Studies Revealed A Surprising Lack Of Controlling

Print   

02 Nov 2017

Disclaimer:
This essay has been written and submitted by students and is not an example of our work. Please click this link to view samples of our professional work witten by our professional essay writers. Any opinions, findings, conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of EssayCompany.

The study of precisely how efficient the market is and what its potential might be has long been a topic of interest and research. More than one mathematical study has been done on the study of the efficient market (Bachelier, 1990).It is of interest to those involved with the market to know what sort of information is accessible and what can be done with that information. Basically an efficient market is one where prices adjust quickly react to new information (Dimson and Mussavian, 2000) thereby reflecting all new information appropriately. In other words, if a good sales report is announced, stock prices should go up and if poor sales are reported, sales should go down.

Early studies revealed a surprising lack of controlling trends or statistical correlations of data as price-trend, time period which led to the belief that the market was indeed efficient (Lo, 2007, cited in Blume and Durlauf, 2007; Reilly and Brown, 2007). The Efficient Markets Hypothesis (EMH) is the theory that major financial markets are reactively efficient. It is of interest to note that while the larger markets are efficient, markets in developing countries with their cumbersome regulations, trading costs, small number of participants and excellence of information are considered to be less efficient. It has long been said that while efficient markets are not necessarily perfect, perfect markets are efficient (Dickinson and Muragu 1994).

The Efficient Markets Hypothesis states that in a truly informationally efficient market, price changes must be not be able to be forecasted if they totally incorporate every expectation of all market participants (Samuelson,1965). Because information is released randomly, prices must also be random. In this way, the hypothesis states that it is not possible to make use of information in predicting future price changes (Campbell et al. 1997, 20).

Fama (1970) reviewed and revised the hypothesis further. Now it has been formally refined to include the fact that a market can only be called efficient if the prices truly reflect all available information (Fama, 1970). Fama (1970) and Findlay and William (2000) determined that three conditions must exist: 1) no transaction costs 2) all relevant information is made available free of cost to market participants and 3) current security prices should reflect all available information.

Robert (1959) and Fama (1970) also introduced three commonly accepted levels of efficiency: weak, semi-strong and strong, all dependent upon the flow of information. A market is called weak if all information about past price movement is known and reflected in the current stock price. This information is known as historical information and should not be useful in predicting future changes of price nor is trend analysis possible. With weak form prices will follow a random walk.

In-between strong and weak market efficiency are instances when information is released privately just slightly before it is released publicly. This provides a slight advantage for some over all other normal investors. Neither technical analysis nor fundamental analysis is effective since they are based on public information. It also assumes that all information has already been reflected in stock prices. A semi-strong efficient market assumes that public information is reflected almost instantaneously and totally in the prices of stocks. It follows that since predictive analysis is not possible in a semi weak efficient market therefore higher profits are also not possible. If the market is semi-strong then it implies the weak form efficiency, the strong efficiency, in addition, implies semi-strong and weak efficiency. If the weak form of the Efficient Market Hypothesis can be rejected, then the semi-strong and strong form Efficient Market Hypothesis can be rejected (Campbell et al. 1997, 22; Fama 1970).

The last form of market efficiency is a strong form. This part of the hypothesis states that all information (public or private) is mirrored in a stock price. In other words, an investor cannot gain the advantage because all interested parties learn the news at the same time. Profits cannot be made that exceed normal returns no matter the amount of information or research provided. No forecast of future prices is possible.

The last level strong form which it is believed that all available information (including insider information) is reflected in stock prices. Prices quickly adjust and investors cannot earn excess returns (above competitive) using information. Forecasting is not possible.

Semi-strong form implies weak form, strong form implies semi-strong and weak form efficiencies. It follows that if the weak form is capable of being rejected then both the semi-strong and strong forms can also be rejected. (Campbell et al. 1997, 22; Fama 1970).

3.1.1 The forms of Efficient Market Hypothesis

Additional information on the different forms of the Efficient Market Hypothesis continue to add that any technical analysis of past price pattern to predict the future is not useful with the weak form since the information has already been incorporated into existing market prices (Haugen, 2001:575). "If the market is efficient, the function of the joint distribution of security prices today has already been incorporated in past price history (Malkiel), it is not possible to develop trading rules based on past prices that will allow anyone to beat the market". (Weston, 1993:338). The conclusion can be made that is no arbitrage in an efficient market because there is no gain from information of utilitarian use.

3.1.2 Weak-form

The weak form EMH asserts that all security market information which is reflected fully in the current stock prices includes the rates of return, historical price series, trading volume and other market-generated information, such as block trades, odd-lot transactions and transactions by exchange specialists. Based on the basic assumption that current market prices already reflect all previous return and other stock’s information, Fama (1970) suggests that there is no relationship between past and future rates of return and it should be independent. It means that the decisions on buying or selling a stock based on the past information and rates of return will bring less profit to investors. The study of Kendall (1953) states that the prices of stock and commodity follow a random walk which indicates no correlation between the change of prices at t and t+1. If the investors can predict price cycles, arbitrage will drive stock prices to their efficient values. Simply, they will invest followed basic trading rule which based on the true value of assets. That is buying the undervalued assets and selling overvalued assets (Malkier, 2003). Stock prices will be only influent by the arrival of new information which is defined as unpredictable; therefore, prices will follow random walk.

The weak form of EMH is believed to be more relevant in developed markets rather than in less developed and emerging stock markets (Mobarek and Keasey, 2000 and Gupta, 2006). Many researchers investigate the weak form of EMH in Asian countries and indicate the inefficiency in those markets such as the studies examining the EMH in India market (Gupta and Basu, 2007), In which, they find significant evidence of autocorrelation which rejects the weak form of EMH in India during the period from 1991 to 2006. The other authors also demonstrate the similar results in Middle East (Abraham et al., 2002), Africa (Smith et al., 2002), Czech Republic (Hajek, 2002), and Spain (Regulez and Zarraga, 2002).

3.1.3 Semi-Strong Form

Additional information on semi-strong form is the presumption that all publicly available information has been reflected in the stock prices. The difference between semi-strong and strong is the speed at which the prices adjust with respect to publicly available information. A test would be to analyze whether the changes happened over a period of a few days or immediately. Ergo, no published information helps in the selection of undervalued securities.

