CHAPTER 2

LITERATURE REVIEW


2.1 Introduction

This chapter reviews the past and recent empirical research of distribution of stock price in relation to information and trading volume in both international and local stock markets.

This study aims to explore the volume-return relation in the Bursa Malaysia. The Bursa Malaysia (a developing stock market) is relatively less liquid and has thinner trading volume compared to developed stock markets. Overall, the studies on volume-return in the stock market are still lacking due to the infancy of the Malaysian capital markets. To this point, no study on dynamic volume-returns relation in the aspect of information asymmetry has yet been carried out in Malaysia. More importantly, this study is done not only to examine the predictability of past returns and trading volume, but also to provide simple and exploitable trading strategy incorporated with the information dissemination.

In this context, Section 2.2 focuses on empirical studies investigating the information and stock price movement. Theoretical explanation of the efficient market hypothesis, mixture of distributions hypothesis, sequential information arrival hypothesis, and information asymmetry are discussed in this section. This is then followed by the review of literature on studies pertaining to volume-return relationship in Section 2.3. Past empirical research in contemporaneous relationship between volume traded and stock returns, linear and nonlinear causality tests as well as dynamic volume and returns autocorrelation in both local and international stock markets are also presented in this section. Finally, the chapter will be concluded in Section 2.4.


2.2 Information and Stock Price Movement

According to Beaver (1968), anything that causes investors to act can be described as information regardless its impact on the underlying valuation of the company. Various researches have focused on attempting to predict future stock prices based on historical prices alone. The financial theories related to this study are discussed as below:

2.2.1 Efficient Markets Hypothesis (EMH)

Samuelson (1965) showed that efficient markets exhibit martingales[10] process and he described the behaviour of martingales process as a weak-form market efficiency. It was then popularised by Fama (1965).

Many investors are attracted to the stock market hoping for high gains. When money is invested into the stock market, it is done with the objective of generating a return on the capital invested. Some not only tried to make a high return but also tried to outperform the stock market. However, market efficiency which was championed in the efficient market hypothesis has suggested that at any point of time, prices will fully reflect all available information about individual stocks and the stock market as a whole (Fama, 1965). The accepted view was that when information arose, the news would spread very quickly and be incorporated into the prices of securities immediately. Thus, according to the EMH, no market player has the advantage in forecasting stock price movements since no one has access to information that not is available to everyone else.

Numerous investors, including highly trained and professional analysts, try to predict trends in the market. They attempt to identify the undervalued stocks that have high chances to increase in value in the future. They believe that they can select those stocks that will outperform the market through fundamental analysis, an analysis of financial information such as company earnings, dividend payout, asset values and so forth, or through technical analysis, a study of past stock prices in an attempt to predict future prices. These analyses enable the investors to achieve returns greater than those that could be obtained by holding a randomly selected portfolio of individual stocks with comparable risk (Malkiel, 2003). However, under the EMH, investors are engaging themselves in a game of chance, not skill, at any time of them buying and selling stocks. If markets are efficient, investors need not waste their time trying to find underpriced or overpriced securities. This means that prices are always reflecting all available information. Therefore, there is no way investors will ever be able to buy mispriced securities. It is impossible to overperform the market too as prices would have incorporated and reflected all relevant information.

The EMH is associated with the theory of random walk (Fama, 1965). This term is used to characterise a price series where all subsequent price changes represent random departures from previous prices. That means "successive price changes are independent" (Fama, 1965, p.36). Since prices have no memory, past and present prices which cannot be used to predict future prices have made it impossible for one to predict which direction the stock or market will move at any point accurately. Therefore, it is not possible to trade profitably simply making decisions based on available price data.

The EMH is not only concerned with the type and source of information, but also the quality and speed of which it is disseminated among inventors. This raises an important question: What types of information are available and incorporated into stock prices? According to Fama (1970, 1991), EMH may exist in three levels: (i) weak form of the EMH, (ii) semi-strong form of the EMH, and (iii) strong-form of the EMH. It should be pointed out that the concept of market efficiency evolved from the concept of perfect competition on assumptions of free and constantly available information, rational investors, zero brokerage commissions, flotation costs, taxes and transaction costs (Fama, 1970). In essence, the more efficient a market is, the more unpredictable future pricing it will be, with the expected value of future prices equal current price. Fama (1970), Reily and Brown (1997), and Jones (1998) indicated that an efficient capital market requires a set of conditions as follows:

  1. There must be a large number of rational, profit-maximising investors who are actively participating in analysing and evaluating securities prices and these participants act independently of each other.

  2. Information is costless and widely available to market participants. Arrival of information regarding securities is in a random fashion. Announcements of new information are independent of investors and news reaches the investors at approximately the same time.

  3. Security prices will adjust to all new information and will fully reflect all available information.

  4. Investors will adjust security prices rapidly to reflect the effect of new information. Sometimes there may be an over-adjustment or an under-adjustment but it cannot be predicted with certainty what it will be.

After the discussion of the EMH, the discussion proceeds to the three forms of EMH (Fama, 1970). The weak form reflects the situation where movements in stock prices follow a random path. Current stock price movements are independent of past price movements. In other words, all information contained in past trading volume, prices of stock, and the rates of return are already reflected in the current stock prices. Hence, the past data on stock and market are of no use in predicting future price changes. The random nature of stock price movements, on the other hand, means that any attempt to study past prices moving in order to detect mispriced stock and to gain above-average profits will fail. One should not be able to gain from using information that everybody else has known. Investors and analysts who practice technical analysis by drawing up charts of past stock prices and trading volume in order to predict future price movements will be wasting their time since it cannot be used to predict and beat a market.

The semi-strong form of the EMH states that the current stock prices not only reflect all past price movement but also all publicly available information (Fama, 1970). Examples of public information are data reported in a company's financial statements, earnings and dividend announcements, announced merger plans, the financial situation of company's competitors, expectations regarding macroeconomic factors and so forth. This information will then be available at random intervals, and are quickly absorbed by the market. Therefore, investors who practice fundamental analysis by studying relevant reports and announcements with the attempt to make above-average returns on a consistent basis, would be disappointed as the stock prices have already reflected such new public information.

The strong-form of the EMH is the strongest version of EMH, which states that current stock prices reflect all pertinent information, both public and private or inside information (Fama, 1970). The current stock price reflects all true or intrinsic value of the share and thus, the stock would be fairly priced in the stock market. There would be no opportunity for investors to have exclusive access to information relevant to stock prices. Even if there were release of new information, it will be easily available to everyone at the same time at no cost. In other words, the strong-form of EMH states that even corporate insiders, such as officers, directors or other privileged individuals within a corporation would find it impossible to systematically gain abnormal returns from inside information. Such information includes detailed information about the financial state and major strategic of the firm, alongside the tactical decisions the company makes that may not be available to shareholders.

2.2.2 Mixture of Distributions Hypothesis and Sequential Information Arrival Hypothesis

The two well-known hypotheses which suggest that volume and price are jointly determined are the mixture of distributions hypothesis (Clark, 1973) and the sequential information arrival hypothesis (Copeland, 1976). Both hypotheses assume an exogenous source of information in the market, and volume variations which correlated with price changes are the ultimate result. Before proceeding to the explanations of these hypotheses, it is necessary to first discuss the ideas of joint behaviour between price and volume.

Volume is a measure of the quantity of shares that change owners for a given stock. For instance, in 2005, the average daily volume was 330.053 million shares, contributing to RM558.583 million of stocks traded each day among the 1021 companies listed on the Bursa Malaysia (Bursa Malaysia, 2007). The amount of daily volume on a stock can fluctuate on any given day depending on the amount of available new information being disseminated to the market. Such information could be a press release or a regular earnings announcement provided by the company, or market-wide news released by the government. For example, Mesdaq-listed Iris Corporation Berhad has an average trading volume of 9,150 shares per day. On May 11, Iris stock was declared by Bursa Malaysia Securities Berhad, in consultation with the Securities Commission (SC), as a designated stock with immediate effect and was given the excessive speculation and unusual patterns observed in the trading of the shares. As a result, Iris' share price rose 1,500% in eight months from RM0.085 on 12th September 2005 to RM1.36 on 11th May 2006 and has gained RM1.275 during the period. It was the highest price since its listing in 2002 with trading of 97,556 shares on that day, about eleven (11) times of the average. Nevertheless, Iris share was requested a trading suspension on the next trading day (15/05/2006) and the news led to a massive drop in trading volume of 60% to 40,028 shares on the second trading day after the announcement was made (Bursa Malaysia, 2006c).

