Index Arbitrage and Futures Pricing Efficiency: Evidence from Thailand1

Satjaporn Tungsong2 and Gun Srijuntongsiri3

Thammasat Business School & Sirindhorn International Institute of Technology, Thailand

October 25, 2011

Abstract

Using both daily and intraday data, this research examines the contribution of the
exchange-traded fund TDEX to pricing efficiency of the SET50 futures with respect to the
SET50 index. In order to analyze the efficiency of the SET50 futures price, frequencies of
mispricing and arbitrage as well as arbitrage profitability are measured. In particular, the
following three arbitrage trades are analyzed: (1) SET50 futures vs. TDEX, (2) SET50 futures
vs. SET50 component stocks, and (3) TDEX vs. SET50 component stocks. Empirical evidence
indicates that the introduction of TDEX does not contribute to pricing efficiency of SET50
futures with respect to the cash index.

Keywords: Index arbitrage; Liquidity; Exchange-traded fund; Index futures; Stock Exchange of
Thailand

JEL classification: G12, G13, G14

1. Introduction
                                                                                                                       
1
Supported by the Stock Exchange of Thailand
2
Corresponding author
Thammasat Business School, Thammasat University, 2 Prachan Road, Pra Nakorn, Bangkok, 10200, Thailand
Email: satjaporn@tbs.tu.ac.th, s.tungsong@gmail.com
3
Sirindhorn International Institute of Technology, Thammasat University, 131 Moo 5 Tiwanont Road, Pathumthani,
12000, Thailand
Email: gun@siit.tu.ac.th  

1
 

Electronic copy available at: http://ssrn.com/abstract=1915282

 Thailand’s derivatives market is at its first stage of development, with its first financial
derivative security, the SET50 futures, begins trading in April 2006. The introduction of the
SET50 futures in April 2006 allows investors to replicate the return of the SET50 index more
efficiently and cost-effectively. Over a year later, Thailand’s first exchange-traded fund, the
ThaiDEX SET50, hereafter TDEX, begins trading in September 2007.
As the completeness of the market for index-tracking securities increases with the
introduction of the TDEX ETF, is intuitive to consider the TDEX trading a contribution to higher
efficiency of the SET50 futures price. Index arbitrage activities are key to reaching such futures
pricing efficiency. In the purest form of market efficiency, mispricing in a futures price will be
eliminated once arbitrageurs step in. Furthermore, arbitrage activities are shown to be crucial in
improving risk sharing amongst investors. Through trading the mispriced securities, arbitrage
induces market clearing even with the presence of large investor diversity (Basak and Croitoru,
2001; Basak and Croitoru, 2006). As such, index arbitrage is an important activity, contributing
to the equilibrium in which the index futures price is efficient with respect to the cash index.
Restrictions to arbitrage thus prevent participation from arbitrageurs, upsetting efficient pricing
mechanism of the index futures.
In developed markets, namely the U.S. S&P 500, and NASDAQ, and the French CAC40,
trading of ETFs is proven to increase index-futures pricing efficiency (Switzer et al., 2000;
Kurov & Lasser, 2002; and Deville et al., 2010). Will this be the case for emerging markets?
Despite a number of literatures on efficiency of index-futures pricing, a few focus on
pricing efficiency of index futures in emerging markets. Mall et al (2011) and Gupta and Singh
(2009), among others, study the efficiency of India’s stock index futures market. To the best of
our knowledge, no existing studies investigate pricing efficiency in other emerging markets
including Thailand. Difficulty in obtaining intraday data on the SET50 futures is a main obstacle
to conducting such study. At the beginning of our research, Thailand’s Futures Exchange
(TFEX) has not made public intraday data on available derivative securities.
The contribution of our research comes in the following four aspects. First of all, the
results strengthen existing literatures on the role of liquidity in pricing efficiency. More
specifically, thin SET50 futures trading volume resulting from insufficient institutional
participation in the Thai market contributes to the persistence of mispricing in the SET50 futures.

2
 

Electronic copy available at: http://ssrn.com/abstract=1915282

 Secondly, this research is first to study the effect of Thailand’s TDEX ETF on the SET50
futures pricing efficiency. In this respect, it is the first study to focus on the effect of ETF
trading on index futures pricing efficiency in an emerging market. As emerging markets are
becoming more attractive especially among global institutional investors, this research will more
or less satisfy the demand for knowledge about emerging markets. Pavabutr and Sirodom (2010)
report a 47 percent growth in the number of Asian country funds from 2002-2009, with the
growth driven by China and India. For Thailand, 23 mutual funds with average net asset of USD
22 million flow into the country in 2009.
Thirdly, the synchronicity of stock prices and the correlation of fund flows in emerging
markets make possible generalization of our study to understanding other emerging economies.
Morck et al. (2000) report high correlation of stock prices in countries with low per capita GDP
and poor investor protection. In particular, China, Malaysia and Poland witness over 80 percent
of stocks move in the same direction in a given week. Moreover, Pavabutr and Sirodom (2010)
find that, except for Malaysia, the flow of funds to Thailand is positively correlated with those to
other Asian markets, especially India and Taiwan with correlation of 0.35 and 0.29 respectively.
Fourthly, the uniqueness of the Thai market allows our study to contribute to existing
literature in various aspects. The current structure of the Thai market is retail dominated, with
approximately 70 percent of total trades made by retail investors (Pavabutr and
Prangwattananon, 2009). The likelihood of passing a law to liberalize financial institutions in a
few years’ time will largely increase the number of securities companies and the scope of their
activities. As Thailand has seen no growth in the number of brokerage firms due to the fixed
availability of brokerage licenses, such liberalization will significantly facilitate the formation of
new brokerage firms and will sharply increase institutional participation, local and foreign, in the
near future.
The aim of this paper is to test the hypothesis that the TDEX ETF contributes to more
efficiency in the pricing of SET50 futures with respect to the SET50 index. Conventional tools
for testing futures market efficiency are employed in this study. More specifically, we measure
the frequency and magnitude of mispricing as well as the frequency and profitability of arbitrage
from the following tests: (1) SET50 futures vs. SET50 index component stocks, (2) SET50
futures vs. TDEX, and (3) TDEX vs. SET50 component stocks. Daily and intraday price data of
the SET50 futures, SET50 index, TDEX, and SET50 index component stocks are used.

3
 

Electronic copy available at: http://ssrn.com/abstract=1915282

Moreover, transaction costs and execution lags are incorporated in the measurement of arbitrage
frequency and profit.
The main findings of this research are as follows. First, both the SET50 index and
SET50 futures are more volatile after the TDEX launch. The SET50 futures are mispriced
against the index more frequently and the average mispricing magnitude increases after the
TDEX launch. In other words, high volatility after the TDEX introduction results in higher
mispricing frequency and larger mispricing magnitude of the SET50 futures.
Second, the TDEX is generally overpriced while SET50 futures are generally underpriced
with respect to the SET50 index. Comparing the SET50 futures to TDEX, one finds that the
futures are often underpriced against the ETF.
Third, sensitivity analysis reveals that, as the TDEX dividend yield increases, the SET50
futures will likely be overpriced against the TDEX. The TDEX dividend yield and arbitrage
profitability are also positively related.
Fourth, arbitrage profit is generally lower after the introduction of TDEX, the period
characterized by higher volatility (of the SET50 index and SET50 futures). Although the SET50
futures are mispriced (with respect to the cash index) more frequently after the TDEX
introduction, the number of profitable trades decreases. The average mispricing magnitude
increases while average profit decreases after the TDEX launch. This empirical evidence aligns
with the theory of limits to arbitrage in that higher volatility of the underlying index increases the
likelihood of adverse price movements, which eventually leads to lower arbitrage profitability
(Yadav and Pope, 1994).
Fifth, arbitrage involving trading all the SET50 stocks against either SET50 futures or
TDEX generally generates no profit. Furthermore, loss from trading the SET50 stocks against
SET50 futures increases after the TDEX introduction. The loss incurred can be attributed to
adverse price movements of the stocks. Once again, higher volatility of the SET50 stocks post
TDEX launch exacerbates adverse price movements, which eventually lead to higher arbitrage
loss. Moreover, as the SET50 is a market value weighted index, the proportion of each stock in
the index varies daily, if not constantly. Arbitrage trades in which the holding proportion of each
stock remains constant from position open to position close are unlikely to generate profit. On
the contrary, arbitrageurs may generate profit if they adjust their holding in each stock
periodically.