3.1.4 Strong Form

All information is reflected in stock prices, including private or inside information as well as public information (Hagen, 2001:575). Strong form tests are not concerned with the full reflection of information in the prices insofar as no individual can expect to receive higher trading profits than others simply due to additional informational access (Coperland, Weston, 1993: 332). Thus the theory dismisses the benefits of insider information in an efficient market. Once an analyst recognizes new information, it is quickly dispersed and almost immediately reflected in market prices. It is important to know if all available information is totally reflected in the stock prices because no one person can have an advantage due to personal access to some information. (Fama, 1969:388). In the strong form tests, there is only limited evidence against the hypothesis (Fama, 1969).

3.2 Empirical Evidence on Efficiency Market Hypothesis

Most of empirical researches on theory of efficient market hypothesis concern whether prices "fully effect" a particular subset of information (Fama 1970). Especially, the empirical studies have been divided in test of weak-form, semi-strong and strong form of market hypothesis.

3.2.1 Evidence from developed markets

EMH is tested in terms of empirical study, and has thus far produces the null hypothesis that there is no serial correlation. When stock returns are measured over periods of days or weeks, the general evidence against market efficiency is that there is a positive correlation in stock returns, specifically in the short run. There has been research recently on autocorrelation. This research was in stock returns. The results have shown mean reversion in the prices of stock. (Engel and Morris 1991) When long run horizons were studied, the results showed that there was found to be serial correlation that measured into the negative in the United States. (Fama and French 1988) Poterba and Summers (1986) have also found positive serial correlation at short horizons and negative serial correlation at long horizons, specifically in the U.S. and 17 other countries. Thus far, negative autocorrelation reflects predictability in the long horizon (Fama 1991), compared to the finding that positive autocorrelation infers predictability of returns in the short horizon.

Early empirical study and research of the EMH had been primarily based on serial correlation and runs tests. Modern tests of market efficiency have used variance ratio test. The break through variance ratio test has it origin in the pioneering works of Lo and MacKinlay (1988) and Cochrane (1988). Here, Lo and MacKinlay studied 1216 weekly observations. These observations were gleaned from the Center for Research in Security Prices (CRSP) daily returns from September 6, 1962 to December 26, 1985. Their conclusions are in opposition to the random walk hypothesis for the sample period (1216-week). Furthermore, their resulting conclusions were also negative for all sub-periods (608-week) for returns indexes and size-sorted portfolios. Conversely, the negative serial correlation that French and Fama (1988) found different results. There, they found that there were positive serial correlation for weekly and monthly holding-period returns. Hence, Fama and French (1988) reveal that long holding-period returns are substantially negatively serially correlated, indicating that 25 and 40 percent of the variation of longer-horizon return is indeed predictable from past returns. Lo and MacKinlay (1988) found that the resulting evidence was against the EMH in stock prices of small firms, but not necessarily for large firms

Lo and MacKinlay (1988) also believe that the rejection of random walk hypothesis cannot fully be explained. They contend that infrequent trading or time varying volatilities are not adequate explanation. Their conclusions are based on the behavior of small stocks. Conversely, these results of Fama and French (1988) are in direct opposition to Lo and MacKinlay (1988). Lo and MacKinlay assert that the rejection of random walk for weekly returns simply does not support a mean reverting model of asset prices. The variance ratio test used by Lee (1992), is used in the examination of whether stock returns of the U.S. and 10 industrialized countries: United Kingdom, Australia, Belgium, Canada, France, Italy, Japan, Netherlands, Switzerland,, and Germany follow a random walk process for the period 1967- 1988. His results found that the random walk model is still appropriate characterization of weekly return series of for majority of these countries.

Choudhry (1994) examines the stochastic structure of individual stock indexes in seven OECD countries: the United States, the United Kingdom, Canada, France, Germany, Japan and Italy. The Augmented Dickey-Fuller and KPSS unit root tests, and Johansen’s co-integration tests were applied to the log of monthly stock indexes from the period 1953 to 1989. His findings reveals that stock markets in seven OECD countries are efficient during the sample period. Their result from both unit root tests show that all seven series have a stochastic trend (unit root) and they are non-stationary in levels. The result of Johansen’s cointegration test shows that there is no evidence of finding that there is a stationary long-run relationship between the seven stock series. There is additional evidence of finding for an efficient market. Namely that there is a lack of long-run multivariate relationships.

Phillips-Peron (PP) employs the unit root and Johansen’s cointegration tests, Chan et al. (1997) tested for the weak-form and the cross-country market efficiency hypothesis of eighteen international stock markets. The markets included are Australia, Belgium, Canada, Denmark, Finland, France, Germany, India, Italy, Japan, Netherlands, Norway, Pakistan, Spain, Sweden, Switzerland, the United Kingdom, and the United States. The resulting data from these countries covers the period from January 1962 to December 1992, with 384 monthly observations for each of the stock series. These markets were analyzed individually and collectively in regions to test for the weak form efficiency. Chan et al. (1997) have come to the conclusions that all stock markets examined are individually weak form efficient and only a small number of stock markets show evidence of cointegration with others.

3.2.2 Evidence from Emerging Stock Markets

Both investors and researchers have noticed that there are new and potentially lucrative emerging stock markets. The sudden and growing interest in these emerging markets is predictable because during early nineties growth of emerging markets were impressive. In addition to its seemingly endless growth, emerging markets attracts their low correlation with primary developed stock markets. Additionally, the stock returns in numerous growing markets are clearly more predictable than developed stock markets because of the exhibiting systematic patterns. Harvey (1995) has studied a number of countries concerning the predictability of volatility and returns. The countries investigated and analyzed are six of the Latin American stock markets, eight of the Asian stock markets, three of the European markets and two African stock markets. The results pointed to a strong serial correlation in the stock returns; and this causes them to be more predictable. There has been a recent and growing liberalization in many developing countries. As a result, studies have focused on predictability of return behaviour and major of the studies on examining the validity of random walk hypothesis in the emerging stock markets.

Laurence (1986) uses price observations of the individual stock from the period 1973 to 1978 for both KLSE and the SES, and applies both the runs and autocorrelation test on the Kuala Lumpur Stock Exchange (KLSE) and the Stock Exchange of Singapore (SES). The results suggest that both markets are not weak form efficient. However, contrary Laurence's findings, Barnes (1986) finds KLSE to be weak form efficient. He used similar tests, and applied them to 30 companies and six sector indexes for the six years period ended 1980. Barnes (1986) determined that the results of both tests show that the KLSE show a high degree of efficiency in the weak-form.

