Weak-Form Efficiency Testimony of Dhaka Stock Exchange
Weak-Form Efficiency Testimony of Dhaka Stock Exchange
Weak-Form Efficiency Testimony of Dhaka Stock Exchange
Shofiqur Rahman*
Mohammad Farhad Hossain**
Abstract:
The focus of this paper is to seek evidence whether Dhaka Stock Exchange
(DSE) is efficient in the weak form or not by hypothesizing normality of
the distribution series and random walk assumption. Both non-parametric
tests [(Kolmogrov-Smirnov goodness of fit test), run test, Lilliefors test],
Q-Q probability plots and parametric-tests [Auto-correlation coefficient
test and ARIMA (0,1,0) for testing random walk model] have been used.
Each statistical test has been performed separately on two data sets. The
data sets include daily All-Share Price Indices (ASPI) and DSE General
Price Indices (DSE-GEN) for 12 years ranging from 1994 to 2005. In
addition, daily stock price data for 33 companies have been used. The daily
return series, in the aspect of skewness and kurtosis, were found nonnormal, which can be categorized as positively skewed distribution. Same
thing resulted from Kolmogorov-Smirnov (K-S) test. As a result, null
hypothesis of normality has been rejected and alternative hypothesis
remained in effect. Run test and auto-correlation results rejected the
randomness of the return series of DSE simultaneously. Overall results
from the empirical analysis suggest that the Dhaka Stock Market of
Bangladesh is not efficient in weak-form.
INTRODUCTION
Capital market efficiency has been a debatable issue since its inception. After Bachelier
(1900), several researchers tried to reach a solid conclusion on it but could not. Because,
efficiency, on a large extent depends on investment patterns, structure of market,
investors risk appetite and practice etc. Unlike past security prices, all other factors make
a capital market more vibrant. As a result, market efficiency has still become within the
continuum of hypothesis. The concept of efficiency is extremely important in finance
because the hypothesis that securities markets are efficient represents the basis for most
research that is made in financial economics. For managers, efficient market implies that
they can receive a fair value for the securities they sell and they can implement their
goal of maximization of shareholders wealth by focusing on the effect each decision will
*
**
1
Electronic copy of this paper is available at: http://ssrn.com/abstract=940811
have on the share price. Another implication of efficient markets is that creative
accounting will not result in price changes because accounting manipulations do not affect
cash flows, thus they are not reflected in prices. EMH also has strong implications for
security analysts. When future returns cannot be predicted from past returns, trading based
on an examination of the sequence of past prices becomes worthless. If the semi-strong
form of the hypothesis is supported, then financial analysts cannot devise trading rules
based on publicly available information. Most of the researches on the capital market
efficiency are concentrated on the major stock markets of the world (Summers (1986),
Fama and French (1988), Lo and MacKinlay (1988) and Poterba and Summers (1988)).
The focus of this study is to testify the efficiency of Dhaka Stock Exchange. The purpose
of this research is to seek evidence whether Dhaka Stock Exchange (DSE) is efficient in
the weak form or not. Before starting the discussion of the study finding, a discussion on
previous empirical studies would be
Mt-1 = the information set used by the market at t 1, at-1 = the specific
information item placed in the public domain at t 1. Defining market efficiency in this
way has two important implications, such as (1) an investor cannot use at-1 to earn
nonzero abnormal return, and (2) in an efficient market, when a new information item is
added to the information set, M, its revaluation implications for ( Ri,t, Rj,t,
2
Electronic copy of this paper is available at: http://ssrn.com/abstract=940811
) are
The concept of market efficiency was originally anticipated by Bachelier (1900). There
were subsequent works such as Working (1934), Cowles and Jones (1937), Kendall
(1953). But Bacheliers contribution receives more highlights after it was published in
English by Cootner (1964). Fama (1965) was, however, the first to use the term efficient
market. Fama extended the process of formalizing the concept of efficiency in economic
terms by defining an efficient market as one in which prices always fully reflects available
information. Fama (1970) stated that the sufficient but not necessary conditions for
efficiency are: (i) there are no transaction costs in trading securities; (ii) all available
information is costlessly available to all market participants, and (iii) all agree on the
implications of current information for the current price and distributions of future prices
of each security. In his review of literature, Fama (1970) divided his work on efficient
market into three categories depending on the types of information and how quickly this
information is impounded in prices: (1) weak-form EMH, (2) semi-strong EMH, and (3)
strong-form EMH. This classification was the first recognition that efficiency must be
defined with respect to a particular information set. Weak-form efficiency exists if security
prices fully reflect all the information contained in the history of past prices and returns. If
capital markets are weak-form efficient, then investors can earn excess profits from
trading rules based on past prices or returns. Under semistrong-form efficiency, security
prices fully reflect all public information. Thus, only traders with access to nonpublic
information (some corporate insiders) can earn excess profits. However, in semistrongform efficiency, the market reacts so quickly to the release of new information that there
are no profitable trading opportunities based on public information. Finally, under strongform efficiency, all information, even apparent company secrets, are incorporated in
security prices; thus, no investor can earn excess profit trading on public or nonpublic
information.
