Limit Order Book

Limit Order Book as a Market for Liquidity
Author(s): Thierry Foucault, Ohad Kadan and Eugene Kandel

Source: The Review of Financial Studies , Winter, 2005, Vol. 18, No. 4 (Winter, 2005),
pp. 1171-1217
Published by: Oxford University Press. Sponsor: The Society for Financial Studies.
Stable URL: https://www.jstor.org/stable/3598019
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
and Oxford University Press are collaborating with JSTOR to digitize, preserve and extend
access to The Review of Financial Studies
This content downloaded from

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
All use subject to https://about.jstor.org/terms
Thierry Foucault
HEC School of Management, Paris and CEPR
Ohad Kadan
John M. Olin School of Business, Washington University in St. Louis
Eugene Kandel
School of Business Administration, and Department of Economics,
Hebrew University
We develop a dynamic model of a limit order market populated by strategic liquidity

traders of varying impatience. In equilibrium, patient traders tend to submit limit
orders, whereas impatient traders submit market orders. Two variables are the key
determinants of the limit order book dynamics in equilibrium: the proportion of
patient traders and the order arrival rate. We offer several testable implications for
various market quality measures such as spread, trading frequency, market resiliency,
and time to execution for limit orders. Finally, we show the effect of imposing a
minimal price variation on these measures.
The timing of trading needs is not synchronized across investors, yet trade
execution requires that counterparties trade simultaneously. Markets
address this problem in one of the three ways: call auctions, dealer
markets, and limit order markets. Call auctions require participants to
either wait or trade ahead of their desired time; no one gets immediacy,
unless by chance. Dealer markets, on the contrary, provide immediacy to
all at the same price, whether it is desired or not. Finally, a limit order
market allows investors to demand immediacy, or supply it, according to
their choice. The growing importance of limit order markets suggests that
this feature is valuable, which in turn implies that traders value order
We thank Bruno Biais, David Easley, Larry Glosten, Larry Harris, Joel Hasbrouck, Frank de Jong, Pete
Kyle, Leslie Marx, Narayan Naik, Maureen O'Hara (the editor), Christine Parlour, Patrik Sandas,
Duane Seppi, Ilya Strebulaev, Isabel Tkach, Avi Wohl, and two anonymous referees for very helpful
comments and suggestions. Comments by seminar participants at Amsterdam, BGU, Bar Ilan, CREST,
Emory, Hebrew, Illinois, Insead, LBS, Stockholm, Tel Aviv, Thema, Toulouse, Wharton, and by
participants at the Western Finance Association 2001 meeting, the CEPR 2001 Symposium at Gerzensee,
and the RFS 2002 Imperfect Markets Conference have been very helpful as well. The authors thank the
J. Nachmias Fund, and The Kruger Center at Hebrew University for financial support. Thierry Foucault
is also affiliated with GREGHEC. He thanks la Fondation HEC for additional financial support.
Finally, we are grateful to Michael Borns for expert editorial assistance. Address correspondence to
Eugene Kandel, Department of Economics, Hebrew University, Mt. Scopus, Jerusalem, Israel, 91905",
and CEPR, or e-mail:[email protected].
? The Author 2005. Published by Oxford University Press. All rights reserved. For Permissions, please email:
[email protected]
doi: 10.1093/rfs/hhi029 Advance Access publication August 31, 2005

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
The Review of Financial Studies I v 18 n 4 2005
execution speed differently.1 In this article, we explore a dynamic model

of limit order trading in which traders differ in their level of impatience.
Limit and market orders constitute the core of any continuous limit
order trading system. A market order guarantees immediate execution at
the best price available upon the order arrival. It represents demand for
the immediacy of execution. With a limit order, a trader can improve the
execution price relative to the market order price, but the execution is
neither immediate nor certain. A limit order represents supply of imme-
diacy to other traders. The optimal order choice ultimately involves a
trade-off between the cost of delayed execution and the cost of immedi-
acy. This trade-off was first suggested by Demsetz (1968, p. 41), who
states: "Waiting costs are relatively important for trading in organized
markets, and would seem to dominate the determination of spreads." He
argued that more aggressive limit orders would be submitted to shorten
the expected time to execution, driving the book dynamics.2
Building on this idea, we study how traders' impatience affects order
placement strategies, bid-ask spread dynamics, and market resiliency.
Black (1971) and Kyle (1985) define a liquid market as being (a) tight-
small spreads, (b) deep-small price impact, and (c) resilient-prices
recover quickly after liquidity demand shocks [see also Harris (2003),
p. 400]. The determinants of spreads and market depth have been exten-
sively analyzed. In contrast, market resiliency, an inherently dynamic
phenomenon, has received little attention in theoretical research. Yet,
mean reversion in the spread following liquidity shocks is an important
feature of data on order driven markets.3 Our dynamic equilibrium
framework allows us to fill this gap.
The model features buyers and sellers arriving sequentially. All are
liquidity traders who would like to buy/sell one unit regardless of the
prevailing price. Traders differ in their costs of delaying execution: they
arrive randomly as either patient (low waiting cost) or impatient (high
waiting cost). Each trader chooses to place a market or a limit order,
conditional on the state of the book, so as to minimize his total execution
cost which includes the cost of waiting. Under several simplifying
assumptions, we derive (i) the equilibrium order placement strategies,
Jain (2002) shows that in the late 1990s, 48% of the 139 stock markets throughout the world are
organized as a pure limit order book, while another 14% are hybrid with the limit order book as the
core engine. Examples of limit order markets include Island and Euronext in equity markets, Reuters D-
2002 in the FX market or MTS in the bond market.
2 Demsetz (1968) focuses on the NYSE. This market is not a pure limit order market since liquidity is
supplied both by limit order traders and dealers (the specialists). Demsetz sees waiting costs as particu-
larly important for traders who choose between limit and market orders.
3For instance, Biais, Hillion and Spatt (1995) find that liquidity demand shocks, manifested by a sequence
of market orders, raise the spread. Then the spread reverts to the competitive level as liquidity suppliers
place new orders within the prevailing quotes. DeGryse et al. (2003) and Coppejans, Domowitz, and
Madhavan (2003) have also studied this phenomenon.
1172

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
(ii) the expected time to execution for limit orders, (iii) the stationary
probability distribution of the spread, and (iv) the expected duration
between trades (conditional on the size of the inside spread). In equili-
brium, patient traders tend to provide liquidity to less patient traders.
A string of market orders, that is, a liquidity shock, increases the
spread. We measure market resiliency by the probability that, after a
liquidity shock, the spread reverts to its former level before the next
transaction. Factors that induce traders to post-aggressive limit orders
make the market resilient. For instance, other things being equal, an
increase in the proportion of patient traders reduces liquidity demand
and lengthens the expected time to execution of limit orders. As a result,
liquidity suppliers submit more aggressive limit orders to reduce their
waiting times, in line with Demsetz's (1968) intuition. Consequently,
when the proportion of patient traders increases, the spread narrows
more rapidly, making the market more resilient. Higher arrival rates
translate into shorter waiting times for limit order traders. The latter
respond with less aggressive limit orders, thus more orders are required
before the spread reverts to its competitive level; consequently, market
resiliency decreases in the order arrival rate.
Interestingly, the distribution of spreads depends on the composition of
the trading population. Large spreads are more frequent in markets
dominated by impatient traders, because these markets are less resilient.
A reduction of the tick size in these markets can result in higher spreads.
Actually, this reduction impairs market resiliency by enabling traders to
bid less aggressively.
We derive several empirical predictions.4 The advent of high frequency
databases has spurred an interest in the role of time in the trading process
[e.g., Easley and O'Hara (1992), Engle and Russel (1998), Hasbrouck
(1999), and Lo, McKinlay, and Zhang (2001)]. The time between trades
in our model is endogenous since a transaction occurs when a trader opts
for a market order. We show that the average time until a transaction,
conditional on the quoted spread for the prior transaction, increases with
the size of the spread. Furthermore, there is a positive relation between
this conditional duration and market resiliency. This result stems from
the fact that, other things being equal, both market resiliency and the
expected duration between trades decrease with the proportion of impa-
tient traders. We also propose to explain intraday liquidity patterns by
time-series variations in the proportion of patient traders. Assuming
traders become more impatient over the course of the trading day, our
model predicts an increase in spreads and trading frequency toward the
The number of empirical papers on limit order markets is growing fast. These include Handa and
Schwartz (1996), Harris and Hasbrouck (1996), Kavajecz (1999), Sandas (2001), Hollifield, Miller, and
Sandas (2004), Hollifield et al. (2003), Kavajecz and Odders-White (2003) and other references which are
mentioned later in the article.
1173

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
The Review of Financial Studies / v 18 n 4 2005
end of the day. This should be concomitant with a decline in limit order
aggressiveness and market resiliency. Whereas the first two predictions
are consistent with the empirical findings, as far as we know the latter has
not yet been tested.
Most of the models in the theoretical literature such as Glosten (1994),
Chakravarty and Holden (1995), Rock (1996), Seppi (1997), Biais,
Martimort, and Rochet (2000), or Parlour and Seppi (2003) focus on
the optimal bidding strategies for limit order traders. These models are
static, which precludes the analysis of the determinants of market resi-
liency. Furthermore, the choice between market and limit orders is exo-
genous in these models. In particular, this choice is not explicitly related
to the level of waiting costs, as it is in our model.5
Parlour (1998), Foucault (1999), and Goettler, Parlour, and Rajan
(2003) study dynamic models.6 Parlour (1998) shows how the order
placement decision is influenced by the depth available at the inside
quotes. Foucault (1999) analyzes the impact of the risk of being picked
off on traders' order placement strategies. Goettler, Parlour, and Rajan
(2003) model limit order trading as a stochastic sequential game and
develop a technique to solve for the equilibrium numerically. In these
models, limit order traders do not bear waiting costs, and time to execu-
tion does not influence traders' bidding strategies.7 In contrast, time to
execution plays a central role here.
There is no asymmetric information among traders in our model. This
approach seems reasonable, as a first cut, since it is very difficult to solve
dynamic models with traders who can strategically choose between mar-
ket and limit orders. Frictions in our model (the bid-ask spread and the
waiting time) are due to (i) the waiting costs and (ii) strategic rent-seeking
by patient traders. Frictions that are not caused by informational asym-
metries appear to be large in practice [see Huang and Stoll (1997) or
Madhavan, Richardson, and Roomans (1997)]. For instance, Huang and
Stoll (1997) estimate that 88.8% of the bid-ask spread on average is due to
non-informational frictions (so-called order processing costs). Given this
evidence, it is important to understand the theory of price formation
when frictions are not due to informational asymmetries.
The article is organized as follows. Section 1 describes the model.
Section 2 derives the equilibrium of the limit order market and analyzes
the determinants of market resiliency. Section 3 discusses in detail the
5 In extant models, traders who submit limit orders may be seen as very patient, while those who submit
market orders may be seen as extremely impatient. We consider a less polar case.
6 Angel (1994), Domowitz and Wang (1994), and Harris (1998) study models with an exogenous order
flow. We use more restrictive assumptions on the primitives of the model that anable us to endogenize the
order flow. Rosu (2004) uses a similar approach."
7 In Parlour (1998) traders' utility does not depend on their execution timing during the market day, i.e.,
there is no cost of waiting.
1174

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
empirical implications of the model. In Section 4, we explore the effect of

a change in tick size on measures of market quality, and Section 5
addresses robustness issues. Section 6 concludes and discusses the limita-
tions of our approach. All proofs are in the Appendix.
1. Model
Consider a continuous market for a single security, organized as a limi

order book without intermediaries. We assume that latent information
about the security value determines the range, [B, A], of admissible prices.
Specifically, a competitive fringe of traders stands ready to sell and buy an
unlimited number of shares at prices A and B (A > B > 0), respectively.
We assume that A and B are constant over time, thus all the prices in the
limit order book stay in the range [B, A].8 The goal of this model is to
investigate price dynamics within this interval; these are determined by
the order submission strategies followed by the traders.
All prices and spreads, but not waiting costs and traders' valuations,
are placed on a discrete grid, and are expressed as a multiple of the tick
size, denoted by A > 0. The inside spread is s = a - b, where a and b are
the best ask and bid quotes in the market, expressed in number of ticks.
By construction, a < A, b > B, and s < K - A - B. Occasionally, we
express prices and spreads in monetary terms, rather than in number of
ticks, using a superscript "m", for example s' = sA.9 We omit time sub-
scripts on variables since we focus on stationary equilibria.
1.1 Timing
This is an infinite horizon model with a continuous time line. Traders
arrive at the market according to a Poisson process with parameter A > 0:
the number of traders arriving during a time interval of length T is
distributed according to a Poisson distribution with parameter AT. As a
result, the inter-arrival times are distributed exponentially, and the
expected time between arrivals is I. We refer to the time elapsed between
two consecutive trader arrivals as a period.
1.2 Patient and impatient traders

Traders are risk neutral. Each trader arrives as either a buyer or a seller for
one share of security. Let Vbuyer and Vseller be buyers' and sellers' valua-
tions. In order to justify our classification to buyers and sellers, we assume
that Vbuyer > AA and Vseller < BA. Upon arrival, traders observe the limit
order book and must submit an order. They can submit either (i) a market
8 Seppi (1997) and Parlour and Seppi (2003) use a similar specification of the admissible price range.
9 For instance, s = 4 means that the spread is equal to four ticks. If the tick is equal to $1, then the
corresponding spread expressed in dollars is s"' = $0.25.
1175