Using monthly prices of individual companies for the period 1974 to 1978, Parkinson (1987) tested the the theory of the weak-form efficiency of the Nairobi Stock Exchange. The runs test clearly showed that of the 50 companies in NSE, 49 exhibited fewer numbers of the runs that expected. Consequently, the hypothesis of random walk does not pan out for these data. Dickinson and Muragu (1994) had similar findings concerning the Nairobi Stock Exchange using the autocorrelation and runs tests. The period of their data continues the work of Parkinson; namely from the period of 1979 through 1989. Their data includes weekly prices of the 30 traded stocks which yield the highest potential.

Unlike Parkinson (1978), Dickinson and Muragu (1994) found that the results support the weak-form of efficient market hypothesis in NSE. Urrutia (1995) uses both variance ratio of Lo and MacKinlay (1988) and the runs tests to research and investigate random walk for the four Latin American growing and emerging markets. He employs the monthly data of index prices in local currency from the

period December 1975 to March 1991 for Argentina, Brazil, Chile, and Mexico. The variance ratio test rejects the random walk hypothesis for all of the four markets. The runs test does not. Based on the runs test results, he found that the four Latin American emerging stock markets are weak form efficient.

Ojah and Karemera (1999) tested random walk for the same four Latin American markets as did Urrutia (1995). They used the single variance ratio of Lo and MacKinlay (1988), multiple variance ratio of Chow and Denning (1993), and runs tests to monthly national stock price indexes in U.S. dollar terms for the period December 1987 to May 1997. Under the variance ratio test, Brazil, Chile and Mexico showed no evidence of following a random walk. The only exception was Argentina. However, when employing the multiple variance ratios, all indications show that all the four markets follow a random walk. Conversely, the runs tests rejected the random walk hypothesis for Chile, but not for Argentina, Brazil or Mexico. Similar to Urrutia (1995), Ojah and Karemera (1999) came to the same conclusion that four Latin American emerging markets are weak-form efficient. Karemera et al. (1999) studied the random walk hypothesis for fifteen emerging stock markets using a series of statistical tests similar to those used by Ojah and Karemera (1999). Their data encompasses monthly national stock price indexes expressed in both local currency and the U.S. dollars for the period 1986 to 1997. They concluded that local currency-based data provide different results, when in comparison with return series expressed in U.S. dollars. With U.S. dollar based data, results of 10 of the 15 emerging stock markets they examined are consistent with the random walk hypothesis under the multiple variance ratios, but 5 of the 15 are consistent with the random walk hypothesis when tested by the single variance ratio.

With local currency-based data, results of 10 (Indonesia, Israel, Jordan, Korea, Malaysia, Mexico, Philippines, Taiwan, Thailand, and Turkey) of the 15 markets follow a random walk under the multiple variance ratios. Only six of the 15 follow, namely (Israel, Jordan, Malaysia, Mexico, and Taiwan) found a random walk under the single variance ratio. The results on Argentina, Brazil, Hong Kong, Indonesia, Mexico, Philippines, Singapore, Taiwan, and Turkey equity returns, however, are not consistent under at least two different currency-based data. Their results of runs test show that the hypothesis of independence cannot be rejected at 5% level of significance for nine of the 15. Thus far, six markets including Chile, Israel, Philippines, Singapore, Taiwan, and Thailand are found not to be weak form efficient based on U.S. dollar data. Hence, based on this, their results support the evidence provide by Urrutia (1995) who finds Argentina, Brazil and Mexico to be weakly efficient. With local currency-based data, 12 of the 15 emerging markets are weak form efficient, only Argentina, Chile and Singapore are found not to be weak-form efficient.

There was a considerable question concerning whether or not from the time period of 1967 to 1993, Chang et al. (1996) tested the weak form of the EMH using monthly data on the Taiwan stock exchange. (Israel, Jordan, Malaysia, Mexico, and Taiwan). When using and applying the Ljung-Box Q, the runs and the unit root tests, they observe that the Taiwan stock market is weak-form efficient. Using the variance ratio test, Chang and Ting (2000) similarly examined the validity of weak form efficiency of the Taiwan stock market for the period 1971-1996 and their findings confirmed those of Chang et al. (1996). Chang and Ting (2000) employed the weekly, monthly, quarterly and yearly returns of the value-weighted stock price index. Their results are in direct conflict with the random walk hypothesis with weekly returns, but not with monthly, quarterly and yearly value-weighted market indexes.

Antoniou et al. (1997) gathered the data from the daily stock prices of the ISE Composite Index for the period 1988 to 1993 to examine the weak form efficiency for the Istanbul Stock Exchange (ISE). They observed that thin trading may lead to serial correlation in the return series, Antoniou (1997) carry out the analysis for both unadjusted and adjusted for thinness returns using a method proposed by Miller et al. (1994). Generally speaking, thin or infrequent trading occurs when stock do not trade at every consecutive interval. The Miller at al. (1994) model suggests that returns should be adjusted in accordance with the results found in the removal of the impact of thin trading, and a moving average model (MA) that reflects the number of non trading days. Despite the improvement with adjusted returns, they nonetheless found serial dependence in returns. The resulting conclusion points to the ISE as being weakly inefficient.

Tas and Dursonoglu (2005) have only recently confirmed the inefficiency result for Turkey using daily stock returns of ISE 30 indexes from the period 1995 to 2004. Their research employed the Dickey-Fuller unit root and runs tests and the results of both tests reject the random walk hypothesis in ISE. In the Middle East, Butler and Malaikah (1992) examined weak-form efficiency for the Kuwait and Saudi Arabian stock markets. This was done by using the autocorrelation test. Their data covers daily stock returns of two stock markets for the time period including 1985 to 1989. They found that that the Kuvait stock market was efficient, but the Saudi Arabian market was not.

