Int. Fin. Markets, Inst.

and Money 30 (2014) 205–219

Contents lists available at ScienceDirect

Journal of International Financial

Markets, Institutions & Money
j o ur na l ho me pa ge : w w w . e l s e v i e r . c o m / l o c a t e / i n t f i n

Anonymity and the Information Content of the

Limit Order Book
Huu Nhan Duong a,∗, Petko S. Kalev b,1
a
Department of Banking and Finance, Faculty of Business and Economics, 900 Dandenong Road,
Caulfield East 3145, Australia
b
School of Commerce, Division of Business, University of South Australia, City West Campus, GPO Box 2471,
Adelaide, South Australia 5001, Australia

a r t i c l e i n f o a b s t r a c t

Article history: We investigate the effect of broker anonymity on the information

Received 25 January 2013 content of the limit order book on the Australian Stock Exchange.
Accepted 19 February 2014 We argue that the move to anonymity has stronger impact on insti-
Available online 13 March 2014
tutional than individual investors. We document that anonymity
increases the informativeness of institutional limit orders about
JEL classification: future volatility but has little effect on individual limit orders. Our
C35
findings imply that anonymity allows institutional investors to dis-
G15
play informative orders in the limit order book. Prices, however,
G25
G29 adjust less sufficiently to institutional investors’ order flow. Over-
all, anonymity may improve liquidity but it comes at the cost of
Keywords: reduced price efficiency.
Anonymity © 2014 Elsevier B.V. All rights reserved.
Institutional and individual investors
Order book slope
Limit order book information
Price volatility

1. Introduction

In this paper we examine the information content of the limit order book and the change in this
information after the removal of broker identifiers (IDs) on the Australian Stock Exchange (ASX). In
doing so, we address three research questions. First, is the limit order book informative about future

∗ Corresponding author. Tel.: +61 39903 2032.

E-mail addresses: huu.duong@monash.edu (H.N. Duong), Petko.Kalev@unisa.edu.au (P.S. Kalev).
1
Tel.: +61 88302 7122.

http://dx.doi.org/10.1016/j.intfin.2014.02.004
1042-4431/© 2014 Elsevier B.V. All rights reserved.
206 H.N. Duong, P.S. Kalev / Int. Fin. Markets, Inst. and Money 30 (2014) 205–219

price volatility? Second, are institutional investors’ limit orders more informative than individual limit
orders about future price volatility? Finally, does the removal of broker IDs have any impact on the
informativeness of the limit order book? If it does, are institutional or individual limit orders more
affected by this change?
We use the order book slope to reflect the information content of the limit order book. Building
on Næs and Skjeltorp (2006), the order book slope can be viewed as an average of various “elasticity”
measures, each describing how the quantity supplied in the limit order book changes as we move from
one price level to the next in the limit order book. Thus, the order book slope provides a summary of
the price and depth dimensions of liquidity provision in the limit order book.
Investigating the informativeness of the limit order book slope about price volatility is important
because volatility is essential for pricing options, determining order choices,2 and making optimal
investment decisions.3 Furthermore, an understanding of the relation between the limit order book
slope and future price volatility can provide evidence regarding the order choices of informed traders,
and thus yields insights about the process through which information is incorporated into prices. Our
investigation of the effect of the removal of broker IDs on the information content of the limit order
book is also motivated by the recent trend of removing broker IDs in equity markets.4 This trend
has taken place despite the conventional wisdom that transparency improves market quality and the
finding of Swan and Westerholm (2006) that the disclosure of broker IDs lowers trading cost and
intraday volatility.5
We investigate the informativeness of the slope of the order book for the constituent stocks of the
S&P/ASX 100 index. Our examination is based on a natural experiment, the switch to an anonymous
trading system by the ASX on November 28, 2005. Prior to this date, the ASX disseminated to the broker
community, in real time, the broker ID associated with every order in the central limit order book.
Although the release of broker IDs information to third parties was strictly prohibited, institutional
clients and very high net worth individuals often requested and received this information from their
brokers (ASX, 2005). This created an informational advantage for those investors using full advisory
broking services over those making their own trading decisions (ASX, 2003). Since November 28,
2005, the ASX has ceased providing broker IDs to the broker community in real time and provided
only market share information at the end of the trading day and releases the full trading history with
broker IDs after a delay of three days.6
Prior studies often find an improvement in liquidity after the move to an anonymous system (see,
for example, Comerton-Forde et al., 2005; Foucault et al., 2007; Frino et al., 2008; Comerton-Forde and
Tang, 2009). We add to this literature by analyzing how a reduction in market transparency affects the
information content of the limit order book. Our study builds on the work of Foucault et al. (2007), who
develop a theoretical model for limit order markets to explain the changing behavior of informed and
uninformed traders after the removal of broker IDs. The authors argue that in a transparent market,
uninformed investors infer information about future price movements from the limit orders submitted
by informed traders. They try to front-run the informed traders to benefit from the information by
setting more competitive quotes than those posted by the informed traders. In response, informed
traders sometimes adopt “bluffing” strategies by posting non-aggressive orders and setting wider
spreads than appropriate.

2
See, for example, Foucault (1999), Bae et al. (2003), Ranaldo (2004), and Wald and Horrigan (2005) for a discussion on the
effect of volatility on investors’ order submission strategies.
3
Fleming et al. (2001, 2003) demonstrate the substantial value of volatility timing in the context of investment decisions.
Fleming et al. (2003) suggest that an investor implementing volatility-timing strategies would be willing to pay on the order of
50–200 basis points per year to capture the incremental gains generated by the realized volatility-based estimator.
4
Several equity markets, such as the Euronext Paris, the Tokyo Stock Exchange, the ASX, and the New Zealand Exchange,
have adopted this change.
5
Market transparency is a broad issue. In this paper, we focus on the identities of the brokers submitting orders into the
limit order book. Other papers focus on another aspect of pre-trade transparency, the disclosure of additional levels of the limit
order book. For evidence on this line of research, see, among others, Boehmer et al. (2005), Madhavan et al. (2005), and Eom
et al. (2007).
6
One can argue that the disclosure/removal of the IDs of brokers’ customers has a more meaningful impact on the market.
This information, however, has never been publicly released on the ASX. Linnainmaa and Saar (2012) further show that broker
identity is a powerful signal about the identity of the trader who initiates trades.
 H.N. Duong, P.S. Kalev / Int. Fin. Markets, Inst. and Money 30 (2014) 205–219 207

An anonymous trading system eliminates the traders’ ability to distinguish informed traders’ orders
from those of uninformed traders. Therefore, in an anonymous trading system, uninformed traders
submit limit orders based on their belief about the identity of the traders with orders in the limit order
book. In this case, if the participation rate of informed traders or the likelihood that existing limit orders
are submitted by informed traders is low (high), uninformed traders will be more (less) aggressive, and
improve on the already posted orders more (less) often. Therefore, if the participation rate of informed
traders is low, a move to anonymity will decrease the bid–ask spread and its informativeness about
future volatility because uninformed traders are more aggressive and thereby set the best quotes more
frequently in the anonymous system. Supporting the prediction of their model, Foucault et al. (2007)
observe a reduction in the bid–ask spread of the constituent stocks of the CAC 40 index after the
move to anonymity in the Euronext Paris. The authors also document that the strength of the relation
between price volatility and the lagged bid–ask spread is lower after the move to anonymity.
Rindi (2008) argues that in a transparent market, uninformed traders can obtain information at no
cost by observing the trading strategies of informed traders. Therefore, the incentive to acquire infor-
mation diminishes in a transparent market. Given that informed traders face lower adverse selection
and thus are more efficient liquidity suppliers, disclosing traders’ identities can reduce the number of
informed traders and, consequently, the amount of liquidity offered to the market. Duong and Kalev
(2013) show that anonymity increases the willingness to submit limit orders for both institutional
and individual investors. The authors conclude that anonymity has a positive effect on the liquidity
supply to the central limit order book.
However, Linnainmaa and Saar (2012) document that the benefits of improvement in liquidity
may come at the cost of reduced price efficiency. They show that market uses broker IDs information
to update prices. A trading system without broker IDs would allow informed investors to hide more
effectively. Thus, price impact would decrease but prices would also react less to order flow infor-
mation (lower price efficiency). The market may become less competitive, with informed investors
realizing greater profits from their informational advantage at the expense of uninformed investors.
We document supportive evidence for the informativeness of the order book slope about future
price volatility in the majority of the constituent stocks of the S&P/ASX 100 index over the period
between July 1, 2005, and June 30, 2006. The informativeness of the order book slope is observed for
both overall volatility and the permanent component of volatility. These findings highlight that the
limit order book is a channel for price volatility information. We also find that the slope of the limit
order book on the buy side is more informative than that on the sell side and institutional limit orders
are more informative than individual limit orders about the future permanent component of volatility.
Similar to Foucault et al. (2007), we show that the removal of broker IDs on the ASX has a significant
impact on the predictive power of the order book slope on future volatility. We next investigate the
changes in the informativeness of institutional and individual limit orders after the removal of broker
IDs. We expect minimal change in the informativeness of individual limit orders since individuals do
not have access to broker IDs information in either transparent or anonymous market. For institutional
investors, prior studies often document that institutional investors are better informed than individual
investors (see, for example, Szewczyk et al., 1992; Alangar et al., 1999; Dennis and Weston, 2001; and
Chakravarty, 2001).7 The anonymous market allows institutions to hide their orders more effectively.
Thus, they have less incentive to submit “bluffing” limit orders (Foucault et al., 2007), and prices
should adjust less efficiently to their order flow (Linnainmaa and Saar, 2012). We therefore expect
institutional limit orders to be more informative after the removal of broker IDs. Our findings confirm
these predictions: Institutional limit orders become more informative after the change in market
transparency, while the impact of anonymity on individual limit orders is minimal.
Our study extends the current literature in several ways. First, we contribute to the literature by
analyzing how a reduction in market transparency affects the information content of the limit order
book. Foucault et al. (2007) present evidence of the declining informativeness of the bid–ask spread,
which reflects the information contained in the first level of the limit order book. We differ from these