4
 
 Sixth, arbitrage mispricing and profit near futures maturity are generally higher than
those in periods further from maturity. A period near futures maturity is characterized by both
higher SET50 futures volatility and trading volume. The theory of limits to arbitrage suggests
that higher volatility will result in lower arbitrage profit due to higher possibility of adverse price
movements (Yadav and Pope, 1994) but higher liquidity will ease the establishment of arbitrage
position, leading to higher arbitrage profitability (Roll et al., 2007). Thus, the effect of liquidity
seems to outdo that of volatility in this case.
Lastly, transaction costs and execution lags negatively affect mispricing frequency and
profitability of index arbitrage. In other words, the mispricing frequency and arbitrage profit
decreases as transaction costs and execution lags increase. For intraday trades, arbitrage profit
disappears after 10 minutes execution lag.
The paper is outlined as followed. Section 2 introduces the three classes of SET50 index-
tracking securities. Section 3 reviews existing literature on index arbitrage and limits of
arbitrage. Section 4 of this paper discusses the research’s hypothesis development. Sections 5
and 6 outline the data and methodology used. Sections 7 and 8 present daily and intraday results.
Section 9 discusses the findings and Section 10 concludes.

2. Three classes of index-tracking securities

The SET50 index, frequently used as the indicator of the Thai equity market
performance, comprises of the top 50 stocks according to the following three criteria: market
capitalization, liquidity, and compliance to the regulations set forth by the Stock Exchange of
Thailand.
Before the introduction of the SET50 futures4 in 2006, investors must trade all 50
individual stocks in order to replicate the performance of the Thai equity market. Trading all the
50 component stocks is costly, however. With the SET50 futures, tracking the Thai equity
market performance becomes easier and less costly.

                                                                                                                       
4
 The SET50 futures contract is a derivative instrument whose price is derived from that of its underlying asset, the
SET50 index.  

5
 
 In 2007, the ThaiDEX SET50 exchange-traded fund (TDEX) was introduced. The
TDEX is an open-ended fund investing mainly in the SET50 component stocks5. Similar to the
SET50 futures, the TDEX tracks the performance of the SET50 index. However, trading a unit
of the TDEX requires significantly lower initial investment than trading a SET50 futures
contract. The price of a unit of TDEX is 0.01 of the prevailing SET50 index level. On the other
hand, the size of a SET50 futures contract is 1000 times the prevailing index level and trading a
futures contract requires putting up an initial margin that is normally 10 percent of the notional
amount. Moreover, trading the TDEX is more convenient6 than trading the SET50 futures.
As commonly known, the liquidity of an ETF comes from two channels: the primary and
secondary markets. In the primary market, authorized participants purchase the underlying
basket of securities and deposit the basket in kind into the ETF. This creates more shares in the
ETF. In the secondary market, liquidity comes from investors buying and selling the ETF.
ETFs are believed to ease the establishment of arbitrage positions due to their lower costs and
risk. As such, the introduction of ETFs would generally result in increased arbitrage activity and
tightened spot-futures relationship7.
The prices of the SET50 futures and TDEX do not always correspond to that of the
SET50 index. The purest form of market efficiency suggests that, when price discrepancy
occurs, arbitrageurs come in and buy the underpriced instruments and simultaneously short-sell
the overpriced instruments. Profit is reaped as arbitrageurs reverse the trades when price
discrepancy disappears. As such, arbitrage activity will correct the mispricing and eventually
bring in market efficiency.
Traditional index arbitrage strategies involve trading the futures and component stocks of
the index. With ETFs in place, index arbitrage can be carried out by trading different
combinations of ETF, futures, and stocks.
With respect to transaction costs, investors are charged with a commission fee between
0.15-0.25 percent of the matched value (price) in a one-way stock transaction. To trade a SET50
futures contract, an investor incurs transaction costs that include exchange and clearing fees
                                                                                                                       
5
A much smaller portion of the fund is invested in bonds and kept in cash.
6
ETF is traded like a stock. Investors can use their equity account to trade. On the contrary, investors need to open
a derivatives account to trade SET50 futures. The account opening process can take from one week to over a month,
as documents need to be verified and available trading amount need to be approved.
7
However, this research’s findings indicate that the TDEX does not contribute to higher SET50 futures pricing
efficiency.
 

6
 
(THB50 per contract or approximately 0.01 percent of contract notional value if the index level
is at 500) and brokerage commission (THB250-450 per contract or approximately 0.06-0.1
percent of notional value if the index level is at 500). In the case of TDEX, investors pay front-
end fee (0.05 percent of NAV per unit), back-end fee (0.05 percent of NAV per unit), brokerage
fee (0.1-0.2 percent of NAV per unit), management fee (0.4 percent of NAV per unit), trustee fee
(0.015-0.025 percent of NAV per unit), and registrar fee (0.01 percent of NAV per unit). The
TDEX management, trustee, and registrar fees are paid per year or pro-rata.
In total, a one-side stock trade incurs 0.15-0.25 percent of traded value (round-trip fee
totals 0.3-0.5 percent). A one-side SET50 futures trade incurs THB300-500 per contract or 0.07-
0.11 percent of notional value if, for example, the index level is at 500 (round-trip fee equals
THB600-1000 per contract or 0.14-0.22 percent of the notional value). A one-side TDEX trade
incurs 0.15-0.25 percent of NAV per unit, excluding management, trustee, and registrar fees
(round-trip fee totals 0.3-0.5 percent of NAV per unit, excluding management, trustee, and
registrar fees).

3. Literature review
3.1 Index arbitrage and futures pricing efficiency
Empirical studies on the spot-futures pricing relationship generally indicate that
deviations of observed futures price from the theoretical price implied by the cost-of-carry model
are significant (Chung, 1991; Klemkosky and Lee, 1991; Neal, 1996; Bae, Chan, and Cheung,
1998). Even after taking into account transaction costs, arbitrage mispricing and profit still exist
(Chung, 1991; Klemkosky and Lee, 1991; Bae, Chan, and Cheung, 1998).
Chung (1991) studies pricing efficiency in the market for stock index futures by
investigating index arbitrage between the Chicago Board of Trade's Major Market (MMI) index
and the MMI futures. His results indicate that the MMI futures-cash index arbitrage became less
profitable as the market matured. Profitable arbitrage trades account for less than 50 percent of
all executed trades.
Klemkosky and Lee (1991) incorporate mark-to-market effects into their investigation of
S&P500 futures pricing efficiency. In measuring intraday arbitrage profit, the authors assume
that mark-to-market cash inflows are invested at the risk-free rate while cash outflows are

7
 
financed at the call loan rate or 1 percent above the call loan rate. The profitability, after taking
into account the proceeds from mark-to-market, of a long arbitrage trade increases. On the
contrary, the profit of short arbitrage decreases. The authors explain such phenomena with the
general uptrend of stock price during most of the test period. They also find that taxes reduce the
mispricing frequency substantially and that arbitrage trades are still profitable 10 minutes
following the mispricing signal.
Neal (1996) examines the characteristics of 837 arbitrage trades between S&P500 Index-
S&P500 futures. He finds that, even though restrictions on short selling can be overcome in the
case of institutional investors who can implement direct sales, the S&P500 index arbitrage is in
fact not risk-free. He estimates the opportunity cost of arbitrage funds to be 88 basis points
exceeding the Treasury bill rate. Furthermore, there is an indication that the index arbitrage will
not be profitable once bid-ask spreads are included. He concludes that such small profit is a
result of market efficiency in which arbitrageurs quickly exploit the price discrepancies. Lastly,
arbitrage positions are mostly reversed prior to expiration; very few are held until expiration.
Bae, Chan, and Cheung (1998) use bid-ask prices to measure index arbitrage profit
involving the Hang Seng Index futures and options in the Hong Kong market. The authors find
that, by using transaction prices (as opposed to bid-ask prices) previous studies overestimate
arbitrage profit. Furthermore, they find that the magnitude of mispricing increases when
liquidity is low (wider bid-ask spread). However, larger mispricing does not necessarily imply
profitable arbitrage opportunity. The appearance of some arbitrage opportunities is due to the
fact that arbitrageurs will not step into the market when bid-ask spread is large.
Among the limited availability of research on efficiency of index futures in emerging
markets, Mall et al (2011) and Gupta and Singh (2009) study the efficiency of India’s stock
index futures market. They find that the Indian futures market is efficient.

