02 Micronotes 070907
02 Micronotes 070907
02 Micronotes 070907
Abstract
In this article, we discuss the value of studying intraday trading data by providing
economic background in the area of market microstructure. We highlight the use
of ultra-high frequency data in the study of market design and quality as well as
asset pricing. To illustrate, we review recent research papers that utilizes intraday
trading data from the Stock Exchange of Thailand (SET) and explore future re-
search opportunities in this area in order to evaluate market policies and enrich our
understanding of the price discovery process on the Thai exchange.
Key words: adverse selection, bid-ask spread, liquidity, transaction costs, tick size
Introduction
Intraday data, once highly guarded by exchanges, has become more readily
available to the public over the years. Apparently, it is not the ever growing
power of the desk-top PC to process large scale datasets alone that has driven
a growing body of research utilizing ultra-high frequency data. Rather, it is
the need for regulators to evaluate market mechanics that best respond to
investors requirements and the need for academics to better understand the
relationships among market participants, trading costs, and trading process
in their quest for an alternative asset pricing paradigm.
Market microstructure deals with the trading of financial assets and the evo-
lution of asset prices by taking into account of transaction costs, incomplete
information, and heterogeneous expectations. The study of market microstruc-
ture requires transaction level data (intraday) that allows examination of
short- run price behavior that can lead to systematic mispricing in the long-
term. It allows assessment of the impact of trading mechanisms, for example,
1
tick rules, trading halts, and upper and lower price bounds, on market quality,
which means transparency and accurate price discovery.
The purpose of this article is to provide readers with basic economic back-
ground on market microstructure. It is not a complete review of all issues in
market microstructure 1 as we are admittedly biased towards topics which re-
lates to the structure of limit order markets and to recent research papers on
the microstructure of the Thai equity market. In the next part of the paper,
we discuss the role of equity markets, market types and trading protocol to
familiarize readers with the institutional set up of organized exchanges. In
section 2, we describe two basic modeling issues in market microstructure,
which are the price discovery process and theory of bid-ask spreads. Sections
3 and 4 provide discussion on market design, asset pricing, and review re-
lated literature in the area, particularly those utilizing SET’s intraday data.
Section 5 concludes the paper and propose directions for future research on
the microstructure of Thai capital markets.
There are two types of trading systems; a quote driven market and an order
driven market. In a quote driven market, buyers and sellers submit their bid
and ask offers to designated market- makers, also known as dealers or special-
ists. Based on information in their order books, the market-makers will post
bid-ask prices. Therefore, a quote driven market only displays the bid and
ask offers of designated market-makers. After individual orders are submitted,
the market maker will either fill in customer orders from its own inventory
or match the orders with another order. Quote driven trading is used in the
major US exchanges, ie. NYSE, AMEX, and NASDAQ and on the London
Stock Exchange (LSE).
Unlike quote driven markets, order driven markets, operate without the inter-
mediation of market makers. Instead, buyers and sellers submit the prices and
quantity which they are willing to buy or sell a security. These buy and sell
orders are displayed and accumulated in a limit-order book and order execu-
tion is usually prioritized based on price and time. A majority of exchanges
1 See O’Hara (1995), Madhavan (2000) and Stoll (2003) for very comprehensive
reviews. Also see special issue of International Review of Financial Analysis Vol 15,
2006 on Asian market microstructure.
2
around the world adopt order driven systems and utilizes computerized order
matching. In some markets such as NASDAQ, a hybrid system, employing
both dealer intermediation and direct crossing of individual orders, is used.
The use of hybrid systems allows participants to enjoy the transparency of
limit order book disclosure and the adequate provision of liquidity for illiquid
assets.
To curb excessive volatility, the SET imposes a 30% band on daily price fluc-
tuations relative to the previous day closing. A circuit-breaker is implemented
if the market index falls by 10% from the previous days’ close. There is also a
minimum price increment requirement for stock quotations. The current tick
rule in use is summarized in the following table.
Group Lower (From Bt) Upper (to less than Bt) Tick Size (Bt)
I <2 - 0.01
II 2 5 0.02
III 5 10 0.05
IV 10 25 0.10
V 25 50 0.25
VI 50 100 0.50
VII 100 200 1.00
VIII 200 400 2.00
IX 400 800 4.00
X 800 - 6.00
3
2 Modeling Issues
pt = pt−1 + εt
where pt is today’s price and pt−1 is yesterday’s price, and the random error
term, εt represents unexpected and unpredictable arrival of new information.
