T
YPES IN A
L
IMIT
O
RDER
M
ARKET
: E
VIDENCE
FROM
T
AIWAN
I
NDEX
F
UTURES
JUNMAO CHIU, HUIMIN CHUNG and GEORGE H. K. WANG*
This study examines the dynamic liquidity provision process by institutional and individual traders in the Taiwan index futures market, which is a pure limit order market. The empirical analysis obtains several interesting empirical results. Wefind that trader type affects liquidity provision in a number of interesting ways. First, although institutional traders use more limit orders than market orders, foreign institution (individual) traders use a relatively higher percentage of market (limit) orders in the early trading session and then switch to more limit (market) orders for the remainder of the day until close to the end of the trading day. Second, net limit order submissions by both institutional and individual traders are positively related to one‐period lagged transitory volatility and negatively related to informational volatility. Third, net limit order submissions by institutional traders are positively related to one‐period lagged spread. Finally, both the state of limit order book and order size significantly influence all types of traders’ strategy on submission of limit order versus market order during the intraday trading session. © 2012 Wiley Periodicals, Inc. Jrl Fut Mark 34:145–172, 2014
1. INTRODUCTION
Electronic limit order market is one of major trading venues in equity, futures, and option exchanges around the world. Because no designated market makers exist in these markets, limit orders supply liquidity and market orders consume liquidity. Thus, liquidity arises endogenously from the orders submitted by market participants in the exchanges. Because liquidity is a major performance measurement for exchanges, understanding the factors affecting the limit order submission rate by different types of traders under different market
conditions is of interest to researchers, exchange officials, and investors.
Junmao Chiu is a Ph.D. Associate at the Graduate Institute of Finance, National Chiao Tung University, Taiwan. Huimin Chung is a Professor of Finance at the Graduate Institute of Finance, National Chiao Tung University, Taiwan. George H. K. Wang is the Research Professor of Finance at the School of Management, George Mason University, Fairfax, Virginia.The authors would like to thank the anonymous referee and Robert Webb (the editor) for their constructive comments and suggestions that significantly improved the quality of the study. An earlier version of this study was presented at the 2011 Financial Management Association Annual Meetings, in Denver, Colorado. The 4th NCTU International Finance Conference, in Hsinchu, Taiwan, and The AsianFA and TFA 2012 Joint International Conference, in Taipei, Taiwan.
A part of this work was done when Junmao Chiu was a visiting scholar in the Finance area, School of Management, George Mason University, Fairfax, VA 22030.
*Correspondence author, School of Management, George Mason University, 4400 University Drive, Fairfax, Virginia 22030. Tel: 703‐993‐3415, Fax: 703‐993‐1870, e‐mail: gwang2@gmu.edu
Received September 2011; Accepted August 2012
The Journal of Futures Markets, Vol. 34, No. 2, 145–172 (2014) © 2012 Wiley Periodicals, Inc.
Published online 28 November 2012 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/fut.21586
Previous literature has approached limit order trading strategy through both theoretical models and empirical analysis. Earlier theoretical models assume that informed traders who
trade on short‐lived, private information are impatient and thus place market orders, whereas
uninformed traders who use limit orders must await execution (Glosten, 1994; Seppi, 1997). Later theoretical models (e.g., Chakravarty & Holden, 1995; Harris, 1998; Kaniel & Liu, 2006) relax this restrictive assumption. They suggest that informed traders use both limited orders and market orders. In general, they show that the time horizon of private information is positively related to the probability of using limit orders by informed traders.
Using an experimental asset market, Bloomfield, O’Hara, and Saar (2005) investigate
empirically the evolution of the liquidity provision by trader type in a pure limit order market
under an experimental market setting. They find that informed and liquidity traders use
reverse strategies: Although informed traders consume liquidity earlier in the trading day, gradually becoming liquidity providers as they increasingly place more limit orders as the trading day progresses, liquidity traders provide liquidity early in the trading day, gradually shifting to consume liquidity as the day progresses. They also report that informed traders use relatively more limit orders. These experimental results challenge the assumptions of the theoretical models on the order choice of informed traders in a limit order market.
Goettler, Parlour, and Rajan (2005) study the dynamics of order choices in a limit order market under asymmetric information. They suggest that the volatility of changes in the fundamental value of an asset affects agents acquiring information about the asset, which in
turn affects the choice of order type of informed traders and market outcomes.1 Keim and
Madhavan (1995) present empirical evidence on the order choices of institutional traders.
They find that informed traders with short‐lived information tend to use market orders
whereas informed traders with long horizon information (e.g., value traders) are more likely to use limit orders.
On the empirical literature, Biais, Hillion, and Spatt (1995) examine the relation
between the limit order book and the order flow in the Paris Bourse. They find that the
conditional probability of submitting limit (market) orders by investors is higher when the spread is wide (tight). Chung, Van Ness, and Van Ness (1999) also show that traders place more limit orders when the intraday spread is wide in New York Stock Exchange (NYSE). Ahn, Bae, and Chan (2001) examine the role of limit orders in providing liquidity in the Stock
Exchange of Hong Kong (SEHK), a pure limit order market. Theyfind that one lagged period
transitory volatility is the major determinant of market depth (due to the submission of limit
orders) and that a rise in market depth is followed by a decrease in volatility.2Volatility also
determines the changing mix of market and limit orders.
Bae, Jang, and Park (2003) examine the trader’s choice between limit and market orders
using a sample from the NYSE SuperDot. Theyfind that the order size, spread, and expected
transitory volatility are positively related with trader’s limit order choice. Using data from the
Moscow Interbank Currency Exchange, Menkhoff, Osler, and Schmeling (2010) investigate the use of aggressive price limit orders by informed and uninformed traders in an ordered logit regression framework. They show that informed traders are more sensitive to changes in the spread, volatility, and market depth than uninformed traders in a pure limit market. We extend this line of research by investigating the difference in market impact on order submission 1
Goettler et al. (2009, p. 68) obtain their results numerically from a theoretical model because they cannot obtain a closed form solution when the relevant frictions of a limit order market are incorporated in the model. The relevant frictions of a limit order market are discrete price staggered trader arrivals and asymmetric information. For other theoretical models on the dynamics of order choice in limit order markets, see Rosu (2009) and Parlour and Seppi (2008).
2
Ahn et al. (2001) do not accurately estimate transitory volatility; they use realized volatility to approximate transitory volatility.
strategy by different trader types in the real world market settings. We then divide traders by type into individual traders and institutional traders with individual (institutional) traders
further categorized as day traders and nonday traders (foreign institutional firms and
proprietary futuresfirm traders). To the best of our knowledge, this analysis is the first study to
examine and compare the dynamic liquidity provision process by institutional and individual traders.
To investigate liquidity provision of these four trader types, we first document the
intraday liquidity provision in a pure limit order market using the actual intraday data from the
Taiwan index futures market for the period from January 2007 to December 2008.3Second,
we examine the impact of various market conditions (i.e., one‐period lagged transitory and
informational volatility, one‐period lagged spread, one‐period lagged same and opposite side
market depth, and limit order size) on the liquidity provision by trader types in a joint
regression framework.4Finally, we compare our empirical results on the changing liquidity
provision by trader type in a natural market setting with the experimental market results of
Bloomfield et al. (2005).
Our study contributes to the literature in several ways. First, our empirical results from a
natural market setting support the experimental market setting results of Bloomfield et al.’s
(2005) study on the intraday trading strategies of informed traders and uninformed traders.
Second, in the influence of market characteristics on the limit and market order choices
decision, we show that net limit order submissions by both institutional and individual traders
are positively related to one‐period lagged transitory volatility and negatively related to
informational volatility. We conduct a direct test on the prediction of Handa and Schwartz
(1996) (vs. Foucault, 1999) on the influence of transitory volatility and informational volatility
on institutional versus individual trader’s decision on selection of limit versus market orders.
To the best of our knowledge, this portion of our analysis adds newfindings to the limit order
literature. Third, wefind that the net limit order submissions by foreign institutional traders
and futures proprietary firm traders are positively related to one period lagged spread;
conversely, no significant relation exists between lagged one period spreads and the limit order
submissions by individual day or noonday traders. Finally, the results also suggest that both
the state of limit order book and order size significantly influence the strategy of all trader types
on submission of limit order versus market order during the intraday trading session. Our study is organized as follows. In Section 2, we present a literature review related to the impact of market conditions on the supply of liquidity by institutional and individual traders in a limit order market. In Section 3, we describe the Taiwan index futures market structure and the data. In Section 4, we present the empirical methodology. In Section 5, we present the empirical results, and Section 6 concludes.
