Control 1 Finanzas

The Singapore Economic Review, Vol. 57, No.
3 (2012) 1250021 (12 pages)

© World Scientific Publishing Company
DOI: 10.1142/S021759081250021X
THE EFFICIENT MARKET HYPOTHESIS REVISITED: EVIDENCE

FROM THE FIVE SMALL OPEN ASEAN STOCK MARKETS
QAISER MUNIR*, KOK SOOK CHING, FUMITAKA FUROUKA

and KASIM MANSUR
School of Business and Economics
by UNIVERSITY OF CALIFORNIA @ DAVIS on 01/22/15. For personal use only.
Universiti Malaysia Sabah, Locked Bag 2073

88999 Kota Kinabalu, Sabah, Malaysia
Singapore Econ. Rev. 2012.57. Downloaded from www.worldscientific.com
Published 18 September 2012
The efficient market hypothesis (EMH), which suggests that returns of a stock market are unpre-
dictable from historical price changes, is satisfied when stock prices are characterized by a random
walk (unit root) process. A finding of unit root implies that stock returns cannot be predicted. This
paper investigates the stock prices behavior of five ASEAN (Association of Southeast Asian Nations)
countries i.e., Indonesia, Malaysia, Philippines, Singapore and Thailand, for the period from 1990:1
to 2009:1 using a two-regime threshold autoregressive (TAR) approach which allows testing non-
linearity and non-stationarity simultaneously. Among the main findings, our results indicate that
stock prices of Malaysia and Thailand are a non-linear series and are characterized by a unit root
process, consistent with the EMH. Furthermore, we find that stock prices of Indonesia, Philippines
and Singapore follow a non-linear series, however, stock price indices are stationary processes that
are inconsistent with the EMH.
Keywords: Efficient market hypothesis; threshold autoregressive model; unit root.
JEL Classification: C22, C52
1. Introduction
The issue of whether stock prices can be characterized as following a random walk or mean
reverting process has important implications for investors in stock markets. Stock market
efficiency implies that prices respond quickly and accurately to relevant information. In-
formation in the efficient market hypothesis (EMH) is defined as anything that may affect
prices, which is unknowable in the present and appears randomly in the future. This
random information is the cause of future price changes. In other words, an efficient stock
market is characterized by a random walk (unit root) process, which indicates that stock
market returns cannot be predicted based on its historical observations (Malkiel, 2003). If
stock price follows a random walk process, any shock to stock price is permanent, and
there is no tendency for the price level to return to a trend path over time. The alternative to
a random walk is that stock prices are mean reverting (trend stationary). If stock prices are
mean reverting, it follows that the price level will return to its trend path over time, making
* Corresponding author.
1250021-1
The Singapore Economic Review
it possible to forecast future movements in stock prices based on past behavior. Thus, there
is potential for investors to get an advantage through utilizing predictability in stock
returns. Hence, testing for mean reversion in stock prices is one avenue for examining
market efficiency (Fama and French, 1988a, 1988b).
The purpose of this study is to investigate the stock prices behavior of five ASEAN
(Association of Southeast Asian Nations) countries i.e., Indonesia, Malaysia, Philippines,
Singapore and Thailand for the period from 1990:1 to 2009:1, to verify the validity of
EMH in the stock exchanges of those countries. There is a large body of the literature that
investigates the existence of the EMH in different countries by using a variety of metho-
dologies, though mixed results were found. On one hand, we believe that there is a need to
provide further insights of the EMH based on country uniqueness. In this study, the
selected countries are five small open economies of ASEAN, and among those ASEAN
countries that have outperformed equity markets during the 2007 financial crisis. For the
period from October 2008 to October 2009, the stock market indices of the JSE Exchange
of Indonesia had increased 57%; the SET Exchange of Thailand had increased 38.7%; the
SGX Exchange of Singapore had increased 33.2%; the PSE Exchange of Philippines had
increased 33.1%; and the KLSE Exchange of Malaysia had increased 30.3% (Bhaskaran
and Ghosh, 2010). It is true that an efficient stock market is likely to preserve confidence
among market players and to gain momentum for sustainability and resilience in facing
crises.
On the other hand, the exploration on the method used is another need to study the
EMH. This study attempts to provide new evidence on the selected five ASEAN countries
by using a two-regime threshold autoregressive (TAR) approach, which allows for
simultaneous testing of non-linearity and non-stationarity.
2. Literature Review
There is a large body of the literature that investigates the EMH using a variety of
methodologies, with mixed results. Many studies have found that stock indexes are not
characterized by a unit root (Lo and MacKinlay, 1988; Poterba and Summers, 1988; Urrutia,
1995; Grieb and Reyes, 1999; Chaudhuri and Wu, 2003; Shively, 2003; Narayan, 2008),
while others have found stock indexes to be a unit root process (Huber, 1997; Liu, Song and
Romilley, 1997; Narayan and Smyth, 2004, 2005; Narayan, 2005, 2006; Ozdemir, 2008;
Qian, Song and Zhou, 2008). Two important features characterize these studies.
First, the majority of these studies are based on univariate unit root tests. However, one
strong criticism of the univariate unit root tests, such as the Dickey and Fuller test used by
most studies, is that it lacks power if the true data generating process of a series exhibits
structural breaks (Perron, 1989). Therefore, the majority of these studies adopt new de-
veloped unit root test with structural breaks (Zivot and Andrews, 1992; Lumsdaine and
Papell, 1997; Lee and Strazicich, 2003; Im, Lee and Tieslau, 2005) to investigate the
stationary property of stock prices. For example, Chaudhuri and Wu (2003) investigate
mean reversion in stock prices in emerging markets, including one break unit root tests.
Their findings, when compared to previous findings, show that there is no consensus
1250021-2
The Efficient Market Hypothesis Revisited
among economists regarding market efficiency. Narayan and Smyth (2004) apply the Zivot
and Andrews (1992) one break and the Lumsdaine and Papell (1997) two break unit root
tests to examine the random walk hypothesis for stock prices in South Korea. Their results
