Efficiency, cointegration and contagion in equity markets: Evidence

from China, Japan and South Korea

Asian Economic Journal 2009; 23(1); 93-118

A.S.M. Sohel Azad

Associate Professor of Finance and Banking
University of Chittagong, Bangladesh
Visiting Research Fellow (2004-05), Kobe Gakuin University, Japan
And
Accounting and Finance
Monash University, Australia
And
School of Asia Pacific Management
Ritsumeikan Asia Pacific University, Japan

E-mail: [email protected]

This version: July, 2008

Acknowledgement

I am grateful to an anonymous reviewer for several constructive criticisms and

suggestions, which improved the value of the paper. I am also indebted to Dipendra Sinha
and Suzuki Yasushi for their comments and suggestions on an earlier version of the
paper. The financial assistance of the Makita Scholarship Foundation is greatly
acknowledged. The usual disclaimer applies for remaining errors, omissions and/or
misinterpretations.

Electronic copy available at: http://ssrn.com/abstract=1366622

 Efficiency, cointegration and contagion in equity markets: Evidence
from China, Japan and South Korea

Abstract

This paper empirically examines whether the three East Asian stock markets, namely those
of China, Japan and South Korea are individually and/or jointly efficient and whether the
cointegrated markets have contagion from one to the other market(s). While the individual
market efficiency is examined through testing for random walk hypothesis, the joint market
efficiency is examined through testing for cointegration and contagion hypotheses. The
study finds that the hypothesis of individual market efficiency is strongly rejected for the
Chinese stock market but not for Japanese and South Korean stock markets. However, the
joint market efficiency is found to be strongly rejected for all these markets under the
cointegration sense. We take a simple case of contagion and find that although there is a
long-term relationship among these three markets, the contagion hypothesis could not be
rejected only between Japanese and South Korean stock markets indicating short-run
portfolio diversification benefits from these two markets.

Key words: Market Efficiency; Unit root; Variance-ratio; Cointegration; Contagion;

Simulation;

JEL classification: C14; C32; G14; G15

1. Introduction

The concept of market efficiency dates back to the theoretical contribution by

Bachelier in 1900 and the pioneering empirical research by Cowles in 1933 (Campbell,

Lo and MacKinlay, 1997). However, Bernstein (1992) points out that the modern

literature of efficiency starts with Samuelson‟s (1965) contribution “Proof that Properly

Anticipated Prices Fluctuate Randomly” where he has neatly introduced the concept of

random walk hypothesis to economics and finance. Fama (1970) also has captured this

Electronic copy available at: http://ssrn.com/abstract=1366622

idea in his work. He points out that, in an informationally efficient market, prices fully

reflect all available information. It is worthwhile to note that an investor behaves

aggressively upon any informational advantage at his/her disposal and thus he/she allows

prices to incorporate new information, which generates returns from his/her investment.

An investor responds to the new information before the profits from trading on the assets

quickly disappear. In the age of rapid growth of IT and economic globalization this

happens instantaneously. Therefore, information-based trading is always risky and most

efficient market would be one in which price changes are random and unpredictable. In

general, the developed markets are found to be informationally efficient in which price

changes are unpredictable and excessive returns are unlikely as because both the prices

are properly set and the risks are appropriately measured.

The study of individual market efficiency helps us to be familiar with the

behavior of that specific market. Such a study is important if the investor focuses on one

market only. However, the increasing levels of trade interaction and the easing of

regulatory rules governing the movement of capital have allowed the investors to look for

international portfolio diversification among the several markets. Consequently, investors

are encouraged to know the investment behavior of other markets to exploit the arbitrage

opportunities, if any. Tai (2001) also argues that in the age of economic globalization and

regional integration, the study of individual market efficiency has limited implications

particularly when investors look for time-varying risk premia in different markets.

The need to study the behavior of several stock markets has thus encouraged the

academics, policy makers and the international fund managers to know whether these

markets are truly interlinked, interdependent, cointegrated and thus contagious to each

3
other. The general notion is that if there is a strong evidence of the cointegration

hypothesis, the markets are susceptible to the shocks in other markets and hence the

volatility in one market does spillover to other1. There are two major assumptions that

empirical literature uses to clarify the relationship between the market efficiency and

cointegration. One of these assumptions is that if asset prices in two different markets are

efficient, then these prices cannot be cointegrated (see also, Granger, 1986). Second

assumption is related to statistical cointegration and market integration. The general

wisdom is that if asset prices in two different markets are integrated of the same order,

i.e., I(1), then these prices are usually cointegrated. Both these two assumptions lead to

contradictory conclusion that the integrated financial markets cannot be efficient markets.

Lence and Falk (2005) have raised these issues and clarified the relationships among

market efficiency, market integration and statistical cointegration more clearly from

theoretical perspectives. They show four possible combinations (concerning market

efficiency and market integration) to demonstrate that the concept of cointegration is

unrelated to market integration or to market efficiency and that the equity prices may be

either cointegrated or not in all four possible combinations. Therefore, it is practically

difficult to prove or refute a direct link among cointegration, market efficiency and

market integration.

This paper focuses on three East Asian equity markets namely, China, Japan and

South Korea to examine whether these markets are individually efficient and/or jointly

efficient and contagious to each other. Question may arise why we confine our study into

these three markets only, while there are many emerging and rapidly developing markets

1
The literature on market contagion typically looks at volatility spillovers using the ARCH/GARCH
approach. Because we focus more on the interrelationships, we do it here investigating the long-run
equilibrium relationship and the causal relationship.

4
in the region. Janakiramanan and Lamba (1998) in this regard argue that markets that are

geographically and economically close exert significant influence over each other. Masih

and Masih (1999) also advocate a strong support of intra-regional impact of fluctuations

in the Asian stock markets.

Three major purposes are identified for this study: to examine (i) the individual

market efficiency (ii) the interdependence (or joint market efficiency) and (iii) the

contagion, each of which is represented by three hypotheses: efficiency hypothesis,

cointegration hypothesis and contagion hypothesis. Each of these hypotheses is tested

using different methodologies. The individual market efficiency is examined through

studying the univariate properties of the markets concerned and the joint efficiency is

examined through multivariate tests of the cointegration and causality. The multivariate

tests elucidate whether the markets are interdependent and interlinked to each other.

While the presence of both individual and joint market efficiencies indicates that the

returns from the markets are unpredictable, the absence of the same implies that the

return from a market can be predicted by its or other‟s past values. On the one hand, the

unpredictability of return from the financial market is commonly interpreted as the

evidence of efficiency. On the other hand, the predictability of return is interpreted as the

evidence against efficiency. The ability to forecast the returns from a market depends on

several factors, for instances, investors‟ ability to gather superior information, economic

integration, the extent of speculation in the market, the extent of policy intervention by

the regulators and entry of foreign investors etc.

Three hypotheses are tested in three phases. First, the individual market efficiency

is examined through the Zivot and Andrew (1992) unit root test and the variance ratio

5
tests of Lo-MacKinlay (1988) and Wright (2000). Second, the joint efficiency is

examined through the Engle-Granger (1987) and Gregory-Hansen (1996) cointegration

tests. Third, the contagion hypothesis is examined through Toda-Yamamoto (1995) tests

of Granger causality.