Based on the above, financial markets are not fully efficient. Stock prices cannot convey all available information to market participants if the markets are not efficient. Many investment professionals such as brokers, investment analysts and also portfolio managers meet the EMH with a great deal of skepticism. For instance, Warren Buffett and Peter Lynch, the world's successful stock market investors, have consistently outperformed the market over long periods of time with techniques that are not supposed to work in a purely random efficient market. Legendary portfolio manager Michael Price did not leave anybody guessing which side he was on: "...markets are not perfectly efficient. The academics are all wrong. 100% wrong. It's black and white." (Tanous, 1997, p.39).

Beaver (1968) noted that volume is a useful tool in determining how much disagreement exists with the arrival of new information. He stated that "an important distinction between the price and volume tests is that the former reflects changes in the expectations of the market as a whole while the latter reflects changes in the expectations of individual investors" (Beaver, 1968, p.69). Therefore, watching at price alone was insufficient to predict future prices. Trading volume was also considered to be incorporated with information in the prediction of stock prices. Therefore, understanding of volume, returns, and information is essential in this study.

Osborne (1959) hypothesised that stock prices could be modelled as a lognormal distribution with the variance term dependent on the quantity of transactions. He implied the possible relation between returns and trading volume. Ying (1966) used trading volume as an exclusive predictor of future prices. He was the first economist to demonstrate the presence of an asymmetric in price change-volume relation. He argued that if there was asymmetry in the behaviour of the ratio of volume change, then the signed price change was more relevant than the absolute price change. He applied a series of statistical tests to a six-year daily series of Standard & Poor's (S&P) 500 Index return and total volume of New York Stock Exchange (NYSE) from January 1957 to December 1962. The results showed that some of these conclusions were inconsistent with the weak form of market efficiency. His main conclusions were i) a small (large) volume is usually accompanied by a fall (rise) in price, ii) a large increase in volume is usually accompanied by a large price change, iii) a large volume is usually followed by a rise in price and iv) if volume has been increasing (decreasing) for a consecutive five trading days, then there will be a rise (fall) in the prices for the next four trading days.

Trading volume facilitates the price discovery process, enables investors to share financial risks and ensures that corporations can raise funds needed for investment. Karpoff (1987) found that there was a positive correlation between absolute price change and volume in both the stocks and futures markets, but positive correlation between price change per se and volume was only found in the stock markets. Karpoff attributed this asymmetry to the greater cost of short positions in the stock markets. He indicated the strength of the relationship between price changes and volume depended on the sign of the price change. For example, for a given price change, a lower level of trading is required for a negative price change rather than a positive price change. This is consistent with the Wall Street adage that volume is relatively heavy in bull markets and light in bear markets.

Karpoff (1987) noted a number of reasons why price-volume relation is important. First, it provides insight into the market structure. This is because price-volume relationship depends on the speed of information flow to the market as well as the ways and the extent of information are disseminated and also the size of the market. Second, understanding of the joint distribution of returns and volume is important for statistical inference in event studies. The structure of tests and the validity of many of the inferences of the event in questions depend on this joint distribution. Third, it helps to understand the empirical distribution of speculative prices. This distribution of rates of return appears kurtotic. The mixture of distributions hypothesis has been used to view this distribution as a finite variance mixture of normal distributions where volume is the proxy for variances at different events. Fourth, price-volume relationship provides significant implications to futures market in addition to equity market. Price variability affects the volume of trade in futures contracts. Therefore, a better understanding of the statistical structure of this relation can provide some explanations to the technical analysis and indicate the importance of private versus public information in determining investors' demand.

The explanations on the mixture of distributions hypothesis (Clark, 1973) and the sequential information arrival hypothesis (Copeland, 1976) are as below:

i. Mixture of Distributions Hypothesis (MDH)

The mixture of distribution hypothesis (MDH) was developed by Clark (1973). He suggested that prices are drawn from a set of distributions of differing variances. This causes the leptokurtotic feature observed in distribution of price changes. The rate of information arrival as a latent common factor affects both the price and volume simultaneously and causes a contemporaneous movement. Therefore, he suggested that price change and trading volume bear a positive relationship due to their joint dependence on a common event. In other words, prices and volume only change when information arrives, and they evolve at a constant speed in event time.

According to Clark (1973), returns and trading volume are positively correlated because the variance of returns is conditional upon the volume of that transaction. In Clark's model, trading volume was a proxy for the speed of information flow, which was regarded as a latent common factor that affects prices and volume simultaneously. The insight of this MDH model is that returns volatilities are changing over time since information is available to investors at a varying rate. Trading is slow and there are only few price changes on days when not much information is available. In contrast, when new unexpected information hits the markets, it results in higher level of market activity than usual, and prices move much more rapidly. The trading volume which is a measure of the level of activity will also increase.

The MDH developed by Clark (1973) was then modified by Tauchen and Pitts (1983) and Andersen (1996) respectively. Tauchen and Pitts assumed all traders to behave unstrategically. Unlike finding of Clark where probability distribution of the daily price change was leptokurtic, they concluded that price volatility and trading volume were both normally distributed and subordinated to the same latent information arrival rate. Andersen incorporated asymmetrically informed strategic traders and liquidity traders into Clark's MDH model. The modified MDH of Andersen argued that the dynamic features of MDH should be managed by information flows that was modelled as a stochastic volatility process. This stochastic volatility process would then generalise autoregressive conditional heteroskedasticity (ARCH) specifications. This will be mainly due to the poisson distribution of stock market microstructure.

At this juncture, it should be noted that this study does not intend to examine the contemporaneous volume and absolute returns as described alone. Instead a discussion of mixture of distributions models that set the stage of evolution of information dissemination in stock market will be made. Besides, the contemporaneous volume and market returns per se and dynamic return-volume relationship (causal) will also be examined in this study. The causal relationship is anticipated in the sequential arrival of information hypothesis, which is next to be discussed.

ii. Sequential Arrival of Information Hypothesis

According to the classic definition explained in Clark's MDH model (1973), in an efficient market setting, volumes should be correlated to contemporaneous returns, but not predictive power on future returns. With new information flowing into the market, prices should adjust to a new equilibrium and increase on volumes should follow.

The sequential arrival of information hypothesis of by Copeland (1976) meanwhile suggested that gradual dissemination of information in such a sequence of temporary market equilibria does exist. The hypothesis states that when the release of new information happens, price changes and volume increases are not necessarily contemporaneous phenomena, as what the market efficiency suggests.

An underlying assumption of this sequential arrival of information hypothesis is that informed investors do not trade to a point where the demand causes the price to adjust to a price that would occur if all investors had the information (Copeland, 1976). Constraints on short selling and borrowing could be factors causing prices to adjust only partially to new information that is asymmetrically distributed. The new information that reaches the market is not disseminated to all participants simultaneously but instead, to one trader at a time. Each trader interprets this information differently. A number of traders will become optimistic while some might become pessimistic. Subsequently, each trader trades in the market based on his or her beliefs.

The sequential arrival of information hypothesis supports that final information equilibrium will only be reached and established after several intermediate equilibria, provided that all traders have received the information (Copeland, 1976). This hypothesis implies the continuation of higher volatility after the initial information shock rather than spikes in volatility. It suggests a feedback relationship between stock returns and trading volume where lagged trading volume is expected to imply current absolute price changes and lagged absolute price changes will imply current volume.

Grammatikos and Saunders (1986) stated that "sequential arrival models imply the possibility of observing lead relations between daily contract price variability and volume" (p.326). This happens when each investor monitors the information sequentially. Following Copeland (1976), McMillan and Speight (2002) echoed that in the sequential information flow where lagged trading volume provides information on current absolute stock returns and lagged absolute returns contain information of current trading volume, a dynamic relationship is developed. This relationship is very important as it gives useful information about trading volume and movement of stock prices and volatility.

In a nutshell, Copeland's (1976) sequential arrival of information hypothesis implies that market is not perfectly efficient since lagged trading volume could have predictive power for current absolute stock returns and vice-versa. This hypothesis however suggests a positive contemporaneous relation between volume and absolute value of the price change, and also a positive causal relationship in either direction.

2.2.3 Information Asymmetry

To this extent, this study has discussed the theoretical foundation of volume-return relation. Overall, the relation is dynamic since there are certain situations where others could have more information than we do. Therefore, before the discussion of review of previous studies on volume-return relation, an explanation of a situation where some market participants have superior information to others will be useful.

