INTRODUCTION
Problem statement
The stock market serves as a barometer for a country's economic health, influencing not only foreign exchange and gold markets but also credit and options markets A robust stock market often prompts investors to convert foreign currencies and gold into cash for stock investments, leading to a depreciation of these assets Conversely, a weak stock market results in tighter cash flow from the government and reduced bank disbursement for stocks Additionally, a strong stock market fosters growth in various options as investors anticipate higher profits.
In Vietnam, the stock market serves as a crucial indicator of the evolving economy, with the VN-index reflecting significant economic changes For example, announcements of rising inflation or bad debts typically lead to a sharp decline in the VN-index Conversely, when the government implements supportive economic policies, the VN-index tends to rise This fluctuation in the VN-index highlights the bi-directional relationship between economic information and stock market performance, signaling periods of growth or recession in the economy.
In mean-variance analysis, investors focus on expected stock returns and return volatility, as these factors indicate the risks associated with their portfolios Additionally, trading volume plays a crucial role in reflecting market information; when investors anticipate higher returns, trading activity increases, leading to higher trading volumes Conversely, expectations of lower returns result in decreased trading volume Therefore, fluctuations in trading volume can serve as indicators of changes in stock returns.
The relationships between stock returns, trading volume, and return volatility are critical in empirical research Numerous studies, such as Karpoff (1987), highlight a positive asymmetric relationship between trading volume and price changes in the equity market Epps (1975) and Jennings et al (1981) further support this asymmetric relationship, linking it to the flow of information Granger et al (1964) and Granger (1968) analyzed data from the New York Stock Exchange, revealing that price changes follow a random walk, indicating that past stock price trends do not predict future movements Additionally, Mehrabanpoor et al (2005) found a positive correlation between market turnover and indices on the Tehran Stock Exchange.
In addition, Michael Long (2007) finds a significantly positive interaction between absolute value of call price changes and trading volume in the option markets
The relationship between volume and return volatility has been extensively studied, beginning with Engle's (1982) Autoregressive Conditional Heteroskedasticity (ARCH) Model, which suggests that stock returns follow a mixture of distributions Bollerslev (1986) expanded this with the Generalized Autoregressive Conditional Heteroskedasticity (GARCH) Model, treating trading volume as an indicator of information flow Subsequent research by Lamoureux and Lastrapes (1990), Brailsford (1996), and others has explored this dynamic further Brailsford (1996) posits a positive correlation between stock return volatility and trading volume, while contrasting studies by Fujihara and Mougoue (1997) and Mestel et al (2003) argue against a causal relationship, suggesting it is too weak to be predictive Okan et al (2009) found that trading volume negatively impacts return volatility using various models In the Vietnamese market, Truong Dong Loc (2009) identified a unilateral causality from the HNX-index to trading volume, a finding replicated by Truong Dong Loc and Dang Thi Thuy Duong (2011) with foreign trading volume data, revealing that the index influences net foreign volume without reciprocal effects.
This thesis addresses the limited research on the relationship between stock returns and trading volume during recent crises, specifically considering the impact of foreign trading volume across various industries using the GARCH model By focusing on listed companies on the Ho Chi Minh City Stock Exchange, this study investigates the effects of trading volume on stock returns and return volatility, employing the GARCH (1, 1) model for analysis.
Research questions
This study aims to clarify the relationships between trading volume, stock returns, and return volatility by analyzing intra-day stock price data The primary objective is to address key research questions regarding these financial metrics.
(i) Is there the relationship between trading volume and return volatility? and (ii) Does trading volume cause stock returns?
Research objectives
To reach above aims, my objectives of the paper are:
(1) To examine the relationship between trading volume and return volatility and,
(2) To understand the impact of trading volume on stock return through GARCH (1,1) model
This study analyzes data from eight publicly traded companies on the Ho Chi Minh City Stock Exchange (HOSE), focusing on their performance before and after the recession caused by the US sub-prime mortgage crisis.
This paper is structured as follows: Section Two provides a concise literature review of empirical studies, while Section Three describes the Vietnamese stock market and highlights its unique data characteristics, including statistics on selected stocks in the HOSE Section Four outlines the methodology and discusses the empirical results, culminating in the Conclusion.
LITERATURE REVIEW
The Efficient Market Hypothesis
The Efficient Market Hypothesis (EMH), introduced by Fama in 1970, is a key theory in financial time series analysis, widely accepted by modern economists It posits that financial markets are highly efficient in reflecting information regarding individual stocks and the overall market According to the EMH, stock prices incorporate all available information, representing investors' beliefs about future performance Consequently, information flows are quickly and effectively integrated into stock prices, meaning that current prices cannot be used to predict future price movements.
According to Keith and Dirk (2005), the Efficient Market Hypothesis (EMH) comprises three versions: weak, semi-strong, and strong forms In weak-form efficiency, stock prices reflect all publicly available historical information, highlighting that past data is the most readily accessible information in the stock markets.
