proposition 3.1. The example assumes (En/E ) ' = 1, t. = 0.75,

t~ = 0.5. V. and V„ are the leader's iso-profit curves for s, = 1.2
and s. = 3, respectively. L is the leader's linear separating equi-
librium strategy. Mailath shows that L must be tangent to V, at
[1.2,L(1.2)] and to V2 at [3,L(3)]. The figure illustrates that I
fulfills this requirement.

To gain insight into the effect of piggybacking by the follower,
consider two special cases.

Corollary 3. 1 . Suppose piggybacking has no value to the follower. In
equilibrium,

(i) The leader trades more than he does with piggybacking.
Further, his trades are a strictly increasing function of his infor-
mation precision.

(ii) Market efficiency (represented by the inf orrnat iveness of the
price system) is a strictly increasing function of information pre-
cision.

(iii) Market liquidity (defined in 2.2) is a strictly decreasing
function of information precision.

Proof : The formal proof is given in the appendix.

-15-

Intuitively, the follower does not benefit from piggybacking if
E(v|s*,s9) = E(v|s9). In other words, observation of the leader's
trade z (and so s*) provides the follower no information about v that
is not already contained in s„. A sufficient (but not necessary) con-
dition for this is s, = s9 = s which means that both insiders have
the same piece of information. It can be shown that the resulting
outcome has the characteristics of a symmetric information Stackelberg
equilibrium. The leader has a strict first-mover advantage, which he
exploits by pre-corami tt ing to a large position in the asset. He
trades more than the follower and obtains higher expected profits.

The result on market efficiency is similar in spirit to the com-
parative static result in Grossman and Stiglitz. Efficiency is
measured by var(v|p) — i.e., the variance of the asset value condi-
tional on prices. This is a measure of the inf ormat iveness of equi-
librium prices. As in Grossraan-Stiglitz , the leader's trades become
more sensitive to changes in his information as t. increases. So,
movements in aggregate trade y become more informative about v. Since,
prices are proportional to y, prices also become more informative.

Part (iii) re-affirms Kyle's result (formalizing Bagehot's [1971]
intuition) that market-makers reduce liquidity to compensate them-
selves for bad trades with insiders.

Strictly speaking, price inf ormat iveness is measured by
£ - var(v jp).

Market-makers face an adverse selection problem due to the
presence of informed traders. On average, they lose money to insiders
which they make up by profiting with respect to noise traders.

-16-

Corollary 3.2. Suppose piggybacking is the follower's only source of
information. Then the following is true.

(i) There is no separating equilibrium under linear trading
strategies.

(ii) Serai -poo ling equilibria and a unique non-linear separating
equilibrium exist. In both kinds of equilibria the leader restricts
his trade relative to the case where there is no value to piggyback-
ing.

Proof : For the formal proof, the reader is referred to Propositions
2.1-2.3 in Sarkar [1988]. Intuitively, under the conditions of the
corollary, the follower's information set is simply {x. } and the
leader's trading strategy X is linear in s ; and part (i) of the
corollary says that there is no x. > 0 which satisfies the (IC) con-
dition 2.11. At the margin, the leader always benefits by reducing
his trade because the gain from reducing the follower's piggybacking
offsets the profits foregone frora not exploiting his information.
This result is similar to Corollary 2 in Gould and Verrecchia where
a specialist with private information about a risky asset sets a
price which is observed by a trader. They show that if the trader
has no information other than his price observation, equilibrium
would not exist unless the specialist pre-coramits to adding noise to
his pricing rule.

Since the follower's piggybacking has strong adverse effects on
the leader's trades, the leader has an incentive to complicate his
trading strategy, making it more difficult for the follower to infer

-17-

his information. Sarkar finds serai-pooling equilibria, where the
leader trades only if his signal exceeds a critical value. In the
unique non-linear equilibrium, his trades are equal to the first-best

Stackelberg level for the best information but are strictly below

8
that level elsewhere.

The two corollaries consider two special cases of Proposition 3.1.
When piggybacking has no value, the leader benefits from the advantage
of trading first, akin to a Stackelberg leader. When the follower is
solely dependent on piggybacking as a source of information about v,
its adverse affect on the leader's trades is so strong that no separ-
ating equilibrium may exist. In general, piggybacking has less ex-
treme but still significant effects on the leader's trades. In par-
ticular, the leader restricts his trading to mislead the follower into
believing that his information is worse than it actually is. In a
separating equilibrium, the leader does not succeed and his informa-
tion is perfectly revealed. But the attempt to hide his information
nevertheless constrains the leader to lower his equilibrium trading
volume relative to a world where piggybacking has no value.

In fact, the next proposition shows that the strategic cost of
using information for the leader can be so high that he may throw
away valuable information.

For the serai-pooling equilibrium, the leader's signal distribu-
tion is truncated at both ends. For the non-linear equilibrium,
the truncation is at the upper end only.