Similarly, Abraham et al. (2002) researched and studied weak-form efficiency in three major Gulf stock markets including Kuwait, Saudi Arabia, and Bahrain. They used the variance ratio and runs tests for the period October 1992 to December 1998. Their data consisted of weekly index values for each of three Gulf stock markets. The random walk hypothesis in all markets was rejected. When taking into consideration all possible infrequent trading in all three markets, they apply a correction to the observed index by a using decomposition of index returns. This decomposition of index returns was first introduced by Beverigde and Nelson (1981). After the correction, they did not reject the random walk hypothesis for the Saudi Arabia and Bahrain markets, but not for the Kuwait market.

Hassan et al. (2003) used a similar method used by Antoniou et al. (1997), and observed that the Kuwait stock market (KSE) is weak-form inefficient. Taking into consideration possible thin trading and non-linearity characterize the Kuwait markets, they used the method proposed by Miller et al. (1994) to correct for possible thin trading; namely the logistic map model to account for possible non-linearity in the generating process of return. They also employ GARCH-M and EGARCH models to examine whether the pattern predictability is present where a measure of time varying risk parameter is included in the model. Their data included a number of daily stock price indexes for the time period of 1995 to 2000. The null hypothesis of market efficiency for the whole sample period is not supported by their results. They explain that the possible reasons for inefficiency is primarily due to the thin trading in the most of the stocks in Kuwait Stock Exchange. They also suggest that the reasons for inefficiency is due to the fact that their study covers the aftermath of various important regulatory reforms carried out in the KSE.

Moustafa (2004) examines the behavior of stock prices in the United Arab Emirates (UAE) stock market using daily prices of 43 stocks included in the UAE market index for the period from October 2, 2001 to September 1, 2003. His findings reveal that the returns of the 43 stocks do not follow normal distribution. However, the results of runs tests show that the returns of 40 stocks out of the 43 are random, and at a 5% level of significance. Despite the fact that the UAE stock market is newly developed and it is still very small, and suffering from infrequent trading, his results find for the UAE being weak-form efficient.

Eleven African stock markets including Botswana, Egypt, Ghana, Ivory Coast, Kenya, Mauritius, Morocco, Nigeria, South Africa, Swaziland, and Zimbabwe were tested out for weak form efficiently by accounting for thin trading in the calculation of returns, and allowing for non-linearity and time-varying volatility in the return generation process. They relied on weekly data of index prices in the local currency for the period 1989-1995, and applied the Miller et al. (1994) model, a logistic map and EGARCH-M model to test the overall efficiency of all of the eleven markets. Their results indicate that with the exception of the markets in Egypt, Kenya, Mauritius, Morocco, and Zimbabwe, the remaining six markets were found not to be consistent with weak form efficiency. Additionally, they found that the return generation process is nonlinear in all of the above listed eleven markets. In five of the markets, investors demanded a time-varying risk premium for the risks they bore. Contrary to past studies, the found that the Nigerian market to be not efficiently weak form. Nonetheless, their modeling approach produces a significant time-varying risk premium for the Nigerian markets that linear models would not have been able to capture. They argue that efficiency test models that do not control for time-varying risk premiums. They also believe that it is very likely that inappropriate models are being employed.

Recently, however, Akinkugbe (2005) found stock markets in Botswana to be weak and semi-strong form efficient. His results were based on information and data that includes 738 weekly observations for the period June 1989 to December 2003. Autocorrelation, and Augmented Dickey-Fuller and Phillip-Perron unit root tests were used to investigate and determine the weak form of EMH in the Botswana stock exchange. In his study, autocorrelation test reveals evidence of no serial correlation. The results of both unit root tests indicate a stationary process for stock returns. This implies weak form efficiency.

Poshakwale (1996) examined the weak form efficiency as well as the daily of the week effect on the Bombay Stock Exchange in India using daily BSE national data for the period January 1987 to October 1994. He found that the frequency distribution of the prices in BSE does not necessarily follow a normal distribution pattern. Additionally, his results of runs and serial correlation tests also provide evidence of non-random behavior of stock prices in BSE. Poshakwale (1996) similarly found evidence that the average returns are different on each day of the week. The result show that the returns achieved on Friday are significantly higher in comparison to rest of the days of the week. As a result of his findings, he concludes that the Indian stock market is not weak-form efficient.

There have been a number of other studies concentrated in the emerging European markets. For example, Gilmore and McManus (2003) examined whether the stock markets in Central European countries including Czech Republic, Hungary and Poland are efficiently weak form. They used a variety of tests including uni-variate methods (unit root, variance ratio, and autocorrelation), multivariate tests (Johansen and Granger causality) and model-comparison approach (Naïve, ARIMA and GARCH). More specifically, they relied on data from weekly Investable and Comprehensive indexes from the International Financial Corporation (IFC) for the period July 1995 through September 2000. Gilmore and McManus (2003) revealed that results of the ADF and PP unit root tests evidenced that all series are integrated of order I(1). The Ljung-Box Qstatistics show that returns are more significant for the Comprehensive series than for the Investable. They contend that this might be derived from the various differences in the behavior of internationally versus domestically traded stocks. Additionally, the Q-statistics similarly show that over time all three markets are moving in the direction of lower levels of autocorrelations in returns. This indicates that there is efficiency improvement in these markets. The variance ratios under the assumption of heteroscedasticity do not necessarily reject random walk hypothesis for either index of the three markets. Their the multivariate tests, however, show mixed evidence, with the Johansen cointegration test indicating the absence of a cointegration relationship between these markets. At the same time, Granger-causality were found to be running from the Czech and Hungarian market to the Polish exchange. They believe that the differences in privatizations methods and economic environments in the three countries could ultimately explain the lack of cointegration during the period. Furthermore, the Granger-causality may be due to the higher levels of foreign investment in the Czech and the Hungarian markets. This would then influence the Polish market. In contrast with the univariate method findings, the model comparison approach is found to provide strong evidence against the random walk hypothesis for these markets. Based on these findings, they conclude that these three markets are not yet weak form efficient.

The random walk behavior in five European emerging markets were investigated by Smith and Ryoo (2003) using variance ratio tests. They used weekly information and data of index prices in local currency for the period April 1991 to August 1998. They found that in four of the markets, Greece, Hungary, Poland and Portugal, the random walk hypothesis is rejected because returns have autocorrelated errors. The positive autocorrelation is found be in four of the markets. In Turkey, the Istanbul stock market is found to follow a random walk. They theory is that the Istanbul stock market is larger and more liquid compared with the other four markets. Evidence from other studies, however, which use variance ratio tests, suggests that a comparatively large size on its own is neither necessary nor sufficient for a market to follow a random walk. Smaller markets like Argentina (Urrutia 1995; Ojah and Karemera 1999) and Indonesia (Huang 1995), and large markets, like Hong Kong, Korea (Huang 1995) and Mexico (Urrutia 1995) do not.

