Non linear model predictability of vietnam stock market price

INTRODUCTION

Problem Statement

Vietnam's stock market has been operational for nearly a decade, significantly influencing the country's economy Recently, financial organizations and investors have increasingly focused on predicting market trends, particularly through the analysis of stock time series and random-walk behavior to maximize investment returns However, there is a notable shortage of qualitative studies aimed at forecasting the Vietnamese stock market This gap is primarily due to a lack of professional expertise and qualitative analysis skills among Vietnamese forecasters Additionally, the challenge of accurate market forecasting is exacerbated by insufficient long-term data on the economy, enterprises, and market conditions.

Recent advancements in adaptive analysis methods for non-linear time series have garnered significant attention from scientists and researchers Various forecasting models, including Smoothing Exponential Regression, Threshold Regression, Artificial Neural Networks, Smooth Transition Regression (STAR), and Logistic Smooth Transition Regression (LSTR), have been developed for analyzing stock market trends However, there are limited studies focused on Vietnam's stock market, with notable research by Hoang Dinh Tuan in 2008, and most current analyses predominantly utilizing the ARIMA model for short-term stock index forecasting.

This study aims to forecast key stock indices in Vietnam, specifically the VN-index and HN-index, as well as other large-cap stock indices, using a non-linear model known as Logistic Smooth Transition Regression (LSTR) The findings are intended to offer valuable insights for investors and financial institutions, ultimately enhancing methods for analyzing and predicting the Vietnamese stock market and stock prices.

Chuyên đề thực tập Tốt nghiệp

This study aims to predict stock market trends and price movements, focusing on the VN-Index, HN-Index, and five major blue-chip stocks with significant market capitalization By utilizing a non-linear model known as Logistic Smooth Transition Regression (LSTR), the research seeks to deliver valuable insights for investors and financial institutions Additionally, the study intends to offer recommendations and policy implications to foster the development of Vietnam's stock market in the near future.

 Question 1: Whether non-linear model will be used to predict VN-Index and HN-

 Question 2: How does blue-chip stock price of HSX (Hochiminh Stock Exchange) and

HNX (Hanoi Stock Exchange) look like by using non-linear model?

Vietnam’s stock market analysis and forecasting are of significant interest to global financial organizations and investors, particularly regarding stock prices Bachelier (1900) established that stock prices fluctuate over time, while Einstein (1923) introduced the Brownian motion process, linking it to stock price movements L Savage noted that stock prices cannot be negative, leading to the development of the Brownian Motion Model to describe market price kinematics (P Samuelson, 1965) This model is represented as S(t) = exp(at + bW(t)) In the late 1960s and early 1970s, Robert C Merton further advanced financial theory, building on Samuelson’s foundational work.

Fischer Black and Myron Scholes (1973) declared their important study which is an

The application of the Brownian Motion Model in generating stock valuation formulas, particularly the Black-Scholes formula, has become essential for investors and businesses in assessing option prices across global markets The groundbreaking work of Merton and Scholes, which earned them the Nobel Prize in 1997, highlights the significance of the Geometric Brownian Motion Model in financial mathematics for evaluating financial options Despite its relative simplicity, this model does not fully capture the complexities of stock price movements.

In today's financial landscape, stock market price analysis and forecasting in developed nations predominantly rely on two key methodologies: Trend Analysis and Artificial Neural Networks Despite their widespread use, these techniques often fall short in accuracy due to significant standard errors.

Empirical studies on stock markets, including those in the United States, United Kingdom, Canada, New Zealand, Ireland, and Japan, reveal important quantitative analyses Nektarios Aslanidis et al (2002) identified that financial and macroeconomic factors, such as GDP, interest rates, inflation, money supply, and US stock prices, significantly influence UK stock returns To explore potential non-linear relationships between UK index returns and these variables, David G McMillan (2002) employed smooth-transition threshold models, finding that investor behavior varies between large and small returns Additionally, Gropp (2004) utilized cross-sectional power from industry-sorted portfolios, demonstrating a significantly positive speed of reversion to long-term equilibrium, with a half-life of approximately four and a half to eight years.