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
order, which gets immediate execution at the best quote or (ii) a limit order,
which results in a better execution price, but delays execution.
Traders bear waiting costs that are proportional to the amount of time
elapsed between their arrival and the completion of their transaction. Hence,
agents face a trade-off between the execution price and the time to execution.
Traders are not required to execute their trade by a fixed deadline but they
cannot choose not to trade [as in Admati and Pfleiderer (1988) for instance].
Both buyers and sellers can be of two types which differ by the magni-
tude of their waiting costs. Type P traders are relatively patient and incur
a waiting cost of 6p per unit of time. Type I traders are relatively impatient
and incur a waiting cost of 6I, where 6I > 6p > 0. The proportion of
patient traders in the population is Op (l>Op>0), and the proportion of
impatient traders is 0 = 1 - Op. These proportions remain constant over
time, and the arrival process is independent of the type distribution.
A patient trader represents, for example, a portfolio manager rebalan-
cing his portfolio due to considerations of long-term fundamental value (a
"value trader"). In contrast, arbitrageurs, technical traders, or indexers,
who seek to mimic the return on a specific stock or index, are impatient
traders. Keim and Madhavan (1995) provide evidence supporting this
interpretation. They find that indexers and technical traders are more likely
to place market orders, while value traders in general place limit orders.
Brokers executing agency trades are also impatient traders, since waiting
may result in a worse price and therefore could lead to claims of negligence.
1.3 Trading mechanism

Limit orders are stored in the limit order book and are executed in
sequence according to price priority. We make the following simplifying
assumptions about the market structure.
Assumption A.I. Each trader arrives only once, submits a market or a limit
order and exits. Submitted orders cannot be canceled or modified.
Assumption A.2. Limit orders must be price improving, that is, narrow the
spread by at least one tick.
Assumption A.3. Buyers and sellers alternate with certainty, for example,
first a buyer arrives, then a seller, then a buyer, and so on. The first trader is
a buyer with probability 0.5.
Assumptions A.1-A.3 facilitate the analysis of the trading game, pri-

marily for two reasons. First, they enable us to solve for the equilibria by
induction (see Section 2.1 for a detailed explanation). Second, they imply
that the order placement strategies depend only on the inside spread (and
not on all the orders in the book). In Section 5, we demonstrate using
1176

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
examples that the main implications and the economic intuitions of the
model persist when assumptions A.2 and A.3 are relaxed. We also explain
why relaxation of these assumptions increases the complexity of the
problem in a way that precludes a general analytical solution. Finally,
we discuss in Section 5 the limitations imposed by Assumption A.1,
namely that traders cannot cancel and resubmit their limit orders.
1.4 Order placement strategies

Let Pbuyer and Pseller be the execution prices of buyers and sellers, respec-
tively. A buyer either (i) submits a market order and pays the lowest ask a
or (ii) submits a limit buy order which narrows the spread. Similarly, a
seller either receives the largest bid b or submits a limit sell order. The
execution prices can be expressed as
Pbuyer = a - ; Pseller = b + j just j C {0,..., S - 1},
withj = 0 for a market order and j > 0 for a limit order creating a spread
of size j. Recall that s - a - b is the inside spread prior to the order
arrival. It is convenient to use j, rather than Pbuyer or Pseller, as the trader's
decision variable. We say that a trader uses a "j-limit order" when he posts
a limit order that creates a spread of j ticks. The expected time to
execution of aj-limit order is denoted by T(j). Thus, the expected waiting
cost of a j-limit order is SiT(j), i E {P,I}. As a market order entails
immediate execution, we have T(0) = 0.
The expected profit of trader i (i c {P, I}) who submits aj-limit order is:
i) | { Vbuyer -buyer\ - i T() = (Vbuyer - aA) +jA - 6i T() for a buyer

PsellerA - Vseller - 6iT) = (bA - Vseller) +jA - 6iT(j) for a seller.
Expressions in parenthesis represent profits associated with market order

submission. These profits are determined by a trader's valuation and the
best quotes in the market. It immediately follows that trader i (i c {P, I})
observing the spread s chooses optimally the order that solves:
max 7ri(') jA - 6iT('), (1)

jc{0,...s-l}
for buyers and sellers alike. An order placement strategy for trader i is a
mapping, oi(.), that assigns a j-limit order, j E {0,... ,s - 1}, to every
possible spread s E { 1,..., K}. If a trader is indifferent between two limit
orders with different prices, we assume that he submits the limit order
creating the larger spread.
Equilibrium definition. An equilibrium of the trading game is a pair of

order placement strategies, op(.) and o (-), such that the orders prescribed
by the strategies solve Program (1) when the expected waiting time, T* (),
is computed assuming that traders follow strategies op(.) and o*(.).
1177

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
The Reviev of Financial Studies I v 18 n 4 2005
Traders' optimal order placement strategies depend on the expected

waiting time function. In turn, the waiting time function is endogenous
and is determined by traders' order placement strategies. We will show
that the equilibrium waiting time function, T* (), is non-decreasing in j;
thus, traders face the following trade-off: a better execution price (larger
value of j) can only be obtained at the cost of a larger expected waiting
time. Finally, notice that we restrict our attention to stationary order
placement strategies and waiting time functions (i.e., the strategies and
waiting times do not depend on the time at which the order is submitted).
Hence, we only focus on the stationary equilibria of the trading game
analyzed in this article. This restriction is natural because all exogenous
parameters are assumed to be stationary.
2. Equilibrium Order Placement Strategies and Market Dynamics
In this section, we first characterize the equilibrium order placement

strategies. Then, we study how spreads evolve in between transactions
and analyze the determinants of market resiliency. We identify three
different patterns for the dynamics of the limit order book: (a) strongly
resilient, (b) resilient, and (c) weakly resilient. The pattern is determined
by the characteristics of the traders' population: (i) the proportion of
patient traders and (ii) the difference in waiting costs between patient and
impatient traders. These parameters also determine traders' bidding
aggressiveness and the resulting stationary distribution of spreads.
2.1 Expected waiting time

Consider a trader who chooses aj-limit order. Let ak 0() be the probability
that the next trader responds with a k-limit order, k e {0, 1,... ,j - 1}
(k = 0 stands for a market order). Lemma 1 characterizes the expected
waiting time for the order placed by the first trader as a function of the
next trader's order placement strategy [described by the ak0()]:
Lemma 1. The expected waiting time for the execution of aj-limit order is:
1
T(/) = if j 1,
* T() = +o0 if ao0() = 0 and C {2,... ,K - 1},
* ) T()-'-o) a 3-k ak(' ) T(k) if ao(/)>0 and j E {2,...
Assumption A.2 implies that a trader, who faces a one-tick

submits a market order, thus T(1) = , that is, the average time
two arrivals. The expected waiting time of a j-limit order th
1178

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
attracts a market order [i.e., such that ao(/) = 0] is obviously infinite. If

aoC) > 0, the expected waiting time of aj-limit order is a function of the
expected waiting times of the subsequent orders that create a smaller
spread. This implies that the expected waiting time function is recursive.
As the expected waiting time function is recursive, we can solve the
game by induction. To see this, consider a trader who arrives when the
spread is s = 2. The trader can submit either a market order or a one-tick
limit order. The latter improves his execution price by one tick and results
in an expected waiting time equal to T(1) = 1/A. Solving Program (1) for
patient and impatient traders, we determine ak(2) for k = 0 and k = 1. If
no trader submits a market order then ao(2) = 0, and the expected wait-
ing time for a j-limit order is infinite for any j > 2 (Lemma 1). It follows
that no spread larger than one tick can be observed in equilibrium. If
instead we find that either patient or impatient traders submit market
orders then aot(2) > 0 and we compute T(2) using the Lemma 1. Next we
proceed to s = 3 and so forth. As we proceed by induction, each type of
trader has a unique optimal order placement strategy, and therefore the
stationary equilibrium is unique.
The expected waiting time function has a recursive structure because
our assumptions yield a simple ordering of the queue of unfilled limit
orders in the book: a limit order is never executed before limit orders that
create a smaller spread. Hence, the waiting time of a j-limit order is a
function of the waiting times of limit orders that create smaller spreads.
This ordering (and therefore Lemma 1) does not hold if buyers and sellers
arrive randomly. Consider a buyer creating a spread of j ticks, followed
by a seller creating a spread ofj'(/'<j) ticks. If the next trader is a second
seller submitting a market order, the buyer is executed before the first
seller. Assumption A.3 rules out this case. We consider the case in which
this assumption does not hold in Section 5.
2.2 Equilibrium strategies

Recall that the payoff of a trader submitting a j-limit order is
7ri() =jA - 6iT(),
and that this payoff is zero for a market order. Hence, a trader submits a
j-limit order only if price improvement, jA, exceeds waiting cost, 6iT(j). A
trader submitting a limit order expects to wait at least one period before
execution. As the average duration of a period is 1, the expected waiting cost
for a trader with type i is at least i. It follows that the smallest spread trader
i can establish is the smallest integer j*, such that 7ri (ii) =jl - - > 0. We
call j the "reservation spread" of a trader with type i. Let CF(x) denote the
ceilingfunction, the smallest integer larger than or equal to x [e.g., CF(2.4) = 3,
and CF(2) = 2]. Then the reservation spread is:
1179

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
ji ( CF(-) i E {PI} (2)

To exclude cases in which no trader submits limit orders, w
jp < K. (3)
We call patient traders'

because no trader will pos
tion spread of a patient
but the two can be equal.
and impatient traders
Otherwise we say that tra
2.2.1 The homogeneous

ders prefer to submit a m
toj* (by definition of the
time for aj*-limit order i
7i('*) > 0 for i e {P,I}. (4)
Now, consider a trader who faces a book with an in

7i(j*) > 0, the trader (patient or impatient) prefers
order to a market order. Hence, a limit order whi
larger thanj* is never executed. It follows that it is ne
spread larger than j*. These observations lead us to
Proposition 1. When traders are homogeneous (i.e.,

equilibrium, all traders submit a market order if s < j*
order ifs > j*.
The equilibrium with homogeneous traders has two distinctive proper-

ties. First, all limit order traders post the competitive spread, j*. Second,
the spread oscillates between K and the competitive spread, and transac-
tions take place only when the spread is competitive. Trade prices are
either A -j* if the first trader is a buyer or B + j*, if the first trader is a
seller. We refer to this market as strongly resilient, since any deviation
from the competitive spread is immediately corrected by the next trader.
The dynamics of the bid-ask spread in the homogeneous case look quite
unusual, but they are not unrealistic. Biais, Hillion, and Spatt (1995)
identify several typical patterns for the dynamics of the bid-ask spread
in the Paris Bourse. Interestingly, they identify precisely the pattern we
obtain when traders are homogeneous (Figure 3B, p. 1681): the spread
alternates between a large and a small size, and all transactions take place
1180

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
when the spread is small. Given that this case requires that all traders
have identical reservation spreads, we anticipate that this pattern is not
frequent. It does, however, provide a useful benchmark for the results
obtained in the heterogeneous trader case.
2.2.2 The heterogeneous case. Now consider heterogeneous traders, that

is, jp<ji. In this case, there are spreads above patient traders' reservation
spread for which impatient traders choose to submit market orders. We
denote by (jl,j2) the set: {i,jl + ,jl + 2,... ,j2}, that is, the set of all
spreads between any two spreads ji < j2 (inclusive).
Proposition 2. Suppose traders are heterogeneous (jp<j1). In equilibrium

there exists a cutoff spread Sc c (ji, K) such that:
1. Facing a spread s E (1 ,j), both patient and impatient traders submit

a market order.
2. Facing a spread s c (jp + 1, sc), a patient trader submits a limit

order and an impatient trader submits a market order.
3. Facing a spread s e (sc + 1, K), both patient and impatient traders
submit limit orders creating a spread of s.
The range of possible spreads is partitioned into three regions:

(i) s < jp, (ii) jp <s < Sc, and (iii) s > sc. The reservation spread of the
patient trader, j,, is the smallest spread observed in the market, while the
cutoff spread s, is the largest spread at which market orders are sub-
mitted. When the spread is larger than sc, all traders submit limit orders.
Limit orders that create a spread larger than sc are never executed, hence
are not submitted in equilibrium. Impatient traders demand liquidity by
submitting market orders at spreads below sc. Patient traders supply
liquidity at spreads above their reservation spread and demand liquidity
at spreads smaller than their reservation spread. It follows that the rate at
which market orders are submitted is decreasing with the size of the spread.
In contrast to the homogeneous case, the inside spread may exceed an
impatient trader's reservation spread, and yet this trader submits a mar-
ket order (because in general sc > j]). The explanation for this result is as
follows. Patient traders do not submit market orders when the inside
spread is larger than jp. Consequently, the expected time to execution of
a j-limit order (j > jp) is strictly larger than one period. In particular
T* (j) > -. For this reason, the expected waiting cost for an impatient
trader posting his reservation spread exceeds, in general, his improvement
in execution price [i.e., 7r (j) =jA - 6 T* (j) < 0]. In this case, the
cutoff spread (s) at which impatient traders can profitably enter limit
orders is strictly larger than their reservation spread. In many cases, the
waiting costs increase so quickly with the size of the spread that impatient
1181