In the semi-strong form of market efficiency, stock prices immediately reflect all available public information, including past prices and relevant data such as financial statements, dividend announcements, mergers and acquisitions, and macroeconomic factors like inflation and interest rates This implies that the current market price serves as the best predictor of a fair stock price, meaning that no investor can gain an advantage by trading on information already understood by others Overall, the semi-strong form is considered more robust than the weak form of market efficiency.
In the strong form of market efficiency, current stock prices swiftly incorporate all available public and private information about companies, making it impossible for investors to consistently identify undervalued stocks or for insiders to gain an advantage Consequently, the prevailing market price serves as the most accurate predictor of a stock's fair value.
Numerous studies support the concept of market efficiency, including research by Thomas E Copeland and Daniel Friedman (1987), who analyzed price behavior, trading volume, and portfolio composition in relation to information arrival on NASDAQ Their comparison of price, volume, and share allocations across three market equilibrium models—telepathic rational expectations, ordinary rational expectations, and private information—revealed that stronger forms of market forecasts are more effective at predicting equilibrium prices than weaker ones.
Maloney and Mulherin (2003) examine the relationship between stock returns and trading volume during the Space Shuttle Challenger crash, focusing on four firms Their findings indicate that price movements occur even during trading interruptions or periods of low trading profits However, they are unable to identify a definitive method by which informed traders can influence these price movements.
While the Efficient Market Hypothesis (EMH) proposed by Fama (1970) has been widely accepted, several researchers challenge its validity Grossman and Stiglitz (1980) argue that true informational efficiency is unattainable due to the high cost of information, which prevents stock prices from fully reflecting available data LeRoy and Porter (1981) highlight the excessive volatility in stock markets as evidence against market efficiency Similarly, Marsh and Merton (1986) reject the EMH, while Summers (1986) critiques the speculative nature of stock markets, suggesting it undermines rational valuation and efficiency In 1990, Laffont and Maskin also dispute the EMH in the context of imperfect competition, and Jegadeesh presents strong evidence against the random walk hypothesis, indicating predictability in stock returns More recently, Lee et al (2010) assert that the stock market lacks efficiency concerning the stationarity of stock prices.
Karpoff (1987) conducts a comprehensive survey of previous studies across various markets to explore the relationship between price changes and trading volume He divides his research into two segments: one focusing on the correlation between absolute price changes and trading volume, and the other on the relationship between price changes and trading volume His findings indicate a positive correlation between trading volume and price changes in the equity market Additionally, he suggests that there may be an asymmetric relationship between price change and volume, challenging the common use of linear models in this area of research.
1 Random Walk Hypothesis: is consistent with the Efficient Market Hypothesis necessary to explain the Mixture of Distribution Hypothesis (Clark, 1973) and the Sequential Information Arrival Hypothesis (Copeland, 1976).
The Mixture of Distributions Hypothesis
The Mixture of Distributions Hypothesis (MDH), proposed by Clark (1973), posits that price changes and trading volume are influenced by the same information flow, leading to a correlation between volume and volatility In this model, return volatility is characterized by a random number of intra-day fluctuations, while trading volume serves as a proxy for the rate of information flow, impacting stock returns The hypothesis suggests that both returns and volume respond simultaneously to new information, indicating a positive correlation between stock prices and trading volume, as the variance in stock price during transactions is dependent on trading volume.
The MDH, developed by Epps and Epps (1976), Tauchen and Pitts (1983), Lamoureux and Lastrapes (1990), and Andersen (1996), explores the relationship between price changes and trading volume Epps and Epps (1976) propose that the logarithmic change in price can follow a mixture of distributions, with trading volume serving as a mixing variable Andersen (1996) enhances this hypothesis by suggesting that trading volume is uncorrelated with the flow of information, emphasizing that volume results from noise or liquidity rather than information arrival In contrast to Tauchen and Pitts (1983), who assume an independent and identically distributed (i.i.d.) arrival process of information, Andersen (1996) indicates that the rate of information arrival is serially correlated, revealing a positive correlation between lagged volumes and volatilities with current volatilities and volumes.
2 i.i.d means independent and identically distributed
Chen and Tse (1993), Omran and McKenzie (2000), Zarraga (2003), Pyun et al (2000), and Bohl and Henke (2003) find supportive evidence from Japanese, UK, Spanish, Korean, and Polish stock markets, respectively
Huson Joher Ali Ahmed, Ali Hassan, and Annuar M.D Nasir (2005) investigate the volatility of the Kuala Lumpur Stock Exchange using the Mixture of Distribution Hypothesis Their study employs the GARCH (1,1) model, demonstrating that this model effectively analyzes return volatility Additionally, they incorporate trading volume as an explanatory variable, revealing that the persistence of volatility remains unchanged with the inclusion of volume.