3.2 Effects of exchange-traded funds (ETFs) on index futures pricing efficiency

A handful of research study the effect of ETFs on index futures pricing efficiency
through measuring mispricing and index arbitrage profit (Switzer, Varson, and Zghidi, 2000;
Kurov and Lasser, 2002; Ritchie, Daigler, and Gleason, 2008; Deville, Gresse, and Severac,
2010). They find that ETFs lead to improvement in pricing relationship between the futures and

8
 
cash index demonstrating in 1) the decrease in size of mispricing, 2) shorter time period in which
mispricing lasts, and 3) the decrease in arbitrage profitability.
Switzer et al. (2000) investigate the effects of the Standard and Poor’s Depository
Receipts (SPDRs) on the pricing efficiency of the S&P500 futures using daily and intraday price
data. They find that the positive mispricing of the S&P500 futures is generally reduced after the
SPDRs begin trading. The authors also conclude that while the dividend yield and time-to-
maturity biases on the S&P500 futures mispricing still exist after the introduction of the SPDRs,
the effects of the biases are reduced.
Kurov and Lasser (2002) use tick-by-tick data to study the impact of the Nasdaq-100
Index Tracking Stock8 on the mispricing relationship between the Nasdaq-100 index and the
Nasdaq-100 futures. They find that the size and frequency of futures-spot mispricing was
reduced and the arbitrage opportunities in the spot-futures markets disappeared more quickly
following the ETF introduction. The authors reason that the ETF allows arbitrageurs to establish
an index arbitrage position more easily and less costly.
Richie, Daigler, and Gleason (2008) extend prior research on arbitrage involving the
SPDRs and the S&P500 futures by incorporating the securities’ actual trade volumes. They find
that using either the SPDRs or the S&P500 index leads to profitable arbitrage trade.
Furthermore, the S&P500 futures is generally underpriced against the SPDRs and overpriced
against the S&P500 cash index. The underpricing disappears more quickly than the overpricing
as transaction costs increase. Moreover, there are more price discrepancies during high volatility
months than low volatility months.
Deville, Gresse, and Severac (2010) investigate index arbitrage between the French
CAC40 index and the Lyxor CAC40 ETF using intraday trade data. They find that the
mispricing in the futures is reduced both in terms of magnitude and frequency following the
inception of the ETF. They conclude that the ETF is a valuable lever for increasing arbitrage
activity and the degree of market efficiency. However, their results indicate that the ETF
turnover does not explain such improvement. They argue that the impact of CAC40 ETF
liquidity may be indirect.

                                                                                                                       
8
The Nasdaq-100 Index Trading Stock is an exchange-traded fund tracking the performance of the Nasdaq-100
index. It is colloquially referred to as Cubes.

9
 
 3.3 Limits to arbitrage, liquidity, and price discrepancy
Financial economists generally rationalize price discrepancies as a consequence of two
phenomena: demand shocks and limits to arbitrage (Gromb and Vayanos, 2010; Baberis and
Thaler, 2003; Attari and Mello, 2006). Demand shocks are defined as market pressures that
drive away an asset’s price from its fundamental value. There are several factors including
institutional frictions, contracting costs, and agency costs that contribute to such demand shocks
(Gromb and Vayanos 2010). Limits to arbitrage are market frictions such as brokerage fees,
commissions, short-selling costs, illiquidity, market impact costs, and execution difficulties, to
name a few. The existence of arbitrage limits prevents arbitrageurs from actively participating in
arbitrage trading. In perfect and complete markets in which there are few to no frictions,
arbitrageurs trade the mispriced securities profitably; this is not the case in markets with high
arbitrage limits. As arbitrageurs only trade to exploit the observed mispricings only if they are
large enough to compensate for transaction and holding costs9 (Tuckman and Vila, 1992), one
witnesses fewer arbitrage activities in high-friction markets than in low-friction markets. As
arbitrageurs trade, the securities’ mispricings are corrected and the market eventually reaches
equilibrium. Without arbitrageurs’ participation, price discrepancies will persist.
Mitchel et al. (2002) conclude that limits to arbitrage causing prices to not converge to
fundamental value will force arbitrageurs who cannot diversify away this risk to invest less.
Furthermore, even if prices eventually converge to fundamental values, the path of convergence
may be long and bumpy. If the arbitrageurs do not have access to additional capital when the
security prices diverge, they may be forced to prematurely unwind the position and incur a loss.
These together imply that limits to arbitrage lead to lower arbitrage activity.

Roll et al. (2007) study the NYSE Composite index and its futures and find that over the
lifetime of a futures contract, the cash-futures basis mean-reverts faster when the market is more
liquid. They conclude that liquidity facilitates the movement of futures and cash prices towards
the frictionless zero cash-futures basis. In other words, persistence in price deviations can be
explained by illiquidity as a limit to arbitrage.
Chordia et al. (2008) examine the predictability of short horizon returns from lagged
order flows of stocks traded on the NYSE. They find that liquidity, manifested in narrower bid-
ask spreads, diminishes the predictability of short horizon return from past order flows and
                                                                                                                       
9
Holding costs are costs associated in establishing an arbitrage positions.

10
 
eventually stimulates market efficiency. They conclude that arbitrage trading should be
extensive and effective during time of high liquidity.
Chen et al. (2009) find that reduction in liquidity, especially market depth (due to
decimalization of the S&P500- and NASDAQ-related index tracking securities), decreases index
arbitrage profit. As a result, arbitrageurs are discouraged from engaging in arbitrage trading and
the market eventually faces lower pricing efficiency on average.

4. Hypothesis development
Existing research finds that trading of the ETFs increases index-futures pricing efficiency
as evidenced in the cases of the U.S. S&P 500 and NASDAQ, and the French CAC40. The
results may or may not hold in emerging markets in which market environments are dissimilar
from those of developed markets. Among other differences, emerging markets are known to be
significantly less complete and less liquid. As liquidity is proven to stimulate market efficiency
and higher liquidity is related to increased arbitrage profit, it is not counterintuitive to presume
that ETF trading may not induce futures pricing efficiency in such illiquid environments as
emerging markets. Therefore, we hypothesize that ETF trading in Thailand does not contribute
to pricing efficiency of the index futures. More specifically, our hypothesis is that we do not
believe that the TDEX trading leads to pricing efficiency of the SET50 futures under Thailand’s
illiquid market condition.

5. Data
Daily closed prices and intraday deal prices of the following financial instruments are
collected from the Stock Exchange of Thailand (SET) and Thailand’s Futures Exchange (TFEX)
databases. The data include:
(1) Daily and intraday prices of SET50 futures from April 200610 to most recent,
(2) Daily and intraday prices of TDEX from September 200711 to most recent,

                                                                                                                       
10
April 2006 is the inception month of the SET50 futures.
11
September 2007 is the inception month of the TDEX.

11
 
 (3) Daily and intraday prices of the SET50 index component stocks from April
2006 to most recent, and
(4) Daily prices of the SET50 index from April 2006 to most recent12.

For futures contracts, only near-maturity contracts are used in order to prevent illiquidity
problem. The average 2007-2009 dividend yield of 2.17 percent for the TDEX is used as a
proxy for expected dividend on the ETF throughout the testing horizons. The average 2006-
2009 dividend yield of 4.46 percent for the SET50 index is used in arbitrage tests involving the
SET50 index. The average 1-month Treasury bill yield from 2006-2009 is used to represent the
risk-free rate, which is the rate at which an arbitrageur can borrow and lend. The average 2006-
2009 T-bill yield is 3.23 percent.

6. Methodology
6.1 Arbitrage test design
The frequency of arbitrage opportunities and size of arbitrage profit are measured in the
following three tests:
(1) SET50 futures vs. SET50 index component stocks
(2) SET50 futures vs. TDEX
(3) TDEX vs. SET50 component stocks

Test period spans from April 2006 when SET50 futures began trading to February 2009
when this study commenced. Daily tests are divided into two sub-periods using the inception
month of TDEX as division point: (1) from April 2006 to August 2007 and (2) from September
2007 to February 2009. Pricing efficiency is inversely related to the frequency of arbitrage
opportunities and the size of arbitrage profit. In other words, the smaller the frequency and size
of index arbitrage, the more efficient the futures pricing is.
Tables A and B illustrate how the three arbitrage tests are implemented and categorized
according to the security types and time periods.

                                                                                                                       
12
Intraday data of the SET50 index are not available in the Stock Exchange of Thailand databases.  

12
 
 Table A
Categorization of Daily Arbitrage Tests by Security Types and Time Periods

Time Period
Apr 06- Aug 07 (17 months) Sep 07- Feb 09 (16 months)
1. SET50 futures vs. SET50 stocks (via SET50 index) 2. SET50 futures vs. SET50 stocks (via SET50 index)
3. SET50 futures vs. TDEX
4. TDEX vs. SET50 stocks

Table B
Categorization of Intraday Arbitrage Tests by Security Types and Time Periods

Time Period
Apr 06- Aug 07 (17 months) Sep 07- Feb 09 (16 months)
5. SET50 futures vs. SET50 stocks (via TDEX)
6. SET50 futures vs. TDEX

6.2 Determining arbitrage opportunity

Following Chung (1991), the signal that arbitrage opportunity exists is determined via the
cost-of-carry model as follows:

ε (t ) = | F (t , T ) − S (t )e( r −δ )(T −t ) | −TC (t ) > 0, (1)

where ε (t ) is the magnitude of mispricing or price discrepancy (that will likely lead to arbitrage
opportunity if greater than zero), t is time at which mispricing is observed, T is expiration time
of SET50 futures contract, F (t , T ) is the price at time t of SET50 futures contract that will expire
at time T , S (t ) is the price at time t of the SET50 index, r is the expected annual risk-free rate, δ
is the expected annual dividend yield to the SET50 index, and TC (t ) is the total transaction costs
incurred in conducting arbitrage trades at time t .
If the mispricing, ε (t ) , is positive, arbitrageurs can sell the overpriced security and
simultaneously buy the underpriced security, then reverse the trade as the overpriced security
becomes underpriced and vice versa. With the presence of transaction costs, an arbitrageur will
make profit if and only if price difference between the securities exceeds the total transaction
costs in a trade. Also, execution lag that may occur between the time an order is made and the

13
 
time it is executed is likely to affect arbitrage profit. This research takes into account both
transaction costs and execution lags. Execution lags are factored in by measuring profit at
execution time, denoted t + , of the arbitrage opportunity observed at time t .

