Efficient Market Hypothesis (EMH) and China's Economy

According to the academic work of Fama (1970), efficient market hypothesis (EMH) becomes one of the most significant study topics in the finance and economic field step by step. It is widely acknowledged that there are three degrees of market efficiency, which are weak-form efficiency, semi-strong efficiency and strong efficiency. From the aspect of theoretical foundation, different types of market efficiency have different criterion to evaluate. For this paper, weak-form EMH indicates that share price is expected to reflect all the information of past price and trading volume, which means stock price tends to follow a random walk. On the other hand, random walk hypothesis is another important notion which defines that share price could not be predicted by modeling past data and consequently abnormal returns could not be generated through technical trading.

On the basis of above theory, a great number of empirical researches have been carried out to understand the efficiency of different capital markets, mainly from aspects of developed markets and emerging markets. However, the findings of existing papers as to the efficiency of relevant target markets are mixed and inconsistent, particularly for the emerging markets. Therefore, this paper chooses Chinese capital market as the study target to investigate the market efficiency.

With conspicuous development of Chinese economy, substantial investors gradually show their great interest in Chinese capital market, and subsequently a lot of studies have been conducted to understand the market nature. However, Chinese capital market owes exclusive characteristics which are essential to be introduced firstly. Followed by the policy of market-oriented development in 1990’s, two security markets, namely Shanghai Stock Exchange and Shenzhen Stock Exchange have been set up in succession. Simultaneously, two types of shares are issued and traded, which are A-share for domestic investors and B-share for international investors. Nevertheless, after 2001 B-share market has been liberalized for domestic investors to participate in. Because Chinese government hopes domestic capital market to be more liberal, pushing forward economic reformation in a desirable pace in turn. Under this scenario, it is worthy evaluating the influence of market-oriented policy on the Chinese market efficiency.

Admittedly, some scholars have already concentrated on the market efficiency issue of China and provided quite of lot findings, such as Long et al. (1999), Ma and Barnes (2001), Seddighi and Nian (2004),Niblock and Sloan (2007) etc. However, currently the existing findings are relatively confusing and there still exist certain limitations in terms of the methodology applied which might exert inevitable influence on the outcomes. First, some of the studies just focus on the time period in the early of 1990s, but Chinese capital market has developed significantly in the latest decade, not only the regulation system, but also the market nature. In addition, Chinese capital market is relatively complicated including Shanghai and Shenzhen Stock Exchange as well as A-share and B share indices, so chosen indices in test is the key to the conclusion. Hence, in order to conduct a comparably comprehensive market efficiency research for China, this paper will cover a relatively longer time serial and incorporate all available indices of China.

The core objective of this paper is to investigate whether the Chinese capital market is efficient in the weak-form degree with the observations including indices of both Shanghai and Shenzhen Stock Exchanges. Besides that, the market behavior of both the two stock exchanges will be examined, analyzed and compared. In details, this paper will concentrate on three main questions: 1) whether random walk hypothesis is valid in Shanghai and Shenzhen Stock Exchanges; 2) whether Chinese capital market becomes efficient gradually in recent years; 3) whether calendar effect presents in Chinese capital market.

This paper choose the testing sample of stock price with weekly frequency covering 17 years from 1992 to 2009 involving both Shanghai and Shenzhen Stock Exchanges. In order to get relatively comprehensive outcomes, the methodology employed here consists of serial correlation coefficient test, unit root test and runs test which aim to investigate the validity of random walk hypothesis in Chinese capital market. At the same time, the developing tendency of Chinese market efficiency, as well as potential seasonal anomalies, could be detected and analyzed from the outcomes of variety tests.

The rest of the paper is organized as follows: A brief literature review of previous theoretical and empirical studies is presented in Chapter 2 from the perspectives of developed market, emerging market as well as Chinese capital market. The methodology employed in this paper is introduced in Chapter 3 involving data collection, research hypothesis and model specification. Based on the empirical tests, results are laid out and analyzed in Chapter 4. Eventually, Chapter 5 draws a conclusion of this study, and limitations and recommendations are stated for further studies.