The Market Dynamics Hypothesis (MDH) faces criticism for its failure to account for volatility in trading volume, which limits its ability to demonstrate volatility persistence when trading volume is included as an explanatory variable Research by Fong (2003) and Xu et al (2006) indicates that the MDH does not adequately address serial dependence concerning return volatility and volume Additionally, a study by B.M Nowbutsing and S Naregadu (2009) analyzing thirty-six stocks and the Stock Exchange of Mauritius (SEM) Index reveals a weakly positive relationship between trading volume and volatility using the ARCH-in-mean model Consequently, the findings do not support the MDH or the Stock Information Asymmetry Hypothesis (SIAH) within the context of the SEM.
A study by Brajesh Kumar, Priyanka Singh, and Ajay Pandey in 2009 analyzed the relationship between price and trading volume using data from 50 Indian stocks The researchers applied the GARCH model to examine the link between conditional volatility and trading volume Their findings revealed a positive asymmetric relationship between volume and unconditional volatility However, the results provided mixed support for the Mixture of Distributions Hypothesis (MDH), neither fully rejecting nor unconditionally confirming it.
The Sequential Information Arrival Hypothesis
The Sequential Information Arrival Hypothesis (SIAH), introduced by Copeland in 1976 and further developed by Jennings et al in 1981, posits that new information is disseminated to informed and uninformed traders in a sequential manner rather than simultaneously This allows informed investors to capitalize on information before others, influencing their portfolio decisions The SIAH suggests a positive bi-directional causality between the absolute value of returns and trading volume, where past returns can predict current trading volume and vice versa Supporting this theory, research by Harris (1987) and Smirlock and Starks (1988) indicates a positive correlation between changes in trading volume and stock prices on the NYSE.
(1994) contend that a flow of sequential information enables lagged trading volume to predict current absolute returns (price changes) and enables lagged absolute returns to predict current trading volume
The MDH and SIAH both explore the connection between trading volume and price changes, but they differ in their approaches The MDH suggests a simultaneous relationship, indicating that price changes and trading volume occur together In contrast, the SIAH presents a sequential dynamic relationship, where past price changes can predict current trading volume and vice versa, as noted by Darrat et al (2003).
The Generalized Autoregressive Conditional Heteroskedasticity
Recent studies under the MDH theory have proposed various models to explore the relationship between price and volatility Engle (1982) introduced the Autoregressive Conditional Heteroskedasticity (ARCH) Model, which is based on the idea that stock return time series originate from a mixture of distributions rather than a single distribution This model is particularly suitable for financial data, especially time series data, as it considers the arrival of information as a stochastic mixing variable, with daily returns representing a mixture of distributions.
In 1986, Bollerslev introduced the Generalized Autoregressive Conditional Heteroskedasticity (GARCH) model, which, alongside the Autoregressive Conditional Heteroskedasticity (ARCH) model, effectively addresses the heteroskedasticity present in time series data These models are essential for measuring portfolio volatility and analyzing asset prices Notably, the GARCH model utilizes trading volume as a proxy for market information flow, enabling a comprehensive analysis of market volatility.
The model developed by Lamoureux and Lastrapes (1990), Brailsford (1996), and Mestel, Gurgul, and Majdosz (2003) focuses on limiting the conditional variance of time series by utilizing past squared residuals Lamoureux and Lastrapes (1990) specifically analyze the volume-return volatility relationship for active US stocks, revealing that daily price variance is heteroskedastic and positively correlated with the arrival of information Their mixed model results indicate that incorporating trading volume into the conditional variance equation reduces the ARCH effect and the persistence of volatility However, they also note that the inclusion of lagged volume in the variance equation is largely insignificant across most cases.
Najand and Yung (1991) use treasure bond futures to analyze as Lamoureux and Lastrapes
(1990) conduct They report that the lagged volume explains volatility better than contemporaneous volume
Timothy J Brailsford (1996) explores the positive relationship between price changes and trading volume using three measures of volume, focusing on the top eight stocks in the Australian Stock Market by market capitalization Employing the GARCH model's conditional variance equation, he finds that the slope of volume and volatility is less for negative returns compared to positive returns, indicating an asymmetric relationship When trading volume is treated as an exogenous variable, the significance and magnitude of the GARCH coefficients diminish This study serves as a foundational analysis of the volume-price change relationship, paving the way for future research in this area.
Research by Roland Mestel, Henryk Gurgul, and Pawel Majdosz (2003) indicates a weak contemporaneous relationship between stock return and trading volume, suggesting that they do not effectively predict one another Conversely, there is a strong correlation between return volatility and trading volume, with volatility serving as a predictor of volume in certain instances.
Safi Ullah Khan and Faisal Rizwan (2008) analyze the connection between trading volumes and return volatility using the GARCH (1,1) model Their study, based on data from the Karachi Stock Exchange (KSE-100 index), reveals a positive contemporaneous relationship between volume and volatility, even after accounting for heteroscedasticity.
Sarika Mahajan and Balwinder Singh (2009) analyze daily data from the Sensitive Index (SENDEX) of the Bombay Stock Exchange, India’s premier stock exchange, covering the period from October 1996 to March 2006 Their research focuses on examining the interrelationship between stock return, trading volume, and market volatility.