All the above procedures are applied to the daily closing prices of the three stock

exchanges: Shanghai Stock Exchange (SSE, China), Tokyo Stock Exchange (TSE, Japan)

and Korea Stock Exchange (KSE, South Korea). The test statistics indicate that the

Chinese stock market is characterized by inter-temporal inefficiency under the hypothesis

of individual as well as joint market efficiency. Not surprisingly, even though the

Japanese and South Korean markets are found to exhibit market efficiency under the

hypothesis of individual market efficiency, these two markets along with the Chinese

markets are found to be jointly inefficient. Since our analysis indicates that the three

equity markets are cointegrated demonstrating a long-term relationship a different

approach of causality (proposed by Toda and Yamamoto, 1995) test is used to detect

whether there is much evidence of contagion hypothesis2. In this regard, we also follow

Tsutsui and Hirayama (2004) to examine the causality from the Tokyo (TSE) and Seoul

(KSE) to Shanghai (SSE) as because the closing stock prices are observed at different

hours of the day in these markets due to time differences. It is to note that among the

three stock markets, SSE closes last during a given day and closing prices in Tokyo and

Seoul on the same day are already available to investors in Shanghai. Thus, in a

2
There is widespread disagreement on the definition and application of the term contagion (for more
details, see, Forbes and Rigobon, 2002; Billio and Pelizzon, 2003). We take a simple case to define the
term and indicate that if the cointegrated markets provide evidence of causality, then there are some
evidences of contagion. Phengpis (2006) euphemistically puts that in most cases we ignore the simple
analytical techniques (e.g., statistical correlation) to explain the linkages and interdependencies among the
markets.

6
regression equation with the SSE price as a dependent variable, we should include prices

in the other two markets on the same day (along with past prices) as regressors.

Notwithstanding, the results from both the previous day and the same day closing prices

have not altered the conclusion. The test statistics indicate that even though the markets

are cointegrated, there is not much evidence of contagion between the markets with the

exception of bi-directional causality/contagion between Japan and South Korea.

The rest of the paper is organized as follows. Section 2 provides a brief review of

literature. Section 3 discusses the background of the markets. Section 4 explains the

methodology. Section 5 discusses the data, summary statistics and empirical results. The

paper ends with some concluding remarks in section 6.

2. Review of literature

The literature review can be divided into two: the ones that relate to the

methodological issues and the ones that relate to the Asian equity markets. The existing

body of literature uses different kinds of econometric methodologies ranging from the

univariate, bi-variate to multivariate tests to examine whether the financial market(s) is

(are) informationally efficient. The univariate tests indicate whether the respective

financial market is individually efficient. The bi-variate and multivariate tests indicate

whether one financial market incorporates sufficient information useful for creating

forecasts of another and to explain the efficiency of the other financial market(s).

The most frequently used univariate tests are the unit root tests, autocorrelation

tests, fractional integration/long memory tests, variance ratio test and so on. Lo and

MacKinlay (1988, 1989), Cecchetti and Lam (1994) and Gilmore and McManus (2003)

argue that the variance-ratio test is more reliable than the other univariate tests like,

7
traditional unit root tests and autocorrelation tests. Also, the variance ratio tests of Lo-

MacKinlay (1988) and Wright (2000) applied in this paper take into account of both the

conditions of homoscedasticity and heteroskedasticity in the relevant time series.

The frequently used bi-variate and multivariate tests are cointegration tests,

causality tests, panel unit root tests, panel cointegration tests and multivariate GARCH.

Of the several cointegration tests available, we apply Gregory-Hansen (1996)

cointegration tests that consider the endogenous breaks in the series 3. There are three

alternative procedures which can be applied when the variables are cointegrated. These

are the ECM-based Granger (1969) causality test, the likelihood ratio test (LR) suggested

by Mosconi and Giannini (1992) and the MWALD test (of Granger causality) suggested

by Toda and Yamamoto (1995) and Dolado and Lütkepohl (1996). The Monte Carlo

experiment, in Zapata and Rambaldi (1997), shows that the MWALD test has comparable

performance in size and power to that of ECM based test and LR tests if there are 50 or

more observations. The major motivation of the use of the Toda-Yamamoto (1995)

approach is attributed to its practical implication in explaining the long-term relationship

while at the same time examining the Granger-causality tests.

We now review some of the relevant literature4. For the individual market

efficiency, the study of Hoque, Kim and Pyun (2007) is a good start since they use the

VR methodology to examine the Asian stock market efficiency. They use the weekly data

and focus on the eight emerging markets in Asia including South Korea (but not Japan

and China) and find that the South Korea does not deviate from the market efficiency.

3
Phengpis (2006) summarizes the advantages and limitations of some of the cointegration procedures with
an application to the study of foreign exchange market efficiency.
4
For a survey of existing literature on individual market efficiency see Hoque, Kim and Pyun (2007) and
for literature on Asian equity price linkages, see Worthington, Katsuura and Higgs (2004).

8
Kim and Shamsuddin (2007) also report the similar results. They apply the VR

methodology to the daily and weekly data and find that the Japanese and South Korean

markets do not deviate from market efficiency. In earlier studies, Ayadi and Pyun (1994),

Ryoo and Smith (2002) among others fail to reject the random walk hypothesis for South

Korea but Huang (1995) rejects the random walk hypothesis for this country. There are,

of course, innumerable research on the Chinese and the Japanese stock markets and, there

are no significant disagreements among the researchers on the results, albeit, the common

wisdom applies to these markets: the former providing the evidence against the market

efficiency and the latter supporting the evidence of market efficiency.

Chan, Gup and Pan (1992, 1997) study the interlinkages among international

equity markets and interpret their findings of no cointegration among these markets as

evidence of joint market efficiency. The term „joint efficiency‟ is documented in other

studies too. Mishra, Rahman and Caples (2002) employ the cointegration and the

associated error-correction model to examine the joint efficiency in forward and futures

markets for foreign currencies. Hassapis, Kalyvitis and Pittis (1999) also use the

cointegration to investigate the joint efficiency in the international commodity markets

for four industrialized countries. Lence and Falk (2005) list a couple of studies that apply

the cointegration procedures to test for the market efficiency in equity markets, security

markets, foreign exchange markets, commodity markets and banking product markets.

However, Lence and Falk (2005) themselves argue that cointegration of asset prices may

not be used for assessing market integration and/or market efficiency but may be used to

draw inferences about preferences and endowment processes. Dwyer and Wallace (1992)

also argue that the cointegration is neither a necessary nor a sufficient condition for

9
market efficiency. Kühl (2007) does not explicitly use the term „joint market efficiency‟

but applies the cointegration procedures to test the informational efficiency in foreign

exchange markets. Worthington and Higgs (2004) and Olienyk, Schwebach and Zumwalt

(1999) interpret the evidence of Granger causal relationships between the cointegrated

markets as a violation of (joint) market efficiency. Lim, Gallo and Swanson (1998) also

argue that investors can devise trading strategies to exploit any inherent inefficiencies

between markets. Contrary to others, Roca (1999) clarifies the statistical cointegration

differently. Roca (1999) identifies the evidence of finding no cointegration (between

equity markets of Australia, Uk, US and other Asian countries) as good for long-term

portfolio diversification. Using three financial market variables namely, exchange rates,

stock price indices and interest rates, Khalid and Kawai (2003) test for contagion

hypothesis in the East Asian financial markets. They apply the Granger causality to nine

East Asian countries including Japan and South Korea and do not find strong support for

contagion. Baig and Goldfijn (1998), Masih and Masih (1999), Reside and Gochoco-

Bautista (1999), Jang and Sul (2004) and Pan, Fok and Liu (2007) among others apply

the VAR, associated ECM techniques and the Granger causality tests to study the

contagion in the East Asian stock markets. Jang and Sul (2002) take the co-movement as

a case of contagion, while Pan, Fok and Liu (2007) use the linkages between exchange

rates and stock prices as a case of contagion. Tai (2007) applies the asymmetric

multivariate GARCH, which provides the evidence of contagion from the stock markets

to the foreign exchange markets in the Asian emerging markets including South Korea.