2. LITERATURE REVIEW AND HYPOTHESES
2.1. Trading Strategies: Informed Versus Uninformed Traders
In a pure limit order market, traders face decisions between limit orders and market orders. Market orders consume liquidity and are executed with certainty at the posted prices in the 3
Thefinancial literature generally agrees that institutional traders are informed traders because they collect and analyze market information more quickly than uninformed traders in index futures markets. However, individual investors often follow their observed market prices pattern as their major inputs for their trading decision. 4
Prior studies include only a subset of our market condition variables in their regression models. For example, Bae et al. (2003) do not include the state of limit order book variable in their regression, and Bloomfield et al. (2005) examine the impact of each market condition variables separately on the submission of limit versus market orders by trader types in their experimental setting.
market. Limit orders supply liquidity and have the advantage of execution at more favorable prices than market orders. However, limit orders face execution uncertainty and an adverse
selection risk because limit order prices arefixed. Limit order traders provide free options to
the arrival of informed traders (Copeland & Galai, 1983).
Earlier theoretical models (e.g., Glosten, 1994; Seppi, 1997) assume that informed traders place market orders because they are impatient and that private information is
short‐lived whereas uninformed traders supply liquidity by submitting limit orders and waiting
for execution. Later theoretical models relax this restrictive assumption. For example,
Chakravarty and Holden (1995) analyze the behavior of the informed trader in a single‐period
call‐type market. They show that in this type of market the informed trader may
simultaneously submit a market buy order and a limit sell order, and limit order acts as a safety net for the market order. In this way, an optimal mix of limit orders and market orders leads to a higher payoff than submitting only market orders when uncertainty exists regarding the price that a market order will fetch.
Harris (1998) develops optimal order submission strategies for trading problems faced by
an informed trader, a uniformed trader, and a value‐motivated trader. He suggests that
informed traders are more likely to use market orders when private information will soon
become public, reflecting the desire of informed traders to realize their valuable private
information. He also predicts that liquidity traders will start by using limit orders and then switch to market orders as the end of trading approaches to meet their trading target. When
informed traders face early deadlines, they also employ market orders. Finally, Harrisfinds
that both informed and uninformed traders submit limited orders when the deadline is distant
and the bid–ask spread is large to minimize transaction costs. In general, Harris (1998)
suggests that informed traders use relatively more market orders than limit orders.
Kaniel and Liu (2006) analyze informed traders’ equilibrium choice of limit and market
orders. They show that the time horizon of private information is positively related to the probability of using limit orders by informed traders. Their empirical results show that informed traders prefer to use limit orders, which are indeed more informative.
Bloomfield et al. (2005) employ experimental asset markets to investigate the evolution
of liquidity provisions by informed and liquidity traders in a pure limit order market. Their study focuses on how trading strategies are affected by trader type, market conditions, and
characteristics of the asset at different time points during a trading day. They find that
informed traders use more market orders than limit orders at the earlier stage of the trading session because informed traders are likely to capitalize on their private information. As the trading progresses, informed traders switch to liquidity provisions. The change in the behavior of informed traders seems to be in response to the dynamic adjustment of price to information.
Informed traders perform better in terms of profit as liquidity suppliers because they face less
adverse selection risk when placing limit orders in comparison to uninformed traders. Their result suggests that informed traders take (provide) liquidity when the value of information is high (low). Uninformed traders supply relatively more liquidity in the earlier stage of the trading session and use relatively more market orders as trading comes to a close
because of their need to meet the target value of their trading purposes. Bloomfield et al.
(2005) also document the difference in the impacts of market conditions (i.e., the volatility, the spread, the state of limit orders) on order choice between informed and uninformed traders. Their experimental results suggest the need to further relax the assumptions of theoretical models and point to an urgent need for a dynamic model on the order choice by trader types in a limit order market.
Anand, Charkravarty, and Martell (2005) empirically investigate the evolution of liquidity and changing of trading strategies of institutional traders (i.e., informed traders) and find that institutional traders use market orders more often in the first half than in the second
half of the trading day. They also document that limit orders placed by institutional traders perform better than those placed by individual traders (i.e., uninformed traders). However, their tests are based on the intraday data for the period from November 1990 to January 1991 obtained from NYSE, which is not a pure limit order market.
2.2. Influence of Market Characteristics: Limit Versus Market Orders
The important market characteristic variables that affect the trader’s choice on limit or market
orders are volatility, spread, the state of limit order book, and order size. Handa and Schwartz (1996) develop a model to explain the rationale of trader choice between markets and limit
order and the profitability of limit order trading. In their model, the trader’s choice depends on
the probability of whether their limit order is executed against an informed trader or an uninformed (liquidity) trader. A limit order suffers a loss when executed against an informed
trader and experiences a profit when executed against a liquidity trader. Thus, traders will
submit more limit orders than market orders when the increase in price volatility is caused by
change in market liquidity reasons. That is, the profitability of limit orders increases as traders
increase in the supply of liquidity. Thus, Handa and Schwartz (1996) predict a positive relation between submission of limit orders and transitory price volatility.
Foucault (1999) develops a model that explicitly incorporates a trader’s decision to
submit market versus limit orders. He theorizes that when the asset volatility increases due to informed traders, the risk of adverse selection increases. Thus, limit order traders must
increase their bid–ask spreads to insure against losses. The cost of trading on market orders is
less attractive, and tradersfind it more cost‐effective to trade using limit orders.
Ahn et al. (2001) use 33 component stocks in the Hang Seng Index (HIS) between July 1996 and June 1997 and show that an increase in market depth follows a rise in transitory volatility due to an increase in submission of limit orders. A decrease in volatility subsequently follows an increase in market depth. These results are consistent with the predication of the theoretical model of Handa and Schwartz (1996). Bae et al. (2003) use a sample of 144 NYSE
list stocks over the period from November 1, 1990 to January 31, 1991 to investigate trader’s
choice between limit and market orders. Theyfind that traders use more limit orders when
they expect an increase in transitory volatility. They find the impact of the asset
(informational) volatility on trader’s choice between limit and market orders is inclusive.
Bloomfield et al. (2005) also find that volatility is one of the major factors affecting both
informed and uninformed traders’ choice between limit and market orders.
Menkhoff et al. (2010) investigate the use of aggressive price limit orders by informed traders and uninformed traders in an ordered logit regression framework with data from the Moscow Interbank Currency Exchange. They show that the volatility variable is negative and
highly significant for informed traders and significant at the 10% level for uninformed traders.
Their results suggest that both types of traders increase their use of limit orders following an increase in volatility.
Menkhoff et al. (2010) alsofind that informed traders are more sensitive to change in the
spreads, volatility, and depths than uninformed traders in a pure limit market. There are two major concerns in their quality of data used in empirical tests: (a) the data lack trader
identification codes on trader type, and thus the authors identify traders as either informed or
uninformed based on their inference from the trade size and location information and (b) the
data cover only a seven‐intraday data period, which may be too short for reliable empirical
tests.
Biais et al. (1995) provide empirical evidence that when the spread is large, the conditional probability increases that investors will place more limit orders than market
orders. In contrast, traders will use more market orders (i.e., hitting the quote) than limit orders when the spread is tight. Similarly, Chung et al. (1999) examine limit order book and
the bid–ask of 144 stocks traded in NYSE. They provide evidence that more traders submit
limit orders when the spread is wide and use market orders when the spread is tight. These results, which suggest that when the spread is wide, traders place more limit orders, may be due to the high cost of submitting market orders or because traders receive higher compensation by executing limit orders.
Previous literature has shown that the state of the limit order book influences a trader’s
order choice. Parlour (1998) provides a theoretical model that suggests that traders are less
likely to use limit orders if the limit book on the same side of the trade is thicker. This so‐called
“crowding out” effect arises because the time priority of orders already in the book lowers the probability of executing of a new order on the same side. However, traders are more likely to
use limit orders if the book on the other side of the trade is thicker. Bloomfield et al. (2005)
examine this hypothesis in an experimental market setting. Their results lend support to
Parlour’s model prediction that traders will use more limit orders as the depth of the other side
increases. However, Bloomfield et al. (2005) find that informed and liquidity traders have
higher limit order submission ratio when the same side of order book is thicker. This result is
inconsistent with Parlour’s model prediction.