The empirical findings reveal a significant positive interaction between trading volume and return volatility, supported by both the Market Dynamics Hypothesis (MDH) and the Stock Information Acquisition Hypothesis (SIAH) Utilizing the GARCH (1,1) model, the study demonstrates a reduction in variance persistence when trading volume is incorporated as a proxy for information flow in the conditional volatility equation.
Tarika Singh and Seema Mehta (2010) explore the connection between trading volume and stock return volatility in Asian markets, utilizing the GARCH (1,1) model to assess volatility and forecast individual stock returns Their findings indicate that all parameters in the return volatility equation are statistically significant, highlighting a clear impact of recession as demonstrated through their figures.
Pratap Chandra Pati and Prabina Rajib (2010) analyzed the persistence of volatility in National Stock Exchange S&P CRISIL NSE Index Nifty Index futures using GARCH and ARMA-EGARCH models Their findings reveal that incorporating trading volume into the GARCH model reduces the persistence of volatility; however, the GARCH effect does not completely vanish.
Chen et al (2001) present findings that contrast with those of Lamoureux and Lastrapes (1990), revealing that incorporating contemporaneous trading volume into the GARCH model does not eliminate volatility persistence.
Numerous empirical studies have explored the relationship between trading volume, stock returns, and price changes, revealing both positive and negative correlations in both developed and emerging markets However, limited research exists on the effects of foreign buy and sell volume, as well as pre- and post-recession impacts, specifically within the GARCH (1,1) model framework on the Ho Chi Minh Stock Exchange Consequently, this study aims to further investigate the connections between trading volume and returns, as well as the relationship between volatility and volume, utilizing data from eight listed companies on the Ho Chi Minh stock market.
VIETNAMESE STOCK MARKET AND LISTED COMPANIES ON
Vietnamese stock market
Numerous studies on the Vietnamese stock market have been conducted by both domestic and international researchers Since the implementation of the "Doi Moi" policy, Vietnam has actively opened and developed its economy, making significant policy adjustments to integrate globally and achieve impressive economic growth As a result, the country's GDP growth rate has averaged around 7 percent annually over the past decade, underscoring the vital role of financial development in this economic progress.
As a result, Vietnamese stock market is born to meet requirement of the economy (Vuong Thanh Long, 2008)
The stock market in Ho Chi Minh City was established in July 2000, initially featuring only two listed stocks During its first two years, trading took place sporadically, occurring only on select days.
As of 2010, the Ho Chi Minh Stock Exchange (HOSE) features 247 listed companies with a market capitalization of approximately $28.28 billion In contrast, the Hanoi Securities Trading Center (HASTC), established by Decision No 127/1998/QĐ-TTg on July 11, 1998, is smaller in scale The Hanoi Stock Exchange (HNX) was created in 2009 through Decision No 01/2009/QĐ-TTg on January 2, 2009, as part of the restructuring of the HASTC by the Prime Minister of Vietnam.
Between 2000 and 2003, the stock market experienced significant fluctuations, starting at 571 points in June 2001 and dropping to 139 points by April 2003 However, from 2004 to 2005, the VN Index rebounded, rising from 213 to 307 points.
Between 2006 and 2009, the VN Index achieved a significant peak of 1,167 points in February 2007, marking an unforgettable moment for investors However, this milestone was followed by a sharp decline in the index later that month.
2009 at 235 points and moves up to 480 on December 31, 2010 (Nguyen Thi Kim Yen,
The stock market has experienced significant growth, with the number of listed companies rising from 164 on the HOSE and 154 on the HNX in 2008 to a total of 627 by 2010, which includes companies across both exchanges and fund management firms (Nguyen Thi Kim Yen, 2011).
Some summarized numbers of the year 2010 as follow:
Market capitalization (USD billion) 35 Average daily trading value (USD million) 80
On May 11, 2007, the Prime Minister signed Decision No 559/2007/QD-TTg, officially transforming the Ho Chi Minh Securities Trading Center into the Ho Chi Minh Stock Exchange (HOSE) As of December 31, 2011, stocks listed on HOSE must adhere to specific criteria to ensure compliance and maintain market integrity.
- The listing companies are joint stock companies with paid-up capital at the time of registration for listing at least VND80 billion at book value
- Business operation in two years before the year of listing has to be profitable and has no accumulated losses up to the year of registration for listing
All debts owed to the company by members of the Board of Directors, Board of Supervisors, General Manager, Deputy Manager, Chief Accountant, and major shareholders are fully disclosed, with no overdue debts that are not in compliance with regulations.
To ensure broad ownership, at least 20% of voting shares in a company must be held by a minimum of 100 shareholders who are not classified as professional investors or major shareholders, unless the company is a state-owned enterprise transitioning into a joint stock company.