3. Background of the market

10
 China: China‟s equity market has been in existence since 1990, when both the

Shanghai Stock Exchange (SSE) and the Shenzhen Stock Exchange (SHSE) were

created. China‟s equity market is characterized as less developed compared to Japan‟s

and South Korea‟s equity markets. Two types of shares are traded in the Chinese stock

markets: A shares for domestic investors and B shares for foreign investors. The stock

market experienced its first peak in the early 1990s led by intense speculative activity and

returned to more moderate levels in the mid-1990s. Wong (2006) finds that China‟s stock

market development during this period was driven primarily by rent-seeking and

speculative activities not by value-driven transactions between investors and fund

seekers. Throughout the 1990s, the market was characterized by frequent price

movements (see the left panel of Figure 1). The market experienced an upward trend

since the late 1990s to 2000. In 2000, the market capitalization, the liquidity and the

trading volume doubled from the previous year.

Figure 1: SSE price and return indices (expressed in logarithms)

5.75 0.1
CHINA
CHINA

-0.0
5.50

-0.1
5.25

-0.2

5.00
-0.3

4.75
-0.4

4.50 -0.5
1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006

Japan: Japan‟s equity market is the second largest in the world and the largest in

Asia Pacific in terms of the market capitalization. The history of her equity trading dates

back to the late 1800s when the Tokyo and Osaka stock exchanges were set up. At

present, Japan features equity trading in six exchanges namely, Tokyo, Osaka, Nagoya,

Fukuoka, Sapporo and JASDAQ. Tokyo Stock Exchange accounts for approximately 80

11
percent of market volume and capitalization, followed by Osaka Stock Exchange (15

percent) and the remaining regional stock exchanges (approximately 1 percent each).

Japan‟s equity market experienced both bubble and burst in the late 1980s, and, for a

brief period from 1989 to1990, market capitalization exceeded that of the US market.

Figure 2: Nikkei 225 price and return indices (expressed in logarithms)

5.50 0.150
JAPAN JAPAN

0.125
5.25
0.100

5.00 0.075

0.050
4.75
0.025

4.50 0.000

-0.025
4.25
-0.050

4.00 -0.075
1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006

South Korea: South Korea‟s equity market has been in existence since 1956 with

the set up of country‟s first exchange, the Daehan Stock Exchange (DSE). The DSE was

reorganized in 1962 as a joint-stock company. In 1963, the DSE became a non-profit

government entity and renamed the Korea Stock Exchange or Korea Exchange (KSE or

KE). South Korea‟s equity market is substantial in size but is very tightly regulated.

Unlike other Asia Pacific stock markets, the South Korean market has managed to sustain

a steady growth in listings, trading volume and market capitalization.

Figure 3: KOSPI price and return indices (expressed in logarithms)

4.6
SOUTHKOREA 0.32
SOUTHKOREA

4.4
0.24

4.2 0.16

4.0 0.08

3.8 0.00

3.6 -0.08

3.4 -0.16

3.2 -0.24
1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006

12
 As demonstrated in Figures 1 through 3, compared to the Japanese and Korean

stock markets, the Chinese stock markets have experienced extremely large price

movements indicating deviations from the market fundamentals. Such price movement or

volatility is attributed to the excessive speculation. Wong (2006) argues that the rapid but

vulnerable price movement/development in China‟s equity market is attributed to three

major factors. First, the government used the stock market as a vehicle of raising funds

for the state-owned enterprises. Second, the repressed financial regime did not allow free

flow of capital (especially capital flight) and competition among the financial assets.

Third, the legal framework was too week to offer shareholders sufficient protection.

Table 1 shows the correlation matrix as a rough measure of the stock price

linkages. The matrix shows the correlation between the price and returns between the

three markets. As the correlation matrix demonstrates, the East Asian stock markets react

both positively and negatively to the other markets. The correlation between the Japanese

stock returns and the Korean stock returns are strongly positive, while the correlation

between the Chinese stock returns with those of the Japanese and Korean stock returns

are strongly negative.

Table 1: Correlation matrix

China Japan Korea China Japan Korea
prices prices prices returns returns returns
China prices 1 .017 .006 -.040(*) .039(*) .024
Japan prices 1 .017 -.031 -.038 .030
Korea prices 1 -.020 -.046(*) .018
China returns 1 -.317(**) -.259(**)
Japan returns 1 .575(**)
Korea returns 1
* Correlation is significant at the 0.05 level (2-tailed).
** Correlation is significant at the 0.01 level (2-tailed).

4. Econometric methodology

13
4.1 Unit root test

We apply the Zivot-Andrew (hereafter, ZA, 1992) test, which allows for a

structural break in a series to mitigate the bias towards non-rejection of the unit root null

hypothesis while the series is in fact stationary but subject to a structural break date. The

ZA test allows the break date to be determined endogenously by the test equation. For a

possible break date, TSB, which ranges from the observation 0.15T to the observation

0.85T where T is the sample size, the unit root test equation can be estimated for each

stock price/return series, yi, as follows:

 y it   i   DU it
  y it 1    y i it  j
  it (1)
j 1

where the dummy variable (modeling for a structural change), DUt=1 for t> TSB, and zero

otherwise, and k is the number of augmented lags. The t-statistic for testing α=0 or tα is

computed for each TSB iteration. The smallest value of tα‟s computed for all TSB iterations

becomes the ZA test statistics under the null hypothesis that the stock price/return series,

yit is I(1) against the alternative hypothesis that it is I(0) with one structural break point. If

the null hypothesis is rejected, the TSB associated with the ZA statistic becomes TB or the

date at which a structural break date in a series transpires.

4.2 Variance-ratio (VR) tests and the calculation of critical values for the VR tests

4.2.1 Lo and MacKinlay VR

According to Lo and MacKinlay (1988), if yt is a time series of stock returns

with a sample of size T, the variance-ratio to test the hypothesis that yt is iid or that it is

an mds is defined as:

 1 T
2  1 T
2 
VR =   ( y t  y t 1 ...  y t  k  k ˆ )  ÷  (y t
 ˆ )  (2)
 Tk t  k  1  T t 1 

14
where, ̂  T  1  t 1 y t . The numerator of VR is 1/k times the variance of yt after
T

aggregation by a factor of investment horizon, k. This statistic should be close to 1 if yt is

iid/mds but not if it is serially correlated. L-M (1988) show that if yt is iid then

 2 ( 2 k  1)( k  1) 
(VR  1)  d N  0 ,
1/ 2
T  (3)
 3k 

So, the test statistic

1 / 2
 2 ( 2 k  1)( k  1) 
M1  (VR  1)   (4)
 3 kT 

follows the standard normal asymptotically under the iid null (homoscedasticity. To

accommodate for conditional heteroskedasticity in yt, Lo and MacKinlay (1988) propose

the following robustified test statistic:

1 / 2
 k 1  2 ( k  j )  2 
M  (VR  1)     j (5)
2
 j 1  k  
 

where

   T 
2
 T 2  
    ( y t  ˆ ) ( y t  j  ˆ )      ( y t  ˆ )  
2 2
j (6)
 t  j 1    t 1  

Lo and MacKinlay (1988) show that if yt is an iid/mds, then M2 is asymptotically

normally distributed with mean zero and standard deviation 1. The usual decision rule for

the standard normal distribution applies to both M1 and M2.