From September 1995 through May 2001, Abrosimova et al. (2005) tested for weak-form efficiency in the Russian stock market. They used daily, weekly, and monthly Russian Trading System (RTS) indexes. They employed Unit root, autocorrelation and variance ratio tests to test the null hypothesis of the random walk. They also used the model-comparison approach in their study. The RTS index series difference are found to be stationary with the ADF and the PP unit root tests. The conclusions drawn from the results of both the autocorrelation and the variance ratio tests reject the null hypothesis of the random walk for the daily and weekly, but not for the monthly data. For monthly data, there remains no rejection of the null hypothesis of random walk of the variance ratio under the assumption of heteroscedasticity increments. They, therefor, study the linear and non-linear dependence in the daily and weekly data using the ARIMA and GARCH models. Their findings revealed that none of the analyzed models outperformed others. The conclusion: there is evidence that supports weak-form efficiency in the Russian stock market.

A test of efficiency in seven European emerging stock markets was employed by Hassan et al. (2006) using the International Finance Corporation’s weekly stock index data for the period December 1988 through August 2002. A number of different methods were used in their studies, including Ljung-Box Q-statistic, runs, and variance ratio tests. According to their results, the markets in Czech Republic, Hungary, Poland and Russia were found to be unpredictable, but there was no such finding in Greece, Slovakia, and Turkey.

Taking into account recent studies that have found that developed markets are completely consistent with weak-form efficiency, the conclusion can still be drawn, mainly based on empirical studies of developed markets, that support the random walk hypothesis, and support the finding of weak form efficiency. This conclusion can not be drawn, however, for emerging stock markets. There is conflicting data regarding emerging markets and their relationship to random walks. There is also a number of conflicting literary sources on emerging stock markets. This is expected because emerging markets are obviously less efficient than are developed markets because of their infancy in the market, and the lack of research for investors. Additionally, the developed markets have well-established institutions and are characterized as having high levels of liquidity and trading activity, substantial market depth and low information asymmetry, amongst other things. The emerging market are observed to exhibit more information asymmetry, thin trading and shallow depth because of their weak institutional infrastructure. (Khaled and Islam 2005) Despite the fact that emerging markets are characterized by the aforementioned imperfections, many emerging markets are becoming increasingly more efficient. Some researchers are optimistic, and have found some of the larger and even smaller stock markets in developing countries to be weak-from efficient.

3.3 Random walk

Explanation of the terms "random walk" dates back to a mathematician, Karl Pearson and has grown to include molecule movement, foraging animals and stock prices. Markov chains utilized the terms to describe statistical models of change. While fluctuating stock prices may not truly be random, the term is used to explain the observed behaviors of stocks. When applied to stocks, it says that the future behavior cannot be accurately predicted by the past behavior. Malkiel, 1996:197). Random walk covers the conundrum that unless the flow of information is stopped, then tomorrow’s prices will only reflect tomorrow’s news. Since this doesn’t appear to be totally true in the market, random walk incorporates the fact that today’s prices may be dependent upon previous prices and yet the returns are independent of each other.

Random walk basically evens out the market and follows the Capital Assets Pricing Model (CAPM) which says that the market is in equilibrium; i.e. it is not possible to gain better performance or larger returns for a stock’s portfolio. However, that being said, the possibility of gaining higher returns comes from the upper end of the beta (possibly riskier portfolios) during a market down turn (Malkiel, 1996:254).

Investors continually look for accurate predictions of future stock prices in order to buy or sell at the most advantageous times. This futile search goes against both technical (stock charts) and fundamental (intrinsic value of stocks) analysis of the market. While chartists deal in 10 percent logic and 90 percent psychological reasoning, fundamental analysts see the market in reverse (Malkiel, 2006:117). Neither viewpoint can overcome the random walk theory that describes the market with respect to predictions of future stock prices. Seasoned investors do not want to believe that new investors will do as well as they based on the random walk theory and prefer their own version non-random walk (against all research).

3.4 Technical trading rules

Technical analysis supposed that past trends in market can be used to predict the future stock prices; therefore, technical analysis is the process of analyzing past prices and other statistics to attempt the future price. Traders use these technical studies to establish target point for buy and sell financial assets, whether to open or close trading positions. (Elder, 2002)

Assumptions that stock prices tend to move in trends for a long period and repeat itself is the key point of technical analysis. Hence, technical analysis also known as chartist, they look at charting and analyze the information as trend , historical high and low prices, a reasonable forecast can be made.

Edwards and Magee (1966), technical analysis assumptions follow by:

The interaction of demand and supply determine market price of securities.

Both rational and irrational governed demand and supply of securities.

Shift in demand and supply are the result of reversal of trends

Securities tend to move persist in trend for a long time period, even there are minor fluctuations in the market.

In charts, can detect any change in demand or supply of securities.

Use of charts and key indicator to project future market movements are essential for developing tools of technical analysis.

EMH and Technical analysis are opposed each other, according to EMH, the market is always efficient, impossible to consistently beat the market because prices reflect all available information, but Technical analysis’ assumption is traders can predict future trend based on past prices. Thus , the acceptance and validity of one is meaning the rejection of other. There are many researches which test efficiency by using Technical trading rules. Trading rule test determine whether technical analysis can used buy and hold strategy. If trading rule test can help investors earn abnormal profit of those earned buy and hold investment, market is inefficiency and vice versa.

Chapter 4 DATA AND RESEARCH METHODOLOGY

4.1 Data

Data will be collected from HoChiMinh Stock Exchange (HOSE), daily, weekly return will be computed. Daily returns will be used to test predictability of stock prices while weekly return to test random walk hypothesis. Raw data i got from FPT Securities Company, all data in period of 2006-2010. Weekly data collected from Jan 2006-March 2010 which subcategorized pre-crisis period of Jan 2006-Oct 2008 and crisis period of Oct 2008-March 2010. These returns are analyzed based on the random walk theory and tested using autocorrelation function (ACF) to investigate the returns predictability.