- Officer-shareholders have to commit to hold 100 percent of their shares in 6 months from the date of listing and 50 percent of them for the following 6 months
This study analyzes data from listed companies on the Ho Chi Minh City Stock Exchange (HOSE) to explore the relationships influenced by HOSE's stringent requirements compared to other exchanges Additionally, it investigates the period from 2007 to 2011 to assess the economic impact on the stock market.
Data
Database includes daily price changes and trading volume of eight stocks listed on the Ho Chi Minh City Stock Exchange
Table 1: Description of stocks (on 17 th Jul 2012)
No Stock Name of Company Sector Chartered capital (mil)
No Stock Name of Company Sector Chartered capital (mil)
5 VIC Vingroup J.S.C Real estates 7,004,621 07 Sep 2007 55,336,633 37.3% 16.2%
6 FPT FPT Corporation IT and
Hinh Hydropower Joint Stock Company
I focus on information from listed companies due to their provision of transparent and reliable data for investors I have selected eight stocks that represent specific industries within the economy, chosen based on high market capitalization, listing duration, firm scale, liquidity, and their significant impact on the VN-index for analysis.
The VN30 index, established by the Ho Chi Minh City Stock Exchange, comprises the 30 stocks with the most significant influence on the VN-index, accounting for approximately 80% of total market capitalization and 60% of total traded value, while also ensuring high liquidity From this selection, I identified 17 stocks as of July 17, 2012, which I categorized into eight distinct industries: processing and manufacturing, mining, finance, consumer goods, real estate and construction, telecommunications and information, transport and warehouse, and electricity Ultimately, I selected eight representative stocks, each from a different industry, based on their market capitalization and impact on the VN-index Notably, I chose SSI for its ability to reflect overall market trends, despite its lower market capitalization compared to STB (Saigon Thuong Tin Commercial J.S.C).
DPM (Petro Vietnam Fertilizer and Chemicals Company) is established in 28 March
Founded in 2003 with a chartered capital of 3,800 billion, DPM specializes in fertilizer production, boasting an annual capacity of 740,000 tons The company also produces ammonia at a daily capacity of 1,350 tons and urea at 2,200 tons per day, while trading liquid ammonia with an annual capacity of 96,000 tons DPM meets 40% of the national fertilizer demand and holds a 50% market share in the South and Central South regions of Vietnam.
PVD (Petro Vietnam Drilling and Well Services Corporation) is established in 1994
The company specializes in contract drilling, well maintenance, and providing conditioning services to petroleum production firms and energy service sectors, including geotechnical and logging services It also focuses on mapping oil fields, controlling oil spills, and leasing drilling equipment and oil rigs As a subsidiary of the Vietnam National Oil and Gas Group (Petro Vietnam), the company operates through three joint ventures in oil exploration and production, alongside six subsidiaries managing various operations.
Established in 1999, Saigon Securities Inc (SSI) is a leading securities firm in Vietnam, offering a range of services including brokerage, portfolio management, and corporate advisory With a commanding 17% market share, SSI stands as one of the largest companies in the industry The firm boasts a significant foreign client base, serving over 100 institutions and managing 2,500 individual accounts, which collectively represent 30% of the market.
VNM (Vietnam Dairy Products J.S.C), established in 1976, offers a diverse range of dairy products, including milk, powdered milk, solid milk, yoghurt, ice cream, fruit juice, and coffee With a robust distribution network of over 1,000 agencies nationwide, VNM also exports its products to key markets such as the US, Germany, Canada, and China The company operates more than 70 stations for fresh milk collection, processing over 260 tons daily, which accounts for 80% of Vietnam's fresh milk supply Additionally, VNM has invested in the construction of 60 fresh milk processing facilities to enhance its production capabilities.
Vincom Corporation (VIC), established in 2002, specializes in the development and management of real estate projects, focusing on commercial and entertainment spaces The company has successfully executed numerous large-scale projects across the country, including the Vincom Twin Towers, comprehensive trade center complexes, and the Vincom Hai Phong Plaza, which features a vibrant trade complex.
FPT (FPT Corporation) was established in 1988 FPT is a multinational company, mainly operates in information technology, telecommunication, distribution, real estates, education, and financial activities FPT has 15 subsidiaries and affiliates, 53 branches, 396
FPT boasts an extensive network of 560 cell phone distributors across the country and collaborates with 60 renowned partners, including industry leaders like IBM, Lenovo, Microsoft, HP, Nokia, Toshiba, Oracle, Samsung, Motorola, Veritas, Apple, and Intel.
Established in 1990, GMD (General Forwarding & Agency Corporation) was a state-owned company that offers comprehensive forwarding and logistics services across the nation and its neighboring regions With a focus on leveraging scale expansion, extensive partnerships, and a team of experienced professionals, GMD ensures efficient and effective service delivery in the logistics sector.
VSH (Vinh Son – Song Hinh Hydropower Joint Stock Company), established in 1991 and equitized in 2005, is the first hydropower company listed on the HOSE Its primary business activities include electricity production and trading, management and maintenance services, as well as advisory and supervisory roles Additionally, VSH invests in hydropower projects with a total capacity of 330 MW.