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
traders never find it optimal to enter limit orders at prices in the eligible
range. In these cases, we set s, = K. This occurs, for instance, when the
cost of waiting for an impatient trader is sufficiently large.10
The cases in which sc < K and the case in which s, = K are qualitatively
similar. The primary difference is that impatient traders never find it
optimal to submit limit orders in the latter case while they do in the
former case (when the spread is K). Henceforth, we focus our attention
on the cases where Sc = K. This restriction has no significant impact on
results, but shortens the presentation.
Proposition 3. In equilibrium, there exist q spreads (K > q > 2),

nl < n2 < ... < nq, with nl =-j, and nq = K, such that the optimal order
submission strategy is.
* An impatient trader submits a market order for any spread in (1, K).
* A patient trader submits a market order when he faces a spread in
(1,nl), and submits an nh-limit order when he faces a spread
in (nh + 1,nh+l)for h- ,...,q - 1.
Thus, when a patient trader faces a spread nh+l (h > 1), he responds by
submitting a limit order which improves the spread by (nh+l - nh) ticks.
This order establishes a new spread equal to nh. This process continues
until a market order arrives.
The next two propositions provide a closed-form solution for the
expected waiting time function and for the equilibrium spreads. Let
p = - be the ratio of the proportions of patient and impatient traders.
Intuitively, when this ratio is smaller (larger) than 1, liquidity is consumed
more (less) quickly than it is supplied since impatient traders submit
market orders, and patient traders tend to submit limit orders. In equili-
brium, the expected waiting time for a limit order is a function of p.
Proposition 4. The expected waiting time function in equilibrium is.
1 * 1 h-1
T*(n) = ; T(nh)=- 1 +2 k=l
and
T*(') = T*(nh) j (nh- + 1, nh) V h = 1,..., q - 1.11
1' Obviously Sc = K ifj > K. It is worth stressing that this condition is sufficient, but not necess
the numerical examples below, j, is much smaller than K, but sc = K.
l We set no = 0 by convention.
1182

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
Recall that a limit order is never executed before limit orders that create
smaller spreads. For this reason, the choice of a spread is tantamount to
the choice of a priority level in a waiting line: the smaller is the spread
chosen by a trader, the higher is his priority in the queue of unfilled limit
orders. Accordingly, the expected waiting time function increases with the
spread. This property is consistent with evidence in Lo, McKinley, and
Zhang (2001) who find that the time to execution of limit orders increases
in the distance between the limit order price and the mid-quote.
Consider a trader facing a spread nh+l (h < q - 1). In equilibrium, he
submits an nh-limit order (Proposition 3). He could reduce his expected
time to execution by submitting an nh_ -limit order, but chooses not to.
Thus the following condition must hold:
nhA - T*(h)P n - T*(nh) > _/- ( )6p, Vh C {2,... q-1},

or
th -nh- h-l > [T*(nh) T*(nh_1)], Vh C {2,...,q-1}. (6)
Now consider a trader facing a spread nh. In equilibrium, this trader

submits an nh_ -limit order. Thus, he must prefer this limit order to a
limit order which creates a spread of (nh - 1) ticks, which imposes
nh-lA - T*(nh-1)6P > (nh - 1)A - T*(nh - 1)6p Vh E {2,..., q}.
In equilibrium, either (i) T*(nh - 1) = T*(nh) or (ii) T*(nh - 1) = T* (nh-).

In both cases T* (nh)> T* (nh-_), hence
6p
th < [T*(nh)- T*(nh_-)] + 1 vh CE 2,..., q}. (7)
Combining Equations (6) and (7), we deduce that
h = CF([T*(nh) - T*(nhl)]) - CF(2ph- )Vh c {2,.., q- 1},
(8)
where the last equality follows from Proposition 4. Using Equation (8),
we derive the set of equilibrium spreads nl, n,..., nq.
Proposition 5. The set of equilibrium spreads is given by:
ni =jp, nq = K,
nh = nl + k= 2 k h=2,..., q- 1,
1183

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
where
"!k CF(2pk- 61 )'
and q is the smallest integer such that.

q
JP + k >K. (9)
k=2
de)"
We refer to Pfh = nh- nh-1 as the spread improvement, when the
spread is nh. The spread improvement is the number of ticks by which a
limit order trader improves upon (undercuts) the best quotes. Thus, it is a
measure of the aggressiveness of the submitted limit order: the larger is
'h, the more aggressive is the limit order.
Proposition 5 presents the determinants of spread improvements.
Spread improvements are larger when (i) the proportion of patient tra-
ders, Op, is large, (ii) the waiting cost, Sp, is large, and (iii) the order arrival
rate, A, is small. In particular, when 25p (0 >AA, a patient trader

improves the spread by more than one tick, that is, 'h>l. Biais, Hillion,
and Spatt (1995), among others, find that many limit orders in the Paris
Bourse improve upon the prevailing bid-ask quotes by more than one
tick.
A common mechanism is driving these results: a rise in expected wait-
ing costs induces liquidity suppliers to bid more aggressively in order to
shorten their execution time. Consider an increase in the proportion of
patient traders. This increase immediately reduces the execution rate for
limit orders as market orders become less frequent. Accordingly, the
expected waiting time, T(.) and, thereby, the expected waiting cost for
liquidity suppliers increase. Patient traders react by submitting more
aggressive orders, thus tPh increases for all h > 1. A similar reasoning
applies when A falls or 6p rises.
2.2.3 Efficiency. What is the efficient outcome in our model? The price
concession paid by liquidity demanders is earned by liquidity suppliers,
net of their waiting costs. Hence, the bid-ask spread is just a transfer
payment, and the waiting costs constitute a dead-weight loss. It follows
that the total welfare-the sum of liquidity suppliers' and liquidity
demanders' expected payoffs-is equal to the expected waiting costs
borne by liquidity suppliers. We denote this expected cost by EC. An
efficient outcome is such that traders use order placement strategies which
result in the smallest possible value for EC.
Liquidity suppliers must wait at least one period before execution and
their waiting cost, per unit of time, is at least 6p; hence EC > 6. We A.
1184

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
deduce that a situation in which order placement strategies are such that
EC = 6 is efficient. This can be attained under the following two condi-
tions: (i) only the patient traders submit limit orders and (ii) all limit
orders generate the competitive spread. In this case, all limit order traders
(i) have the smallest possible waiting cost per unit of time (6p) and (ii)
always wait exactly one period before execution. Thus EC -= .
In general, the efficient outcome is not obtained in equilibrium. First,
consider the homogeneous case. All the liquidity suppliers post the com-
petitive spread, but some of them are impatient. As impatient traders bear
relatively large waiting costs per unit of time (6,>6p), the equilibrium
outcome is inefficient (i.e., EC> 6). In the heterogeneous case, only
patient traders provide liquidity; however, they strategically post spreads
larger than the competitive spread. As a consequence, liquidity suppliers'
average time to execution is strictly larger than one period and, for this
reason, EC> 6. In summary, our model uncovers two possible sources of
inefficiency in a limit order market: (i) impatient traders sometimes sub-
mit limit orders and (ii) patient traders post spreads larger than the
competitive spread. Both features induce excessive waiting costs com-
pared to the efficient outcome.
2.3 The determinants of market resiliency

Suppose that a liquidity shock (a succession of market orders in our
model) causes the spread to increase. In a resilient market, it takes a
small number of quote updates for the spread to revert to its former level.
Hence, market resiliency can be measured by the average number of
orders observed before the spread reverts to its competitive level. The
smaller is this number, the larger is market resiliency. Degryse et al.
(2003), for instance, use this method. It is difficult to derive analytically
a general expression for this order-based measure of market resiliency in
our model. Instead, we use another closely related measure. We measure
market resiliency as the probability that the spread reverts to its compe-
titive level before the next transaction occurs. The larger is this probabil-
ity, the larger is market resiliency. We have checked numerically that all
our implications regarding our measure of market resiliency (for instance
Corollary 1) hold for the order-based measure of resiliency as well.12
Suppose that initially the spread is at its competitive level, jp, and that a
succession of market orders enlarges the spread to its maximal level of K.
Let R be the probability that the spread reverts to the competitive level,
jp, before the next transaction. We will take R as being our measure of
market resiliency. When traders are homogeneous, a deviation from the
12 Our measure of market resiliency is based on the inside spread. In reality, the speed at which depth at
given quotes recovers after a liquidity shock is another dimension of market resiliency [see Coppejans,
Domowitz, and Madhavan (2003) and Degryse et al. (2003)]. We cannot study this dimension, since all
orders are of the same size in our model, and we do not allow orders to queue at the same price.
1185

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
competitive spread is immediately corrected and R = 1. When traders are

heterogeneous, a streak of q - 1 consecutive patient traders is required to
narrow the spread to the competitive level conditional on a current spread
of K ticks (see Proposition 3). Thus R = Oq-l<l. Notice that q is endo-
genous and is a function of all the exogenous parameters [see Equation (9)].
Consequently, market resiliency is determined jointly by the proportion of
patient traders, the order arrival rate, trader's waiting costs, and the tick size.
Corollary 1. When traders are heterogeneous, the resiliency of the limit

order book, R, increases in the proportion of patient traders, Op, and the
waiting cost, Sp, but decreases in the order arrival rate, A.
The factors which enlarge (lower) spread improvements have a positive

(negative) effect on the resiliency of the limit order book. As the proportion
of patient traders rises, or as waiting costs increase, limit order traders
improve upon the inside spread by larger amounts, thereby increasing the
market resiliency. With an increase in the order arrival rate, limit order
traders become less aggressive in their price improvements. More orders
are required to bring the spread to the competitive level, and thus resiliency
declines.13 We postpone the analysis of the effect of the tick size on market
resiliency to Section 4.
2.4 Numerical examples

We study three examples representing three cases of interest: (a) traders
are homogeneous, (b) traders are heterogeneous and patient traders
dominate the trading population (p = > 1), and (c) traders are hetero-
geneous and impatient traders dominate the trading population (p < 1).
Our goal here is to show that the dynamics of the limit order book are
strikingly different in these three cases.
In each example, the tick size is A - $1 and the arrival rate is A = 1. The
lower and upper price bounds of the limit order book are BA = $20
and AA = $21.25. The maximal spread is K = 20 (KA = $1.25). The
remaining parameters are presented in Table 1.
In Example 1, the waiting costs are such that traders are homogeneous
(jp =-j = 2). In Examples 2 and 3, the waiting costs are such that traders
are heterogeneous (j/ = land j] = 3). The composition of the trading
population is different in Examples 2 and 3: patient traders dominate in
Example 2 (p = 1.22), whereas impatient traders dominate in Example 3
(p = 0.81). In Example 1, the competitive spread is two ticks. In Exam-
ples 2 and 3, the competitive spread is one tick.
13 The average time elapsed until the spread reverts to its competitive level could also be used to measure
resiliency. However, the effect of increasing the order arrival rate on this time-based measure of market
resiliency is ambiguous. While it increases the number of quote updates before the spread reverts to its
former level, at the same time it reduces the average time between orders.
1186

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
Table 1
Parameter values for the three examples
Example 1 Example 2 Example 3
6p 0.075 0.050 0.050

6I 0.100 0.125 0.125
Op Any value 0.55 0.45
Table 2
Equilibrium order placement strategies
Example 1 Example 2 Example 3
Current spread Type P Type I Type P Type I Type P Type I
1 0 0 0 0 0 0
2 0 0 1 0 1 0
3 2 2 1 0 1 0
4 2 2 3 0 3 0
5 2 2 3 0 3 0
6 2 2 3 0 5 0
7 2 2 6 0 6 0
8 2 2 6 0 7 0
9 2 2 6 0 8 0
10 2 2 9 0 9 0
11 2 2 9 0 10 0
12 2 2 9 0 11 0
13 2 2 9 0 12 0
14 2 2 13 0 13 0
15 2 2 13 0 14 0
16 2 2 13 0 15 0
17 2 2 13 0 16 0
18 2 2 13 0 17 0
19 2 2 18 0 18 0
20 2 2 18 0 19 0
2.4.1 Order
Table 2
gies for pat
derive from
arrives. Each
the equilibri
the placemen
spread is eig
spread of six
In Example
mit a limit o
than their
Thus, the i
ticks and th
(heterogene
11

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
traders demand liquidity. In contrast to Example 1, limit order traders do

not necessarily post the competitive spread (ji = 1). In fact, in Example
2, the spreads on the equilibrium path are: {1, 3, 6, 9, 13, 18, 20}; while in
Example 3 these are: {1, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 16, 17, 18, 19,
20}. Traders place more aggressive limit orders when patient traders
dominate the trading population (Example 2). Actually, spread improve-
ments are larger than one tick for all spreads on the equilibrium path in
Example 2. In contrast, in Example 3, spread improvements are equal to
one tick in most cases.
2.4.2 Expected waiting time. Figure 1 presents the expected waiting

times of limit orders in Examples 2 and 3 as functions of the spread. In
each example, the expected waiting time increases with the spreads on the
equilibrium path and remains constant over spreads that are not reached
in equilibrium. The expected waiting times are uniformly smaller in
Example 3 than in Example 2. This explains the differences in the optimal
strategies. When the proportion of patient traders is small, as in Example 3,
patient traders are not aggressive, because they expect a fast execution.
30-
o example 2
25- example 3
0 20 0 0 0
E
I-
10-
_0 ~ ~
0
~ ~ ~ ~ ~ ~ , *
.
0
0 I
1 2 3 4 5 6 7 8 910111213141 1
0 I i I t i I i t I I I I ! ! I i t !
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
Submitted Spread (j )
Figure 1
Expected waiting time
The figure reports the expected waiting times of limit orders as a function of the spread they create in
Examples 2 and 3. Expected waiting times for Example 2 are depicted by White Diamonds and for
Example 3 by Black Squares.
1188