Alternatively, we can rewrite the above representation that the daily change
in price levels or rate of return, rt is driven by the random error term as in,
rt = ∆pt = εt
4
at once at the likelihood of a substantial premium. Thus, he is likely to split his
buys into small orders to reduce market price impact. The splitting of large
orders is one reason among others that leads to persistence in order flows
and continued price trends. To examine this problem more formally, suppose
investors receive public information set Ωt−1 at time t − 1 and decide to trade
on that information in the subsequent period, the expected quantity of trade
is defined as Et−1 [xt |Ωt−1 ] . Now suppose the actual trade quantity at period
is xt , then the difference xt − Et−1 [xt |Ωt−1 ] should reflect trade innovation
in response to unanticipated news during that trade period. Therefore, one
simple way to measure how fast the stock responds to new information is to
estimate the size of intraday quote revisions in response to order flows.
Table 1 reports the impact of signed trade on SCC and LANNA over three
lagged half hour time interval using 2003 data. Since we view the signed trade
volume as trade innovations to new information, the cumulative quote revi-
sions can be interpreted as information revealed in the trade. In sum, Table 1
shows the impact of trade innovation on SCC value is stronger particularly at
the contemporaneous level where a 1,000 share trade increases quote revision
in the same direction by 0.4%. By the third half hour after the trade, the cu-
mulative quote revision is down to 0.2%. On the other hand, LANNA’s quote
revisions appears less related to the direction of trade innovations.
Limit order markets function like an auction system where the buyer with the
highest bid and seller with lowest asking price wins the deal. An investor’s
order submission price depends on the bidder’s perception of the asset value.
In general, the value of assets can depend on independent private values where
other people’s bids and offers have no influence on a person’s valuation. For
example, at an art auction, an investor’s acquired taste for, let’s say, Aboriginal
3 The signed trade volume receives a positive (negative) value if it is buyer (seller)
initiated.
5
painting determines his private valuation and the maximum price he is willing
to pay to acquire the art piece. Other people’s bid prices for the same art piece
at the auction does not impact his private valuation. In contrast, affiliated
values are personal valuations that are positively related to the private values
of other bidders. The hype for internet stocks in the US in the late 1990s and
for Thai energy stocks during the oil price run-up in recent years are good
examples. The third type of valuation is common values which are intrinsic
values that are unknown to investors but will be the same across all bidders.
For example, the exploration rights for oil and gas in the gulf of Thailand.
E (V |s1 = 30 and s2 , s3, s4 < 30) . Conditional on the new information that the
average signal must be lower than 30 or E (s|s < 30) = 15, the first investor
adjusts his expectations down such that E (V |s1 = 30 and s2 , s3, s4 < 30) =
(30 + 15 + 15 + 15) /4 = 18.75. Although this example oversimplifies the process
of dynamic learning, it illustrates that past prices can provide valuation in-
formation about common values. As a matter of fact, when we see execution
price for any stock rising and at heavy trading volume and open interest, we
tend to raise our expectations of security values.
There exists a number of models that explains the dynamic learning process.
Earlier models based on dealership market can be found in the important
work of Glosten and Milgrom (1985) and Kyle (1985). These two models pro-
vide a framework of how a market maker learns about the intrinsic value of a
security over time by trading with informed traders in a sequential and contin-
uous trading model, repectively. Thus, the focus of such models is on Bayesian
6
learning and the risk of trading with informed traders. In a limit order struc-
ture, like the SET, the problem of trading with better informed investors still
holds 4 . At the same time, there are two additional complications, the risk
that a limit order fails to fill (execution risk) and the risk that the order takes
time to fill (picking-off risk). The latter can cause unexpected loss since the
investor fails to monitor his limit order and revise his order when there is a
change in security valuation.
vt = yt + ut . (2)
The common value changes as the investor learns new information, δ t , which
satisfies, Et−1 (δ t ) = 0 with ,
∆yt = δ t . (3)
h i
ψBU
t
Y
= Et Qt,T /qt |dBU
t,b
Y
= 1, qt . (4)
Note that the term Qt,T /qt , is the proportion of total order quantity filled.
