Working Paper Series
Working Paper No. 05-01
October 2005
Are the UAE Financial Markets
Efficient?
By
Jay Squalli*
* EPRU, Zayed University
Are the UAE Financial Markets Efficient?
Jay Squalli∗
Working Paper No. 05-01
Abstract
This paper tests for market efficiency in the represented sectors of the Dubai Financial Market (DFM)
and the Abu Dhabi Securities Market (ADSM). Using daily sectoral indices between 2000 and 2005,
variance ratio tests reject the random walk hypothesis in all sectors of the UAE financial markets except
in the banking sector of the DFM. Returns in the two financial markets are negatively serially correlated,
thus suggesting the presence of a Bull market. Runs tests find insurance in the ADSM to be the only weakform efficient sector. Cointegration and Granger Causality tests provide evidence of one-way causality
between the common sectors of the ADSM and the DFM except for the insurance sector, thus suggesting
that efficiency does not necessarily spill over across markets.
∗ Assistant Professor of Economics, Zayed University, Economic & Policy Research Unit, P.O. Box 19282, Dubai, UAE,
Phone: +971 4 208 2465, Fax: +971 4 264 0394, E-mail: [email protected]
1
1
Introduction
Around the time market efficiency theory came to life, a country was born: the United Arab Emirates
(UAE). For the past three decades, the UAE has experienced jaw-dropping growth rates that have made
the country one of the most sought-after investment hubs in the gulf. Increased oil revenues and a stronger
focus on the diversification of economic activities have created an environment conducive for viable and
lucrative investments. As the country embraces international economic integration and alignment with
global economic and financial standards, the need for highly structured financial markets has gained growing
attention. Since the official inception of the only UAE’s financial markets: the Abu Dhabi Securities Market
(ADSM) and the Dubai Financial Market (DFM) in 2000, capitalization in the two respective markets has
grown by 343% and 1,238%.1
The rising interest in investment opportunities in emerging economies has raised questions about the
efficiency of their financial markets. Why is it so important that financial markets are efficient? When
financial markets are (weak-form) efficient, the prices paid for stocks reflect past prices and the trading
history of a security at each point in time and thus reflect the true value of stocks and result in the optimal
allocation of private and social resources. Efficiency eliminates market distortions and arbitrage opportunities
based on asymmetric information. Therefore, there are no opportunities for earning abnormal returns and
no successful systematic attempts at deriving financial forecasts. Financial markets have long been known
as the driving force of the economy and a strong stimulant for economic growth. Among many of their
roles, they serve as a means for firms and the government to raise capital for expansion and infrastructural
development, improve the allocation of resources through the mobilization of savings into more productive
economic activities, and improve corporate governance through shareholder participation in the management
of corporate activities. Weak-form efficiency is therefore a very desirable characteristic for financial markets
to exhibit.
Are the UAE financial markets (weak-form) efficient? The literature on the efficiency of the UAE financial
markets has been relatively meager and provides mixed results (Ebid, 1990; Moustafa, 2004). This paper
attempts to fill this void by investigating the random walk hypothesis and weak-form efficiency using the
1 Source:
Central Bank of the UAE, Annual Report, 2004.
2
variance-ratio and the runs tests. These procedures are robust and have been widely used in the international
academic literature (Butler and Malikah, 1992; Urrutia, 1995; Grieb and Reyes, 1999; Ojah and Karamera,
1999; Abeysekera, 2001; Moustafa, 2004). This paper is organized as follows: Section 2 summarizes the
relevant literature. Section 3 describes the UAE financial markets. Section 4 describes the data and empirics.
Section 5 summarizes the main results and concludes.
2
Previous Literature
There is a large literature on the efficiency of emerging financial markets. Among the most relevant papers are
studies on the Latin American financial markets and the middle east. Urrutia (1995) assesses the efficiency of
the financial markets of Argentina, Brazil, Chile, and Mexico. The author rejects the existence of a random
walk when using a variance-ratio test and finds all four markets to be weak-form efficient when using a runs
test. Urrutia further adds that the rejection of the random walk hypothesis suggests the presence of positive
serial correlation in returns. He explains that the positive correlation does not necessarily imply that the
markets are inefficient but that it could be indicative of economic growth (especially in emerging markets).
His results are consistent with Lo and MacKinlay (1988) and Poterba and Summers (1988) (for short time
horizons). Ojah and Karamera (1999) later confirm Urrutia’s results using multiple variance-ratio and autoregressive fractionally integrated moving-average tests. Grieb and Reyes (1999) revisit the Brazilian and
Mexican markets using variance-ratio tests and find evidence of a random walk only for Brazil.
One of the leading studies of market efficiency in the middle east by Butler and Malikah (1992) finds
the Saudi and Kuwaiti markets not to be weak-form efficient using serial correlation and runs tests. AlLoughani (1995) finds further evidence of an inefficient Kuwaiti stock market using various tests. More
recently, Abraham et al. (2002), using the Beveridge-Nelson (1981) decomposition of index returns to
control for thin trading, runs tests, and variance-ratio tests, find evidence of weak-form efficiency in the
Saudi, Kuwaiti, and Bahraini equity markets.
The published literature on UAE stock markets is relatively meager and, to my knowledge, is limited to
Ebid (1990) and Moustafa (2004). Ebid (1990) uses data preceding the official inception of the UAE financial
markets by about 14 years and finds the UAE financial markets to be weak-form inefficient using parametric
3
serial correlation tests. However, this procedure is not robust unless supplemented with tests verifying that
returns are normally distributed.2 A more recent study by Moustafa (2004) uses data between 2001 and 2003
to investigate weak-form efficiency of the UAE financial markets. The author implicitly describes the “UAE
stock market” as one market composed of all firms in the two UAE financial markets and completes runs
tests on individual firms’ returns. Moustafa (2004) concludes that most firms are weak-form efficient. The
assumption that all firms from two independent financial markets form a single “UAE stock market” weakens
the argument for weak-form efficiency since the regulatory environment, the market structure, the degree of
competition, the firms and corresponding listed industries are far from being common across the two distinct
ADSM and DFM markets. Furthermore, a runs test is insufficient in providing conclusive answers about the
efficiency of the UAE financial markets as it fails to test for serial correlation.