Famas (1970) review has stimulated a definitional debate that extended for almost two
decades. LeRoy (1976) criticized Famas definition of market efficiency, because it allows
any feasible set of return definition to be consistent with efficiency. Fama (1976) revised
the definition by requiring that the market correctly uses all available information and
thus the joint distribution of future prices established by the market is identical to the
correct distribution implied by all available information at that time. Fama (1991)
reviewed again the literature on EMH but changed the previous classification to reflect the
new trend in academic research. He replaced the weak-form efficiency with tests for
3
predictability of returns, the semi-strong form with the event studies and the strong-form
with tests of private information.
empirical studies on developed market show no profitability from using past records of
price series supports the weak-form efficiency of the EMH in general.
On the other hand, the research findings of weak-form efficiency on the market of
developing and less developed markets are controversial. Most of the less developed
market suffers with the problem of thin trading. In addition, in smaller markets, it is easier
for large traders to manipulate the market. Though it is generally believed that the
emerging markets are less efficient, the empirical evidence does not always support the
thought. There are two groups of findings; the first group finds weak-form efficiency in
developing and less developed markets [Branes (1986) on the Kuala Lumpur Stock
Exchange; Chan, Gup and Pan (1992) in major Asian markets; Dickinson and Muragu
(1994) on the Nairobi Stock Exchange; and Ojah and Karemera (1999) on the four Latin
American countries markets] despite the problems of thin trading. On the other hand, the
latter group, who evidenced that the market of developing and less developed markets are
not efficient in weak-sense are Cheung, Wong and Ho, (1993) on the stock market of
Korea and Taiwan. In a world bank study, Claessens, Dasgupta and Glen (1995) report
significant serial correlation in equity returns from 19 emerging markets and suggest that
stock prices in emerging markets violates weak form EMH; similar findings are reported
by Harvey (1994) for most emerging markets. Kababa (1998) has examined the behaviour
of stock price in the Saudi financial market seeking evidence that for weak-form efficiency
and find that the market is not weak-form efficient. He explained that the inefficiency
might be due to delay in operations and high transaction cost, thinness of trading and
illiquidity in the market. Roux and Gilberson (1978) and Poshakwale (1996) find the
evidence of non-randomness stock price behaviour and the market inefficiency (not weakform efficient) on the Johannesburg stock Exchange and on the Indian market.
autocorrelation is observed. The results also suggest that it may take close to one month
for new information to be completely incorporated into stock prices on DSE. Chowdhury,
Sadique, and Rahman (2001) investigates the mean daily returns of DSE around the turn
of the week, turn of the month, turn of the year, and around the holidays. Using the
dummy variable regression model, they found that there is robust evidence of seasonal
return regularities. Regarding the days of the week effect the findings of this study are:
Average return on the last trading day of the week produces significant negative return (at
10% level), no significant abnormal returns on the last day of the week (5% level), no
significant average return on the opening day of the week, Mondays produce persistent
significant average negative returns. In their study, the researchers used market index,
which is very common and widely used, to find out the calendar effect. Hossain (2004),
using a different methodology, finds that when portfolios of bank stocks are constructed
on day one of the week with the opening prices and sold on Monday or the last day of the
week, there is considerable chance that the portfolio will provide significantly higher mean
daily return (MDR). Even when the portfolios were produced randomly, the MDRs came
as significantly higher in D1M (buy on day one of the week and sell on Monday of the
same week) and D6M (buy on day six of the week and sell on next Monday) strategies.
According to Hassan, Islam and Basher (2000), DSE equity returns show positive
skewness of 0.11 and 22.93 (excluding July 1, 1996 - December 31, 1996), excess kurtosis
of 49.66 and 992.65 (excluding 1996) and deviation from normality. Total sample period
was September 1986 - November 1999 and the two sub-periods of September 1986
December 1990 and January 1991 November 1999. The returns display significant serial
correlation of -0.07, implying stock market inefficiency. The results also show a
significant relationship between conditional volatility and the stock returns, but the riskreturn parameter is negative (-0.1072) and statistically significant. Mobarek (2000) sought
evidence supporting existence of at least weak-form efficiency of DSE. The sample
included the daily price indices of all listed securities on the DSE for the period of 1988 to
1997. The results provide evidence that the share return series do not follow random walk
model and the significant autocorrelation co-efficient at different lags reject the null
hypothesis of weak-form efficiency. Kader and Rahman (2004), by basically using k%
filter rule, find that abnormal profit is possible on regular basis by trading at a specific
pattern, which violates the random walk hypothesis.
METHODOLOGY
6
This paper gives an indication of the efficiency of the Dhaka stock market. We have
chosen only to test efficiency of the Dhaka Stock market in weak form. There are several
statistical tools to carry out such a study.
examined serial correlations of daily and weekly stock returns, which largely conclude that
the stock market is efficient. i. e., the stock prices follows a randomly process. For
example, Fama (1965) examined serial correlation coefficients for successive price
changes and had concluded that the behavior of the stock prices follows a randomly
process that was labeled as the random walk model. The hypothesis of a pure random walk
model is given as pt = + pt-1+ t , where pt is the natural logarithm of the securities price
series under consideration at time t, is a drift parameter and t is the random error term.