The picking-off risk is defined as,
h i
ξ BU
t
Y
= Et (Qt,T /qt ) (yT − yt ) |dBU
t,b
Y
= 1, qt . (5)
Here the term, (Qt,T /qt ) (yT − yt ) captures potential profit or loss from change
in common values over the period T −t during which the trader fails to monitor
4 In their comparative study of Singapore and Thai market, Bailey, Mao, and
Sirodom (2007) show that foreign investors have better information processing than
local investors in the Thai market. Moreover, Bailey et al. (2006), find that both
foreign and local investors who trade cross-market possess superior information.
7
his limit order and update new information that affects the security’s common
value. The trader’s expected pay-off conditional on his information set is then,
h i ³ ´
Et (Ut ,T ) |dBUY
t,b = 1, qt , ut = qt ψBU
t
Y
vt − pBUY
t,b + qt ξ BUY
t − qt c (6)
From equation 6) if on average, the entire order is filled and common values
moves favourably during the execution period, the trader earns a positive
pay-off worth the submitted order quantity times excess value on immediate
execution plus the order quantity times the positive change in common values
minus the trading cost, which is order quantity times commissions rate, c. In
contrast, the trader can also earn a loss if common values move out of his
favour. The key to ensuring positive expected payoff depends on the accuracy
of the bidder’s private information and how this relates to future change in
common values. The higher the private signal, the more aggressive is the bid
and hence, the higher the probability of execution and the expected pay-off.
Investors usually frame market frictions only in the context of explicit trading
costs, which are commissions and taxes. In reality, market frictions consists
of implicit costs in microstructure. These includes bid-ask spreads, thin limit
order book, tick size, and price impact. We discuss here first the role of bid-
ask spreads. By definition the bid-ask spread is the difference between ask and
bid price. Since the SET displays the best (highest) three bids and and best
(lowest) offers available to the public, we typically refer to the inside spread
as the difference between the most aggressive bid and ask as shown in the box
below.
Inside Spread
| {z }
When submitting an order, the size of the spread tells us the price of imme-
diacy. Consider stock A with best bid at Bt 200, best offer at Bt 212, and
thus a spread of Bt 12. An investor whose private assessment of stock A is Bt
206 and requires immediacy will have to buy stock A at Bt 212 which is Bt 6
above his private valuation or sell stock A at Bt 200 or Bt 6 below his private
valuation. On the other hand, if stock A’s best bid and offer had been Bt 204
and Bt 208, respectively, his cost of immediacy would only have been Bt 2.
8
2.2.1 The determinants of bid-ask spreads
There are various reasons why bid-ask spreads exist. First, explicit trading
costs (commissions and taxes) generate spreads. From an economic perspec-
tive, an investor with private valuation of stock i worth vi will submit a bid
price bi such that vi −bi ≥ 0. In a market with no other frictions, bi should only
be a function of private value vi and fixed rate commissions, ψ, hence bi (vi , ψ)
. For instance, an investor’s private valuation of a stock is Bt 100. Let com-
missions be 0.5% of transaction value. The investor’s maximum bidding price
will be Bt 99.50 (100 - (100*0.005)). By symmetry, if the investor is a seller,
his optimal order submission strategy will be such that ai − vi ≥ 0.Therefore,
his minimum’s asking price will be Bt100.50 (100 + (100*0.005)). Unless pri-
vate valuations change, the switch between buying and selling activity will
cause observed trading prices to swing between Bt 99.50 and Bt 100.50. In
microstructure literature, this is known as the bid-ask bounce.
At this stage it is useful to discuss the intuition behind the size of ξ i . The
adverse selection premium or discount depends on how much investors know
about the stock in general. Larger firms that receive more visibility from more
analyst and news coverage tend to have lower asymmetric information prob-
5 Unlike dealership markets usually modeled in the US, the spread in a continuous
double auction market as the SET, need not be symmetric. The symmetry in spread
setting of dealership market is justified in Glosten and Milgrom (1985) famous
sequential trade model.