-18-

Propositlon 3.2. Suppose the follower has less than perfect inforraa-
tion. Then there exists a critical level of information precision t
for the leader (which is a function of the follower's information pre-
cision t ) such that for t less than (greater to) t , the leader's
trades and expected profits are a strictly increasing (decreasing)
function of t •

Proof: See appendix.

Increases in the accuracy of the leader's information t. also in-
creases the value of piggybacking to the follower who reacts very
strongly to the leader's trades. For t > t , increases in the
leader's trades prompt the follower to put so much volume on the
market that market-makers set a very high price. Therefore, the

■k

leader finds it more profitable to reduce his trades when t > t to
restrain the follower's piggybacking. Figure 3 illustrates that the
leader's trades are a single-peaked function of t .

Proposition 3.2 has an important implication for market efficiency,
as described below. But, first, it is necessary to inquire about the
behavior of liquidity. Figure 4 shows that market liquidity may in-
crease with respect to both t and t . Due to the constraint on the
1 I 2

leader's trades brought about by piggybacking, market-makers face a
lower probability of bad trades with insiders. They are, therefore,
less inclined to reduce liquidity to compensate for insider trades.

-19-

Proposition 3.3. When both insiders are well-informed (t. and t~
close to I), market efficiency is a decreasing function of the
leader's information precision.

Proof : See appendix.

Market efficiency is positively related to both liquidity and the
covariance between asset value v and total trades y. For low values
of t., there is little piggybacking, markets are liquid and efficiency
increases with t. (Corollary 3.1). For high values of t. and low
values of t„ , piggybacking is effective so that trades are not infor-
mative (Proposition 3.2) but liquidity is increasing with t, (see
Figure 4) so efficiency also increases. When both t. and t? are high,
further increases in t. decreases both the covariance term and
liquidity, so that prices become less information. Figure 5 illus-
trates.

The discussion above suggests that (i) piggybacking acts as a
mechanism for reducing the volume of insider trading but (ii) the
strategic behavior induced by piggybacking on insiders may cause a
reduction in market efficiency.

Section 4. The Model With Noisy Piggybacking
Suppose that, in the model described in Section 2, the follower
observes the sum of the leader's trade and the noise trade y. = (x +u)
instead of just x. . Noise traders' activities, therefore, camouflage
the leader's trades and the follower can no longer infer the leader's
information perfectly. If, for example, the follower observes a high

-20-

y. he does not know whether the leader has good information and x. is
high or whether the leader has poor information but noise trading u is
high. In this sense, the follower acts like Kyle's market-makers.

To solve the model assume that the leader's trading strategy is
linear and satisfies X, = A.s,. Given an observation of y, and his
private information s9, the follower selects to trade an amount x9
that maximizes his expected profits.

Max E{([v-ry1-rx2]x2)|y1,s2}
X2
The first-order condition for this problem yields

(4.1) x = [E(v|y s„) - fy ]/2T when

(4.2) E(v|y ,s ) = a„s9 + a.y (1-a )

a, =

t.t A,

1 u 1

1 2 '

t, (l-tu)+t A~
1 V 1

t (1-a A )

a9 = -—■ -~ , t = Sn/(E. + E ), t. = Sn/(^+E ),

2 l-t0a.A, m 0 0 u i 0 0 e.

2 11 i

i = 1,2.

Details of the computation are given in the appendix. t can be
interpreted as a measure of the noisiness of the follower's observa-
tion of y.. (This interpretation of t is consistent with Mathews and

J 1 ' u

Mirraan.) For example, if £ =0 then t =1 and this is the same as

u u

the follower observing x. . Conversely, T. infinity implies t =0
and y. conveys no information to the follower.

-21-

Substituting x„ from (4.1) back into the leader's expected profit
function and solving gives the leader's equilibrium trading strategy:

(4.3) Xj = A^, Ax = r+ai(l-a2)

Given T, A. , a. and a~ can be solved as a simultaneous equation
system. The equilibrium is characterized below.

Proposition 4.1. For t. e (0 1), t~ > 0 and t e (0 1), there is a
unique equilibrium under linear strategies X., X~ , and P where:

(4.4) XjCsj) = A1s1, X2(y1,s2) = B1y1 + 32s2 , P(y) - Ty where

t.T a (1-a ) a?

a and a are given in (4.2) and U is a constant terra determined by

t, , t„ , and t . U is described in equation (A10) of the appendix.
12 y ' l r

T is defined in (A.14) of the appendix.

Proof: See appendix.

As before, some intuition maybe obtained by looking at special

cases of the result. First, when t = 1 in equations (4.1)— (4.3)

u

A,, B, , and B~ have the same expressions as in the model where the
foLlower observes x. perfectly. In this case, of course, T. =0 and
so no equilibrium exists. Second, suppose t =0. Then the follower's
observation is perfectly noisy, B. = 0 and there is no value to piggy-
backing. In fact, a stronger result can be proved.

