Effects of stock attributes, market
structure, and tick size on the speed of
spread and depth adjustment
(An earlier version of
“The Dynamics of Quote Adjustments”)
Kee H. Chung
State University of New York (SUNY) at Buffalo
Chairat Chuwonganant
Indiana University-Purdue University at Fort Wayne
1
Motivation


The bid-ask spread is an important
measure of market quality because it
represents the cost of trading in
securities markets.
Marketmakers adjust the bid-ask
spread in response to new
information embedded in order flow,
trades, and return volatility, among
other factors.
2


We know very little about the
dynamics of the bid-ask spread. Prior
studies offer little evidence as to the
speed at which new information is
impounded into the bid-ask spread.
There is also limited evidence
regarding how market structure and
trading protocol, such as tick size,
affect the speed at which new
information is incorporated into the
bid-ask spread. In this study, we
provide such evidence.
3
7/4/2003
1/4/2003
7/4/2002
1/4/2002
7/4/2001
1/4/2001
7/4/2000
1/4/2000
7/4/1999
1/4/1999
7/4/1998
1/4/1998
7/4/1997
1/4/1997
7/4/1996
1/4/1996
7/4/1995
1/4/1995
7/4/1994
1/4/1994
7/4/1993
7/4/2003
1/4/2003
7/4/2002
1/4/2002
7/4/2001
1/4/2001
7/4/2000
1/4/2000
7/4/1999
1/4/1999
7/4/1998
1/4/1998
7/4/1997
1/4/1997
7/4/1996
1/4/1996
7/4/1995
1/4/1995
7/4/1994
1/4/1994
7/4/1993
1/4/1993
0
1/4/1993
Percentage Spread
Dollar Spread
Figure 1
0.3
0.25
0.2
0.15
0.1
0.05
Figure 2
0.02
0.018
0.016
0.014
0.012
0.01
0.008
0.006
0.004
0.002
0
4
Research Questions


How quickly do specialist/dealer quotes
incorporate new information? Do
specialist quotes reflect changes in
stock attributes more quickly than
dealer quotes?
How is the speed of quote adjustment
related to stock attributes? Do stocks
with greater information-based trading
exhibit faster quote adjustments
toward optimal spreads and depths?
5


Do stocks that are traded in less
competitive markets (e.g., fewer
dealers) exhibit slower quote
adjustments?
Do liquidity providers move more
quickly to optimal spreads and
depths for stocks with more frequent
trading and higher return volatility?
6


Do smaller tick sizes result in faster
quote adjustments to new information?
What is the relation between quote
adjustment speeds and variable
measurement intervals?
7


Answers to these questions would be
of significant interest to market
regulators because they could help
design better market structures.
Because marketmaker quotes (i.e.,
bid-ask spreads) determine trading
costs, the speed at which specialists/
dealers adjust quotes to new
information is also of concern to
traders.
8
How our study differs from
previous studies?


Hasbrouck (1988, 1991a, 1991b),
Hasbrouck and Sofianos (1993),
Madhavan and Smidt (1993), Dufour and
Engle (2000) examine how marketmakers
adjust quote midpoints to signed trades.
Our study examines how quickly
marketmakers adjust quote width (i.e.,
the bid-ask spread) and depth (i.e.,
number of shares at the bid and ask) to
their optimal values in response to new
information.
9



Huang and Stoll (1996), Barclay (1997),
Bessembinder (1999, 2003c), and Chung,
Van Ness, and Van Ness (2001) compare
the execution costs of dealer and auction
markets.
Amihud and Mendelson (1987), Stoll and
Whaley (1990), Masulis and Ng (1995)
investigate the impact of market structure
on return volatility.
Affleck-Graves, Hedge, and Miller (1994)
compare components of the bid-ask spread
between auction and dealer markets.
10



Heidle and Huang (2002) examine the
impact of market structure on the
probability of trading with an informed
trader.
Garfinkel and Nimalendran (2003)
compare the impact of insider trading on
effective spreads between NYSE and
NASDAQ stocks.
However, none of these studies examine
how market structure affects quote
adjustment speeds on the NYSE and
NASDAQ.
11

Damodaran (1993) estimates the
speed of price adjustment for a
sample of NYSE and NASDAQ
securities using the partial adjustment
model of Amihud and Mendelson
(1987).