Based on order and transaction intraday data from the Swiss stock exchange, Ranaldo (2004) also demonstrates that patient traders become more order aggressive when their own (opposite) side book is thicker (thinner). Using limit order book information from the Australian Stock Exchange (ASX), Cao, Hansch, and Wang (2008) also provide additional empirical evidence that traders use more market orders when the same side of limit order book is thicker.
In general, traders have strong motives to minimize their trading cost when the order size is relatively larger. Bae et al. (2003) divide their sample into two order groups based on size and find that, on average, traders in large order size group use more limit orders, ranging from 66%
to 79% of the total orders in a trading day. In small order size group, 28–36% of the orders are
limit orders. These results provide evidence that traders tend to use limit orders when the order size is relatively large.
Building on the results from previous literature, we use unique real world data to examine the differences among institutional, individual day and nonday traders in providing liquidity in response to changes in market conditions during a trading day in a joint regression model.
3. TAIWAN INDEX FUTURES MARKET STRUCTURE AND THE DATA
The Taiwan Futures Exchange (TAIFEX) is a pure order‐driven market. Investors submit limit
and market orders through brokers to the automated trading systems (ATSs). Limit orders are
consolidated into the electronic limit‐order book. The ATS order matches and executes orders
continuously following a price–time priority rule and setting a single transaction price.
Markets buy (sell) orders hit the best ask (bid) prices. The buy (sell) order with higher (lower) limit price than the set transaction price is executed at the transaction price. Market participants can also submit cancel orders at any time prior to matching. The preopen session is from 8:30 a.m. to 8:45 a.m. During this period, investors can submit limit and market orders
to the ATS system through brokers, and the exchange uses the single‐price auction system to
establish the opening prices of regular trading hours. The regular trading hours conducted on weekdays excluding public holidays are 8:45 a.m. to 1:45 p.m. Limit orders are automatically
canceled at the end of trading day; thus, we work with a one‐day limit order book. No hidden
TAIFEX disseminates order and transaction prices to the public in real time. Investors
can observe on the screen the specific anonymous best five bid and best five ask prices with the
number of contracts. Because no designed market makers exist, market participants generate liquidity endogenously by placing orders.
We use intraday tick‐by‐tick data of Taiwan stock index futures (FITX) obtained from
TAIFEX in our analysis. Our sample period covers from January 1, 2007 to December 31, 2008. The contract size is the index value of FITX multiplied by 200 New Taiwan Dollars (NT$). The maximum of each order size of TIFX is 100 contracts. We use nearby futures contracts in our analysis, and trading volume in the delivery month is used as the indicator to
switch fromfirst deferred contract to nearby futures contract. In our data‐editing process, we
eliminate price limit days, time periods without limit order information and days with missing
trading data.5The data set contains the detailed history of orderflows, order book, transaction
data, and the identity of the traders. For each order, the date and time of arrival of the order, its direction (buy or sell initiation), the quantity demanded or supplied, and the trader
identification are recorded. The trader identification enables us to categorize four types of
traders: individual traders, domestic institution traders, futures proprietaryfirms, and foreign
institutional traders.
Panel A of Table I shows that the daily average trading volume is about 93,684 contracts.
Individual traders account for 61% of the total daily average volume. Futures proprietaryfirms
are different from futures brokers in that they trade for their own accounts to make profits and
also make commissions by trading for clients. Their trading activity accounts for 23.34% of daily average total volume. Foreign Institutional traders executed about 12.26%, and domestic
TABLE I
Daily Trading Volume Statistics by Trader Type Individual Traders (%) Day Trader (%) Nonday Trader (%) Domestic Institutional Traders (%) Foreign Institutional Traders (%) Futures Proprietary Firms (%) Total Daily Average Panel A: Percentage of Total Volume by Trader Type
Trading volume 30.40 30.31 3.69 12.26 23.34 93,683.69
Panel B: Percentage of Total Volume of Day Trading Versus Nonday Trading by Trader Type
Day trading 94.78 1.54 2.83 0.85 100 (50.06) (13.40] (7.41) (1.17) (32.07) [30,044.36] Nonday trading 44.64 4.70 16.71 33.96 100 (49.94) (86.60] (92.59) (98.83) (67.93) [63,639.33] Total 60.72 3.69 12.26 23.34 100 (100) (100) (100) (100) (100) [56,875.37] [3,456.93] [11,485.62] [21,865.77] [93,683.69]
Note. In this table, we provide daily trading volume statistics by trader type in the Taiwan Stock Exchange index futures from January 1, 2007 to December 31, 2008. Panel A shows the percentage of daily trading volume for individual day traders, individual nonday traders, domestic institutional traders, foreign institutional traders, and futures proprietaryfirms traders. Panel B separates trading volume into day trading and nonday trading. The numbers in parentheses represent the percentages of day trading and nonday trading for each trader type. For example, among foreign institutional traders, 7.41% engage in day trading and 92.59% engage in nonday trading. A trader is defined as a day trader when the amounts of contracts purchased and sold are the same in a specific day. The numbers in brackets are the total number of average trades by trader type.
5
institutional traders account for only 3.69% of daily average trading volume. Due to their low trading activity, which leads to frequent inadequate observations, we drop domestic
institutional traders from our sample.6 Our analysis assumes that foreign institutional
investors and futures proprietaryfirms are members of institutional traders and that individual
traders are uninformed or liquidity traders.7Panel A shows that day trading in total trading
volume accounts for about 30.4%, whereas individual nonday trading accounts for 30.31% of
total volume.8Our results are similar to the results reported by Barber et al. (2009), whofind
that day trading by individual traders is over 20% in the Taiwan stock market.
4. METHODOLOGY
Our empirical analysis consists of two steps. First, we use one‐way analysis of variance model
to estimate the intraday submission patterns of limit orders, market orders, and limit order
submission ratios. Second, we use regression models to estimate the influences of market
condition variables (i.e., Transitory_Volatilityt1, Informational_Volatilityt1, Spreadt1,
Same_Side_Deptht1, Opposite_Side_Deptht1, and Limit_Sizet) on net limit order
submis-sion by institutional and individual traders.
In the analysis of intraday variation patterns of order choices by trader types, we follow two principles to select the length of the time interval. First, we are interested in short time
variations in limit and market order submissions. Second, the time interval must be sufficient
to provide reliable estimates of intraday patterns. Balancing these two guidelines, we select a
15‐minute interval.
The one‐way analysis of variance regression model is specified as follows:
Yt¼ d0þ
X19
j¼0
bjDj;tþ et: ð1Þ
The dependent variable Ytis equal to the sum of limit orders in a 15‐minute time interval,
the sum of market orders in a 15‐minute time interval, or the limit order submission ratio in a
15‐minute interval. The value of intercept d0 is equal to the daily average as the basis of
comparison. For this reason, we impose the restrictionP20j¼0bj¼ 0. Dj,tis a dummy variable
that equals 1 if it is in jth interval, j¼ 0,1,2,…,19; zero, if it is not in the jth interval; and equal
to1 if it is in 20th time interval.9The error term is et. The coefficient of bjis equal to the
difference between the mean of jth time interval and the value of d0, the daily average. The
sample mean of jth time interval is equal to the sum of the values of bjþ d0. This model allows
us to examine the influences of the role of the time interval on order submissions by trader
type.
The regression model used to examine the influence of characteristics of market
conditions lagged one period (i.e., Transitory_Volatilityt1, Informational_Volatilityt1,
6
In the rest of our analysis, we concentrate only on activities of individual traders, foreign institutional traders, and futures proprietaryfirm traders because the trading activity of domestic traders only accounts 3.69% of average daily trade volume. In addition, domestic institutionalfirms do not trade very frequently. As a result, we often face inadequate observations of domestic institutionalfirms in our 15‐minute time interval.
7
Goettler et al. (2009, p. 68) suggest institutional traders are informed traders who view the current expected value of cashflow on the instrument. This finding implies that informed traders perform research on the value of the instrument while uninformed agents estimate the value of the instrument based on market observables.