In order to test using technical data rules, daily return are required. Daily returns calculated in natural logarithm of index for two trading days. As a requirement of technical trading rule test, a series of non-overlapping 10 day returns will be computed Brock et al. (1992). The multi-day returns allow comparing short-run and long-run effects. Variance ratios have been used in market efficiency tests (Poterba and Summers 1988; Lo and MacKinlay 1988). Ten days returns (Rt) are cumulative rates of return calculate for 10 days holding period:

Rt = ln(pt)-ln(pt-9)

The weekly and daily returns used in this study do not include dividend yield, because the dividend payments in listing companies of HOSE are paid semi-annual. Hence, extrapolation of dividend yield to daily and weekly frequencies induces a smoothing term into stock return, so increase their persistent and introduce a bias in variance ratio test toward non-rejection of null hypothesis of random walk.

4.2 Research approach

4.2.1. Deductive and Inductive

In business researches, building the research approaches following logical and clear ways is one of the basic steps to ensure the researches’ structure, and then achieve the researches’ purposes. The two main approaches used in business researches are known as deductive and inductive approaches. According to Burns and Burns (2008), inductive approach bases on the process from specific descriptions or observations to develop theory. In contrast, deductive reasoning will follow the test of a theory or hypothesis; and consequently, based on the basis of results, the researchers can make conclusions that will support or reject that initial theory or hypothesis. In the other words, inductive approach is a bottom-up process with an open-end attribute. The nature of inductive reasoning is open-ended and exploratory, especially at the beginning of the approach (Saunders et al., 2009). Specially, in inductive research, there is no fixed structures construct in advance and assumptions about the relation of data are made without any previous observation; conversely, deductive research has a fixed framework built on the basis of mathematical analysis and logical explanation.

Based on these natures of the two approaches and the attribute of this piece of research, this dissertation will follows a deductive research methodology in mind. That is because this research will test the month of the year effect through a set of data from the two stock markets, namely FTSE 350 index and VN Index. Therefore, testing theories by using statistical descriptions and mathematical technique is obviously an example of deductive research.

4.2.2. Positivism and Interpretivism

As a matter of fact, this research is also performed under the viewpoint of positivism. As defined by Coates and Sloan (2008), positive research is predominant in the field of science due to its independence with human unpredictable behaviours, which enhances the reliability of the research. Hence, beside quantitative methods, positive research methods, with the emphasis on quantifiable data, becomes more and more popular.

As opposed to positivism, interpretivism holds different viewpoint. According to interpretivists, natural and social sciences are dramatically distinguished, which then implies that in social science, it is hardly to observe data to approve or negate theories; rather than, the data need to be interpreted in order to achieve sound understanding of the results (Bryman, 2001)

In the view of interpretivists, scientific positive methods in doing research are oversimplified; therefore, there is not enough data to draw thorough understanding about the complexity of human behaviours. Taking these two methodologies into consideration, the author decide to use mathematic and logic method on the basis of positivism to justify social theories.

4.2.3. Quantitative and Qualitative research

Referred to the requirements and needs of the researcher, quantitative and qualitative research methodologies have both advantages and disadvantages due to different techniques used in data collection and data analysis in each method. Quantitative approach emphasizes on the development and collection of a set of valid data, the significance of which is then examined by a variety of descriptive statistic analysis techniques (Cooper and Schindler, 2006; Robson, 2002; Shields and Twycross, 2003). In the contrary, the focus of qualitative approach, which includes phenomenology or ethnography strategies, is on the use of quasi-statistical data collection techniques, template data analysis techniques, editing approaches and immersion approach, which are distinguished with quantitative analytical tools (Cooper and Schindler, 2006; Robson, 2002; Shields and Twycross, 2003)

As differentiated by Coates and Sloan (2008, pp. 192), quality, "the what", is considered as the core feature of something whereas quantity, "the how-much", is only the amount; therefore, qualitative methodology aims at profound understanding of human behaviours and their causes while quantitative methodology focuses on the measurement in order to justify hypotheses about human behaviours. Being properly used, quantitative strategy enables the researcher to obtain results of high validity consisting of internal validity, content validity, criterion-related validity and construct validity which are important considerations that researchers need to take in designing and implementing the study (Cooper and Schindler, 2006, pp. 318-321). In addition, as added by Cooper and Schindler (2006), reliability, stability, equivalence, internal consistency, practicality, economy, convenience, and interpretability are also needed to be satisfied in the study.

According to Firestone (1987, pp. 16), in the positive conception, quantitative research is assumed to be able to explain behaviours via objective facts. As referred to the positive philosophy and principles of deductive research methodology, the current study is believed to be most properly executed in the light of quantitative methodology. Furthermore, the collected data is the quantified and numeric data of the value of the Vietnam stock market. Lastly, this research aims at testing different theories on a set of sample in order to generate a conclusion in larger scope, which employs statistical experimental investigation, a property of quantitative methodology.

4.3 Methodology

In this part, there are two approaches of random walk hypothesis will be employed to test randomness: tests for stationary or non-stationary, and tests for serial correlation of time series.

4.3.1 Tests of Randomness

Two significant approaches appear throughout the writings concerning the random walk hypothesis. These two approaches have been employed to test for randomness. First of all are the tests for stationarity of nonstationarity. Second are the tests for serial correlation of the time series. In the first approach, A stationary or nonstationary process is characterized by a time series. Specficially, a stochastic process can be stationary when its mean, variance and autocovariances are constant over time, at various lags. In other words if they are time invariant.

However, when a time series is not stationary, it is referred to as a nonstationary time series, having a time varying mean, variance, or possibly both (Gujarati,2003). The difference between stationary and nonstationary possesses important implications as to whether the trend of a time series is stochastic or deterministic. Specifically a deterministic trend in a time series is predictable and not variable, it is then said to be a stationary time series. However, if the trend is not predictable, it is then stochastic and the time series is nonstationary. In essence then, if a series of price changes in stock is random, it would be characterized by a process that is nonstationary. Another aspect is the independent structure of a time series is somewhat related to randomness as well, it means if price changes after that are uncorrelated to those of the past, then stock return series is subsequently random. If any relationship exists, or a correlation exists in stock prices changes in succession, the return series will be dependent or sometimes said to be serially correlated. Therefore, serial correlation is the tendency of stock returns to be related in certain parameters to past returns.