provide strong evidence that stock prices in South Korea are characterized by a unit root,
which is consistent with the EMH. Lean and Smyth (2007) apply univariate and panel
Lagrange Multiplier (LM) unit root tests with one and two structural breaks (Lee and
Strazicich, 2003; Im, Lee and Tieslau, 2005) to examine the random walk hypothesis for
stock prices in eight Asian countries. The results from the univariate LM unit root tests and
panel LM unit root test with one structural break suggest that stock prices in each country
is characterized by a random walk, but the findings from the panel LM unit root test with
two structural breaks suggest that stock prices in the eight countries are mean reverting.
Narayan (2008) provide evidence on the unit root hypothesis for G7 stock price indices
using the LM panel unit root test that allows for structural breaks. His main finding is that
stock prices are stationary processes, inconsistent with the EMH. Narayan and Smyth
(2005) applied the sequential trend break test besides the panel data unit root test to
estimate the data of 22 OECD countries, and Narayan and Narayan (2007) used five panel
unit root tests to estimate the G7 countries data. Both studies proved the absence of the
mean reverting of stock prices, and advocated the random walk process. However, Narayan
and Smyth (2007) provide mixed evidence on the random walk hypothesis employing
monthly data on G7 stock price indices over the period 1960–2003, using unit root tests
with one and two structural breaks in the trend. They find stock prices are stationary in
Japan while the stock price indices in the other countries follow a random walk. Another
efficient hypothesis related study (Narayan, Narayan and Zheng, 2010) tested on the long-
run relationship between gold and oil spot and futures markets, and proven that the two
commodity markets are inefficient as predictions might set in by investors. Inflationary
pressure may follow a rise in oil price and investors tend to use gold as an inflationary–
hedging commodity.
Second, however, following the works of Abhyankar, Copeland and Wong, (1995,
1997); Atchison and White (1996); Kohers, Pandey and Kohers (1997); Schaller and van
Norden (1997); Qi (1999); Kanas (2001); Sarantis (2001); Shively (2003); Narayan (2005);
Qian, Song and Zhou (2008), among others, who find stock prices to be consistent with a
non-linear data generating process, the reliability of the findings from existing studies is
questionable. Shively (2003) examines the six stock prices (CAC 40, DAX 30, FTSE 100,
Nikkei 225, S&P 500 and TSE 300) for the period 1970:1–2000:12. He applies Tsay’s
(1998) chi-squared test and find that all six stock-price indexes are all highly consistent
with threshold non-linearity. Then he applies Tsay’s (1998) threshold modeling technique
to partition each stock-price index into three regimes using the corresponding stock-return
series as the stationary threshold variable and finds the series to be a regime-reverting
process. This non-linear regime-reverting process implies a violation of the EMH. In
contrast, Narayan (2006) investigates the behavior of US stock prices using an unrestricted
two-regime threshold model for the period 1964:06 to 2003:04. He finds that the stock
prices are a non-linear process and characterized by a unit root process, consistent with the
EMH.
1250021-3
Lean and Smyth (2007) suggest that, in terms of future research, there is growing
evidence that univariate unit root tests lack the power to find mean reversion in stock
prices. Perhaps a more promising approach might be to examine whether ASEAN-5 stock
prices are non-linear with a unit root. Thus, this paper contributes to the existing literature
on the random walk hypothesis, by providing additional evidence on the selected five
ASEAN countries, namely, Indonesia, Malaysia, the Philippines, Singapore and Thailand,
by using a two-regime TAR approach developed by Caner and Hansen (2001). The main
reason for the selection of these five countries is that the stock markets in these countries
are rapidly developing and more well established than other remaining ASEAN countries,
such as Vietnam, Cambodia, Laos, Myanmar and Brunei. Reliable and sufficient data in
these countries’ stock market is lacking. There is strong evidence of an increase in the level
integration and interdependence between these developing ASEAN markets after the 1997
financial crisis (Lim, 2009).1 Furthermore, the basic motivation to analyze the EMH in
these countries is that, despite its increasing importance, there is still a lack of systematical
analysis on the ASEAN stock market. This paper employs the TAR method, suggested by
Caner and Hansen (2001), to examine the EMH and aims to add some new empirical
evidence to the existing literature which are dominated by the research in developed
countries’ stock markets.
The Caner and Hansen (2001) methodology is applicable if a non-linear process has unit
root. The main advantage of the TAR model is that it allows us to discriminate non-
stationarity from non-linearity in data simultaneously. Furthermore, their methodology
allows testing for a partial unit root process in two regimes. Due to its novelty, many
studies used TAR model, for instance, Alba and Park (2005), Basci and Caner (2005) and
Ho (2005) among others, have applied the Caner and Hansen methodology to exchange
rates and Purchasing Power Parity (PPP). Basci and Caner (2005) suggest that this
methodology helps us to disentangle the non-linearity from the non-stationarity rigorously
for the first time in the literature.
The rest of the paper is organized as follows: Section 2 outlines the empirical meth-
odology. Section 3 presents the data and empirical results. Finally, Section 4 provides a
conclusion.
3. Empirical Methodology
Following the work of Caner and Hansen (2001), we adopt a two-regime TAR (k) model
with an autoregressive unit root as follows:
Δyt ¼ θ 01 xt1 I{Zt1<λ g þ θ 02 xt1 I{Zt1 ‚ λ g þ et , ð1Þ
where y is the logarithm of the stock price index for t ¼ 1,…, T, xt1 ¼ ðyt1 , r t0 ,
Δyt1 ,…, Δytk Þ 0 ; I{g is the indicator function; et is an independent and identically
distributed (i.i.d.) error term; Zt ¼ yt ytm for m represents the delay order and some
1 In the Finance Thailand 2010 Conference, a proposal for the integration of ASEAN capital markets has been put forth,
which is translated as part of the Pan-ASEAN Vision (Phuvanatnaranubala, 2010).
1250021-4
1 • m • k. rt is a vector of deterministic components including an intercept and a possible