8Day trader is defined as a trader who satisfies the following rule: The amount of contracts purchased is equal to the amount of contracts sold in the same trading day.
9
Spreadt1, Same_Side_Deptht1, and Other_Side_Deptht1) and order size on liquidity provision by institutional and individual traders is
NLMt¼ a þ b1Spreadt−1þ b2Transitory Volatilityt−1
þ b3Informational Volatilityt−1þ b4Same Side Deptht−1
þ b5Other Side Deptht−1þ b6Limit Sizetþ
X19
j¼1
b7; jDjþ et: ð2Þ
The dependent variable, the net sum of limit order (NLMt), denotes the sum of limit
orders minus market orders and marketable limit orders during the 15‐minute interval.10
Spreadt1is the average of all dollar quote spreads during t 1 time period. Same_Side_Deptht1
(Other_Side_Deptht1) is measured as the average number of limit orders at the best bid (ask)
just prior to a buy order’s submission and as the number at the best ask (bid) just prior to a sell
order’s submission at a given time in the t 1 time interval.
Previous literature has reported a positive relation between total price volatility and submissions of limit orders by traders. Handa and Schwartz (1996) hypothesize that an increase in transitory volatility will attract new limit orders and that an increase in informational volatility will discourage the submission of new limit orders due to an increase in adverse selection risk. Conversely, Foucault (1999) argues that when informational volatility increases, traders will submit more limit orders even though they face increasing adverse selection risk. During periods of increased informational volatility, traders face higher trading
costs due to higher bid–ask quotes. Thus, market order trading is even more expensive than
limit order trading, and more tradersfind it optimal to implement their trades using limit orders.
To test these two competing hypotheses, we decompose total volatility into transitory volatility and informational volatility. To estimate transitory variance and informational variance, we assume transaction price follows a random walk model with transitory noise. The
local‐level model is specified as11
Pt ¼ mtþ jt jt NID 0; s2j
mt¼ mt−1þ yt yt NID 0; s2y
; ð3Þ
where Ptis transaction price; mtis unobserved equilibrium (efficient) price that follows a
random walk model; and jtis transitory component. We use the Kalmanfilter technique to
estimate the parameters of the Model (3) for each 15‐minute interval.
We use sjas our measure of transitory volatility in each 15‐minute interval and syas our
measure of informational volatility in each 15‐minute interval. Bae et al. (2003) use Model (3)
to estimate intraday‐efficient price and transitory price for each day and then employ high–low
price range in 30‐minute intervals to estimate the transitory and informational volatility,
respectively, for each time interval. In our case, we obtain the estimates of transitory volatility
and informational volatility from the empirical results of Model (3) applied to each 15‐minute
interval. We employ Transitory_Volatilityt1 (transitory volatility lagged one period) and
Informational_Volatilityt1 (informational volatility lagged one period) to approximate a
trader’s view on expected transitory and informational volatility in next time period.
10
Marketable limit orders are limit orders that come with better quotes than the current best quotes in the order book. 11See Harvey (1989) for further discussion on this unobserved component (local level) model. Hasbrouck (1996) discusses this type of model with application tofinance, and Bae et al. (2003) apply this model to decompose the transactions into efficient and transitory price components.
We measure Limit_Sizetas the average size of all limit orders for all traders during the tth
time interval. The dummy variable Di,t, as defined in the Equation (1), controls intraday
variation of limit order submission patterns with respect to time.
We estimate both Equations (1) and (2) for each type of traders using ordinary least squares. Newey and West (1987) heteroskedasticity and autocorrelation covariance procedure is used to calculate the consistent standard errors of estimates.
5. EMPIRICAL RESULTS
5.1. Intraday Variation of Limit and Market Orders by Trader Types
Panels A–C of Table II present the average daily market and limit order submissions by trader
type for whole sample period, for the prefinancial crisis period (January 2007 to July 2007) and
for thefinancial crisis period (August 2007 to December 2008), respectively.12We sort all
orders into pure market order, marketable limit order, and limit order. The numbers in parentheses for each row represent the percentages of order types for individual day traders,
individual nonday traders, foreign institutional traders, and futures proprietaryfirm traders.
For example, for the whole sample period, the total daily average order submissions of day traders is composed of 17.26% of pure market orders, 8.94% of marketable limit order and 73.80% of limit orders. The numbers in brackets represent the total number of average trades by each trader type.
Panel A (whole sample) provides several interesting observations: (a) the sum of the average pure market order and marketable limit order submissions is only 16.76%, compared to 83.24% for limit order submissions; (b) individual day traders and nonday traders submit 73.80% and 73.09%, respectively, of their total orders in limit orders whereas 94.08% and 92.62% of the total order submission of foreign institutional traders and futures proprietary firms, respectively, are limit orders. These results confirm that, in general, institutional traders use more limit orders than market orders. Our results are consistent with Kaniel and Liu
(2006) and Bloomfield et al. (2005) but do not support Harris (1998) who predicts that
informed traders use more market orders than limit orders.
Panel C of Table II shows that the order submissions by trader types during thefinancial
crisis period are very similar to the order submissions by trader types during whole sample period.
This result is not surprising because the time period of financial crisis period accounts for
three‐fourths of the whole sample period. Panel B shows that during the prefinancial crisis period
(January 2001 to July 2007) individual day traders use slightly less market orders and marketable
limit orders and relatively more limit orders than during thefinancial crisis period. Also, during
the prefinancial crisis period, the sum of foreign institutional traders and futures proprietary firm
accounts for 33% of total average daily order submission, and yet the sum of their daily trading
volume accounts for total average daily order submission duringfinancial crisis period. These
results suggest that individual traders are trading more active during the prefinancial crisis period
and that institutional traders are more active during thefinancial crisis period.
In Table III, we present the regression analysis of the intraday variation of limit and
market orders by trader types on 15‐minute time intervals. See Appendix for further details.13
12Following Brunnermeier (2009) as well as Melvin and Taylor (2009), we define the beginning of the subprime crisis period as August 2007. We thus divide our sample period into prefinancial crisis period (January 2007 to July 2007) andfinancial crisis period (August 2007 to December 2008).
13Appendix is a supplement to Table III. It presents the means of the numbers of limit and market orders submitted by trader type on a 15‐minute time interval. Limit order submission ratio in the Appendix is the ratio of the mean of the number of limit orders to the sum of limit orders, market orders, and marketable limit orders.
TABLE II
Daily Order Book Statistics by Trader‐Type Categories
Individual Trader Day Trader (%) Nonday Trader (%) Foreign Institutional Traders (%) Futures Proprietary Firm Traders (%) Total Daily Average Orders Panel A: Full Sample Period (January 2007 to December 2008)
Pure market order 49.15 47.15 1.57 2.13 100
(17.26) (17.48) (0.57) (0.82) (9.17)
[22,465.46]
Marketable limit order 30.80 30.74 17.75 20.71 100
(8.94) (9.43) (5.35) (6.56) (7.59)
[18,581.06]
Limit order 23.16 21.73 28.48 26.63 100
(73.80) (73.09) (94.08) (92.62) (83.24)
[203,841.30]
Total daily order average 26.13 24.74 25.20 23.93 100
(100) (100) (100) (100) (100)
[63,981.86] [60,590.68] [61,705.38] [58,609.91] [244,887.84]
Panel B: Prefinancial Crisis—January 2007 to July 2008
Pure market order 38.87 56.41 1.50 3.22 100
(12.73) (15.53) (0.95) (1.80) (9.95)
[14,278.45]
Marketable limit order 22.93 37.36 17.16 22.54 100
(5.81) (7.95) (8.43) (9.73) (7.69)
[11,040.01]
Limit order 30.04 33.58 17.23 19.14 100
(81.46) (76.52) (90.62) (88.47) (82.36)
[118,185.89]
Total daily order average 30.38 36.15 15.66 17.82 100
(100) (100) (100) (100) (100)
[43,589.69] [51,871.18] [22,472.08] [25,571.40] [143,504.35]
Panel C: During Financial Crisis—August 2007 to December 2008
Pure market order 51.48 45.06 1.57 1.89 100
(18.37) (18.13) (0.52) (0.67) (9.01)
[25,818.09]
Marketable limit order 32.44 29.36 17.88 20.32 100
(9.72) (9.92) (4.98) (6.10) (7.57)
[21,673.68]
Limit order 21.77 19.32 30.76 28.15 100
(71.91) (71.95) (94.49) (93.22) (83.42)
[238,950.09]
Total daily order average 25.25 22.40 27.16 25.19 100
(100) (100) (100) (100) (100)
[72,340.28] [64,164.67] [77,785.05] [72,151.86] [286,441.86]
Note. In this table, we present a daily order book statistics by trader type in the futures contract FITX from whole sample period, prefinancial crisis, and during financial crisis periods. We divide all order books into the pure market order, marketable limit order, and limit order and give the percentages of order types by individual day traders, individual nonday traders, foreign institutional traders, and futures proprietaryfirm traders. The numbers in parentheses represent the percentages of order types by each trader types. For example, during whole sample period, the total daily average orders of day traders, 17.26% are pure market orders, 8.94% are marketable limit order, and 73.80% are limit orders. The numbers in brackets are the total number of average trades by trader type.