Chapter 2 Literature Review

Based on the seminal work of Fama (1970), efficient market hypothesis has been acknowledged and investigated by substantial scholars, mainly conducted for the purpose of three different forms of EMH tests, namely weak-form, semi-strong and strong EMH. Theoretically, weak-form EMH, the core topic of this paper, maintains that the current security prices reflect all available information of past prices and volume, which implies that any investment strategy derived from analysis of historical price is unprofitable. According to this foundation, lots of empirical studies have been carried out focused on different markets and time periods.

Previously, a large group of studies have found evidence for the random walk hypothesis of stock returns, among which serial correlation and runs test are frequently resorted to. Theoretically, for serial-correlation test, if the autocorrelation of the chosen price serial was zero, the stock returns are supposed to follow a random walk, based on which Jaffe (1974) reported positive evidence for weak-form EMH of the US security market. However, a positive autocorrelation has been exhibited during the testing period which is considered to be insignificant. For runs test, a plenty of research conclusions indicate that random walk behavior of stock returns in the prior studies.

Contrast to previous findings, more recent studies provide opposite conclusions which implies that abnormal profit could be generated by analyzing historical stock prices to predict current prices. For instance, investors’ overreaction (Debondt et al. 1979), momentum trading (Jegadeesh and Titman 1993) and calendar effects (Ariel 1990) all conflict with the validity of weak-form EMH.

2.1 Researches for Developed Markets

During the early stage, a large proportion of EMH studies are focused on developed markets and evidence of stock returns’ random nature have been consistently documented for major researches, especially taking the transaction cost into consideration (Kendal 1953, Cootner 1962). Admittedly, some predictable components of stock returns have been detected particularly during recent years’ studies, while there is no consensus regarding the arbitrage strategy for abnormal profit.

A group of researchers report positive evidence for the weak-form EMH. Lee (1992) picks up 10 equity markets of develop countries, such as Australia, Canada and Belgium, to test whether the stock prices support random walk hypothesis during 1976 to 1988. The findings of this paper imply that these ten developed markets are weak-form efficient. Along with this conclusion, Chan et al. (1997) resort to both unit root and cointegration tests to discover the nature of market efficiency from 1961 to 1992 of 18 developed security markets, including German, France, Demark and so on, through which the validity of weak-form EMH has been confirmed greatly for each of the international equity market. Furthermore, by analyzing the results of cointegration tests, another important conclusion has been reached that international diversification could be manipulative and beneficial for investors because of the collectively inefficiency of stock prices among different stock markets. More recently, in 2004, the market efficiency of Hong Kong and Singapore stock market has been examined by Lima and Tabak through variance ratio and multiple variance ratio tests, the results of which suggest that Hong Kong stock market is weak-form efficient while the same evidence does not exist in Singapore equity market. Moreover, the potential reason for this situation is related to the diversified capitalization and liquidity of these two markets.

Nevertheless, some scholars provide negative evidence to against the random walk hypothesis for developed market. Early in 1988, Lo and MacKinlay have detected the positive serial correlation by using variance ratio tests to investigate both weekly and monthly stock returns in US market, while Fama and Frence report a significant negative correlation for US market in the same year. To same extent, the findings of the two papers violate the random walk hypothesis, particularly for the stock price of relatively small companies. Consistently, though modeling the daily statistics of FTSE 30 index from 30/06/1983 to 16/11/1989, Al-Loughani and Chanppel (1997) discovered the non-random nature of daily stock returns for FTSE 30 index.

In sum, conclusions of most empirical studies focused on develop countries are in favor of random walk hypothesis, and the weak-form EMH exists at least.

2.2 Researches for Emerging Markets

With the conspicuous growth of emerging market, substantial academicians have paid increasing attentions to the movement of stock price in developing equity markets. As to the market nature, Havey (1993) states that the index of emerging markets are correlated to the developed countries in a really low degree and the stock returns tends to be more predictable. In order to understand the truth, a plenty of studies have been conducted from the perspective of different markets.