-22-

Corollary 4.1. Suppose the follower had no Information other than his
noisy observation y. . Then in equilibrium,

(i) The follower does not trade and makes zero profits,
(ii) The leaders makes monopoly profits and trades the monopoly
quantity.

Proof : Substitute t = 0 in equations (4.1 )— ( 4.3). The proof that
these are the monopoly trade and quantity levels is in the appendix.

The above discussion suggests that adding noise to the follower's
observation mitigates the effect of piggybacking on the leader's
trades. Without noise, Corollary 3.2 stated that there was no equi-
librium with t? = 0. With some noise, the leader can effectively be a
monopolist when t~ = 0.

It may be expected that the leader can exploit his information
more freely now that the threat of piggybacking is diluted. The next
proposition shows that this is indeed true.

Proposition 4.2. If t > 0, then

(i) The leader's trades are a strictly increasing function of his
information precision.

(ii) For a given quantity and quality of information, the leader's
trades are a strictly increasing function of noise.

Since the complete solution (i.e., including T and the pricing
function) to the model with noisy piggybacking cannot be obtained in
closed-form, Proposition 4.2 cannot be proved analytically. However,

-23-

A in (4.4) and T can be solved numerically as a system of two non-
linear equations in A., and T. Figure 6 shows how A. behaves when t.
varies over the unit interval, given t„ and t . What's striking is
the fact that arbitrarily small amounts of noise is sufficient for the
comparative static effects of piggybacking to be reversed. Part (ii)
of the proposition is intuitive since noise hides the leader's infor-
mation from the follower. Figure 6 also illustrates that, given t.
and t , the leader's trades are a decreasing function of t„. When the
follower's information is more precise, he is less dependent on the
noisy observation and so noise is a less effective masking device.

Proposition 4.2 implies that market efficiency is also an increas-
ing function of the leader's information precision. The behavior of
liquidity depends upon the level of t„. For low values of t?, the
follower's trade is not sensitive to his own information. Therefore,
market-makers view changes in total trades as due to either changes
in the leader's information precision or noise trades. Kyle's analy-
sis applies and liquidity declines with t. . For high values of t„ ,
the follower has little need to piggyback and behaves more like a
Stackelberg follower with a downward sloping reaction function.
Changes in the leader's trades x. causes the follower's trader to
change by some proportion of (x.+u) in an offsetting manner. Total
insider activity may decline and cause liquidity to increase with t..

Section 5. Other Extensions to the Model
In this section, I consider some further extensions to the model
with perfect piggybacking. These extensions consider different

-24-

situations under which the effect of piggybacking on the leader's
trades are diminished.

( a ) Information Sharing by Leader

If the leader could credibly convey his information s. to the

follower, the latter would have no further need to piggyback on the

9
leader's trade. The follower's information set is now I9 = {s.,s«}.

Suppose the leader trades z. Solving the follower's problem in the

usual way:

(5.1) x2 = [e(v|s1,s2) - rz]/2r

where E(v|s,,s2) = t,(l-T)s, + Ts2.

The equilibrium trade for the leader is simple to obtain. Remember
that the leader's expected profit function V in (2.10) was defined as
a function of his information s,, the follower's inference s* and his
trade z. Under direct revelation s* is replaced by s. . So his equi-
librium trading strategy is simply the solution to:

(5.2) V (s. ,s. ,z) = 0.

z 1 1

V is the partial derivative of V with respect to z. The follow-
z ' '

ing proposition describes the direct revelation equilibrium and shows
that the leader always does better from revealing his information
directly as compared to having it inferred indirectly through his
t rades .

9

In other words, moral hazard problems associated with sales and

purchases of information are ignored.

-25-

Proposition 5.1. (i) The linear strategy equilibrium of the direct
revelation model is given by

KjCsj) = tlSl/2r, X2(Sl,s2) = [s1(0.5-T)+Ts2]/2r,

r = (E.M/E )1/2/2
0 u

3tj Ttl-tj) T (t1t2-l)

where M = —i— + „• + and T is defined in (2.3).

2 t2

(ii) The leader's equilibrium trades and expected profits are
always greater than their levels in the model with perfect piggy-
backing.

Proof : Part (i) follows directly from solving (5.1) to get the

equilibrium X (s ) and then substituting back into (5.1) to get the

equilibrium X„ . Y is obtained using (2.4). Part (ii) is proved in
the appendix.

( b ) Many Followers

Here 1 consider the effect of having several followers piggyback
on the leader's trade. For expositional simplicity, T consider the
case where n followers observe the leader's trade x„ but have no
information of their own. Assume that all trading strategies are
linear and take the following form.

(5.3) x„ = A s

< . - Cx^ , i - 1 , . . . , n.
l 0 ' ' '

-26-

Given a trade z by the leader and trades x. by followers j * i ,
follower i chooses x. to maximize his expected profits conditional on
his inference about s, . In other words, after observing z, the n
followers play a Cournot game when choosing their trades. Then, x.
solves :

Max [E(v|s*) - Tx - Tx. - T Z x.]x..

x. j*i

1 J

Assume that the n followers are symmetric so that x. = x for each

i = 1,2, ...,n. Solving for the followers' and leader's trades in the

usual way,

ts (1-n)
(5.4) x = — iyjr

(5.5) xQ = x/(l-n).