Thoebald and Yallup (2004)

We focus on spreads and depths
12
Some Conflicting Results


Jones and Lipson (1999) find that quotes
in NYSE- and AMEX-listed stocks adjust
more quickly to the information contained
in order flow than quotes in NASDAQlisted stocks.
Masulis and Shivakumar (2002) show that
price adjustments are faster by as much
as one hour on NASDAQ using a sample of
seasoned equity offering announcements
by NYSE/AMEX and NASDAQ companies.
13
Our Main Findings



The speed of quote adjustment on the NYSE is
faster than the speed of quote adjustment on
NASDAQ.
In both markets, quote adjustment speed is
faster for stocks with a larger number of
trades, higher share prices, greater return
volatility, and smaller trade sizes.
Stocks with greater information-based trading
and in more competitive trading environments
exhibit faster quote adjustments.
14



The speed of quote adjustment after
decimal pricing is significantly faster than
the corresponding figure before decimal
pricing in both markets.
Quote adjustment speed increases with the
length of variable measurement intervals.
On the whole, our study provides evidence
that stock attributes, market structure, and
tick size exert a significant impact on the
speed of quote adjustment.
15
Market Structure and Speed of
Quote Adjustment

Masulis and Shivakumar (2002) hold that quote
adjustment speed is likely to be slower on the
NYSE for several reasons.
• Limit orders on the NYSE cannot be updated
instantaneously or conditioned on public information (such
as the stock’s last transaction price) and this slow updating
of limit orders can delay revisions in the specialist’s bid
and ask quotes.
• NYSE specialists may buy stocks when prices are falling
because of their affirmative obligation to stabilize prices
and this behavior can slow quote adjustment process. The
specialists’ obligation to provide price continuity can
reinforce this effect because it requires them to go tick by
tick through the limit order book.
16
• Based on these observations and
the fact that NASDAQ is essentially
an electronic market in which
dealers do not have affirmative
obligations, Masulis and
Shivakumar conjecture that quote
adjustments on the NYSE are likely
to be slower than those on
NASDAQ.
17


Chung, Chuwonganant, and
McCormick (2004) show that a large
portion of order flow on NASDAQ is
either internalized or preferenced.
NASDAQ dealers do not have strong
incentives to make quick quote
adjustments in response to
information shocks.
18



Specialists on the NYSE can adjust
quotes quickly after each trade because
all order flow in a stock goes through
one specialist.
On NASDAQ however, dealers are less
able to make quick quote adjustments
to informed trading because one
informed trader can trade
simultaneously with several different
dealers before the quotes are adjusted.
Hence, NASDAQ dealers may be slower
in detecting information-based trading
than their counterparts on the NYSE.
19

Specialists on the NYSE may be able
to respond more quickly to changes
in informed trading because they
have face-to-face contact with floor
brokers while such contact is not
available to NASDAQ dealers because
NASDAQ operates on an electronic
screen-based system.
20

Garfinkel and Nimalendran (2003)
find that there is less anonymity on
the NYSE specialist system compared
to the NASDAQ dealer system. They
find that when corporate insiders
trade medium-sized quantities, NYSElisted stocks exhibit larger changes in
proportional effective spreads than
NASDAQ stocks.
21
Data



The NYSE’s Trade and Quote (TAQ)
database.
We use the trade and quote data for
the three-month period from
September 2002 to November 2002.
Applied various filters to minimize
data errors
22
23
Methodology

We estimate the following partial adjustment
model for each stock:
$SPREADi,t – $SPREADi,t-1
= a1i[$SPREAD*i,t – $SPREADi,t-1] + 1i,t; (1)
$SPREADi,t = the mean dollar spread of stock i
at time t
$SPREAD*i,t = the equilibrium dollar spread of
stock i at time t
24
Equilibrium Spread
$SPREAD*i,t = 0i + 1ilog(NTRADEi,t)
+ 2ilog(TSIZEi,t) + 3ilog(PRICEi,t)
+ 4iRISKi,t;
(2)
NTRADEi,t = the number of transactions
TSIZEi,t = the average trade size
PRICEi,t = the average share price
RISKi,t = the standard deviation of quote
midpoint returns
25
Substituting Eq. (2) into Eq. (1) and
after rearrangement, we obtain
$SPREADi,t – $SPREADi,t-1
= 0i a1i
– a1i$SPREADi,t-1
+ 1i a1ilog(NTRADEi,t)
+ 2i a1ilog(TSIZEi,t)
+ 3i a1ilog(PRICEi,t)
+ 4i a1iRISKi,t + 1i,t. (3)
26
We then estimate the model for each stock
using time-series observations:
$SPREADi,t – $SPREADi,t-1
= 0i + 1i$SPREADi,t-1
+ 2ilog(NTRADEi,t)
+ 3ilog(TSIZEi,t)
+ 4ilog(PRICEi,t)
+ 5iRISKi,t + 1i,t.

(4)
We measure the speed of quote adjustment
by the estimate of –1i.
27
28
29
30
Matched Sample

We calculate MS for each NYSE stock against each
of the 2,888 NASDAQ stocks in our study sample:
MS = [(XkN - XkY)/{(XkN + XkY)/2}]2,
where Xk represents one of the four stock
attributes and N and Y, refer to NASDAQ and
NYSE, respectively.

Then, for each NYSE stock, we select the NASDAQ
stock with the smallest MS.