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
2.4.3 Book dynamics and resiliency. Figure 2 reports the evolution of

the limit order book for a sequence of 40 trader arrivals in Examples 2
and 3. In each period, it gives the state of the limit order book after the
order submission of the trader arriving in this period. Figure 3 depicts the
corresponding dynamics of the inside spread and its mid-point (black
dots). Initially the spread is equal to K = 20 ticks. This is the state of the
book after the arrival of several market orders. How fast does the spread
revert to the competitive level (one tick)?
Example 2 - A Resilient Book (p = 1.222)
Period 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
Trader EB SPBP SI Bl Si BPSP Bi SP BI SP BPSP Bl SPBPSPBP SI BPSPBPSP BI SI BP Si BPSPBP SI B! SP BI SI B8 SP BP SI
21 1/4 o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o
21 3116
21 1/8 o o o
21 1/16
21
20 15/16 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
20 7/8
20 13/16
20 3/4 0 0o 0 0o 0o 0o 0o 0o 0o 0 0o 0 0 0 0 0 0o0 0
20 11/16
20 5/8 o o o o
20 9/16 b bb b b bbb b b b b b
20 1/2
20 7/16
20 3/8 bbbbbbbbbbbbbbbbbbbbbbb b
20 5/16 b
20 1/4
20 3/16
20 1/8 b bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb
20 1/16
20 bbbbbbbbbbbbbbbbbbbb bb b b bbbb b b b bb b b b
Example 3 - A Weakly Resilient Book (p = 0.818)
Period 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
Trader 81 SPBP SI BI SI BPSP BI SP Bi SP BSP SP BP SPP SI PSP BP SP BP S BP SI BP SP BP SI 1B SP B SI BI SP BP SI
21 114 o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o
21 1/16 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
213/16 00 0O0O0OO0000000 0 00 0 00000o oo
21 1/16 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2015116 0o o o0 0
20 7/8
20 13/16
20 3/4
20 11/16
20 5/8
20 9/16
20 1/2
20 7/16
20 3/8 b
20 5/16 b b b b b b b b b b b b
20 1/4 b b b b b b b b b b b b b b b b b b b b b
20 3/16 bbbbbbbbbbbbbbbbbbbbbbbb
20 1/8 b bb b b b b b b b b b b b b b b b b b b b b b b b bb
201/16 bb bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb
20 bbbbbbbbbb bbbbbbbbbbbbbbbbbbbbbbbbbbbb b
Figure 2
Book simulation
The figure presents the evolution of the limit order book in Examples 2 and 3 for a given sequence of 40
trader arrivals. For each period, the figure indicates the type of the trader arriving in this period. "BP" and
"BI" indicate the arrival of a "Patient Buyer" and an "Impatient Buyer" respectively. "SP" and "SI" indicate
the arrival of a "Patient Seller" and an "Impatient Seller" respectively. For each period, the figure gives the
state of the book after the order submission by the trader arriving in this period. Letter "b" (respectively 'o')
at a given price indicates the presence of a buy (respectively sell) limit order at this price.
1189

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
Exanrpe 2 - A Reslient Book
21.500
21.375
-------------------------------------------------------------- ---------- --------------
21.250 'I ------------- ----------------------------------------------------------------- -------
21.125 ------------ ---------------- ------------------ ------------------------- -------
21.000
- ----------------------------------- ---------- ---------- -------
20.875 ---------- - ---------------------------------------------------- -------
20.750
20.a25 -T- 4- - I
20.500
- --- -------- ------------ --- -------- --- ---
20.375
- --- --- -------- ------------ --- -------- ---
20.250 ----- ------------------------------------------------------ -------
20.125
.......... -------- ----------------------------------------------------
20.000 .......... ---------------------------------------------------------------------------
19.875 ----------------------------------------------------------------------------------------
- - - - - - - - - - -
19.750 I t I I I I I I t I t I I I I I I I I I I I I t I
1 2 3 4 5 6 7 8 9 10111213141510171819202122232425252728293031323334353637383940
Example 3 - A WeeMy Resilient Book
21.500
21.375
------ --- ,----- - ------- l) II-
21 250
-------I -- - - - - - -- ---------- --^------------- - -- - -- - -- - -- - -- - -
21.125
21.000 ?I ---rr -- - ---------- ---r------- --r- *- - --------r - --- ------- --------r- - -?-
20.875 I
20.750 I
20.625 I
- ---------- - ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ -? --
20.500 . - . . . . . . . ,. -. - - . . ... - - . . - .. -.. - - .. -.. - . . - -.. - .. - -.. -.. - . . - -.. -.. - . . - -.. - .. - -.. - .. -.. - - . . -.. - .. - -.. - .. - - .. -.. - .. - -.- . 4 . . , . .
20.375 ~^-r--r- r-rrrrr--x --- -r ^---

,. --- -
........--- ------
..... ------
. . .. . . . ... . .
20.250 ?-
. .... . . .. . . . . . - . . .. - . . .. - . . . . - . . .. - . . .. - . . .. - . . - - - - - - - - - - - - - - - - - - .- ..............- ..........
20.125 I
20.000
19.875
19.750 I I I I I l t I I I I I l I I l t I I I I I l I I I I I I I It i t I I 1 i
2 3 4 5 6 7 89 011 12 13 14 15 1016 7 18 19 20 21 2223 24 25 20 27 28 29 30 31 32 33 34 35 38 37 38 39 40
Figure 3
Spread evolution
The figure presents the evolution of the size of the inside spread (vertical line) and the mid-quote (dots) in
Examples 2 and 3 for the sequence of trader arrivals depicted in Figure 2.
It is apparent from Figure 3 that the competitive spread is reached much

more quickly in Example 2 than in Example 3. In fact, in Example 3, the
quoted spread remains much larger than the competitive spread during all
40 periods depicted in Figures 2 and 3. In contrast, in Example 2, the inside
spread reaches the competitive level for the first time after 18 periods and
then it remains close to this level. As traders' types in each period are
identical in each example, this is due to the fact that in Example 2, patient
traders use more aggressive limit orders in order to speed up execution.14
This bidding behavior explains why the market appears much more resi-
lient in Example 2 than in Example 3. Accordingly, our measure indicates
that the resiliency of the market is much larger in Example 2,
R = 0.556 - 0.02, than in Example 3, where R = 0.4517 _ 1.27 x 10-6.
14 If realizations for traders' types were not held constant, an additional force would make small spreads
more frequent when p > 1. In this case, the liquidity offered by the book is consumed less rapidly, since the
likelihood of a market order is smaller than when p<l. Thus, the inside spread has more time to narrow
between market order arrivals.
1190

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
Summary. When traders are homogeneous, any deviation from the com-
petitive spread is immediately corrected. This is not the case in general
when traders are heterogeneous. In the latter case, the market is more
resilient when p > 1 than when p < 1. Thus, although the equilibrium of
the limit order market is unique, three patterns for the dynamics of the
spread emerge: (a) strongly resilient, when traders are homogeneous, (b)
resilient, when traders are heterogeneous and patient traders dominate
(p > 1), and (c) weakly resilient, when traders are heterogeneous and
impatient traders dominate (p < 1). Market resiliency clearly depends
on the composition of the traders' population.
2.5 Distribution of spreads

In this section, we derive the equilibrium probability distribution of the
bid-ask spread when traders are heterogeneous (when they are homoge-
neous, all transactions take place at the competitive spread). We show
how the distribution of spreads depends on the composition of the trading
population.
Proposition 3 shows that the equilibrium spread takes q different
values, n <n2<... <nq. A patient trader submits a nh_l-limit order when
the spread is nh (h = 2, ..., q) and a market order when the spread is n1. An
impatient trader always submits a market order. These strategies yield the
following dynamics for the bid-ask spread. If the spread is
nh (h = 2,...., q - 1), there is a probability Op that it becomes nh1_ and
a probability 01 that it becomes nh+l when a new trader arrives in the
market. If the spread is nl, the inside spread becomes n2 when a new
trader arrives in the market. If the spread is K, it remains unchanged with
probability 0z or it becomes nq_l with probability Op, when a new trader
arrives in the market. Hence, the random sequence of observed spreads is
a finite Markov chain with q > 2 states. The transition matrix of this
Markov chain is:
/0 1 0 0 O\
Op 0 0I ... 0 0
O Op 0 ... 0 0
W= : :
0 0 0 ... 0 01
\ 0 0 * p OI/
The jh entry in the hth row of this matrix is t

of the spread changes to nj conditional on the
We denote the stationary probabilities of this M
where Uh is the stationary probability of a spre
1191

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
*ExanVIe 2
0 Example 3
0.25
i
0.2
0.15
0.1
0.05 I II - 1
0 Tl ,l, ,n,1, n,n
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Figure 4 Spreads
Spread distribution
The figure presents the stationary probab
and 3 (white bars). For each inside sprea
occurrence of this inside spread.
Lemma 2. The spread has a uniq
oq-1 oq-hoh-2
u? '0 ,and Uh 1 P I
ul :Se
T i+ i-20q1
Ei=2P +P~q
+ o_oi, n-io
I p i=2 _ i-2
I
h 2,..., q. (10)
Observe that for h
uh h uh 1
- - ph'h and - (11)
Uhl ( 11I I61
This remark yields the following propositio
Proposition 6. For a given tick size and waiting
1. If p<l, uh>uh, for 1 < h'<h < q. Thus, the

skewed toward higher spreads in weakly re
2. If p>1, uh < uh, for 2 < h'<h < q.15 Thus,
is skewed toward lower spreads in resilient
Hence, small spreads are more frequent wh

This reflects the fact that markets dominated
resilient than markets dominated by impatie
15 The inequality, Uh<Uh', does not necessarily hold for h'= 1,

spread can only be reached from higher spreads, while other spr
(nq = K can be reached either from nq_1 or from nq itself). This
the smallest possible spread is relatively small for all values of p.
1192

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
the stationary distribution in Examples 2 and 3. Clearly, the distribution

of spreads is skewed toward high spreads in Example 3 where p < 1 and
toward low spreads in Example 2 where p > 1.
The expected dollar spread is given by:16
q
ESm =>l Uhnh. (12)

h=l
Using this expression, the expected spread is $0.525 in Example 2 and $1

in Example 3. The average spread is smaller in the more resilient market,
because small spreads are more frequent in this case.
3. Implications for Empirical Analysis
In this section, we discuss the implications of our model for empirical

studies of limit order markets. We restrict attention to the heterogeneous
case. We study the relationships between measures of trading activity (the
order arrival rate and the time between trades) and market resiliency.
Then we analyze the impact of the state of the book on order aggressive-
ness. Finally, we examine the implications of changes in the proportion of
patient traders on intraday patterns and discuss the liquidity effects of
changes in ownership.
3.1 Trading activity, spreads and market resiliency

3.1.1 Fast versus slow markets. We compare market resiliency and the
expected spread across two markets "Fast" (F) and "Slow" (S), which
differ only in the order-arrival rates, with AF > XS.17 Thus, the average
waiting time between orders in market F is less than in market S. We
denote by nh(Ak) the hth smallest spread in the set of spreads on the
equilibrium path in market k c {F, S} and by qk the number of spreads
in this set. Using Proposition 5 and Corollary 1, we obtain the following
result.
Corollary 2. In two markets F and S with order-arrival rates, AF > AS,
1. The equilibrium spreads satisfy.
(a) nh (AF) < nh (Xs), for h < qs and

(b) nh (AF) < nqs (XA), for qs < h < qF.
2. The Slow market is more resilient than the Fast market.
16 Recall that a superscript "m" indicates variables expressed in monetary terms, rather than in number of
ticks (i.e., n' = nhA).
17 In our model, equilibrium spreads are determined by the ratio M (see Proposition 5). For this reason, the
results for an increase in the arrival rate translate immediately to results for a decline in the waiting costs
6p.
1193

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
Table 3
Expected dollar spreads and order arrival rates
Op
A 0.45 0.475 0.5 0.525 0.55
1 1 0.9 0.71 0.57 0.525

4/5 1.01 0.91 0.73 0.67 0.6
2/3 1.01 0.86 0.73 0.65 0.59
1/2 1.02 0.85 0.78 0.68 0.67
1/3 0.97 0.84 0.84 0.78 0.78
1/5 0.99 0.91 0.91 0.89 0.90
The first part of

spreads in the Fast m
of possible spreads
pare two limit order
Slow market, with
expected waiting ti
limit order trader
taking a given posi
spreads tend to be n
tion). On the other h
narrows more quick
discussion followin
resilient.
These two effects
cannot determine a
lest spread. Compu
reduces the expect
which suggests tha
illustration of this c
spread for various p
in Example 3 (i.e.,
correlation betwee
negative and equal
suggests that the
increases.
In summary, the model generates two predictions. First, the average
spread tends to be inversely related to the order arrival rate. More
surprising, maybe, the model implies a negative relationship between
market resiliency and the order arrival rate.
18 The condition Sc = K holds for all parameter values considered in this table. Hence, we use Proposition 5,
Lemma 2, and Equation (12) to compute the equilibrium expected spreads.
1194

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
We interpret A as the long-run arrival rate of orders, including both

market and limit orders. It is a measure of trading activity in a given
stock. Demsetz (1968) suggests that the primary determinant of the long-
run order arrival rate is the number of shareholders. We conjecture that
another related factor is the public float of the stock. As these variables
are not affected by variations in the state of the book, it is natural to
consider A as independent from the state of the limit order book as well,
just as it is exogenous in our model.19
We recognize that, in reality, the intraday arrival rate deviates from the
long-run arrival rate, and that these deviations may in part be explained
by transient changes in the state of the book. We are not aware of
empirical evidence from limit order markets pointing out in this direction.
At any rate, if such a relation exists, it is not captured by our model.
3.1.2 Durations between trades and market resiliency. The order arrival
rate is one measure of market activity. Market activity can also be
measured by the average time between trades. There has been a consider-
able interest recently in modeling the time between trades. For instance,
the Autoregressive Conditional Duration approach, pioneered by Engle
and Russell (1998), postulates the expected duration between trades as a
function of pre-determined variables including past realizations of the
duration between trades. There is very little theory that endogenizes the
time between trades in limit order markets.20 This scarcity makes it
difficult to specify and interpret conditional duration models.
In our model, the time between trades is endogenous, and it depends on
the size of the spread. Let Dh denote the expected time elapsing between
two consecutive transactions, conditional on the first transaction taking
place when the spread is nh. We call this variable a conditional duration.
Corollary 3. In equilibrium, the conditional duration is.
1 - 1h+I I -_ 4
Dh = L P for 1 < h < q; and D = " P. (13)
Hence, the conditional duration (i) increases with the size of the inside
spread, (ii) decreases with the order arrival rate, and (iii) increases in the
proportion of patient traders.
The conditional duration decreases as the order arrival rate rises, other
things being equal, and is positively related to the spread: the larger is the
19 The rates at which market and limit orders arrive are endogenous since traders optimally choose between
these two types of orders. As expected, they depend on the spread (see Propositions 1 and 2). This is the
overall order arrival rate which is exogenous in our model.
20 Easley and O'Hara (1992) endogenize the time between transactions in a dealer market. In their model,
the timing of trades is driven by the existence of new information.
1195

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
The Review o' Financial Studies / v 18 n 4 2005
spread at which a trade occurs, the larger is the average time until the next
transaction. Notice that there is an interesting contrast with the findings
obtained in Easley and O'Hara (1992). In their model, the spread depends
on the time elapsed since the last transaction while in our model, the
expected time until the next transaction depends on the size of the inside
spread. This suggests that the spread should be used as an explanatory
variable in empirical models of conditional duration. Finally, the condi-
tional duration increases with the proportion of patient traders, Op. As Op
rises, the probability that a trader submits a market order declines which
delays the next transaction.
In our model, conditional duration and market resiliency are governed
by the same factors. These factors include the order arrival rate or the
proportion of patient traders. Consider an increase in the proportion of
patient traders. Other things being equal, it leads to (i) a larger condi-
tional duration (Corollary 3) and simultaneously to (ii) greater market
resiliency (Corollary 1). Similarly an increase in the order arrival rate
results in (i) a smaller conditional duration (Corollary 3) and in (ii) lower
market resiliency (Corollary 1). Consequently, our model predicts a posi-
tive association between the average time between trades, conditional on
the size of the spread, and market resiliency.
3.2 State of the book and order aggressiveness

The amount by which limit order traders improves upon prevailing
quotes ("the spread improvement") is a measure of order aggressiveness.21
Proposition 5 has the following implication.
Corollary 4. Spread improvements depend on the size of the inside spread.