The usual stochastic assumptions of is that E() = 0 and E(2) = 2.
Summers (1986) challenges the way as the efficient market hypothesis is tested in the
early tests. He argues that the commonly used tests to evaluate market efficiency have
very low power. If a market is inefficient it means that prices have slowly decaying
stationary components. He shows that serial correlations of short-horizon returns cannot
give a clearer impression of the importance of these mean-reverting price components,
because the slow mean reversion can be missed with the short return horizons commonly
used in the efficiency tests. Based on these findings, Fama and French (1988) conduct
efficiency tests which try to identify if the behavior of long-horizon returns can shows the
importance of mean-reverting price components. They use a model for stock prices that is
the sum of a random walk and a stationary component. In their tests they find a pattern
that is consistent with the hypothesis that stock prices have a slowly decaying stationary
component. At the short return horizons the negative serial correlations generated by a
slowly decaying component is weak, but it becomes stronger as the return horizon
increases. So, it is clear that the random walk properties of securities returns are crucial for
the efficient market hypothesis. Indeed, if a security price series follows a random walk
process, it manifests significant permanent components and hence there is no meanreversion tendency. But on the other hand, if a security price series do not follows a
randomly process and presents significant temporary components, then it is possible to
predict the future security prices based on historical prices. In this case, it is possible to
design a profitable trading strategy based on historical data.
The subsequent research about the market efficiency has used a new methodology to test
the random walk nature of stock prices, which is known by unit root tests. This
methodology is used to examine the stationarity of the time series and was developed by
Dickey and Fuller (1979, 1981), among others. The most commonly used test to examine
the existence of a unit root is the augmented Dickey-Fuller (ADF) test. For example,
Kppi (1997) uses this procedure to test the stationarity of yield series before examine the
co-integration between yields on bonds and futures contracts on coupon bonds.
Due to the lack of power of the unit root tests and its failure to detect some deviations
from the random walk nature of time series, it was developed another type of tests for the
market efficiency hypothesis labeled as the variance ratio tests. This kind of tests was
originated from the pioneering works of Cochrane (1988), Lo and MacKinlay (1988,
1989) and Chow and Denning (1993). For example, Urrutia (1995) used the variance ratio
methodology to test the hypothesis that Latin American emerging equity market prices
(Argentina, Brazil, Chile and Mexico) follow a random walk.
The methodology of variance ratio test has been also used to analyze the efficiency of
other financial assets. For example, Liu and He (1991) use the procedure of Lo and
MacKinlay (1988) to provide empirical evidence of the random walk hypothesis in five
pairs of foreign exchange rates (CAN/USD, FRF/USD, DEM/USD, JPY/USD and
GBP/USD) and they reject the random walk hypothesis.
As stated above, serial correlation, unit root tests and variance ratio tests can be used to
test the efficiency hypothesis. Although the complementary of these tests, they can be used
all together to get a higher robustness of the conclusions. Lee, Gleason and Mathur (2000)
used all of them to test the efficiency of four financial futures contracts and have obtained
overwhelming evidence that the random walk hypothesis cannot be rejected for all of the
contracts. As suggested by the empirical evidence cited above, it should be mentioned that
we cannot conclude that any market is efficient or not without making an empirical test.
In our study, we choose some common tools only. These standpoints are in conformity
with Claessons (1987) reasoning concerning studies in the field of efficient markets. We
used both non-parametric tests [(Kolmogrov-Smirnov goodness of fit test), run test,
Lilliefors test], Q-Q probability plots and parametric-tests [Auto-correlation coefficient
8
test and Auto-Regressive Integrated Moving Average Model (ARIMA)]. Each statistical
test was performed separately on two index data. In some cases, total observation was
divided into two sub-samples for making the test result consistent and verified. In order to
avoid the possible bias originated from thin or infrequent trading we have used a longer
time-period, which reduces the problem of non-trading bias. Non-parametric test and
parametric-tests have been used to compare the results so that non-normal distribution can
not bias the findings.
The study includes daily All-Share Price Indices (ASPI) and DSE General Price Indices
(DSE-GEN) for 12 years. We used daily price indices of the Dhaka Stock Exchange for
the period of 1st January 1994 to 30th March 2005. Daily price indices were collected
from the Dhaka Stock Exchange (Research and Library Center). All the data came from
two indices for a certain period of time. The data of January 1994 to December 2001 are
all share price index (ASPI) data, and the data of January 2002 to March 2005 are general
price index (DSE-GEN) data. The sample included total 3026 daily observations for the
total sample period from 1994 to 2005. In addition, we also considered daily closing prices
of 33 companies. These companies represent Banks, Insurance and Pharmaceuticals
industries and are active in stock market. We executed run test on daily market returns of
these 33 companies.