3
The UAE Financial Markets
Following the approval of the stock exchange law in June 1999, two financial markets were created in the
UAE: the ADSM and the DFM. The inception of these markets would turn the city of Abu Dhabi into the
base for the Securities and Commodities Commission and would introduce electronic trading on the two
trading floors of Abu Dhabi and Dubai. Prior to the creation of the ADSM and the DFM, the exchanging
of stocks took place in an unofficial parallel market. Since electronic trading was only used after the official
opening of the two UAE financial markets, full disclosure of trade volume and prices would play an important
role in market stability.
The ADSM, the largest of the two, began operations in November 2000. As of September 2005, there are
50 companies listed in the ADSM (see table (1)). The number of firms and their corresponding sectors are:
14 firms in banking, 2 firms in hotels, 14 firms in industry, 11 firms in insurance, and 9 firms in services.
Company listings have grown steadily over the past five years, going from 53% between 2001 and 2002, to
17%, 15%, and 61% yearly onwards, as reported in table (2). As a result of this spike in stock market listings,
market capitalization has increased by 343%.
The DFM began its operations in March 2000 solely as a securities trading market before expanding
to stock exchange. The DFM was initially expected to operate as a secondary market for the trading of
2 Examples
of such tests include a Jarque-Bera test and the Kolmogorov-Smirnov test.
4
securities issued by public shareholding companies and bonds issued by the local or federal government,
public institutions, and financial and investment institutions. As of September 2005, there are 28 companies
listed in the DFM (see table (3)). The number of firms and their corresponding sectors are: 7 firms in
banking, 9 firms in insurance, 4 firms in investment, and 7 firms in services.3 Company listings have grown
by 50% between 2001 and 2002, and 11%, 60%, and 75% yearly onwards, as reported in table (4). This
substantial increase in market listings has resulted in a 1,238% increase in capitalization.
4
Data and Methodology
4.1
Description of the Data
The data used for this study were obtained from the ADSM and DFM and include daily sectoral indices for
the ADSM from September 30, 2001 through July 19, 2005 and for the DFM from March 26, 2000 through
September 17, 2005. The sectors included are banking, hotels, industry, insurance, and services for the
ADSM and banking, insurance, investment, and services for the DFM. Since the sectoral indices were not
used immediately after the official opening of the DFM, the 4 sectors of the DFM include data from August
28, 2003 through September 17, 2005 while general index data range between March 26, 2000 and December
30, 2004.
The importance of a random walk resides in its ability to provide support for market efficiency. In fact,
nonstationary series suggest the absence of temporal dependence in the series or information building. This
is consistent with price fluctuations that are not subject to a deterministic time trend and in which updates
occur only with new information. Lo and MacKinlay (1988) suggest the use of a variance-ratio (VR) statistic
to test the random walk hypothesis. However this procedure is not sufficient on its own to assess weak-form
efficiency. In fact, when the random walk hypothesis is rejected, the alternative hypotheses are that the
series analyzed are serially correlated. Therefore, further testing must be completed to provide an accurate
assessment of weak-form efficiency. This has been commonly done with a runs test. In the next sections,
I first complete a variance ratio test to test for the random walk hypothesis. Second, I assess weak-form
efficiency with a runs test. Third, because of the close proximity between the two UAE financial markets, I
test for causation across the two markets.
3 The
industrial sector does not report its index as it is represented by only one firm in the DFM.
5
4.2
Variance-Ratio Tests
The VR procedure is motivated by the fact that the variance of a random walk increases linearly with time.
The VR approach has gained popularity and has become the standard tool in random-walk testing. The VR
is calculated as follows:
V R(q) =
σ 2 (q)
σ 2 (1)
(1)
where σ 2 (q) is the unbiased estimator of 1/q of the variance of the qth difference and σ 2 (1) is the variance
of the first difference:
nq
σ 2 (q) =
X
1
(xt − xt−q − q µ̂)2
q(nq − q + 1)(1 − q/nq) t=q
(2)
and
nq
σ 2 (1) =
where µ̂ =
1
nq (xnq
1 X
(xt − xt−1 − µ̂)2
nq − 1 t=1
(3)
− x0 ). Under homoscedasticity, the asymptotic variance of the variance-ratio is expressed
as follows:
v(q) =
2(2q − 1)(q − 1)
3q(nq)
(4)
Under heteroscedasticity, the asymptotic variance can be expressed as:
q−1
X
2(q − k) 2
[
v (q) =
]
q
∗
Pnq
− xt−1 − µ̂)2 (xt−k − xt−k−1 − µ̂)2
Pnq
[ t=1 (xt − xt−1 − µ̂)2 ]2
t=k+1 (xt
k=1
(5)
The homoscedasticity and heteroscedasticity consistent Z-statistics are respectively denoted by Z(q) and
Z ∗ (q) and expressed as follows:4
Z(q) =
V R(q) − 1 a
p
→ N (0, 1)
v(q)
(6)
Z ∗ (q) =
V R(q) − 1 a
p
→ N (0, 1)
v ∗ (q)
(7)
and
where the null hypothesis is that V R(q) = 1 or that the chosen index follows a random walk. When the
random walk hypothesis is rejected and V R(q) > 1, returns are positively serially correlated and consistent
with the findings of Urrutia (1995), Lo and MacKinlay (1988) and Poterba and Summers (1988) (for short
time horizons). For emerging markets, positive serial correlation in returns could simply describe market
4 Because
of the inevitable heteroscedasticity in the data, the latter statistic is given the most attention.
6
growth (Urrutia, 1995). When the random walk hypothesis is rejected and V R(q) < 1, returns are negatively
serially correlated. This situation is often described as a mean-reverting process and consistent with the
findings of Summers (1986) and Fama and French (1988). This has been interpreted as an efficient correction
mechanism in mature markets (Fama and French, 1988) and as a sign of a “bubble” in emerging financial
markets (Summers, 1986).
As reported in tables (5) and (6), the null hypothesis of a random walk is rejected for all sectors of the
ADSM under homoscedasticity and heteroscedasticity. The results for the DFM are similar except for the
banking sector in which the null hypothesis of a random walk is not rejected under the heteroscedasticity
assumption. These results are consistent with the results presented by Urrutia (1995). However, because
the reported values of the variance-ratio V R(q) are below 1, there appears to be negative serial correlation
in the series, consistent with mean-reversion and the likelihood of a “bubble”.