Finally, T is solved using the efficiency condition (2.4). A case
of particular interest is the equilibrium outcome when the number of
followers increases without bound.

Proposition 5.2. As the number of followers go to infinity,
(i) Each follower's trades go to zero,
(ii) Total trades of all followers increase without bound.

proof : Follows directly from (5.5).

( c ) Non-Linear Trading Strategies

So far, 1 have only worked with trading strategies which are
linear in signals because of their tractability when working out

-27-

coraparative static results. However, Theorem 1 in Mailath [1987] pro-
vides a general method of solving for separating equilibria when
strategies are dif f erentiable.

Consider again the V function defined in (2.10).

. t.(l-T)x"1(z) t.Ts. p
VlSpX. (z),z) = (t1s1 2 2 2~ *

Then z(s.) is a separating equilibrium strategy if it satisfies:

-V (s ,s ,z)
(5'6) T-' v t \~

Assume that s. e (-00 s.] and z(s, ) = t,s1/2r. Solving the dif-
ferential equation (5.6) yields the following unique non-linear
separating strategy equilibrium.

(5.7) z(Sl) = [CK1/T + t^TSjim

where K = 2T - (t.Ts./z),

C - [t1s1(l-T)]/[2r(l-T)]1/T

Details of the solutions are given in the appendix. Note, first,
that the linear solution z(s.) = t.Ts./2r is also a solution to (5.7)
and, second, that the leader plays his Stackelberg strategy when
s. = s. . For s, < s., trades are strictly below the Stackelberg level.
In general, therefore, the properties of this equilibrium is con-
sistent with those described for the linear separating equilibrium.
For all but the "best" information, piggybacking causes a reduction in
the leader's trades.

-28-

( d ) Randomization by the Leader

Since the leader's information is perfectly revealed in a separ-
ating equilibrium, incentives to mask information exist. In the semi-
pooling equilibria described in Corollary 3.2, information is revealed
imperfectly. Other papers in the literature allow the first-mover to
pre-commit to a randomization mechanism. Examples are Gould and
Verrecchia and also Grinblatt and Ross. In both these papers the
informed agent's pricing rule includes a noise terra which prevents
perfect information revelation.

Gould and Verrecchia justify the pre-commitment assumption by
referring to a noisy telephone connection between the informed
specialist and the trader. In this paper, the leader has not been
allowed to pre-commit to noise because it is hard to tell a plausible
story that could justify pre-commitment in the present context. The
noisy piggybacking model can be viewed as an alternative means of
introducing these considerations whereby life-cycle traders mas the
information of insiders.

Section 6. Policy Implications for Regulating
Ins id e_r^ Trad ing

How does trading by an insider affect uninformed outsiders? What

are the implications for a policy of regulating insider trading? The

model described previously can suggest answers to these questions if

it's assumed that the leader is not able to extract the full value of

his information within a short time span. Otherwise he would not

care about piggybacking. Also, since all traders are risk-neutral,

-29-

issues of risk-sharing addressed by Allen are ignored here. With
risk-averse traders, information-release may be sub-optimal.

6 . 1 Pre-Trade Disclosures

Observation of public statistics (such as the SEC's "Official
Summary of Insider Trading") is a noisy signal of insider information
to uninformed outsiders. I use the noisy piggybacking model of
Section 4 and put t = 0 (the follower becomes an outsider). This
will be the benchmark case of no regulation. Following Grinblatt [12],
I consider two regulatory regimes — pre-trade information disclosure
and disclosure of trading intentions.

Proposition _6_.l . With risk-neutral traders, a policy of mandating
information disclosure (ID) is superior to trade intentions (TI)
disclosure.

Proof: When t = 0, the ID policy leads to the Stackelberg outcome
described in Corollary 3.1 which is the first-best outcome in a
symmetric information world. The TI policy can be interpreted as
mandating the leader to reveal x . If traders follow linear trading
strategies, Corollary 3.2 says that markets break down. Under non-
linear trading strategies, Proposition 2.1 in Sarkar shows that
trades are below the Stackelberg volume almost everywhere.

If the ID policy cannot be credibly implemented due to moral
hazard problems, then a policy of not regulating insiders is second-
best because it avoids the piggybacking induced by the TI policy.
Without regulation, Corollary 4.1 states that the insider obtains

-30-

expected monopoly profits. To ensure participation of the outsider
he can be given a lump-sum grant equal to the expected Stackelberg
profit of the follower, financed by a tax on the leader's monopoly
profits. This still leaves the leader with profits higher than his
Stackelberg amount, so he has an incentive to remain the leader. The
scheme is feasible because expected monopoly profits exceed total
Stackelberg profits of the leader and follower combined.