Variables
We used daily market returns as individual time series variable. Market returns are
calculated from the daily price indices without adjustment of dividend, bonus and right
issues. Many researchers confirm that their conclusions remain unchanged whether they
adjusted their data for dividend or not [for example, Lakonishok and Smidt, (1988); Fishe,
Gosnell and Lasser, (1993)]. Daily market returns (Rmt ) are calculated from the daily price
indices as follows:
Rmt = Ln (PI t / PI t-1)
(1)
Where, Rmt refers to market return in period t; PIt, price index at day t; PIt-1, the price
index at period t-1 and Ln refers to natural log. We used logarithm just because, lognormal
returns are more likely to be normally distributed, which is prior condition of standard
statistical techniques (Strong, 1992). For individual companies, the daily return, (Rmtj)
equals
Ln [Pt / Pt-1]
(2)
Where, Pt refers to daily price per share at time t; Pt-1 is the daily price per share at time t 1; and j is the individual security. Table 1 shows the variables used in this study.
Table 1: Variables and Description
Name of the Variables
Descriptions
Hypotheses
This paper tests two hypotheses to determine the efficiency of the DSE in the weak form.
This concept of using hypotheses is similar to that of Moustafa (2004) who tested weak
form efficiency on United Arab Emirates stock market. Unlike Mobarek (2000), the first
hypothesis involves determining whether the stock returns follow a normal distribution or
not. The null and alternative hypotheses are:
H0: The stock returns in DSE stock market follow a normal distribution.
H1: The stock returns in DSE stock market do not follow a normal distribution.
The second hypothesis involves determining, whether the stock returns are random across
time. The null and alternative hypotheses are:
H0: The stock returns in DSE are random over the time period of the study.
H1: The stock returns in DSE are not random over the time period of the study.
Though hypothesis of normality and randomness are complementary, we used them
simultaneously in order to establish the robustness of the analysis. Besides, some specified
hypotheses have taken into consideration while using several parametric and nonparametric tests.
10
Figure 1 and 2 illustrate daily market return series and daily price index series for getting
presumption.
Figure 1: Daily Market Return Series
Daily Market Return Series (2002-2005)
30
20
10
-2
-10
-4
-20
-6
-30
-8
Test of Normality
We start with test of normality. Descriptive statistics, Kolmogrov-Smirnov (K-S) test and
Lilliefors Coefficient test had been used to test the hypothesis of normality. Daily market
return data (Rmt) from both indices are used in this goodness of fit test. In order to test the
distribution of the return series, the descriptive statistics of the log of the market returns
are calculated and presented in Table 2. Result shows that the returns are not normally
distributed. Rather, characterized by significantly high skewness and kurtosis. In a
symmetrical distribution the values of mean, median and mode are alike. As the value of
mean is greater than the mode, so market return series follow positively skewed
distribution. Generally, values for skewness (zero) and kurtosis (3) represents that the
observed distribution is perfectly normally distributed. The kurtosis value of 61.754 for
ASPI and 7.815 for DSE-GEN exhibits extreme leptokurtic distribution. So, skewness and
leptokurtic frequency distribution of stock return series on the DSE rejects our null
11
30-Dec-04
4-Sep-04
20-May-04
20-Jan-04
30-Sep-03
16-Jun-03
1-Aug-02
17-Apr-02
1-Jan-02
16-Sep-01
17-Jan-01
13-Jun-00
2-Mar-99
17-Oct-99
14-Jul-98
15-Nov-97
13-Mar-97
08-Aug-96
09-Nov-95
20-Apr-95
27-Aug-94
01-Jan-94
2500.00
2000.00
1500.00
1000.00
500.00
0.00
4-Mar-03
4000.00
3500.00
3000.00
2500.00
2000.00
1500.00
1000.00
500.00
0.00
13-Nov-02
hypothesis of normality as well as contradicts with the random walk model. We further
justify the hypothesis with K-S goodness of fit test.
Table 2: Descriptive Statistics of Log of the Daily Market Return
Statistic
Variable
Daily Market
Return (Rmt)
Description
Number of observations
Mean
Median
Mode
Variance
Std. Deviation
Minimum
Maximum
Range
Skewness
Kurtosis
General price
index (20022005)
903
.0922
.0480
.00
.85391
.92407
-7.36
4.83
12.19
-.428
7.815
Mean
Std. Deviation
Absolute
Positive
Negative
Kolmogorov-Smirnov Z
Asymp. Sig. (2-tailed)
12
General price
index
(2002-2005)
903
.0922
.92407
.078
.070
-.078
2.354
.000
Lilliefors coefficient
Table 4 displays the Kolmogorov-Smirnov (K-S) statistic with a Lilliefors significance
level for testing normality. If non-integer weights are specified, the Shapiro-Wilk (S-W)
statistic is calculated when the weighted sample size lies between 3 and 50. For no weights
or integer weights, the statistic is calculated when the weighted sample size lies between 3
and 5000. The probability of (.000) for both (K-S) and (S-W) implies test statistics are
significant and asserts that market return series do not follow normal distribution.