In line with the statement of Havey (1993), some of the findings in terms of different emerging markets demonstrate predictable factors of the stock price which conflict with the weak-form EMH. Cheung (1993) conducts a study to test the stock price behaviour of Korean and Taiwanese equity markets by applying the Capital Asset Pricing Model. However, the results reject the weak-form EMH in both of the two security markets. Similarly, in 1996 Poshakwale investigate the stock prices’ performance of Bombay Stock Exchange by running serial correlation coefficient test and non-parametic KS test. One of the significant findings of this paper is that the distribution of stock price is not expected to be normal one. One the other hand, the outcomes of correlation coefficient and runs tests indicate share prices do not follow random walk. Therefore, Poshakwale concludes that the market of Bombay Stock Exchange is not efficient in the weak-form degree. As to the Latin American stock markets, including Argentina, Brazil, Chile and Mexico, Claessens, Dasgupta and Glen (1993) have carried out variance ratio and runs tests to investigate the market efficiency. Based on monthly data from 1975 to 1991, the results of variance ratio test and runs test are contrary with each other. Specifically, supporting evidence could be generated by runs test while the outcomes of variance ratio test imply the non-random nature of stock prices of emerging markets in Latin America. More recently, Mobarek and Keasey (2002) choose share daily index during the period from 1988 to 1997 to discover the efficiency of Dhaka Stock Market. This paper concludes that stock prices of Dhaka Stock Market demonstrate a violation of random walk hypothesis, the possible causes of which might be associated with the less developed regulation and communication system.

Nevertheless, some of other studies provide evidence that the researching target markets are efficient at least on the weak-form level by applying testing methods and data different from above empirical studies. Such as Ojah and Karemera (1999), in order to develop the study of Claessens et al. (1993), they apply additional multiple variance ratio test, the results of which support weak-form EMH of Latin American markets as a whole. Additionally, contrast to the study of Mobarek and Keasey (2002), previous paper by Alam et al (1999) indicates that Dhaka Stock Market is generally efficient on the weak-form sense by testing the monthly data during the time period from 1986 to 1995. To the same market, Khaled and Islam (2005) develop the study methodology and take daily, weekly and monthly statistics simultaneously. Finally, several significant findings have been concluded which are different from previous studies. Firstly, the random walk hypothesis just has been supported by monthly data, while the Dhaka Stock Market was efficient on the weak form level only before 07/1996 by considering daily and weekly data. What is more, another reason for the findings which violate the random walk hypothesis by Mobarek and Keasey (2002) has been point out by Khaled and Islam, because their testing outcomes are exclusively on the basis of Box-Pierce Q test, which do not take the heteroskedastic errors into account.

Apart from that, another group of investigations put their weight on the emerging markets of European countries. For example, Hall and Urga (2002) carry out their research of weak-form EMH for Russian Stock Exchange from 1995 to 2001. The outcomes of majority testing methods, including unit root test, variance ratio test and autocorrelation test, indicate the stock prices of this market follow the random walk in general, particularly the daily statistics. Meanwhile, Cilmore and McManus (2003) look into the stock markets of Poland, Hungary and Czech Republic. They find that, in general, these three stock markets are weak-form efficient by comparing the results from serial correlation test, variance ratio test, univariate test as well as multivariate test.

By analyzing the existing findings of empirical studies as to the emerging markets, relatively complicated results could be concluded, especial compared with the developed markets. To some extent, the degree of market efficiency of emerging markets are comparably lower than that of develop markets, which probably associated with the immature regulation system and thin trading feature. To the mixed study conclusions, a possible cause has been suggested by Buguk and Brorsen (2003) that is related to the less developed methodologies applied by some of researchers such as serial-correlation test and runs test. Therefore, for the emerging markets, ongoing study of EMH is desirable and meaningful.

2.3 Researches for Chinese Capital Market

Nowadays, with the prominent development of China, increasing attentions has been paid to Chinese economy, especially Chinese security market, among which the efficiency of Chinese market is a significant study perspective.

In 1997, as a pioneer Liu et al. conduct a research to investigate the efficiency of Chinese capital market, including both Shanghai and Shenzhen Stock Exchanges, by choosing the observations sample from 05/1992 to 12/1995. The results of both causality and cointegration methods reveal that Shanghai and Shenzhen Stock Exchanges are efficient on the weak-form level separately rather than collectively. Similarly, the weak-form EMH has been verified by Long et al. (1999) which using a sample including 99 weekly returns of shares in Shanghai Stock Exchange. This paper eventually suggests Shanghai Stock Exchange is even more efficient than the US market, but further studies are necessary to test this hypothesis.