The tests for stationarity or nonstationarity and for serial correlation speak to totally different aspects of randomness. Though the tests are designed to detect patterns of trends in a time series, the tests for serial correlation are supposed to detect the correlating relationship among and between the different observations in a series. Just because there is a stationarity in a series does not mean there are correlations, but rather the deterministic pattern of the trend.

Considering the short time span of this study, various aspects of randomness probably should be tested to allow a reliable and comprehensive conclusion regarding the conformity of the Vietnamese stock market compared to the random walk hypothesis. Therefore, I will use the following tests of randomness of the series of returns by the week to look at the results of stock prices in the stock market in three different tests: 1) Portmanteau 2) Unit Root; and 3) Lo and MacKinlay’s variance ratio.

Explanation of methods chosen

Per Campbell, Lo and MacKinlay (1997), arguably the most direct and intuitive test assessing the random walk hypothesis for an individual time series is one that determines serial correlation—the correlation of two observations on different dates, but of the same series. Taking that into account, the Portmanteau, or autocorrelation test, will be done first. The most commonly used measure of serial dependence in time series analysis is the autocorrelation. If all autocorrelations are at zero, then a time series will be considered to be random. To examine the dependence structure of the data to be researched then, a test statistic developed by Ljung and Box (1978) to detect departure from zero autocorrelations. But, note that the Ljung-Box test statistic is robust only to linear dependence. If the data, therefore, is characterized by non-linear dependence, other tests robust to such as non-linear dependence, for example the BDS tests (Brock, Dechert and Sheinkman, 1987) is preferable to be used along with it.

To rephrase, when the Ljung-Box test denies the hypothesis of no autocorrelations contained in the data then the BDS test is not conducted either, because of the presence of linear dependence. If Ljung-Box does not reject the null hypothesis, then the BDS is conducted to test for non-linear dependence. Using it this way, methodology of the BDS test will be presented later in this paper.

Subsequent to checking for the presence of serial dependence, unit root tests would then be performed on the data. Additionally, statistical tests for unit root might be of interest to help evaluate if the return series is nonstationarity or stationarity. Specifically, such tests are supposed to reveal if the trend is toward stochastic (nonstationary) because of the presence of a unit root, or possibly deterministic (stationary) because of the presence of a polynomial time trend (Phillips and Perron, 1988).

The unit root tests are somewhat similar to the portmanteau tests because that stationarity may be an explanation for the dependence structure of the time series. In that way, unit root hypothesis is then tested using the Augmented Dickey-Fuller (ADF) test (Said ad Dickey, 1984) along with the Phillips-Perron test (Phillips and Perron, 1988). The rationality of utilizing both of the above test statistics is in the robustness of the different tests in different conditions. Particularly, the Phillips-Perron test is more advantageously used for models with positive moving average errors, when the ADF test works more powerfully for models with moving average errors and negative serial correlation.

The variance ratio tests as put forth by Lo and MacKinlay (1988) have been used and have been very productive and robust for looking at the behavior of stock price indices where the returns are many times not normally distributed. Considering the fact that the series of weekly return reported in this study is not conformative to normality, the Lo and MacKinlay’s variance ratio test (LOMAC) should be used. It had been well documented in literature that most of stocks returns are heteroskedastic in a conditional way regarding time. In this way the LOMAC test if better by far than many others such as autocorreltions and unit root tests. Mainly, the LOMAC test statistics are robust with both homoskedasticity and heteroskedasticity. Therefore the LOMAC being used means a single variance ratio, along with the other two tests, yields a much more comprehensive and reliable result concerning the degree of randomness of stock returns in the stock market.

Autocorrelations Tests

Ljung-Box test

One way to test the random walk hypothesis is to test all at the same time the autocorrelations are zero. Specifically stock prices will "follow a random walk" if the returns are not correlated at all leads and lags. The formula to measure dependence or independence of different observations in a time series is:

ScreenHunter_571

Where ρ (k) is the kth order autocorrelation, rt is the return at time t, k is the time lag.

The autocorrelation function is simply the expected value of a product. Assume we have a pair of random variables from the same process,X1=X(t1) and X2=X(t2), then the autocorrelation is often written as

ScreenHunter_610

Above equation is valid for stationary and nonstationary random process. For stationary process , we can modify this equation wider , it can be proven that expected values from our random process will be independent of origin of our time function. So autocorrelation will depend on the time difference and not absolute time.(CNX, 2005) We will let  τ=t2−t1 so we have modified equation:

ScreenHunter_611

The discrete time case of autocorrelation function:

ScreenHunter_612

For a large sample, Ljung-Box statistics follows the chi-square distribution with m degree of freedom:

LB=n(n+2)

In this research, i calculated rt as natural logarithm weekly returns, to test hypothesis that all autocorrelation coefficient are zero, i have to applied Ljung-Box Q test. Developed by Box and Pierce (1970) , Ljung-Box modified and proposed Q(m) statistic:

ScreenHunter_572

M: is the number of lags being tested

T: Sample size

Ljung-Box Q(m) statistic used as a test statistic for null hypothesis

Ho = ρ (1) = ρ (2) = ….= ρ (m) = 0

Against the alternative hypothesis:

H1: ρ(k)≠0 for some k {1,…,m}

Decision rule rejects Ho if Q(m)>, where refers to 100(1-α)th percentile of a chi-squared distribution with m degree of freedom. The selection of m may effect to the performance of Q(m) statistic. The presence of higher order autocorrelation may be missed if too few are used, and the test may not have much power if too many are used, due to significant higher order autocorrelation (Campbell et al. 1997). To solve this problem, there’s a suggested that the choice of m ≈ ln(T) will provide the better power performance Tsay (2005), so this research will follow these rules.

BDS Test

Brock, Dechart and Scheinkman (1987) suggests the test statistic that provides a test of long run non-linear dependence within a time series primarily to be based on correlation dimension. The BDS test is an alternate portmanteau test for time based series dependence.

This alternate test is dependent on a null hypothesis of independent and identical distribution (IID) and is often utilized to look at non-linear dependence.

{xt} be a time series of length T whose observations may embedded in m-dimensional space by named m-history.