Table 4: K-S & S-W statistics for market return series
Kolmogorov-Smirnov (a)
Statistic
df
Sig.
1.71
2123
.000
.078
903
Shapiro-Wilk
Statistic
df
.691
2123
.000
.921
903
Sig.
.000
.000
3
2
-1
-1
Expected Normal
Expected Normal
7. Test
of Randomness on the Return Series
2
-2
-3
-4
-30
-20
-10
10
20
30
-2
-3
-4
-6
Observed Value
-4
Observed Value
-2
Detrended Normal Q-Q Plot (All Share Index Detrended Normal Q-Q Plot (General Price Ind
20
10
1
-10
13
Normal
Normal
-1
Runs Test
We used run test to determine statistical dependencies or randomness, which may not be
determined by auto-correlation test. This test is regarded as powerful test to prove randomwalk model as it disregards the properties of distribution. The null hypothesis of this test is
that the observed series is a random series.
We can define a run as an uninterrupted sequence of one symbol or attribute. The number
of runs is computed as a sequence of the price changes of the same sign (such as; + + or
). In other words, length of a run is the number of elements in it. When the expected
number of run is significantly different from the observed number of runs, the test rejects
the null hypothesis that the daily returns are random as Poshakwale (1996) stated, A
lower than expected number of runs indicates markets overreaction to information,
subsequently reversed, while higher number of runs reflects a lagged response to
information. Either situation would suggest an opportunity to make excess returns.
The run test converts the total number of runs into a Z statistic. For large samples, the Z
statistics gives the probability of difference between the actual and expected number of
runs. The Z value is greater than or equal to 1.96; reject the null hypothesis at 5% level
of significance (Sharma and Kennedy, 1977).
Table 5: Result of Run test on the return series of DSE for the period 1994 to 2005
14
Result shows that the Z statistics of daily market return of (-11.672) and (-6.077) for both
indices are greater than 1.96 and negative (Table 5), which means the observed number
of runs is fewer than the expected number of runs with observed significance level. In
addition to runs test of market returns, we further proceed to runs test of company-specific
stock returns. Z values of daily stock returns of 27 companies out of 33 individual
companies, which are from the most active sectors at DSE, are negative and grater than
1.96 (Appendix 1), which supports our previous findings that the return series do not
follow random walk. Thus, we can accept the alternative hypothesis that the return series
on the DSE do not follow random walk. In brief, the results of runs test on the Dhaka
Stock Exchange indicate that the daily stock returns of DSE are not random as the Z
statistic does not fall in between 1.96. These results are all the same to the findings of
Mobarek (2000).
Autocorrelation Tests
An autocorrelation measures the association between two sets of observations of a series
separated by some lags. It tests whether the correlation coefficients are significantly
different from zero. Ratio of ACF and Standard Error (T-ratio) has been employed to
determine this. On the other hand, Ljung-Box statistic is used here to diagnose white noise
(purely random) of the series.
The autocorrelation coefficients, computed for the log of the market return series, shows
significant auto-correlation at different lags for the whole sample period comprising all
share and general price index. Appendix 2 and 3 show the correlogram of market return
series for both indices. It is clearly evident that the return series are not purely random,
rather it shows weak stationarity and there is significant positive auto-correlation
coefficient at 1st, 4th, 5th, 10th, 11th, 21st, 24th, 25th, 26th, and 27th lag for all share price
index and 1st, 3rd, and 22nd lag for general price index. The presence of non-zero auto-
15
correlation coefficients in the log of market returns series suggests serial dependence
between the values.
The nonzero auto-correlation of the series is further tested by Ljung-Box Q statistics,
which are jointly significant at 5% level at 30 degrees of freedom.
As Ljung-Box
diagnoses white noise series which is completely random and has pattern less ACFs, a
constant variance and a mean of zero. We commonly infer the series is not white noise
when the level of significance is less then 0.05 and that it is consistent with white noise
when the level of significance is greater than of equal to 0.05 (DeLurgio, 1998).
All the probability falls below the tolerance level of .05 and all the LB statistics are greater
then the table value of Ljung-Box formula (such as for LB30, table value of 43.77 is
associated with 5% level of significance and 30 degrees of freedom, considering that it
follows chi-square distribution). We reject the hypothesis of white noise. In conclusion,
ACF patterns are statistically, significantly different than those of white noise. Hence the
results in Appendix 2 and 3 suggest that return series do not support random walk model.
Let us check the random walk model with graphical representation of ACFs. Figure 5 is
the depiction of ACFs for both all share and general price index. Random Walks are
characterized by extremely high autocorrelations. That is, adjacent observations are highly
associated with each other (DeLurgio, 1998). Figure 5 of random walk model depicts
ACFs declining gradually.