Time series forecasting has gained significant attention in recent studies, particularly in analyzing and predicting global stock price indexes The STAR model has emerged as a popular tool for this purpose, as demonstrated by Rodrigo Aranda and Patricio Jaramillo's research in 2008, which focuses on estimating smooth returns.

The study of Transition Autoregressive (STAR) models and Markov-Switching Vector Autoregressive (MS-VAR) models reveals nonlinear patterns in both trading volume and stock returns within the Chilean Stock Market, indicating distinct series characteristics Research by Marcella Niglio (2002) on the German stock market employs the Logistic Double Smooth Transition (LDST) model to analyze daily returns of the Dax 30 stock index, demonstrating significant forecast improvements in conditional variance compared to the Double Threshold ARCH (DTARCH) model Additionally, F Chan and M McAleer evaluate the forecasting performance of the STAR-Generalized Autoregressive Conditional Heteroscedasticity (STAR-GARCH) model and the STAR-Smooth Transition GARCH model, highlighting that different algorithms yield varying parameter estimates, which in turn affect forecast accuracy and performance.

As Vietnam's economy increasingly integrates into the global market, its stock market experiences significant influence from international fluctuations, particularly during economic transitions Despite being in the early stages of development, Vietnam's stock market presents considerable potential for investors, making the analysis and forecasting of stock indices highly appealing due to the prospect of substantial profits However, the use of quantitative models for analyzing Vietnam's financial market remains limited, with many investors relying on information from newspapers, media, or personal networks, especially in the context of less transparent data According to Hoang.D Tuan (2008), some random processes can be observed in this market behavior.

(such as Geometric Brownian Motion and Mean Reversion Model) in order to describe stock price moves and technical testing.

This research presents a method utilizing Logistic Smooth Transition Regression (LSTR) to forecast stock market prices in Vietnam, specifically focusing on the VN-Index, HN-Index, and five blue-chip stocks from January 1, 2009, to October 31, 2011 The objective of this thesis is to provide valuable reference information for both domestic and international investors.

This thesis explores the modeling of nonlinear relationships through the STR model with a transition function, aiming to address key research questions It outlines a comprehensive modeling cycle that encompasses three critical stages: specification, estimation, and evaluation, as detailed in Teravirta (1994) These stages are thoroughly discussed in the subsequent section.

2 of Chapter 2 After that, forecasting stock indexes also are implemented in this thesis.

The article begins with an introduction, followed by Chapter 2, which outlines an analytical framework that includes key concepts and definitions related to the stock market, along with a predictive model for stock prices Chapter 3 provides an overview of the Vietnam stock market and analyzes the performance of blue-chip stocks set for forecasting In Chapter 4, the estimated results of stock indexes are presented using the quantitative methods discussed in Chapter 2 Finally, Chapter 5 summarizes the thesis's main findings and offers recommendations.

Research questions

 Question 1: Whether non-linear model will be used to predict VN-Index and HN-

 Question 2: How does blue-chip stock price of HSX (Hochiminh Stock Exchange) and

HNX (Hanoi Stock Exchange) look like by using non-linear model?

Literature review

Vietnam's stock market analysis and forecasting have garnered significant attention from global financial organizations and investors, particularly concerning stock market prices Bachelier (1900) demonstrated that stock prices change over time, with the equation S(t+Δ) - S(t) = Δ 1/2 indicating this continuous variation In 1923, Albert Einstein introduced the Brownian motion W(t) as a framework for studying kinematic elements, establishing that stock prices can be modeled as components of this process L Savage noted that stock prices are never negative, leading to the development of the Brownian Motion Model to describe the kinematics of various stock prices (P Samuelson, 1965), exemplified by the equation S(t) = exp(at + bW(t)) Further advancements in financial theory were made by Robert C Merton in the late 1960s and early 1970s, building on Samuelson's foundational model.

Fischer Black and Myron Scholes (1973) declared their important study which is an

The application of the Brownian Motion Model in generating stock valuation formulas, particularly the renowned Black-Scholes formula, has become essential for investors and businesses worldwide in evaluating option prices across various markets The groundbreaking work of Merton and Scholes, which earned them the Nobel Prize in 1997, highlights the significance of the Geometric Brownian Motion Model in financial mathematics for assessing the value of financial options Despite its relative simplicity, this model does not fully capture the complexities of stock price movements.