4.3
The Runs Test
Despite the rejection of the random walk for all sectors of the UAE financial markets (except banking in the
DFM), weak-form efficiency remains a possibility (Lucas, 1978). Weak-form efficiency is when all past prices
and information are fully captured in securities prices. In other words, there is no temporal dependence in
the prices and thus technical predictions and forecasts based on historical data are rendered useless. This
would be true if returns are distributed randomly. To determine whether this is the case for the UAE
financial markets and to identify the proper testing procedure, I complete a Jarque-Bera (JB) test to assess
whether returns are normally distributed. The JB statistic is calculated as follows:
JB =
ns2
(k − 3)2
+
6
4
(8)
where s is the skewness, k is the kurtosis, and n is the number of observations. The lower the JB statistic, the
more likely a distribution is normal. As evidenced in tables (7) and (8), the JB test rejects the null hypothesis
that series for returns are normally distributed for all sectors of the two financial markets. Therefore,
a parametric serial correlation test is inappropriate and can be replaced with a nonparametric runs test
(Abraham et al., 2002).
The Runs test procedure tests whether the order of occurrence of two values of a variable is random. In
7
general, a run involves the sequencing of similar events separated by different events, such as increases in
returns separated by decreases. For instance, when returns increase (decrease) sequentially three times in a
row, the runs test treats them as +++ (- - -). Once this sequence is broken with a decrease or no change
in returns, then a new runs count begins again. A sample with too many or too few runs suggests that the
sample is not random. In fact, too few runs would suggest a time trend or a systematic arrangement due
to temporal dependence while too many runs would suggest cyclical or seasonal fluctuations or clustering.
The randomness of a particular series can therefore be assessed after an analysis of the distribution of the
duration of specific runs. Furthermore, this test is very useful as it does not require series to be normally
distributed.
Let n represent the number of observations, na and nb respectively represent observations above and
below the sample mean (or median), and r represent the observed number of runs. The expected number of
runs is represented by:
n + 2na nb
n
(9)
2na nb (2na nb − n) 1
]2
n2 (n − 1)
(10)
E(r) =
The standard error can therefore be written as:
σ(r) = [
The asymptotic (and approximately normal) Z-statistic can be written as follows:
Z(r) =
r − E(r)
σ(r)
(11)
The null hypothesis for this test is for temporal independence in the series (or weak-form efficiency).
The runs tests are completed with a mean and a median as a base. The mean is generally effective in
measuring the central tendency for symmetrical distributions but can be weak when outliers exist. Since the
series analyzed do not exhibit symmetry (e.g. normal), the median can represent a more effective measure
of central tendency especially when distributions are skewed. As evidenced in tables (9) and (10), the actual
number of runs r is substantially lower than the expected number of runs E(r) for all sectors of the two
financial markets. When the median is used as a base, the null hypothesis of weak-form efficiency is rejected
for all sectors of the two markets except the insurance sector of the ADSM. When the mean is used as a base,
the null hypothesis is rejected for all sectors of the two financial markets. These results are not consistent
8
with the findings of Moustafa (2004) and therefore the ADSM and DFM do not possess the characteristics
of weak-form efficiency.
4.4
Unit Root and Cointegration Tests
The relatively close proximity of the ADSM and the DFM coupled with the fact that both markets are
new and not well developed raises questions about whether the performance of one market may affect the
other. In fact, the existence of co-movements in geographically or economically related stock markets has
been evidenced in the London stock market and the New York Stock Exchange (King and Wadhwani, 1990)
and in specific geographic clusters of Europe, Asia, and the U.S. (Groenen and Franses, 2000; Heaney et al.,
2000). These market co-movements have been generally associated with several observable and unobservable
factors including increased market volatility (King and Wadhwani, 1990; Ramchand and Susmel, 1998),
consumer sentiment (King et al., 1994), equity risk premium (Ammer and Mei, 1996), and dividend yields
and short-term interest rates (Longin and Solnik, 1995).
As for most time series studies, stationarity plays an important role in data analysis. When series are not
stationary, cross estimations can result in spurious results erroneously establishing a meaningful statistical
relationship. Time trends make economic variables nonstationary. Graphs are important because they reveal
deterministic time trends but are not good at identifying stochastic time trends. This is a serious problem
because the correct way to remove the trend depends on the source. The two types of time trends are:
Purely deterministic time trends and Stochastic time trends (unit roots). To remove a purely deterministic
time trend (PDTT), the variable must be detrended, that is the variable must be regressed on time. To
remove a stochastic time trend (STT), the variable must be differenced. If the wrong method is used, serious
problems will arise. Detrending a unit root variable fails to make it stationary, and differencing a purely
deterministic trend makes it stationary, but introduces a noninvertible moving average into the variable.
When series are not stationary, they are integrated of order d and are denoted by I(d). This means that
series must be differenced d times to become stationary.5
5 Stationary
series are I(0).
9
4.4.1
Unit Root Tests
The time series literature has been dominated with unit-root tests such as the Dickey-Fuller (DF), Augmented Dickey-Fuller (ADF), Phillips-Perron (PP), and Kwiatkowski-Phillips-Schmidt-Shin (KPSS) tests,
respectively introduced by Dickey and Fuller (1981), Said and Dickey (1984), Phillips and Perron (1988),
and Kwiatkowski et al. (1992). Within the context of bivariate modeling, there has been a general consensus
that the ADF test is appropriate (Engle and Granger, 1987; Baghestani, 1991; Elliott, 2001). A unit-root
test is generally completed by estimating the following equation:
yt = αyt−1 + βxt + t
(12)
where xt is a vector of optional exogenous regressors (e.g. intercept, trend, seasonal variables) and t is
assumed to be white noise. If |α| ≥ 1, then y is not stationary. If |α| < 1, then y is stationary. The ADF
test is completed by estimating the following equation:
∆yt = γyt−1 + βxt +
p
X
δi ∆yt−i + t
(13)
i=1
where γ = α − 1 and the hypotheses tested are:
H0 : γ = 0 (STT)
Ha : γ < 0 (PDTT)
Equation (13) is first estimated with the inclusion of a linear trend and an intercept term. In all cases,
the trend is statistically insignificant and is dropped from subsequent estimations. This indicates that the
series are following a stochastic time trend. As reported in table (11), the series of all sectoral indices become
stationary after being differenced once. In fact, the ADF test statistics exceed the MacKinnon critical values
for all reasonable levels of significance. Once the series are differenced, the ADF test statistics exceed the
MacKinnon critical values, thus indicating that the series are I(1).