P(k)
Lag(k)
16
10
11
12
.2
.1
.1
0.0
0.0
ACF
Confidence Limits
-.1
-.1
1
Coefficient
1
9 11 13 15 17 19 21 23 25 27 29
9 11 13 15 17 19 21 23 25 27 29
Lag Number
Lag Number
Upon closer inspection of Figure 5, we note that ACF starts with positive correlation and
remained positive and negative at different lags. ACFs in the Figure 5 do not show gradual
declination as random walk does. Hence it is proved that DSE market return series does
not support random walk.
The characteristics of ACFs of random series tell that if a series is completely random,
then 50 percent of the ACFs will be above and 50 percent below zero. Above figures do
not maintain this property properly. Thus, there may be some pattern left in the ACFs.
Once again it is proved that DSE market return series is not a white noise series as such
series is patternless.
17
different from zero asserts dependency of the series, which violates the assumption of
random walk model and deviates from weak-form efficiency.
Table 6: Results of ARIMA (0,1,0) for all share & general price index
ARIMA (0,1,0)
Constant (ASPI)
Constant (DSE-GEN)
Coefficient
.203
1.203
SE
.640
.429
T-ratio
.318
2.800
Prob.
.750
.005
18
a part of the observations, it can forecast the future values of the series and will also match
with the rest of the observations.
15
Fit_1
10
lcl_1
ucl_1
0
-5
-10
-15
Figure 7: Building up Predictive Model [ARIMA, (1,0,1)] for the Historical Period (1451) and Forecasting the Validation Period (452-903). [For General Price Index]
6
4
2
RMT2(451-903)
0
fit_1
lcl_1
-2
ucl_1
-4
-6
-8
19
Hence, we employ ARIMA (3,0,2) and ARIMA (1,0,1) models to generate data for
validation period and examine how far the fitted value deviates from the actual value.
With a closer inspection in Figures 6 and 7, it is evident that the fitted value derived from
the models and actual value is all but well fitted.
We are now in a conclusion that validation period is the reflection of historical period.
More specifically, past price can be used to determine the future price of the stocks. Thus,
ARIMA model confirms the previous findings and results of various parametric and nonparametric tests, and also these models are consistent all over the sampling period, subsample period for daily market return series of all share price index and general price
index.
CONCLUSION
Overall results from the empirical analysis suggest that the Dhaka Stock Market of
Bangladesh is not efficient in weak-form. No analysis was undertaken on technical trading
rule or adjusting transaction cost. Though all the tests revealed the rejection of weak form
of efficiency, nevertheless emphasis should be given to interpretation and realization
process. Simply saying, evidence of auto-correlation, which violates the assumption of
random walk does not necessarily prove inefficiency of the market. There might some
other factors to be considered which will bring the capital market in the efficiency
spectrum.
As mentioned earlier, this research mainly hunted for the evidence of weak form
efficiency by hypothesizing normality of the distribution series and random walk
assumption. In other words, daily market return series were compared with the white noise
series, keeping in mind that white noise series is purely stationary. In the aspect of
skewness and kurtosis, the daily return series were found non-normal, which can be
categorized as positively skewed distribution and having a little bit resemblance with chisquare distribution. Same thing resulted from Kolmogorov-Smirnov (K-S) test. As a
result, null hypothesis of normality has been rejected and alternative hypothesis remained
in effect. Run test and auto-correlation results rejected the randomness of the return series
20
of DSE simultaneously. Finally, ARIMA model, as the part of the time series forecasting,
strengthen the non-random walk situation of Dhaka Stock Exchange.
The absorption of good and bad news or any other price forming information may take late
effect on share price because of available advance technology, control system and
publication of business journals. So, before denouncing an inefficient market, above
factors should get priority. However, DSE deviated from weak form EMH. But it would
not be wise to label it as inefficient, because market efficiency changes over time and
capital market is subject to be tested continuously.
References
Ahmed, F. (2002). Market Efficiency in Emerging Stock Markets: The Case of Dhaka
Stock Exchange, www.fgda.org/html/savings_2002-1.htm.
Bachelier, L. (1900). Thorie de la Spculation. (Doctoral dissertation in Mathematical
Sciences, Facult des Sciences de Paris, defended March 29, 1900), Annales de /cole
Normale Suprieure, Vol. 3, pp. 21 -86. Translated with permission of GauthierVillars, Paris, France, as Chapter 2 in Cootner, Paul H. (1964), Ed. The Random
Character of Stock Market Prices, Cambridge, MA: The MIT Press.
Branes, Paul (1986). Thin trading and stock market efficiency: A case of the Kuala
Lumpur Stock Exchange. Journal of Business Finance & Accounting, Vol. 13, No. 4
(Winter), pp. 609-617.
Chan, Kam C., Gup, Benton E., and Pan, Ming-shiun (1992). An Empirical Analysis of
Stock Prices in Major Asian Markets and United States. The Financial Review, Vol.
27, No. 2 (May), pp. 289-307.
Cheung, Yan Leung, Wong, Kie-Ann, and Ho, Yan-Ki (1993). The pricing of risky assets
in two emerging Asian markets- Korea and Taiwan. Applied Financial Economics,
Vol. 3, No. 4 (December), pp.315-324.