9
lem, and thus lower percentage bid ask spreads, ceteris paribus. In addition,
firms with larger price volatility, tend to have larger spreads. 6 This is because
volatility reflects uncertainty about values. In Figure 3, we show the per-
centage bid-ask spreads 7 of SET50 and non-SET50 groups by month during
the year 2003. The bar charts reveals a consistant trend that the percentage
spreads for non-SET50 stocks are about twice as high as the SET50 group.
10
can be classified as intra-main board trades and cross-board trades, It is pos-
sible to rewrite the above equation as,
³ ´ ³ ´
vt = vt−1 + αM S M /2 QM
t−1 + α
C
S C /2 QC
t−1 + εt (9)
where
QM
t = trade indictor variable for intra main board trades,
taking the value of +1/-1 if the trade is buyer (seller) initiated.
= 0 otherwise
C
Qt = trade indictor variable for cross-board trades,
taking the value of +1/-1 if the trade is buyer (seller) initiated.
= 0 otherwise
The adjusted equation allows values of adverse selection α’s to differ. We can
also show that the security price, pt is determined by its fundamental value
vt , its implicit trading costs or spreads and error term, ηt such that,
³ ´ ³ ´
pt = vt + S M /2 QM C C
t + S /2 Qt + η t (10)
³ ´ ³ ´ ³ ´³ ´
∆pt = αM S M /2 QM
t−1 + α
C
S C /2 QC M
t−1 + S /2 QM M
t − Qt−1
³ ´³ ´
+ S C /2 QC C
t − Qt−1 + εt + ∆η t (11)
The first two terms in equation (11) captures the information effect from
adverse selection. The larger the value of α’s implies a larger amount of asym-
metric information revealed by the trade. The 3rd and 4th terms measures the
bid-ask bounce. Rearranging equation (11) yields,
³ ´ ³ ´³ ´ ³ ´
∆pt = S M /2 Qt + αM − 1 S M /2 QM C
t−1 + S /2 Qt
³ ´³ ´
+ αC − 1 S C /2 QC
t−1 + et (12)
11
∆pt = α (Qt − ρQt−1 ) + β (Qt − Qt−1 ) + ut (13)
12
Taiwanese Stock Exchange and the SET to examine the characteristics of
stocks that frequently hit price limits. They find that small, actively traded,
and volatile stocks tend to hit price limits more often. In particular, they
note that actively traded stocks which carries more information flow hit price
limit frequently is further evidence that these price limits could potentially
be obstructing faster price adjustment. The work of Kim and Limphapayom
(2000) can be extended by exploring whether the same characteristics hold for
stocks with high intraday volatility.
Turning to the issue of tick size, a few recent papers, Pavabutr and Prangwat-
tananon (2007), henceforth PP (2007) and Boonvorachote, Charoenwong, and
Sirodom (2006a, 2006b), hence forth BCS (2006a, 2006b) examine the impact
of tick size change on transaction costs and liquidity. The study on tick size
merits analysis because its existence has both benefits and costs. 10 On one
hand, it creates additional transaction costs for traders on top of commission
fees. Suppose a stock has intrinsic value of Bt 100. Taking into account of
commission fees, the ask price that clears the fee should be Bt 100.50. How-
ever, since the minimum price variation for stocks between Bt 100 - Bt 200 is
Bt 1, the asking price must be set higher at Bt101. This new price reflects an
additional Bt 0.50 of discreteness related cost that a market order buyer must
pay. On the positive side, minimum price variations discourages front-running
and reduce bid-ask spreads ((Harris (1994), Goldstein and Kavajecz (2000),
Lau and McInish (1995), Niemeyer and Sandas (1994)).
PP (2007) evaluates the impact of exogeneous change in tick size when the
SET reduced minimum price variations for stocks priced below Bt 25 on No-
vember 25, 2001. The paper finds that tick reduction is associated with a
decline in spreads, and in quoted and accumulated market depths. However,
using relative comparisons to a control group, it appears that tick reduction is
only successful in increasing volume of stocks in the range from Bt 10 to less
than Bt 25, but not for those trading below Bt 10. BCS (2006 a, b) questions
the effectiveness of the SET’s multiple tick size regime by examining endoge-
neous changes in tick size on transaction costs as well as the order submission
strategy. Their findings support PP (2007) that small ticks are more effective
in improving market liquidity. However, since their study includes stocks in all
tick ranges, they also find evidence that endogeneous tick reduction in stocks
in the high price range of Bt 200 - Bt 400 or those with tick size above Bt1.0
also leads to substantial reductions in bid ask spreads. At the same time, they
show that tick size above Bt 1 enlarges the proportion of limit order submis-
10 Angel (1997) and Anshuman and Kalay (2002) provide an economic model for
optimal tick size. The models share the conclusion that the optimal tick must min-
imize trading costs. In Angel (1997), the optimal tick size is increasing in firms’
market values and idiosyncractic risks. In Anshuman and Kalay (2002) the optimal
tick depends on the trade-off between adverse selection and the price of discreteness.