linear time trend. The threshold value λ is unknown and takes the values in the compact
interval λ 2 Λ ¼ ½λ1 , λ2 , where λ1 and λ2 are picked according to PðZt • λ1 Þ ¼ 1 > 0 and
PðZt • λ2 Þ ¼ 2 < 1. It is convenient to show the components of θ1 and θ2 as follows:
0 1 0 1
1 2
@
θ1 ¼ β1 A and θ2 ¼ β2 A,
@ ð2Þ
α1 α2
where 1 and 2 are slope coefficients on yt1 , β1 and β2 are scalar intercepts, and α1 and
α2 are K 1 vectors containing the slope coefficients on dynamic regressors
( Δyt1 ,…, Δytk ) in the two regimes. In order to calibrate Equation (1), the concentrated
least squares (LS) approach is usually utilized. For each λ 2 Λ, Equation (1) is as follows:
Δyt ¼ θ^1 ðλÞ 0 xt1 I{Zt1<λ g þ θ^2 ðλÞ 0 xt1 I{Zt1 ‚ λ g þ e^t : ð3Þ
P
Let ^ 2 ðλÞ ¼ T 1 T1 e^t ðλÞ 2 be the OLS estimate of 2 for fixed λ. The LS estimate of
threshold parameter (λ) is found by minimizing the residual variance, 2 ðλ):
^λ ¼ arg min ^ 2 ðλÞ: ð4Þ
λ2Λ
Estimating the TAR model in Equation (1), the two central issues are whether or not there
is a threshold effect and whether the process yt (stock price index) is stationary or not. In
this paper, standard Wald test statistics, WT ¼ WT ð^λÞ ¼ sup WT ðλÞ, proposed by Caner and
λ2Λ
Hansen (2001), is used to test the null hypothesis of no threshold effect (i.e., the process is
linear) H0 : θ1 ¼ θ2 , against the alternative of threshold effect (i.e., the process is non-
linear). If the null hypothesis cannot be rejected, there is no threshold effect, in which case
the two vectors of coefficients are identical between the two regimes (θ1 ¼ θ2 Þ. Caner and
Hansen find that WT has a non-standard asymptotic null distribution with critical values
that cannot be tabulated. Hence they propose a bootstrap method to compute asymptotic
critical values and p-values.
The stationarity of the process yt depends on the parameters 1 and 2 . For regime one,
we can reject the null hypothesis of unit roots in favor of the alternative hypothesis of level
stationarity if 1 is significantly different from zero. We can do the same for regime two if
2 is significantly different from zero. If the null hypothesis: H0 : 1 ¼ 2 ¼ 0 holds, the
process yt has a unit root and model (1) can be expressed in terms of the stationary
difference Δyt . The obvious alternative to H0 is H1 : 1 < 0 and 2 < 0, in which case the
process yt is stationary in both regimes. We also have to consider the intermediate partial
unit root case H2 : 1 < 0 and 2 ¼ 0 or 1 ¼ 0 and 2 < 0, in which case the process yt
have a unit root in one regime and is stationary in other showing mean reversion behavior.
The null hypothesis is tested against the unrestricted alternative 1 6¼ 0 or 2 6¼ 0 using
the Wald statistics and expressed as R2T ¼ t 12 þ t 22 , where t1 and t2 are the t-ratios for ^1
and ^2 , respectively, from the OLS estimation. However, Caner and Hansen (2001) note
that this two-sided Wald statistics may have less power than a one-sided version of the test.
1250021-5
As a result, they recommend the following one-sided Wald statistics:

R1T ¼ t 12 I{^1 <0g þ t 22 I{^2 <0g , ð5Þ
which tests H0 against the one-sided alternative 1 < 0 or 2 < 0. A statistically significant
R1T justifies rejecting unit roots in favor of stationarity. However, it does not allow us to
discriminate between the stationary case H1 and the partial unit root case H2 . This requires
further examining the individual t-statistics t1 and t2 . Only one of t1 or t2 being
significant would be consistent with the partial unit root case.
4. Data and Empirical Results

4.1. Data
We use the stock price indices for five Asian countries, i.e., Indonesia [Jakarta Composite
Index (JKSE)], Malaysia [Kuala Lumpur Composite Index (KLCI)], Philippines [Phi-
lippine Composite Index (PCE)], Singapore [Straits Times Index (STI)] and Thailand
[Stock Exchange of Thailand Index (SET)] in our empirical study. Monthly averaged data
over the period from 1990:1 to 2009:1 are utilized for analysis and taken from Asian
Development Bank (ADB), Asia Regional Integration Centre (ARIC) database (www://
aric.adb.org/aric_database), which gives the time series yt defined in the preceding section.
Log transformation for stock prices are used throughout the study.
4.2. Conventional univariate unit root test

We first use the traditional univariate unit root tests such as those in Dickey and Fuller
(ADF, 1979) and Phillips and Perron (PP, 1988). Table 1 summarizes the results based on
conventional unit root tests. We include an intercept in the model and use the Schwartz
Information Criterion to select the optimal lag length for ADF test. The bandwidth is
selected automatically using the Newey and West (1994) method for PP and KPSS tests.
Our main finding from ADF and PP (which tests the null hypothesis of non-stationarity)
can be summarized as follows. Based on the ADF tests, we find the calculated t-statistics to
Table 1. ADF and PP Unit Root Tests
ADF (t-stat) LL PP (t-stat) Bandwidth
Indonesia 1.011 3 0.972 3

Malaysia 2.544 1 2.087 1
Philippines 2.178 1 1.943 3
Singapore 2.051 0 2.115 1
Thailand 1.615 7 1.598 3
Notes: LL denotes the lag length. The finite sample critical values
for the ADF and PP test are 2.573 and 2.874 at the 10% and
5% levels, respectively, and are extracted from MacKinnon
(1996).
1250021-6
be 1.01, 2.54, 2.17, 2.05 and 1.61 for Indonesia, Malaysia, Philippines, Singa-
pore and Thailand stock prices respectively. Given the 10% level critical value of 2.57,
we are unable to reject the unit root null hypothesis for all countries. Further, according to
the PP test, we find the calculated t-statistics to be 0.97, 2.08, 1.94, 2.11 and 1.59
for Indonesia, Malaysia, Philippines, Singapore and Thailand stock prices respectively. As
like ADF test, we are unable to reject the unit root null hypothesis for all countries. This
implies that all of the countries have a unit root in stock indices according to ADF and PP
and suggests that all the countries’ stock markets are efficient. This finding is not sur-
prising, since ADF and PP tests have almost no power when the alternative is a non-linear
process. In order to circumvent this problem, we next apply the non-linear threshold unit
root test as in Caner and Hansen (2001).
4.3. Non-linear threshold unit root test