6.3 Calculating arbitrage profit

Once price discrepancy is observed, the trading rules are:

• If F (t , T ) − S (t )e( r −δ )(T −t ) is positive, an arbitrageur can make profit by selling the SET50
futures and buying SET50 component stocks. The number of shares of each stock to buy is
proportional to its weight in the SET50 index. The arbitrageur holds his position until the
reverse signal is observed at time s , that is, until F ( s, T ) − S (s)e( r −δ )(T −s ) becomes negative.
He then takes the opposite position by buying the SET50 futures and selling the stocks. Note
that the proportion of each stock that an arbitrageur holds remains constant throughout the
holding period. Profit is then calculated as follows:

π (t + ) = ( F (t + , T ) − w1S (t + ) − ... − w50 S (t + ) ) + ( − F ( s + , T ) + w1S ( s + ) + ... + w50 S ( s + ) )

1 50 1 50

(2)
(
− TC (t + ) + TC ( s + ) . )

• If F (t , T ) − S (t )e( r −δ )(T −t ) is negative, an arbitrageur makes profit by buying the SET50 futures
and short-selling the stocks, holding the position until the reverse signal is observed, then
closing the position by selling the SET50 futures and buying the stocks. Profit is calculated
as follows:

π (t + ) = ( − F (t + , T ) + w1S (t + ) + ... + w50 S (t + ) ) + ( F ( s + , T ) − w1S ( s + ) − ... − w50 S ( s + ) )

1 50 1 50

(3)
− (TC (t + ) + TC ( s + ) ) .

In equations (2) and (3), t + is the time (after t ) at which a trade is executed to open an
arbitrage position, s is the time (after t + ) at which the first reversed arbitrage signal is observed,
s + is the time (after s ) at which the closing trade is executed, F (t + , T ) is the SET50 futures price at

14
 
time t + , wi is the weight of the ith stock in the SET50 index, and Si (t + ) is the price of the ith

stock at time t + . A stock’s weight in the SET50 index is approximately equal to the proportion
of its market capitalization to the total SET50 market capitalization.
The aforementioned equations (1), (2), and (3) are modified to accommodate arbitrage
tests involving the TDEX by substituting appropriate TDEX parameters (prices, dividend yield,
and transaction costs) into the equations.
The authors acknowledge that marking-to-market creates cash flows over the life of the
arbitrage. However, such cash flows are relatively small (Ramaswamy, 1988). As such,
marking-to-market effects can be disregarded without significantly changing the results.
Dividend yield is another important ingredient of the arbitrage tests in this paper as it
determines the fair price of the futures contract. An unexpected dividend increase will increase
the profits for short-futures arbitrage positions while an unexpected dividend decrease will
decrease the profits for long-futures arbitrage positions. The accuracy of dividend yield
estimation is more crucial for daily tests than intraday tests because dividend yield generally
does not fluctuate significantly over a short period of time.
Short sale13constraints create risk for arbitrageurs because (1) the borrowed stocks can be
called back anytime at the owner’s discretion and (2) not all SET50 stocks can be short-sold.
Such limitations prevent arbitrageurs to participate in short-stock arbitrage positions unless they
currently own the stocks in their portfolios. As short sale restrictions depend on individual
stocks, this study does not incorporate short-selling costs.

7. Daily Results
Daily price levels of the SET50 index and SET50 futures in the period prior to TDEX
introduction (from April 2006 to August 2007) are illustrated in Figure 1 (Panel A). The graph
shows that the SET50 futures generally track the SET50 index but in some periods, price
discrepancies appear. The SET50 futures are generally underpriced with respect to the SET50
index. Figure 1 (Panel B) shows daily prices of the SET50 index, SET50 futures, and TDEX
after TDEX introduction (from September 2007 to February 2009). Most of the time, the price
of SET50 index and those of its tracking instruments are closely aligned. When mispricing
                                                                                                                       
13
Thai practitioners colloquially refer to the act of short selling as stock borrow/loan (SBL).

15
 
occurs, TDEX is generally overpriced while SET50 futures are generally underpriced with
respect to the SET50 index. Comparing SET50 futures to TDEX one finds that SET50 futures
are often underpriced against TDEX.
Table 1 illustrates daily volatility of SET50 index and SET50 futures in different sub-
periods within the test period. Both the SET50 index and SET50 futures are more volatile after
the TDEX launch with 2.25% and 2.70% daily volatility, compared to 1.71% and 1.65%
volatility before TDEX launch. The October-December quarter is the most volatile quarter,
followed by July-September quarter.
Figure 2 graphs daily volatility of the SET50 index and SET50 futures over the sample
months. December, January, and June are the three most volatile months. Price discrepancies
generally occur within those highly volatile quarters and months. That is, when the SET50 index
is volatile, it is highly likely that the tracking securities will be mispriced against the cash index.
Moreover, the SET50 futures is mispriced against the SET50 index more frequently after the
TDEX launch.
Table 2 contains more details on price discrepancies between SET50 futures and
respective benchmark securities and statistical test results on the price discrepancies. The SET50
futures are mispriced (with respect to the cash index) slightly more frequently with increasing
average mispricing magnitude after the TDEX introduction. The increase in mispricing is due to
higher volatility of the SET50 index post TDEX introduction. Furthermore, buy and sell signal
frequencies indicate that SET50 futures are generally overpriced with respect to the SET50 index
but underpriced with respect to TDEX. The p-value slightly improves after the TDEX
introduction, implying that mispricing is more significant.
Table 3 illustrates arbitrage profit from trading SET50 futures against respective
benchmark securities. It shows that even though SET50 futures are mispriced more frequently
after the TDEX introduction, profitable trades occur less frequently. Moreover, even though the
average mispricing magnitude increases, average profit per trade decreases after the TDEX
introduction. This empirical evidence aligns with the theory of limits to arbitrage in that higher
volatility of the underlying index increases the likelihood of adverse price movements, which
eventually leads to lower arbitrage profitability. Moreover, some unprofitable trades may occur
due to forced exit of arbitrage position due to non-convergence of prices. False signal is

16
 
calculated as a percentage of unprofitable trades to total trades. In general arbitrage profit is
lower after the introduction of TDEX.
Table 4 shows arbitrage profit from trading the SET50 component stocks against each of
the index tracking securities. In general, trading the SET50 stocks against either SET50 futures
or TDEX does not generate profit. Moreover, the loss magnitude from trading the SET50 stocks
against the SET50 futures increases after the TDEX introduction. The loss can be attributed to
adverse price movements, that is, prices of the SET50 component stocks at position close are
significantly different from the levels at position open. Higher volatility of the SET50 stocks
post TDEX launch causes more adverse price movements, which eventually lead to higher loss
for arbitrage trades. Furthermore, as the SET50 index is market value weighted, the proportion
of each stock in the index varies daily, if not constantly. Implementing arbitrage trades in which
the proportion of each stock being held remains constant is unlikely to generate profit. Arbitrage
trading may be profitable if the holding in each stock is adjusted periodically.
Table 5 demonstrates the impact of transaction costs on profitability (loss) from trading
the SET50 component stocks against each of the index tracking securities. The mispricing
frequency decreases as transaction costs increase while loss increases as transaction costs
increase.

A. Time series of prices of the SET50 index and SET50 futures prior to TDEX launch

17
 
 B. Time series of prices of the SET50 index, SET50 futures, and TDEX post TDEX launch

 
Fig. 1 Panel A plots daily prices of SET50 index and SET50 futures before TDEX launch (Apr 06 - Aug 07). Panel
B plots daily prices of SET50 futures, SET50 index, and TDEX after TDEX launch (Sep 07 - Feb 09).

Fig. 2 plots daily volatility of the SET50 index and SET50 futures over the sample period (Apr 06 - Feb 09).

18
 
 Table1
Daily volatility of the SET50 index and SET50 futures in different time periods

This table reports daily volatility of the SET50 index and SET50 futures over the sample period. Panel A
demonstrates the increase in volatility of the SET50 index and SET50 futures after the TDEX launch. Panel B
reports volatility in each sample month. In Panel C, volatility in each sample quarter is reported.