7
Duong and Kalev (2013) also show that institutional orders have larger price impacts than individual orders in the Australian
market.
208 H.N. Duong, P.S. Kalev / Int. Fin. Markets, Inst. and Money 30 (2014) 205–219

authors by examining the effect of anonymity on the informativeness of the limit order book slope,
which incorporates the information beyond the best quotes. The importance of the limit order book
beyond the best quotes, as demonstrated in Cao et al. (2008, 2009) and Pascual and Veredas (2010),
provides further support for our examination.
Second, our analysis on the informativeness of the limit order book slope about future price volatil-
ity complements prior studies on the information content of the limit order book (see, for example,
Ahn et al., 2001; Harris and Panchapagesan, 2005; Cao et al., 2009; Kalay et al., 2009; Pascual and
Veredas, 2010). Given that the order book slope describes how the quantity supplied in the order book
changes as a function of prices, our investigation extends prior work that examines the informative-
ness of the limit order book about volatility based on the quantity supplied or the order imbalance
between the demand and supply sides of the order book (see, for example, Ahn et al., 2001; Pascual
and Veredas, 2010).
Third, we analyze the impact of anonymity on the information content of the slope of the order
book based on either institutional or individual limit orders. To the best of our knowledge, this is the
first study to investigate the impact of anonymity on the information content of the order submissions
of two classes of investors with different levels of information. Our investigation complements Rindi
(2008), Linnainmaa and Saar (2012), and Duong and Kalev (2013) by showing that institutional limit
orders become more informative about future volatility after the move to anonymity. Our finding
highlights that the benefits anonymity brings, in the form of improvements in liquidity, may come at
the cost of reduced informational efficiency.
The remainder of the paper is organized as follows. Section 2 discusses details about the data and
variables measurements. Section 3 discusses the results and their implications and Section 4 concludes
the paper.

2. Data and variables measurements

2.1. Data description

Our samples include one year of a proprietary order book dataset from July 1, 2005, to June 30, 2006,
for companies making up the S&P/ASX 100 index. The proprietary order book dataset records informa-
tion on each order, including the order type (order submission, order revision, order cancelation, and
order execution), date and time to the nearest 100th of a second, stock code, order price, order volume,
and order direction (buy or sell order). Each new order is assigned a unique identification number so
that we can track it from its initial submission through to any revision, cancelation, or execution. A
unique feature of this dataset is the provision of the confidential dummy variable indicating whether
the order is submitted by an institutional or an individual investor. This confidential classification of
institutional and individual orders is released by the ASX and provided to us via SIRCA.8
We investigate the informativeness of the limit order book based on a 30-minute interval. Since the
ASX’s staggered opening procedure takes up to 10 minutes to complete, the data for the first 10 minutes
of each day are excluded from our sample to avoid any potential bias.9 Based on the proprietary order
book dataset, we reconstruct the limit order book at the end of every 30-minute interval. Finally, we
examine only stocks that have not been merged or acquired by other companies and those whose data
are available for the entire sample period. The final sample includes 90 stocks.

2.2. Return series

Return series are constructed based on the mid-quote prices at the end of every 30-minute interval.
As suggested by Roll (1984), we use mid-quote prices instead of transacted prices because they reduce
the measurement errors due to bid–ask bounce, which can result in substantial spurious volatility. We
calculate the return for each interval as the difference of the natural logarithm of the mid-quote prices

8
For a more detailed description and analysis on this dataset, see Duong et al. (2009).
9
This practice is also used in other studies on the ASX (see, for example, Kalev et al., 2004 and Duong et al., 2009).
 H.N. Duong, P.S. Kalev / Int. Fin. Markets, Inst. and Money 30 (2014) 205–219 209

at the end of the current interval and those at the end of the previous interval. Consistent with Foucault
et al. (2007), overnight returns are excluded. Therefore, the returns for the first interval in each trading
day are calculated as the difference of the natural logarithms of the mid-quote prices at the end of the
interval and those at the beginning of the interval.
Given that intraday patterns can severely corrupt traditional volatility models based on raw (unad-
justed) high-frequency series (see, for example, Andersen and Bollerslev, 1997b; Martens et al., 2002;
Andersen et al., 2003), we follow the method of Andersen et al. (2003) in constructing our seasonally
adjusted intraday returns. More specifically, for each intraday interval, we compute the “seasonal fac-
tor” by calculating the interval’s squared return for each day and then averaging the squared returns
over the sample period. The seasonally adjusted intraday returns are then calculated by dividing the
unadjusted intraday returns for each intraday interval by the square root of the “seasonal factor” for
that interval.

2.3. Slope of the order book

Following Næs and Skjeltorp (2006), we measure the order book slope as the average of the order
book slope for the demand (bid) and supply (ask) sides. The order book slope for the bid side for firm
i in interval t (DEi,t ) is
 
1 vB1  vB /vB − 1
NB −1
DEi,t =  +  +1  , (1)
NB pB /p0 − 1 pB /pB − 1
1 =1 +1

Similarly, the order book slope for the ask side (SEi,t ) is defined as
 
1 vA1 
NA −1 A
v+1 /vA − 1
SEi,t = + , (2)
NA pA1 /p0 − 1 pA /pA − 1
=1 +1

where NB and NA are the total number of bid and ask prices (tick levels) containing orders, respectively;
 denotes the tick levels, with  = 1 representing the best ask (bid) quote; p0 is the best bid–ask mid-
point; and vA and vB are the natural logarithms of the accumulated total share volume at the price level
 (p ) on the ask and bid sides, respectively. In other words, vA (vB ) is the natural logarithm of the total
share volume supplied (demanded) at p or lower (higher). At the end of each 30-minute interval, we
use the 10 best bid and ask quotes together with the share volume at these quotes for the calculation
of the order book slope for that particular interval. We also remove all undisclosed (hidden) orders in
our calculation of the limit order book slope.10 Finally, similar to Næs and Skjeltorp (2006), we divide
the slope measure by 100 to scale the parameter estimates in our econometric models.

3. Findings and discussions

3.1. Descriptive statistics

Table 1 provides summary statistics for the 90 stocks investigated in the current study. The results
are obtained by averaging across 30-minute intervals for each stock and then averaging across stocks.
We observe an increase in the overall order book slope and the order book slope on the demand
(buy) and supply (sell) sides after the removal of broker IDs on the ASX. This finding indicates that
limit orders are more concentrated and placed closer to the best quotes after the move to anonymity.