TABLE III Regression Analysis of Intraday Variation Patterns of Limit and Market Orders by Type of Traders Indiv idual Day Trade rs In dividual Non day Tra ders Foreign Institutiona l Trad ers Futures Proprieta ry Firm Trade rs Time Interval Lim it Marke t Lim it Ord er Submis sion Ra tio (%) Lim it Marke t L imit Ord er Sub mission Ratio (%) Limit Marke t Lim it Order Submission Ra tio (%) Lim it Mar ket Limit Or der Su bmission Ratio (%) 0. 08:30 – 1,168. 70 684.7 1 0.1485 655 .68 414.46 0.0700 2,6 28.85 116.9 5 0.2 364 2,155. 82 149.27 0.0348 08:45 hours ( 21.13) ( 25. 36) (46.70 ) ( 14. 37) (18.0 4) (23.09 ) ( 18. 84) ( 12. 27) ( 47.42) ( 20.06) (17.44) ( 13.93) 1. 08:45 – 152.44 152.80 0.0 415 1, 102.33 539.62 0.0 256 298.62 79.55 0.0166 403.85 100.71 0. 0155 09:00 hours (2.75) ( 5. 64) (13.00 ) (24.0 6) (23.3 9) ( 8. 41) ( 2.1 3) (8.31 ) ( 3.3 1) (3.7 4) (11.72) ( 6.19) 2. 09:00 – 805.39 295.9 5 0.0 118 649.6 8 406.30 0.0 407 1,0 31.12 128.3 1 0.0 021 964.08 133.07 0. 0154 09:15 hours (14 .52) (10.93 ) ( 3. 71) (14.2 0) (17.6 3) ( 13. 38) (7.37) (13.42 ) (0.43) (8.9 5) (15.50) ( 6.16) 3. 09:15 – 664.03 227.1 2 0.0 050 451.8 1 215.21 0.0 194 1,0 35.06 57.47 0.0 157 687.73 66. 97 0. 0056 09:30 hours (11 .97) (8.39 ) ( 1. 58) (9.87 ) (9.34 ) ( 6. 37) (7.40) (6.01 ) (3.14) (6.3 8) (7.80) ( 2.23) 4. 09:30 – 401.64 125.8 0 0.0 022 225.1 8 49.37 0.0 019 723.01 16.76 0.0 204 406.30 20. 00 0. 0030 09:45 hours (7.25) (4.65 ) ( 0. 67) (4.92 ) (2.14 ) ( 0. 61) (5.17) (1.75 ) (4.09) (3.7 7) (2.33) ( 1.19) 5. 09:45 – 333.06 119.5 2 0.0 033 164.0 6 39.79 0.0 035 491.88 0.71 0.0 180 246.97 15. 78 0.0006 10:00 hours (6.01) (4.41 ) ( 1. 05) (3.58 ) (1.73 ) ( 1. 16) (3.52) ( 0.07) (3.60) (2.2 9) (1.84) (0.24 ) 6. 10:00 – 177.62 84.70 0.0 075 27.14 28. 82 0. 0039 240.23 11. 49 0.0 165 42.28 3.7 7 0.0038 10:15 hours (3.20) (3.12 ) ( 2. 36) (0.59 ) ( 1.25) (1.27 ) (1.72) ( 1.20) (3.29) (0.3 9) ( 0.4 4) (1.52 ) 7. 10:15 – 65. 48 48.80 0.0 074 85. 27 88. 19 0. 0108 20. 64 28. 78 0.0 179 40.80 19. 85 0.0065 10:30 hours (1.18) (1.79 ) ( 2. 32) ( 1.86) ( 3.81) (3.52 ) ( 0.1 5) ( 3.00) (3.57) ( 0.38) ( 2.3 0) (2.58 ) 8. 10:30 – 146.42 24. 76 0.0 070 272.59 158.2 7 0. 0127 149.69 39. 72 0.0 231 148.5 3 50. 27 0.0139 10:45 hours ( 2.63) ( 0. 91) ( 2. 20) ( 5.94) ( 6.85) (4.18 ) ( 1.0 7) ( 4.15) (4.62) ( 1.38) ( 5.8 5) (5.54 ) continued
TABLE III (Continued) Indiv idual Day Trade rs In dividual Non day T raders Foreign Institutiona l Trad ers Futures Proprieta ry Firm Trade rs Time Interval Lim it Marke t Lim it Ord er Submis sion Ra tio (%) Lim it Marke t Limit Ord er Sub mission Ratio (%) Limit Marke t Lim it Order Submission Ra tio (%) Lim it Mar ket Limit Or der Su bmission Ratio (%) 9. 10:45 – 139.04 1.99 0.0 104 224.96 133.8 8 0. 0134 157.45 46. 67 0.0 246 69.53 31. 88 0.0115 11:00 hours ( 2.50) (0.07 ) ( 3. 25) ( 4.91) ( 5.80) (4.39 ) ( 1.1 2) ( 4.87) (4.90) ( 0.64) ( 3.7 1) (4.59 ) 10. 11: 00 – 231.09 53. 07 0.0 072 335.43 181.3 1 0. 0168 313.47 43. 35 0.0 183 223.5 7 50. 84 0.0132 11:15 hours ( 4.15) ( 1. 96) ( 2. 23) ( 7.30) ( 7.84) (5.51 ) ( 2.2 3) ( 4.52) (3.64) ( 2.07) ( 5.9 0) (5.25 ) 11. 11: 15 – 125.02 8.82 0.0 146 218.28 112.1 9 0. 0095 281.79 35. 50 0.0 200 93.56 24. 74 0.0092 11:30 hours ( 2.24) (0.32 ) ( 4. 55) ( 4.75) ( 4.84) (3.10 ) ( 2.0 0) ( 3.70) (3.99) ( 0.86) ( 2.8 7) (3.64 ) 12. 11: 30 – 24. 15 35.66 0.0 084 176.89 99. 71 0. 0089 127.11 32. 63 0.0 203 9.3 3 20. 99 0.0081 11:45 hours (0.43) (1.31 ) ( 2. 61) ( 3.85) ( 4.31) (2.91 ) ( 0.9 1) ( 3.40) (4.05) (0.0 9) ( 2.4 4) (3.24 ) 13. 11: 45 – 4.42 41.34 0.0 123 186.26 81. 67 0. 0023 15. 58 37. 59 0.0 214 36.41 21. 67 0.0108 12:00 hours ( 0.08) (1.52 ) ( 3. 84) ( 4.05) ( 3.52) (0.76 ) ( 0.1 1) ( 3.91) (4.25) ( 0.34) ( 2.5 1) (4.30 ) 14. 12: 00 – 66.85 13.64 0.0 074 304.69 114.7 3 0. 0050 82. 20 38. 12 0.0 234 97.33 28. 90 0.0134 12:15 hours ( 1.20) (0.50 ) ( 2. 29) ( 6.61) ( 4.94) (1.65 ) (0.58) ( 3.96) (4.66) ( 0.90) ( 3.3 4) (5.29 ) 15. 12: 15 – 16. 67 26.51 0.0 048 165.60 73. 01 0. 0066 142.17 26. 35 0.0 190 9.4 5 11. 88 0.0084 12:30 hours (0.30) (0.97 ) ( 1. 50) ( 3.59) ( 3.15) (2.17 ) (1.01) ( 2.74) (3.77) (0.0 9) ( 1.3 7) (3.31 ) 16. 12: 30 – 52.80 4. 79 0.0 026 213.02 113.4 9 0. 0081 30. 09 42. 21 0.0 265 12.47 31. 65 0.0106 12:45 hours ( 0.95) ( 0. 18) ( 0. 82) ( 4.63) ( 4.90) (2.66 ) (0.21) ( 4.39) (5.27) ( 0.11) ( 3.6 7) (4.21 ) 17. 12: 45 – 37. 86 6.68 0.0 026 95. 13 56. 68 0. 0008 123.99 37. 12 0.0 239 34.81 15. 28 0.0068 13:00 hours (0.68) (0.24 ) ( 0. 80) ( 2.06) ( 2.44) (0.24 ) (0.88) ( 3.85) (4.74) (0.3 2) ( 1.7 6) (2.67 ) continued
TABLE III (Continued) Indiv idual Day Trade rs In dividual Non day Tra ders Foreign Institutiona l Trad ers Futures Proprieta ry Firm Trade rs Time Interval Lim it Marke t Lim it Ord er Submis sion Ra tio (%) Lim it Marke t L imit Ord er Sub mission Ratio (%) Limit Marke t Lim it Order Submission Ra tio (%) Lim it Mar ket Limit Or der Su bmission Ratio (%) 18. 13: 00 – 97. 15 30.12 0.0 016 31.86 0.94 0.0 007 361.89 9.41 0.0 138 123.85 3. 79 0.0017 13:15 hours (1.74) (1.10 ) ( 0. 50) (0.69 ) (0.04 ) ( 0. 24) (2.56) (0.98 ) (2.74) (1.1 4) (0.44) (0.68 ) 19. 13: 15 – 13:30 hours 333.42 164.84 0.0 063 89. 07 57. 83 0. 0029 254.02 70.92 0.0051 236.2 3 33. 69 0.0081 ( 5.95) ( 6. 03) (1.96) ( 1.93) ( 2.48) (0.95 ) (1.80) (7.34 ) ( 1.0 1) ( 2.17) ( 3.8 9) (3.22 ) C 2,272. 27 807.0 0 0.7 555 2, 129.46 780.19 0. 7466 2,8 00.14 1, 74.91 0.9 066 2,615. 02 205.30 0.9168 (182 .53) (132.7 7) (1,055.3 8) (207.2 9) (150 .82) (1,093 .45) (89.15) (81.51 ) (80 7.97) (108 .11) (10 6.56) (1,630 .38) Observ ation 10,042 10, 042 10, 042 10, 042 10, 042 10, 042 10, 042 10, 042 10, 042 10,042 10, 042 10,042 Adj. R 2 0.0 90 0.082 0. 225 0.119 0.155 0.132 0.0 48 0.084 0.2 03 0.050 0.1 02 0.078 F ‐test 50. 88 46.07 146.51 68.72 93.27 77.14 26. 14 47.25 128.82 27.37 58. 28 43.39 Note . In this table, we present one ‐way analysis of variance model (see Equation (1) to estimate intraday submission patterns of limit order, market order, and limit order submission ra tio by all types of traders in the futures contract FITX from January 1, 2007 to December 31, 2008. The limit and market order are divided into four types of traders: indivi dual day traders, individual nonday traders, foreign institutional traders, and futures proprietary firm traders. The dependent variable is the mean of limit order sum, market order sum, or limit order submission ratio sum for each trader types during the 15 ‐minute intraday interval, which is regressed on the time ‐of ‐day dummy variables for each 15 ‐minute interval (i.e., 8:30 – 8:45 a.m. to 13:15 – 13:30 p.m.). The value of intercept C is daily average and is used as the basis for comparison. The t‐ statistic is reported in parentheses for each estimate. The preopen trading period for each trading day is denoted by 0. The preopen session is from 8:3 0 a.m. to 8:45 a.m. , , and indicate signi ficance at the 1%, 5%, and 10% levels, respectively.