Spread improvements increase with the size of the spread when p > 1 and
decrease with the size of the spread when p < 1.
As an illustration, compare the bidding strategies in Examples 2 and 3 in

Table 2. When they face a spread of six ticks, limit order traders undercut
the posted spread by three ticks in Example 2 (p > 1) and only one tick in
Example 3 (p < 1). When they face a spread of three ticks, limit order
traders undercut by two ticks in Examples 2 and 3. Hence, their aggres-
siveness when they face a spread of three ticks is (a) smaller (than when
they face a spread of six ticks) in Example 2 but (b) larger in Example 3.
The explanation for this result is as follows. Consider a patient trader
who faces a spread nh. The spread improvement chosen by the trader
results from the following trade-off. The cost of a large spread improve-
ment is that it results in a worse execution price for the trader. The benefit
21 Several empirical papers have analyzed the relationship between the state of the book and the aggres-
siveness of incoming orders. See, for instance, Griffith et al. (2000), and Biais, Hillion, and Spatt (1995).
1196

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
is that it results in a smaller waiting cost. Hence, the trader's decision

hinges upon a comparison between the size of the spread improvement
and the resulting reduction in waiting cost. In equilibrium, the optimal
spread improvement is equal to the expected reduction in the waiting cost,
rounded up to the nearest integer. This follows from Equation (8):
"h = CF([T*(nh)- T*(nh^l)] ).
When p > 1, the difference T*(nh) - T*(nh_l) increases in nh (see Propo-

sition 4). It follows that liquidity suppliers are willing to offer larger
spread improvements when the spread is large. In contrast, when p < 1,
the difference T*(nh) - T*(nh-1) decreases in nh; thus, liquidity suppliers
are willing to make larger spread improvements at small spreads.
For stocks listed on the NYSE, Engle and Patton (2003) find that the
change in the log of the best ask (bid) price is negatively (positively) related
to the size of the spread. This means that the amount by which traders
improve upon prevailing quotes is related to the size of the spread, as
predicted by Corollary 4.22 This corollary also implies that the direction
of this relationship is affected by the proportion of patient traders. This is
an additional prediction, which could be tested in future empirical work.
3.3 Changes in the mix of patient versus impatient traders

In our model, market resiliency, spread improvements and the distribu-
tion of spreads are functions of the proportion of patient traders. Unfor-
tunately, the proportion of patient traders cannot be directly observed.
Hence, it is hard to directly test the model predictions on the effects of a
change in the proportion of patient traders.
One approach is to use a proxy for the proportion of patient traders. The
model suggests that the proportion of limit orders in the flow of market and
limit orders as a proxy. In equilibrium, patient traders tend to submit limit
orders whereas impatient traders tend to submit market orders. In fact,
conditional on a spread above the competitive spread, the probability of
observing a limit order is Op, the proportion of patient traders. Another
approach is to exploit predictable changes in the proportion of patient
traders. We describe this approach in the following subsections.
3.3.1 Intraday variations in the proportion of patient traders and intraday

patterns. Traders' impatience is likely to increase toward the end of the
trading day. The inability to trade overnight is a binding constraint for
many investors which makes them eager to trade as the end of the day
22 A caveat is in order here since the NYSE is a hybrid market. There is a possibility that Engle and Patton's
findings are driven by the actions of NYSE specialists rather than those of limit order traders. We are not
aware of evidence on this issue from pure order-driven markets.
1197

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
approaches.23 Furthermore, many institutions mark their positions to

market at the end of the day; thus, they prefer to trade closer to that
deadline. Also, some traders act as implicit market markers in limit order
markets and may be keen to unload their inventory before the end of the
day to avoid an exposure to the overnight risk.
For these reasons, we conjecture that the proportion of impatient traders
is larger at the end of the trading day than in earlier periods. Under this
conjecture, comparing measures of market liquidity in the last period of
the day (say the last half-hour) with these measures in an earlier period is
like analyzing the impact of an increase in the proportion of impatient
traders in our model (holding other parameters fixed). Several testable
implications follow from this remark:
* Limit order traders submit less aggressive orders at the end of the
day than in an earlier period. This prediction derives from Proposi-
tion 5 which establishes that spread improvements are small when
the proportion of impatient traders is large.
* As a result, market resiliency is smaller at the end of the day than in
an earlier period (see Corollary 1 and the discussion following the
corollary).
* The spread is larger at the end of the day than in an earlier period
(see Proposition 6 and the discussion in Section 2.4).
* The conditional durations between trades are smaller at the end of
the day than during the day (see Corollary 3).
Testing these predictions empirically is not entirely straightforward.

The information asymmetry is high at the beginning of the trading day
and thus is likely to influence measures of market liquidity at this time.
For instance, Madhavan, Richardson, and Roomans (1997) show empiri-
cally that the adverse selection component of the spread is large at the
beginning of the day and declines thereafter. Hence, in order to avoid
confounding effects due to asymmetric information, our predictions
should be tested by comparing measures of market liquidity (limit order
aggressiveness, resiliency, spreads, and conditional durations) in the last
period of the trading day and in an earlier period, which is not too close to
the opening of the trading session.
Some of our implications are consistent with stylized facts. It is well
known that spreads and trading frequency decline from the opening of
the trading day on but peak again at the end of trading sessions. For
instance, Biais, Hillion, and Spatt (1995) and Chung et al. (1999) observe
this pattern for the Paris Bourse and the NYSE, respectively. Although
the quotes in the NYSE are in part determined by the specialist, Chung
23 Recent experimental findings by Bloomfield, O'Hara, and Saar (2002) show that when liquidity traders
are assigned a trading target, they switch from limit to market orders at the end of trading sessions.
1198

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
et al. (1999) find that the increase in the spread is driven by the order
placement decisions of limit order traders. The joint peak in spreads, and
trade rates at the end of the trading day has proved difficult to explain in
asymmetric information models. Actually, as pointed out by Foster and
Viswanathan (1993), these models predict an inverse relationship between
spreads and trading activity. Our model offers a complementary explana-
tion. At the end of the day, the proportion of impatient traders increases.
This translates into a decline in the conditional duration and thereby
more frequent trades. Patient traders exploit this impatience by bidding
less aggressively which results in larger spreads.
Pagano and Schwartz (2003) analyze the impact of the introduction of
a closing call auction in the Paris Bourse. This closing auction offers
another trading opportunity and should decrease traders' impatience
toward the end of the day. Our model predicts that this should lead to a
decline in the spread in the last half-hour of the trading day. This is
precisely what is observed by Pagano and Schwartz (2003).24 Notice
that traditional theory would rather predict the opposite, as the closing
auction is likely to draw liquidity away from the continuous market.
Finally, Tkach and Kandel (2004) show that the time to execution of
limit orders in Tel Aviv Stock Exchange declines toward the end of the
trading day, in line with our predictions.
3.3.2 Liquidity effects of changes in ownership structures. We expect

cross-sectional variations in the mix of patient and impatient traders to
be related to variations in institutional ownership. Stocks that are pre-
dominantly owned by index funds should feature a larger proportion of
impatient traders, since their managers must trade rapidly to minimize
their fund's tracking error. This line of reasoning suggests to test the
model by analyzing long-run liquidity effects on stocks that are added
to (removed from) a widely followed index. Beneish and Whaley (1996),
among others, document a substantial increase in the proportion of index
funds owning a stock when it becomes listed in such an index. Our model
predicts that the resulting increase in the proportion of impatient traders
should manifest itself in a decline in market resiliency (Corollary 1) and
smaller spread improvements (Proposition 5). Naturally, an inclusion in
the index may have other effects on the stocks as well.
4. Tick Size and Market Resiliency
The tick size, that is, the minimal price variation, has been reduced in
many markets in recent years. The rationale for this move was to reduce
the trading costs of investors. In this section, we examine the effect of a
24 "Kaniel and Liu (2000) can also explain this result."
1199

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
change in the tick size, and show that a reduction in the tick size impairs
market resiliency, and may have adverse effects on the spread.25
We assume that a change in the tick size does not affect the monetary
values of the boundaries: Am = A/ and Bm - BA, which implies that Km
is fixed as well.
To better convey the intuition, it is useful to consider the polar case in
which there is no minimum price variation, that is, A = 0. In this case,
prices and spreads can be expressed solely in monetary terms; in what
follows, we index all spreads by a superscript "m" to indicate that they are
expressed in dollars. When the tick size is zero, a trader's reservation
spread is exactly equal to his expected waiting cost until the arrival of the
next trader, i.e., j?m = (i c {P,I}). We denote by T'"') the expected
waiting time for a limit order trader who creates a spread of jm dollars.
Let
cfe Km X - 6p
p KmA- 6p (14)
Notice that 0 < pC < 1, since jl' < Km by assumptio
next proposition extends Propositions 4 and 5 to the
no mandatory minimum price variation, but p > p
Proposition 7. Suppose that A = 0. If p > pc and 6p >

is as follows:
1. The impatient traders never submit a limit order

2. There exist qo spreads
2 n' < n'
qo qo< ... n, with n
such that a patie
spread in (nh, n
than or equal to
3. The spreads a
h = 2,... qo - 1 a
as given in Secti
4. The expected waiting time function is as follows:
T* (/m) _ T* (nm) Vjm e (n n1,n].
25 See Seppi (1997), Harris (1998), Cordella and Foucault (1999), Goldstein and Kavajecz (2000),
Kadan (2005) for arguments in favor and against the reduction in the tick size in various mar
structures.
26 If p < p', then the spread improvements are so small that the competitive spread is never re
discuss this case later.
27 A closed-form expression for qo is given in the proof of this proposition.
1200

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
Proposition 7 shows that when p>pc, the equilibria with and without a
minimum price variation are qualitatively similar. When A = 0, the
smallest possible spread is patient traders' per period waiting cost, i.e.,
6. In contrast, when A > 0, it is equal to this cost rounded up to the nearest
tick. Thus, not surprisingly, the competitive spread is larger when a
minimum price variation is enforced. This rounding effect propagates to
all equilibrium spreads. To make this statement formal, let n (A) denote
the hth smallest spread in the set of spreads on the equilibrium path when
the tick size is A > 0, and let qA be the number of spreads in this set. The
following holds.
Corollary 5. "Rounding effect": Suppose p > pc. In equilibrium: (I) qA < qo,
(2)n'(O) < nh(A), for h < qa, and (3)nm(0) < n (A) for qA < h < qo.
This means that the support of possible spreads when the tick size is zero is
shifted to the left compared to the support ofpossible spreads when the tick size
is strictly positive.
Given this result, it is tempting to conclude that the average spread is

always minimized when there is no minimum price variation. We show
below that this reasoning does not constitute the whole picture because it
ignores the impact of the tick size on the dynamics of the spread between
transactions.
When p > pC and 6p > 0, in zero-tick equilibrium, traders improve the
spread by more than an infinitesimal amount, 4 (0) > 0.28 Intuitively,
patient traders improve the quote by a non-infinitesimal amount to speed
up execution. However, as p decreases, spread improvements become
smaller and smaller: traders bid less aggressively since market orders
arrive more frequently (see the discussion following Proposition 5).
When A > 0, spread improvements can never be smaller than the tick
size; thus, for small values of p traders improve prices by more than they
would in the absence of a minimum price variation. We refer to this effect as
being the "spread improvement effect." The spread improvement
effect works to increase the speed at which spread narrows in between
transactions. For this reason, imposing a minimum price variation helps to
make the market more resilient. This intuition can be made more rigorous
by using the measure of market resiliency, R, defined in Section 2.2.2.
Corollary 6. Other things being equal, the resiliency of the limit order
market, R, is always larger when there is a minimum price variation than
in the absence of a minimum price variation. Furthermore, the resiliency of
28 Traders must improve upon prevailing quotes (Assumption A.2). However when the tick size is zero, they
can improve by an arbitrarily small amount. Proposition 7 shows that they do not take advantage of this
possibility when p > p'.
1201