As such condition above, a so-called information asymmetry will occur. Becker related information asymmetry as a "condition in which at least some relevant information is known to some, but not all, parties involved" and "information asymmetry causes markets to become inefficient, since all the market participants do not have access to the information they need for their decision-making processes (Becker, 2008, p.1312). Likewise, information asymmetry is a situation where "informed investors have material, firm-specific information related to future public announcements and uninformed investors do not" (Chae, 2005, p.413).

The terminology of information asymmetry is derived from the Akerlof's "lemons" car market.[11] Akerlof (1970) showed that informational asymmetries could give rise to adverse selection in markets. He related quality and uncertainty using an example of used cars to capture the essence of the information asymmetry problem in automobile markets. He illustrated how markets malfunctioned when different participants had different information about the qualities of assets being traded. In the absence of adequate mechanisms to assure quality, the symmetric information problem could either cause an entire market to collapse or contract it into an adverse selection of low-quality products. He also provided examples and implications of information asymmetry for the organisation besides regulation of important real markets such as the labour, health insurance, and financial commodities markets.

The impact of information asymmetries on firm value and stock price has been analysed extensively over the years in the finance literature. Merton (1987) and Easley, Hvidkjaer and O'Hara (2002) who uses a two-period model of capital market equilibrium with incomplete information environment found information asymmetry an active factor in determining asset returns and cost of capital. Merton noted that asymmetric information among investors has caused some institutional banks and retail investors not to invest at all in certain firms, for instance, small firms with few stockholders. Using data of individual stocks listed on New York Stock Exchange from 1983 to 1998, Easley et al. found that the probability of informed trading was more important than the beta in determining asset price. They incorporated estimates of the probability of information-based trading into a Fama and French (1992) assest-pricing framework and discovered that an increase in the probability of informed trading of 10 percentage points was due to a 2.5 percentage point decrease in expected return per year.

In poorly-regulated markets, corporate insiders like directors, executives, and large shareholders may represent a significant proportion of superiorly-informed traders. Their presence may increase the amount of information asymmetry, and subsequently affect the volume traded and stock prices movement. Although paragraphs 3.14, 3.15, 4.07, 4.22, 9.19, 15.02 of Malaysia listing rules require a company to appoint an independent corporate adviser to advise minority shareholders and prevent directors, substantial shareholders or interested parties from voting on the transaction, there are still no specific regulations in governing insider trading (Bursa Malaysia, 2009). Therefore, the study of dynamic volume-returns on the extent of degree of information asymmetry is extremely important for the Malaysian stock market which is an illiquid emerging market.

2.2.3.1 Proxies of Information Asymmetry

Since the degree of information asymmetry is not directly observable, econometric and empirical methods have been developed in the microstructure literature to detect and measure information asymmetry in security markets. An explanation of the different proxies variables used to measure information asymmetry will be provided, as follows:

i. Firm size

Firm size is an important factor in determining the degree of information aysmmetry of a firm. Conrad, Gultekin, and Kaul (1991) categorised 300 NYSE/American Stock Exchange (AMEX) listed firms into three market value-weighting portfolios, namely large, medium, and small firm size. They noted that firm size was a measure of information asymmetry and firms of larger size were expected to be more information efficient. They explained that different findings obtained for the three market value-weighting portfolios were probably due to the impacts of aggregate information on large capitalisation firms before the information was impounded into small capitalisation companies. Diamond and Verrecchia (1991) differentiated small and large firms using a model of liquidity. They found that small firms faced more information asymmetry problem due to lower disclosure quality and they practiced increased information asymmetry. Unlike large firms, small firms tended to receive less attention from the market and regulators.

By far, market capitalisation has been commonly used to measure firm size due to the direct function of a company's stock price. Gervais, Kaniel and Mingelgrin (2001) investigated whether unusual high or low trading volume over a period of a day to a week tended to appreciate or depreciate in the following month. They utilised daily and weekly data on NYSE stocks from August 1963 to December 1996 and divided them into large, medium, and small market capitalisation groups. This was to show different degree of information asymmetry in the respective groups.

On the other hand, �lvwarez-Otero and Gonz�lez-M�ndez (2006) used price data of 111 firms going public on the Madrid Stock Exchange from 1985 to 1997 to investigate the effect of firm size on firm's performance. They employed the end of the first trading day's market capitalisation of those companies to measure their firm size. They found that larger firms tended to show a greater initial return to investors and performed better over a long period of time. The finding also indicates the existence of a size would affect the measurement of market capitalisation on the initial return and financial performance of newly listed firms. Following the finding, market capitalisation would be adopted as the proxy to the level of information asymmetry in this study.

ii. Microstructure Measures

According to O'Hara (1995), "market microstructure is the study of the process and the outcomes of exchanging assets under explicit trading rules" (p.1). It analyses how specific trading mechanisms affect the price formation process. Lerner (1994) defined bid price as "the price at which dealers are willing to buy the stock" and ask price as "the price at which dealers are willing to sell the stock" (p.64). Hence, the bid-ask spread is the quoted difference between how much money a dealer is willing to pay to buy a stock and how much money a dealer is willing to accept to sell a stock.

The idea that the bid-ask spread could be caused by the presence of asymmetric information across market participants was first illustrated by Demsetz (1968) and Bagehot (1971). Demsetz noted that not all liquidity-motivated traders who seek for immediacy would arrive at the same time. He stressed the importance of NYSE market makers in providing immediacy to traders. He also measured bid-ask spreads as compensation for market makers who provides immediacy to traders. It was a function of the market maker's waiting and inventory holding costs. Then, he measured the arrival of a trader or time rate of transactions for a stock through the number of transactions per day and number of shareholders of the stock. He showed that the transaction rate of a stock depended on the number of shareholders and there was a negative relationship between bid-ask spreads and time rate of transactions. In other words, an increase in the number of shareholders would cause a reduction in bid-ask spreads. He explained that it was due to economies of scale in transactions.

Bagehot (1971), meanwhile, has first discovered the relation between information and liquidity traders. He suggested that information costs could affect prices and the quote process. He linked bid-ask spreads to market makers' informed trading risk due to trades of informed traders, noise traders and uninformed or liquidity traders. When a market maker is lost to informed traders, he would gain from liquidity traders. Hence, a market maker ought to balance his sources of income in setting the bid-ask spreads.

The concept of bid-ask spreads was formalised by Copeland and Galai (1983) through developing a quantitative model for determining the bid and ask prices with the presence of both liquidity and information motivated traders. They did not directly measure informed trading risk. Instead, the market maker determined the bid and ask spreads based on the probability of encountering an informed trader in a market. The expected return earned from liquidity traders was maximised and the expected losses from trading with informed traders was minimised in order to make the most profit of all.

Glosten and Milgrom (1985) noted that "bid-ask spread can be purely informational phenomenon, occurring even when all specialist's fixed and variable transaction costs (including his time, inventory costs, etc.) are zero and when competition forces the specialist's profit to zero" (p.72). They showed that a bid-ask spread increased when the degree of information asymmetry became higher. The findings concluded a positive relationship among bid-ask spread, price level, and return variance. At this juncture, both the models of Copeland and Galai (1983) and Glosten and Milgrom (1985) would have illustrasted that asymmetric information could induce a bid-ask spread. It is echoed in Raymond and Venkatesh (1988) that "a higher degree of information asymmetry leads to a larger bid-ask spread" (p. 1041). In other words, bid-ask spread is a function of the degree of information asymmetry.

More recently, How, Verhoeven, and Huang (2005) have used bid-ask spread of a sample of 109 firms listed on the Australian Stock Exchange (ASX) as the proxy of information asymmetry. The results of multiple linear regression showed that the degree of information asymmetry was reduced after earnings and dividend announcements. How et al. (2005) indicated that the information content in the announcement of earnings and dividends have left the informed investors with no better knowledge in the prospect of the firm and thus, the reduction in the spread.

Chan, Menkveld, and Yang (2008) then employed data of 76 A- and B-stocks traded on the Shanghai Stock Exchange (SHSE) and the Shenzhen Stock Exchange (SZSE) from January 2000 to November 2001 to examine the impact of information asymmetry on local A- and foreign B- share markets in China. They used adverse selection component of the bid-ask spread decomposition and applied it into market microstructure models as a measure of information asymmetry. The results showed that A-shares in China stock market tended to exhibit more informed trading and traded more actively than B-shares. Information asymmetry have also reflected the extent of private information available to investors. Domestic investors in China were more willing to pay a higher price compared to foreign investors and this resulted in a lower price in B-shares. Hence, information asymmetry plays a more important role than trading volume in determining stock price in China.