When two variables are I(1) and they are regressed against one another in a particular way, their
combination can be I(0). More specifically, assume that xt and yt are I(1) variables that are regressed
against one another as follows:
yt = α + βxt + t
(14)
For xt and yt to be cointegrated, the residuals t must be I(0), thus suggesting a long-run equilibrium
10
between xt and yt . This essentially implies that the relationship between xt and yt , as measured by t , does
not wander away from a long-run equilibrium. Therefore, when the residuals are not stationary then xt and
yt are not cointegrated. Furthermore, cointegration tests can only be completed on series that are integrated
of the same order. As evidenced previously, all sectoral indices and the general index are I(1) which makes
them good candidates for cointegration testing. This can be completed once residuals are verified to be I(0).
The residuals are obtained from estimating the levels of the sectoral indices across the two UAE financial
markets. As reported in table (12), the dependent variable originates from the regression in which the listed
variable is estimated against its corresponding variable in the other market. For instance, the ADSM index
for the banking sector is estimated against the DFM index of the same sector and residuals are generated
and tested for stationarity. The reported ADF test statistic of −2.12 is below the MacKinnon critical value
of −1.94, thus resulting in the rejection of the null hypothesis for nonstationary series at the 5% level.
Therefore, the residuals of the equation in which the index of the ADSM banking sector is tested against the
index of the DFM banking sector are I(0). This is also verified by estimating a similar equation with the
index of the DFM banking sector as a dependent variable. Results for the services sector and the general
index indicate that the residuals corresponding to their estimated equations are also I(0). However, this
is not necessarily true for the insurance sector in which the ADF test statistic is below the MacKinnon
critical value at the 10% significance level. Furthermore, when the index of the DFM insurance sector is the
dependent variable, the residuals become I(1), thus making cointegration testing unfeasible for this sector.
Based on the results reported in table (12), the following sectors are used for cointegration testing: Banking,
services, and the general index.6
4.4.2
Cointegration Tests
Cointegration tests have been completed using several methods, including error correction modeling and the
Johansen-Juselius (JJ) procedure. I use the latter procedure as it has been established to be superior to
other competing tests (Cheung and Lai, 1993; Gonzalo, 1994; Enders, 1995). The JJ procedure is completed
by estimating the following equation using a vector autoregressive procedure:
∆yt = γyt−1 +
p−1
X
δi ∆yt−i + t
i=1
6 The
insurance sector is excluded because of inconsistent results from the ADF residuals tests.
11
(15)
where yt and t are vectors and γ is a matrix of parameters. The rank of the matrix is tested and the
parameters are estimated via maximum likelihood. The following statistics are derived:
n
X
ln(1 − λ̂i )
(16)
LRm (r0 ) = −T ln(1 − λ̂r0 +1 )
(17)
LRt (r0 ) = −T
i=r0 +1
where LRt (r0 ) is the trace statistic, LRm (r0 ) is the eigen-max statistic, λ̂i denotes the estimated eigenvalues,
and T is the number of usable observations. The null hypothesis tested in LRt (r0 ) is for no cointegration.
In fact, for bivariate cointegration tests, up to two null hypotheses can be tested. If the null that r0 = 0 is
rejected then at least one cointegrating vector may exist and the second hypothesis that r0 ≤ 1 is subsequently
tested. If the latter is rejected then there may be two cointegrating vectors.
As reported in table (13), there is statistical evidence of at least one cointegrating vector in the banking
index, services index, and the general index. In fact, the trace and eigen-max statistics far exceed the reported
critical values when the null hypothesis is that r0 = 0. However, the null that r0 ≤ 1 is not rejected as shown
by critical values exceeding the trace and eigen-max statistics. This implies that there exists one long-run
equilibrium between the ADSM index and the DFM index. Therefore, when one of the two markets suffers
a short-term shock, convergence toward this equilibrium will be achieved through self-correcting internal
market forces. However, although the two markets are cointegrated, it is not clear which of the two markets
is causing this long-run equilibrium. In other words, which of these two markets is the driving force of this
long-run equilibrium? To answer this, it is important to determine the direction of any potential causation.
This can be done by completing a Granger Causality test as described in the next section.
4.5
Granger Causality Tests
Granger Causality (GC) tests attempt to identify whether fluctuations in a particular market affect another
market. They represent a crucial supplement to cointegration tests by determining the specific direction of
the causation flow. The GC test can be completed by running the following bi-variate regressions:
yt =
p
X
xt =
i=1
p
X
i=1
αi yt−i +
p
X
βi xt−i + t
(18)
αi xt−i +
i=1
p
X
βi yt−i + ut
(19)
i=1
12
where yt (xt ) is assumed to be a function of past values of itself and past and contemporaneous values of xt
(yt ) and where yt and xt represent indices for the common represented sectors of the ADSM and DFM. The
reported F-statistics are for the joint hypothesis that:
β1 = β2 = . . . = βi = 0
The null hypothesis is that x does not Granger-cause y in equation (18) and that y does not Granger-cause x
in equation (19). I test for Granger non-causality across the general index, banking, insurance, and services
sectors of the two UAE financial markets. In order to allow for common samples across the different sectors,
I use data ranging between September 30, 2001 and December 30, 2004 for the general indices and ranging
between September 28, 2003 and December 30, 2004 for the sectoral indices.7
As reported in table (14), for a lag value of i = 2, the null hypothesis that the ADSM index does not
Granger-cause the DFM index is rejected for the banking sector, services, and the general index and not
rejected for the insurance sector. This is consistent with the findings of the previous section in which the
insurance sectors of the two financial markets are not cointegrated. Moreover, the null hypothesis that the
DFM index does not Granger cause the ADSM index is not rejected for all sectoral indices and the general
index.