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
the market approaches zero as p approaches pC from above in the absence of

a minimum price variation, whereas it is always strictly greater than zero
when a minimum price variation is imposed.
Intuitively, as p approaches pc from above, spread improvements

become infinitesimal when the spread is large (e.g., equal to K). Thus,
the limit orders are submitted arbitrarily close to the largest possible ask
price, A, or the smallest possible bid price, B. This explains why, in the
absence of a minimum price variation, the resiliency of the market
vanishes when p goes to pc. Imposing a minimum price variation is a
way to avoid this pathological situation, because it forces traders to
improve by non-infinitesimal amounts to get price priority.
Thus, intuitively, imposing a minimum price variation can reduce the
expected spread, despite the rounding effect, because it makes the market
more resilient. We demonstrate this claim by providing a numerical
example. The parameter values are as in Example 3 except that Km=
0.4375 (K = 7) so that the condition p > pc is satisfied (pc = 0.79). Table 4
gives all the monetary spreads on the equilibrium path for two different
values of the tick size: (1) A = 0 and (2) A = 1. The two last lines of the
table give the expected spread and the resiliency obtained for each regime.
First, observe the "rounding effect"-the five smallest spreads are lower
when A = 0, than in the case of A = 1i. Second, observe the "spread
improvement effect"-the spread reduction is quicker for every spread
level if a minimum price variation is enforced. This explains why market
resiliency is smaller when there is no minimum price variation. For this
reason, the expected spread turns out to be larger in this case ($0.3675
instead of $0.3265).
So far we have compared a situation with and without a mandatory
minimum price variation. More generally, the "spread improvement"
Table 4
Rounding and spread improvement effects
h n,"(A = 0) n` (A= )
1 $0.05 $0.0625
2 $0.1318 $0.1875
3 $0.1988 $0.3125
4 $0.2535 $0.3750
5 $0.2983 $0.4375
6 $0.3350 NA
7 $0.3650 NA
8 $0.3896 NA
9 $0.4096 NA
10 $0.4260 NA
11 $0.4375 NA
Expected spread $0.3676 $0.3265
resiliency 3.4 x 104 0.041
1202

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
Table 5
The tick size minimizing the expected spread
p 0.85 0.95 1 1.05 1.15 1.25
A*
16 16
I20
1
100
I
100
1i
100
effect implies
the tick size
demonstrates
expected spre
meters are ch
the smallest
spread for lo
make large im
becomes less e
aggressive lim
ond order com
which minimi
Finally, we br
upon large sp
arbitrarily c
possible bid p
mum price v
improvement
Notice that t
very similar t
(Section 3.1).
frictional cos
between arriv
out that fac
improvement
spreads.
To sum up, reducing or even eliminating the tick size may or may not
reduce the average spread. The impact depends on the proportion of
patient traders in the market. Many empirical papers have found a decline
in the average quoted spreads following a reduction in tick size. These
papers, however, do not control for the ratio of patient to impatient
traders. In Section 3, we have argued that the proportion of patient
traders is likely to decrease over the trading day. In this case, the impact
29 This would also be the case if patient traders' waiting cost were equal to zero (6p = 0). When p < pc or
6p = 0, the equilibrium (when there is no minimum price variation) is difficult to describe formally since
traders improve upon prevailing quotes by an infinitesimal, but strictly positive, amount.
1203

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
The Reviesw of Financial Studies / v 18 n 4 2005
of a decrease in the tick size on the quoted spread may not be uniform
throughout the trading day. Specifically, a decrease in the tick size may
increase the average spread at the end of the trading day. To the best of
our knowledge, there are no papers testing this hypothesis.
5. Robustness
Recall our assumptions regarding the trading process: A.1-no order

cancelations and resubmissions; A.2-limit orders cannot queue at or
behind the best quotes; and A.3-buyers and sellers alternate. In this
section, we present conditions under which our non-queueing restriction,
A.2, is not binding. We also show, using examples, that the main proper-
ties of the model persist when we relax the assumption that buyers and
sellers alternate. Overall, these robustness tests show that our main results
are not driven by the technical assumptions, and that the economic
intuitions are still valid when these assumptions are relaxed. For brevity,
we omit the proofs of the results given in this section. They can be
obtained from the authors upon request.
5.1 Cancelations and resubmissions

Although in practice traders frequently cancel and resubmit their limit
orders, our model does not allow them to do so. Hence, it cannot explain
why traders actively manage their orders. Clearly, modeling cancelations
and resubmissions is important. But it is also very difficult. One possible
approach, followed by Hollifield et al. (2003) or Goettler, Parlour, and
Rajan (2003) assumes that cancelations occur exogenously at random
points in time. In our model, all orders must eventually be executed.
Thus, in order to follow this approach, we would need to arbitrarily
specify the payoff to a trader when his order is canceled. For this reason,
we do not engage in this exercise.30
Most cancelations appear as the result of a particular behavior that we
do not seek to capture in this study. Hasbrouck and Saar (2002) find that
the majority of cancelations on Island ECN occurs very quickly after
order submission (about 60% are canceled within 30 seconds). They argue
that these "fleeting" orders seek liquidity rather than provide it. Tkach
(2002) studies the limit order submission in the 100 most liquid stocks on
the Tel Aviv Stock Exchange. She shows that the median time to cancela-
tion is 11 minutes, and over 12% of all cancelations occur within a minute
of submission. This is a short period of time, especially given the low
volume in many of these stocks. We do not expect that these cancelation
30 Rosu (2004) considers a dynamic model of price formation in limit order markets. As in our model,
traders value speed of execution. Assuming that limit orders resubmission is costless and instantaneous,
he allows traders to cancel and resubmit their orders. Some of his results are qualitatively similar to ours.
In particular, he finds that resiliency increases with the proportion of patient traders.
1204

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
strategies affect our conclusions. They are phenomena outside the scope
of this model.
5.2 Queuing at the inside quotes

We assume that traders cannot place limit orders at or behind the inside
quotes. In fact, such orders are allowed and used.31 Queuing, however,
should not invalidate the findings that (i) an increase in the proportion of
patient traders or (ii) a decrease in the order arrival rate yield more
resilient and more competitive limit order markets. Actually, limit order
traders' incentive to jump ahead of the queue is greater when time to
execution increases, that is, when the proportion of patient traders rises or
the order arrival rate declines. This means that spreads narrow more
rapidly in these two cases, even when queuing is an option.
This reasoning suggests that if the proportion of patient traders
is sufficiently large or the order arrival rate is sufficiently small
then traders will choose not to queue. In this case, the equilibrium is
exactly as described in Section 2. This result is established in the next
proposition.
Proposition 8. Suppose traders are heterogeneous and are allowed to queue

at the inside quotes subject to time priority (i.e., limit orders enteredfirst at
a given price are executedfirst). If
< 2[1 + 0p(2 - Op)], (15)

Op
then the equilibrium when traders are not allowed to queue is an equilibrium
in this setting.
Suppose that traders use the order placement strategies described in

Section 2 and give them the freedom to queue at the best quotes. Under
condition (15), traders prefer to submit limit orders improving upon the
inside quotes rather than queuing. Hence, traders' strategies form an
equilibrium even though traders have the possibility to queue. Condition
(15) is satisfied in all the numerical examples in the article. It follows that
the possibility of queuing does not per se invalidate our comparative
statics.
Inequality (15) is satisfied when the order arrival rate is sufficiently
small. Furthermore, a rise in the proportion of patient traders increases
the right-hand side of Inequality (15) and thereby helps to satisfy this
31 For instance, Biais, Hillion, and Spatt (1995) report that about 50% of new buy limit orders are placed at
the best bid price. They find a similar frequency for sell limit orders.
1205

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
inequality. This means that traders are less likely to queue when the order
arrival rate is small and the proportion of patient traders is large, as
conjectured. Finally, Inequality (15) is satisfied when the tick size, A, is
sufficiently small. In this case, queuing is never optimal because liquidity
providers can jump ahead of the queue at a low cost. This reasoning
suggests that the number of limit orders placed at the same price should
have decreased following tick size reductions.
5.3 Buyers and sellers arrive randomly

We have assumed that buyers and sellers alternate. Clearly, it would be
more realistic to assume that the sequence of arrivals for buyers and
sellers is random. Unfortunately, in this case, the model becomes intract-
able for two reasons. First, it is not possible to solve for the equilibrium
by induction because the waiting time function is not recursive. Second,
the waiting time for a limit order depends on the state of the limit order
book when the order is placed and not simply on the inside spread.
Under these conditions, it is very difficult to compute the equilibrium
analytically. Such a computation involves the following steps. First, con-
jecture equilibrium order placement strategies for patient and impatient
traders. Second, use the conjectured equilibrium strategies to calculate the
expected waiting time for each possible limit order in each possible state
of the book. This task requires solving a number of simultaneous linear
equations which grows quickly with K because the waiting time function
depends on the entire state of the book. Third, check that the "conjec-
tured" strategies are optimal given the expected waiting times computed
in the second step. If these strategies are not optimal, the steps are
repeated until an equilibrium is found.
This procedure is tedious even for small values of K. It can be imple-
mented for specific values of the parameters, however. Thus, we use
examples to demonstrate that the economic intuitions of our model persist
when buyers and sellers arrive randomly. We assume that each trader is a
buyer or a seller with equal probabilities. We focus on the case K = 4. This
choice allows for different levels of spread improvements. For example,
when the spread has four ticks, a limit order trader can improve upon
prevailing quotes by one tick (small improvement), two or three ticks (large
improvements). We consider 3 different sets of values for the parameters.
In Example 4, the parameters are such that traders are homogeneous while
in Examples 5 and 6, the parameters are such that traders are heteroge-
neous. The proportion of patient traders is larger in Example 5 than in
Example 6. We describe below the equilibrium obtained in each example.32
32 A detailed derivation of the claims in these examples can be obtained from the authors upon request.
1206

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
5.3.1 Example 4-A Strongly Resilient Book (homogenous traders). Set

K = 4, A = 1, 8p = 61 = 0.025 (traders are homogeneous). The follow-
ing order placement strategy constitutes an equilibrium: (i) when the
spread is larger than one tick, buyers and sellers of both types submit a
1-limit order and (ii) when the spread is equal to one tick, both submit a
market order. Following a transaction, the spread increases to four ticks,
but then reverts to the traders' reservation spread of one tick before the
next transaction. This market is therefore strongly resilient: R = 1. As
traders are homogeneous, the equilibrium is not affected by Op, the
proportion of patient traders.
5.3.2 Example 5-A Resilient Book (heterogenous traders, large 0p). Set:
A = 16, K = 4, Op = 0.7, A = 1, 6p = 0.01, and 56 - 0.07. In this case
traders are heterogeneous. The following order placement strategies consti-
tute an equilibrium. An impatient trader always submits a market order. A
patient trader submits (i) a 2-limit order when the spread is equal to three or
four ticks, (ii) a 1-limit order when the spread is equal to two ticks, and (iii) a
market order when the spread is equal to one tick. The resiliency of the
market is R = 0.49.
5.3.3 Example 6-A Weakly Resilient Book (heterogenous traders, small

Op). Set: A-=1, K=4, Op=0.3, A= , 6p=0.01, and 6 = 0.07.
The following order placement strategies constitute an equilibrium. An
impatient trader always submits a market order. When the spread is larger
than one tick, a patient trader places a limit order improving the spread by
one tick. When the spread is equal to one tick, a patient trader places a
market order. The resiliency of the market is R = 0.027.
Clearly, the equilibrium obtained in each example has the same proper-
ties as the equilibrium obtained when buyers and sellers alternate. In
Example 4, the spread oscillates between a large level and a small level.
This is expected since this pattern epitomizes the homogeneous case (see
Proposition 1). Furthermore, as in the baseline model, limit order traders
use a more aggressive bidding strategy when the proportion of patient
traders is large (i.e., in Example 5). To see this, consider the case in which
the spread is equal to four ticks and a patient trader arrives in the market.
In Example 5, the trader improves upon prevailing quotes by two ticks
whereas in Example 6 he improves by only one tick. The economic
intuition is exactly the same as when buyers and sellers alternate. Limit
order traders bid more aggressively when Op is large because their waiting
times are larger, other things being equal. It follows that the resiliency of
the market increases in the proportion of patient traders. Accordingly, we
find that the stationary distribution of spreads is skewed toward small
spreads in Example 5 and large spreads in Example 6. For instance, the
1207

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
probability that the inside spread is equal to four ticks is 4% in Example 5

and 38% in Example 6. Again, this is expected since the proportion of
patient traders is larger in Example 5.
These findings suggest that relaxing the alternating arrival assumption
does not change the conclusions obtained when buyers and sellers alter-
nate. The driving force of our model is that limit order traders react to
exogenous increases in their total waiting costs by submitting more
aggressive orders. This basic economic intuition does not hinge on the
assumption that buyers and sellers alternate. However, relaxing this
assumption prevents us from solving the model in general. We view our
model as a way to bypass this problem without losing much of the
economic intuitions.
6. Conclusion
We construct a model of price formation in a limit order market. Agents

in our model are strategic, trade for liquidity reasons, and differ in their
impatience. Upon arrival, they must decide on whether to submit a
market order or a limit order. Their choice is driven by a trade-off
between the cost of immediacy (the spread) and the cost of delayed
execution, as first suggested by Demsetz (1968).
We derive the equilibrium order placement strategies. We find that the
proportion of patient traders in the population and the order arrival rate
are the key determinants of the limit order book dynamics. Traders
submit aggressive limit orders (improve upon quoted spreads by large
amounts) when the proportion of patient traders is large or when the
order arrival rate is low. For this reason, markets with a high proportion
of patient traders or a small order arrival rate are more resilient. Also, a
reduction in the tick size reduces market resiliency, and in some cases
increases the average spread.
The analysis yields several testable predictions: (i) a positive relation-
ship between inter-trade durations (conditional on the spread) and mar-
ket resiliency; (ii) a negative relationship between the order arrival rate
and market resiliency; (iii) a joint decline of limit order aggressiveness and
market resiliency at the end of trading sessions; and (iv) limit order
traders submit more (respectively less) aggressive orders when the spread
is large if patient (respectively impatient) traders dominate the trading
population.
Future research will focus on relaxing some assumptions that limit the
scope of our results. We assume that the proportion of patient traders in a
given market is exogenous. It would be interesting to endogenize the
composition of the trading population to gain insights on the sources of
cross-sectional variations in this composition. The order arrival rate is
independent of the state of the limit order book in our model. In practice,
1208