This study analyzes data collected from January 1, 2007, to December 31, 2011, focusing on the pre- and post-recession periods surrounding December 2008 By dividing the analysis into these two intervals, the research aims to highlight the significant effects of macroeconomic and external factors on the stock market during and after the recession.
We use the daily closing prices to estimate daily returns And the percentage of stock return is identified as:
R t = (P t - P t-1 / P t-1 )*100 Which P t and P t-1 are daily price of stock on two continuous day t-1 and t
This article examines the influence of trading volume and foreign trading activity on stock returns and volatility in the Vietnamese stock market It highlights the significant impact of foreign trading on stock prices and overall market dynamics Additionally, the analysis includes a discussion on the combined effects of foreign buy and sell volumes, underscoring their importance in understanding market behavior.
In my dataset, there may be instances of null or invalid inputs in the columns representing foreign buy and sell volume due to days when foreign investors do not engage in transactions To ensure that all data series remain relevant for model analysis, I treat “zero volume” as “0.001 volume.”
Finally, I will use Eviews and Stata software to analyze the relationship between stock return, trading volume, and volatility.
Summary statistics
I start to examine the relationship among stock return, trading volume and volatility within initial analysis of statistical description of time series
Stocks Variables Mean Std Dev Skewness Kurtosis Jarque -
Stocks Variables Mean Std Dev Skewness Kurtosis Jarque -
Source: Author’s own calculation based on dataset
Between 2007 and 2011, an analysis of eight stocks on the HOSE reveals key descriptive statistics, including Mean, Standard Deviations, Skewness, Kurtosis, and Jarque-Bera of daily returns, trading volumes, foreign buy volume, and foreign sell volume Most stocks exhibit negative average returns, with the exception of VIC at 0.04% and VNM at 0.017%, while SSI records the lowest return among all analyzed stocks.
The analysis of trading volume percentages reveals that SSI leads with a mean of 13.722, followed closely by DPM at 13.031 In contrast, VIC and VNM exhibit the lowest trading volumes, recorded at 11.439 and 11.487, respectively Stocks such as GMD, PVD, VSH, and FPT fall in the mid-range, with percentages of 11.785, 11.932, 12.147, and 12.307 Notably, foreign investors demonstrate a stronger purchasing power in DPM.
(11.296) and FPT (11.035), but lower power of GMD (6.932) and VIC (6.889)
Among the eight stocks analyzed, SSI and VIC exhibit the highest volatility in returns, with standard deviations of 0.034 and 0.033, respectively Conversely, DPM shows the least volatility, with a standard deviation of just 0.025 The standard deviations for the other stocks remain relatively stable, ranging from 0.027 to 0.030.
Most stock returns exhibit negative skewness, indicating higher risk, except for DPM, which has a positive skewness of 0.1013 This negative skewness suggests that returns are asymmetric and non-normal, largely due to the behavior of risk-averse investors (Moolman, 2004) Additionally, stock returns are predominantly leptokurtic, characterized by excess kurtosis greater than three, implying increased risk A more leptokurtic distribution correlates positively with return volatility, as noted by Tauchen and Pitts (1983), Karpoff (1987), and Gallant et al (1992) Conversely, DPM's negative excess kurtosis indicates lower risk compared to other stocks Consequently, SSI and VIC are identified as the most hazardous stocks, while DPM is recognized as the safest option due to its negative excess kurtosis.
The Jarque-Bera test indicates that the data series of nearly all stocks are non-normal, as the test results show JB values exceeding the critical χ² value of 4 This failure to accept the null hypothesis suggests that these series do not follow a normal distribution, which is essential for establishing weak-form market efficiency, as outlined by Fama.
(1965), Stevenson and Bear (1970), Reddy (1997), and Kamath (1998)) There is only case that stock DPM is normal distribution at one percent significant level
3 Excess kurtosis: a probability (return distribution) has a kurtosis parameter larger than parameter with normal distribution around 3.
4 χ 2 critical of degree of freedom of 2 at 1% significant level is 9.21, 5% level is 5.99 and 10% level is 4.6
Overall, from above statistics, VIC and SSI are the most dangerous for investors whereas DPM is the most safety stock for investment.
Graphical analysis
Source: Author’s graphs based on dataset
As seen from graphs, I can infer that data series for eight listed stocks are fully stationary This will be also affirmed by some following tests in the next sessions.
ECONOMETRIC MODELS AND DISCUSSION
Test for stationarity in stock return and trading volume
Granger (1974) highlights that estimating relationships among non-stationary variables can yield misleading results due to the challenge of distinguishing between temporary and permanent relationships in non-stationary time series To mitigate the risk of spurious correlations, I test the stationarity of stock returns and trading volume percentages Additionally, Su (2003) and Chen, Firth, and Rui (2001) discuss the presence of time trends in raw trading volume on the Chinese Stock Market, prompting me to incorporate a time trend in the Augmented Dickey-Fuller equation For the stationarity test, I utilize the Augmented Dickey-Fuller (ADF) test as proposed by Dickey and Fuller (1979).