10
The ASX allows traders to hide the total size of their limit orders by submitting undisclosed limit orders. The total value
of these orders must exceed a disclosure threshold. This threshold was initially set at $10,000. It was increased to $25,000
in October 1994 and further increased to $100,000 in October 1996 (Aitken et al., 2001). In our sample period, the threshold
was $200,000 (Lepone and Mistry, 2011). Undisclosed limit orders do not lose time priority in the order book when they are
executed and exist as disclosed limit orders when their size falls below $200,000. The current threshold for undisclosed limit
orders is $500,000 (ASX, 2013). Since June 2010, traders can also hide the size of their orders by using iceberg orders. See ASX
(2013) for a detailed description of this type of order.
210 H.N. Duong, P.S. Kalev / Int. Fin. Markets, Inst. and Money 30 (2014) 205–219

Table 1
Descriptive statistics.

Whole sample Pre-anonymity Post-anonymity Pre–post

Slope 15.2145 14.3764 15.7860 −1.4096***

Buyslope 15.1650 14.3091 15.7487 −1.4396***
Sellslope 15.2640 14.4438 15.8233 −1.3795***
Institutional slope 52.7625 52.7266 52.7870 −0.0604
Individual slope 17.2212 16.0449 18.0232 −1.9783***
NT 56.1148 46.0714 62.9744 −16.9030***
ATS 3345.9470 3710.3489 3094.1778 616.1711***
Spread 0.2131 0.2188 0.2092 0.0096***

This table presents summary statistics for the 90 stocks investigated in this study. The sample period is from July 1, 2005 to June
30, 2006, excluding four weeks before and after the changes in transparency on November 28, 2005. Slope is the slope of the
limit order book based on the 10 best quotes; Buyslope and Sellslope are the slopes of the limit order book on the buy (demand)
and sell (supply) sides, respectively; Institutional slope and Individual slope are the slopes of the limit order book based on orders
submitted by institutional and individual investors, respectively; NT and ATS are the number of trades and average trade size for
each 30-minute interval, respectively; Spread is the bid–ask spread, measured as the percentage of the difference between the
best ask and bid quotes over the bid–ask mid-point. The results are obtained by averaging over all 30-minute intervals for each
stock and then averaging across stocks. The column heading “pre–post” indicates the difference between the pre-anonymity
and post-anonymity periods.
***
Significance at the 1% level.

Consistent with Foucault et al. (2007) and Comerton-Forde and Tang (2009), we find a statistically
significant decline in the bid–ask spread following the move to an anonymous market. Investors trade
more often, on average, after the move to anonymity but they decrease the size of their transactions.
From the overall limit order book, we also create two “smaller” limit order books: the institutional
limit order book, which contains only institutional limit orders, and the individual limit order book,
which consists of only individual limit orders. We calculate the slopes of the institutional and indi-
vidual limit order books using Eqs. (1) and (2). The slope of the institutional limit order book is called
Institutional Slope and the slope of the individual limit order book is referred to as Individual Slope. We
find that Institutional Slope is about three times larger than Individual Slope. Consistent with Aitken
et al. (2007), this finding implies that institutional limit orders are placed closer than individual limit
orders to the best quotes. The variable Individual Slope increases after the move to anonymity but we
observe no significant change in Institutional Slope.

3.2. Information content of the order book slope

Since Generalized Autoregressive Conditional Heteroskedasticity (GARCH)-type models are widely

used in studies that deal with volatility modeling (Bollerslev et al., 1992), our analysis of the informa-
tion content of the order book slope is based on estimating the following first-order autoregressive
GARCH(1,1) (AR(1)-GARCH(1,1)) model for each firm in our sample11 :

rt =  + rt−1 + εt ,

t2 = ω + ˛ε2t−1 + ˇt−1

2
+ ı0 Dpost + (ı1 + ı2 Dpost )SLOPEt−1 + ϕ1 NTt−1 + ϕ2 ATSt−1 , (3)

where rt is the seasonally adjusted stock return at interval t,  is a constant, the εt are the serially
uncorrelated errors (innovations) of stock returns with mean zero, t2 is the conditional variance of εt ,
SLOPEt−1 is the limit order book slope at the end of interval t − 1, and Dpost is a dummy variable that takes
on the value of one for the anonymous period (from November 28, 2005, onward), and zero otherwise.
Consistent with Jones et al. (1994), the average trade size, ATSt−1 , and the total number of trades, NTt−1 ,
for the (t − 1)th interval are included as control variables for price volatility. We expect the order book
slope to be informative about future volatility (i.e., ı1 to be negative and significant). We include an
interaction term between the lagged value of the order book slope and the dummy variable Dpost for

11
Bollerslev (1987) suggests that the GARCH(1,1) model is the simplest model that captures the dynamics of volatility well.
 H.N. Duong, P.S. Kalev / Int. Fin. Markets, Inst. and Money 30 (2014) 205–219 211

Table 2
Predictive power of the order book slope.

ı0 ı1 ı2 ϕ1 ϕ2 DoF Q2 (12)

Panel A: GARCH model

0.6114 −0.1248 −0.0798 0.0077 −2.40 × 10−5 10.5775 14.6970
5% level (20.00%), 21.11% (81.11%), 0% (30.00%), 15.56% (5.56%), 84.44% (44.44%), 5.56% 10.00%
10% level (21.11%), 26.67% (84.44%), 0% (40.00%), 20.00% (5.56%), 93.33% (50.00), 5.56% 15.56%

Panel B: CGARCH model

0.0868 −0.0299 −0.0102 0.0041 −1.40 × 10−5 6.8531 9.7797
5% level (12.22%), 24.44% (83.33%), 0% (31.11%), 14.44% (0.00%), 96.67% (31.11%), 12.22% 4.44%
10% level (21.11%), 28.89% (87.78%), 0% (42.22%), 23.33% (0.00%), 97.78% (41.11%), 13.33% 7.78%

This table presents the results of investigating the predictive power of the order book slope on future volatility. The results
are obtained from estimating the following models. AR(1)-GARCH(1,1): rt =  + rt−1 + εt , t2 = ω + ˛ε2t−1 + ˇt−1
2
+ ı0 Dpost + (ı1 +
ı2 Dpost )SLOPEt−1 + ϕ1 NTt−1 + ϕ2 ATSt−1 . AR(1)-CGARCH(1,1): rt =  + rt−1 + εt , t2 = qt + ˛(ε2t−1 − qt−1 ) + ˇ(t−1
2
− qt−1 ), qt =  +
(qt−1 −  ) +
(ε2t−1 − t−1
2
) + ı0 Dpost + (ı1 + ı2 Dpost )SLOPEt−1 + ϕ1 NTt−1 + ϕ2 ATSt−1 , where rt is the seasonally adjusted return
of the stock at the tth interval; t2 is the conditional variance of the error term process εt , which follows a Student’s t-distribution,
t2 − qt is the transitory volatility component; and qt is the time-varying permanent (long-run) volatility. SLOPEt−1 , NTt−1 , and
ATSt−1 are, respectively, the order book slope, the total number of trades, and the average trade size for the interval t − 1. The
variable Dpost is a dummy variable that takes on the value of one from November 28, 2005, onward, and zero otherwise; DoF is
the average degree of freedom for the GARCH and CGARCH models; and Q2 (12) is the average of the Ljung–Box portmanteau
test statistics for serial correlation in the squared residuals with 12 lags. The mean estimates of the coefficients are reported in
Panels A and B. The percentages in parentheses indicate the numbers of estimates that are negative and significant, whereas
the percentages outside parentheses indicate the numbers of estimates that are positive and significant. The rows “5% level”
and “10% level” show significance at the 5% and 10% levels of significance, respectively.

the purpose of analyzing the effect of removing broker IDs on the information content of the limit order
book slope. If the removal of broker IDs on the ASX has a significant impact on the informativeness
of the limit order book slope, we should observe the majority of the coefficient estimates for ı2 to be
statistically significant. If ı2 is positive (negative) and significant, the move to anonymity has resulted
in a decrease (an increase) in the informativeness of the limit order book slope.
The results of this investigation are given in Panel A of Table 2. Similar to Jones et al. (1994), we
find that the number of trades has a more significant effect on volatility than the average trade size.
More importantly, we document strong support for the predictability of the limit order book slope
over the price volatility in the next 30-minute interval. Specifically, the coefficient estimate for the
order book slope in the previous interval is negative and significant at the 5% (10%) level in 81.11%
(84.44%) of the stocks under investigation. The move to anonymity has resulted in significant changes
in the informativeness of the limit order book slope, with the coefficient estimates for ı2 significant
at the 5% (10%) level in 45.56% (60.00%) of our sample stocks.12 In addition, the coefficient estimates
of ı2 are negative (positive) and significant at the 10% level in 40.00% (20.00%) of the stocks under
examination. This evidence suggests that the limit order book slope tends to be more informative
after the move to anonymity.
Since intraday volatility can contain different components (see, for example, Andersen and
Bollerslev, 1997a; Muller et al., 1997), we also utilize the component GARCH (CGARCH) model of Engle
and Lee (1999) to investigate the relation between the order book slope and the long-run permanent
component of volatility. We estimate the following AR(1)-CGARCH(1,1) model:

rt =  + rt−1 + εt , t2 = qt + ˛(ε2t−1 − qt−1 ) + ˇ(t−1

2
− qt−1 ),

qt =  + (qt−1 −  ) +
(ε2t−1 − t−1
2
) + ı0 Dpost + (ı1 + ı2 Dpost )SLOPEt−1

+ ϕ1 NTt−1 , +ϕ2 ATSt−1 , (4)