5
Summary and Conclusions
In order for financial markets to achieve their purpose, it is important that they demonstrate weak-form
efficiency. Using sectoral data for the two UAE financial markets, ADSM and DFM, I complete varianceratio tests to test for a random walk and runs tests to investigate whether weak-form efficiency exists in
these markets. The variance-ratio tests reject the random walk hypothesis in all sectors of the UAE financial
markets when assuming homoscedasticity and fail to reject it only for the banking sector of the DFM when
assuming heteroscedasticity. The runs tests yield statistical evidence of weak-form efficiency only in the
insurance sector of the ADSM.8 However, returns in all sectors appear to fit a mean-reverting process which
7 The industrial and hotel sectors of the ADSM and the investment sector of the DFM are excluded since they are not
represented in both markets.
8 Insurance firms represent about 22% of the total firms listed in the ADSM. The insurance sector in the UAE is the largest
in the GCC and is expected to double by 2010. It is free from government regulation and generates most of its revenues from
logistics due to high inflows and outflows of goods. This sector is highly liquid and, as of July 2005, has reached a capitalization
of about $3.5 billion.
13
may suggest high market volatility and a potential bubble.9 This is quite alarming as it may be indicative of
a bull market (Summers, 1986).10 The major concern with such market conditions is that if market prices
are persistently rising before dying out, then stock markets could suffer bubbles that can lead to heavy losses
or a market crash as a correction mechanism.11 The repercussions on economic activity can be disastrous
and long lasting.
In light of this evidence in support of weak-form inefficiency, it is important to examine whether the
performance of a market influences the other. Evidence of such a spillover effect would tend to raise further
concerns about the efficiency of the UAE financial markets. Cointegration tests reveal evidence of a long-run
equilibrium between the banking, services, and the general index across the two UAE markets. Using a
Granger Causality test, I find that the indices for the ADSM banking sector, services, and the general index
Granger cause their corresponding DFM indices whereas the insurance sector of the ADSM does not Granger
cause that of the DFM. As the ADSM appears to be the driving force of the long-run equilibrium that exists
across the two markets, it is important to note that weak-form inefficiency spills over while efficiency does
not. This is perhaps explained by the large disparities between the two markets.12 Further research should
analyze the specifics of this sector across the two markets.
The policy and economy implications of this inquiry are that the ADSM and DFM, as emerging markets,
must be closely monitored to achieve an optimal maturity level. Investors must be aware that, in inefficient
markets, heavy gains are just as likely as heavy losses. Furthermore, in anticipation of the dreaded bursting
of the bubble, the central bank should take a leading role in regulating abnormal financial activity.13 In the
meantime, an inefficient Bull market could suffer overinflated stock prices, speculation, and insider-trading,
all potentially intensified by herding behavior. Several measures can be taken to improve the efficiency of the
9 Mean reversion can suggest market corrections, corporate exploitation of competitive advantage, or corporate reaction to a
competitor’s competitive advantage (Haugen, 2001).
10 A Bull market refers to a condition in which price increases are persistent. A Bear market is a condition in which prices
are falling or are expected to fall.
11 In fact, the explosive increase in capitalization that the UAE financial markets are experiencing could be indicative of a
naturally maturing market but at “bubbly” proportions.
12 This is consistent with the literature discussed in the previous section in which observable and unobservable factors in a
particular market could spill over to other markets. Because the DFM insurance sector is relatively thinly traded while that
of the ADSM is very liquid, efficiency cannot be expected to “spill over” from the ADSM to the DFM. Rather, the absence of
causality may suggest that the markets are very distinct and thus, contrary to Moustafa (2004), must be analyzed separately.
13 In late August 2005, the DFM suffered a major setback as a result of bogus trading of the Dubai Islamic Bank (DIB)
shares. Two individuals allegedly initiated the trading of about $2.6 billion of DIB stocks on the same day, resulting in an
increase of 7.6% in DIB shares. As this represented about 87% of the entire market volume, the DFM and Emirates Security
and Commodities cancelled all DIB transactions for that day and referred the case to the local authorities (Source: Oxford
Business Group, http://www.oxfordbusinessgroup.com).
14
UAE financial markets, including (and not limited to) the imposition of complete transparency in corporate
financial reporting, the development of private and corporate accountability measures, and the adoption
and enforcement of international accounting standards and legislation consistent with international financial
standards.
References
[1] Abeysekera, S. P. (2001) Efficient Markets Hypothesis and the Emerging Capital Market in Sri Lanka:
Evidence from the Colombo Stock Exchange - A Note, Journal of Business Finance and Accounting,
28, 249-61.
[2] Abraham, A., Seyyed, F. J. and Alsakran, S. A. (2002) Testing The Random Walk Behavior and
Efficiency Of Gulf Stock Markets, The Financial Review, 37, 469-480.
[3] Al-Loughani, N. E. (1995) Random Walk in Thinly Traded Stock Markets: the Case of Kuwait, Arab
Journal of Administrative Science, 3, 198-209.
[4] Ammer, J. and Mei, J. (1996) Measuring International Economic Linkages with Stock Market Data,
Journal of Finance, 51, 1743-63.
[5] Baghestani, H. (1991) Cointegration Analysis of the Advertising-Sales Relationship, Journal of Industrial Economics, 39, 671-81.
[6] Beveridge, S. and Nelson, C. R. (1981) A New Approach to Decomposition of Economic Time Series
into Permanent and Transitory Components with Particular Attention to Measurement of the Business
Cycle, Journal of Monetary Economics, 7, 151-174.
[7] Butler, K. C. and Malaikah, S. J. (1992) Efficiency and Inefficiency in Thinly Traded Stock Markets:
Kuwait and Saudi Arabia, Journal of Banking and Finance, 16, 197-210.
[8] Cheung, Y. L. and Lai, K. S. (1993) Finite-Sample Sizes of Johansen’s Likelihood Ratio Tests for
Cointegration, Oxford Bulletin of Economics and Statistics, 55, 313-28.
[9] Dickey, D. A. and Fuller, W. A. (1981) Likelihood Ratio Statistics for Autoregressive Time Series with
a Unit Root, Econometrica, 49, 1057-72.