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
traders time their arrivals during the day according to market conditions,
and variations in the size of the inside spread may then trigger changes in
the intraday order arrival rate. This relationship is not captured by our
model. Finally, we have observed that the equilibrium outcome is in
general inefficient in our model. This result raises the possibility that
introducing designated intermediaries in order driven markets could be
efficiency enhancing, pointing to another interesting direction for future
work.
Appendix
Proof of Lemma 1.
Step 1. Suppose a trader (say a buyer) submits a j-limit order when the spread is s. By A.3
the following trader is a seller. We claim that at the time the j-limit order is cleared, the
spread will revert to s. We prove this claim by induction onj. If j = 1 then by A.2 the next
order is a sell market order, and the spread immediately reverts to s. Suppose now that j > 1,
and assume that our assertion is true for all k = 1,..., j - 1. By A.2, the seller must either
submit a market order or submit a k-limit order with k = 1,..., j - 1. If the seller submits a
market order, then the spread reverts s. If, on the other hand, the seller submits a k-limit
order with k c { 1,..., j- 1, then by the induction hypothesis, when that seller's k-limit
order is cleared the spread reverts to j. It follows that when the j-limit order is cleared, the
spread reverts to s as required.
Step 2. Consider a trader, say a buyer, who submits a j-limit order. The expected waiting
time of this order from this moment on is T(j). By A.2, this buyer acquires price priority (he
posts the best bid price). Suppose that the next trader (a seller by A.3) submits a k-limit order
with k { 1,..., j - 1}. When this k-limit order will be executed, the spread will revert to j
(Step 1). As traders do not cancel their orders or do not submit orders behind the best
quotes, the state of the book will then be exactly as when the buyer initially posted the j-limit
order. In particular, the buyer will have price priority. Thus, when the spread reverts to j, the
buyer's expected waiting time from that moment on is T(j) as well.
Step 3. We have explained in the text why T(1) = r. Now, consider a trader (say a buyer)
who submits a j-limit order with j > 1. The next trader (a seller) must choose among j
options. With probability ao0(), he submits a market order that clears the buyer's limit order.
In this case, the expected waiting time of the buyer is 1. With probability ak(/), the seller
submits a k-limit order (k = 1,..., j- 1). In the latter case, the original buyer's expected
waiting time is + T(k) + TU). Indeed, he has to wait (1) 1/A-for the seller to arrive, (2)
T(k)-until the seller's order is cleared and the spread reverts toj (by Step 1), and (3) another
T(j) as we are back to the original position (by Step 2). Overall the original buyer's expected
waiting time, T(j), is given by:
T(') =A c-
k=l
+ E k(') [+ T(k) + T(l) (16)
If ao(') > 0, we obtain the second part of the lemma by solving for T(j) and using t
that Yk=o cak() = 1. As for the third part of the lemma: If ao0() = 0, then the sell
submits a market order when the spread is j. Thus, the waiting time of the buyer who
the j-limit order is infinite: T(O) = +oo. U
Proof of Proposition 1. It follows immediately from the arguments preceding the prop
1209

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
Proof of Proposition 2. We first prove the following lemma.
Lemma 3. Suppose that facing a spread of size s (s E {1,... ,K - 1}), trader i (i E {P,I})
submits aj-limit order with 0 < j < s. Then, facing a spread of size s + 1, he either submits an
s-limit order or submits a j-limit order.
Proof. By assumption, trader i submits aj-limit order when he faces a spread of size s. Thus:
ri() > 7ri(k) k0, . ..j- 1,j+l . . ., s - 1.
Now, suppose that trader i faces a spread of size s + 1. If 7sr(s) < 7ri(/), then trader i w
submit a j-limit order since 7ri,() > 7ri(k) for all k = ,..., s. If 7ri(s) > 7ri(/), then trader
submits a s-limit order since Iri(s) > 7ri () > Tri(k) for all k = ,..., s - 1.
By definition of the reservation spread, and since 6p < 6b, it follows that:
7rl(/) < p(j) < O, Vj <j.
Thus, all traders submit a market order when they face a spread which is smaller than
equal to patient traders' reservation spread. This implies that T* () = T* (2) = ... T* ('P) =
Now suppose a patient trader faces a spread of sizejp + 1. Lemma 3 implies that he will eit
submit a jp-limit order or submit a market order. He obtains a larger payoff with a jp-li
order since
7rp(p) =ijpA - T (i)6 =j;A - > o,
where the last inequality follows from the definition ofjp. Then, we deduce from Lemm
that the patient type submits limit orders for all spreads s E (jp + 1, K). As for the impat
type there are two cases:
Case 1: The impatient type submits a market order for each s E (jp + 1, K) in which case
set sc = K.
Case 2: There are spreads in (1, K) for which the impatient type submits limit orders. In
case, let Sc be the smallest spread that an impatient trader creates with a limit order. B
definition of sc, the impatient trader submits a market order when he faces a spread s E (1,
and a sc-limit order when he faces a spread of size Sc + 1. Then, we deduce from Lemma 3 th
impatient traders submit a limit order when they face a spread in (sc + 1, K) and a mar
order otherwise. Finally, it cannot be optimal for an impatient trader to submit a limit orde
which creates a spread smaller than his reservation spread. This implies Sc > j*.
Proof of Proposition 3. Since we assume that Sc = K, the impatient type always subm
market orders. From Proposition 2, a patient trader submits a market order when he fac
spread in (l,jp) and a j;-limit order when he faces a spread of size jp + 1. Repeat
application of Lemma 3 (see the proof of Proposition 2) shows the existence of spre
nl < n2 < ... < nq such that facing a spread in (nh + 1, nh + 1) the patient trader subm
an nh-limit order for h = 1,... ,q - 1. Clearly, nl =jjp and nq = K. U
Proof of Proposition 4. When they observe a spread of size n , all the traders submit a mark
order. Therefore T*(n) =-. Let h E {2,...,q}. Suppose that the posted spread is
s E (nh- + 1,nh). When he observes this spread, a patient trader submits an nh_l-li
1210

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
order, and an impatient trader submits a market order (Proposition 3). Therefore when the
posted spread is s E (nh-1 + 1, nh), we have ao(s) = 1- Op, a,,, (s) = 0p, and
ck(s) = 0,V k ? {O, nh-_}. Thus, Lemma 1 (second part) yields
Tr(s) = - + OpT*(nh-) , Vs E (n/h- + l, n). (17)

1 - Op A
Hence, T*(.) is constant for all s E (nh-1 + 1, nh). Using Equation (17), we obtai
T*(n+l) - T*(nh) -p(T*(nh) - T*(nh-)) for h > 2, (18)
where p = . Furthermore, using Equation (17) and the fact that T*(nl) = , we obtain
T*(n2)- r*(nl) = > .
The claim follows now by repetitive application of Equation (18) and from the fact that
T*(nl) = . U
Proof of Proposition 5. Since nh = nh- + 'h, we immediately get that nh= nl + - =2 2k

Furthermore, since nq = K, it must be the case that q is the smallest integer such that
1n + Et=2 ;k > K. The expression for 'Jh is given by Equation (8). U
Proof of Corollary 1. Recall that q is the smallest integer such that nl + k=2 'k > K. It
follows that q (a) decreases with Op and bp and (b) increases with A, since 'k increases with
Op and 6p and decreases with A, for all k { 1,...,q - }. The result is then immediate. U
Proof of Lemma 2. We first show that the Markov chain given by Wis (a) irreducible and (b
a-periodic.
The Markov chain is irreducible. Observe that given any two statesjl ,j2 with 1 jl < j2 <
there is a positive probability that the chain will move from jl to j2 after a sufficiently larg
(though finite) number of transitions. This implies that any two states in the chain commu
nicate; hence, the chain is irreducible.
The Markov chain is a-periodic. Notice that Wq,q = O > 0. This means that when the chain is
in state q, there is a probability equal to Oj that it will stay in this state for the next
transitions, Vn > 1. Since state q communicates with all the other states of the chain, i
follows that no state has a period greater than 1. Thus the chain is a-periodic.
These properties imply that the Markov chain is ergodic. Hence, the induced Markov
chain has a unique stationary probability distribution of spreads (see Feller 1968). Let u
(ul,..., uq) denote the row vector of stationary probabilities. The stationary probability
distribution is obtained by solving q + 1 linear equations given by:
uW= uand u = 1, (19)
where e stands for the unit column vector. It is

ities given by Equation (10) are a solution of t
Proof of Proposition 6. The proof follows imm
1211

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
Proof of Corollary 2. In the proof of Corollary 1, we have established that q increases with A.
Thus qxs < q\x. Using Proposition 5, we obtain
nk+l(A) = nk(A) + CF(2P ),for A C {As,AF} and k < qxs - 2.

\C AA /
Thus, if nk(AF) < nk(Xs), then nk+l (AF) < nk+l (As) for k < qAs - 2. Now, observe that for
k = 1, we have (using Proposition 5):
n (A)= CF( .
We deduce that nl (AF) < nl (As) and conclude that nk((AF) < nk(AS) for k < qx - 1. Furth
more, nqAs (As) = nqAr (AF) = K. Consequently, nk(AF) < nqAs (AS) for qAs < k < qAF. T
proves the first part of the corollary. The second part follows from Corollary 1. U
Proof of Corollary 3. Let Nh denote the random variable describing the number of trad
arrivals between two consecutive transactions, conditional on the event that the first tr
action took place when the spread was nh. The conditional duration is:
E(N11)
Dh= h 1 l,..., q, (20)
since the expected waiting time between

Suppose that the last transaction took pla
this transaction, the new spread in equilibriu
then a new transaction takes place and N
submits a limit order which creates a spre
order since all traders submit market orde
deduce that the probability distribution for
Pr(Ni = 1) = 0i and Pr(N = 2
More generally, the same type of reasoning

1 < h < q. The largest possible value for N
Pr(Nh =j)= 01pl for j
and
Pr(Nh = h + 1) = h.
We deduce that
E(Nh) = 0, Zj161' + (h + l)Oe,

j=l
1212

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
which simplifies as
E(Nh) = 9 , for 1 < h < q. (21)
The expression for Dh follows from this equation and Equation (20). Finall
when the last transaction takes place at the largest possible spread, nq t
following this transaction remains nq. Hence, the situation is as if the last tr
place at spread, nq_1. It follows that the probability distributions of N
identical. Therefore E(Nq) = E(Nq_ ). The expression for Dq follows. The l
proposition follows directly from the expression for Dh, h < q. I
Proof of Corollary 4. The size of spread improvement (in number of ticks) w

spread is nh is given by h = CF(2 .-A Thus when p < 1, 9h decreases with
p > 1, Th increases with h. This means that when p < 1, spread improvemen
related to the inside spreads on the equilibrium path. In contrast, when
improvements are positively related to the inside spreads on the equilibrium
Proof of Proposition 7.
Step 1. We first derive the expected waiting time function associated with t
ment strategies described in Parts 1, 2, and 3 of the proposition. All traders s
order when they face a spread equal to n'. It follows that T* (n-) = I. Now su
posted spread is sm E (n^l, np ] with h > 2. When he observes this spread, a
submits an n _-limit order and an impatient trader submits a market or
ao(s") = 0O and ac_, (s) = Op. It follows that
T*(s) = +Op + T (n ) + T (sm) Vsm E (n I, nm] for h

A \A h- h- h
which yields
Tr(sm) = [ Op T (^ V], E (nm , n p ] for h > 2.
Hence T*(.) is constant for all sm c (n^ 1,n ] with h > 2. Then, following the last part of
the proof of Proposition 4, it is straightforward to show that the expected waiting time
function is:
1 h-
T*(nm) = 1+2 pk V h = 2,...,qo-1.
- kl
k=l A
This proves the last part of the proposition. Observe that the last equation implies:
(T*(nm) - Tr(nhm )) P = 2ph- = m - nm (23)
where the second equality follows from Part 3 of the proposition.

Step 2. Now we show that the order placement strategies described in Parts 1, 2, and 3 of the
proposition are optimal given the expression of the waiting time function given in Part 4.
Equation (23) implies that:
n nh
- T*(n)6ph
= n_-
- T*h
(n_) 6p =.. n= T* (n) p for h = 2,..., qo - 1. (24)
1213

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
As nl =P and T* (n) = , we have
n'n - T* (n')p = 0. (25)

Now Equations (24) and (25) imply that:
nh - T*(n)6p = 0 for h = l,...