12
These figures are obtained by summing the percentages of the positive and significant and negative and significant coefficient
estimates for ı2 in Table 2. We also perform a t-test and find that the mean of the ␦2 coefficients is significantly different from
zero.
212 H.N. Duong, P.S. Kalev / Int. Fin. Markets, Inst. and Money 30 (2014) 205–219

where the conditional variance t2 is mean reverting around a permanent component qt , with the speed
of mean reversion determined by the parameters ˛ and ˇ and qt being the time-varying, long-run
permanent component of volatility. The results of this analysis are given in Panel B of Table 2.
Similar to the results obtained for the GARCH model, the coefficient estimates for the lagged order
book slope are negative and significant at the 5% (10%) level in 83.33% (87.78%) of our sample stocks.
This finding provides strong support for the predictive power of the order book slope on the future
permanent component of volatility. Furthermore, consistent with the results obtained when using the
GARCH model, the limit order book slope tends to be more informative in an anonymous market. All
of the GARCH and CGARCH models are also well specified, as reflected by the insignificance of the
majority of the Ljung–Box portmanteau tests for serial correlation in the squared residuals with 12
lags at the 5% level of significance.
Overall, the results presented in Table 2 support the arguments of Foucault et al. (2007) that the
limit order book is a channel for price volatility information. The results in Table 2 also indicate that
the move to anonymity has a significant impact on the information content of the limit order book
slope for future price volatility. Among stocks that experience a significant impact of anonymity on
the informativeness of the limit order book slope, the limit order book slope tends to become more
informative about future volatility after the removal of broker IDs on the ASX. Our findings extend the
results of Foucault et al. (2007) that the informativeness of the bid–ask spread declines after the move
to anonymity. Since the order book slope summarizes limit order book information at and beyond the
best quotes, our results suggest that investors place more informative limit orders in the anonymous
market, especially those orders below the best quotes.

3.3. Information content of the order book slope on the demand and supply sides

Since the limit order book slope is the average of the slopes of the demand and supply sides, we
also analyze the informativeness of the slope of the limit order book on the buy side (buy slope) versus
the slope of the limit order book on the sell side (sell slope). The results of this investigation are given
in Table 3.

Table 3
Predictive power of the buy (demand) and sell (supply) side slopes.

ı0 ı1 ı2 ϕ1 ϕ2 DoF Q2 (12)

Panel A: Buy side

0.0309 −0.0359 −0.0062 0.0066 −3.50 × 10−5 6.6535 9.8617
5% level (10.00%), 18.89% (75.56%), 0% (23.33%), 13.33% (0%), 96.67% (32.22%), 7.78% 5.56%
10% level (15.56%), 21.11% (86.67%), 0% (32.22%), 20.00% (0%), 97.78% (35.56%), 7.78% 6.67%

Panel B: Sell side

0.0338 −0.0252 −0.0033 0.0040 −1.40 × 10−5 7.6513 10.2865
5% level (4.44%), 17.78% (63.33%), 0% (22.22%), 11.11% (0%), 95.56% (40.00%), 5.56% 5.56%
10% level (16.67%), 24.44% (73.33%), 0% (31.11%), 15.56% (0%), 95.56% (43.33%), 8.89% 10.00%

This table presents the results of investigating the predictive power of the slope of the order book on the demand
(buy) and supply (sell) sides on future volatility. The results are obtained from estimating the following AR(1)-
CGARCH(1,1) models. AR(1)-CGARCH(1,1) model for the buy side: rt =  + rt−1 + εt , t2 = qt + ˛(ε2t−1 − qt−1 ) + ˇ(t−1
2
− qt−1 ),
qt =  + (qt−1 −  ) +
(ε2t−1 − t−1
2
) + ı0 Dpost + (ı1 + ı2 Dpost )BUYSLOPEt−1 + ϕ1 NTt−1 + ϕ2 ATSt−1 . AR(1)-CGARCH(1,1) model
for the sell side: rt =  + rt−1 + εt , t2 = qt + ˛(ε2t−1 − qt−1 ) + ˇ(t−1
2
− qt−1 ), qt =  + (qt−1 −  ) +
(ε2t−1 − t−1
2
) + ı0 Dpost +
(ı1 + ı2 Dpost )SELLSLOPEt−1 + ϕ1 NTt−1 + ϕ2 ATSt−1 , where rt is the seasonally adjusted return of the stock at the tth interval;
t2 is the conditional variance of the error term process εt , which follows a Student’s t-distribution; t2 − qt is the transitory
volatility component; qt is the time-varying permanent (long-run) volatility; BUYSLOPEt−1 , SELLSLOPEt−1 , NTt−1 , and ATSt−1 are,
respectively, the order book slope on the buy side, the order book slope on the sell side, the total number of trades, and the aver-
age trade size for the interval t − 1; Dpost is a dummy variable that takes on the value of one from November 28, 2005, onward,
and zero otherwise; DoF is the average degree of freedom for the CGARCH model; and Q2 (12) is the average of the Ljung–Box
portmanteau test statistics for serial correlation in the squared residuals with 12 lags. The mean estimates of the coefficients are
reported in Panels A and B. The percentages in parentheses indicate the numbers of estimates that are negative and significant,
whereas the percentages outside parentheses indicate the numbers of estimates that are positive and significant. The rows “5%
level” and “10% level” show significance at the 5% and 10% levels of significance, respectively.
 H.N. Duong, P.S. Kalev / Int. Fin. Markets, Inst. and Money 30 (2014) 205–219 213

The findings in Table 3 indicate that buy limit orders are more informative than sell limit orders
about the future permanent component of volatility. The coefficient estimates for the lagged buy slope
are negative and significant at the 5% (10%) level of significance in 75.56% (86.67%) of the stocks in
our sample. In contrast, the predictive power of the sell slope on the future permanent component of
volatility is evident at the 5% (10%) level of significance in only 63.33% (73.33%) of the stocks under
investigation. Our results complement the findings of Griffiths et al. (2000) that aggressive buy orders
are more likely to be motivated by information than aggressive sell orders are.13 We also document
that the number of negative and significant ı2 is larger than the number of positive and significant ı2
for both the buy and sell slopes. This finding suggests that, among stocks that experience significant
changes in the informativeness of the buy and sell slopes, the buy and sell slopes tend to be more
informative after the move to anonymity.

3.4. Information content of institutional and individual slopes

Panels A and B of Table 4 report the results of the examination of the predictive power of
Institutional Slope and Individual Slope over the future permanent component of volatility. The
lagged Institutional Slope is informative about the future permanent component of volatility in
57.78% (70.00%) of the stocks under investigation at the 5% (10%) level of significance. We find
evidence supporting the predictive power of Individual Slope on the future permanent compo-
nent of volatility at the 5% (10%) level of significance in only 32.22% (38.89%) of the stocks in
our sample. These results imply that institutional investors are more informed than individual
investors and institutional limit orders do convey institutions’ informational advantage over future
volatility.
We document that 60.00% of the stocks in our sample experience significant changes in the informa-
tiveness of Institutional Slope. In contrast, the majority of the coefficient estimates for ı2 for Individual
Slope are not statistically significant. Institutional limit orders are also more informative in the anony-
mous market, where 40.00% (20.00%) of our sample stocks have negative (positive) and significant
coefficient estimates for ı2 at the 10% level of significance. The CGARCH models for Institutional
Slope and Individual Slope are also well specified, as evident from the insignificance of the major-
ity of the Ljung–Box portmanteau test statistics for serial correlation in the squared residuals with 12
lags.
We also run separate CGARCH models for changes in the institutional and individual slopes. From
Panels C and D of Table 4, we find that the changes in the institutional and individual slopes are
significantly related to future volatility in 31.11% and 26.66% of the stocks in our sample, respec-
tively. In contrast, from Panels A and B, the institutional and individual slopes are informative for
about future volatility in 70.00% and 38.89% of the stocks under investigation, respectively. We
conclude that the changes in slope are less informative than the overall levels of the order book
slope.14
In general, the results presented in Tables 2–4 suggest that the move to anonymity often results
in an increase in the informativeness of the overall limit order book slope. Institutional limit orders
are also more affected by this change than individual limit orders. After the move to anonymity, insti-
tutional investors are more willing to incorporate their informational advantage in their limit orders
submissions. If anonymity reduces the incentive for better-informed investors to submit “bluffing”
limit orders (Foucault et al., 2007), institutional limit orders (order book slope) should be more infor-
mative after the removal of broker IDs. Our finding of an improvement in the informativeness of