13
sion, which tends to result in higher transaction costs for market order traders.
In combination, these findings suggest that the SET apply a uniform tick size
of Bt 1 for stocks above Bt 100 to reduce transaction costs of stocks in the
high price range.
Asset pricing has many dimensions. With parsimony in mind, asset pricing can
be viewed as three dimensional consisting of risk, return, and liquidity. While
we find the concepts of risk and return unambiguous to describe, liquidity is
actually a broader and more elusive concept than the familiar trading volume,
trading value, and turnover. This is because liqudity encompasses a number of
transactional properties of markets (see Kyle (1985)). These include tightness
(the cost of turning around a position over a short period of time), depth
(the size an order flow innovation required to change prices), and resiliency
(the speed which prices recover from a random, uninformative shock). Table
2, provides a list of possible measures of liquidity. These liquidity proxies
are all correlated to each other that it is difficult to pin-point what aspect of
liquidity each of them capture. Panel A of Table 3 reports the average selected
daily and intraday liquidity information of SET50 and non-SET 50 stocks
for the period 2003. The numbers clearly show that SET50 stocks are more
liquid than non-SET50 by both intraday and daily measures. While it is not
surprising that SET50’s trading value and volume is substantially greater, the
smaller percentage bid-ask spreads and intraday price impact, λ quantifies the
implicit trading costs. Panel B of Table 3 summarizes the correlation among
these liquidity variables and firm market capitalization. As shown, there is
clearly a negative relationship between market capitalization and measures
of trade friction ie. bid-ask spreads and intraday price impact. This finding
reminisces the work of Chaiyachantana et al., which uses institutional trading
data of 37 countries to analyze cross-country differences in price impact. They
find that price impact is an important composition of transaction costs and
that it is negatively correlated to firm size and positively correlated to the size
of the order.
Researchers include liquidity risk as part of equity pricing. Amihud and Mendel-
son (1986), Brennan and Subrahmanyam (1998), and Amihud (2002)) finds
that firms’ liquidity characteristics determines cross-sectional return varia-
tion. More specifically, expected return is decreasing in liquidity. This means
investors require compensation for holding illiquid stocks. Chordia, Roll, and
Subrahmanyam (2000) Huberman and Halka (2001) document strong liquidity
commonality suggesting that liquidity risk is systematic.
14
idence on the role of liquidity on the Thai Exchange is quite limited. On the
Thai exchange, Visalthanachoti et al (2006) examines the liquidity distribu-
tion or the shape of the limit order book 11 and its determinants. They find
that the convexity 12 in the shape of the order book is related to firm size.
Their finding is indicative that small firms with high return volatility is likely
to be more sensitive to adverse selection problems. Gorkittisunthorn et al.
(2006) extends this result by examining how bid-ask spreads change follow-
ing stock split events. They find that stocks with high insider ownership do
not exhibit significant improvement in bid-ask spreads, consistent with the
conjecture that the risk of adverse selection is negatively related to liquidity.
These findings is indication that we would expect firm size and their liquidity
characteristics should be related to their required returns.
Pavabutr and Sirodom (2007) also use split events in their study. However,
their focus is more specific on evaluating the impact of stock splits trading
activities and trading costs. They argue that existing empirical finding on stock
split impact on liquidity is ambiguous because liquidity needs to be assessed
from different aspects. Thus, the study provides analysis of pre- and post-split
changes in trading activities and price impact measures of trades. The study
finds successful splits, which are the ones that reduce price impact and raise
adjusted stock price, tend to be those that bring stock prices down to investors’
preferred trading range of Bt 25-50. This finding combined with PP (2007)
suggests that the clustering of stock trading around this price range could be
a result of it creating a balance between reducing the cost of discreteness while
at the same time preserving a reasonable margin for trading profits.