[10] Ebid, S. (1990) Characteristics and Behavior of UAE Stock Market, Journal of Economic and Administrative Sciences, 6, 19-61.
[11] Elliott, C. (2001) A Cointegration Analysis of Advertising and Sales Data, Review of Industrial Organization, 18, 417-26.
[12] Enders, W. (1995) Applied Econometric Time Series, New York: John Wiley & Sons.
15
[13] Engle, R. F. and Granger, C. W. J. (1987) Cointegration and Error Correction: Representation, Estimation, and Testing, Econometrica, 55, 251-76.
[14] Fama, E. (1970) Efficient Capital Markets: A Review of Theory and Empirical Work, Journal of Finance,
25, 383-417.
[15] Fama, E. and French, K. (1988) Permanent and Temporary Components of Stock Prices, Journal of
Political Economy, 96, 246-73.
[16] Grieb, T. and Reyes, M. G. (1999) Random Walk Tests for Latin American Equity Indices and Individual
Firms, Journal of Financial Research, 4, 371-83.
[17] Groenen, P. J. F. and Franses, P. H. (2000) Visualizing Time-varying Correlations across Stock markets,
Journal of Empirical Finance, 7, 155-72.
[18] Haugen, R. A. (2001) The Inefficient Stock Market: What Pays Off and Why, Prentice Hall, 2nd edition.
[19] Heaney, R., Hooper, V. and Jagietis, M. (2000) Regional Integration of National Stock Markets, Asian
Review of Accounting, 8, 55-65.
[20] King, M. A. and Wadhwani, S. (1990) Transmission of Volatility between Stock Markets, Review of
Financial Studies, 3, 5-33.
[21] King, M. A., Sentana, E. and Wadhwani, S. (1994) Volatility and Links Between National Stock Markets,
Econometrica, 62, 901-33.
[22] Kwiatkowski, D., Phillips, P. C. B., Schmidt, P. and Shin, Y. (1992) Testing the Null Hypothesis of
Stationarity Against the Alternative of a Unit Root, Journal of Econometrics, 54, 159-78.
[23] Lo, A. and MacKinlay, A. C. (1988) Stock market prices do not follow random walks: Evidence from a
simple specification test, Review of Financial Studies, 1, 41-66.
[24] Longin, F. and Solnik, B. (1995) Is the Correlation in International Equity Returns Constant: 19601990?, Journal of International Money and Finance, 14, 3-26.
[25] Lucas, R. E. (1978) Asset Prices in an Exchange Economy, Econometrica, 46, 1429-45.
[26] Moustafa, M. A. (2004) Testing the Weak-Form Efficiency of the United Arab Emirates Stock Market,
International Journal of Business, 9, 309-25.
[27] Ojah, K. and Karemera, D. (1999) Random Walks and Market Efficiency Tests of Latin American
Emerging Equity Markets: A Revisit, The Financial Review, 34, 57-72.
[28] Phillips, P. C. B. and Perron, P. (1988) Testing for Unit Roots in Time Series Regression, Biometrika,
75, 335-46.
[29] Poterba, J. M. and Summers, L. H. (1988) Mean-Reversion in Stock Prices: Evidence and Implications,
Journal of Financial Economics, 22, 27-59.
16
[30] Ramchand, L. and Susmel, R. (1998) Volatility and Cross Correlation across Major Stock markets,
Journal of Empirical Finance, 5, 397-416.
[31] Said, S. E. and Dickey, D. (1984) Testing for Unit Roots in Autoregressive Moving-Average Models with
Unknown Order, Biometrika, 71, 599-607.
[32] Summers, L. H. (1986) Does the Stock Market Rationally Reflect Fundamental Values? Journal of
Finance, 41, 591-601.
[33] Urrutia, J. L. (1995) Tests of Random Walk and Market Efficiency for Latin American Emerging
Markets, Journal of Financial Research, 18, 299-309.
17
Table 1: Companies Listed in the ADSM
Sector
Banking
Hotels
Industry
Insurance
Services
Symbol
ADCB
ADIB
BOS
CBI
CIB
FGB
FH
INVESTB
NBAD
NBQ
RAKBANK
SIB
UAB
UNB
ADNH
NCTH
BILDCO
FOODCO
ADSB
AGTHIA
FCI
GCEM
JULPHAR
RAKCC
RAKCEC
RAPCO
RAKWCT
SCIDC
QCEM
UCC
ABNIC
ADNIC
TKFL
ALAIN
AKIC
AWNIC
AMAN
EIC
RAKNIC
UIC
UNION
ADAVIATION
TAQA
ALDAR
ETISALAT
GMPC
NMDC
OILC
QTEL
SUDATEL
Company
Abu Dhabi Commercial Bank
Abu Dhabi Islamic Bank
Bank of Sharjah
Commercial Bank International
Commercial International Bank (Egypt)
First Gulf Bank
Finance House
Invest Bank
National Bank of Abu Dhabi
National Bank of Umm Al-Qaiwain
RAK Bank
Sharjah Islamic Bank
United Arab Bank
Union National Bank
Abu Dhabi National Hotels
National Co. for Tourism & Hotels
Abu Dhabi National for Building Materials
Abu Dhabi National Foodstuff
Abu Dhabi Ship Building Co.
Emirates Foodstuff & Mineral Water Company
Fujairah Cement Industries
Gulf Cement
Gulf Pharmaceutical Industries
Ras Al Khaimah Cement Co.
RAK Ceramics
RAK Poultry & Feeding
RAK White Cement Co.
Sharjah Cement & Industrial Development Co.
Umm Al-Qaiwain Cement Industries
Union Cement Co.
Al-Buhaira National Insurance Co.
Abu Dhabi National Insurance Co.
Abu Dhabi National Takaful Co.
Al-Ain Ahlia Insurance
Al-Khazna Insurance
Al-Wathba Insurance
Al-Dhafra Insurance
Emirates Insurance
RAK National Insurance Co.
United Insurance
Union Insurance
Abu Dhabi Aviation Co.
Abu Dhabi National Energy Co.
Al Dar Properties
Emirates Telecommunications
Gulf Medical Projects Co.