6.2 Regulatory Impact on Efficiency, Liquidity,
and Price Volatility

Sequential trading with symmetrically informed leader and follower
(the Stackelberg outcome) leads to liquid markets, both in terras of
depth (prices are relatively insensitive to volume fluctuations) and
volume. Better information improves efficiency but reduces depth due
to the adverse selection effect on market-makers. Price volatility is
inversely related to depth and positively related to the informative-
ness of trades. So volatility also increases in response to better
inf orraat ion.

With the leader as insider and follower as uninformed outsider,
markets are less liquid (lower volume and depth) and less efficient
but with more stable prices. These effects arise due to strategic
behavior by the insider. Regulation of insider trading has ambiguous
effects on market parameters, however. The Tl policy, for example,
induces piggybacking which serves to restrict trading volumes and
lower market efficiency but may increase liquidity since market-makers
feel less threatened by insider trading. Further, as Grossman [1986]
has argued in the context of futures trading, if information is

-31-

acquired at a cost piggybacking allows traders to free-ride on infor-
mation collected and processed by other traders.

6 . 3 Are Piggybackers Ins id e rs?

There is some controversy as to whether piggybackers should fall
under current insider trading regulations since they may not have a
fiduciary duty to shareholders. In this model both the piggybacker
and the corporate insider profit at the expense of the liquidity
traders. More importantly, in the non-linear equilibrium, the piggy-
backer makes profits without having any independent information. So
the piggybacker also profits from the leader's information. The act
of piggybacking makes him an insider.

Section 7. Conclus ion

Does the intense scrutiny of insider trades act as a mechanism for
regulating insider trading? The question is explored in the context
of an asymmetric information Stackelberg model with the insider as the
leader and the piggybacker as the follower. Noise traders mask the
insiders' information and provide liquidity to the market. Market-
makers observe total trades and set prices which clear the market.

When piggybacking is perfect, the insider's trade is observed
perfectly. The insider tries to mislead the follower and signal poor
information by trading low. In a separating equilibrium, however the
leader does not succeed and his information is perfectly revealed.
But the attempt to do so restricts his equilibrium trading volume. So
perfect piggybacking does regulate insider trades but at a cost. Be-
cause of the strategic cost of using information, the insider may

-32-

throw away valuable information. An increase in the precision of
information may cause insiders to trade less, decreasing the inforraa-
tiveness of prices.

Under noisy piggybacking, the follower observes the sum of noise
trades and the leader's trade. So piggybacking conveys only statis-
tical information. I find that a small amount of noise is sufficient
to drown the piggybacking effect. The insider's trades increase with
respect to both his information precision and the amount of noise
trading.

Although intuitively appealing, 1 suggest why randomizing by the
leader is not plausible in this model. Also, if there is a credible
way for the leader to reveal his information to the follower, he could
avoid piggybacking and be better off.

The results of the models are used to analyze pre-trade disclosure
laws. It is suggested that a policy of requiring trade intentions
disclosure from insiders will have adverse market effects. Instead,
a policy of deregulating insider trading along with a tax on insiders'
profits used to subsidize outsider traders is proposed.

-33-

APPENDIX

Sections 2 and 3

First, (2.7) and (2.8) are derived. Since v, s. , and s? are
normally distributed E(v|s*,s9) will be linear in s* and s?. Suppose
that E(v|s*,s~) - c.s* + c?s„, where c. and c„ are given by the
normal equation c = Z cov(v,s). c,s are 2x1 vectors [c. c_] and
[s* s„ ] and Z is the variance-covariance matrix of s. The result is
easily obtained using E(v|s.) = t.s., i = 1,2.

Second, T in (3.3) is derived. T is the regression coefficient of
the linear projection of v on y, where y = x.+x9 + u.

[A1(1+B1)+B2]EQ

[A1(l+B1)+32]2^i+B22CS2 + Eij+2A1(l+B1)B2E0

Substituting the equilibrium values for A., B,, and B~ from (3.3)
in the text gives the following quadratic equation in V

(Al) 2r2VV-tl " T2(l"5t1t2/4)/t2]/2 = 0

Since the second-order condition requires r > 0, T is obtained as the

2
positive square root of T~ that solves (Al).

Proof of Corollary 3. 1

s. = s„ = s and define t = Zn/ (Z +Z ). It is immediate that
12 0 0s

E(v]s*,s) = E(v|s) = ts. The follower's maximization problem yields
the first-order condition (given some trade z by the leader)
x„ = [ts - r z]/2T . Solving the leader's maximization problem yields
the following equilibrium solution.

-34-

(3tE /E )1/2

(A2) z = ts/2r x2 = ts/4r , rg = ~-^

Finally, solve for the inf orraat iveness of the price system.

Var(v|p) = E[v-E(v|p)]2 = E(v-p)2 = EQ( 1-1 lt/16). Part (iii) follows

because T is increasing in t. Var(v|p) is decreasing in t which

proves part (ii). z is increasing in t, as stated. To show that z

is higher than its value in (3.3) note that (1) t < 1 and (2) r > V

— s

3t t (1-T2)

because Q = —j— + -. + T( 1-t, ) > 3t. /4.