5 Conclusion
The SET has recently made intraday data available to academic research.
The quality and the nature of intraday data disclosure allows a number of
interesting research agendas. On issues of market design, further study on
the optimal trading price range and optimal tick size can provide us better
understanding in our search for optimal price variation rules that minimizes
trading costs and ensure orderly trading. To our knowledge, there is no existing
research that evaluates the effectiveness of SET’s use of daily price limit of
30%. The daily price limit is non-existing in some markets whereas others
15
impose an even tighter daily price variation than the SET.
On issues of asset pricing, the role of liquidity, their intraday proxies, and
their relationship between price discovery can be assessed in greater detail.
In particular, a number of stocks on the exchange is liquidity-driven given
the relatively low free float and small market capitalization. The disclosure
of trader types allows exploration into traders’ strategic behavior, their re-
sponses to information, and their trade impact on the exchange. We have only
just begun. Studies in market microstructure accommodates for the unique
characteristics of the Thai market that will surely enrich our understanding of
our own market where traditional economic theories fail to explain observed
empirical anomalies.
References
[1] Admati, A.R, and P. Pfleiderer, 1988. A theory of intraday patterns: Volume
and price variability, Review of Financial Studies 1, 3-40.
[2] Ahn, H.J., Y. Hamao, and J. Ho, 2002. The components of the bid-ask spread
in a limit-order market: Evidence from the Tokyo Stock Exchange, Journal of
Empirical Finance, 399-430.
[3] Amihud, Y., and H. Mendelson, 1986. Asset pricing and the bid ask spread,
Journal of Financial Economics, 17, 223-249.
[4] Amihud, Y., 2002. Illiquidity and stock returns: cross-section and time-series
effects, Journal of Financial Markets 5, 31-56.
[5] Angel, J., 1997. Tick size, share prices, and stock splits, Journal of Finance, 52,
655-681.
[6] Anshuman, V.R., A. Kalay, 2002. Can splits create market liquidity? Theory
and evidence, Journal of Financial Markets, 5, 83-125.
[9] Bailey, W., C. Mao, and K. Sirodom, 2007. Investment restrictions and
the cross-border flow of information: Some empirical evidence, Journal of
International Money and Finance, 26, 1-25.
16
[10] Bailey, W., C. Mao, and K. Sirodom, 2006. Local, foreigners, and multi-market
trading of equities: Some intraday evidence. Cornell University Working paper.
[12] Chen, H., 1998. Price limits, overreactions, and price resolutions in futures
markets, Journal of Futures Markets, 18, 243-263.
[15] Glosten, L., and Milgrom, 1985. Bid, ask, and transaction prices in a
specialist market with heterogeneously inMormed traders, Journal of Financial
Economics, 13, 71-100.
[16] Goldstein, M., and K. Kavajecz, 2000. Eighths, sixteenths, and market depth:
Changes in tick size and liquidity provision on the NYSE, 56, 125-149.
[18] Harris, L., 1994. Minimum price variations, discrete bid-ask spreads, and
quotation sizes, Review of Financial Studies, 7, 149-178.
[19] Hasbrouck, Joel, and Duane Seppi, 2001. Common factors in prices, order flows,
and liquidity, Journal of Financial Economics 59, 383-411.
[20] Hollifield, B., R. Miller, and P. Sandas, 2004, Empirical analysis of limit order
markets, Review of Economic Studies, 71, 1027-1063.
[21] Huang R.., and H.Stoll, The components of the bid-ask spread: A general
approach, Review of Financial Studies, 995-1034
[22] Huberman, Gur, and Dominika Halka, 2001. Systematic liquidity, Journal of
Financial Research 26, 161-178.
[23] Kim, K.A., Rhee, S.G., 1997. Price limit performance: evidence from the Tokyo
Stock Exchange, Journal of Finance 52, 885-901.
[24] Kim, K.A., and P. Limpaphayom, 2000. Characteristics of stocks that frequently
hit price limits: Empirical evidence from Taiwan and Thailand, Journal of
Financial Markets, 3, 315-322.