Chow, V. K., and Denning, K. D. (1993). A Simple Multiple Variance Ratio Test. Journal
of Econometrics, Vol. 5, pp. 385-401.
Chowdhury, Shah S. H., Sadique, M. Shibley, and Rahman, M. Arifur (2001). Capital
Market Seasonality: The Case of Dhaka Stock Exchange (DSE) Returns. South Asian
Journal of Management, Vol. 8, No. 3 & 4 (July-Dec.), pp. 1-7.
Claessens, Stijin, Dasgupta, Susmita, and Glen, Jack, (1995). Return behaviour in
emerging Stock Market. The World Bank Economic Review, Vol. 9, No.1, pp. 131151.
Claesson, K. (1987). Effektiviteten p Stockholms fondbrs, Stockholm: EFI.
Cochrane, J. H. (1988). How Big is the Random Walk in GNP? Journal of Political
Economy, Vol. 96, pp. 893-920.
21
Cootner, Paul H. (1962). Stock Prices: Random Vs. Systematic Changes. Industrial
Management Review, Vol. 3 (Spring), pp. 24 45.
Cootner, Paul H. (1964). Ed., The Random Character of Stock Market Prices. Cambridge,
MA: The MIT Press.
Cowles, A. III, and Jones, H. (1937). Some a Posteriori Probabilities in Stock Market
Action. Econometrica, Vol. 5, pp. 280-294.
DeLurgio, Stephen A. (1998). Forecasting Principles and Applications. First Edition,
USA: Irwin McGraw-Hill.
Dickey, D. A., and Fuller, W. A. (1979). Distribution of the Estimators for Autoregressive
Time Series with a Unit Root. Journal of American Statistical Association, Vol. 74,
No. 366, pp. 427-431.
Dickey, D. A., and Fuller, W. A. (1981). Likelihood Ratio Statistics for Autoregressive
Time Series with a Unit Root. Econometrica, Vol. 49, pp. 1057-1072.
Dickinson and Muragu (1994). Market Efficiency in Developing Countries: A case study
of the Nairobi Stock Exchange. Journal of Business Finance & Accounting, Vol. 21,
No. 1 (January), pp. 133-150.
Fama, Eugene (1965). The Behavior of Stock Market Prices. Journal of Business, Vol. 38
(Jan), pp. 34-105.
Fama, Eugene (1970). Efficient Capital Markets: A Review of Theory and Empirical
Work. Journal of Finance, Vol. XXV, No. 2 (March), pp. 383 417.
Fama, Eugene (1976). Foundation of Finance. New York: Basic Books.
Fama, Eugene (1991). Efficient Capital Markets II. Journal of Finance, Vol. 26, No. 5
(Dec.), pp. 1575-1617.
Fama, Eugene, and French, Kenneth R (1988). Permanent and Temporary Components of
Stock Prices. Journal of Political Economy, Vol. 96 (April), pp. 246 273.
Fishe, R., Gosnell, T., and Lasser, D. (1993). Good news, bad news, volume and the
Monday effect. Journal of Business Finance & Accounting, Vol. 20, 881-892.
Harvey, Campbell R. (1994). Conditional Asset allocation in Emerging Markets. Working
Paper, No. 4623, Cambridge, MA.
Hassan, M. K., Islam, A. M. and Basher, S. A. (2000). Market Efficiency, Time-Varying
Volatility and Equity Returns in Bangladesh Stock Market.
http://econwpa.wustl.edu:80/eps/fin/papers/0310/0310015.pdf.
Hossain, M. Farhad (2004). Days of the Week Effect in Dhaka Stock Exchange: Evidence
from Small Portfolios of Banking Sector. The Jahangirnagar Review, Part II: Social
Science, Vol. XXVIII (Printed in 2005), pp. 73-82.
Hudson, R., Dempsey, M., and Keasey, Kevin. (1994). A note on the weak-form
efficiency of capital markets: The application of simple technical trading rules to UK
Stock prices-1935 to1994. Journal of Banking & Finance, Vol. 20, pp. 1121-1132.
Kababa, Nourrendine. (1998). Behavior of stock prices in the Saudi Arabian Financial
Market: Empirical research findings. Journal of Financial Management & Analysis,
Vol. 11, No. 1 (Jan-June), pp. 48-55.
22
23
Poterba, James, and Summers, Lawrence (1988). Mean Reversion in Stock Prices:
Evidence and Implications. Journal of Financial Economics, Vol. 22, No. 1 (Oct.), pp.
27 59.
Roux and Gilberson (1978). The behavior of share prices on the Johannesburg Stock
Exchange. Journal of Business Finance and Accounting, Vol. 5, No. 2, pp. 223-232.
Strong, N. (1992). Modelling Abnormal Returns: A Review Article. Journal of Business
Finance and Accounting, Vol. 19, No. 4 (June), pp. 533-553.
Summers, Lawrence H. (1986). Does the Stock Market Rationally Reflect Fundamental
Values? Journal of Finance, Vol. 41, No. 3 (July), pp. 591-601.