Compared to the above studies, more recent papers have presented contrary findings that Chinese capital market is not weak-form efficient and calendar effect is apparently observed. In 2000, the daily stock prices of both Shanghai and Shenzhen markets have been modeled by Darrat and Zhong to understand the markets’ efficiency. By employing variance ratio test, they discover that the share prices present non-random characteristics, meaning the weak-form inefficiency particularly for A-share index. Moreover, potential reasons have been suggested like thin trading, limited stock supplying and so on. One year later, Ma and Barnes collect daily, weekly and monthly stock price statistics from both Shanghai and Shenzhen Stock Exchanges over 1990-1998 to test the efficiency of Chinese equity markets by applying runs, serial-correlation as well as variance testing methods. This paper detects three important evidences. First of all, stock returns of B-share index could be predicted more easily than A-share’s. Secondly, the efficiency level of Chinese market is relatively lower than that of developed market at the proportion of 36.36% which could be explained by the gap of the information-spreading speeds between Chinese market and developed market. Lastly, Ma and Barnes point out that the inefficiency of Chinese market is possibly resulted by the non-transparent information available in the market. The same finding has been proved by Seddighi and Nian (2004) for Shanghai Stock Exchange, which provide evidence of the non-random feature of the share prices of eight listed corporations. By incorporating 955 daily stock price data over the period of 05/2002-10/2005, Niblock and Sloan (2007) suggest the Chinese capital market is not weak-form efficient, but it presents a tendency to become more and more efficient later. What is more, the integration between of the Chinese market with the developed markets tends to be increasing intensive and extensive.

In addition, the market behavior of both Shanghai and Shenzhen Stock Exchanges has been observed intensively from the perspective of calendar effect. Take the study conducted by Gao and Kling (2004) as an example, they focus on the daily and monthly stock returns of Shanghai and Shenzhen markets to detect the existence of weekend and January effects. Anomalies have been identified in this paper with certain different characteristics from other markets. More specifically, stock returns tend to be relatively higher on Friday which might be explained by the short-term trading behavior of most Chinese amateur investors. Apart from that, compared to phenomenon of January effect, generally remarkable higher returns could be expected in March or April because of the Chinese Festival in February.

From the above literature review of previous empirical studies, the following features of Chinese equity market could be summarized. First, although there is no consensus on the issue that whether Chinese market is weak-form efficient, majority studies report inefficient evidence and calendar effect could be detected noticeably with exclusive characteristics. Secondly, from the study results on hand, stock returns of B-share index could be modeled more exactly than that of A-share index, which means A-share market is more efficient on the weak sense in corresponding. Eventually, the inefficient factor of Chinese market could be associated with some less developed aspects compared to other industrialized markets, such as immature regulation system, non-transparent information communication and so no.

Chapter 3 Methodology

3.1 Sample Selection and Data Collection

The data is collected from Datastream covering the time span from 21/02/1992 to 30/06/2009, consisting of 5364 closing prices in weekly frequency [1] and belonging to six branch indices from Shanghai and Shenzhen markets, as demonstrated in the following table.

The involving indices are considered to be the most representative and reliable market indices in general academic study of Chinese economy. Primarily, it is essential to state that, prices in SHSE-C and SZSE-C indices are the weighted average one of all the stocks in Shanghai and Shenzhen markets respectively.

As shown in Table 1, observations of SHSE-A, SHSE-B, SHSE-C and SZSE indices are available from 21/02/1992 to 30/06/2009, while share closing prices of SZSE-A and SZSE-B are collected from 05/10/1992 to 30/06/2009, mainly caused by restriction of statistics collection. Besides that, there are two significant reasons for the starting year 1992’s choosing. On one hand, China Securities Regulation Commission, the main supervision department for capital market, was established in 1992. Before that time, Chinese security market was full of chaos without uniform regulation and supervision. On the other hand, before 1992 share prices of SHSE and SZSE markets were dramatically volatile.

In order to minimize the bias incurred by the time structural changes, the sample of each index has been divided into four subgroups with different and subsequent time spans (21/02/1992-29/12/1995, 01/01/1996-30/12/2000, 02/01/2001-30/12/2005, 01/01/2006-30/06/2009). Along with the entire sample, the data of the four subsamples are run by testing tools as well. According to Laurence et al. (1997), stock returns are calculated by the following function:

Where:

Pt = price index at time t;

Pt-1 = price index at time t-1;

Rt = stock return in logarithmic form

After calculating the stock return series, the following figures present the movements of price and return indices for SHSE and SZSE respectively.