13
Our results can arguably be driven by rising market conditions during our sample period. We control for this factor by
using detrended returns in our CGARCH models. We observe a decline in the difference between the number of negative and
significant coefficient estimates for the buy and sell slopes. The buy slope is still more informative than the sell slope. Thus,
our finding that the buy slope is more informative than the sell slope can be partially driven by general market conditions. We
thank the referee for suggesting this alternative explanation for our results. See Chiyachantana et al. (2004) for a discussion on
the effect of underlying market conditions on the price impact of buy and sell orders.
14
Our results imply that omitted broad firm characteristics may be more important than limit order book characteristics. We
thank the referee for suggesting this analysis and implication.
214 H.N. Duong, P.S. Kalev / Int. Fin. Markets, Inst. and Money 30 (2014) 205–219

Table 4
Predictive power of the institutional slope and individual slope.

ı0 ı1 ı2 ϕ1 ϕ2 DoF Q2 (12)

Panel A: Institutional slope

0.2622 −0.0058 −0.0052 0.0168 −3.20 × 10−5 9.4569 11.9708
5% level (16.67%), 18.89% (57.78%), 0% (25.56%), 16.67% (0%), 88.89% (40.00%), 7.78% 7.78%
10% level (28.89%), 25.56% (70.00%), 0% (40.00%), 20.00% (0%), 91.11% (48.89%), 10.00% 12.22%

Panel B: Individual slope

0.0397 −0.0139 −0.0025 0.0052 −1.40 × 10−5 6.1576 11.0838
5% level (8.89%), 12.22% (32.22%), 0% (13.33%), 8.89% (0%), 88.89% (34.44%), 6.67% 10.00%
10% level (10.00%), 18.89% (38.89%), 0% (18.89%), 10.00% (0%), 91.11% (46.67%), 6.67% 11.11%

Panel C: Changes in institutional slope

−0.0009 −0.0025 0.0006 0.0030 −1.70 × 10−5 5.5980 10.8776
5% level (23.33%), 8.89% (20.00%), 6.67% (10.00%), 6.67% (0%), 83.33% (46.67%), 1.11% 4.44%
10% level (26.67%), 11.11% (23.33%), 7.78% (12.22%), 7.78% (0%), 83.33% (53.33%), 2.22% 11.11%

Panel D: Changes in individual slope

0.0093 −0.0203 −0.0081 0.0021 0.0001 5.6887 11.8097
5% level (26.67%), 8.89% (20.00%), 2.22% (6.67%), 2.22% (0%), 80.00% (43.33%), 6.67% 10.00%
10% level (35.56%), 12.22% (24.44%), 2.22% (10.00%), 5.56% (0%), 86.67% (50.00%), 6.67% 15.56%

This table presents the results of investigating the predictive power on future volatility of the slope of the order book based
on the orders submitted by institutional and individual investors. The results are obtained from estimating the following
AR(1)-CGARCH(1,1) models. AR(1)-CGARCH(1,1) model for the Institutional slope: rt =  + rt−1 + εt , t2 = qt + ˛(ε2t−1 − qt−1 ) +
ˇ(t−1
2
− qt−1 ), qt =  + (qt−1 −  ) +
(ε2t−1 − t−1
2
) + ı0 Dpost + (ı1 + ı2 Dpost )INSTITUTIONALSLOPEt−1 + ϕ1 NTt−1 + ϕ2 ATSt−1 .
AR(1)-CGARCH(1,1) model for the individual slope: rt =  + rt−1 + εt , t2 = qt + ˛(ε2t−1 − qt−1 ) + ˇ(t−1
2
− qt−1 ),
qt =  + (qt−1 −  ) +
(ε2t−1 − t−1
2
) + ı0 Dpost + (ı1 + ı2 Dpost )INDIVIDUALSLOPEt−1 + ϕ1 NTt−1 + ϕ2 ATSt−1 , AR(1)-
CGARCH(1,1) model for the changes in institutional slope: rt =  + rt−1 + εt , t2 = qt + ˛(ε2t−1 − qt−1 ) + ˇ(t−1
2
− qt−1 ),
qt =  + (qt−1 −  ) +
(ε2t−1 − t−1
2
) + ı0 Dpost + (ı1 + ı2 Dpost )INSTSLOPECHANGEt−1 + ϕ1 NTt−1 + ϕ2 ATSt−1 . AR(1)-
CGARCH(1,1) model for the changes in individual slope: rt =  + rt−1 + εt , t2 = qt + ˛(ε2t−1 − qt−1 ) + ˇ(t−1
2
− qt−1 ),
qt =  + (qt−1 −  ) +
(ε2t−1 − t−1
2
) + ı0 Dpost + (ı1 + ı2 Dpost )INDISLOPECHANGEt−1 + ϕ1 NTt−1 + ϕ2 ATSt−1 , where rt is
the seasonally adjusted return of the stock at the tth interval; t2 is the conditional variance of the error term process εt , which
follows a Student’s t-distribution; t2 − qt is the transitory volatility component; qt is the time-varying permanent (long-run)
volatility; INSTITUTIONALSLOPEt−1 , INDIVIDUALSLOPEt−1 , INSTSLOPECHANGEt−1 , INDISLOPECHANGEt−1 , NTt−1 , and ATSt−1 are,
respectively, the order book slope based on institutional orders, the order book slope based on individual orders, the changes
in the order book slope based on institutional orders, the changes in the order book slope based on individual orders, the
total number of trades, and the average trade size for the interval t − 1; Dpost is a dummy variable that takes on the value of
one from November 28, 2005, onward, and zero otherwise; DoF is the average degree of freedom for the CGARCH model; and
Q2 (12) is the average of the Ljung–Box portmanteau test statistics for serial correlation in the squared residuals with 12 lags.
The mean estimates of the coefficients are reported in Panels A and B. The percentages in parentheses indicate the numbers of
estimates that are negative and significant, whereas the percentages outside parentheses indicate the numbers of estimates
that are positive and significant. The rows “5% level” and “10% level” show significance at the 5% and 10% levels of significance,
respectively.

(institutional) limit orders is also consistent with the suggestion of Linnainmaa and Saar (2012) that
prices adjust less efficiently to order flow information when the degree of anonymity increases. For
individual investors, in the transparent market, the broker ID information on the ASX is disseminated
only to the broker community (ASX, 2003). Thus, the move to anonymity does not significantly change
individual investors’ information environment. Therefore, the removal of broker IDs has a minimal
impact on the information content of individual limit orders.

3.5. Robustness tests

We perform additional robustness tests for the results regarding the informativeness of the limit
order book slope about future volatility. The first set of robustness tests involves the use of an alterna-
tive measure of the order book slope. In Table 5, we present the results using two different measures
of the limit order book slope. In Panel A, we use the measure of the limit order book slope proposed by
 H.N. Duong, P.S. Kalev / Int. Fin. Markets, Inst. and Money 30 (2014) 205–219 215

Table 5
Robustness tests – alternative measures of the limit order book slope.