* 4 4 11

Proof of Proposition 3.2

I show that (I) 6A./6t. > 0 at t, - 0 (2) SAj/St < 0 at t. - 1

and (3) A, is concave.

-1/2
A, = t,TQ ignoring the terms not involving T. and Z
* ° e p

12

Differentiating with respect to t. :

At t = 1, 5A /5t = -t2 < 0 because T = 0, Q = 1, and 5T/5t1 = -t0

1 /2
At t{ = 0, 5A./5t. = t9 > 0 because T = Q = t2 and &T/St{ = 0.

To show concavity, calculate the second derivative of A,:

r , "SA ?

(A3) T~ (7- * TT-) = -l/2t/ + (T-5T/5t.)/t,t9Q -
ot A. ot . I 1 I Z

(T2-5Q/5t1)/2t1t2Q2 + (ST/St^/TQ - [t l (5T/6t1 )2] /T2Q -

[t1(6Q/5t1)-(5T/5t1)]/TQ2

wh

ere (-SQ/St^/Q = 1/t. + (2T • 5T/ 5t { /Qt2 - ( 5t jT -5T/ 5t [ ) /2Q.

-35-

Substitute the above expression into the RHS of (A3) and rewrite the
RHS of A3:

(A4) - l/2tl2 + T4/(2tl2t22Q2) - T2/(2tl2t2Q)

+ 6T/5t1[T/tlt2Q - T3/t1t22Q + (5T3) /4tj t2Q2 ]
+ (5T/5tl)2 [5tl2/2 - 2t1/t2 - t1/T2]/Q2

Consider the first three terras of (A4):

(A5) --4- [1 +f- ■ (1 J T >] < 0

2tL 2^ n2M

2
Proof: Q >_ 0 implies t2Q - T" >_ t x t „ ( I |-).

2
If T < 4/5, then we are done.

Suppose not and assume T = 1, the maximum possible value of T.

T2 (I - T2) Cl

Then 1 + — -- • — -±-^ = 1 - 77-= .. . > 0.

t2Q t2Q 4(l-t1/4)2

Next, consider the terras involving 5T/5t. in (A4).

(A6) T.5T/5t1-[(t2Q - T2 + - t { t^2 ) /t 1 t2V ]

2 2
= T«5T/6t -t t2/t t2 Q < 0 as 5T/5t. < 0.

Finally, arrange the last three terras of (A4);

(\7) (6T/6tl)2 • [(5tl2t2T2 - 4tlT2 - 2t { t2 ) /2t 2T2 ] /Q2

2 2

The numerator factors to 4t T (t t -1) + t t (t .T -2) < 0 because

C , < 1 , t < 1 and T < 1 .

9 2
(\5)-(A7) together proves 5*"A./et. < 0.

-36-

Proof o£ Proposition 3.2

Var(v|p) = E(v-p)2 = EQ( l-T/2 ) ( 1-t { 11 )
Define price inf orraativeness PI as PI = Z Q[ l-( l-T/2) ( 1-t /2) ]
(A8) 6Pl/6t1 = [l-T/2 + 6T/6t1.(l-t1/2)].EQ/2

where 5T/6t1 = [ t2( t2>l ) ] /( 1-t { t2 )2

Fix t~ and let t. approach 1. In the limit, the RHS of (A8) tends to
[1 - j , . — A • -t— which is negative for t~ sufficiently close to I.

Section 4

(4.2) is derived using the same technique for deriving (2.8) in
the Section 2 appendix.

Proof of Proposition 4.1

Substituting a. and a? from (4.2) into (4.3) yields the following
cubic equation in \:

(\9) A. ' + pAj" + q\ + tt = 0

where p - -t,T/2T, q = t . (1-t )/t (l-t.t0)

1 1 ' 1 u u 1 2

tt = -t1q(i-t2)/2r

The standard cubic solution (using reduced form) is applied to
(A9) to give A = + t T/6T where

-37-

(A10) = L(l-m/3), L = [-n/2 + (n2/4 + m3/27 )1/2 ] 1/3 ,

m = (3q - p2)/3, n = (2p3 - 9pq + 27tt)/27

Given A. , the variables a. , a» , B, , and B~ are also defined.