Initially, the study does not intend to elaborate on the theory of the bid-ask spread. Nevertheless, the application of the bid-ask spread as one of the proxies to information asymmetry will be explained further in Chapter 3.

iii. Analysts' Forecasts Measures

Financial analysts mostly study on the financial performance of firms and recommend investment decisions. Elton, Gruber, and Gultekin (1984) however discovered that analyst forecast errors could be a reasonable proxy for information asymmetry. As observed, about 84 per cent of the analyst forecast errors was mainly caused by misestimation on firm-specific factors rather than industry or economy factors. They also found that firms with higher levels of information asymmetry between managers and investors on a firm's cash flows tended to have higher forecast errors than the rest.

Krishnaswami and Subramaniam (1999) used analysts forecast dispersion as a proxy for asymmetric information to investigate the change in information environment before and after a spin-off. A sample of 118 voluntary corporate spin-offs from January 1979 and December 1993 was used. They found that the forecast variability became smaller than a matched industry sample and decreased significantly after spin-off. In other words, levels of information asymmetry have decreased significantly following the completion of a spin-off.

iv. Number of Analysts Following a Stock

Bhushan (1989) noted that a decrease in the number of analyst following a firm would usually be accompanied by an increase in the number of business segments within a company. Brennan and Subrahmanyam (1995) employed a sample of 1,550 listed firms on NYSE for the calendar year 1988. They applied a simultaneous equations framework by using number of analysts following the stock as a proxy for the supply of information about the firm. They found that greater analyst coverage tended to improve in American market's liquidity. They attributed this to more information that was revealed to the public when there were more analysts following a firm. Hence, less asymmetric information the firm would suffer. Overall, market makers anticipated lesser losses to informed traders due to lower degree of private (asymmetric) information on large firms.

Besides that, Chung, Wood, and Wyhowski (1995) have used 469 firms listed on NYSE and AMEX from 1984 to 1988 to construct 2,345 pooled time-series and cross-sectional observations. They found that analyst coverage was positively correlated with the level of asymmetric information of a firm. As opposed to Brennan and Subrahmanyam (1995), they insisted that more analysts were attracted to firms with high informational asymmetries. Thus, the value of private information has increased with the degree of information asymmetry. Larger firms were thought to have lower levels of information asymmetry since they normally release more information than small firms do.

On the other hand, D'Souza and Jacob (2000) used a sample of 35 firms from 1990 to 1996 to examine any significant increase in coverage after a firm issues a tracking stock. They defined analyst following before the issuance of targeted stock as "the number of analysts who issued at least one forecast of earnings for the company during the year" (p.478). They employed the sign and Wilcoxon signed-rank tests and found no statistically significant change in analyst following after the issuance of targeted stock. They noted that the valuation task of an analyst may be facilitated by how detailed the financial statements that firms must disclose for each targeted stock segment. In other words, the degree of information asymmetry of a firm is likely to have a positive effect on the firm valuation in short term.

v. Investment Opportunity Set Measures

Smith and Watts (1992) stated that the asymmetric information problem was more severe for firms with significant growth opportunities. This was because managers of high growth firms had more superior information about their firm's investment opportunities and better knowledge on the expected future cash flows from their firm's existing assets compared to outsiders. Based on this reasoning, a number of studies have used proxies for a firm's investment opportunity set as measures of information asymmetry.

For instance, McLaughlin, Safieddine, and Vasudevan (1998) have used market-to-book ratio of equity as a proxy for firm's investment opportunities to measure its degree of information asymmetry. They investigated the relation between information asymmetry and long run performance of firm following a seasoned equity offering using a sample of 1,967 equity offerings and 960 debt offerings on listed on the NYSE and the AMEX. They found that firms with higher degree of information asymmetry tended to have more negative abnormal performance following a seasoned equity offering. They also found that an increase in information about a firm could lead to a convergence of opinions regarding the firm's expected future earnings.

Thomas (2002) meanwhile used a sample of 2,677 firms listed on 224 segments on American stock market from July 1985 to June 1996 to examine the relation between corporate diversification and asymmetric information. He proxied the market-to-book value and leverage to control the growth opportunity and risk of a firm. He found a negative relation between market-to-book variable with the forecast error and leverage with the forecast error. Firms with better growth opportunity set tended to receive more accurate forecasts than firms with low growth potential. Larger industrial firms were also found to have less information asymmetry since profitability was negatively associated with analyst forecasts error. Overall, larger firms would tend to reveal more information to investors and therefore, display a milder information asymmetry problem.

2.3 Review of Previous Studies

The contemporaneous relation between trading volume and stock returns serves as the first issue to be investigated in this study which is then incorporated with varying volatility. It should be noted that this study intends to examine the dynamic causal relation by considering the possibility of predictive power of trading volume. The second issue, the effect of information asymmetry in this relation will also be examined in this study. Hence, this section reviews all the empirical research done on these issues and some relevant ones. Section 2.3.1 focuses on the discussion of previous research about on the contemporaneous relationship between trading volume and returns with regard to the mixture of distributions hypothesis. Then, the causal relation of that two variables consistent with the sequential arrival of information hypothesis is discussed in Section 2.3.2. An examination of past empirical research on the effect of information asymmetry in such volume-returns relation in both local and international stock markets will be presented in Section 2.3.3. This is then followed by a conclusion in Section 2.4.

2.3.1 Contemporaneous Volume-Returns Relation

The mixture of distribution hypothesis (Clark, 1973) states that the information arrival rate as a latent common factor affects both the stock price and volume simultaneously and causes a contemporaneous movement. A positive relation between price change and trading volume was suggested.

Empirical studies on the return-volume relation in developed financial markets began in the 1960s when initiated by Ying (1966). Besides Ying, Epps (1975) adopted a two-parameter portfolio selection model to investigate the impact of information arrival on return-volume relationship. He characterised all investors as either bears or bulls and assumed they exhibited reinforcing-interpretative behavior. He theoretically predicted and found that the ratio of transaction volume to price changes on upticks and exceeded absolute value of this ratio on downticks. This implied that stock price-volume relationship was asymmetry when the ratio of volume to returns was greater in price increases than in price decreases. Epps (1977) then statistically examined the price-volume on the stock market using price and volume for each transaction among 20 NYSE securities for five trading days in the week of 4th January 1971. The study showed inconclusive results of asymmetric relation using cumulative volume.

Apart from that, Tauchen and Pitts (1983) examined the relationship between variability of daily price change and trading volume on the speculative markets using data of 90-day Treasury bills futures contracts traded at the Chicago Merchantile Exchange from 6th January 1976 to 30th June 1979. They used maximum likelihood estimation and noted that the relationship of variance in daily price change and mean daily trading volume depended much on the average daily date at which new information flows to the market, to the extent where traders disagree to new information given and to the number of active traders in the market. Their findings resolved a conflict in those of Clark (1973).

Meanwhile, Smirlock and Starks (1985) used 131 NYSE stocks for the period from 15th June to 21st August 1981 to examine the price change-volume relationship implied by Epps' (1975) theoretical model. They employed nonparametric Wilcoxon and Fisher's Pooling tests and found that volume was higher on upticks than on downticks on the day of informational arrival. The ratio of volume to absolute price change on price increases was higher than the ratio on equivalent price decreases. In a way, this asymmetry of volume between bulls and bears was very much dependent on informational arrival. They explained that this reversal of results could be due to the combined effect of positive transaction costs and null informational arrival. This further supported the findings of Epps (1975) that the return-volume relation was asymmetric.

Jain and Joh (1988) provided evidence on the intraday joint characteristics of hourly trading volume on NYSE and hourly returns on S&P 500 index for the year 1979 to 1983 using autoregressive integrated moving average (ARIMA) models. The findings showed that there was a strong contemporaneous relation between trading volume and absolute value returns and were consistent with the mixture of distributions hypothesis as proposed by Clark (1973).