It can be seen from Figure 1 that from 1992 to 2009 the share price movement of SHSE-A and SHSE-C indices are roughly matched. To be specific, share price fluctuate moderately from 1992 to 2005 with a rising trend. But the next two years see a remarkable soar to reach a peak, and then the indices drop dramatically in 2008. Unlike the developing trend of SHSE-A and SHSE-C, share price of SHSE-B index experience a rapid climb in 2001, and then develop similarly to SHSE-C. For the comparison of stock return charts, SHSE-A and SHSE-C indices see a relative noticeable fluctuation by approximate positive or negative 20 percent in the first several years, and in the rest of time the returns fluctuate within a small range. Compared with this, return of SHSE-B index demonstrates an extensive variation during the testing period.

As presented in Figure 2, the situation of SZSE-A and SZSE-C indices exhibit roughly identical to SHSE-A and SHSE-C indices, while SZSE-B perform similar to SHSE-B. However, in terms of the stock returns, A-share, B-share and Composite indices in Shenzhen all demonstrate a relatively greater fluctuation compared to Shanghai Stock Exchange.

3.2 Hypothesis

There are two key objectives of this study, one of which is to examine the validity of weak-form EMH for Chinese capital market. The other one is to detect the existence of seasonal anomalies by applying Ordinary Least Square (OLS) Regression. According to these two purposes, the following hypotheses are proposed:

Hypothesis 1

H0: Weak-form EMH and random walk hypothesis are valid in Chinese capital market.

H1: Weak-form EMH and random walk hypothesis are not valid in Chinese capital market.

Hypothesis 2

H0: Chinese stock market returns do not follow a seasonal pattern.

H1: Chinese stock market returns do follow a seasonal pattern.

3.3 Model Specification

3.3.1 Random Walk Model

With the intention of investigating the randomness of stock price in Chinese capital market, random walk model is applied, which is modeled as follows:

Where:

Pt = share price of the tested index at time t;

β1 and β2 are parameters, while μt is an error term.

Based on the random walk theory, there are two prerequisites, which are:

The error term μt is a white noise, namely

Under the condition that β2 = 1, symbolically

or

When β1 = 0, it is considered to be random walk without drift.

3.3.2 Models for Weak-form EMH

In order to understand the nature of Chinese capital market comprehensively and accurately, four different tools have been applied respectively in this paper, which are serial correlation coefficient, runs, unit root and OLS regression tests. The first two testing methods are traditional ones in EMH studies, but the assumptions of these two measures are too rigorous to satisfy. Therefore, the unit root test is expected to compensate the outcomes in order to generate a relatively more reliable conclusion. Apart from that, the calendar effect could be investigated by the tool of OLS regression for further understanding of Chinese stock market.

3.3.2.1 Serial Correlation Coefficient Test

Serial correlation coefficient test is one of the most prevent and basic tool in EMH study. From the theoretical aspect, the share prices are believed to follow a random walk if there is no correlation between stock return and any lags. The basic function of serial correlation coefficient test is provided as follows:

Where:

Rt = stock return in logarithmic form at time t;

ρ(k) = the autocorrelation coefficient of Rt;

Moreover, ρ(k) satisfies the following distribution:

The null hypothesis that the real ρ(k) is equal to zero, should be accepted when the calculated drops inside the confidence interval of chosen significant level.

Another test, called Box and Pierce Q statistics, could be applied to test whether all the ρ(k) are equal to zero at the same time.

Where Q is a special distribution ofwith the df equal to m. The null hypothesis here is that not all ρ(k) are zero and we reject it when the Q value is greater than the critical value of χ2 distribution.

In addition, Ljung-Box statistics is more appropriate for small sample, which could be expressed as:

Where m and n stand for the lag length and sample size respectively.