ı0 ı1 ı2 ϕ1 ϕ2 DoF Q2 (12)

Panel A: Using the measure of the limit order book slope of Kalay et al. (2004)
0.0135 −4.6428 −0.0590 0.0032 −7.50 × 10−6 6.1480 10.1826
5% level (8.89%), 14.44% (82.22%), 0% (15.56%), 12.22% (0%), 92.22% (20.00%), 13.33% 4.44%
10% level (13.33%), 18.89% (90.00%), 0% (28.89%), 20.00% (0%), 95.56% (21.11%), 15.56% 7.78%

Panel B: Using all orders up to 100 levels of the order book

0.0092 −0.0372 −0.0034 0.0042 −1.57 × 10−5 8.2660 11.6817
5% level (10.00%), 15.56% (50.00%), 0% (21.11%), 11.11% (0%), 90.00% (43.33%), 6.67% 10.00%
10% level (13.33%), 21.11% (58.89%), 0% (30.00%), 16.67% (0%), 92.22% (51.11%), 6.67% 11.11%

This table presents the results of the robustness tests on the predictive power of the limit order book slope. The follow-
ing AR(1)-CGARCH(1,1) models are estimated Robustness test 1: using the measure of the limit order book slope of Kalay
et al. (2004) rt =  + rt−1 + εt , t2 = qt + ˛(ε2t−1 − qt−1 ) + ˇ(t−1
2
− qt−1 ), qt =  + (qt−1 −  ) +
(ε2t−1 − t−1
2
) + ı0 Dpost + (ı1 +
ı2 Dpost )SLOPEKSWt−1 + ϕ1 NTt−1 + ϕ2 ATSt−1 , Robustness test 2: using the order book slope calculated from all orders, up to
100 ticks away from the best quotes: rt =  + rt−1 + εt , t2 = qt + ˛(ε2t−1 − qt−1 ) + ˇ(t−1
2
− qt−1 ), qt =  + (qt−1 −  ) +
(ε2t−1 −
t−1
2
) + ı0 Dpost + (ı1 + ı2 Dpost )SLOPEALLt−1 + ϕ1 NTt−1 + ϕ2 ATSt−1 , where rt is the seasonally adjusted return of the stock at the
tth interval; t2 is the conditional variance of the error term process εt , which follows a Student’s t-distribution; qt is the time-
varying permanent (long-run) volatility; SLOPEKSWt−1 is the order book slope, calculated based on the methodology of Kalay
et al. (2004), at the end of interval t − 1; SLOPEALLt−1 is the order book slope, calculated from all orders up to 100 ticks away from
the best quotes, at the end of interval t − 1. NTt−1 and ATSt−1 are the total number of trades and the average trade size for interval
t − 1; Dpost is a dummy variable that takes on the value of one from November 28, 2005, onward, and zero otherwise; DoF is the
average degree of freedom for the CGARCH model; and Q2 (12) is the average of the Ljung–Box portmanteau test statistics for
serial correlation in the squared residuals with 12 lags. The mean estimates of the coefficients are reported. The percentages in
parentheses indicate the numbers of estimates that are negative and significant, whereas the percentages outside parentheses
indicate the numbers of estimates that are positive and significant. The rows “5% level” and “10% level” show significance at the
5% and 10% levels of significance, respectively.

Kalay et al. (2004). This measure of the slope on the supply side of the order book for firm i in interval
t (SEKSWi,t ) is defined as
N 
1 A A
(V+1 − VA )/NOSHi,t
SEKSWi,t = , (5)
NA pA+1 /pA − 1
=0

where NOSHi,t is the number of shares outstanding for firm i in interval t, VA is the total share volume
at ask price level  (pA ), and NA is the total number of bid prices (tick levels) containing orders. The
order book on the buy side is calculated similarly. The overall limit order book slope is then calculated
as the average of the slope on the buy (demand) and sell (supply) sides. The 10 best bid and ask quotes
together with the share volume at these quotes are used in the order book slope calculation.
Consistent with the results obtained in Panel B of Table 2, the results in Panel A of Table 5 provide
strong support for the informativeness of the limit order book slope about the future permanent
component of volatility. The coefficient estimates for the lagged order book slope are negative and
significant for the majority of stocks under investigation. In addition, the move to anonymity has a
significant impact on the informativeness of the order book slope, with the order book slope tending
to become more rather than less informative after the removal of broker IDs on the ASX.
In Panel B of Table 5, instead of using only the 10 best levels in the limit order book, we calculate
the limit order book slope using all orders in the limit order book, up to 100 ticks away from the best
quotes. The results show that when all orders in the limit order book, up to 100 ticks away from the
best quotes are used to calculate the order book slope, the order book slope is informative about the
future permanent component of volatility in 58.89% of the stocks in our sample. We still find evidence
that the move to anonymity has a significant impact on the informativeness of the limit order book
slope. Comparing these results with the results obtained from Panel B of Table 2, we conclude that the
limit order book slope calculated using the 10 best levels in the limit order book is more informative
than the limit order book slope calculated using all orders up to 100 ticks away from the best quotes.
This finding implies that including orders beyond the 10th best level in the limit order book results in
216 H.N. Duong, P.S. Kalev / Int. Fin. Markets, Inst. and Money 30 (2014) 205–219

Table 6
Robustness tests – alternative measures of returns.

ı0 ı1 ı2 ϕ1 ϕ2 DoF Q2 (12)

Panel A: Seasonally adjusted returns calculated from ask quotes

0.3009 −0.0348 −0.0248 0.0071 −7.72 × 10−6 7.7703 11.1587
5% level (8.89%), 26.67% (74.44%), 0% (27.78%), 11.11% (0%), 93.33% (34.44%), 6.67% 11.11%
10% level (13.33%), 30.00% (81.11%), 0% (40.00%), 13.33% (0%), 93.33% (40.00%), 8.89% 14.44%

Panel B: Seasonally adjusted returns calculated from bid quotes

0.1094 −0.0333 −0.0118 0.0044 −1.19 × 10−5 7.7545 10.5150
5% level (12.22%), 21.11% (72.22%), 0% (28.89%), 15.56% (0%), 95.56% (35.56%), 10.00% 5.56%
10% level (15.56%), 30.00% (82.22%), 0% (40.00%), 18.89% (0%), 96.67% (40.00%), 11.11% 6.67%

Panel C: Seasonally adjusted returns calculated from transaction prices

0.1547 −0.0312 −0.0103 0.0040 −3.71 × 10−5 9.2303 9.9499
5% level (10.00%), 12.22% (66.67%), 0% (14.44%), 11.11% (80.00%), 0% (24.44%), 0% 7.78%
10% level (13.33%), 26.67% (82.22%), 0% (32.22%), 17.78% (83.33%), 0% (33.33%), 2.22% 10.00%

Panel D: Unadjusted returns calculated from bid–ask midpoints

6.13 × 10−7 −3.63 × 10−7 −1.06 × 10−7 9.32 × 10−8 −5.00 × 10−11 17.0170 23.1705
5% level (22.22%), 18.9% (85.56%), 0% (41.11%), 22.22% (0%), 100.00% (28.89%), 15.56% 45.56%
10% level (24.44%), 28.89% (90.00%), 0% (53.33%), 26.67% (0%), 100.00% (36.67%), 18.89% 48.89%

Panel E: Overnight returns

0.0308 −0.0138 −0.0030 0.0027 −1.25 × 10−5 10.8075 10.4727
5% level (8.89%), 22.22% (78.89%), 0% (26.67%), 11.11% (0%), 96.67% (37.78%), 7.78% 4.44%
10% level (12.22%), 28.89% (85.56%), 0% (35.56%), 18.89% (0%), 96.67% (41.11%), 8.89% 7.78%

This table presents the results of the robustness tests on the predictive power of the limit order book slope. The first three
robustness tests use seasonally adjusted returns, with returns calculated based on ask quotes, bid quotes and transaction
prices, respectively. The forth robustness test uses returns calculated from the bid–ask midpoint, without adjusting for
seasonal patterns. The fifth robustness test incorporates the overnight period in the calculation of returns. The following
AR(1)-CGARCH(1,1) models are estimated for these robustness tests. rt =  + rt−1 + εt , t2 = qt + ˛(ε2t−1 − qt−1 ) + ˇ(t−1
2
− qt−1 ),
qt =  + (qt−1 −  ) +
(ε2t−1 − t−1
2
) + ı0 Dpost + (ı1 + ı2 Dpost )SLOPEt−1 + ϕ1 NTt−1 + ϕ2 ATSt−1 , where rt is the return of the
stock at the tth interval (seasonally adjusted return based on ask quotes, bid quotes, and transaction prices for the first three
robustness tests; unadjusted returns based on the bid–ask midpoint for the forth robustness test; and seasonally adjusted
returns based on the bid–ask midpoint, including the overnight period); t2 is the conditional variance of the error term process
εt , which follows a Student’s t-distribution; qt is the time-varying permanent (long-run) volatility. SLOPEt−1 , NTt−1 , and ATSt−1
are, respectively, the order book slope, the total number of trades, and the average trade size for the interval t − 1. Dpost is a
dummy variable that takes on the value of one from November 28, 2005, onward, and zero otherwise; DoF is the average degree
of freedom for the CGARCH model; and Q2 (12) is the average of the Ljung–Box portmanteau test statistics for serial correlation
in the squared residuals with 12 lags. The mean estimates of the coefficients are reported. The percentages in parentheses
indicate the numbers of estimates that are negative and significant, whereas the percentages outside parentheses indicate the
numbers of estimates that are positive and significant. The rows “5% level” and “10% level” show significance at the 5% and 10%
levels of significance, respectively.