This procedure results in 539 pairs of NASDAQ and
NYSE stocks with similar attributes.
31
Speed of Quote Adjustment and
Stock Attributes

Hypothesis 1: The speed of quote
adjustment is positively related to
both the number of trades and return
volatility.
Insofar as trades convey information on asset
values, liquidity providers may update quotes
more quickly for stocks that are more actively
traded and have greater return volatility.
32

Hypothesis 2: The speed of quote
adjustment is positively related to
share price.
Chung and Chuwonganant (2002) show that
the minimum price variation is more likely to
be a binding constraint on absolute spreads
for low-price stocks.
Liquidity providers make slower adjustments
toward optimal spreads for low-price stocks
because the binding constraint prevents
them from making such quote revisions.
33

Hypothesis 3: The speed of quote
adjustment is positively related to
both adverse-selection risks (and
costs) and the level of dealer
competition.
Liquidity providers are likely to make faster
quote adjustments to new information for
stocks with greater adverse-selection risks
(and costs). This is because the dealer cost of
quoting sub-optimal spreads is probably
greater for such stocks.
Similarly, we hold that liquidity providers
make faster quote revisions to optimal
spreads when competition is high
34
Measurement of adverse-selection
costs and risks


We use the spread component models
developed by Glosten and Harris (1988),
George, Kaul, and Nimalendran (1991),
and Lin, Sanger, and Booth (1995) to
measure adverse-selection cost.
We use the algorithm in Easley, Hvidkjaer,
and O’Hara (2002) to estimate adverseselection risk.
35
Glosten and Harris (GH) model
George, Kaul, and Nimalendran (GKN) model
Lin, Sanger, and Booth (LSB) model
36
Easley, Hvidkjaer, and O’Hara
(EHO)’s model
37
The likelihood function:
38
Regression Model
39
Regression Model
40
41
42
Does tick size affect quote
adjustment speed?


The NYSE initiated a pilot decimalization
program on August 28, 2000 with seven
listed issues. The NYSE converted all 3,525
listed issues to decimal pricing on January
29, 2001.
The NASDAQ Stock Market began its
decimal test phase with 14 securities on
March 12, 2001. All remaining NASDAQ
securities converted to decimal trading on
April 9, 2001.
43


Although there is extensive literature
on the effect of tick size on market
quality, there is little evidence as to
how tick size affects quote
adjustment speed.
In this study, we contribute to
existing literature by investigating
the impact of tick size on quote
adjustment speed using data before
and after decimal pricing.
44
Hypothesis: The speed of quote adjustment
during the post-decimal period is faster than
the speed during the pre-decimal period.
The penny tick size would be a binding
constraint less frequently than the predecimal tick size, allowing liquidity
providers to move toward optimal
spreads more quickly.
A smaller tick size results in greater
price competition because it implies a
smaller cost of both front running by
sell-side intermediaries and stepping
ahead of the existing queue by buyside traders.
45


For NYSE stocks, we consider the
three-month period from May 28,
2000 to August 27, 2000 as the predecimal period and January 30, 2001
to April 29, 2001 as the post-decimal
period.
For NASDAQ stocks, we consider the
three-month period from December
12, 2000 to March 11, 2001 as the
pre-decimal period and April 10, 2001
to July 9, 2001 as the post-decimal
period.
46
47
48
49
Regression Model
50
51
Speed of depth quote adjustment



Marketmakers post both the price and the
quantity of shares that they are willing to
trade.
The analysis of price quotes alone is likely
to show an incomplete picture of
marketmaker behavior.
We analyze how adjustment speed in
depth quotes varies with stock attributes
and tick size.
52
Estimating the Speed of Depth
Adjustment
53
Regression Model
54
55
56
Regression Model
57
58
59
Summary


Market structure exerts a significant impact
on the speed of quote adjustment. Liquidity
providers on the NYSE react more quickly
to new information than liquidity providers
on NASDAQ.
Liquidity providers make faster quote
adjustments new information for stocks
with greater adverse-selection costs and
quote competition.
60



Stocks with a greater number of trades,
greater return volatility, higher prices, and
smaller trade sizes exhibit faster quote
adjustments to new information.
Liquidity providers on both the NYSE and
NASDAQ react more promptly to new
information after decimalization. Large
tick sizes create friction in exchange
markets, and thus delaying price
discovery.
Quote adjustment speed increases with
variable measurement intervals.
61
Limitations and Future Studies


We assume that optimal spreads and
depths are determined by four stock
attributes and that liquidity providers
make quote adjustments accordingly.
To the extent that optimal spreads
and depths are also functions of
other variables, our empirical models
are subject to misspecification.
62


We assume that liquidity providers make
quote adjustments every 30 minutes
according to the target liquidity level (i.e.,
the optimal spread and depth) projected by
the value of four stock attributes during
the same 30-minute interval.
It would be a fruitful area for future
research to estimate the speed of quote
adjustment using more frequent
observations (e.g., every five minutes) and
to assess the robustness of the results.
63