To examine the stationarity in the possible presence of non-linearities, we apply the Caner
and Hansen (2001) procedure described earlier. The first issue we must address is the
presence of the threshold effects, and hence non-linearity. In other words, we try to find
whether the stock indices are non-linear or not. As stated previously, the appropriate test
this purpose is the standard Wald statistic WT . In Table 2, we report the Wald test, boot-
strap critical values at 1% conventional level, bootstrap p-values (using 10,000 replica-
tions) for threshold variables of the form Zt ¼ yt ytm and the corresponding optimal
delay parameter, m, along with the percentage of observations that falls into each regime.
Since the data is at monthly intervals, we set the maximum lag equal to 12. Caner and
Hansen (2001) recommended making m endogenous, which is achieved by selecting an m
value that minimizes the residual variance of the LS estimates. This is also the value that
maximizes WT since WT is a monotonic function of the residual variance. In Table 2, we
find very strong evidence for threshold effect, hence non-linearity of stock indices, at 1%
level. The significant bootstrap p-values corresponding to the Wald tests WT indicate that
we can reject the null hypothesis of linearity in favor of the alternative that there is a
threshold effect in the monthly stock prices of each country. To provide an example in the
Table 2. The Wald Tests for a Threshold-WT
Countries Wald Bootstrap Bootstrap Optimal Threshold Number

Statistics Critical p-Value Delay Parameter of Observations
(WT ) Values (1%) Parameter m (^λ) (6)
(1) (2) (3) (4) (5) Regime One Regime Two
Indonesia 59.7 50.0 0.000 11 0.268 40 (19%) 176 (81%)

Malaysia 74.9 54.1 0.000 10 0.187 31 (14%) 185 (86%)
Philippines 76.2 50.5 0.000 11 0.208 40 (19%) 176 (81%)
Singapore 74.7 52.1 0.000 8 0.190 35 (16%) 181 (84%)
Thailand 57.8 50.5 0.000 6 0.109 57 (26%) 159 (74%)
Note: Threshold parameter is given in absolute terms.
1250021-7
case of Thailand, the Wald statistics is maximized (WT ¼ 57:8, corresponding p-value ¼
0.000) when m ^ ¼ 6. Hence we take m ^ ¼ 6 as the preferred model for Thailand. For
a second example, consider Singapore. The Wald statistics is WT ¼ 74:7 (corresponding
p-value ¼ 0.000) when optimal delay parameter m ¼ 8. Hence m ^ ¼ 8 is the preferred model
for Singapore. Second, in Table 2, we also report the threshold parameters at the corre-
sponding optimal delay parameter and the percentage of observations in each regime. For
example, for Thailand, the optimal delay parameter is m ¼ 6 and the threshold parameter ^λ is
0.109. this means that in the first regime, Thailand composite stock price (SET) index has
either decreased, remained constant or increased by less than 10.9% (because the model is
specified in logs, the threshold parameter can be interpreted as the percentage change) within
six months. For Thailand 57 observations or 26% of the sample belong to first regime (inside
the band). In the second regime (outside the band of 0.109) SET index increases by more
than or equal to 10.9% within six months. Approximately 74% of the sample belongs to the
second regime. When we analyze Table 2, we see that the second regime is dominant in all of
the countries. Specifically, the share of the “outside the band” regime fluctuates between
74% and 86% of the observations in the countries analyzed. Our findings thus lend very
strong support for a TAR model and the presence of threshold effects.
Next, having established that stock price index is a non-linear process for each of the
countries, we examine whether stock market indices contains a unit root. To examine the unit
root properties, we first calculate threshold unit root test statistic for the one-sided R1T ,
which tests unit root against a two-regime stationary non-linear mode. Caner and Hansen
(2001) suggest that one-sided Wald test has superior properties to the two-sided test. We do
so for the optimal value of the delay parameter for each country. The R1T test statistic,
bootstrap critical values (at 1% significance level) and the bootstrap p-values, are reported in
Table 3. At the optimal value of the delay parameter for each country, we are able to reject
the null of a unit root at 5% level for Indonesia and at 1% level for Philippines and Singapore.
For Malaysia and Thailand, we were unable to reject the unit root null hypothesis. For
example, the Wald statistic WT obtained from R1T is statistically insignificant at the 5% level
for Malaysia and Thailand. For Malaysia, the WT test statistics of 9.40 is less than the 5%
critical value (13.8). For Thailand, The WT test statistics of 2.89 is less than the 5% critical
value (12.0). As a result, the null hypothesis of a unit root for Malaysia’s and Thailand’s
stock price series cannot be rejected. The Wald statistic WT obtained from R1T is statistically
significant at the 5% level for Indonesia, Philippines and Singapore. To give a specific
example, in the case of Singapore, the WT test statistics of the remaining countries are greater
than the 5% critical values. As a result, we are able to reject the null of a unit root at 5% level
for Indonesia and at 1% level for Philippines and Singapore.
Further, we test for a partial unit root in regime one and two. For those countries for
which the R1T test rejects the null of a unit root, it is possible there might still be a partial
unit root in either the first or second regime. The results for the t1 test (for a partial unit root
in regime one) and t2 test (for a partial unit root in regime two), together with the corre-
sponding p-values, are reported in Table 3. For countries for which the R1T test rejected the
unit root null, the results in Table 3 indicate that there is a partial unit root in the first
regime for Indonesia at the 5% level, for Philippines and Singapore at the 1% level.
1250021-8
Table 3. One-Sided and Partial Unit Root Tests
Countries R1T Bootstrap Critical Bootstrap t1 Statistic Bootstrap t2 Statistic Bootstrap

