آیا بازارهای سهام GCC قابل پیش بینی هستند؟

Weak-form efficiency in the stock markets of the Gulf Cooperation Council is examined using daily, weekly, and monthly index data for the 10-year period 2000–2009. Various variance ratio test specifications with specific homo- and heteroscedasticity assumptions found evidence of nonlinear dependence for the daily data, supporting the evidence in favor of a rejection of the random walk. A correction procedure for thin and nonsynchronous trading was applied but failed to produce significantly different results. Following an ARCH based model building procedure, conditional heteroscedasticity models are applied to the log return series. Significant differences in forecasting performance cannot be detected. The random walk hypothesis is generally rejected for daily but differences appear across markets using weekly and monthly data. The increased involvement of foreign institutional investors may play a role in the increased serial correlation in stock returns in the most recent period.

GCC1 countries produce about 20% of all world oil, control 36% of world oil exports and possess 47% of all known oil reserves. Economic growth rates have been averaging almost 16% over the last 10 years, making the Gulf region one of the hottest emerging economies.2 In recent years countries in the Gulf Cooperation Council (GCC) region have dramatically grown their cross-border financial asset holdings from just under 10% of GCC Gross Domestic Product in 1980 to over 100% in 2007. The capitalization of its stock markets has increased almost tenfold, from $102bn in 2000 to $1.07tr in 2007 before contracting significantly in the wake of the global financial crisis.3 In spite of its growing importance, stock markets in the Middle East have been widely ignored by international investors until very recently, due to imposed restrictions on foreign stock ownership, the lack of common accounting standards and corporate transparency, or dismissed simply on the basis of economic and political uncertainty. As a consequence, and due to the difficulty of obtaining sufficient and reliable market data, researchers did not focus on the regional financial markets.4 Only recently has a set of research papers emerged that focus on various aspects of GCC stock market behavior, mainly its reaction to oil price shocks, the status of integration, and the level of efficiency. Given the economic importance of the region's energy sector, several authors analyzed the linkage between oil and stock prices. Arouri et al. (2011) find stock market sensitivity to changes in oil prices to differ across GCC countries, while Bashar (2006) determines the predictive power of oil prices on stock prices to be the highest for Saudi Arabia and Oman. The time varying nature of the relationship is emphasized by Fayyad and Daly (2011), who conclude that overall stock market sensitivity to oil price shocks has increased since the beginning of the global financial crisis. To the mean variance optimizing international investor, GCC stock markets are a welcome portfolio addition, given the low correlation with the US market (Bley and Chen, 2006). The authors, however, found an increase in the number of co-integrating vectors within the GCC stock markets region as an indication of ongoing attempts to harmonize market economies in preparation for an economic union and eventually the introduction of a single currency. Still, most of the GCC region's volatility persistence is domestic in nature (Rao, 2008). While signs of an economic decoupling of the GCC economies from the US economy exist (Genc et al., 2010) regional stock markets seem to become more synchronized with US stock markets as a result of the growing influx of non-GCC capital. In fact, Bley and Saad (2010) discover strong evidence of tax-selling induced spillover effects in regional stock market segments with high foreign ownership, a surprising finding in a tax-free environment. The region's stock markets are predominantly segmented and overly sensitive to regional political events (Arouri and Rault, 2010). And while economies differ in degree of energy dependency and efforts to diversify (Arouri and Lahiani, 2010), recently imposed capital requirements to fund budget deficits and boost economic activity in the wake of the global financial crisis have driven the Gulf country governments to embark on capital market liberalization, privatization, and broad-ranging structural reforms, allowing foreign investors greater access to their stock markets (Arouri and Nguyen, 2010). However, the stock markets of the Gulf Cooperation Council (GCC) differ in degree of openness to foreign ownership and are still dominated by local retail investors. Compared to developed markets, emerging stock markets, like those in the Gulf region, often experience thin trading and are easier manipulated by few large traders. Relatively weaker accounting standards and publication rules add to the hindrance of efficient information transmission through these markets. Whether the consequence is a higher level of return predictability has yet to be answered conclusively.5Butler and Malaikah (1992) study MENA market efficiency and find serial autocorrelation in Kuwait and reject the RWH for Saudi Arabia. The authors offer low trading volume as a possible explanation. El-Erian and Kumar (1995) find the Jordanian and the Turkish stock market to be inefficient, Buguk and Brorsen (2003), however, dispute these findings. Focusing on the GCC markets, Abraham et al. (2002) find evidence of a random walk (RW) in Bahrain, and Saudi Arabia but not in Kuwait. Al-Khazali et al. (2007) reject the RWH for eight MENA markets (Bahrain, Egypt, Jordan, Kuwait, Morocco, Oman, Saudi Arabia, and Tunisia) based on variance ratio tests on raw data, supporting similar findings by Rao and Shankaraiah, 2003, Sharma, 2005 and Elango and Hussein, 2008. Corrected for thin trading, however, all eight markets appear to be weak efficient. Asiri (2008) find the stock market of Bahrain to follow a random walk. Most recently, Al-Janabi et al. (2010) test for ARCH effects and find GCC stock markets to be informational efficient with regard to gold and oil price index changes. Yet again, Abdmoulah, 2010 paper against weak-form efficiency in all GCC stock markets except Saudi Arabia. This study intents to add to the inconclusive evidence reported thus far by examining weak-form efficiency in the stock markets of the (GCC) using daily, weekly, and monthly index data for the 10-year period 2000–2009. Any market that is found not to follow a random walk using a battery of autocorrelation and variance ratio tests, is subjected to an ARCH based model building procedure. Following an ARCH based model building procedure, conditional heteroscedasticity models GARCH, GARCH-M, TARCH, and PARCH are applied to the log return series. To incorporate the widely documented asymmetric price reaction of financial assets to new information especially in emerging markets, the fit of EGARCH and EGARCH-M models is also investigated. A dynamic forecasting approach, in which the testing sample is successively enlarged by the forecast values (Pesaran and Pesaran, 1997), is then taken using the best fitting model for each market. The forecasting performance of the AR, ARMA, and GARCH-type models is compared based on three symmetric error statistics and relative forecasting performance is determined. The main contribution of this study is found in the following five aspects: (1) Data frequency: by adding weekly and monthly data series to commonly analyzed daily returns, the potential effect of white noise (French and Roll, 1986) on weak-form efficiency is mitigated. (2) Data period: With a gradual market opening to foreign investors, non-GCC institutional share ownership has increased which may affect market efficiency, as suggested, e.g., by Badrinath et al. (1995). By breaking the 10-year data set into two 5-year subperiods, possible changes in market efficiency can be investigated. (3) Index methodology: by using SHUAA stock market indices for all six country markets, as opposed to local market indices, any potential confounding effects of differences in index methodology are avoided. (4) Robustness: as trading volume in the GCC stock markets is highly concentrated, a correction procedure for nonsynchronous trading (as suggested by Miller et al., 1994) is incorporated. (5) Model comparison: rankings of forecasts are sensitive to the choice of error statistics (Brailsford and Faff, 1996). Following a dynamic forecasting approach, model performance is determined based not on one but a set of error statistics.