Lamoureux and Lastrapes (1990) examined on the behaviour of joint behavior of volatility and trading volume under a situation of high volatility persistence. They realised that the contemporaneous correlation between returns volatility and volume is likely to be incorrectly rejected in case of presence of long memory in prices. A market can be perfectly efficient but still exhibit the varying volatility in price changes if information arrives at uneven intervals. Thus, GARCH model was introduced by them to take account of long memory in testing the contemporaneous correlation between returns volatility and volume. They argued that the presence of the GARCH effects was due to the incorporation of varying rate of information arrival on daily return. They noted that information which led to a change in expectations would in turn lead to a change in prices. Consequently, they used daily trading volume and returns for 20 actively traded stocks whose options were traded on the Chicago Board Options Exchange (CBOE) to examine whether the GARCH effects were really due to time dependence in the process of generating information flows. Trading volume was basically used as a measure for the amount of information that flows into the market using the GARCH model. The results indicated that the GARCH effects in the majority of securities have largely removed after incorporating volume into conditional variance equation and they concluded that GARCH effects were the manifestation of the daily time dependence in the rate of information arrival. Bollerslev and Jubinski (1999) supported these findings and also showed likelihood that news arrival process had a long memory property and could subsequently affect both the volatility and volume.

On the other hand, Saatcioglu and Starks (1998) examined the contemporaneous relation between price changes and volume on a set of emerging markets in Latin American, namely Argentina, Brazil, Chile, Columbia, Mexico and Venezuela by using ordinary least squares (OLS). They used monthly indices data of these six Latin American stock markets over a ten-year sample period from January 1986 to April 1995 in the local currency and United States dollars. They found some evidence supporting a positive price-volume relation in both the magnitude of price change and price change itself.

Lee and Rui (2000) investigated the contemporaneous relationship between six-year daily data of trading volume, stock returns and return volatility from December 1992 to December 1997 in China's four stock exchanges, namely A and B indexes each of Shanghai and Shenzhen markets. They employed GARCH (1,1) model and found a strong positive contemporaneous relationship between return and volume after taking heteroskedasticity into account.

Chen, Firth and Rui (2001) investigated the contemporaneous correlation of return-volume relations using also OLS besides exponential generalised autoregressive conditional heteroscedasticity (EGARCH) for nine markets comprising New York, Tokyo, London, Paris, Toronto, Milan, Zurich, Amsterdam, and Hong Kong. The sample consisted of daily market index and trading volume series from 1973 to 2000. They found a positive contemporaneous relation between trading volume and returns in Japan, Switzerland, the Netherlands, Hong Kong and France and a positive relation between volume and absolute returns in all nine markets. They also found that volatility was better explained by pervious volatility compared to volume.

Lee and Rui (2002) applied generalised method of moments (GMM) and GARCH models on daily data to investigate the information content of volume and the contemporaneous relations among stock market trading volume, returns and volatility in the three largest stock markets namely, New York, Tokyo, and London. The S&P 500 index covered the period of from 2nd January 1973 to 1st December 1999; Tokyo Stock Exchange Price Index (TOPIX) covered from 7th January 1974 to 1st December 1999 and Financial Times Stock Exchange (FTSE) 100 index covered 27th October 1986 to 1st December 1999. A positive contemporaneous relation was found between trading volume and returns in all these three stock markets after taking heteroskedasticity into account. This implies that the volatility of returns was not solely explained by trading volume.

Most recently, Floros and Vougas (2007) investigated the contemporaneous relationship between trading volume and returns in Greek stock index futures market, namely, FTSE/ Athens Stock Exchange (ASE)-20 and FTSE/ASE Mid 40 using GARCH models and GMM system. Daily closing prices and volume for FTSE/ASE-20 index from September 1999 to August 2001 were utilised besides FTSE/ASE Mid 40 from January 2000 to August 2001. The results showed a positive and significant effect which suggested that volume was vital in explaining the conditional variance for FTSE/ASE-20 using GARCH models while GMM system estimation demonstrated a significant relationship between lagged volume and absolute returns. As for FTSE/ASE Mid 40, both the methods provided no evidence of positive relationship between volume and returns.

Table 2.1: Evidence from Prior Studies on Contemporaneous Relation between Trading Volume and Returns

Study

Sample Data

Time period

Methodology

Verdict (contemporaneous relation)

Epps (1977)

New York Stock Exchange

Price and volume for 20 stocks during five trading days in the week of 4th January 1971

Ordinary least squares

Inconclusive results of the asymmetric relation using cumulative volume 

Tauchen and Pitts (1983)

Chicago Merchantile Exchange

Daily price change and trading volume on 90-day Treasury bills futures contracts from 6th January 1976 to 30th June 1979

Maximum likelihood estimation

Relationship depended upon the average daily date at which new information flew to the market

Smirlock and Starks (1985)

New York Stock Exchange

131 stocks from 15th June to 21st August 1981

Nonparametric Wilcoxon and Fisher's Pooling tests

Asymmetry relation of volume-returns dependent on informational arrival

Jain and Joh (1988)

New York Stock Exchange and  Standard and Poor's 500 index

Hourly trading volume for NYSE and  hourly returns for Standard and Poor's 500 index from 1979 to 1983

Autoregressive integrated moving average

Significant contemporaneous relation

Table 2.1 (continued)

Lamoureux and Lastrapes (1990)

Chicago Board Options Exchange

Daily price change and trading volume of 20 actively traded stocks for which options are traded on the CBOE

Generalised autoregressive conditional heteroscedasticity

GARCH effects were manifestation of the daily time dependence in the rate of information arrival

Saatcioglu and Starks (1998)

Emerging markets in Latin American, namely Argentina, Brazil, Chile, Columbia, Mexico and Venezuela

Monthly indices data from January 1986 to April 1995

Ordinary least squares

Positive price-volume relation

Lee and Rui (2000)

China's four stock exchanges, namely, A and B indexes each of Shanghai and Shenzhen markets

Daily data of trading volume, stock returns and return volatility from December 1992 to December 1997

Generalised autoregressive conditional heteroscedasticity (1,1)

Positive contemporaneous price-volume relation

Chen, Firth and Rui (2001)

New York, Tokyo, London, Paris, Toronto, Milan, Zurich, Amsterdam, and Hong Kong stock markets

Daily market index and trading volume series from 1973 to 2000

Ordinary least squares and exponential generalised autoregressive conditional heteroscedasticity

Positive contemporaneous relation in Japan, Switzerland, Netherlands, Hong Kong and France

Lee and Rui (2002)

New York, Tokyo, and London stock markets

Daily market index and volume from January 1973 to  December 1999

Generalised method of moments and generalised autoregressive conditional heteroscedasticity

Positive contemporaneous relation

Table 2.1 (continued)

Floros and Vougas (2007)

Financial Times Stock Exchange / Athens Stock Exchange-20

Financial Times Stock Exchange / Athens Stock Exchange Mid 40

September 1999 to August 2001

January 2000 to August 2001

Generalised method of moments and generalised autoregressive conditional heteroscedasticity

Positive contemporaneous relation

No evidence of positive relationship between volume and returns 


2.3.2 Causality Relation between Trading Volume and Returns

Based on the sequential arrival of information hypothesis, the market is not perfectly informational efficient whereby the price changes and volume increases are not necessarily contemporaneous when there is new information releases to markets. Hence, trading volume could have predictive power for current returns and vice-versa.

Rogalski (1978), Jain and Joh (1988), and Smirlock and Starks (1988) reported an evidence of unidirectional Granger causality from returns to trading volume in the case of United States markets. Monthly data from year 1968 to 1973 for ten stocks and their associated warrants in United States markets were used to empirically examine the causal relation between security prices and volume. Rogalski discovered a positive relationship between price change per se and volume for individual securities after conducting linear filters and causality tests. A positive contemporaneous dependence at lag zero between security price change and volume was also found. He also commented that volume did contain useful information that could improve forecast of warrant price change and common price change.

Jain and Joh (1988), on the other hand, tested the intraday joint characteristics of hourly common stock trading volume and returns on NYSE for the year 1979 to 1983 using Granger-Sims causality tests. The result indicated a somewhat weak evidence of unidirectional relation running from trading volume to return lagged up to four hours.

Smirlock and Starks (1988) investigated the lagged relation between price changes and volume by employing transaction individual daily stock price and volume data on New York Stock Exchange for 49 consecutive trading days from 15th June through 21st August 1981. Using linear Granger causality tests, they documented a strong positive lagged relation between absolute price changes and volume. Linear causality tests however showed no predictive power for lagged volume for returns per se. According to Smirlock and Starks, the relationship was not entirely contemporaneous since lagged volume has forecasted absolute returns. This relationship would be even more significant in short periods preceding and immediately following firm's announcement of quarterly earnings. The results illustrated that information arrival to investors tended to follow a sequential process as suggested by sequential information arrival hypothesis.