The intercept is the daily average, which is used as the basis of comparison. The results show that during the preopening session (i.e., 8:30 a.m. to 8:45 a.m.), individual day traders and nonday traders actively submit limit orders whereas foreign institutional traders and futures
proprietaryfirms are relatively inactive in submitting limit orders. The second time interval
(9:00–9:15 a.m.) has the highest average number of order submissions for all type of traders.
Given the results in Table III, Figure 1 shows that the intraday average numbers of order
submissions for all trader types is V‐shaped for both market and limit orders. Our intraday
pattern of order submissions is very similar to the patterns reported by Biais et al. (1995) and Bae et al. (2003).
FIGURE 1
Intraday average numbers of limit and market order submission by four types of traders. The graph
depicts the average number of orders submitted during the 15‐minute intervals for each trading day for
the futures contract FITX from January 1, 2007 to December 31, 2008. The limit and market order are divided into four types of traders: individual day traders, individual nonday traders, foreign institutional
In Table III, we also show that the limit order submission ratio ranges in the regular trading period from 83.97% to 93.12% for foreign institutional traders and from 86.45% to
93.07% for futures proprietaryfirm traders. The limit order submission ratio of individual day
(noonday) traders is in the range from 67.53% to 79.7% (66.67–76.34%). These results
support Bloomfield et al. (2005) and Kaniel and Liu (2006) who find that informed traders use
more limit orders than market orders and do not support Harris (1998) who predicts that informed traders use relatively more market than limit orders.
Panel A of Figure 2 shows that the limit order submission ratio of institutional traders is
an inverted U‐shaped during whole sample period; that is, institutional traders use relatively
more market orders at the beginning and closing time intervals.14Thisfinding is expected, in
that institutional traders use more market orders to capture the value of private information in the early trading process and use relatively more market orders to close their positions as trading comes to an end. Limit order submission ratios of individual day and nonday traders
are somewhat L‐shaped with a sudden drop in the last two time intervals, suggesting that
individual traders (i.e., uninformed traders) use relatively more limit orders in the early trading and relatively more market orders in late trading. These results suggest that individual traders provide relatively greater liquidity in the early session and consume relatively greater liquidity toward the end of trading session. Panel B shows that limit order submission ratios of all types
traders during the prefinancial crisis are very similar to limit order submission ratios of all
trader types for whole sample period. Our data clearly show that limit orders are preferred to market orders for all types of traders. Relatively, some trader may submit more market orders during certain time of the day versus other time of the day (but not versus limit orders), but by and large, the order type used is predominantly limit orders.
In Table IV, we report regression results on intraday variation of the size of limit orders
and market orders submitted by all types of traders over a trading day.15Panels A and B of
Figure 3 show the time‐series patterns of the limit order size and market order size by trader
type, respectively. First, we find that limit orders submitted by individual day traders,
individual nonday traders, and foreign institutional traders are larger in size than their
corresponding market orders. These results affirm the results for all traders reported by Bae
et al. (2003). Futures proprietaryfirms however show the exact reverse pattern on submission
of limit order size versus market order size.
Second, wefind that the limit order size and market order size intraday patterns for both
foreign institutional traders and futures proprietaryfirms are clearly L‐shaped and that the
same patterns areflat for individual traders. The larger order sizes of institutional traders
suggest that they try to capture as much value as possible from their market information in the early stage of the trading process. These differences in intraday order size submission between institutional and individual traders are new to the limit order market literature. In general foreign institutional traders, compared to the other three trader types, use larger limit and market orders.
In sum, wefind that institutional traders use relatively more market orders in the early
stage of trading process and switch to relatively more limit orders as the trading process progresses. Individual traders submit relatively more limit orders in the early trading and use relatively more market orders as trading come to a close.
14Limit order submission ratio is defined as the ratio of the number of limit orders to the sum of limit and market orders during each 15‐minute interval.
15
Based on the regression results of Table 4, we can estimate the means of the size of limit and market orders submitted by trader types on a 15‐minute interval. For example, the mean of limit order size at 8:45–9:00 a.m. time interval submitted by individual day trader is equal to 2.2791 (intercept) 0.0347 (the coefficient of limit order size regression of individual day trader at 8:45–9:00 a.m.) ¼ 2.244.