And Philips – Perron (1988) (PP) test: x t = α 0 + αx t-1 + u t
Where x stands for stock return and trading volume percent, and ρ0, ρ, and δ are model parameters, ε t represents white noise error term, respectively
ADF test Level and Intercept 1st difference and Intercept
No trend With trend No trend With trend
ADF test Level and Intercept 1st difference and Intercept
No trend With trend No trend With trend
ADF test Level and Intercept 1st difference and Intercept
No trend With trend No trend With trend
Source: Author’s own calculation based on dataset
The ADF statistics for DPM returns are significantly low, at 0.000, indicating that both the level and first difference tests, with or without trend, reject the null hypothesis of non-stationarity This suggests that the DPM return series is stationary overall, a trend that is consistent across all eight stocks analyzed Consequently, the stationary nature of these returns enhances their suitability for statistical analysis.
PP test Level and Intercept 1st difference and Intercept
No trend With trend No trend With trend
PP test Level and Intercept 1st difference and Intercept
No trend With trend No trend With trend
PP test Level and Intercept 1st difference and Intercept
No trend With trend No trend With trend
Source: Author’s own calculation based on dataset
Table 4 indicates that DPM stock has a PP statistic of less than one percent across all cases, leading us to reject the null hypothesis of a unit root in returns Consequently, DPM returns are stationary This conclusion extends to the other stocks analyzed, confirming that all stock returns exhibit stationarity.
In general, from two tests above, we can conclude that all stock returns in the paper are stationary absolutely which are necessary for below regressions.
Trading volume and return volatility
In this paper, I employ the GARCH model to evaluate and forecast the stock returns on the Vietnamese Stock Exchange
The GARCH model, introduced by Bollerslev in 1986, stands out as the most effective model for addressing excess kurtosis in stock returns Subsequent research by Lamoureux and Lastrapes (1990), Brailsford (1996), and Mestel, Gurgul, and Majdosz has further validated its success in financial modeling.
The GARCH (1,1) model, initially developed in 2003, has been widely applied to financial time series data Over the past two decades, its use has expanded beyond just measuring volatility to encompass various forms of returns (Engle, 2001).
This section utilizes the GARCH (1,1) model, based on the methodology of Lamoureux and Lastrapes (1990), to analyze how trading volume impacts mean returns and conditional return volatility The modified GARCH (1,1) equation is expressed as: r_t = α_0 + α_r r_{t-1} + α_{Vol} lnVol_t + α_{FBVol} lnFBVol_t + α_{FSVol} lnFSVol_t + α_d d + ε_t.
t = β 0 + β ε ε 2 t 1 + β σ t 2 1 + β Vol lnVol t + β FBVol lnFBVol t + β FSVol lnFSVol t + β d d + e t
(a): the mean stock return equation (b): the conditional return volatility equation Where r t is daily return of stock; r t-1 is conditional return on past information;
The conditional variance (volatility) of ε_t at day t is denoted as σ_t², where ε_t represents a sequence of independent and identically distributed random variables with a mean of zero and a variance of one The percentage of trading volume at day t is indicated by lnVol_t, while lnFBVol_t and lnFSVol_t represent the percentages of foreign investors’ purchasing and selling volumes, respectively A dummy variable, d, is defined as d=0 for the period from January 1, 2007, to December 12, 2008, and d=1 for the period from January 1, 2009, to December 31, 2011 The white noise component is represented by e_t The constant β₀ is greater than zero, while β_ε and β_σ are coefficients that measure the dependence of current volatility on past squared residuals and past volatility, respectively, with both coefficients being greater than or equal to zero.
The condition (β ε + β σ ) < 1 indicates the persistence of conditional volatility in financial markets The coefficients α Vol, α FBVol, α FSVol, and α d represent the effects of trading volume percent, foreign buy volume percent, foreign sell volume percent, and a dummy variable on average returns Similarly, the coefficients β Vol, β FBVol, β FSVol, and β d measure the impacts of these same variables on conditional volatility, highlighting their significant roles in market dynamics.
Table 5: GARCH (1,1) model without LnVol
DPM PVD SSI VNM VIC FPT GMD VSH
DPM PVD SSI VNM VIC FPT GMD VSH
Source: Author’s own calculation based on dataset Note: (*), (**), (***) indicate 10%, 5%, and 1%, respectively
Excluding the percentage of trading volume (LnVol) as an explanatory variable significantly increases volatility persistence for most stocks, with SSI showing a notable decrease from 44% to 32% However, this outcome raises concerns, as FPT and VIC exhibit persistence greater than one, indicating explosive variance and undermining the assumption of stationarity, which leads to model instability Additionally, the persistence levels for VSH and PVD approach unity, suggesting that volatility shocks maintain a high degree of persistence Therefore, it is evident that the percentage of daily trading volume plays a crucial role as an explanatory variable in volatility analysis.