Blume, Easley and O'Hara (1994) was first to examine the informational role of volume by developing a new equilibrium model in which traders can learn valuable information about a security through observing both past price and past volume information. In their model, the aggregate supply was fixed and traders would receive signals with differing quality. They proved the relation of volume, information and price movements, as well as demonstrated the case that sequences of volume and prices can be informative. Traders who used information contained in market statistics did better than those who did not. Overall, the volume captures the important information contained in the quality of the traders' information signals, which price alone does not. This is due to the fact that the statement of stock prices are noisy and it cannot convey all available information to market participants. Hence, volume could be used as an informative statistic to observe that signal. Basically, Blume et al. (1994) have shown that lagged volume could be useful to predict price movements and they too supported a positive correlation between price changes and trading volume found by Smirlock and Starks (1988).

Two exercises in the price-volume area were carried out by Long, Payne and Feng (1999). They used daily index data for both A and B shares traded on the Shanghai market for the period from February 1992 to January 1994 and discovered a strong contemporaneous relationship between price change and volume for both A and B shares. This relationship is much stronger than those usually found in the United States (US). On the contrary, they found only weak evidence of causality in either direction after linear Granger causality tests.

Lee and Rui (2000) tested the causal relationships of daily data in trading volume, stock returns and return volatility in China's four stock exchanges, namely A and B indexes each of Shanghai and Shenzhen markets, and across China, US and Hong Kong stock markets for a sample period from December 1992 to December 1997. Using linear Granger causality test, they found little evidence of predictability of returns by volume or vice versa, either within the China domestic market or China in combination with the two overseas markets for a causal relation. They found that trading volume did not Granger-cause each of those China's domestic market returns. As for the cross-market causal relations, they discovered that US return helped in predicting returns of Shanghai A and Shanghai B stocks. US and Hong Kong volumes did not Granger-cause either return or volatility in China's stock markets. They concluded that information contained in returns, volatility, and volume from financial markets in the US and Hong Kong had very weak predictive power for China's financial market variables.

Chen, Firth and Rui (2001) meanwhile investigated the causal relations of return-volume for nine markets namely New York, Tokyo, London, Paris, Toronto, Milan, Zurich, Amsterdam, and Hong Kong using daily market index and trading volume series from 1973 to 2000 through bivariate linear Granger causality test. The results found a feedback system in Switzerland, the Netherlands, and Hong Kong and indicated that the returns in these three counties were influenced by volume and vice versa. Hence, trading volume has added some significant predictive powers for future returns in these three countries.

Lastly, Lee and Rui (2002) applied linear vector autoregression (VAR) model on daily data of three largest stock markets, namely New York, Tokyo, and London to test for the information content of volume besides examining the causal relations and the sign and magnitude of dynamic effects among stock market trading volume, returns and volatility for both domestic and cross-country markets. The S&P 500 index covered the period from 2nd January 1973 to 1st December 1999; Tokyo stock Price Index (TOPIX) covered from 7th January 1974 to 1st December 1999 and FTSE 100 index from 27th October 1986 to 1st December 1999. The findings suggested that trading volume did not Granger-cause stock market returns on these three stock markets. They also found that the New York trading volume did contain an extensive predictive power for London and Japanese financial market variables and there was a positive feedback relationship between trading volume and return volatility in all the three markets. For the cross-country relationships, US trading volume contained an extensive predictive power for United Kingdom and Japanese financial market variables.

In the case of emerging markets, there is a limited literature that concentrates mostly on Asian and Latin American stock markets. Moosa and Al-Loughani (1995) employed Granger causality tests to examine the price-volume relation for four emerging Asian stock markets. Using monthly aggregate price and volume data, they found that there was a strong evidence for causality running from volume to absolute price changes and from price changes to volume in the case of Malaysia, Singapore and Thailand. Furthermore, evidence from volume to price changes is also found in Singapore and Thailand. However, they reported no causality for the Philippines.

Saatcioglu and Starks (1998) tested the causal relation between price changes and volume on a set of emerging markets in Latin American, namely Argentina, Brazil, Chile, Columbia, Mexico and Venezuela. Data of monthly indices over the sample period from January 1986 to April 1995 in the local currency and in US dollars from the six Latin American stock markets were examined through OLS. They used a linear vector autoregression (VAR) analysis to test for Granger causality and found a causal relationship from volume to stock price changes but not vice versa in four of the six markets, namely Brazil, Colombia, Mexico and Venezuela. This was due to the different institutions and information flows to emerging markets rather than the developed markets.

Gurgul and Majdosz (2005) examined the predictive power of trading volume for stock return and return volatility using daily market index for New York, Frankfurt and Vienna over a period of ten years from January 1994 to August 2003. Applying bivariate linear Granger causality test, they found a significant feedback relationship between returns and trading volume and also from return volatility to trading volume in Polish stock market. For cross-country causality, their findings indicated an improvement in short-run forecasts of current and future trading volume of the Polish stock market due to the knowledge of past stock price movement in the German and US stock markets.

Table 2.2: Evidence from Prior Studies on Causal Relation between Trading Volume and Returns

Study

Sample Data

Time period

Methodology

Verdict (has linear cause-effect/             No cause-effect)

Rogalski (1978)

United States markets

Monthly data for ten stocks and their associated warrants from year 1968 to 1973

Linear filters and causality tests

Positive contemporaneous dependence at lag zero

Jain and Joh (1988)

New York Stock Exchange

Hourly common stock trading volume and returns from 1979 to 1983

Granger-Sims causality tests

Weak evidence of unidirectional relation running from trading volume to return lagged up to four hours

Smirlock and Starks (1988)

New York Stock Exchange

Daily stock price and volume data for 49 consecutive trading days from 15th June to  21st August 1981

Linear Granger causality tests

Strong positive lagged relation between absolute price changes and volume and no predictive power for lagged volume for returns per se

Long, Payne and Feng (1999)

Shanghai market

Daily index data for both A and B shares from February 1992 to January 1994

Linear Granger causality tests

Weak evidence of causality in either direction

Lee and Rui (2000)

China stock exchanges

Daily data of trading volume, stock returns and volatility from December 1992 to December 1997

Linear Granger causality test

Little evidence of predictability of returns by volume or vice versa

Table 2.2 (continued)

Chen, Firth and Rui (2001)

New York, Tokyo, London, Paris, Toronto, Milan, Zurich, Amsterdam, and Hong Kong

Daily price and volume indices  from 1973 to 2000

Bivariate linear granger causality test

Feedback system was found in Switxerland, the Netherlands, and Hong Kong

Lee and Rui

(2002)

New York, Tokyo, and London stock markets

Daily data from January 1973 to December 1999

Linear vector autoregression model

No Granger-cause relation found in all stock markets except New York

Moosa and Al-Loughani (1995)

Malaysia, Singapore and Thailand

Monthly aggregate price and volume data

Granger causality tests

Causality from volume to absolute price changes and from price changes to volume in Malaysia, Singapore and Thailand 

Causality from volume to price changes in Singapore and Thailand 

Saatcioglu and Starks (1998)

Argentina, Brazil, Chile, Columbia, Mexico and Venezuela

Data of monthly indices from January 1986 to April 1995

Linear vector autoregression model

Causal relationship from volume to stock returns found in Brazil, Colombia, Mexico and Venezuela

Gurgul and Majdosz

(2005)

New York, Frankfurt and Vienna

Daily market index from January 1994 to August 2003

Linear Granger causality test

Significant feedback relationship between returns and trading volume and from return volatility to trading volume in Polish stock market

The existence of nonlinear behaviour allows the development of richer and more representative models of asset behaviour. Antoniou, Ergul, Holmes, and Priestly (1997) incorporated possibility of nonlinearity and changes of regulatory in their study on Istanbul Stock Exchange. They listed five reasons to why nonlinearity may be observed in financial markets: (1) difficulties in carrying out arbitrage transactions; (2) nonlinear feedback adjustments, for instance, overreaction to bad news (Da Costa, 1994); (3) market imperfections like transactions costs; (4) differences in the frequency of important announcements relative to that of observations (5) irrationality of investors.

Gallant, Rossi and Tauchen (1992) and Hiemstra and Jones (1994) showed that volume and returns per se could have nonlinear linkages in addition to linear linkages. Both of them found a unidirectional linear causation from volume to stock returns but bidirectional nonlinear Granger causality between stock returns and volume, even after controlling for volatility persistence. Gallant, Rossi and Tauchen (1992) employed the nonlinear impulse response functions on NYSE by using daily S&P 500 stock index returns and trading volume data from 1928 to 1987. They found that more about the stock market could be learned through studying the joint dynamics of stock prices and trading volume than by focusing on the univariate dynamics of stock prices alone. They also found evidence for returns leading volume and concluded that volume did not forecast returns and large price movements were followed by high volume.