5.2. Regression Analysis
We report the regression analysis of the influences of market conditions on liquidity provision
by trader types in Table V. To conserve space, we do not present the results of dummy
variables. For all traders (see column 2), the coefficient of Spreadt1is positive and highly
significant at the 1% level. This result confirms that when the spread is wide, traders place
FIGURE 2
The means of limit order submission ratios during the 15‐minute intervals of each trading day for the
futures contract FITX for full sample period and prefinancial crisis period in Panels (A) and (B),
respectively. The preopen session is from 8:30 a.m. to 8:45 a.m. Submission ratio is defined as the ratio of
the number of limit order to the sum of his limit and market orders during each 15‐minute interval. The
four types of traders are individual day traders, individual nonday traders, foreign institutional traders,
TABLE IV Regression Analysis of Limit and Market Order Size Regression by Trader Type Indiv idual Day Tra der Individ ual Nonday Trade r For eign Institutiona l Tra ders Fu tures Proprie tary Firm Trade rs Tim e Inte rval Lim it Mar ket Limit Marke t Lim it Marke t Lim it Mar ket 0. 08: 30 – 08:45 hour s 0. 1170 0.0 084 0. 1954 0.1300 5.7 214 24. 6161 3.3099 3.1 426 ( 12.40) (0.49) (19.87 ) (9.4 2) (91.90) (71.05 ) (190 .26) (29.04) 1. 08: 45 – 09:00 hour s 0. 0347 0.1662 0.0 902 0. 1687 0.0 757 1. 7080 0.2423 0.1 143 ( 5.25) ( 22.77) ( 14. 01) ( 23.80) (6.26) (18.96 ) (49.4 3) (3.09) 2. 09: 00 – 09:15 hour s 0.1385 0.0 305 0.0 895 0. 0740 0.3852 0.1 752 0. 0528 0.3530 (22.6 6) (5.08) ( 12. 96) ( 9.76) ( 39.84) ( 2. 36) ( 12.22) ( 10.49) 3. 09: 15 – 09:30 hour s 0.0402 0.0 066 0.0 970 0. 0536 0.3218 0.8 310 0. 1208 0.1510 (6.5 7) (1.07) ( 13. 57) ( 6.48) ( 33.12) ( 10. 38) ( 27.30) ( 3.9 8) 4. 09: 30 – 09:45 hour s 0.0156 0.0163 0.0 675 0. 0571 0.2670 0.7 203 0. 1442 0.1114 (2.4 5) ( 2.5 3) ( 9. 01) ( 6.36) ( 26.42) ( 8. 19) ( 31.45) ( 2.6 8) 5. 09: 45 – 10:00 hour s 0.0154 0.0115 0.0 226 0. 0205 0.3376 0.9 115 0. 1676 0.0289 (2.3 9) ( 1.7 8) ( 2. 95) ( 2.25) ( 32.61) ( 10. 06) ( 35.79) ( 0.6 8) 6. 10: 00 – 10:15 hour s 0. 0345 0.0182 0. 0031 0. 0062 0.3094 0.9 704 0. 2368 0.0522 ( 5.24) ( 2.7 8) (0.40 ) ( 0.65) ( 28.78) ( 10. 44) ( 49.47) ( 1.1 8) 7. 10: 15 – 10:30 hour s 0. 0635 0.0367 0.0 056 0. 0188 0.3653 1.3 504 0. 2497 0.2025 ( 9.49) ( 5.5 2) ( 0. 69) ( 1.90) ( 32.77) ( 14. 21) ( 51.41) ( 4.4 7) 8. 10: 30 – 10:45 hour s 0. 1311 0.0593 0.0 579 0. 0123 0.5853 1.3 928 0. 2643 0.2525 ( 19.03) ( 8.5 9) ( 6. 88) ( 1.19) ( 52.52) ( 14. 20) ( 53.62) ( 5.1 7) 9. 10: 45 – 11:00 hour s 0. 1473 0.0304 0.0 106 0. 0077 0.5099 1.6 362 0. 2198 0.0565 ( 21.51) ( 4.4 4) ( 1. 27) ( 0.76) ( 45.37) ( 16. 61) ( 44.90) ( 1.1 9) 10. 11:00 – 11: 15 hour s 0. 1186 0.0236 0. 0039 0.0086 0.5268 1.2 922 0. 2321 0.1532 ( 16.81) ( 3.3 2) (0.45 ) (0.8 1) ( 45.62) ( 12. 88) ( 46.04) ( 3.0 8) 11. 11:15 – 11: 30 hour s 0. 0905 0.0025 0. 0346 0.0374 0.5282 1.6 840 0. 2288 0.1734 ( 13.05) ( 0.3 6) (4.09 ) (3.6 7) ( 46.02) ( 17. 82) ( 46.42) ( 3.8 4) 12. 11:30 – 11: 45 hour s 0. 0232 0.0 096 0. 0164 0.0419 0.4300 1.5 553 0. 2023 0.7042 ( 3.41) (1.42) (1.96 ) (4.1 5) ( 38.15) ( 16. 45) ( 41.67) ( 17.25) 13. 11:45 – 12: 00 hour s 0. 0313 0.0 271 0.0 252 0.0482 0.3033 1.7 834 0. 2335 0.4007 ( 4.57) (3.98) ( 3. 03) (4.8 2) ( 27.03) ( 18. 87) ( 47.89) ( 9.5 5) continued
TABLE IV (Continued) Indiv idual Day Tra der Individ ual Nonday Trade r For eign Institutiona l Tra ders Fu tures Proprie tary Firm Trade rs Tim e Inte rval Lim it Mar ket Limit Marke t Lim it Marke t Lim it Mar ket 14. 12:00 – 12: 15 hour s 0.0078 0.0 490 0.0 618 0.0364 0.2338 1.8 204 0. 2815 0.1933 (1.1 1) (7.05) ( 7. 25) (3.5 6) ( 21.01) ( 19. 36) ( 57.56) ( 4.1 5) 15. 12:15 – 12: 30 hour s 0.0489 0.0 459 0. 0179 0.0411 0.1981 1.9 944 0. 2556 0.0222 (7.0 3) (6.65) (2.14 ) (4.1 4) ( 17.92) ( 22. 33) ( 53.00) ( 0.4 9) 16. 12:30 – 12: 45 hour s 0.0597 0.0 491 0. 0167 0.0149 0.2677 2.0 648 0. 1987 0.1621 (8.4 5) (6.99) (1.97 ) (1.4 7) ( 23.94) ( 22. 10) ( 40.66) ( 3.4 4) 17. 12:45 – 13: 00 hour s 0.0915 0.0 609 0. 0297 0.0374 0.2730 2.1 274 0. 2054 0.0548 (13.1 1) (8.67) (3.61 ) (3.8 2) ( 24.73) ( 23. 30) ( 42.33) ( 1.2 0) 18. 13:00 – 13: 15 hour s 0.1281 0.0 607 0. 0412 0.0289 0.1729 2.3 978 0. 2105 0.0644 (18.4 2) (8.77) (5.13 ) (3.0 8) ( 16.06) ( 30. 74) ( 44.04) ( 1.4 8) 19. 13:15 – 13: 30 hour s 0.1063 0.0 076 0. 0516 0. 0033 0.0 245 1.9 318 0. 2189 0.0391 (13.9 6) (0.98) (6.23 ) ( 0.34) (2.20) ( 26. 84) ( 42.95) ( 0.8 1) C 2.2791 1.9 928 2. 4798 2.3579 5.4 160 6. 8632 2.6302 4.2 399 (1,455 .77) (1,16 5.16) (1,367.9 8) (1,083 .70) (1,39 5.77) (265.6 5) (1,944 .13) (38 9.84) Obs ervation 10, 039,866 4,078, 962 8,6 90,712 3, 374,647 5,489, 755 305,87 7 10, 624,149 511,533 Adj. R 2 0.0003 0.0 002 0. 0002 0.0003 0.0 026 0.0 239 0.0048 0.0023 F ‐test 157.68 48. 71 70. 26 44.43 710.04 376.1 8 2544 .45 59. 72 Note . In this table, we present one analysis of variance model (see Equation (1) to estimate intraday variance pattern of limit order size and market order s ize submitted by trader types over a trading day in the futures contract FITX from January 1, 2007 to December 31, 2008. The trader types are classi fied by individual day traders, individual nonday traders, foreign institutional traders, and proprietary firm traders. The dependent variable is the tick ‐by ‐tick limit orders by trader types, which is regressed on the time ‐of ‐day dummy variables for each 15 ‐minute interval (i.e., 08:30 – 08:45 a.m. to 13:15 – 13:30 p.m.). The value of intercept C represents daily average and is used as the basis for comparison. The t‐ statistic is reported in parentheses for each estimate. The preopen trading period from 8:30 a.m. to 8:45 a.m. for each trading day is denoted by 0. , , and indicate signi ficance at the 1%, 5%, and 10% levels, respectively.
more limit orders either because submission of market orders is costly or because compensation is higher if limited orders are executed (e.g., Bae et al., 2003; Chung
et al., 1999). The coefficient of Transitory_Volatility lagged one period has a positive sign and
the coefficient of the Informational_Volatility has a negative sign; both of these coefficients
are highly significant at the 1% level. Our empirical evidence is consistent with the prediction
Handa and Schwartz’s (1996) theoretical model but does not support the implications of the
model proposed by Foucault (1999).