BDS test derived from correlation integral C m(ε):

ScreenHunter_581

(T-m+1) is the maximum number of overlapping vectors that can be formed with a time series length T, and I ε is an indicator function that equal 1 if or else equal 0 , being sub norm.

BDS statistic given by:

ScreenHunter_582

Under the assumption of independence, this statistic is expected be close to zero

Unit root test

Augmented Dickey-Fuller (ADF) Test

The regular ADF test is used to look carefully at the nature of stationarity or nonstationarity of the researched data. The unit root test, as put forth by Said and Dickey (1984) is a type of the Dickey-Fuller (1979) test (DF test) but rather for a bigger and more complex set of time series models. Particularly, using the DF test, the error term is assumed to be uncorrelated. But it is in fact, probably the case that the error is indeed correlated. So, the ADF test was supposed to account for correlated error term by the addition of the lagged values of the dependent variable ΔYt - The difference in values.

ADF based on formula:

ScreenHunter_585 (1) With constant, no trend

And

ScreenHunter_586(2) With constant, with trend

Δ : the first different , yi is natural logarithm of price index, εt is a pure white noise error term , T is the deterministic time trend, k is lag length which selected using Akaike Information Criterion (AIC). (1) equation examine for null hypothesis of unit root against mean stationary alternative in yi,. (2) equation examine the null hypothesis of unit root against trend stationary alternative. Null hypothesis H0 is β equal 0, it point out that the series is nonstationary or unit root exist in time series. Alternative hypothesis (H1) is that β<0 , ADF test , τ statistics, larger in absolute value than critical value, then null hypothesis of nonstationary will be rejected. Τ statistic is a negative number , if it more negative, rejection is stronger of hypothesis that there is a unit root at some level of confidence. If failing to reject H0, it means that time series has property of random walk.

Phillips-Perron (PP) test

Phillips and Perron (1988) in testing for unit roots had a proposal for another nonparametric statistical method to account for the serial correlation in the error terms without the addition of lagged difference terms. Their approach was to calculate the above unit root tests from regression equations with k=o. The resulting statistics are then changed to remove the serial correlation effects on the asymptotic distribution of the test statistic. Using this method, Phillips and Perron provide new statistics. They definitely prove that these statistics have not only identical asymptotic power as the regular ADF test, but in addition allow a general class of error processes.

Using the Phillips-Perron (PP) test, the null, the critical values and the alternative hypotheses are identical to those in the ADF test. The PP statistic is attained using Eviews 5.0. The null hypothesis of nonstationarity will not be accepted if the PP statistic is more in absolute value than the critical values.

Variance Ratio test

Lo and MacKinlay (1988) proposed the variance ratio test based on the idea that under the random walk hypothesis, the different increment in asset price series are normally serially uncorrelated and that the changes in random walk increments in any finite sample in the sampling interval is linear. To state it another way, if the natural logarithm of a time series is a random walk, the variance of its q-differences grows in relation to the difference q. For instance, with weekly data, the variance of the weekely series should be five times the variance of the daily series if the random walk is truly the process generating the stock price series.

Therefore, if we have a series of (nq + 1) observations including X0, X1, X2,..., Xnq

at equally spaced intervals, the variance of (Xt – Xt-q) should be q times the variance of (Xt

– Xt-1) according to the random walk hypothesis. Lo and MacKinlay (1988) gives one test

for this hypothesis using the single variance ratio, denoted by VR(q).

The variance ratio of q observations, VR(q) is defined simply as

VR(q)

Where (q) is 1/q times the variance of q-differences and (1) is the variance of the first differences. Lo and MacKinlay variance ratio test null hypothesis that variance ratio VR(q) equals 1. Besides, values for VR(q) that are more than 1 imply positive serial correlations while values that are less than 1 imply negative serial correlations or mean reversion.

Values of (1) & (q) are given by Lo and McKinlay as follows:

ScreenHunter_214

Where

And

Under the assumption of homoscedasticity and heteroscedasticity increments, two standard normal test-statistics, Z(q) and Z*(q) respectively, developed by Lo and MacKinlay (1988), are calculated by Equation:

ScreenHunter_215

Where is the asymptotic variance of variance ratio under the assumption of homoscedasticity and is the asymptotic variance of the variance ratio under assumption of heteroscedasticity

ScreenHunter_216

Where is the heteroscedasticity – consistent estimator and computed as follow:

ScreenHunter_217

Using the Monte-Carlo simulations in further research, Lo and MacKinlay (1989) have demonstrated that the asymptotic distribution of Z*(q) performs well in finite sample, and the variance ratio test is better performing than either the Ljung-Box test of serial correlation or the ADF test of unit roots. In order to empirically look at the random walk hypothesis in the stock market the above battery of tests will be applied on the weekly series of the V-index’s returns.

4.3.2 Tests of Technical analysis

There are various trading rules that need to be studied. The effectiveness of academic researches depends on the implementation of moving averages, filters, momentum, residence rules and support. For this dissertation the moving average oscillator will be chosen as the method for analysis (Brock et al, 1992). The rational for this selection is that this technique is consistent. The versions for this analysis technique include the Variable Length Moving Average (VMA) and the Fixed Length Moving Average (FMA). A moving average simply refers to the recursive but dated average of past prices. The moving averages yield insight into the underlying trend of the series in addition to smoothening of the volatility. The moving averages under consideration are those of the short run and long run price indices. A comparison of the short run and long run moving averages, there is the generation of the buy and sell signals. The decision rule is that a buy decision is reached when the short run average is above the long run average while a sell decision is if the short run average is lower that the long run average (Brock, et al, 1992, p 1735). The short run moving average of an order with n observations is given by

The long run average is given by



rev

Our Service Portfolio

jb

Want To Place An Order Quickly?

Then shoot us a message on Whatsapp, WeChat or Gmail. We are available 24/7 to assist you.

whatsapp

Do not panic, you are at the right place

jb

Visit Our essay writting help page to get all the details and guidence on availing our assiatance service.

Get 20% Discount, Now
£19 £14/ Per Page
14 days delivery time

Our writting assistance service is undoubtedly one of the most affordable writting assistance services and we have highly qualified professionls to help you with your work. So what are you waiting for, click below to order now.

Get An Instant Quote

ORDER TODAY!

Our experts are ready to assist you, call us to get a free quote or order now to get succeed in your academics writing.

Get a Free Quote Order Now