Furthermore, we know (see Step 1) that T*(.) is constant for
T*(s-") =T* (n7) for sm E (n'_ , n<] and h = 2,
which implies that for h E {2, ..., o}:
s - T*(s"') <i < nh T*(nh )6i for s" E (n,_, n
Consider an impatient trader who faces a spread with size j

spread can be on the equilibrium path (" = n') or not (i' < n
implies
n' - T*(n')6 < O for k=l,...h - 1.
Using Equation (27), we deduce that
s" - T*(sm")6 < O, Vs'" E (0, j").
Thus any limit order yields a negative payoff to the impatient trader. It follows that he
submits a market order (which has a zero payoff).
Now consider a patient trader who faces a spread with size j' E (ni_, n"'], with h > 2.
From Equations (26) and (27), we know that:
m - T*(s")6p < 0, Vsm ? {0, nm, n ,... nh }. (28)
Furthermore we know from Equation (26) that:
n -T*(n)Sp =O, for kE{1,2,...,h-l}. (29)
We deduce from Equations (28) and (29) (i) that the patient trader's be
the set {n", n7 ..., nI_l } and (ii) that he is indifferent between any sprea
is a best response to choose nh_ for the patient trader when he fa
i' E (nml , n ]. Now consider a patient trader who faces a spread
follows from Equation (28) that the patient trader cannot profitab
spread. Therefore he chooses a market order.
Step 3. Finally, we compute the expression for qo. Since n' = n'_l + 'T
get that nh = n + Eh=2 i' (0). Furthermore since nq, = K", it must
the smallest integer such that n"T + k- 2 k (O) > Km". As k(O)
qo is the smallest integer such that:
6p , (2Pk- )6P > K

Now> K"'. (30)
k=2
Now the smallest integer qo which satisfies Condition (30) is given by:
1214

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
CF('" n(
qo =) ifp 1 and
n(p) p> p, (31)
(31)
CF(KmP)+ p) if
There is no finite solution if

straightforward to check th
Proof of Corollary 5. Using
2ph-16e
nk l(0) = nm'(0) + A
and
nkl(A) = nm (A) + CF (2A )A,
for 1 < k < min{qo - 2, q - 2}. Thus if nm(0) < n5m(A), then nk+l(O) < nm+l (A) for
1 < k < Min{qo - 2, qa - 2}. Now observe that for k = 1, we have (using Propositions 5
and 7):
n(0)= A and n (A)= CF E^ A.
Hence n'(O) < nm(A) since P6 < CF(P). We deduce that n'(0) < n'(A) for
k < Min{qo - 1, q - 1}. Recall that qo and qa are the smallest integers such that:
nm_ (0)+2pq0-1 > qo-

AKm and nqm (A)
\
+ CF 2pq-1
AA
p A >Km
Since nm (0) < nnl -(A

the corollary. The las
Proof of Corollary 6.
write this measure in

given tick size, A and
qA < q0. We deduce th
(31) (see proof of Pro
limpp R(0, p) = 0. Wh
larger than K, that
References
Admati A., and P. Pfleid
Review of Financial Studi
Angel, J., 1994, "Limit ve
Beneish, M., and R. Wha

Journal of Finance, 51, 1
Biais, B., P. Hillion, and C

Flow in the Paris Bourse,
1215

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
Biais, B., D. Martimort, and J. C. Rochet, 2000, "Competing Mechanism in a Common Value Environ-
ment," Econometrica, 68, 799-837.
Black, F., 1971, "Towards a Fully Automated Exchange, Part I," Financial Analysts Journal, 27,
29-34.
Bloomfield R., M. O'Hara, and G. Saar, 2002, "The 'Make or Take' Decision in Electronic Markets:
Evidence on the Evolution of Liquidity," Journal of Financial Economics, 55, 425-459.
Chakravarty, S., and C. Holden, 1995, "An Integrated Model of Market and Limit Orders," Journal of
Financial Intermediation, 4, 213-241.
Chung, K., B. Van Ness, and R. Van Ness, 1999, "Limit Orders and the Bid-Ask Spread," Journal of
Financial Economics, 53, 255-287.
Coppejans, M., I. Domowitz, and A. Madhavan, 2003, "Dynamics of Liquidity in an Electronic Limit
Order Book Market," working paper, Duke University.
Cordella, T., and T. Foucault, 1999, "Minimum Price Variations, Time Priority and Quote Dynamics,"
Journal of Financial Intermediation, 8, 147-173.
Degryse, H., F. deJong, M. Ravenswaaij, and G. Wuyts, 2003, Aggressive Orders and the Resiliency of a
Limit Order Market, Mimeo, Leuven University.
Demsetz, H., 1968, "The Costs of Transacting," Quarterly Journal of Economics, 82, 33-53.
Domowitz, I., and A. Wang, 1994, "Auctions as Algorithms," Journal of Economic Dynamics and Control,
18, 29-60.
Easley, D., and M. O'Hara, 1992, "Time and the Process of Security Price Adjustment," Journal of
Finance, 47, 577-605.
Engle, R., and A. Patton, 2003, "Impacts of Trades in an Error-Correction Model of Quote Prices,"
Journal of Financial Markets, 7, 1-25.
Engle, R., and J. Russel, 1998, "Autoregressive Conditional Duration: A New Model for Irregularly
Spaced Transaction Data," Econometrica, 66, 1127-1162.
Foster, F., and S. Viswanathan, 1993, "Variations in trading Volume, Return Volatility and Trading
Costs: Evidence on Recent Price Formation Models," Journal of Finance, 48, 187-211.
Feller, W., 1968, An Introduction to Probability Theory and its Applications (3rd ed.), Wiley, New York.
Foucault, T., 1999, "Order Flow Composition and Trading Costs in a Dynamic Limit Order Market,"
Journal of Financial Markets, 2, 99-134.
Glosten, L., 1994, "Is the Electronic Order Book Inevitable," Journal of Finance, 49, 1127-1161.
Goettler R. L., C. A. Parlour, and U. Rajan, 2003, "Equilibrium in a Dynamic Limit Order Market,"
forthcoming in Journal of Finance, 60, 2149-2192.
Goldstein M., and K. A. Kavajecz, 2000, "Eighths, Sixteenths, and Market Depth: Changes in Tick Size
and Liquidity Provision on the NYSE," Journal of Financial Economics, 56, 125-149.
Griffith, M., B. Smith, D. Turnbull, and R. W. White, 2000, "The Costs and Determinants of Order
Aggressiveness," Journal of Financial Economics, 56, 65-88.
Handa, P., and R. Schwartz, 1996, "Limit Order Trading, Journal of Finance," 51, 1835-1861.
Harris, L., 1998, "Optimal Dynamic Order Submission Strategies in Some Stylized Trading Problems,"
Financial Markets, Institutions and Instruments, 7, 1-91.
Harris, L., 2003, Trading and Exchanges, Oxford University Press, New York.
1216

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC
Harris, L., and J. Hasbrouck, 1996, "Market versus Limit orders: the Superdot evidence on Order
Submission Strategy," Journal of Financial and Quantitative Analysis, 31, 213-231.
Hasbrouck, J., 1999, Trading Fast and Slow: Security Market Events in Real Time, Mimeo, NYU.
Hasbrouck, J., and G. Saar, 2002, "Limit Orders and Volatility in a Hybrid Market: The Island ECN,"
working paper, NYU.
Hollifield, B., A. Miller, and P. Sandis, 2004, "Empirical Analysis of Limit Order Markets," Review of
Economic Studies, 71, 1027-1063.
Hollifield, B., A. Miller, P. Sandas, and J. Slive, 2003, "Estimating the Gains from Trade in Limit Order
Markets," Mimeo, Carnegie Mellon University.
Huang, R., and H. Stoll, 1997, "The Components of the Bid-Ask Spread: A General Approach," Review
of Financial Studies, 10, 995-1034.
Jain, P., 2002, "Institutional Design and Liquidity on Stock Exchanges," working paper, Indiana
University.
Kadan, 0., 2005, "So Who Gains From a Small Tick Size?," forthcoming in Journal of Financial
Intermediation.
Kaniel R., and H. Liu, 2005, "So What Orders Do Informed Traders Use?", forthcoming in the Journal
of Business.
Kavajecz, K., 1999, "A Specialist's Quoted Depth and the Limit Order Book," Journal of Finance, 54,
747-771.
Kavajecz, K., and E. Odders-White, 2003, "Technical Analysis and Liquidity Provision," 17,
1043-1071.
Keim, D., and A. Madhavan, 1995, "Anatomy of the Trading Process: Empirical Evidence on the
Behavior of Institutional Traders," Journal of Financial Economics, 37, 371-398.
Kyle, R., 1985, "Continuous Auctions and Insider Trading," Econometrica, 53, 1315-1335.
Lo, A., C. McKinlay, and J. Zhang, 2001, "Econometric Models of Limit Order Executions," Journal of
Financial Economics, 65, 31-71.
Madhavan, A., M. Richardson, and M. Roomans, 1997, "Why do Securities Prices Change? A Transaction-
Level Analysis of NYSE Stocks," Review of Financial Studies, 10, 1035-1064.
Pagano M., and R. Schwartz, 2003, "A Closing Call's Impact on Market Quality at Euronext Paris,"
Journal of Financial Economics, 68, 439-484.
Parlour, C., 1998, "Price Dynamics in Limit Order Markets," Review of Financial Studies, 11,
789-816.
Parlour, C., and D. Seppi, 2003, "Liquidity-Based Competition for Order Flow," Review of Financial
Studies, 16, 301-343.
Rock, K., 1996, "The Specialist's Order Book and Price Anomalies," working paper, Harvard University.
Rosu, I., 2004, "A Dynamic Model of the Limit Order Book," Mimeo, MIT.
Sandas, P., 2001, "Adverse Selection and Competitive Market Making: Evidence from a Limit Order
Market," Review of Financial Studies, 14, 705-734.
Seppi, D., 1997, "Liquidity Provision with Limit Orders and a Strategic Specialist," Review of Financial
Studies, 10, 103-150.
Tkach, I., 2002, "Liquidity Provision on the Tel Aviv Stock Exchange," Mimeo, Hebrew University.
Tkach, I., and E. Kandel, 2004, "Demand for Immediacy - Time is Money," Mimeo, Hebrew University.
1217

134.58.253.30 on Thu, 12 May 2022 13:21:04 UTC

Limit Order Book

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

Limit Order Book

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Limit Order Book

Uploaded by

Copyright:

Available Formats

Limit Order Book as a Market for Liquidity

Author(s): Thierry Foucault, Ohad Kadan and Eugene Kandel

Stable URL: https://www.jstor.org/stable/3598019

This content downloaded from

We develop a dynamic model of a limit order market populated by strategic liquidity

This content downloaded from

execution speed differently.1 In this article, we explore a dynamic model

This content downloaded from

This content downloaded from

This content downloaded from

empirical implications of the model. In Section 4, we explore the effect of

Consider a continuous market for a single security, organized as a limi

1.2 Patient and impatient traders

This content downloaded from

1.3 Trading mechanism

Assumptions A.1-A.3 facilitate the analysis of the trading game, pri-

This content downloaded from

1.4 Order placement strategies

i) | { Vbuyer -buyer\ - i T() = (Vbuyer - aA) +jA - 6i T() for a buyer

Expressions in parenthesis represent profits associated with market order

max 7ri(') jA - 6iT('), (1)

Equilibrium definition. An equilibrium of the trading game is a pair of

This content downloaded from

Traders' optimal order placement strategies depend on the expected

2. Equilibrium Order Placement Strategies and Market Dynamics

In this section, we first characterize the equilibrium order placement

2.1 Expected waiting time

* T() = +o0 if ao0() = 0 and C {2,... ,K - 1},

* ) T()-'-o) a 3-k ak(' ) T(k) if ao(/)>0 and j E {2,...

Assumption A.2 implies that a trader, who faces a one-tick

This content downloaded from

attracts a market order [i.e., such that ao(/) = 0] is obviously infinite. If

2.2 Equilibrium strategies

7ri() =jA - 6iT(),

This content downloaded from

ji ( CF(-) i E {PI} (2)

We call patient traders'

2.2.1 The homogeneous

7i('*) > 0 for i e {P,I}. (4)

Now, consider a trader who faces a book with an in

Proposition 1. When traders are homogeneous (i.e.,

The equilibrium with homogeneous traders has two distinctive proper-

This content downloaded from

2.2.2 The heterogeneous case. Now consider heterogeneous traders, that

Proposition 2. Suppose traders are heterogeneous (jp<j1). In equilibrium

1. Facing a spread s E (1 ,j), both patient and impatient traders submit

2. Facing a spread s c (jp + 1, sc), a patient trader submits a limit

The range of possible spreads is partitioned into three regions:

This content downloaded from

Proposition 3. In equilibrium, there exist q spreads (K > q > 2),

Proposition 4. The expected waiting time function in equilibrium is.

T*(') = T*(nh) j (nh- + 1, nh) V h = 1,..., q - 1.11

This content downloaded from

nhA - T*(h)P n - T*(nh) > _/- ( )6p, Vh C {2,... q-1},

th -nh- h-l > [T*(nh) T*(nh_1)], Vh C {2,...,q-1}. (6)

Now consider a trader facing a spread nh. In equilibrium, this trader

nh-lA - T*(nh-1)6P > (nh - 1)A - T*(nh - 1)6p Vh E {2,..., q}.

In equilibrium, either (i) T*(nh - 1) = T*(nh) or (ii) T*(nh - 1) = T* (nh-).

h = CF([T*(nh) - T*(nhl)]) - CF(2ph- )Vh c {2,.., q- 1},

T(') = T(nh) j (nh- + 1, nh) V h = 1,..., q - 1.11

nhA - T(h)P n - T(nh) > _/- ( )6p, Vh C {2,... q-1},

th -nh- h-l > [T(nh) T(nh_1)], Vh C {2,...,q-1}. (6)

nh-lA - T(nh-1)6P > (nh - 1)A - T(nh - 1)6p Vh E {2,..., q}.

In equilibrium, either (i) T(nh - 1) = T(nh) or (ii) T(nh - 1) = T (nh-).

h = CF([T(nh) - T(nhl)]) - CF(2ph- )Vh c {2,.., q- 1},