4.2.2 Wright’s alternative VR

Wright (2000) proposes ranks (R1 and R2) and signs (S1 and S2) as alternatives to

M1 and M2 of Lo and MacKinlay (1988):

15
  1 

T
( r1 t  r1 t 1     r1 t  k )
2
  1 / 2
t  k 1  2 ( 2 k  1)( k  1) 
R1    1  
Tk
 (7)
 1   3 kT 

T 2
 t 1 1t
r 
 T 

and

 1 

T
( r2 t  r2 t  1     r2 t  k )
2
  1 / 2
t  k 1  2 ( 2 k  1)( k  1) 
R2   
Tk
1   (8)
 1   3 kT 

T 2
 t 1
r2 t 
 T 

where,

T 1 (T  1)( T  1)
r1 t  r ( y t )  and r2 t    1  r ( y t ) (T  1) , (9)
2 12

Where, r (yt) be the rank of yt among y1, y2,…., yT.  is the standard normal cumulative

distribution function. The sign based tests are defined as follows:

 1 

T
( s t  s t 1     s t  k )
2
  1 / 2
t  k 1  2 ( 2 k  1)( k  1) 
S1   
Tk
1   (10)
 1   3 kT 

T 2
 t 1
s t 
 T 

Under the assumptions that the time series of stock return is an mds, if μ =0, then S1 has

the same distribution as:

 1 

T   
( s t  s t 1     s t  k )
2
  1 / 2

 Tk
t  k 1  2 ( 2 k  1)( k  1) 
 1    (11)
 1   3 kT 

T 2
 t 1
s t 
 T 

where, s t t 1 is an iid sequence, each element of which is 1 with probability 0.5 and -1
T

otherwise. The test S2, which is related to the conservative test that a series is a random

walk with drift (see, Campbell and Dufour, 1997), controls for the probability of Type I

error in finite samples and is robust to conditional heteroskedasticity. Wright (2000)

16
shows that both R1 and R2 have better power than either of the M1 and M2 tests. He also

shows that even though the sign-based tests generally have less power than the rank-

based tests, they still have more power than the Lo-MacKinlay tests.

We obtain the critical values for Wright‟s (2000) R1, R2 and S1 tests by simulating

their exact sampling distributions. Wright‟s (2000) S2 is not considered, because in Monte

Carlo simulation, Wright (2000) finds that the size and power properties of S2 are inferior

to those of S1. Table 2 shows the critical values for R1, R2 and S1 test statistics associated

with the sample sizes and holding periods.

Table 2: Critical values for Wright’s (2000) R1, R2 and S1 tests

k T=2500 T=3000
R1
2 -1.970 1.915 -1.973 1.964
5 -1.997 1.919 -1.971 1.971
10 -1.950 1.928 -1.948 1.957
20 -1.979 1.888 -1.974 1.916
30 -1.972 1.837 -1.974 1.967
40 -1.975 1.807 -1.885 1.935
R2
2 -1.965 1.919 -1.985 1.971
5 -2.025 1.933 -1.933 1.973
10 -1.965 1.913 -1.922 1.944
20 -1.974 1.852 -1.949 1.934
30 -1.975 1.848 -1.968 1.972
40 -1.950 1.828 -1.981 1.961
S1
2 -1.920 1.960 -2.010 1.972
5 -1.950 1.980 -1.953 1.953
10 -1.910 1.995 -1.932 1.923
20 -1.883 2.015 -1.922 1.899
30 -1.876 1.999 -1.913 1.908
40 -1.866 2.011 -1.885 1.935
Notes: The critical values were simulated with 10,000 replications in each case. For each entry, the
numbers in column 2 and 4 give the 2.5 percentile of the distribution of the test statistics (for specified
value of T and k) and the numbers in column 3 and 5 give 97.5 percentile of that distribution. The critical
values for other percentile of the distribution can be obtained from the author on request.

4.3 Engle-Granger and Gregory-Hansen cointegration tests

17
 The Engle-Granger (1987) cointegration test with no structural change (i.e., with

the cointegrating relationship being time-invariant) can be estimated as follows:

Y it   i   X t   it t = 1,…,n, (12)

where, Xt are the explanatory variables all with I(1) and εit is equilibrium error if

cointegration exists and hence I(0). Parameters μi (intercept) and β (slope) are the m-

dimensional hyperplane towards which the vector process Yit tends over time. This test

presumes that the parameters μi and β are time-invariant. Nevertheless, these parameters

may be time-invariant over fairly long period of time and then form a new “long-run”

relationship at some unknown point, which requires a change/shift in either the intercept

μi or the slope β or both intercept and slope. Gregory and Hansen (GH, 1996) consider

three types of such (structural) shifts in the cointegrating relationship: (i) level shift or

level break denoted by C (ii) level break with trend denoted by C/T and (iii) regime

shift/full break denoted by C/S. In the case of level shift, a structural change/break is

allowed in the intercept ui, while considering the slope coefficients β as constant. In the

second case, the structural change allows a time trend,  t in the level shift model. In the

third case, the structural change allows the slope vector to shift as well. Thus, the GH

procedure determines the breakpoint by finding the minimum value for the augmented

Dickey-Fuller (ADF) statistic. The three cases are as follows:

(i) Level shift (C)

Y it   i1   i 2  t   
T
X t   it t = 1,…,n, (13)

(2) Level shift with trend (C/T)

Y it   i1   i 2  t    t   X t   it
T
t = 1,…,n, (14)

(3) Regime shift (C/S)

18
 Y it   i1   i 2  t    1 X t   2 X t  t    it
T T
t = 1,…,n, (15)

Although the null hypothesis is same in both the tests, the alternative hypothesis

of the Gregory-Hansen (1996) test takes the Engle-Granger (1987) test as a special

subclass when the cointegrating vector shifts at one unknown break point.

4.4 The MWALD test for Granger causality in cointegrated system

The contagion hypothesis is tested using one of the several alternative procedures

available when the variables in the system are cointegrated. We adopt the Toda and

Yamamoto (1995) tests for Granger causality to examine the contagion hypothesis in

these three equity markets. To simplify the Toda and Yamamoto procedure, we follow

Rambaldi and Doran (1996).

Let us take that dmax is the maximum order of integration in the system (i.e., return

series in our case). In cointegrated systems, the Wald test for linear restrictions on the

parameters of a VAR(k) has an asymptotic χ2 distribution when a VAR(k+dmax) is

estimated. Preliminary tests are performed to determine the lag length, k (=3), in the VAR

using Akaike Information Criteria (AIC). Let China, Japan and South Korea be denoted

by y, z and w respectively. Since dmax =1, we need to estimate a VAR (4) using the

following system of equations:

 yt   y t 1   yt2   y t3   y t 4  e y 
           
z  A 0  A1 z t  1  A 2 z t  2  A 3 z t  3  A 4 w t  4   e z  (16)
 t         
 w t   w t 1   w t  2   w t  3   z t  4   e w 
 

The above system of equations is estimated by using the method of seemingly

unrelated regression (SUR). If we want to test that Japan does not Granger-cause China,

the null hypothesis becomes H0: a 12(1 )  a 12( 2 )  a 12( 3 )  0 where a 12( i ) are the coefficients of zt-

19
i, i=1, 2, 3, in the first equation of the system. The other null hypothesis can be defined

similarly.

5. Data, summary statistics and empirical results

The study covers the daily closing price indices namely, Shanghai SE Composite

(China), Nikkei 225 Stock Average (Japan) and Korea SE Stock Price Index (South

Korea) from July 2, 1996 to December 24, 20065 giving a total of 2650 observation. All

data are collected from the Global Financial Data.

It is worthwhile to explain a bit about the price indices that are used in the

analysis. First, the Shanghai Stock Price Index that we used is a capitalization-weighted

index. This index tracks the daily price performance of all A-shares and B-shares listed

on the Shanghai Stock Exchange. The index was developed on December 19, 1990 with a

base value of 100. The Global Financial Data series ID of this index is _SSECD. Second,

for the Japanese stock market, the Nikkei 225 stock average is taken. The Global

Financial Data series ID of this index is _N225D. The index was developed on January 4,

1968 with a base value of 100. Third, for the South Korean stock market, the Korea SE

Stock Price Index (KOSPI) is taken. The Global Financial Data series ID of this index is

_KS11D. For details, the relevant series of the Global Financial Data can be consulted.