Heimstra and Jones (1994) studied the dynamic relationship between Dow Jones Industrial Average (DJIA) index returns and aggregate trading volume using linear and nonlinear Granger causality tests in NYSE from year 1915 to 1940. As a new finding to the literature, they found a significant positive bilateral nonlinear causal relation between stock returns and trading volume. They used nonlinear Granger causality test based on nonparametric estimators of temporal relations in time series, which was a modification of the model suggested earlier by Baek and Brock in 1992. The modified Beak and Brock test is a nonparametric test designed to detect linkages that cannot be uncovered by conventional linear test statistics. In addition, they found an evidence of significant bidirectional nonlinear causality between stock returns and trading volume.

Tauchen, Zhang, and Liu (1996) used dynamic impulse response analysis to investigate the interrelationship among price volatility, trading volume, and the leverage effect. They employed daily data of four active stocks listed on NYSE, namely Boeing, International Business Machine, Cocal-Cola, and Minnesota Mining and Manufacturing from 1982 to 1987. They found that volume responded nonlinearly to price shocks that could not be discovered by using standard symmetric parametric models. They also noted that volume did not affect the serial correlation properties of the stock price series.

Silvapulle and Choi (1999) noted that the richer types of asset behaviour such as large stock price swings and abrupt changes in stock market volatility can only be properly modelled with nonlinear structures. Silvapulle and Choi investigated the linear and nonlinear Granger causality between returns and trading volume by using daily closing prices and trading volume on the Korean Stock Exchange (KSE) for the period January 1980 to December 1994. They extended the study by dividing the total sample period into three sub-sample periods, those corresponding from January 1980 to December 1985, January 1986 to December 1989, and January 1990 to December 1994. They found evidence of significant bidirectional linear and nonlinear causality between stock returns and volume after controlling for volatility persistent in both series and filtering for linear dependence. The finding of the strong bidirectional stock price-volume causal relationships implied that knowledge of current trading volume did improve the ability to forecast stock prices. This evidence is not supportive of the efficient market hypothesis though. Anyway, another finding from Silvapulle and Choi was that the nonlinear relationship was sensitive to institutional, organisational, and structural factors.

Table 2.3: Evidence from Prior Studies on Nonlinear Impulse Response Functions

Study

Sample Data

Year

Sample

Verdict (has nonlinear cause-effect/ No cause-effect)

Gallant, Rossi and Tauchen

Daily S&P 500 stock index returns and trading volume data from 1928 to 1987

1992

New York Stock Exchange

No evidence of nonlinear causality from volume to returns

Heimstra and Jones

Dow Jones Industrial Average index returns and aggregate trading volume from 1915 to 1940

1994

New York Stock Exchange

Significant positive bidirectional nonlinear relation between stock returns and trading volume

Tauchen, Zhang, and Liu

Daily data of four stocks, namely Boeing, International Business Machine, Cocal-Cola, and Minnesota Mining and Manufacturing from 1982 to 1987

1996

New York Stock Exchange

Volume responded nonlinearly to price shocks

Silvapulle and Choi

Daily closing prices and trading volume from January 1980 to December 1994

1999

Korean Stock Exchange (KSE)

Significant bidirectional linear and nonlinear causality between stock returns and volume


2.3.3 Dynamic Relation between Trading Volume and Serial Correlation of Stock Return

Further efforts have been devoted to the relation between return dynamics and trading volume, indicating that the volume-return relation depends on trade volume as well as motives of traders. Epps (1975) who formalised and modelled two Wall Street was considered the first to contribute on speculative trading. Two old Wall Street adages are that volume makes prices move, and volume is relatively heavy in bull markets but light in bear markets.

Morse (1980) examined price and trading volume relationships that were consistent with the use of asymmetrically distributed information by using a model of investor demand. Morse (1980) used daily price and volume data for four-year period from 1973 to 1976 for 25 securities traded on NYSE and American Stock Exchange (ASE) and 25 securities traded over the counter (OTC). It was found that the investors with the private information would trade on this information until the price reflected the information. This process may occur over several days and lead to a joint occurrence of excessive trading and monotonic price movement. Therefore, he suggested that serial correlation of returns residual in high volume and high volume periods tended to have positive autocorrelated returns. Furthermore, the incentives to trade would be greater if the private information revealed that the price change would be large when price was adjusted. This would then lead to greater trading prior to and in participation of a large price change. Morse (1980) concluded that the findings were due to the existence of asymmetrical information in marketplace.

Kyle (1985) assumed there was a market maker who informed traders and liquidity traders in an auction-based model. He indicated that trading volume was increased in information asymmetry when liquidity trading was exogenous and inelastic to price. This was because informed traders attempted to exploit their private information. Nevertheless, Admati and Peiderer (1988) as well as Foster and Viswanathan (1990) used the auction-based model of Kyle and provided evidence that trading volume can decrease in information asymmetry if liquidity traders had timing discretion. According to them, when discretionary liquidity traders received exogenous trade demands prior to announcements, they would postpone trading until the announcement was made and the information asymmetry would resolve and eventually the traders would not bear the trading costs with informed traders. Therefore, total volume can decrease before announcements and increase correspondingly after that.

LeBaron (1992a), on the other hand, tested the relation between serial correlation and volatility of market return on NYSE from 1928 to 1991 at daily and weekly frequencies using GARCH (1,1) process. He found that serial autocorrelations were time-varying and were related to stock return volatility.

Sentana and Wadhwani (1992) used EGARCH model and nonparametric methods to present evidence on the links between volatility and returns autocorrelations based on both NYSE hourly data for the period of October 1987 crash and daily data from 1885 to 1988. They reported that when volatility was low, stock return at short horizon exhibited positive serial correlation and vice versa. They attributed this to the effect of nonsynchronous trading and application of feedback strategies of traders.

Almost simultaneously, Duffee (1992) began to establish the relation between serial correlation and trading volume in aggregate monthly market returns and trading volume data for a 75-year period from 1915 to 1989. A significant relationship between NYSE volume shocks and return reversals was found. The findings showed that a one-standard-deviations shock to trading volume has caused a reversal of 40 per cent to 50 per cent of stock return.

Wang (1993) tested a theoretical equilibrium model of asset-pricing model under information asymmetry that involved investors who were separately informed about the state of economy. He found that stock prices became more variable when investors had less information about future cash flows and the informed investors could destabilise the price. He commented that information asymmetry can increase the price volatility and negative serial correlation in returns. It was demonstrated that large trading volume tended to induce negative return autocorrelations when the main motive of traders was liquidity and these autocorrelations would be positive if speculation was the primary motive of traders.

Campbell, Grossman, and Wang (1993) applied linear and nonlinear regressions on daily data for the sample period from July 1962 to December 1988 to investigate the relationship between aggregate stock trading volume and the serial correlation of daily stock returns in NYSE and ASE. They found that for both stock indexes and individual large stocks, the first-order daily return autocorrelation tended to decline with volume and a stock prices decline on a high-volume day was more than a stock price decline on a low-volume day which was then associated with an increase in the expected stock return. They assumed that the demand curves for stocks were not perfectly elastic and the price changes could be affected by informational and non-informational trading. If there was a general agreement regarding the new valuation of the security from the information disseminated to the public, a price change may occur to reflect the news with low trading. Conversely, price changes due to noninformational traders have resulted in heavier volume due to an imbalanced relation between the intrinsic value and market value of a stock. Hence, investors' need for liquidation purpose could cause a significant drop or rise at stock price in a period of heavy trading. As a result, according to Campbell et al. (1993), the prices had to be reversed in the following day as no new information was revealed and the intrinsic value of the stock has not changed since hedging motivated trades and informational motivated trades accompanied with low trading volume should not revert. Both Duffee (1992) and earlier Gallant, Rossi and Tauchen (1992) reported intuitively similar empirical results in the US markets.