[25] Kyle, A. 1985, Continuous auctions and insider trading, Econometrica, 53, 1315-
1336.
17
[26] Lau, S.T., and T. McInish, 1995, Reducing tick size on the Stock Exchange of
Singapore, Pacific Basin Finance Journal, 3, 485-496.
[30] Neimeyer , J., and P. Sandas, 1994, Tick size, market liquidity, and trading
volume: Evidence from the Stockholm Stock Exchange, Working paper.
[31] Pavabutr, P. and S. Prangwattananon, 2007, Tick size change and clientele
perception of transaction costs on the Stock Exchange of Thailand. Stock
Exchange of Thailand Working Paper
[32] Pavabutr, P., and S. Sirodom, 2007, What stock splits do to liquidity and
valuation of Thai firms, Stock Exchange of Thailand Working Paper.
[33] Visaltanachoti, N., C., Charoenwong, and D. Ding, Liquidity distribution in the
limit order book on the Stock Exchange of Thailand, forthcoming, International
Review of Financial Analysis.
18
Figure 1: Components of bid-ask spreads
This figure shows the basic components of the bid-ask spreads around stock i’s common private value,
vi. The spread consists of transaction costs, ψ and adverse selection cost, ξ. In reality, the spread need
not be symmetric as it is possible for ξ b ≠ ξ s in a double auction.
ai
Seller’s transaction cost, ψ
vi
Buyer’s adverse selection cost, ξb
Figure 2: Percentage Bid-ask spreads of SET 50 vs Non SET 50 stocks for 2003
This figure shows the monthly average percentage bid ask spreads by SET50 and non-SET50 groups.
The percentage bid-ask spread is computed from Quoted spread divided by the mid-point (bid+ask)/2.
4.50%
4.00% SET50
Non-SET50
3.50%
3.00%
2.50%
%BAS
2.00%
1.50%
1.00%
0.50%
0.00%
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Source: Data source from Stock Exchange of Thailand, spreads computed by authors
19
Table 1: Intraday Quote Revisions
This table provides the estimates for corresponding the vector autoregressions corresponding to the
models
rt = c0 xt + c1 + xt −1 + c 2 + xt −2 + c3 + xt −3 + u t
xt = d1 xt −1 + d 2 xt −2 + d 3 xt −3 + wt
where rt is the average percentage quote revision in each half hour time interval and xt is the signed
trade volume in thousand shares receiving values of for buy (sell) initiated trades. The cumulative
quote revision is the sum of ci through the third time interval. Two stocks are selected, Lanna Lignite
(LANNA) and Siam Cement (SCC). Only 2003 data is used.
LANNA SCC
Time Quote Cumulative Quote Cumulative
Interval revision P value Quote rev revision P value Quote rev
0 -0.009 0.782 -0.009 0.004 0.000 0.004
1 -0.002 0.961 -0.010 -0.001 0.484 0.003
2 0.002 0.961 -0.009 -0.001 0.360 0.002
3 -0.007 0.827 -0.015 0.000 0.867 0.002
This table lists common measures of daily and intraday liquidity. For the daily price impact measure,
Di is the number of trading days in the sample, Ri is stock i return, and TVALi trading value of
t
stock i. For intraday price impact measure, q t is the order flow, Dt is signed trade, which receives
value of +1 if it is buyer initiated and -1 if seller initiated. These proxies of liquidity are likely to be
related to resilience. Alternatively, we can measure resilience by volatility persistence.
Intraday
Bid-ask spreads Difference between ask and bid Transaction costs, tightness, information
outstanding (see details in glossary) asymmetry
Intraday price impact Δpt = λ ⋅ qt + ϕ ⋅ (Dt − Dt −1 ) + y t Transaction costs, Inverse of market
(λ) depth
Limit order book depth Amount of buy and sell order Tightness, flow of information,
quantities at the inside spread. information asymmetry
20
Table 3: Selected Daily and Intraday Measures of Liquidity (2003 only)
Panel A of this table shows selected average daily and intraday measures of liquidity for SET50 and
non-SET50 stocks. Data from 2003 only. Due to missing data or insufficient number of observation in
2003, the number of firms in the SET50 and non-SET 50 is reduced to 38 and 203, respectively. Panel
B shows the correlation among the liquidity variables.
Panel A
Panel B
21