Table 3 shows the descriptive statistics for these three markets.

Table 3: Summary statistics (N=2650)

SSE Nikkei 225 KOSPI
Sample Mean 0.00039 -0.000114 0.00021
Variance 0.000358 0.000270 0.000642
Skewness -5.72 0.18 0.022
(0.000) (0.00016) (0.64)
Kurtosis (excess) 137.20 2.68 10.93

5
For missing data owing to holidays in one market while operational days in other markets, the previous
day‟s closing price is used. The tests are also used on the data set by deleting the missing observations.
However, the inference from the tests is not altered.

20
 (0.000) (0.000) (0.000)
Jarque-Bera 2092958.96 807.56 13194.169
(0.000) (0.000) (0.000)
Ljung-Box Q-Statistics (lag = 12) 22.58 23.73 137.63
(0.032) (0.022) (0.024)
L-M 3.459 1.315 5.507
(0.000) (0.000) (0.000)
The mean return in SSE is positive. However, the negative skewness reveals the

fact that the Chinese stock return is non-symmetric, non-normal and has a long tail to the

left. The large excess kurtosis indicates that the underlying data is leptokurtic and the

Chinese stock market is speculative in nature. The JB test indicates that the return from

the Chinese stock market is non-normal and the Q statistics indicates that there is

evidence of autocorrelation in the stock return series. Unlike the SSE return, the mean

return from the Japanese stock market is negative, while the skewness is positive

indicating that the return series is symmetric and normally distributed. But the JB and Q

stats are significant at 1 and 5 percent levels of significance, respectively. The kurtosis is

less than 3, which means that there is not much evidence of speculation in the Japanese

stock market. For the KOSPI, we find that both the mean and the skewness of the return

series are positive. Nonetheless, the kurtosis tends to have a distinct peak near the mean,

declines rather rapidly, and has heavy tails. This indicates that the Korean stock market is

speculative in nature. Both the JB and Q stats are statistically significant at 1 percent

level of significance. The LM indicates the Lagrange multiplier test for the presence of

the conditional heteroskedasticity (ARCH) effect with up to 30 lags. The LM statistics for

all the three markets are significant at the 1% level.

Other univariate statistics, namely, the unit root test statistics and variance ratio

tests results are shown in Table 4 and Table 5, respectively. The unit root test statistics

indicate that the three equity markets examined identify a unit root component in the

21
price series but not the return series. The results for the return series are significant at 1%

level indicating the stationarity of the stock price returns from these markets. The ZA

statistics also gives the structural break date TSB. Since the Chinese markets are

characterized by high price movements, the ZA test statistics indicate a structural shift in

Chinese markets long before the structural shift in other markets. However, as the price

series features a unit root component in all series, the return series should be taken to

explain the structural break date more meaningfully. As evidenced from table 4, South

Korea has undergone a structural break in July 29, 1998 followed by Japan in June 6,

2000 and China in December 6, 2000. This indicates that South Korea might have led the

regulatory reforms in relation to the equity markets in the East Asia. We find that Korea

Exchange systematically introduced the reforms before it fully lifted the foreign

ownership restrictions in 1998. In the long way to this, the order-routing system in KSE

was automated in 1983 enabling member firms to electronically transmit orders to the

trading floor. The trading system was fully automated in 1997 allowing the market to

operate without the trading floor. Since the early 1980s, the Korean stock market was

gradually opened to foreign investors with the introduction of international investment

trusts and country funds (e.g., Korea Fund). In 1992, the Korean stock market was

opened to foreign investors with certain restrictions, and the foreign share ownership

restrictions were gradually lifted and were fully eliminated in 19986. This major reform

can be linked to the structural break date identified for the KSE in 1998. In the SSE case,

there are some reforms information, which can be connected to the break period

identified. In the early 2000, the Chinese government undertook a series of reforms to

privatize listed enterprises, remove the restrictive barriers in the financial sector, and
6
Related to the reforms in the Korean stock markets, see the KSE homepage.

22
improve the legal protection for shareholders. The Chinese stock market responded very

well to such reforms. By the end of 2000, stock market capitalization in Chinese stock

markets rose to more than US$507 billion to make it the second largest in Asia, after

Japan. The trading volume almost doubled in 2000 before it continued to slump again

since 2001. Interestingly, such a rapid growth in the liquidity and market capitalization

occurred during the period of an opposite picture in many other transitional economies,

which were rather plagued by low market capitalization and low liquidity7.

There is also relevant information associated with break period identified for the

Tokyo Stock Exchange (TSE). On 1st March 2000, Hiroshima and Niigata stock

exchanges merged into TSE. However, the effect of such merger was only vivid on

March 10 of the same year. Because of the merger, the trading volume on March 10,

2000, was shown to be two times higher than the previous day. Nonetheless, the break

date identified by the ZA test differs from that of the actual change in the relevant time

series. We argue that this difference originates from the data mining problem.

Table 4: The ZA unit root test statistics on daily stock prices and returns
Markets Price series Return series
ZA test statistics TSB ZA test statistics TSB
China -4.17959 1999:09:22 -15.803** 2000:12:06
Japan -2.40775 2000:10:27 -16.451** 2000:06:06
South Korea -1.93358 2003:04:07 -16.856** 1998:07:29
Critical Values are -5.34 and -4.80 at 1% and 5%, respectively

Let‟s look at the variance ratio test statistics of these stock markets. The variance-

ratio tests are examined for several values of k (holding period of investment). Although

the random walk hypothesis can be strongly rejected when the test statistics are rejected

for all k, the hypothesis can also be soundly rejected if there are more than two rejections

(Hoque, Kim and Pyun, 2007). The inferences from the L-M variance ratio tests are
7
See Wong (2006) for more details on the reforms in Chinese stock markets.

23
different from those of the Wright‟s tests. The results from these tests for the three equity

markets are shown in Table 5. The hypothesis that the Chinese stock markets exhibit

iid/mds null cannot be rejected under the L-M tests but can be strongly rejected according

to the Wright‟s tests. Yet, the analysis of Class A and Class B shares in the Chinese

equity markets may give different results. Lima and Tabak (2004) report such findings.

Lima and Tabak (2004) also use the variance-ratio tests and find that Class A shares are

weak-form efficient but Class B shares do not follow the random walk hypothesis. In

such a case, liquidity and market capitalization may play an important role in explaining

the inter-temporal efficiency of the developing and emerging equity markets. The

empirical result on the random-walk process of the SSE is consistent with earlier studies.

Darrat and Zhong (2000) Ma and Barnes (2001) and Seddighi and Nian (2004) among

others demonstrate the similar evidence (that the Chinese stock markets do not follow

random walk).

For the Japanese and Korean markets, both the conventional and rank & sign

based variance-ratio tests give fairly consistent results, rejecting the iid/mds null at some

(for instance, for Japan, at k <10 and, for Korea at k =2 and 40) but not all k. Thus, the

variance-ratio tests indicate that the Japanese and South Korean markets are efficient

indicating the evidence of random walk behavior. Finding the Japanese market efficient

is not in disagreement with the general notion that the developed markets, by and large,

constitute strong evidence of random walk behavior. But empirical literature on the

efficiency of South Korean market is divided. The results from this paper relating to the

South Korean market are in agreement with Hoque, Kim and Pyun (2007) but differ with

24
Mun and Kee (1994)8, Huang (1995) and Chaudhury and Wu (2003). Hoque, Kim and

Pyun (2007) find that the South Korean market appears on balance unaffected even by

further opening of its equity market following the Asian financial crisis. The timing and

intensity of Asian crisis is not considered in the present study since there is a vast

literature on this and it is not expected to add significant contributions to the literature.