Wang (1994) analysed dynamic relations between volume and returns based on a theoretical equilibrium model of stock trading in which investors are heterogeneous in their information and private investment opportunities, and trade rationally for both informational and noninformational reasons. He introduced information asymmetry as an additional motive to trade apart from liquidity motive. He showed that informational trading and noninformational trading could lead to different dynamic relations between trading volume and stock returns. He noted that trading was always accompanied by price changes since investors were risk-averse. For instance, when a group of investors sells their shares to rebalance their portfolio and to induce other investors to buy, the price of the stock will drop. As information asymmetry increases, the uninformed investors demand higher discounts in price when they buy the stock from the informed investors. Thus, these investors are able to cover the risk of trading against private information. Therefore, trading volume may provide information about expected future returns and is always positively correlated with absolute prices changes, and the correlation increases with information asymmetry gradually.

Cooper (1999) tested a contrarian strategy using different filters on past returns and lagged volume changes based on a sample of top 300 largest market capitalisation NYSE and AMEX stocks from 2nd July 1962 to 31st December 1993 at weekly horizons. He found that highly traded stocks showed greater price reversal and positive return autocorrelation. He supported the asymmetric information model of Wang (1994) in which price continuations were accompanied by high trading volumes when informed investors conditioned their trades on private information.

Rational expectations models show disagreements generated by private information. These models generally involve trading among privately informed traders, uninformed traders, and liquidity or noise traders. He and Wang (1995) used a multiperiod rational expectations model of stock trading in which investors had different information concerning the underlying value of the stock. They showed that the volume pattern over time was closely related to the flow and the nature of private and public information. Based on the information, investors traded for either to accommodate supply shocks or to speculate on the future price changes. According to them, private information led to current and future trading whereas public information generated current trading. In addition, volume generated by new private or public information was always accompanied by large price changes and this did not happen to the volume generated by existing information.

Suominen (2001) introduced a model in which private information about equity returns is available in any given period with some probabilities that changes stochastically over time. He developed a market microstructure model where trading volume was used by uninformed traders as a signal for private information in the market and could therefore help in overcoming information asymmetries. He indicated that trading volume not only described market behaviour but also actually affected it since it directly entered into the decision process of market participants. Traders estimated the availability of private information using lagged volume and modified their trading strategies as the probability of private information entering the market increased. Therefore, the trades of informed traders have revealed to private market the information that has affected the prices volatility. A strong contemporaneous and causal relationship between volume and return volatility was derived. This implied that volume conveys information to the market that cannot be obtained from price alone because prices are noisy.

Chordia and Swaminathan (2000) examined the interaction between trading volume and the predictability of short-term stock returns over a period from 1963 to 1996 using VAR and Dimson beta regressions. The findings indicated that trading volume was a significant determinant of lead-lag patterns observed in stock returns. Daily and weekly returns of stocks with high trading volume significantly led to daily or weekly returns stocks with low trading volume in NYSE and ASE. High volume stocks tended to respond promptly to market-wide information and differential speed of adjustment to information was a significant source of the cross-autocorrelation in short horizon stock returns. The results indicated that trading volume played a significant role in the dissemination of market-wide information and higher trading volume helped prices to reflect quicker full information.

Apart from that, Gervais, Kaniel and Mingelgrin (2001) used daily data on NYSE stocks from August 1963 to December 1996 to study the role of trading activity in explaining future prices. Using methods of average return of zero investment and reference return portfolio formation strategies, they provided evidence that a high trading activity had an informational content on future returns and they explained that price run-ups following large changes in trading volumes were consistent with the visibility hypothesis. They acknowledged this effect as high-volume return premium. The unusual high (low) trading volume over a day or a week tended to appreciate (depreciate) over 20 to 100 trading days regardless the firm size.

Last but not least, Llorente, Michaely, Saar, and Wang (2002) investigated the volume-induced return autocorrelation with the level of disclosed insider trading using a theoretical model with heterogeneously informed agents over a sample period from 1st January 1993 to 31st December 1998. Using 2226 individual firms listed on NYSE and AMEX, they found that daily returns generated by risk-sharing trades tended to reverse themselves and those generated by speculative trades tended to continue themselves. The results showed that dynamic return and volume were measures of information asymmetry. The dynamic volume-return was dependent on the investors' motive to trade. The information asymmetry based speculation motive in addition to hedging was established.

Table 2.4: Evidence from Prior Studies on Trading Volume and Autocorrelation of Returns

Study

Sample Data

Year

Sample

Verdict

Morse

Daily price and volume for a period from 1973 to 1976

1980

New York Stock Exchange, American Stock Exchange, and Over-the-Counter

Serial correlation of returns residual in high volume and high volume periods tended to have positive autocorrelated returns

LeBaron

Daily and weekly data from 1928 to 1991

1992a

New York Stock Exchange

Autocorrelation of daily stock returns changed with the variance of returns 

Santana and Wadhwani

Hourly data for the period of October 1987 crash and daily data from 1885 to 1988

1992

New York Stock Exchange

When volatility was low, stock return at short horizon exhibited positive serial correlation and vice versa 

Duffee

Aggregate monthly market returns and trading volume data from 1915 to 1989 

1992

New York Stock Exchange

Significant relationship between volume shocks and return reversals

Campbell, Grossman, and Wang

Daily stock indexes and individual stocks from July 1962 to December 1988

1993

New York Stock Exchange and American Stock Exchange

First-order daily return autocorrelation tended to decline with volume for stock indexes and individual large stocks

Stock prices decline on a high-volume day was more than stock prices decline on a low-volume day

Cooper

Weekly volumes and returns of 300 stocks from 2nd July 1962 to 31st December 1993

1999

New York Stock Exchange and American Stock Exchange

Highly traded stocks showed greater price reversal and positive return autocorrelation

Table 2.4 (continued)

Chordia and Swaminathan

Daily and weekly returns and trading volume of stocks

from 1963 to 1996

2000

New York Stock Exchange and American Stock Exchange

High volume stocks tended to respond promptly to market-wide information

Gervais, Kaniel and Mingelgrin

Daily data from August 1963 to December 1996

2001

New York Stock Exchange

High trading activity had an informational content on future returns

Llorente, Michaely, Saar, and Wang

Daily returns and trading volume of 2226 individual forms from 1st January 1993 to 31st December 1998

2002

New York Stock Exchange and American Stock Exchange

Stocks with high proportion of private information trading volume exhibited return continuations


2.4 Conclusion

This chapter has provided a review of the theories and empirical evidence on the information and price movement, efficient market hypothesis, mixture of distributions hypothesis, sequential information arrival hypothesis, information asymmetry, contemporaneous and causality models, and volume-autocorrelation return relation in both developing and developed stock markets. To this extent, the empirical findings have shown that on the existence of contemporaneous and causal relation between trading volume and return incorporated with information, it is clear that markets are not informationally efficient.


[10] First appearance of martingale was in Bernstein (1926) under the name of 'variables encha�n�es'. The word 'martingale' was first used in probability and random processes as the name of a betting system. A martingale is the gambling system of "double-or-nothing". It doubles the gambler's stake every time he loses in order to recoup the loss. The fact following this gambling system that neither helps nor hurts the player's expected fortune is a particular case of basic theorem in martingale theory. Martingales process is also known as diffusion process.

[11] George Akerlof, a winner of the Nobel Prize in Economics in 2001 for his essay entitled 'The Market for Lemons" with analyses of markets with asymmetric information in a "lemons" market phenomenon. He illustrated information asymmetry in the sale of used cars, and subsequently introduced the Lemons Principle in 1970. Spence (1973) and Grossman and Stiglitz (1980) further explored the area of information asymmetry. Spence pointed out the possibility of different expectation-based equilibria for gender and races in education and pay aspects. Grossman and Stiglitz analysed efficiency on financial markets. Their key result was known as the Grossman-Stiglitz paradox that says if a market is informationally efficient in which all relevant information is reflected in market prices, then no single agent will have sufficient incentive to acquire the information on which prices are based.

Source: Essay UK - http://turkiyegoz.com/free-essays/finance/distribution-of-stock-price.php


Not what you're looking for?

Search:


About this resource

This Finance essay was submitted to us by a student in order to help you with your studies.


Rating:

Rating  
No ratings yet!

  • Order a custom essay
  • Download this page
  • Print this page
  • Search again

Word count:

This page has approximately words.


Share:


Cite:

If you use part of this page in your own work, you need to provide a citation, as follows:

Essay UK, distribution of stock price. Available from: <http://turkiyegoz.com/free-essays/finance/distribution-of-stock-price.php> [19-12-18].


More information:

If you are the original author of this content and no longer wish to have it published on our website then please click on the link below to request removal:


Essay and dissertation help

badges

 
Adelheid und ihre Mörder | Crest of the Stars (13) | Akcji / Strzelaniny