the inclusion of stale limit orders, which reduces the overall informativeness of the limit order book
slope.
In the second set of robustness tests, we present the results using alternative measures of returns.
In Panels A to C of Table 6, we calculate the returns based on ask quotes, bid quotes, and transaction
prices. Specifically, we calculate the return for each interval as the difference of the natural logarithms
of the closing bid quote (or closing ask quote, or last traded price) of the current interval and of that
of the previous interval. The return for the first interval is calculated as the difference of the natural
logarithms of the closing bid quote (or closing ask quote, or last traded price) and of the opening bid
quote (or opening ask quote, or first traded price) of the interval. Similar to our main finding in Table 2,
we show that the slope of the limit order book is negatively related to future price volatility. The move
to anonymity also has a significant effect on the informativeness of the limit order book slope.
In Panel D of Table 6, we present the CGARCH results using returns without adjustments for intraday
patterns. Our results are qualitatively similar to the findings in Table 2. Note that usage of unadjusted
returns increases the number of CGARCH models that experience significant serial correlation in the
squared residuals, as evident by the Ljung–Box test statistics. This finding illustrates the importance
 H.N. Duong, P.S. Kalev / Int. Fin. Markets, Inst. and Money 30 (2014) 205–219 217

of adjusting returns for intraday patterns, as highlighted in Andersen and Bollerslev (1997b), Martens
et al. (2002) and Andersen et al. (2003).15
Finally, in Panel E of Table 6, we present our results with the inclusion of overnight returns. In
addition to the 12 intraday intervals (10:10 am–10:30 am, 10:30 am–11 am, 11 am–11:30 am, 11:30
am–12 pm, 12 pm–12:30 pm, 12:30 pm–1 pm, 1 pm–1:30 pm, 1:30 pm–2 pm, 2 pm–2:30 pm, 2:30
pm–3 pm, 3 pm–3:30 pm, 3:30 pm–4 pm), we add one more interval for the period 4 PM of the previous
day to 10:10 AM of the day after. Similar to our main findings in Panel B of Table 2, we document that
the limit order book slope is informative about future volatility. The removal of broker IDs also has a
significant impact on the information content of the limit order book slope for future price volatility.

4. Conclusions

We examine the information content of the order book slope in explaining future price volatility
on the ASX. We also investigate whether the relation between the order book slope and future price
volatility is affected by the removal of broker IDs on the ASX. Analyzing the stocks included in the
S&P/ASX 100 index for the period between July 1, 2005, and June 30, 2006, we find that the order
book slope is informative in explaining the future price volatility of the majority of stocks under
investigation. The predictive power of the order book slope is present in both the overall volatility
and the permanent component of volatility. These findings support the importance of limit orders
in the order submission strategies of informed investors and the notion that the limit order book is
a channel for volatility information. In addition, when decomposing the limit order book into limit
orders submitted by institutional and individual investors, we observe that institutional limit orders
are more informative than individual limit orders. This finding implies that the informativeness of the
limit order book is mainly due to limit orders submitted by institutional investors. We also document
that the slope of the order book on the buy side is more informative than that on the sell side, and the
limit order book beyond the best quotes contains significant information on future volatility beyond
that contained in the best quotes of the limit order book.
The removal of broker IDs on the ASX has a significant impact on the predictive power of the limit
order book slope. Moreover, institutional limit orders become more informative about future volatility
after the move to anonymity, whereas anonymity has a minimal impact on individual limit orders.
This finding implies that the move to anonymity has reduced the risk of front-running activities and
better-informed investors are more willing to submit and expose their limit orders in the limit order
book. Our results also suggest that prices react less to order flow information from informed investors
after the switch to an anonymous trading system. Similar to Linnainmaa and Saar (2012), we highlight
that the benefits of withholding broker IDs come at the cost of reduced informational efficiency. Thus,
the market may be less competitive, with informed investors being now able to realize greater profits
at the expense of uninformed investors.
Our analysis raises several new issues. Given the proliferation of algorithmic trading (with high
frequency trading as the main subset) in recent years, it is important to examine whether our result
on the informativeness of institutional and individual orders is robust in the new “fast trading”
environment.16 In addition, other recent phenomena, such as dark pools, make/take fees, and mar-
ket fragmentation, are likely to affect institutional and individual order submission strategies, and
eventually the informativeness of their orders.17 We analyze the changes in pre-trade transparency

15
We perform additional robustness tests such as removing intraday seasonal factor for intraday returns, the order book slope,
and other control variables, including hidden orders in the order book slope calculation, and using one-hour interval instead of
a 30-min interval. The results of these investigations (available upon request) are consistent with those presented in Table 2.
16
Chordia et al. (2011) show that there has been a sharp upward trend in turnover for stocks traded on the New York Stock
Exchange. Trade size has declined while the number of transactions has increased substantially. They show that these trends in
trading activity are consistent with the notion that algorithms have allowed institutions to trade more cheaply and frequently.
In an unreported analysis, we observe similar patterns for the number of trades and trade size in the Australian market. These
results are available on request.
17
For recent research works on algorithmic trading, market fragmentation, make/take fees, dark pool, and short selling, see
Hendershott et al. (2011), O’Hara and Ye (2011), Foucault et al. (2013), Zhu (2014), and Beber and Pagano (2013), respectively.
We thank an anonymous referee for suggesting these future research directions.
218 H.N. Duong, P.S. Kalev / Int. Fin. Markets, Inst. and Money 30 (2014) 205–219

based on the removal of broker IDs. It will also be useful to study the effect on market quality of other
changes in pre-trade transparency rules, such as the disclosure of limit order book levels or the intro-
duction of limit orders with undisclosed volume, and changes in post-trade transparency rules, such
as the disclosure of block trades, portfolio positions, or short selling positions. Finally, our study is
based on a limit order market without designated market makers. Future studies can address similar
issues in markets with the presence of market makers or for different groups of stocks with and with-
out designated market makers.18 All the issues discussed above provide fruitful avenues for future
research.

Acknowledgments

The previous version of the paper was circulated under the title “Order Book Slope and Price Volatil-
ity”. We thank the Financial Management Association (FMA) for awarding this paper the best PhD
paper in the 2009 FMA European Conference. We gratefully acknowledge helpful comments from an
anonymous referee. We also thank Henk Berkman, Alain Chaboud, Dobrislav Dobrev, Robert Elliott,
Kingsley Fong, Doug Foster, Bruce Grundy, Michael McKenzie, Richard Roll, Gordon Sick, Erik Theis-
sen, Clara Vega, Terry Walter, and participants at the 20th Australasian Finance & Banking Conference,
the 2008 Accounting & Finance Association of Australia and New Zealand Conference, the 4th Annual
Central Bank Workshop on the Microstructure of Financial Markets, the 2009 FMA European Confer-
ence, the seminar programs of Cambridge University, The University of Calgary, Massey University,
The University of Bonn, The University of Melbourne, The University of Technology, Sydney, and the
Division of International Finance of the Board of Governors of the Federal Reserve System, Washing-
ton, DC. We are grateful to the Australian Stock Exchange and the Securities Industry Research Centre
of Asia-Pacific for providing the confidential data used in this study. The usual caveat applies.