Malliaropulos and Priestley (1999) relate the failure to reject the random walk hypothesis

(for the South East Asian markets including Japan and South Korea) to mean-reversion of

expected returns rather than market inefficiency.

Table 5: VR test statistics on daily stock returns

k M1 M2 R1 R2 S1
China
2 2.62** 1.57 2.78** 2.87** 1.18
5 2.69** 1.67 2.76** 2.71** 0.91
10 1.15 0.74 2.82** 2.15** 1.01
20 0.90 0.61 4.11** 2.86** 2.18**
30 0.86 0.60 4.44** 2.99** 2.55**
40 0.75 0.54 4.75** 3.12** 2.89**
Japan
2 -2.10** -1.82 -2.61** -2.40** -2.46**
5 -2.24 ** -1.95 -2.34** -2.32 ** -2.24**
10 -1.50 -1.31 -1.80 -1.62 -1.70
20 -1.30 -1.15 -1.17 -1.25 -0.98
30 -1.16 -1.03 -0.89 -1.06 -0.70
40 -1.04 -0.93 -0.50 -0.82 -0.26
South Korea
2 4.50** 2.58** 4.04** 4.48** 2.51**
5 -1.25 -0.60 1.32 0.90 0.94
10 -0.38 -0.18 0.77 0.62 0.92
20 0.62 0.30 1.04 1.14 1.74
30 0.79 0.39 0.85 1.04 1.77
40 0.88 0.45 0.87 1.05 2.07*
Notes: M1 and M2 are asymptotically distributed standard normal, while M1 is under the conditions of
homoskedasticity and M2 is under the conditions of heteroskedasticity. **rejections are significant at the
1% level, *rejections are significant at the 5% level.

8
Mun and Kee (1994) apply L-M variance-ratio tests along with other tests.

25
 Based on the findings from the univariate tests, the hypothesis of (individual)

market efficiency can be strongly rejected for the Chinese stock markets and, we firmly

say that the returns from the Chinese stock markets are predictable. But the hypothesis

(of individual market efficiency) could not be rejected for the Japanese and South Korean

stock markets and hence, returns from these markets are unpredictable.

Now, we report the results of cointegration tests to see whether the Chinese

markets are jointly efficient with those of Japanese and Korean markets. To facilitate this,

both the Engle and Granger (1987) and the Gregory and Hansen (1996) cointegration

tests are considered. Table 6 shows the test statistics of the Engle and Granger

cointegration test that does not consider the structural break and the Gregory and Hansen

cointegration tests that consider the structural break in the cointegration space. Although

a cointegrating relation is evident in both cases, Gregory and Hansen tests imply that this

cointegrating relation is time-variant and has shifted at some break point.

Table 6: Cointegration tests: Engle-Granger vs. Gregory-Hansen

Equation Engle- Gregory-Hansen Gregory-Hansen
Granger (Level break) (Full break)
(ADF test ADF test TSS ADF test TSS
statistics) statistics statistics
China -14.013** -14.060** 2001:09:12 -14.063** 2001:03:13
Japan -16.667** -16.760** 2003:06:19 -16.751** 1998:03:16
South Korea -16.257** -18.448** 1998:01:14 -16.493** 1998:08:04
Notes: 7 lags selected automatically; Critical values for the ADF test statistics obtained from MacKinnon
(1996) are -4.68 and -4.12 at the 1% and 5% levels, respectively; Critical values for GH cointegration test
are -5.44 and -4.92 at 1% and 5% level of significance for the level break and -5.97 and -5.50 at 1% and
5% level of significance at full break.

In table 6, column 1 indicates cointegration test equation in which the country

listed is a dependent variable. Column 2 shows the ADF test statistics under the null

hypothesis of no cointegration based on the Engle-Granger cointegration procedure in

which a possible shift in a cointegration relation is not considered. For the Gregory-

26
Hansen test, the cointegration equation is allowed to have an endogenous shift in its

intercept (level break) and a shift in the slope vector (full break). The ADF test statistics

in the case of Gregory-Hansen cointegration is under the null hypothesis of no

cointegration against the alternative hypothesis of a cointegrating relation with a

structural shift. It is the smallest value of the ADF statistics calculated for all possible

dates for a structural shift. TSS is the possible structural shift associated with the ADF.

Both the Engle-Granger and Gregory-Hansen tests strongly reject the null hypothesis of

no cointegration between the markets. Noted earlier, multivariate tests like the

cointegration tests elucidate whether one financial market incorporates sufficient

information useful for creating forecasts of another and to explain the efficiency of the

other financial market(s). The evidence of cointegrating relationship indicates that there

are common stochastic trends shared by the three markets and that returns from one are

predictable in terms of information in another.

The cointegration tests confirm that there is a long-term equilibrium relationship

between the three markets but the relationship is not time-invariant rather subject to some

structural shifts. Worthington, Katsuura and Higgs (2004) and Worthington and Higgs

(2004) find the similar results in terms of the long-run relationship between these

markets. While Worthington, Katsuura and Higgs (2004) find a significant relationship

among three developed and six emerging markets in the Asia Pacific, Worthington and

Higgs (2004) find a causal relationship between China-Japan and a unidirectional

causality running from Japan to Korea. According to them, these causal relationships

suggest opportunities for international portfolio diversification in Asian equity markets

still exist. In an earlier study, Pan, Liu and Roth (1999), who use the modified

27
cointegration test with GARCH effect, demonstrate that the six Asia Pacific stock

markets including the United States are integrated through the second moments of stock

returns but not the first moments. The evidence of cointegrating relationship is

characterized as joint market inefficiency and implies that at least one of the markets

within the VAR is inefficient even though the rest of the markets may be individually

efficient (see, Hasapis, Kalyvitis and Pittis, 1999). Not surprisingly, even though

Japanese and South Korean stock markets are found to be individually efficient (from the

unit root and variance ratio tests) but are found to violate the market efficiency under the

system of cointegration. This is due to the individual market inefficiency of the Chinese

market. The evidence of joint market inefficiency suggests the potential for short-term

arbitrage opportunities and the potential for international portfolio diversification.

However, Allen and MacDonald (1995), Richards (1995), DeFusco et al. (1996) and

others, argue that if the equity markets are cointegrated, the long-run international

diversification potential are limited and hence, diversifying between the integrated

markets over long time is not likely to generate large benefits through risk reduction.

Responding to the last (contagion) hypothesis of this paper, since the

cointegration tests identify a long-term relationship among the markets, we proceed to the

application of Toda-Yamamoto procedure of Granger causality in order to identify

whether these markets are contagious to the other markets during any volatility shock.

The test statistics identify bi-variate causality between South Korea and Japan. Therefore,

these two markets are interdependent and at the same time contagious to the volatility

shocks in its counterpart. Even though Chinese market is cointegrated with Japanese and

South Korean markets, the Chinese market is not contagious to the volatility shocks in

28
other markets. The null hypothesis and the results from Toda-Yamamoto tests of Granger

causality are shown in table 7.

Our results from the causality are consistent with the findings of Baig and

Goldfijn (1998), who also demonstrate less supportive evidence for stock market

contagion in the region. However, our results are not in agreement with those of Masih

and Masih (1999), who rather confirm a contagion within Asian stock markets. It is to

note that of the three markets considered, trading hours of the Japanese and the South

Korean stock markets are one hour ahead of that of the Chinese stock markets (SSE).