Statistic Values (5%) p-Value p-Value p-Value
Indonesia 12.7 12.5 0.049 3.46 0.017 0.29 0.749

Malaysia 9.40 13.8 0.153 2.21 0.178 2.11 0.221
Philippines 36.4 12.2 0.000 5.90 0.000 1.15 0.504
Singapore 18.6 12.2 0.007 4.15 0.003 1.18 0.468
Thailand 2.89 12.0 0.606 1.46 0.360 0.87 0.568
Furthermore, for Indonesia, Philippines and Singapore, the bootstrap p-values for the in-
dividual t ratios t1 and t2 are 0.017, 0.000, 0.003 in first regime and in the second regime,
0.749, 0.504, 0.468, respectively, which implies that we can reject the unit root hypothesis in
favor of 1 < 0 but we are unable to reject 2 ¼ 0. Moreover, for Malaysia and Thailand, the
bootstrap p-values for the individual t ratios t1 and t2 are not statistically significant in each
regime as indicated in Table 3. As a result, we are unable to reject the unit root null
hypothesis, implying that, unlike the stock prices of Indonesia, Philippines and Singapore,
Malaysia’s and Thailand’s stock prices are non-stationary processes.
Taken as whole, we find at best mixed evidence that stock prices are stationary in
Indonesia, Philippines and Singapore while the stock price indices in the Malaysia and
Thailand follow a random walk. In this regard, Chaudhuri and Wu (2003) examined the
random walk hypothesis for stock prices in 18 emerging markets using monthly data from
1985 to 1997. They find that the Zivot and Andrews (1992) test rejects the random walk
null hypothesis for 10 of the 18 markets which they studied. Narayan and Smyth (2007),
also provides the mixed evidence on the random walk hypothesis in G7 stock price indices
using unit root tests which allow for one and two structural breaks in the trend. Of the
seven countries they find, at best, evidence of mean reversion in the stock price index of
Japan. Thus, overall, their results support the random walk hypothesis. Worth noting is that
the results here are not consistent with those of Shively (2003) who examined the six stock
prices (CAC 40, DAX 30, FTSE 100, Nikkei 225, S&P 500 and TSE 300) for the period
1970:1–2000:12. He finds the series to be a regime-reverting process. This non-linear
regime-reverting process implies a violation of the EMH. Lean and Smyth (2007) exam-
ined the random walk hypothesis for stock prices in eight Asian countries. The results from
panel LM unit root test with two structural breaks suggest that stock prices in the eight
countries are mean reverting, suggesting the acceptance of EMH.
5. Conclusions
An important area of research in the financial economics literature has centered on the
issue of mean reversion in stock prices. Despite the number of studies dating back to the
1960s, the extant literature has not reached a consensus on whether or not stock prices
follow a unit root process. This information is crucial for investors, for if stock prices can
be characterized as a unit root process, then it implies that shocks to prices have a
1250021-9
permanent effect, in that stock prices will attain a new equilibrium and future returns
cannot be predicted based on historical movements in stock prices. On the other hand, if
stock prices are mean reverting, then shocks to prices will have a temporary effect, en-
suring that one can forecast future movements in stock prices based on past behavior and
trading strategies can be developed so as to earn abnormal returns.
In this paper, we have investigated EMH in five Asian countries’ stock price indices by
applying the two-regime TAR approach on monthly data over the period 1990:1 to 2009:1.
Three important results derive from our empirical analysis. First, we find that stock prices
of five ASEAN countries are non-linear series. Second, there is evidence for partial unit
root regime (for Indonesia, Philippines and Singapore), which consists of one stationary
and one non-stationary regime. Third, we apply a suite of tests for unit roots and find
Malaysia’s and Thailand’s stock price series are characterized by a unit root process. In
other words, KLCI and SET series are characterized by a random walk (unit root) process,
consistent with the EMH. This implies that returns on the Malaysia and Thailand stock
markets cannot be predicted using its own history of stock prices. In contrast to Malaysia
and Thailand stock prices, we find evidence of mean reversion (stationary) for the Indo-
nesia, Philippines and Singapore stock prices, rejecting the EMH. As suggested by
Ozdemir (2008), in a developing country, the financial system is not well established;
second, the commercial banks have limited ability to borrow in their own currency; lastly,
there may be frequent economic and political instability. Hence, we cannot expect the stock
markets to be continuously efficient. The stock market of such an economy can go from an
inefficient to efficient state, and vice-versa, depending on the developments in these three
conditions.
Acknowledgments
Helpful comments and suggestions have been received from the two anonymous referees
of this journal on the earlier version of this paper. The usual disclaimer applies.
References
Abhyankar, A, LS Copeland and W Wong (1995). Nonlinear dynamics in real-time equity market
indices: Evidence from the United Kingdom. Economic Journal, 105, 864–880.
Abhyankar, A, LS Copeland and W Wong (1997). Uncovering non-linear structure in real-time
stock market indexes: The S&P 500, the DAX, the Nikkei 225 and the FTSE-100. Journal of
Business and Economics Statistics, 15, 1–14.
Alba, JD and D Park (2005). An empirical investigation of purchasing power parity (PPP) for
Turkey. Journal of Policy Modeling, 27, 989–1000.
Atchison, MD and MA White (1996). Disappearing evidence of chaos in security returns: A
simulation. Quarterly Journal of Business and Economics, 35, 21–37.
Basci, E and M Caner (2005). Are real exchange rates non-linear or non-stationary? Evidence from
a new threshold unit root test. Studies in Nonlinear Dynamics and Econometrics, 9, 1–19.
Bhaskaran, M and R Ghosh (2010). Global economic and financial crisis impact on developing
Asia and immediate policy implications, prepared for ADB (Asian Development Bank).
1250021-10
Regional Forum on the Impact of Global Economic and Financial Crisis, Manila, 14–15 January
2010.
Caner, M and B Hansen (2001). Threshold autoregression with a unit root. Econometrica, 69,
1555–1596.
Chaudhuri, K and Y Wu (2003). Random walk versus breaking trend in stock prices: Evidence from
emerging markets. Journal of Banking and Finance, 27, 575–592.
Dickey, DA and WA Fuller (1979). Distribution of the estimators for autoregressive time series with
a unit root. Journal of the American Statistical Association, 74, 427–431.
Fama, EF and KR French (1988a). Dividend yields and expected stock returns. Journal of Financial
Economics, 22, 3–25.
Fama, EF and KR French (1988b). Permanent and temporary components of stock prices. Journal
of Political Economy, 96, 246–273.
Grieb, TA and MG Reyes (1999). Random walk tests for Latin American equity indexes and
individual firms. Journal of Financial Research, 22, 371–383.