National Marine Dredging
Oasis International Leasing
Qatar Telecom (Qtel)
Sudan Telecommunications
Source: Abu Dhabi Securities Market (self-compiled).
18
Table 2: Number of Companies Listed in the ADSM
Sector
Banking
Hotels
Industry
Insurance
Services
Total
Growth Rate
2001
3
2
2
5
3
15
-
2002
5
2
7
5
4
23
53%
2003
6
2
9
6
4
27
17%
2004
7
2
10
8
4
31
15%
2005
14
2
14
11
9
50
61%
Table 3: Companies Listed in the DFM
Sector
Banking
Industry
Insurance
Investment
Services
Symbol
AEIBANK
CBD
DIB
EBI
EIB
MASQ
NBD
JEEMA
ALLIANCE
ARIG
ASCANA
DIN
DNIR
AMAN
IAIC
NGI
OIC
DI
GLOBAL
GGICO
IFA
AMLAK
ARMX
ARTC
EMAAR
TABREED
SHUAA
UPP
Company
Arab Emirates Investment Bank
Commercial Bank of Dubai
Dubai Islamic Bank
Emirates Bank International
Emirates Islamic Bank
MashreqBank
National Bank of Dubai
Jeema Mineral Water Company
Alliance Insurance Company
Arab Insurance Group
Arabian Scandinavian Insurance Company
Dubai Insurance
Dubai National Insurance & Reinsurance
Dubai Islamic Insurance & Reinsurance
Islamic Arab Insurance Company
National General Insurance
Oman Insurance Company (PSC)
Dubai Investments
Global Investment House
Gulf General Investment Company
International Financial Advisors
Amlak Finance
Arab International Logistics Company
Arab Technical Construction Company
EMAAR Properties
National Central Cooling Company
SHUAA Capital
Union Properties
Source: Dubai Financial Market.
19
Table 4: Number of Companies Listed in the DFM
Sector
Banking
Industry
Insurance
Investment
Services
Total
Growth Rate
2001
2
1
0
1
2
6
-
2002
4
1
0
2
2
9
50%
2003
4
1
1
2
2
10
11%
2004
5
1
2
4
4
16
60%
2005
7
1
9
4
7
28
75%
Table 5: ADSM Variance-Ratio Tests
Sector
Banking
Hotels
Industry
Insurance
Services
General Index
q
2
5
10
20
40
2
5
10
20
40
2
5
10
20
40
2
5
10
20
40
2
5
10
20
40
2
5
10
20
40
V R(q)
0.6368
0.2335
0.1203
0.0613
0.0321
0.5080
0.2206
0.0933
0.0508
0.0264
0.5915
0.2410
0.1207
0.0707
0.0319
0.4952
0.1993
0.1015
0.0488
0.0237
0.7468
0.2577
0.1351
0.0746
0.0359
0.7317
0.2528
0.1336
0.0741
0.0360
Z
−11.7122***
−11.2821***
−8.4025***
−6.0913***
−4.3558***
−15.8662***
−11.4725***
−8.6605***
−6.1590***
−4.3817***
−13.1723***
−11.1722***
−8.3987***
−6.0302***
−4.3568***
−16.2779***
−11.7860***
−8.5821***
−6.1725***
−4.3935***
−8.1662***
−10.9266***
−8.2612***
−6.0045***
−4.3386***
−8.6521***
−10.9990***
−8.2749***
−6.0082***
−4.3381***
Z*
−6.7729***
−6.8529***
−5.2359***
−4.0784***
−3.1945***
−6.2196***
−4.9705***
−3.9456***
−3.0306***
−2.3736**
−6.3317***
−6.4892***
−5.4994***
−4.4284**
−3.5457**
−6.6527***
−5.7450***
−4.6140***
−3.6737***
−2.8181***
−3.3921***
−4.6126***
−3.8482***
−3.2238***
−2.6121***
−3.5685***
−4.6698***
−3.8221***
−3.2079***
−2.6556***
Asterisks, ** and ***, denote respectively statistical significance at the 0.05 and 0.01 level.
20
Table 6: DFM Variance-Ratio Tests
Sector
Banking
Insurance
Investment
Services
General Index
q
2
5
10
20
40
2
5
10
20
40
2
5
10
20
40
2
5
10
20
40
2
5
10
20
40
V R(q)
0.3795
0.1560
0.0806
0.0266
0.0117
0.5572
0.2683
0.1471
0.0779
0.0411
0.7603
0.2518
0.1237
0.0711
0.0337
0.8577
0.2547
0.1500
0.0818
0.0404
0.5551
0.2186
0.1091
0.0556
0.0284
Z
−15.2875***
−9.4908***
−6.7085***
−4.8256***
−3.3979***
−10.9103***
−8.2287***
−6.2237***
−4.5713***
−3.2969***
−5.9060***
−8.4136***
−6.3944***
−4.6048***
−3.3221***
−3.5069***
−8.3815***
−6.2021***
−4.5516***
−3.2994***
−16.5938***
−13.3027***
−9.8409***
−7.0874***
−5.0569***
Z*
−1.5134
−1.2427
−1.1812
−1.1691
−1.1385
−4.9231***
−4.2621***
−3.8046***
−3.2127***
−2.5242***
−4.0341***
−5.8787***
−4.7608***
−3.6694***
−2.7715***
−2.1412**
−5.2782***
−4.1310***
−3.1935***
−2.4592***
−2.0501**
−2.2386**
−2.2395**
−2.2030**
−2.1406**
Asterisks, ** and ***, denote respectively statistical significance at the 0.05 and 0.01 level.
Table 7: ADSM Descriptive Statistics
Mean
Maximum
Minimum
Std. Dev.
Skewness
Kurtosis
Jarque-Bera
Probability
Observations
Banking
0.0016
0.0534
−0.0582
0.0097
0.1983
9.8676
2052.58
0.000
1041
Hotels
0.0008
0.1411
−0.0769
0.0138
1.8130
23.1527
18186.29
0.000
1041
Industry
0.0012
0.0594
−0.0650
0.0095
0.2493
11.3667
3047.11
0.000
1041
21
Insurance
0.0012
0.0522
−0.0417
0.0079
0.7448
11.7950
3451.40
0.000
1041
Services
0.0011
0.1457
−0.1041
0.0143
1.7388
24.0354
19717.60
0.000
1041
General Index
0.0014
0.1003
−0.0680
0.0095
1.2672
23.3786
18291.86
0.000
1041
Table 8: DFM Descriptive Statistics
Mean
Maximum
Minimum
Std. Dev.