Since the Shanghai Stock Exchange (SSE) closes one hour after the closing of the Tokyo

Stock Exchange (TSE) and the Korea Stock Exchange (KSE) and thus, today‟s closing

prices in Tokyo/Seoul are already known to investors in Shanghai when the SSE closes,

we examine whether the inference is altered due to time differences. Even after taking the

effect of the same day‟s closing prices of the TSE/KSE on the SSE, we do not find

significant causality running from the TSE/KSE to the SSE. A similar result is reported

by Jang and Sul (2002). Adjusting time difference considering the closing hours, Jang

and Sul (2002) find no significant causality in the East Asian markets. The evidence of

non-causality (no contagion) indicates that knowledge of past return behavior in one

market is unlikely to improve forecasts of returns in another, except for some causality

running from South Korea to Japan and Japan to South Korea.

Table 7: Toda-Yamamoto tests of Granger causality

Null hypothesis Test statistics P value
(χ2)
China does not Granger cause Japan 0.984 0.805
Japan does not Granger cause China [taking no effect of same 0.609 0.894
day‟s closing prices of the TSE on SSE- up to VAR(4)]
Japan does not Granger cause China [taking the effect of same 1.493 0.684
day‟s closing prices of the TSE on SSE - up to VAR(4)]

29
Japan does not Granger cause China [taking the effect of same 1.551 0.671
day‟s closing prices of the TSE on SSE - up to VAR(5)]
China does not Granger cause South Korea 1.508 0.680
South Korea does not Granger cause China [taking no effect 0.156 0.984
of same day‟s closing prices of the KSE on SSE- up to
VAR(4)]
South Korea does not Granger cause China [taking the effect 0.261 0.967
of same day‟s closing prices of the KSE on SSE - up to
VAR(4)]
South Korea does not Granger cause China [taking the effect 0.669 0.955
of same day‟s closing prices of the KSE on SSE - up to
VAR(5)]
South Korea does not Granger cause Japan 17.855 0.0005
Japan does not Granger cause South Korea 11.391 0.0098
Related to the above findings on individual and joint market efficiency, the

investors in these markets may try to exploit the cases of market inefficiency to generate

market returns. But they are advised to be much cautious in forecasting the returns. It is

because, other investors may also try to exploit the opportunity of informational

inefficiency as observed in these markets. Although an identification of the short-term

relationship through ECM based Granger causality might bring some profits for the

investors, the (long-term) causal relationship (identified through Toda-Yamamoto level

VAR) between Japan and South Korea indicates that there would be no long range

benefits from pair-wise portfolio diversification between these two markets. For other

non-causal relationships, for example, China-Japan and China-Korea, where both the

long-run and short-run differences may exist, the knowledge of one of these markets

would seldom help forecast the return from the other market(s) and therefore, the gain

from the portfolio diversification is generally limited.

6. Concluding remarks

This paper explains market efficiency, interdependences and contagion of the

three East Asian stock markets using both univariate and multivariate tests. The study

30
indicates that the Chinese stock market is (informationally) inefficient while Japanese

and South Korean stock markets are efficient. Yet, the three markets are jointly

inefficient in the cointegrating sense. Thus, even though the Japanese and South Korean

stock markets are individually efficient, these markets are not jointly efficient under the

system of cointegration due to inefficiency of the Chinese stock market. Further, while

these three markets are cointegrated, the contagion effect exists only between Japanese

and South Korean stock markets. The market inefficiencies, both in individual and

cointegration sense, give the speculators the chance to manipulate the prices. Any short-

term price movements may persuade the weak market players to wrongfully estimate the

returns from their stocks. Therefore, we suggest the regulators to control unrealistic price

movements in order to protect the interests of the weak market players and to facilitate

the potential development of the capital markets. At the same time, the investors need to

be cautious about the information flows, the noise and clear/overall understanding of the

markets they are interested in. The results in this study are not against the general wisdom

that the developed and emerging markets, by and large, constitute the market efficiency,

while the underdeveloped markets do not. Since we find that the long-term relationships

between the East Asian equity markets are not stable over time, it is not possible that the

identification and the use of an error correction model will indicate short-term arbitrage

opportunities. Although the causal relationship between Japan and South Korea suggests

market inefficiencies and possible short-term arbitrage opportunities, the evidence of

non-causality (no contagion) for other pairs namely, China-Korea and China-Japan

indicates that knowledge of past return behavior in one market is unlikely to improve

forecasts of returns in another. The future researchers may use more high frequency data

31
and try to explore the underlying reasons why some of the East Asian markets provide

the evidence of market efficiency and why some others do no, under the univariate and/or

multivariate or in both univariate and multivariate framework.

References

Allen, D.E. and G. MacDonald, 1995, The long-run gains from international equity
diversification: Australian evidence from cointegration tests. Applied Financial
Economics, 5, pp. 33-42.
Baig, T. and I. Goldfijn, 1998 (November), Financial market contagion in the Asian
Crisis. IMF Working Paper No. WP/98/155. Washington, DC: International
Monetary Fund.
Bernstein, P., 1992, Capital Ideas: The Improbable Origins of Modern Wall Street. Free
Press, New York.
Billio, M. and L. Pelizzon, 2003, Contagion and interdependence in stock markets: Have
they been misdiagnosed? Journal of Economics and Business, 55, pp. 405-426.
Campbell, B. and J.M.Dufour, 1997, Exact nonparametric tests of orthogonality and
random walk in the presence of a drift parameter. International Economic Review,
38, pp. 151-173.
Campbell, J.Y., A.W. Lo and A.C. MacKinlay, 1997, The Econometrics of Financial
Markets. Princeton University Press, New Jersey.
Cecchetti, S.G. and P. Lam, 1994, Variance ratio tests: small-sample properties with an
application to international output data. Journal of Business and Economic
Statistics, 12, pp. 177-186.
Chan, K.C., B.E. Gup and M.S. Pan, 1992, An empirical analysis of stock prices in major
Asian markets and the United States. Financial Review, 27, pp. 289-307.
Chan, K.C., B.E. Gup and M.S. Pan, 1997, International stock market efficiency and
integration of eighteen nations. Journal of Business Finance Accounting, 24, pp.
803-813.
Chaudhury, K. and Y. Wu, 2003, Mean Reversion in Stock Prices: Evidence from
Emerging Markets. Managerial Finance, 29, pp. 22-27.
Darrat, A.F. and M. Zhong, 2000, On testing the random-walk hypothesis: a model
comparison approach. The Financial Review, 35, pp. 105–24.
DeFusco, R.A., J.M. Geppert and G.P. Tsetsekos, 1996, Long-run diversification
potential in emerging stock markets. The Financial Review, 31, pp. 343–363.
Dolado, J.J. and H. Lütkepohl, 1996, Making Wald test work for cointegrated VAR
systems. Econometric Reviews, 15, pp. 369-386.
Dwyer, G.P. Jr. and M.S. Wallace, 1992, Cointegration and market efficiency. Journal of
International Money and Finance, 11, pp. 318-327.
Engle, R.F. and C.W.J. Granger, 1987, Cointegration and error correction:
Representation, estimation, and testing. Econometrica, 55, pp. 251-276.