Ho, T (2005). Investigating the threshold effects of inflation on PPP. Economic Modelling, 22,
926–948.
Huber, P (1997). Stock market returns in thin markets: Evidence from the Vienna stock exchange.
Applied Financial Economics, 7, 493–498.
Im, KS, J Lee and M Tieslau (2005). Panel LM unit root tests with level shifts. Oxford Bulletin of
Economics and Statistics, 67, 393–419.
Kanas, A (2001). Neural network linear forecasts for stock returns. International Journal of Finance
and Economics, 6, 245–254.
Kohers, T, V Pandey and G Kohers (1997). Using nonlinear dynamics to test for market efficiency
among the major US stock exchanges. Quarterly Review of Economics and Finance, 37,
523–545.
Lean, HH and R Smyth (2007). Do Asian stock markets follow a random walk? Evidence from LM
unit root tests with one and two structural breaks. Review of Pacific Basin Financial Markets and
Policies, 10, 15–31.
Lee, J and MC Strazicich (2003). Minimum LM unit root tests with two structural breaks. The
Review of Economics and Statistics, 85, 1082–1089.
Lim, LK (2009). Convergence and interdependence between ASEAN-5 stock markets. Mathe-
matics and Computers in Simulation, 79, 2957–2966.
Lo, AW and AC MacKinlay (1988). Stock market prices do not follow random walks: Evidence
from a simple specification test. Review of Financial Studies, 1, 41–66.
Liu, CY, SH Song and P Romilley (1997). Are Chinese stock markets efficient? A cointegration and
causality analysis. Applied Economic Letters, 4, 511–515.
Lumsdaine, R and D Papell (1997). Multiple trend breaks and the unit-root hypothesis. Review of
Economics and Statistics, 79, 212–218.
MacKinnon, JG (1996). Numerical distribution functions for unit and cointegration tests. Journal of
Applied Econometrics, 11, 601–618.
Malkiel, BG (2003). The efficient market hypothesis and its critics. Journal of Economic
Perspectives, 17, 59–82.
Narayan, PK (2005). Are the Australian and New Zealand stock prices nonlinear with a unit root?
Applied Economics, 37, 2161–2166.
Narayan, PK (2006). The behavior of US stock prices: Evidence from a threshold autoregressive
model. Mathematics and Computers in Simulation, 71, 103–108.
Narayan, PK (2008). Do shocks to G7 stock prices have a permanent effect? Evidence from panel
unit root tests with structural change. Mathematics and Computers in Simulation, 77, 369–373.
Narayan, PK and S Narayan (2007). Mean reversion in stock prices: New evidence from panel unit
root test. Studies in Economics and Finance, 24, 233–244.
1250021-11
Narayan, PK and R Smyth (2004). Is South Korea’s stock market efficient? Applied Economics
Letters, 11, 707–710.
Narayan, PK and R Smyth (2005). Are OECD stock prices characterized by a random walk?
Evidence from sequential trend break and panel data models. Applied Financial Economics, 15,
547–556.
Narayan, PK and R Smyth (2007). Mean reversion versus random walk in G7 stock prices evidence
from multiple trend break unit root tests. Journal of International Financial Markets, Institutions
and Money, 17, 152–166.
Narayan, PK, S Narayan and X Zheng (2010). Gold and oil futures markets: Are markets efficient?
Applied Energy, 87, 3299–3303.
Newey, W and K West (1994). Automatic lag selection in covariance matrix estimation. Effects of
Economics Studies, 61, 631–653.
Ozdemir, ZA (2008). Efficient market hypothesis: Evidence from a small open-economy. Applied
Economics, 40, 633–641.