This study's intent is to add to the inconclusive evidence collected on stock market efficiency in the Middle East. Weak-form efficiency in the stock markets of the Gulf Cooperation Council is examined using daily, weekly, and monthly index data for the 10-year period 2000–2009. The analysis is performed for the full 10-year and two 5-year subperiods. Using daily data, the hypothesis of a random walk is rejected for every single country market. On the basis of weekly and especially monthly data could traces of a random walk be found in Bahrain, Qatar, and Saudi Arabia in individual subperiods. Various variance ratio tests with specific homo- and heteroscedasticity assumptions found evidence of nonlinear dependence for the daily data, supporting the evidence in favor of a rejection of the random walk. A correction procedure for thin and nonsynchronous trading was applied but failed to produce significantly different results. Following an ARCH based model building procedure, conditional heteroscedasticity models GARCH, GARCH-M, TARCH, and PARCH are applied to the log return series of any market that is found not to follow a RW, using a battery of autocorrelation and variance ratio tests. To incorporate the widely documented asymmetric price reaction of financial assets to new information especially in emerging markets, the fit of EGARCH and EGARCH-M models is also investigated. The best fitting model for each market is then used to take a dynamic forecasting approach. The forecasting performance of the AR, ARMA, and GARCH-type models is compared based on three symmetric error statistics. Overall, the GARCH-type models seem to have provided the best fit across the GCC market indices. Only in the case of Saudi Arabia is the best forecasting performance generated by a simple AR(2) model. None of the analyzed models, however, seem to be producing superior forecasts, as the values of the error statistics are fairly close. After the first day, forecasts converge to zero. Applying a relative forecasting performance methodology cannot detect significant forecasting performance differences. The rejection of the RWH for daily market return series can be explained by the mean reverting tendency of stock market prices as indicated by VR test results and significant first and second order autocorrelations. To fund budget deficits and boost economic activity in the wake of the global financial crisis, Gulf economies have embarked on capital market liberalization, privatization, and broad-ranging structural reforms and have gradually become more accepting of foreign stock ownership. Non-GCC institutional investors assuming a more active role may have altered the behavioral characteristics of the retail investor-driven local markets. The well documented herding behavior of institutional investors is a potential explanation for the increased serial correlation in stock returns in the most recent period. This may provide an opportunity for traders to predict future prices and generate abnormal returns by applying mechanical trading rules. The more pronounced serial correlation in stock returns in the most recent period bears the question, however, whether local governments should intensify efforts to further financial market liberalization or be wary of potential side effects? Specifically, does opening the GCC financial markets to foreign investors negatively impact regional market efficiency?