32
Fama, E.F., 1970, Efficient capital markets: A review of theory and empirical work.
Journal of Finance, 25, pp. 383-417.
Forbes, K.J. and R. Rigobon, 2002, No Contagion, Only Interdependence: Measuring
Stock Market Comovements. Journal of Finance, 62, pp. 2223-2261.
Gilmore, C.G. and G.M. McManus, 2003, Random walk and efficiency tests of Central
European equity markets. Managerial Finance, 29, pp. 42-61.
Granger, C.W.J., 1969, Investigating Causal Relations by Econometric Methods and
Cross-Spectral Methods. Econometrica, 34, pp. 424-438.
Granger, C. W. J., 1986, Developments in the study of cointegrated economic variables.
Oxford Bulletin of Economics and Statistics, 48, pp. 213-228.
Gregory, A.W. and B.E. Hansen, 1996, Residual-based tests for cointegration with model
in regime shifts. Journal of Econometrics, 70, pp. 99-126.
Hassapis, C., S. Kalyvitis and N. Pittis, 1999, Cointegration and joint efficiency of
international commodity markets. The Quarterly Review of Economics and
Finance, 39, PP. 213-231.
Hoque, H.A.A.B., J.H. Kim and C.S. Pyun, 2007, A comparison of variance-ratio tests of
random walk: A case of Asian emerging stock markets. International Review of
Economics and Finance, 16, pp. 488-502.
Huang, B. N., 1995, Do Asian stock market prices follow random walks? Evidence from
variance ratio test. Applied Financial Economics, 5, pp. 251–256.
Janakiramanan, S. and A.S. Lamba, 1998, An Empirical Examination of Linkages
Between Pacific-Basin Stock Markets. Journal of International Financial
Markets, Institutions and Money, 8, pp. 155-173.
Jang, H. and W. Sul, 2002, The Asian financial crisis and the co-movement of Asian
stock markets. Journal of Asian Economics, 13, pp. 94-104.
Khalid, A.M. and M. Kawai, 2003, Was financial market contagion the source of
economic crisis in Asia? Evidence using a multivariate VAR model. Journal of
Asian Economics, 14, pp.131–156
Kim, J.A. and A. Shamsuddin, (forthcoming), Are Asian stock markets efficient?
Evidence from new multiple variance ratio tests. Journal of Empirical Finance.
Kühl, M., 2007, Cointegration in the foreign exchange market and market efficiency
since the introduction of the Euro: Evidence based on bivariate cointegration
analyses. Centre for European Governance and Economic Development
Discussion Paper # 68, October.
Lence, S. and B. Falk, 2005, Cointegration, market integration and market efficiency.
Journal of International Money and Finance, 24, pp. 873-890.
Lim, E.S., J.G. Gallo and P.E. Swanson, 1998, The relationship between international
bond markets and international stock markets. International Review of Financial
Analysis, 7, pp. 181-190.
Lima, E. J. A. and B.M. Tabak, 2004, Testing of the random walk hypothesis for equity
markets: Evidence from China, Hong Kong and Singapore. Applied Economics
Letters, 11, pp. 255-258.
Lo, A. and A.C. MacKinlay, 1988, Stock market prices do not follow random walks:
Evidence from a sample specification test. Review of Financial Studies, 1, pp. 41-
66.

33
Lo, A. and A.C. MacKinlay, 1989, The size and power of the variance-ratio test in finite
samples: A Monte Carlo investigation. Journal of Econometrics, 40, pp. 203-238.
Ma, S. and M.L. Barnes, 2001, Are China‟s stock markets really weak-form efficient?
Discussion Paper, No # 0119, Adelaide University.
MacKinnon, J.G., 1996, Numerical distributions functions for unit root and cointegration
tests. Journal of Applied Econometrics, 11, pp. 601-618.
Malliaropulos, D. and R. Priestley, 1999, Mean reversion in Southeast Asian stock
Markets. Journal of Empirical Finance, 6, pp. 355-384.
Masih, A.M.M. and R. Masih, 1999, Are Asian stock market fluctuations due mainly to
intra-regional contagion effects? Evidence based on Asian emerging stock markets.
Pacific-Basin Finance Journal, 7, pp. 251-282.
Mishra, B., M. Rahman and S. Caples, 2002, Cointegration, error-correction, joint
efficiency in forward and futures markets for major foreign currencies. Global
Finance Journal, 3, pp. 171-180.
Mosconi, R. and C. Giannini, 1992, Non-causality in cointegrated systems:
Representation, estimation and testing. Oxford Bulletin of Economics and
Statistics, 54, pp. 399-417.
Mun, F.W. and K.S. Kee, 1994, Do Asian stock market prices follow martingales?
Evidence from spectral shape tests. Asia-Pacific Journal of Management, 11, pp.
345-359
Olienyk, J.P., R.G. Schwebach and J.K. Zumwalt, 1999, WEBS, SPDRs, and country
funds: an analysis of international cointegration. Journal of Multinational
Financial Management, 9, pp. 217-232.
Pan, M.-S., R.C.-W. Fok and Y.A. Liu, 2007, Dynamic linkages between exchange rates
and stock prices: Evidence from East Asian markets. International Review of
Economics and Finance, 16, pp. 503-520.
Pan, M.-S., Y. A. Liu and H.J. Roth, 1999, Common stochastic trends and volatility in
Asia-Pacific Equity Markets. Global Finance Journal, 10, pp. 161-172.
Phengpis, C., 2006. Market efficiency and cointegration of spot exchange rates during
periods of economic turmoil: Another look at European and Asian currency
crises. Journal of Economics and Business, 58, pp. 323-342.
Rambaldi, A. and H. Doran, 1996, Testing for Granger non-causality in cointegrated
systems made easy. Working Papers in Econometrics and Applied Statistics # 88,
University of New England.
Reside, R.E., Jr. and M.S. Gochoco-Bautista, 1999, Contagion and the Asian currency
crises. The Manchester School, 67, pp. 460-474.
Richards, A.J., 1995, Co-movements in national stock market returns: evidence of
predictability, but not cointegration. Journal of Monetary Economics, 36, pp. 631-
654.
Roca, E.D. 1999, Short-term and long-term price linkages between the equity markets of
Australia and its major trading partners. Applied Financial Economics, 9, pp. 501-
511.
Seddighi, H. R. and W. Nian, 2004, The Chinese stock exchange market: Operations and
efficiency. Applied Financial Economics, 14, pp. 785-797.
Samuelson, P.A., 1965. Proof that properly anticipated prices fluctuate randomly.
Industrial Management Review, 6, pp. 41-49.

34
Tai, C.-S., 2001, A multivariate GARCH in mean approach to testing uncovered interest
parity: Evidence from Asia-Pacific foreign exchange markets. The Quarterly
Review of Economics and Finance, 41, pp. 441-460.
Tai, C.-S., 2007, Market integration and contagion: Evidence from Asian emerging stock
and foreign exchange markets. Emerging Markets Review, 8, pp. 264-283.
Toda, H.Y. and T. Yamamoto, 1995, Statistical inference in vector autoregressions with
possibly integrated processes. Journal of Econometrics, 66, pp. 225-250.
Tsutsui, Y. and K. Hirayama, 2004, Appropriate lag specification for daily responses of
international stock markets. Applied Financial Economics, 14, pp. 1017-1025.
Wong, S.M.L., 2006, China‟s stock market: A marriage of Capitalism and Socialism.
Cato Journal, 26, 389-424.
Worthington, A.C., M. Katsuura and H. Higgs, 2003, Price linkages in Asian equity
markets: Evidence bordering the Asian economic, currency and financial crises.
Asia-Pacific Financial Markets, 10, pp. 29-44.
Worthington, A. and H. Higgs, 2004, Co-movements in Asia-Pacific equity markets:
Developing patterns in APEC. Asia-Pacific Journal of Economics and Business,
8, pp. 79-93.
Wright, J.H., 2000, Alternative variance-ratio tests using ranks and signs. Journal of
Business and Economic Statistics, 18, pp. 1-9.
Zapata, H.O. and A.N. Rambaldi, 1997, Monte Carlo evidence on cointegration and
causation. Oxford Bulletin of Economics and Statistics, 59, pp. 285-298.
Zivot, E. and D.W.K. Andrews, 1992, Further evidence on the great crash, the oil-price
shock, and the unit-root hypothesis. Journal of Business & Economic Statistics,
10, pp. 251-270.

35