Panel A: Difference in SET50 index and SET50 futures volatility before and after TDEX launch
Daily volatility
Period
SET50 index SET50 futures
Apr06-Aug07 (before TDEX launch) 1.71% 1.65%
Sep07-Feb09 (after TDEX launch) 2.25% 2.70%
 
Panel B: SET50 index volatility in each sample quarter
Daily volatility
Quarter
SET50 index SET50 futures
May-Jun 2006 1.49% 1.46%
Jul-Sep 2006 1.70% 1.10%
Oct-Dec 2006 1.31% 2.08%
Jan-Mar 2007 0.95% 1.26%
Apr-Jun 2007 0.95% 1.32%
Jul-Sep 2007 0.82% 2.22%
Oct-Dec 2007 0.81% 2.11%
Jan-Mar 2008 4.93% 1.86%
Apr-Jun 2008 1.59% 1.47%
Jul-Sep 2008 0.97% 2.26%
Oct-Dec 2008 0.62% 4.69%
Jan-Feb 2009 0.63% 2.93%
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 Table 2
14
Price discrepancy between the SET50 futures and respective benchmark securities*

This table illustrates that price discrepancy between SET50 futures and the SET50 index occurs slightly more
frequently after the launch of TDEX. The mean and standard deviation of price discrepancy are also higher post
TDEX launch. Percentages of buy and sell signals indicate that while the SET50 index is generally underpriced
with respect to the futures, the TDEX is generally overpriced.

Before TDEX launch After TDEX launch

Apr 28, 2006 -Sep 9, 2007 Sep 10, 2007 -Feb 9, 2009
Benchmark SET50 index SET50 index TDEX
Number of observations 332 362 362
Number of mispricings 268 304 317
Percentage of mispricings 80.72 83.98 85.77
Percentage of buy signals15 16.42 29.01 63.72
Percentage of sell signals 83.58 70.99 36.28
Mean 3,538.80 4,712.20 6,254.30
Standard deviation 3,814.20 4,860.20 5,397.60
Min -1,639.00 -1,646.90 -1,720.90
Max 21,244.00 22,080.00 27,200.00
p-Value 0.50 0.49 0.49
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
                                                                                                                       
14
Price discrepancy is measured in Thai Baht per one futures contract.
15
A buy signal occurs when the SET50 futures is underpriced against the benchmarked security whereas a sell
signal occurs when the SET50 futures is overpriced against the benchmarked security.

20  
 
 Table 3
Arbitrage profit for trading SET50 futures against respective cash securities**

Before TDEX launch After TDEX launch

Apr 28, 2006 -Sep 9, 2007 Sep 10, 2007 -Feb 9, 2009
Benchmark SET50 index SET50 index TDEX
Number of observations 332 362 362
Number of mispricings 268 304 317
Percentage of mispricings 80.72 83.98 85.77
Profitable trades (%)16 97.39 88.49 56.78
False signals (%) 2.61 11.51 43.22
Profitable buy trades (%) 16.04 23.03 20.50
Profitable sell trades (%) 81.34 65.46 36.28
Conditional profit (Baht)17 6,578.30 4,812.30 5,511.70
Conditional return (%) 1.22 0.99 0.83
Conditional buy profit (Baht) 9,178.20 8,667.80 4,816.00
Conditional sell profit (Baht) 6,065.50 3,456.00 5,905.00
Conditional buy return (%) 1.45 1.49 0.80
Conditional sell return (%) 0.98 0.50 0.85
Unconditional profit (Baht)18 6,566.80 4,812.30 5,511.70
Unconditional return (%) 1.22 0.99 0.83
Unconditional buy profit (Baht) 9,178.20 8,667.80 4,816.00
Unconditional sell profit (Baht) 6,065.50 3,456.00 5,905.00
Unconditional buy return (%) 1.45 1.49 0.80
Unconditional sell return (%) 0.98 0.50 0.85
 

                                                                                                                       
16
Profitable trades and false signals are calculated as percentage of total trades.
17
Conditional profit is average profit of profitable trades.
18
Unconditional profit is average profit of both profitable and unprofitable trades.  

21  
 
 Table 4
Arbitrage profit for trading SET50 component stocks against respective index-tracking securities

Before TDEX launch After TDEX launch

Apr 28, 2006 -Sep 9, 2007 Jan 2, 2008 -Jun 30, 2008
Index-tracking security SET50 futures SET50 futures TDEX
Number of observations 328 122 122
Number of mispricings 228 87 119
Percent of mispricings 69.51 71.31 97.54
Profitable trades (%) 0 0 0
False signals (%) 100 100 100
Profitable buy trades (%) 0 0 0
Profitable sell trades (%) 0 0 0
Conditional profit (Baht) 0 0 0
Conditional return (%) 0 0 0
Conditional buy profit (Baht) 0 0 0
Conditional sell profit (Baht) 0 0 0
Conditional buy return (%) 0 0 0
Conditional sell return (%) 0 0 0
Unconditional profit (Baht) -2,102.3 -2,365.4 -2,361.0
Unconditional return (%) -0.33 -0.37 -0.37
Unconditional buy profit (Baht) -2,090.4 -2,348.9 0
Unconditional sell profit (Baht) -2,149.4 -2,436.8 -2,361.0
Unconditional buy return (%) -0.33 -0.37 0
Unconditional sell return (%) -0.34 -0.38 -0.37
 

22  
 
 Table 5
Profit from arbitrage trades between the SET50 component stocks and respective index-tracking
securities

Jul 1, 2008 -Dec 30, 2008 (After the TDEX launch)

Index-tracking security SET50 futures TDEX
Number of observations 124 124
Transaction costs Low High Low High
Number of mispricings 106 99 124 124
Percentage of mispricings 85.48 79.84 100 100
Percentage of buy signals 99 99 0 0
Percentage of sell signals 1 1 100 100
Profitable trades (%) 0 0 0 0
False signals (%) 100 100 100 100
Profitable buy trades (%) 0 0 0 0
Profitable sell trades (%) 0 0 0 0
Conditional profit (Baht) 0 0 0 0
Conditional return (%) 0 0 0 0
Conditional buy profit (Baht) 0 0 0 0
Conditional sell profit (Baht) 0 0 0 0
Conditional buy return (%) 0 0 0 0
Conditional sell return (%) 0 0 0 0
Unconditional profit (Baht) -1,789.2 -3,007.2 -1,770.9 -2,951.4
Unconditional return (%) -0.33 -0.56 -0.33 -0.55
Unconditional buy profit (Baht) -1,792.2 -3,012.7 0 0
Unconditional sell profit (Baht) -1,486.3 -2,477.3 -1,770.9 -2,951.4
Unconditional buy return (%) -0.33 -0.56 0 0
Unconditional sell return (%) -0.28 -0.46 -0.33 -0.55

23  
 
8. Intraday Results
As the SET50 index intraday quotes are unobtainable from the Stock Exchange of
Thailand intraday database at the time of research, intraday arbitrage tests in this study include
the following: (1) SET50 futures vs. TDEX and (2) SET50 futures vs. SET50 stocks. Thus,
TDEX replaces the SET50 index in the mispricing equation (Equation 1). Test period is divided
into sub-periods of 5 days. Tables 6-11 illustrate the results from selected sub-periods. For
consistency, the sub-periods shown are the same hereafter. They are January 2-8, 2008, March
10-14, 2008, May 2-8, 2008, and June 2-6, 2008. The sub-periods of January 2-8 and May 2-8
are further away from SET50 futures expiration than those of March 10-14 and June 2-619. As
such, results in the former two sub-periods can be viewed as representations of what happens in
periods far away from futures expiration. Similarly, results in the latter two sub-periods
represent the situation in near-expiration periods.
Table 6 shows mispricing characteristics in the aforementioned four sub-periods. Results
indicate that mispricing occurs more frequently in the periods closer to maturity of SET50
futures. Except for the March 10-14 sub-period, the average mispricing magnitude is higher in
periods closer to SET50 futures maturity. Standard deviation of mispricing is higher in periods
closer to the SET50 futures maturity. Moreover, mispricing is statistically more significant
(having lower p-values) in the periods closer to futures maturity. Note that the higher mispricing
magnitude and volatility coincide with higher futures volatility and liquidity in the periods closer
to maturity.
Table 7 illustrates arbitrage profit from trading the SET50 component stocks against
SET50 futures. Arbitrage trades involving the SET50 component stocks and SET50 futures are
generally unprofitable. The only sub-period in which profitable arbitrage trades are still possible
is January 2-8, 2008, with low probability of 0.0051. In this regard, intraday results are similar
to daily results, that is, arbitrage trades involving the SET50 component stocks are generally
unprofitable due to adverse price movements of the stocks. More specifically, by the time a
reverse signal (calculated using TDEX price) is observed, stock prices have diverged from
TDEX price to the point that trading all individual stocks does not lead to the same result as
trading TDEX.