Proof of Corollary 4.1

For a monopolist insider, the problem is

max E[ (v-rx-Ty)x | s]
x

which gives as the first-order condition x = t,s/2r. The monopoly

equilibrium is then:

(All) X(s) = A.s where A. = t,/2r, r(t. ZjZ )1/2/2

1 11 1 0 u

T is derived using the condition Y = cov«(v,y)/E . The (uncondi-
tional) expected profits of the insider is:

(A12) EIT1 = E[(v-rA1s-rp)A1s] - (tjE EQ)1/2/2

Now substitute t„ = 0 in equations (4.1)-(4.4) of the text. This
gives :

(All) a0 =0, Aj - t.(r+a.), B? = 0, and B + a,/2T

Z n[A,(l+B, ) + 3„]

(Ai4) r= 2° 1 2 -2-

Ey^l+B^ + B2 Es2 + 2B2(1+B1)A1S0

VAi

Using (A13), T = ■= — 7TTZ—<; which can be rewritten as T = a.(l+B.)
i.y . (, i+d ) ll

using the definition a, = k.ZQ/Zy . Solve (A13) and (A14) to get

-38-

B =0, r = a. , A. = t.s/2T, B„ = 0. Finally, solve the equation

r = a, using the definition of a. and t in (4.2).
1 1 u

(Ai5) r = (t1tu)1/2/s(l-tu)1/2 = (tl?:0/s0)1/2/2

Comparing with (All), both V and A. are at their monopoly levels.
3 = B„ = 0 so the follower does not trade. It is easily checked
that the leader's expected profits are also at their monopoly levels

Proof of Proposition 5.1

,P =

To prove part (ii), remember that AF. = t.T/2r in the model with

perfect piggybacking with A . = t./2F in the information-sharing

model. The result follows because (1) T < 1 and (2) T > TT .

- P

Claim 1: T < 1

Proof: T = (t„-t t2)/(l-t.t2) ( 1 as t„ < 1

Claim 2: T > TT
p - I

Proof: From the defi

nition of T in (3.3) and TT in Proposition

P L

5.1, need only show M < Q.

M = 3t1/4 + Td-t^/2 + T (t1t2-l)/t2

< 3t./4 + Td-tp/2 as tt < 1, i = 1, 2

Q = 3tj/4 + T( l-tj) + tjCl-T )/4

> 3tj/4 + KL-tp as T < 1

> M

-39-

Proof of Non-Linear Equilibrium

The equilibrium is obtained by solving the following differential
equation:

(A19) dz/ds1 = t.Cl-D-z/CtjSj-Zrz)

(A19) is of the form dz/ds. = f(z/s.). Make the substitution
g = z/s. and solve.

(A20) Sj = ceG, G = /dg/[f(g)-g] where f(g) is the RHS of (A19).

G = /dg(t1-2rg)/[g(2rg-t1T)]

= i log[(2rg-t1T)/g] - log[2rg-tlT]

Lastly, C is solved using the boundary condition z = t.s ./2V at

si " V

-40-
REFERENCES

Allen, B., 1981, "Generic Existence of Completely Revealing Equilibria
for Economies with Uncertainty when Prices Convey Information,"
Econometrica, 1173-1199

Allen, Franklin, 1987, "Risk-Sharing, Taxing Information and Insider
Trading," Working Paper, Department of Finance, University of
Pennsylvania.

Bagehot, W. , pseud. 1971, "The Only Game in Town," Financial Analyst
Journal , 2, 12-14.

Diamond, Douglas W. and Robert E. Verrecchia, 1981, "Information

Aggregation in a Noisy Rational Expectations Economy," Journal of
Financial Economics , 9, 221-235.

Dye, Ronald A., 1984, "Insider Trading and Incentives," Journal of
Business, Vol. 57, No. 3.

Easley, David and Maureen O'hara, 1987, "Price, Trade Size, and
Information in Securities Markets," Journal of Financial
Economics, 19, 69-90.

Finnerty, J. E., 1976, "Insiders and Market Efficiency," Journal of
Finance, 31, 1141-1148.

Gal-Or, E. , 1987, "First Mover Disadvantages with Private Infor-
mation," Review of Economic Studies, LIV , 279-292 .

Givoly, Dan and Palmon, Dan, 1985, "Insider Trading and the Exploita-
tion of Inside Information — Some Empirical Evidence," Journal of
Business, Vol. 58, No. I.

-41-

Glosten, L. L. and P. R. Milgrom, 1985, "Bid, Ask and Transaction

Prices in a Specialist Market with Heterogeneous Informed

Traders," Journal of Financial Economics, 14, 71-100.
Gould, John P. and Robert E. Verrecchia, 1985, "The Information

Content of Specialist Pricing," Journal of Political Economy, Vol

93, No. 1, 66-81.
Grinblatt, Mark S., 1987, "On the Regulation of Insider Trading,"

Working Paper, Graduate School of Management, UCLA.
Grinblatt, Mark S. and Stephen A. Ross, 1985, "Market Power in a

Securities Market with Endogenous Information," Quarterly Journal

of Economics, November.
Grossman, Stanford J., 1986, "An Analysis of the Role of 'Insider

Trading' on Futures Markets," Journal of Business, 59, No. 2,

S129-S146.
Grossman, S. J. and J. E. Stiglitz, 1981, "On the Impossibility of

Inf orraationally Efficient Markets," American Economic Review, 70,

393-408.
Hellwig, M. F., 1980, "On the Aggregation of Information in

Competitive Markets," Journal of Economic Theory, 22, 477-498.
Jaffe, Jeffrey, 1974, "Special Information and Insider Trading,"

Journal of Business , 48, 410-428.
Kihlstrom, Richard and Andrew Postlewaite, 1987, "Equilibrium in a

Securities Market with a Dominant Trader Possessing Inside

Information," Working Paper, Department of Economics, University

of Pennsylvania.