Perron, P (1989). The Great Crash, the oil price shock, and the unit root hypothesis. Econometrica,
57, 1361–1401.
Phillips, PCB and P Perron (1988). Testing for a unit root in time series regression. Biometrika, 75,
335–346.
Phuvanatnaranubala, T (2010). The pan-ASEAN Vision: Driving the integration of ASEAN capital
markets, Speech by the Secretary-General, Securities and Exchange Commission, Chairman,
ASEAN Capital Market Forum (ACMF), at the Finance Thailand 2010 Conference, Bangkok,
Thailand, 22 February.
Poterba, JM and LH Summers (1988). Mean reversion in stock prices: Evidence and implications.
Journal of Financial Economics, 22, 27–59.
Sarantis, N (2001). Nonlinearities, cyclical behavior and predictability in stock markets: Interna-
tional evidence. International Journal of Forecasting, 17, 459–482.
Schaller, H and van S Norden (1997). Regime switching in stock market returns. Applied Financial
Economics, 7, 177–191.
Shively, PA (2003). The nonlinear dynamics of stock prices. The Quarterly Review of Economics
and Finance, 43, 505–517.
Qi, M (1999). Nonlinear predictability of stock returns using financial and economic variables.
Journal of Business and Economic Statistics, 17, 419–429.
Qian, X-Y, F-T Song and W-X Zhou (2008). Nonlinear behavior of the Chinese SSEC index with a
unit root: Evidence from threshold unit root tests. Physica A, 387, 503–510.
Tsay, R (1998). Testing and modeling multivariate threshold models. Journal of the American
Statistical Association, 93, 1188–1202.
Urrutia, JL (1995). Test of random walk and market efficiency for Latin American emerging equity
markets. Journal of Financial Research, 18, 299–309.
Zivot, E and DWK Andrews (1992). Further evidence on the Great Crash, the oil-price shock and
the unit root hypothesis. Journal of Business and Economic Statistics, 10, 251–270.
1250021-12

Control 1 Finanzas

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Control 1 Finanzas

Uploaded by

Copyright:

Available Formats

The Singapore Economic Review, Vol. 57, No.

3 (2012) 1250021 (12 pages)

THE EFFICIENT MARKET HYPOTHESIS REVISITED: EVIDENCE

QAISER MUNIR*, KOK SOOK CHING, FUMITAKA FUROUKA

Universiti Malaysia Sabah, Locked Bag 2073

Published 18 September 2012

Keywords: Efficient market hypothesis; threshold autoregressive model; unit root.

JEL Classification: C22, C52

1 • m • k. rt is a vector of deterministic components including an intercept and a possible

As a result, they recommend the following one-sided Wald statistics:

4. Data and Empirical Results

4.2. Conventional univariate unit root test

Table 1. ADF and PP Unit Root Tests

ADF (t-stat) LL PP (t-stat) Bandwidth

Indonesia 1.011 3 0.972 3

4.3. Non-linear threshold unit root test

Table 2. The Wald Tests for a Threshold-WT

Countries Wald Bootstrap Bootstrap Optimal Threshold Number

Indonesia 59.7 50.0 0.000 11 0.268 40 (19%) 176 (81%)

Note: Threshold parameter is given in absolute terms.

Table 3. One-Sided and Partial Unit Root Tests

Countries R1T Bootstrap Critical Bootstrap t1 Statistic Bootstrap t2 Statistic Bootstrap

Indonesia 12.7 12.5 0.049 3.46 0.017 0.29 0.749

individual firms. Journal of Financial Research, 22, 371–383.

Economics, 40, 633–641.

You might also like