                                                                                                                       
19
There are four SET50 futures contracts having expirations in March, June, September, and December.

24  
 
 Table 8 illustrates arbitrage profit from trading SET50 futures against TDEX. The results
indicate that arbitrage involving trading the two index-tracking securities is profitable, although
the probability that a trade is profitable is quite low (2.52%, 2.86%, 5.53%, and 0.84% in the
reported January, March, May, and June sub-periods, respectively). Profit in the periods closer
to futures maturity is generally higher.
While Tables 6-8 show results from the control setup, Tables 9-11 are outcomes from
sensitivity analyses with changing variable. The control setup features institutional-bracket
(low) transaction cost, no execution lag, and base dividend yield. Each of the three variables is
varied in our sensitivity analyses as shown in Tables 9-11.
The sensitivity analysis in Table 9 illustrates the effects of transaction costs on mispricing
and profitability from trading SET50 futures against TDEX. The results indicate that mispricing
occurs more frequently under the low transaction cost bracket. The average magnitude and
standard deviation of mispricing are also higher under the low transaction cost bracket.
Moreover, under the low transaction bracket, profitability is higher.
Table 10 contains sensitivity analysis results illustrating the effects of execution lags on
mispricing and profitability from trading SET50 futures against TDEX. In general, mispricing
occurs more frequently if execution lag is shorter. The average magnitude and standard
deviation of mispricing are also higher if execution lag is shorter. However, mispricing is more
statistically significant in scenarios of longer execution lag. That is, if mispricing persists despite
longer execution lag, it must be a true price discrepancy, demonstrating higher significance
(having lower p-value). Profitability, on the other hand, is higher for lower execution lags.
Notice that arbitrage profit disappears after 10 minutes execution lag.
Table 11 shows the effects of varying dividend yields on mispricing and profit from
arbitrage trades between SET50 futures and TDEX. There is a decreasing trend in mispricing
frequency as dividend yield increases. Dividend yield also affects the proportions of buy and sell
signals, that is, the proportion of buy (sell) signals decreases (increases) as dividend yield
increases. The higher is the dividend yield, the more likely SET50 futures will be overpriced
against TDEX. Also, the average magnitude and standard deviation of mispricing are lower for
higher dividend yield and the mispricing is less statistically significant as dividend yield
increases. Index arbitrage profitability increases as dividend yield increases.

25  
 
 Table 6
Intraday frequency of mispricing in different time periods

Price discrepancies occur more frequently during periods closer to expiration of SET50 futures. The average
magnitude and standard deviation of price discrepancies are higher in the periods closer to the SET50 futures
expiration. Price discrepancies are more statistically significant in the periods closer to futures expiration.

Jan 2-8, 2008 Mar 10-14, 2008 May 2-8, 2008 Jun 2-6, 2008
SET50 futures vs. TDEX
20
Transaction cost Low Low Low Low
Execution lag (min) 0 0 0 0
Dividend yield (%) 2.17 2.17 2.17 2.17
Number of observations 4900 2081 2289 3111
Number of mispricings 4334 2081 623 3110
Percentage of mispricings 88.45 100 27.22 99.97
Percentage of buy signals 100 100 100 100
Percentage of sell signals 0 0 0 0
Mean 148.21 141.38 -5.19 138.70
Standard deviation 453.50 987.12 117.84 736.55
Min -2,371.20 0 -2,395.40 -77.29
Max 5,363.90 11,437.00 2,503.00 7,422.6
p-Value 0.39 0.24 0.38 0.24

                                                                                                                       
20
Institutional investors normally face low transaction cost bracket, which includes THB 300 per futures contract
and 0.15% of TDEX trading value. Retail investors face high transaction cost bracket, which is THB 500 per
futures contract and 0.25% of TDEX trading value.

26  
 
 Table 7
Intraday arbitrage profit for trading SET50 futures vs. SET50 component stocks

Using TDEX in signal assessment, this table illustrates arbitrage profit from trading the SET50 component stocks
against SET50 futures. Arbitrage trades involving the SET50 stocks and SET50 futures are generally unprofitable.

After TDEX launch

Jan 2-8, 2008 Mar 10-14, 2008 May 2-8, 2008 Jun 2-6, 2008
Securities SET50 futures vs. SET50 component stocks
Transaction cost Low Low Low Low
Execution lag (min) 0 0 0 0
Dividend yield (%) 2.17 2.17   2.17   2.17  
Number of observations 4900 2081 2289 3111
Number of mispricings 4334 2081 623 3110
Percentage of mispricings 88.45 100 27.22 99.97
Percentage of buy signals 100 100 100 100
Profitable trades (%) 0.51 0 0 0
False signals (%) 11.72 0 4.80 0
Profitable buy trades (%) 0.51 0 0 0
Profitable sell trades (%) 0 0 0 0
Conditional profit (Baht) 583.11 0 0 0
Conditional return (%) 0.09 0 0 0
Conditional buy profit (Baht) 583.11 0 0 0
Conditional sell profit (Baht) 0 0 0 0
Conditional buy return (%) 0.09 0 0 0
Conditional sell return (%) 0 0 0 0
Unconditional profit (Baht) -158.46 0 -1,255.10 0
Unconditional return (%) -0.03 0 -0.20 0
Unconditional buy profit (Baht) -316.92 0 -1,255.10 0
Unconditional sell profit (Baht) 0 0 0 0
Unconditional buy return (%) -0.05 0 -0.20 0
Unconditional sell return (%) 0 0 0 0

27  
 
 Table 8
Profit from arbitrage trades between SET50 futures and TDEX

This table shows arbitrage profit from trading SET50 futures against TDEX. Arbitrage involving trading the two
index-tracking securities is profitable, although the probability that a trade is profitable is slightly low. Profit in the
periods closer to futures maturity is generally higher.

After TDEX launch

Jan 2-8, 2008 Mar 10-14, 2008 May 2-8, 2008 Jun 2-6, 2008
Securities SET50 futures vs. TDEX
Transaction cost Low Low Low Low
Execution lag (min) 0 0 0 0
Dividend yield (%) 2.17 2.17   2.17   2.17  
Number of observations 4900 2081 2289 3111
Number of mispricings 4334 2081 623 3110
Percentage of mispricings 88.45 100 27.22 99.97
Percentage of buy signals 100 100 100 100
Profitable trades (%) 2.52 2.86 5.53 0.84
False signals (%) 0 0 0 0
Profitable buy trades (%) 2.52 2.86 5.53 0.84
Profitable sell trades (%) 0 0 0 0
Conditional profit (Baht) 2,332.00 10,673.00 2,555.20 3,323.00
Conditional return (%) 0.37 1.85 0.42 0.56
Conditional buy profit (Baht) 2,332.00 10,673.00 2,555.20 3,323.00
Conditional sell profit (Baht) 0 0 0 0
Conditional buy return (%) 0.37 1.85 0.42 0.56
Conditional sell return (%) 0 0 0 0
Unconditional profit (Baht) 2,332.00 10,673.00 2,555.20 3,323.00
Unconditional return (%) 0.37 1.85 0.42 0.56
Unconditional buy profit (Baht) 2,332.00 10,673.00 2,555.20 3,323.00
Unconditional sell profit (Baht) 0 0 0 0
Unconditional buy return (%) 0.37 1.85 0.42 0.56
Unconditional sell return (%) 0 0 0 0

28  
 
 Table 9
Sensitivity analysis: Effects of transaction costs on mispricing & profit for SET50 futures vs. TDEX

This table shows that transaction costs affect mispricing frequency, mispricing magnitude, and arbitrage profit.
Mispricing frequency and average profit vary negatively with transaction costs.

Jun 2-6, 2008 SET50 futures vs. TDEX

Transaction costs Low High
Execution lag (min) 0 0
Dividend yield (%) 2.17 2.17
Number of observations 3111 3111
Number of mispricings 3111 3071
Percentage of mispricings 100 98.71
Percentage of buy signals 100 100
Percentage of sell signals 0 0
Mean 166.44 97.60
Standard deviation 877.05 531.35
Min 0 -1,677.2
Max 8,097.90 5,401.10
p-Value 0.24 0.24
Profitable trades (%) 2.52 2.30
False signals (%) 0 0
Profitable buy trades (%) 2.52 2.30
Profitable sell trades (%) 0 0
Conditional profit (Baht) 2,332.00 1,008.43
Conditional return (%) 0.37 0.16
Conditional buy profit (Baht) 2,332.00 1,008.43
Conditional sell profit (Baht) 0 0
Conditional buy return (%) 0.37 0.16
Conditional sell return (%) 0 0
Unconditional profit (Baht) 2,332.00 1,008.43
Unconditional return (%) 0.37 0.16
Unconditional buy profit (Baht) 2,332.00 1,008.43
Unconditional sell profit (Baht) 0 0
Unconditional buy return (%) 0.37 0.16
Unconditional sell return (%) 0 0

29  
 
 Table 10
Sensitivity analysis: Effects of execution lags on mispricing & profit for SET50 futures vs. TDEX

This table shows how varying execution lags affect mispricing frequency, mispricing magnitude, and arbitrage
profit. Mispricing frequency and average arbitrage profit vary negatively with execution lag. However, mispricing
is more statistically significant for longer execution lag.