3.3.2.2 Runs Test

Contrast to serial correlation coefficient test, runs test is a particular one without any parameter, which under the assumption that stock returns might not follow the normal distribution but should be independent, according to Worthington and Higgs (2004). To be specific, the test mechanism is to investigate whether the observed-runs of the sample proportion is equal to the expected-runs of the population proportion, where the runs refers to the changes of stock price (negative, zero or positive changes). If it relationship was established, we believe the runs number of the sample proportion could be calculated by using the following formula:

Where:

n = the amount of observations

ηi = the amount of changes of stock price

In addition, in terms of large sample number of observation of which is more than 30, M roughly follows normal distribution (Fama 1965) and Z-statistic is:

3.3.2.3 Unit Root Test

As to , when β is equal to one, then we can get , which indicates that Pt is not stationary. So in order to test the stationary of Pt, to examine whether β is equal to one is significate, called unit root test. Nevertheless, there still exist some problems in unit root test. On one hand, it requires certain critical values, because the t-values of the coefficient of Pt-1 sometimes are out of the expected t distribution. On the other hand, under different forms of I(1), the according critical values vary from each other.

However, another important contribution has been made by Dickey and Fuller in 1976, who discovered τ-statistic, namely DF test. In detail, when a time series follows AR(1) process, DF test is available to apply. On the other hand, in terms of the AR(p) process where p >1, DF test is not applicable and ADF test could be employed, which adds lagged different term into the equation. There are three different conditions as follows:

Pt is a random walk:

Pt is a random walk with drift:

Pt is a random walk with drift around a stochastic trend:

The null hypothesis that the tested time series follows a random walk process could be accepted when the calculated absolute τ-statistic is greater than the critical value of τ according to the chosen significant level.

3.3.2.4 Tests of Monthly and Daily Effect

Theoretically, a capital market is considered to be weak-form efficient when the stock prices are random without any predictable clues. Nonetheless, a great number of previous empirical studies have reported substantial anomalies in stock market, particularly the seasonal pattern of share price. Therefore, in other to comprehensively understand Chinese capital market, the calendar anomalies will be examined as well through modeling monthly and daily data as follows (the first equation is to investigate monthly effect, while the second one is for daily effect).

Where:

Rt = the mean return on month (day) t;

ai = the mean return for different month (day) of the year (week);

D1t …D12t = dummy variables (either equal to 1 or 0)

The null hypothesis here is all of the coefficients of all the dummy variables are equal to zero. So when any one of the coefficients is significantly different from zero, the null hypothesis could not be accepted, implying the existence of monthly or daily effect.

Chapter 4 Empirical Results

4.1 Descriptive Statistics

The above table shows relative descriptive statistics of weekly stock returns for both Shanghai and Shenzhen Stock Exchanges. As presented in Table 2, the greatest average return could be saw in SHSE-A. Moreover, SHSE-A has the maximum stock return of 0.928, especially compared with the comparably lower maximum values of SZSE. On the other hand, SZSE-B experiences the lowest minimum stock return (-0.388). In consequence, it is obvious that compared to SZSE, SHSE tends to generate relatively higher returns in general.

Besides that, the highest standard deviation of stock return exists in SHSE-A (0.066) while SZSE-C has the lowest one (0.053), which implies that the volatility of stock return in Shanghai market is more dramatic than the corresponding one of Shenzhen market.

What is more, as indicated in Table 2, it is reasonable to conclude that the stock return of none of the security market follows normal distribution. First of all, for the skewness values, it could be seen that the stock returns of all target indices are distributed asymmetrically, but skew to the right side, in line with the findings of Mookerjee and Yu (1999). Secondly, none of the p-value could result in the accepting of the null hypothesis that the stock return is normal distributed. Therefore, for the six return series, none of the indices follows standard normal distribution.

4.2 Results of Serial Correlation Coefficient Test

The following Table 3 demonstrates the empirical outcomes of serial correlation tests for both Shanghai and Shenzhen Stock Exchanges. The relationship of weekly returns for six indices could be clearly detected. As show in Table 1, it could be seen that, as to SHSE-C, the autocorrelation coefficients are significantly different from zero for lag 1 to lag 7 at least at the significance level of 10%. Contrast to that, for the index of SZSE-C, the correlation coefficients of lag 4, lag 6 and lag 8 are statistically non-zero at the significance level of 10%, while the ones for lag1, lag 2, lag 3, lag 5 and lag 7 are non-zero at the significance level of 5%.

In addition, the association between the return pattern of each A- or B-share index and the one of composite index could be point out respectively. In another word, for both Shanghai and Shenzhen Stock Exchanges, the effects on the market composite index from A- and B-share markets are remarkably diversified, which exhibits from two aspects. On