-42-

Kreps, David A. and Robert Wilson, 1982, "Sequential Equilibria,"

Econometrica, 50, 863-894.
Kyle, A. S., 1985, "Continuous Auctions and Insider Trading,"

Econometrica, 53, 1315-1335.
Madhavan, Anath, 1988, "Trading Mechanisms in Securities Markets,"

Working Paper, The Wharton School, University of Pennsylvania.
Mailath, George, 1987, "Incentive Compatibility in Signalling Games

with a Continuum of Types," Econometrica, 55, 1349-1365.
Manne , H. G., 1966, Insider Trades and the Stock Market, Free Press,

New York.
Mathews, Steven A. and Leonard J. Mirraan, 1983, "Equilibrium Limit

Pricing: The Effects of Private Information and Stochastic

Demand," Econometrica, 51, No. 4, 981-996.
Milgrom, Paul and John Roberts, 1981, "Limit Pricing and Entry Under

Incomplete Information," Econometrica, 50, 443-459.
Sarkar, Asani , 1988, "Piggybacking on Insider Trades II: Some

Additional Results," Working Paper, Department of Economics,

University of Pennsylvania.
Scholes, M. S., 1972, "The Market for Securities: Substitution versus

Price Pressure and the Effects of Information on Share Prices,"

Journal of Business , 4 5, 179-211.
Seyhun, Nejat H., 1986, "Insiders' Profits, Costs of Trading and

Market Efficiency," Journal of Financial Economics, 16, 189-212.

D/135

LIST OF FIGURE LEGENDS

Page

1. Figure 1: Sequence of Moves 43

2. Figure 2: Linear Separating Equilibrium 44

3. Figure 3: Trading Strategy of Leader Versus Tl 45

4. Figure 4: Market Liquidity Versus Tl 46

5. Figure 5: Price Inf orraativeness Versus Tl 47

6. Figure 6: Leader Trade Versus Tl 48

-43-

1

2

3

S\ IS

observed

Leader chooses
xj . Noise-
traders choose
U

Fol lower
observes
xi and S2

Follower
chooses

*2

Market-
makers
observe
(XJ+X2+U)
and choose
P

v is
realized

Figure 1. SEQUENCE OF MOVES

-44-

(M

m

(N

Q
<
04

H

oi
W
Q
<

w

o

o
o

OS

w

c
•J
►J
o

u

z

Cs3

t—

u

w

z

<

w

w

OS

OS

u:

u

u.

z

UZ

t— 1

-.£.

>

-45-

>

Q
<

o

o
w

<
K

E-<
CO

o

I— I

Q
<

E-

cn
O

o>

o

oo
o

r--
o

o

o

CI

o

fN
O

in
O

w

E-i
<
OS
-
to

2

M

Q
<

K
H

OS
U
Q
<

in

ON

c

o

in

o

m

o

in

o

00

r^

r-

VO

V£>

in

in

in

T '
O

T '

in

T

o

m

m

CM

(N

— i — ' — r
in o
o o

-46-

o

CO

o

o

o

o

>- in

[ '

in

T " '

O

in

o

■ 'T ■

m

^

v£>

U1

in

T

- 1 —
in

— i —
in

~[ —
o

-I —
in

— i —
o

-i —
in
o

r^o

H

■<r

^2

o

22
2M

o

M

o

o

3
O

2

-47-

m

D
w

>

w
w
w

>

E-
<

s

o
i— i
o

I— I

K

a.

m

o

i— i

En

ON

o

CO

o

o

o

o

m
o

O

n —
c/5 in
to r~
U •
2 O
U
>
H

o

— i —
in

l£>

-i —
o

—I —

in

— i —
o
in

in

o

-i —
o

-i —
m

-i —
o

fN

in
o

T
O

O

O

o

O

Cu

2

M

u
u

M

OS
0,

-48-

(T\

o

<J\

9

?

CNJ

Eh

(£

«r

en

o

,9

o>
o

CO

o

o

o

T-

M O
Q fN

< •
as o

OS
U

Q
<
U
►J

o

-I —
O
(N

O
fN

- 1 —
O

-I —
O
fN

- 1 —

o

fN

-I ' 1 —

O O

(N fN

r"> fN

-1 —
O
fN

O
fN

"1 —
03

o

— I —
vo
o

fN
O

o
o

OS

w

o
J

O
b

O

W
2Q

o<

mqS
«H

H

CJW
WW

OS HI

0,0
2
2

ODu

MO

H

<W

SOS

K3

<D</>

Du<

ZH

m2

4fN3

CKMAN IXl

)ERY INC. |§|

JUN95

o-lW N. MANCHESTER.
INDIANA 46962