I discuss more the results of ARCH and GARCH effects in Table 5 The table reveals that
The study reveals that the ARCH effect (β ε) is significantly present at a 1 percent level, while over 80 percent of the GARCH effect (β σ) also shows significance at the same level Notably, the SSI exhibits no GARCH effect, indicating that this effect is removed when trading volume percentage is excluded Conversely, when considering trading volume percentage, both the ARCH and GARCH effects are significant These findings challenge the conclusions drawn by Lamoureux and Lastrapes (1990a) regarding the US stock market.
Furthermore, I compare the results of GARCH (1,1) model including percentage of volume (in Table 6) to model excluding percentage of volume (in Table 5) by the likelihood ratio (LR) test
The LR test is calculated as follows:
Where the probability distribution of the test statistic is approximately a chi-squared distribution with degrees of freedom of 1 (q=1)
L U : The log likelihood of the unrestricted model (with percentage of trading volume)
L R : The log likelihood of the restricted model (without percentage of trading volume) q = 1
After calculation, I get LRs of six stocks (excluding FPT and VIC) as follows:
Table 6: Likelihood ratios of stocks
Stocks DPM PVD SSI VNM GMD VSH
Source: Author’s own calculation based on dataset
The analysis indicates that the likelihood ratios (LRs) for six stocks exceed the critical value of χ²₁ (3.841), leading to the selection of the unrestricted model that incorporates trading volume percentage Notably, the model that omits trading volume percentage significantly reduces the GARCH effect of the Stock Market Index (SSI) Additionally, the persistence of stocks in the model without trading volume does not satisfy the stationary assumption present in the model that includes trading volume percentage These factors justify the choice of the model with trading volume percentage for further analysis.
Table 7: GARCH (1,1) model with LnVol
DPM PVD SSI VNM VIC FPT GMD VSH
DPM PVD SSI VNM VIC FPT GMD VSH
Source: Author’s own calculation based on dataset Note: (*), (**), (***) indicate 10%, 5%, and 1%, respectively
The Mixture of Distribution Hypothesis suggests that the GARCH effect is explained when β Vol is significantly positive, and the sum of (β ε + β σ) is less than the model's persistence magnitude, excluding the impact of trading volume.
Table 6 reveals that the ARCH and GARCH effects for each stock are significant at the one percent level, with the exception of SSI, which is significant at the five percent level The significant ARCH term indicates that current return volatility is influenced by lagged error terms, while the GARCH term suggests that past variance significantly impacts conditional variance A higher GARCH effect implies a stronger influence of past variance on current volatility, indicating inefficiency in the Vietnamese stock market's weak form Most stocks exhibit ARCH effects ranging from 20 to 30 percent, with VIC notably high at 77 percent GARCH effects are particularly pronounced in GMD, FPT, and VSH, peaking at 61 percent for DPM and exceeding 59 percent for FPT and VSH, while VNM and DPM show GARCH parameters around 40 percent Other stocks display GARCH effects between 10 and 20 percent, with VIC, PVD, and SSI at 21 percent, 16 percent, and 13 percent, respectively Overall, both ARCH and GARCH coefficients are significant across all stocks.
Conditional volatility for all stocks, represented by the sum (β ε + β σ ), is less than one, indicating stationarity and the impact of volatility shocks on returns This relationship also highlights volatility clustering, which contributes to asymmetry and inefficiency in emerging markets Higher persistence signifies a slower decrease in return volatility, while lower persistence reflects a quicker return to the mean Specifically, the conditional variance of VIC demonstrates very high volatility persistence, exceeding 90 percent, suggesting that volatility shocks are long-lasting In contrast, the persistence levels for PVD and SSI are notably low, falling below 90 percent.
50 percent) The volatility persistence of the rest of stocks is moderate from 70 to 80 percent
In the mean equation, all stock constants are negative and significant at the one percent level, indicating a general decrease in returns Conversely, the coefficient of lag returns is consistently positive and statistically significant at the same level Notably, VIC (0.24), SSI (0.23), and GMD (0.21) exhibit a greater influence on returns compared to FPT (0.15), VSH (0.12), VNM (0.12), and DPM (0.11), while PVD shows a minimal impact with a coefficient of 0.09 Thus, VIC and SSI significantly affect current returns through lag returns, whereas PVD demonstrates the opposite effect.
Over 80 percent of stocks exhibit a significant positive relationship between return and trading volume, indicating that higher trading volume often correlates with better returns However, for SSI, this relationship is not significant It's important to note that while trading volume percentage has a slight impact on average returns, its overall effect remains limited.
Volatility clustering indicates that significant price movements are often succeeded by further large fluctuations, while minor changes are typically followed by more modest shifts In the stock market, trading volumes generally range from 0.001 to 0.003, and notably, the volume percentage of SSI does not have a significant impact on returns.
In many instances, the percentage of foreign buying volume does not significantly affect returns due to minor influencing factors However, the VIC coefficient shows a statistically significant positive relationship at the one percent level, albeit with a minimal impact of only 0.00047.