Jun 2-6, 2008

SET50 futures vs. TDEX
Transaction cost Low Low Low Low
Execution lag (min) 0 2 5 10
Dividend yield (%) 2.17 2.17 2.17 2.17  
Number of observations 3111 3111 3111 3111
Number of mispricings 3110 1882 957 0
Percentage of mispricings 99.97 60.50 30.76 0
Percentage of buy signals 100 100 100 0
Percentage of sell signals 0 0 0 0
Mean 138.70 85.50 44.41 NA
Standard deviation 736.55 610.87 458.58 NA
Min -77.29 0 0 NA
Max 7,422.60 6,201.10 6,001.10 NA
p-Value 0.24 0.15 0.08 NA
Profitable trades (%) 0.84 0.81 0.72 NA
False signals (%) 0 0.12 0.10 NA
Profitable buy trades (%) 0.84 0.81 0.72 NA
Profitable sell trades (%) 0 0 0 NA
Conditional profit (Baht) 3,323.00 3,428.47 3,605.80 NA
Conditional return (%) 0.56 0.58 0.61 NA
Conditional buy profit (Baht) 3,323.00 3,428.47 3,605.80 NA
Conditional sell profit (Baht) 0 0 0 NA
Conditional buy return (%) 0.56 0.58 0.61 NA
Conditional sell return (%) 0 0 0 NA
Unconditional profit (Baht) 3,323.00 3,192.02 3,118.10 NA
Unconditional return (%) 0.56 0.54 0.52 NA
Unconditional buy profit (Baht) 3,323.00 3,192.02 3,118.10 NA
Unconditional sell profit (Baht) 0 0 0 NA
Unconditional buy return (%) 0.56 0.54 0.52 NA
Unconditional sell return (%) 0 0 0 NA

30  
 
 Table 11
Sensitivity analysis: Effects of dividend yield on mispricing and profit for SET50 futures vs. TDEX

This table shows that dividend yields affect mispricing frequency, mispricing magnitude, and arbitrage profitability.
As dividend yield increases, mispricing occurs more frequently with higher average magnitude and standard
deviation. However, the mispricing is less statistically significant as dividend yield increases. Dividend yield is the
only variable affecting the mispricing direction–as dividend yield increases, SET50 futures are more likely to be
overpriced against the cash index. Lastly, profit generally increases as dividend yield increases.

Jun 2-6, 2008

SET50 futures vs. TDEX

Transaction cost Low Low Low Low Low
Execution lag (min) 0 0 0 0 0
Dividend yield (%) 0 2.17 4 7 15
Number of observations 3111 3111 3111 3111 3111
Number of mispricings 3111 3110 3105 2889 106
Percentage of mispricings 100 99.97 99.81 92.9 3.41
Percentage of buy signals 100 100 100 100 62.3
Percentage of sell signals 0 0 0 0 37.7
Mean 166.44 138.70 115.35 77.14 -15.18
Standard deviation 877.05 736.55 619.33 431.85 101.27
Min 0 -77.29 -984.59 -2,301.00 -2,348.10
Max 8,097.90 7,422.60 6,096.20 4,599.30 1,639.30
p-Value 0.24 0.24 0.24 0.24 0.76
Profitable trades (%) 0.74 0.84 0.85 0.89 6.71
False signals (%) 0.18 0 0 0 2.64
Profitable buy trades (%) 0.74 0.84 0.85 0.89 6.71
Profitable sell trades (%) 0 0 0 0 0
Conditional profit (Baht) 3,519.40 3,323.00 3,427.90 3,715.90 7,300
Conditional return (%) 0.59 0.56 0.58 0.62 1.23
Conditional buy profit (Baht) 3,519.40 3,323.00 3,427.90 3,715.90 7,300
Conditional sell profit (Baht) 0 0 0 0 0
Conditional buy return (%) 0.59 0.56 0.58 0.62 1.23
Conditional sell return (%) 0 0 0 0 0
Unconditional profit (Baht) 2,792.90 3,323.00 3,427.90 3,715.90 4,529.80
Unconditional return (%) 0.47 0.56 0.58 0.62 0.76
Unconditional buy profit (Baht) 2,792.90 3,323.00 3,427.90 3,715.90 7,300
Unconditional sell profit (Baht) 0 0 0 0 -2,502.20
Unconditional buy return (%) 0.47 0.56 0.58 0.62 1.23
Unconditional sell return (%) 0 0 0 0 -0.42

31  
 
9. Discussion

Daily results indicate that mispricing in the SET50 futures against the SET50 index
occurs more frequently after the TDEX launch (83.98% vs. 80.72%). The volatility of the
SET50 index post TDEX launch is higher (2.25% vs. 1.71%). Despite higher frequency of
mispricing post TDEX launch, the proportion of profitable trades is lower as reverse signals are
rarely observed. In other words, arbitrageurs cannot close their position profitably many times
due to the divergence of futures price. In such case, arbitrage trades are unwound unprofitably at
the end of the study horizon. False signals indicate the percentage of forced closed arbitrage
trades. The proportion of profitable trades (SET50 index vs. SET50 futures) is 88.49% post
TDEX launch, an 11% drop from prior to TDEX launch. The theory of limits to arbitrage put
forward by Yadav and Pope (1994) suggests that higher volatility of the underlying index
increases the likelihood of adverse price movements, which eventually leads to lower probability
of making profit.

When TDEX is used as a proxy for the SET50 index, average arbitrage return from
trading TDEX vs. SET50 futures is lower than the “theoretical” arbitrage profit from trading
SET50 index vs. SET50 futures (0.83% vs. 0.99%). Proportion of profitable trades is also lower
(56.78%) for TDEX vs. SET50 futures trades. That is, the SET50 futures on average align more
closely with the TDEX than with the SET50 index.

Generally, the SET50 futures is overpriced compared to the SET50 index but is
underpriced compared to the TDEX. The percentage of sell signals in the case of SET50 futures
vs. SET50 index (70.99%) is higher than in that of SET50 futures vs. TDEX (36.28%). The
higher TDEX price and volume reflect more positive market sentiment with respect to the
TDEX.

TDEX trading leads to lower arbitrage return from SET50 futures vs. SET50 index trades
(0.99% post TDEX launch vs. 1.22% before TDEX launch). As mentioned before, an arbitrage
trade is closed either when a reverse signal is observed or at the end of test horizon. Therefore, a
trade may or may not make profit. When a trade is closed following a reverse signal, it generates
profit but when it is forced closed at the end of test period, it normally incurs a loss. An
arbitrage trade that incurs a loss is considered a false signal. The number of false signals for

32  
 
SET50 futures vs. SET50 index trades increases from 2.61% to 11.52% post TDEX launch. The
increase in false signals occurs because fewer reverse signals are observed post TDEX launch so
trades are unwound unprofitably. Slower price adjustment of the SET50 futures contributes to
few to no reverse signals. Once the SET50 futures are mispriced against the SET50 index, it
takes longer time for the futures price to be corrected.
The slower SET50 futures price adjustment post TDEX launch is attributed to the effect
TDEX trading has on the pricing of SET50 futures. The TDEX, compared to SET50 futures, is a
more popular security among investors because trading TDEX is less risky and requires lower
initial investment. As such, TDEX is traded more often than the SET50 futures, resulting in
significantly higher trading volume. With low trading volume, it is unlikely that mispricing in
the SET50 futures will be corrected instantaneously.
Intraday results indicate that mispricing frequency, its average magnitude, standard
deviation, and statistical significance are higher in the period closer to expiration of SET50
futures. Moreover, arbitrage profit is higher in the period closer to futures maturity. The period
near futures maturity is characterized by high futures volatility and trading volume. Theory of
limits to arbitrage suggests that higher volatility will result in lower arbitrage profit due to higher
possibility of adverse price movements (Yadav and Pope, 1994) but higher liquidity will ease the
establishment of arbitrage position, leading to higher arbitrage profitability (Roll et al., 2007).
Therefore, the fact that profit is higher near futures maturity indicates that the effect of liquidity
outweigh that of volatility.
Sensitivity analyses show that varying transaction costs, execution lags, and dividend
yields affect mispricing frequency, mispricing magnitude, and arbitrage profitability. More
specifically, mispricing occurs less frequently for higher transaction costs. For shorter execution
lags, mispricing occurs more frequently with higher average magnitude and standard deviation.
However, mispricing is more statistically significant for longer execution lags. Dividend yield
has opposite effect to execution lag on mispricing. As dividend yield increases, mispricing
occurs more frequently with higher average magnitude and standard deviation. However, the
mispricing is less statistically significant as dividend yield increases. Moreover, dividend yield
is the only variable affecting the mispricing direction. That is, as dividend yield increases, the
SET50 futures are more likely to be overpriced against the cash index. Finally, profit generally
increases as dividend yield increases.