Skewness
Kurtosis
Jarque-Bera
Probability
Observations
Banking
0.0030
0.5281
−0.3541
0.0292
6.6050
208.5039
1074295
0.000
608
Insurance
0.0043
0.1187
−0.1186
0.0234
1.1262
12.7705
2546.92
0.000
608
Investments
0.0050
0.1209
−0.1061
0.0276
0.7458
6.1210
303.13
0.000
608
Services
0.0052
0.1244
−0.1276
0.0268
0.7213
7.1142
481.54
0.000
608
General Index
0.0011
0.2419
−0.0814
0.0106
7.7760
193.4829
2118484
0.000
1392
Table 9: Runs Tests with the Mean as a Base
Sector
ADSM Banking
ADSM Hotels
ADSM Industry
ADSM Insurance
ADSM Services
ADSM General Index
DFM Banking
DFM Insurance
DFM Investments
DFM Services
DFM General Index
n
1041
1041
1041
1041
1041
1041
608
608
608
608
1392
na
451
255
329
369
384
453
237
129
250
233
596
nb
590
786
712
672
657
588
371
479
358
375
796
E(r)
512
386
451
477
486
513
290
204
295
288
683
r
412
355
311
448
401
414
228
156
233
198
546
σ(r)
15.836
11.925
13.939
14.757
15.014
15.853
11.719
8.229
11.929
11.645
18.262
Z(r)
−6.328***
−2.606***
−10.046***
−1.992**
−5.641***
−6.229***
−5.310***
−5.864***
−5.231***
−7.764***
−7.481***
Asterisks, *, ** and ***, denote respectively statistical significance at the 0.10, 0.05 and 0.01 level.
Table 10: Runs Tests with the Median as a Base
Sector
ADSM Banking
ADSM Hotels
ADSM Industry
ADSM Insurance
ADSM Services
ADSM General Index
DFM Banking
DFM Insurance
DFM Investments
DFM Services
DFM General Index
n
1041
1041
1041
1041
1041
1041
608
608
608
608
1392
na
521
829
744
718
557
521
304
459
304
304
696
nb
520
212
297
323
484
520
304
149
304
304
696
E(r)
521
339
426
447
519
521
305
226
305
305
697
r
442
321
310
427
441
428
267
175
243
210
564
σ(r)
16.124
10.454
13.148
13.800
16.045
16.124
12.318
9.110
12.318
12.318
18.648
Z(r)
−4.930***
−1.688*
−8.786***
−1.417
−4.857***
−5.798***
−3.084***
−5.594***
−5.033***
−7.711***
−7.132***
Asterisks, *, ** and ***, denote respectively statistical significance at the 0.10, 0.05 and 0.01 level.
22
Table 11: Augmented Dickey Fuller Tests of the Sectoral Indices
Dependent Variables
Banking
Insurance
Services
General Index
ADSM Levels
3.95
2.37
−2.35
5.04
ADSM Differenced
−13.80*
−9.68*
−8.57*
−4.78*
DFM Levels
3.85
0.07
4.17
4.78
DFM Differenced
−14.74*
−10.60*
−9.91*
−10.40*
The ADF test statistics are reported above. The MacKinnon critical values are −2.57, −2.86, and −3.44 respectively for a 10%,
5%, and 1% significance level. An * denotes the rejection of the null hypothesis of a unit root at a 5% level of significance. The
ADF tests are completed with a constant and a linear trend then with just a constant. In all cases, the trend was dropped due
to being statistically insignificant.
Table 12: Augmented Dickey Fuller Tests of the Residuals
Dependent Variable
ADSM Banking
DFM Banking
ADSM Insurance
DFM Insurance
ADSM Services
DFM Services
ADSM General Index
DFM General Index
Residual Levels
−2.12*
−2.51*
−1.64
−1.23
−2.79*
−2.44*
−2.41*
−2.59*
Differenced Residuals
−19.78*
−12.12*
The ADF test statistics are reported above. The MacKinnon critical values are −1.61, −1.94, and −2.57 respectively for a 10%,
5%, and 1% significance level. An * denotes the rejection of the null hypothesis of a unit root at a 5% level of significance. The
ADF tests are completed with a constant which was ultimately dropped due to being statistically insignificant.
Table 13: Johansen-Juselius Cointegration Tests
Sectors
Banking
Services
General Index
Null Hypothesis
r0 = 0
r0 ≤ 1
r0 = 0
r0 ≤ 1
r0 = 0
r0 ≤ 1
LRt (r0 )
37.56*
5.33
46.28*
3.32
75.70*
6.61
Critical Values
19.96
9.24
19.96
9.24
19.96
9.24
LRm (r0 )
32.23*
5.33
42.96*
3.32
69.09*
6.61
Critical Values
15.67
9.24
15.67
9.24
15.67
9.24
An * denotes the rejection of the null hypothesis of no cointegration at a 5% level of significance. The insurance sector is
dropped since the residuals from the regression of the levels are I(1)
23
Table 14: Granger Causality Tests
Null Hypothesis
DFM Banking does not Granger cause ADSM
ADSM Banking does not Granger cause DFM
DFM Insurance does not Granger cause ADSM
ADSM Insurance does not Granger cause DFM
DFM Services does not Granger cause ADSM
ADSM Services does not Granger cause DFM
DFM Index does not Granger cause ADSM
ADSM Index does not Granger cause DFM
Banking
Banking
Insurance
Insurance
Services
Services
Index
Index
Observations
393
393
393
393
393
393
963
963
An * denotes the rejection of the null hypothesis at a 0.01 significance level.
24
F-Statistic
1.84558
6.23949
1.88909
2.14711
0.82400
11.7037
2.07786
27.1986
Probability
0.15932
0.00215*
0.15260
0.11821
0.43944
1.2E − 05*
0.12576
3.2E − 12*