References

Ahn, H.J., Bae, K.H., Chan, K., 2001. Limit orders, depth, and volatility: evidence from the Stock Exchange of Hong Kong. Journal
of Finance 56, 767–788.
Aitken, M., Almeida, N., Harris, F.H., McInish, deB.T.H., 2007. Liquidity supply in electronic markets. Journal of Financial Markets
10, 144–168.
Aitken, M., Berkman, H., Mak, D., 2001. The use of undisclosed limit orders on the Australian Stock Exchange. Journal of Banking
and Finance 25, 1563–1589.
Alangar, S., Bathala, C., Rao, R., 1999. The effect of institutional interest on the information content of dividend-change announce-
ments. Journal of Financial Research 22, 429–448.
Andersen, T.G., Bollerslev, T., 1997a. Heterogeneous information arrivals and return volatility dynamics: uncovering the long-run
in high frequency returns. Journal of Finance 52, 975–1005.
Andersen, T.G., Bollerslev, T., 1997b. Intraday seasonality and volatility persistence in financial markets. Journal of Empirical
Finance 4, 115–158.
Andersen, T.G., Bollerslev, T., Diebold, F.X., Labys, P., 2003. Modeling and forecasting realized volatility. Econometrica 71,
579–625.
Australian Securities Exchange, 2013. New Order Types. Overview Document of the Australian Securities Exchange. Australian
Securities Exchange, pp. 1–11.
Australian Stock Exchange, 2003. ASX Market Reforms – Enhancing the Liquidity of the Australian Equity Market. Market
Consultation Paper. Australian Stock Exchange, pp. 1–41.
Australian Stock Exchange, 2005. Enhancing the Liquidity of the Australian Equity Market. Decisions on Reform Paper. Australian
Stock Exchange, pp. 1–14.
Bae, K.-H., Jang, H., Park, K.S., 2003. Traders’ choice between limit and market orders: evidence from NYSE stocks. Journal of
Financial Markets 6, 517–538.
Beber, A., Pagano, M., 2013. Short-selling bans around the world: evidence from the 2007–2009 crisis. Journal of Finance 68,
343–381.
Boehmer, E., Saar, G., Yu, L., 2005. Lifting the veil: an analysis of pre-trade transparency at the NYSE. Journal of Finance 60,
783–815.
Bollerslev, T., 1987. A conditionally heteroscedastic time series model for speculative prices and rates of return. Review of
Econometrics and Statistics 69, 542–547.
Bollerslev, T., Chou, R.Y., Kroner, K.F., 1992. ARCH modeling in finance: a review of the theory and empirical evidence. Journal
of Econometrics 52, 5–59.
Cao, C., Hansch, O., Wang, X., 2008. Order placement strategies in a pure limit order book market. Journal of Financial Research
31, 113–140.

18
See Menkveld and Wang (2013) for a discussion on the benefits of designated market makers for small-cap stocks.
 H.N. Duong, P.S. Kalev / Int. Fin. Markets, Inst. and Money 30 (2014) 205–219 219

Cao, C., Hansch, O., Wang, X., 2009. The information content of an open limit order book. Journal of Futures Markets 29, 16–41.
Chakravarty, S., 2001. Stealth trading: which traders’ trades move stock prices? Journal of Financial Economics 61, 289–307.
Chiyachantana, C.N., Jain, P.K., Jiang, C.X., Wood, R.A., 2004. International evidence on institutional trading behavior and price
impact. Journal of Finance 59, 869–898.
Chordia, T., Roll, R., Subrahmanyam, A., 2011. Recent trends in trading activity and market quality. Journal of Financial Economics
101, 243–263.
Comerton-Forde, C., Frino, A., Mollica, V., 2005. The impact of limit order anonymity on liquidity: evidence from Paris, Tokyo
and Korea. Journal of Economics and Business 57, 528–540.
Comerton-Forde, C., Tang, K.M., 2009. Anonymity, liquidity and fragmentation. Journal of Financial Markets 12, 337–367.
Dennis, P., Weston, J., 2001. Who’s Informed: An Analysis of Information and Ownership Structure, Working Paper. Rice Uni-
versity.
Duong, H.N., Kalev, P.S., 2013. Anonymity and order submissions. Pacific Basin Finance Journal 25, 101–118.
Duong, H.N., Kalev, P.S., Krishnamurti, C., 2009. Order aggressiveness of institutional and individual investors. Pacific Basin
Finance Journal 17, 533–546.
Engle, R.F., Lee, G.G.J., 1999. A permanent and transitory component model of stock return volatility. In: Engle, R.F., White, H.
(Eds.), Cointegration, Causality and Forecasting: A Festschrift in Honor of Clive WJ Granger. Oxford University Press, New
York, USA.
Eom, K.S., Ok, J., Park, J.-H., 2007. Pre-trade transparency and market quality. Journal of Financial Markets 10, 319–341.
Fleming, J., Kirby, C., Ostdiek, B., 2001. The economic value of volatility timing. Journal of Finance 56, 329–352.
Fleming, J., Kirby, C., Ostdiek, B., 2003. The economic value of volatility timing using realized volatility. Journal of Financial
Economics 67, 473–509.
Foucault, T., 1999. Order flow composition and trading costs in a dynamic limit order market. Journal of Financial Markets 2,
99–134.
Foucault, T., Kadan, O., Kandel, E., 2013. Liquidity cycles and make/take fees in electronic markets. Journal of Finance 68, 299–341.
Foucault, T., Moinas, S., Theissen, E., 2007. Does anonymity matter in electronic limit order markets? Review of Financial Studies
20, 1707–1747.
Frino, A., Gerace, D., Lepone, A., 2008. Limit order book, anonymity and market liquidity: evidence from the Sydney Futures
Exchanges. Accounting and Finance 48, 561–573.
Griffiths, M.D., Smith, B., Turnbull, D.A.S., White, R.W., 2000. The costs and determinants of order aggressiveness. Journal of
Financial Economics 56, 65–88.
Harris, L., Panchapagesan, V., 2005. The information content of the limit order book: evidence from the NYSE specialist trading
decisions. Journal of Financial Markets 8, 25–67.
Hendershott, T., Jones, C.M., Menkveld, A.J., 2011. Does algorithmic trading improve liquidity? Journal of Finance 66, 1–33.
Jones, C., Kaul, G., Lipson, M., 1994. Transaction, volume and volatility. Review of Financial Studies 7, 631–651.
Kalay, A., Sade, O., Wohl, A., 2004. Measuring stock illiquidity: an investigation of the demand and supply schedule at the TASE.
Journal of Financial Economics 74, 461–486.
Kalay, A., Sade, O., Wohl, A., 2009. Detecting liquidity traders. Journal of Financial and Quantitative Analysis 44, 29–54.
Kalev, P.S., Liu, W.-M., Pham, P.K., Jarnecic, E., 2004. Public information arrival and volatility of intraday stock returns. Journal
of Banking and Finance 28, 1441–1467.
Lepone, A., Mistry, M., 2011. The information content of undisclosed limit orders around broker anonymity. Australasian
Accounting Business and Finance Journal 5, 6–18.
Linnainmaa, J.T., Saar, G., 2012. Lack of anonymity and the inference from order flow. Review of Financial Studies 25, 1414–1456.
Madhavan, A., Porter, D., Weaver, D., 2005. Should securities markets be transparent? Journal of Financial Markets 8, 265–287.
Martens, M., Chang, Y.C., Taylor, S.J., 2002. A comparison of seasonal adjustment methods when forecasting intraday volatility.
Journal of Financial Research 25, 283–299.
Menkveld, A.J., Wang, T., 2013. How do designated market makers create value for small-caps? Journal of Financial Markets 16,
563–571.
Muller, U.A., Dacorogana, M.M., Dave, R.D., Olsen, R.B., Pictet, O.V., von Weizsacker, J.E., 1997. Volatilities of different time
resolution – analyzing the dynamics of market components. Journal of Empirical Finance 4, 213–239.
Næs, R., Skjeltorp, J.A., 2006. Order book characteristics and the volume–volatility relation: empirical evidence from a limit
order market. Journal of Financial Markets 9, 408–432.
O’Hara, M., Ye, M., 2011. Is market fragmentation harming market quality? Journal of Financial Economics 100, 459–474.
Pascual, R., Veredas, D., 2010. Does the open limit order book matter in explaining informational volatility? Journal of Financial
Econometrics 8, 57–87.
Ranaldo, A., 2004. Order aggressiveness in limit order book markets. Journal of Financial Markets 7, 53–74.
Rindi, B., 2008. Informed traders as liquidity providers: anonymity, liquidity and price formation. Review of Finance 12, 497–532.
Roll, R., 1984. A simple implicit measure of the effective bid–ask spread in an efficient market. Journal of Finance 39, 1127–1139.
Swan, P.L., Westerholm, J., 2006. Market Architecture and Global Exchange Efficiency, Working Paper. University of New South
Wales.
Szewczyk, S., Tsetsekos, G., Varma, R., 1992. Institutional ownership and the liquidity of common stock offerings. Financial
Review 27, 211–225.
Wald, J.K., Horrigan, H.T., 2005. Optimal limit order choice. Journal of Business 78, 597–619.
Zhu, H., 2014. Do dark pools harm price discovery? Review of Financial Studies 27, 747–789.