33  
 
 Both daily and intraday results indicate that it is almost impossible to make arbitrage
profit from trading the SET50 component stocks against either the SET50 futures or TDEX.
Furthermore, the magnitude of loss from trading the SET50 stocks against SET50 futures is
smaller than that from trading the SET50 stocks against TDEX. Arbitrage trades involving the
SET50 component stocks are generally unprofitable due to adverse price movements of the
stocks. More specifically, by the time a reverse signal calculated using either the SET50 index
or TDEX prices is observed, individual stock prices have diverged from SET50 index or TDEX
prices. Because the proportion of each stock traded at position close is the same as at position
open, a loss is incurred.

10. Conclusion

The main question of how the introduction of TDEX contributes to pricing efficiency of
the SET50 futures has been answered. TDEX trading does not have positive contribution to
pricing efficiency of the SET50 futures as the futures diverges from the SET50 index more
frequently after TDEX launch. The frequent divergence of SET50 futures price from the cash
price post TDEX launch could result from lower liquidity of the SET50 futures compared to that
of the TDEX. While both the TDEX and SET50 futures allow investors to replicate the SET50
performance, TDEX is generally more attractive to an investor. As a riskier security, which also
requires higher initial investment, the SET50 futures are traded less often compared to TDEX.
With lower trading volume and turnover, the SET50 futures, once mispriced, is unlikely to be
corrected instantaneously.

Insufficient institutional trading in Thailand could lead to the persistence of the SET50
futures price deviations. As a retail-dominated market, institutional trades in Thailand account
for only 30 percent of all trades. Barrier to entry into the securities and brokerages industry
demonstrated in the form of fixed number of brokerage licenses is a contributing factor to the
limited participation of institutional investors in Thailand. As arbitrageurs are typically
institutional investors, the limited number of institutional investors implies limited index
arbitrage activity. It is the inadequate participation of arbitrageurs that leads to lower trading
volume and turnover of the SET50 futures. Thus, when the SET50 futures are mispriced, it takes
significantly long time to revert.

34  
 
 Existing literatures provide ample evidence indicating that limits to arbitrage including
illiquidity are what cause persistence of index futures mispricing (Roll et al., 2007; Chordia et
al., 2008; Chen et al., 2009). Empirical results in our research strengthen the theory of limits to
arbitrage in two following ways. First, daily results indicate that higher volatility of the SET50
index and futures post TDEX launch leads to higher frequency of mispricing and lower arbitrage
profit. This is consistent with Yadav and Pope (1994) in that higher volatility will result in lower
arbitrage profit due to higher possibility of adverse price movements. With lower expected
profit, arbitrageurs are unwilling to enter into an arbitrage trade. Consequently, mispricing will
persist and efficiency of the SET50 futures price will be unattainable.
Second, intraday results indicate that the effect of liquidity on futures mispricing
outweigh that of volatility. While Yadav and Pope (1994) suggest that higher volatility will
result in lower arbitrage profit due to higher possibility of adverse price movements, Roll et al.
(2007) finds that higher liquidity will ease the establishment of arbitrage position, leading to
higher arbitrage profitability. We find that the periods near futures maturity are characterized by
higher futures volatility and trading volume. The mispricing and arbitrage profit in these periods
are also higher than those further from maturity. Such empirical evidence implies that the effect
of liquidity outdo that of volatility in the SET50 futures case.
In sum, this research diverges from some existing literatures, which conclude that ETFs
improve pricing efficiency of the corresponding index futures (Chung, 1991; Switzer et al., 2000;
Kurov and Lasser, 2002; Cummings and Frino, 2008). More specifically, the TDEX does not
contribute to pricing efficiency of the SET50 futures. However, the results of this research
strengthen existing literatures on liquidity and pricing efficiency. That is, lower SET50 futures
liquidity, demonstrated in thinner trading volume, contributes to the persistence of mispricing in
the SET50 futures.

11. Acknowledgements

The authors thank the Stock Exchange of Thailand for the research grant and invaluable
support, the Thailand’s Futures Exchange for useful data, participants at the Stock Exchange of
Thailand Capital Market Research Forum 2/2011 for constructive comments.

35  
 
12. References

Attari, M. and Mello A., 2006, Financially constrained arbitrage in illiquid markets, Journal of
Economics and Dynamical Control 30(2), 793-882.
Bae, K.H., Chan, K., Cheung, Y. L., 1998, The profitability of index futures arbitage: evidence
from bid-ask quotes, Journal of Futures Markets18, 743-763.
Barberis, N. and Thaler, R., 2003, A survey of behavioral finance, Handbook of the Economics
of Finance: Financial Markets and Asset Pricing (Vol. 1B), Amsterdam: North Holland.
Basak, S., Croitoru, B., 2000, Equilibrium mispricing in a capital market with portfolio
constraints, Review of Financial Studies 13(3), 715-748.
Basak, S., Croitoru, B., 2006, On the role of arbitrageurs in rational markets, Journal of
Financial Economics 81, 143-173.
Chen, W.P., Chou, R. K., Chung, H., 2009, Decimalization, ETFs and futures pricing efficiency,
Journal of Futures Markets 29(2), 157-178.
Chordia, T., Roll, R., Subrahmanyam, A., 2008, Market liquidity and trading activity, Journal of
Finance 56(2), 2001.
Chung, P. Y., 1991, A transactions data test of stock index futures market efficiency and index
arbitrage profitability, Journal of Finance 46, 1791-1809.
Cummings, J. R., Frino, A., 2008, Index arbitrage and the pricing relationship between
Australian stock index futures and their underlying shares, 21st Australasian Finance and
Banking Conference 2008 proceedings.
Deville, L., Gresse, C., Severac, B., 2010, Direct and indirect effects of index ETFs on spot-
futures mispricing and illiquidity, Working paper, EDHEC Business School.
Gagnon, L., Karolyi, G.A., 2010, Multi-market trading and arbitrage, Journal of Financial
Economics 97, 53-80.
Gromb, D., Vayanos, D., 2010, Limits of arbitrage, Annual Review of Financial Economics 2,
251-275.
Klemkosky, R. C., Lee, J. H., 1991, The intraday ex post and ex ante profitability of index
arbitrage, Journal of Futures Markets 11, 291-311.

36  
 
Kurov, A. A., Lasser, D. J., 2002, The effect of the introduction of Cubes on the Nasdaq-100
index spot-futures pricing relationship, Journal of Futures Markets 22, 197-218.
MacKinlay, A. C., Ramaswamy, K., 1998, Index-futures arbitrage and the behavior of stock
index futures prices, Review of Financial Studies 1(2), 137-158.
Mall, M., Pradhan, B.B., Mishra, P.K., 2011, The efficiency of India’s stock index futures
market: An empirical analysis, International Research Journal of Finance and Economics
69, 2011.
Morck, R., Yeung, B, Yu, W., 2000, The information content of stock markets: why do emerging
markets have synchronous stock price movements?, Journal of Financial Economics 58,
215-260
Mitchell, M., Pulvino, T., Strafford, E., 2002, Limited arbitrage in equity markets, Journal of
Finance 57, 551-584.
Neal, R., 1996, Direct tests of index arbitrage models, Journal of Financial and Quantitative
Analysis 31, 541-562.
Pavabutr, P., Sirodom, K., 2010, Mutual fund flows in Asian equity markets, Working paper,
Capital Market Research Institute of the Stock Exchange of Thailand.
Pavabutr, P., Prangwattananon, S., 2009, Tick size change on the Stock Exchange of Thailand,
Review of Quantitative Finance and Accounting 32, 351-371.
Richie, N., Daigler R., Gleason K., 2008, The limits to stock index arbitrage: Examining S&P
500 futures and SPDRS, Journal of Futures Markets 28(12), 1182-1205.
Roll, R., Schwartz, E., Subrahmanyam, A., 2007, Liquidity and the law of one price: The case of
the futures/cash basis, Journal of Finance 62, 2201-2234.
Switzer, L. N., Varson, P. L., Zghidi, S., 2000, Standard and Poor’s depository receipts and the
performance of the S&P 500 index futures market, Journal of Futures Markets 20, 705–
716.
Tuckman, B., Vila, J. L., 1992, Arbitrage with holding costs: a utility-based approach, Journal of
Finance 52, 35-55.
Yadav, P.K., Pope, P.F., 1994, Stock index futures mispricing, profit opportunities or risk
premia?, Journal of Banking and Finance 18, 921-53

37