FIGURE 3
Thesefigures plot the means of order sizes of limit and market order by trader type during the 15‐minute
intervals of each trading day for the futures contract FITX from January 1, 2007 to December 31, 2008. The preopen session is from 8:30 a.m. to 8:45 a.m. The four trader types are individual day traders,
In Handa and Schwartz’s (1996) model, traders lose when they execute orders with
informed traders due to adverse selection risk and profit when they execute limit orders with
uninformed (liquidity) traders. Thus, traders will submit more limit orders than market orders when the expected (one period lagged) transitory volatility increases, and they will submit fewer limit orders when expected informational volatility increases. Bae et al. (2003) also report that traders will increase their submission of limit orders when transitory volatility is
TABLE V
Regression Analysis on the Influences of Market Conditions on Liquidity Provision by Trader
Types
Time Interval All
Individual Day Trader Individual Nonday Trader Foreign Institutional Traders Futures Proprietary Firm Traders Spreadt1 16.76 0.03 0.34 8.20 5.12 (4.87) (0.06) (1.04) (5.38) (3.71) Transitory_Volatilityt1 13.99 2.34 1.66 4.36 5.94 (4.70) (5.46) (4.77) (3.64) (4.41) Informational_Volatilityt1 60.13 8.34 5.52 22.27 19.77 (2.96) (3.78) (3.44) (2.78) (3.04) Same_Side_Deptht1 875.05 104.67 87.19 353.73 256.46 (8.46) (7.46) (7.12) (8.51) (7.18) Other_Side_Deptht1 630.84 85.39 86.99 211.12 187.72 (7.19) (6.64) (7.84) (6.00) (6.46) Limit_Sizet 1,355.56 – – – – (5.89) Dayt – 233.54 – – – (10.77) Non_Dayt – – 454.38 – – (13.36) Foreignt – – – 911.26 – (157.74) Proprietaryt – – – – 143.69 (2.41) Observation 9,204 9,204 9,204 9,204 9,204 Adj. R2 0.13 0.19 0.28 0.37 0.09 F‐test 55.49 87.94 147.57 215.71 36.13
Note. In this table, we present the regression analysis results that examine whether the lagged spread, lagged volatility, lagged same side depth, lagged other side depth, or limit order size by the trader‐type variables affect limit orders in the futures contract FITX. The regression analysis model is specified as
NLMt¼ a þ b1Spr eadt−1þ b2T r ansit or y V ol at il i tyt−1þ b3Inf or mat i onal V ol at il i t yt−1þ b4Same Si d e Deptht−1 þ b5Other Si d e Dept ht−1þ b6Li mi t Si zetþ
X19 j¼1
b7;jDjþ et:
The dependent variable NLMtis equal to the sum of limit orders minus market orders and marketable limit orders for each trader
types during each 15‐minute interval. The trader types are classified as individual day traders, individual nonday traders, foreign institutional traders, and proprietaryfirm traders. Spreadt1is the average of dollar quote spread during time interval t 1;
Transitory_Volatilityt1denotes transitory volatility lagged one period; Informational_Volatilityt1represents informational volatility
lagged one period; Same_Side_Deptht1(Other_Side_Deptht1) is measured as the average number of limit orders at the best bid
(ask) just prior to a buy order's submission, and as the average number of limit order at the ask (bid) just prior to a sell order's submission during time interval t 1; Limit_Sizetis the average of limit orders during time interval t for all traders, individual day
traders, individual nonday traders, foreign institutional traders, and proprietaryfirm traders; Djis the time‐of‐day dummy variables
for each 15‐minute interval (i.e., 8:45–9:00 a.m. to 13:15–13:30 p.m.). The specification of Djis discussed in Equation (1). To save
the space, we do not report the dummy variables results. The t‐statistics are reported in parentheses.
expected to increase, but the impact of informational volatility on submission of limit order is
inconclusive.16
In Table V, we also show that the parameter of same side depth at best bid (ask) lagged
one period has a negative sign and is significant at the 1% level, and the parameter of opposite
side depth lagged one period has a positive sign and is also significant at 1% level. As we expect,
this result confirms that all traders will submit fewer limit orders when the state of the same
side order book is thicker and more limit orders when the book is thinner. The impact of the state of the opposite side order book on limit order submissions by all traders has exactly the
reverse effect of the state of the same side order book. This result confirms the theoretical
prediction of Parlour (1998) and is also consistent with the experimental results obtained by
Bloomfield et al. (2005). The positive and significant coefficient of the limit order size
confirms that traders prefer to use more limit orders to minimize their trading cost when order
sizes are relatively large.
In columns 3–6 of Table V, we report the regression results of the influence of market
conditions on the liquidity provision by each type of traders. We summarize their differing
provision of liquidity given changes in market conditions as follows. First, the coefficients of
the spread lagged one period of individual day traders and individual nonday traders are
positive but not significant. The coefficients of Spreadt1of institutional traders (i.e., both
foreign institutional traders and futures proprietary firms) have positive signs and are
significant at the 1% level or greater. The insignificant impact of change in spreads on the
decision of individual day traders may be due to individual traders typically engaging in quick
turn‐around trading.
Second, the coefficient of the limit order size of for futures proprietary firms is negative
and significant at the 1% level. One possible explanation is that because futures proprietary
firms often have access to order flow information, they may use market orders to capture
the value of short‐lived information.17 Third, the coefficients of Transitory_Volatilityt1,
Informational_Volatilityt1, Same_Side_Deptht1, and Opposite_Side_Deptht1 of all four
trader types have the same expected signs and are significant at the 1% level. However, their
responses to the net submission of limit orders differ due to changes in these market variables. Based on empirical results of Table 5, we estimate the elasticity of the limit order submission with respect to market condition variables and limit order size variable; these results are reported in Table VI.
In general, in Table V, we show that institutional traders are more elastic to changes
in these four lagged one‐period variables (i.e., Transitory_Volatilityt1, Informational_
Volatilityt1Spreadt1, Same_Side_Deptht1, and Other_Side_Deptht1) than individual day
and nonday traders. For example, the elasticity of spreads lagged one period of foreign
institutional traders and futures proprietaryfirms is 0.99 and 0.69, respectively. The elasticity
of spreads lagged one period is less than 0.01 and 0.08 for individual‐day and nonday traders,
respectively. The elasticity of Informational_Volatility is0.17 for both foreign institutional
traders and futures proprietaryfirms while the same elasticity for individual day and nonday
traders is0.12 and 0.008, respectively. As expected, the elasticity of these market variables
16
We use Transitory_Volatility and Informational_Volatility as explanatory variables in the regression model, whereas Bae et al. (2003) use dummy variables to denote four combination cases of high and low transitory versus informational volatility cases.
17We interviewed several traders from futures proprietary futuresfirms. They report that they often hire a large number of traders to monitor orderflow from the order book and use relative large market order size to capture the instant trading opportunity. Traders from futures proprietaryfirms often use relatively larger market order size than limit order size to implement their momentum trading strategy. The anonymous referee suggests the difference in the submission of limit and market order size between foreign institutional investor and futures proprietaryfirms may be due to difference in information in nature. Further research will be required to resolve this issue.