Urrutia, J. L. (1995). Tests of Random Walk and Market Efficiency for Latin American
Equity Markets. Journal of Financial Research, Vol. 18, No. 3, pp. 299-309.
Working, H. (1934). A Random Difference Series for Use in the Analysis of Time Series.
Journal of the American Statistical Association, Vol. 29, pp. 11-24.
24
Appendices
Appendix 1: Result of Run test of some selected DSE companies return
for the period 1994 to 2005
Individual Companys Daily Return
No. Run
Asymp. Sig.
(2-tailed)
421
1087
1143
1242
1223
957
1054
442
426
1030
908
833
773
323
335
470
476
921
854
266
321
291
288
331
548
35
725
685
691
559
874
303
157
-8.316*
-3.623*
-4.752*
-1.526
-2.770*
-4.277*
0.040
-0.746
-3.237*
-6.646*
-14.597*
-17.060*
-16.792*
-25.791*
-.482
-4.373*
-6.330*
-3.005*
-4.289*
-8.331*
-2.711*
-2.682*
-3.100*
-5.925*
-8.489*
-.153
-7.425*
-9.638*
-9.833*
-2.297*
-9.925*
-11.881*
-1.416
.000
.000
.000
.127
.006
.000
.968
.455
.001
.000
.000
.000
.000
.000
.630
.000
.000
.003
.000
.000
.007
.007
.002
.000
.000
.875
.000
.000
.000
.022
.000
.000
.157
No.
Name
1
1st ICB
2
ACI
3
Aftab
4
Bata
5
Monno Ceramic
6
BGIC
7
Ambee
8
GLAXO
9
Renata
10
Square
11
Pharmaco
12
Reckitt
13
IBN Sina
14
Pharma Aids
15
Keya Cosmetics ltd.
16
Libra Pharma
17
Orion Pharma
18
Bexmico Infusion
19
Eastern Insurance
20
Peoples Insurance
21
National Life Insurance
22
Pragati General
23
Sandhani Life insurance
24
Delta life isurance
25
Green Delta
26
Pubali Bank
27
Rupali Bank
28
UCBL
29
Uttara Bank
30
Al-Arafah
31
Islami Bank
32
Oriental Bank
33
Social Investment Bank
Mean is used as cut point.
25
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
AutoCorr.
.142*
-.002
.032
.068*
.053*
-.011
.000
.019
.032
.065*
.065*
.028
.031
.009
-.010
-.016
.020
-.025
-.015
.002
.048*
.031
.022
.057*
.052*
.055*
-.048*
.033
.001
-.016
Stand.
Err.
-1
-.75
-.5 -.25
.022
.022
.022
.022
.022
.022
.022
.022
.022
.022
.022
.022
.022
.022
.022
.022
.022
.022
.022
.022
.022
.022
.022
.022
.022
.022
.022
.022
.022
.022
.25
.5
.75
. .**
.*.
. *
. *
. *
.*.
.*.
.*.
. *
. *
. *
. *
. *
.*.
.*.
.*.
.*.
.*.
.*.
.*.
. *
. *
.*.
. *
. *
.*.
* .
.*.
.*.
.*.
26
BoxLjung
42.789
42.794
44.954
54.711
60.651
60.931
60.931
61.721
63.886
72.832
81.899
83.547
85.572
85.731
85.950
86.474
87.342
88.655
89.149
89.160
94.189
96.276
97.301
104.405
110.140
110.186
115.232
115.246
115.249
115.768
Prob.
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
AutoCorr.
Stand.
Err.
.148*
-.022
.087*
-.001
-.012
-.009
.062
-.004
.025
.027
-.020
.037
.046
.021
.036
.051
-.012
.009
.038
.013
.002
.070*
.035
-.057
-.010
.045
-.012
.042
.029
-.026
.033
.033
.033
.033
.033
.033
.033
.033
.033
.033
.033
.033
.033
.033
.033
.033
.033
.033
.033
.033
.033
.033
.033
.033
.033
.033
.033
.033
.033
.033
-1
-.75
-.5 -.25
.25
.5
.75
. .**
.*.
. .*
.*.
.*.
.*.
. *
.*.
.*.
. *
.*.
. *
. *
.*.
. *
. *
.*.
.*.
. *
.*.
.*.
. *
. *
* .
.*.
. *
.*.
. *
. *
* .
27
BoxLjung
19.720
20.141
26.997
26.998
27.127
27.193
30.735
30.749
31.315
31.967
32.333
33.615
35.577
35.967
37.164
39.597
39.719
39.799
41.124
41.293
41.297
45.827
46.986
50.021
50.120
52.005
52.148
53.830
54.638
55.284
Prob.
.000
.000
.000
.000
.000
.000
.000
.000
.000
.000
.001
.001
.001
.001
.001
.001
.001
.002
.002
.003
.005
.002
.002
.001
.002
.002
.003
.002
.003
.003
Overall
results
Randomness Test
Normality Test
Hypothesis-1
DSE return series are not random or supports Random Walk Model.
Hypothesis-2
COMMENTS
28