In contemporary finance, stock market price analysis and forecasting in developed nations primarily utilize two key methodologies: Trend Analysis and Artificial Neural Networks Despite their popularity, these techniques often struggle with accuracy, resulting in significant standard errors in predictions.

Empirical studies have conducted quantitative analyses of stock markets in various countries, including the United States, United Kingdom, Canada, New Zealand, Ireland, and Japan For instance, research by Nektarios Aslanidis et al (2002) indicates that financial and macroeconomic factors such as GDP, interest rates, inflation, money supply, and US stock prices impact UK stock returns David G McMillan (2002) explored the potential non-linear relationship between UK index returns and these variables using smooth-transition threshold models, revealing that investor behavior varies between large and small returns Additionally, Gropp (2004) found that industry-sorted portfolios exhibit a significant positive speed of reversion, with a half-life of approximately four and a half to eight years to return to long-term equilibrium.

In recent years, time series forecasting has garnered significant attention in various studies, particularly for analyzing and predicting global stock price indexes A prominent approach in this field is the STAR model, which has been widely utilized in contemporary research Notably, Rodrigo Aranda and Patricio Jaramillo (2008) contributed to this area by estimating smooth return patterns.

The study of the Chilean Stock Market utilizing Transition Autoregressive (STAR) and Markov-Switching Vector Autoregressive (MS-VAR) models reveals nonlinear patterns in trading volume and stock returns, highlighting the series characteristics of this market In a related analysis by Marcella Niglio (2002) on the German stock market, the Logistic Double Smooth Transition (LDST) model was employed to assess daily returns of the Dax 30 stock index, demonstrating significant forecast improvements in conditional variance over the Double Threshold ARCH (DTARCH) model Furthermore, research by F Chan and M McAleer on the STAR-Generalized Autoregressive Conditional Heteroscedasticity (STAR-GARCH) and STAR-Smooth Transition GARCH models indicates that varying algorithms yield different parameter estimates for similar likelihood values, ultimately affecting forecasting performance.

In the context of increasing international economic integration, Vietnam's economy is becoming more open to global markets, significantly impacting its stock market, particularly during economic transitions As Vietnam's stock market is still in its early stages of development, it presents various potential risks, making stock index analysis and forecasting highly appealing to financial organizations and investors seeking profitability However, the use of quantitative models to analyze Vietnam's financial market remains limited, with many investors relying on information from newspapers, media, or personal networks, especially when faced with a lack of transparency.

(such as Geometric Brownian Motion and Mean Reversion Model) in order to describe stock price moves and technical testing.

This research introduces a method for utilizing Logistic Smooth Transition Regression (LSTR) to forecast stock market prices in Vietnam, focusing specifically on the VN-Index, HN-Index, and five blue-chip stocks from January 1, 2009, to October 31, 2011 The aim is to provide valuable reference information for both domestic and foreign investors.

Thesis Methodology

This thesis explores the modeling of nonlinear relationships through the STR model with a transition function to address the specified research questions It outlines a three-stage modeling cycle comprising specification, estimation, and evaluation, as detailed in Teravirta (1994) Further information on these stages can be found in Section [insert section number].

2 of Chapter 2 After that, forecasting stock indexes also are implemented in this thesis.

Thesis Structure

This article begins with an introduction and proceeds to present an analytical framework in Chapter 2, which includes key concepts and definitions related to the stock market, alongside a model for predicting stock prices Chapter 3 offers an overview of the Vietnam stock market, focusing on the performance of blue-chip stocks that will be forecasted Chapter 4 discusses the estimated results of stock indexes using the quantitative methods outlined in Chapter 2 Finally, Chapter 5 summarizes the main findings of the thesis and provides recommendations.

ANALYTICAL FRAMWORK

Concept and Definition

This section focuses on giving some concepts and definitions regarding to stock market and security market which are mentioned in this thesis.

Common stock represents ownership in a corporation and serves as a claim on its earnings and assets By issuing or selling stock to the public, corporations can effectively raise funds to support their operations and growth.

(ii) Stock market is the market in which claims on the earnings of corporations

(shares of stock) are traded, is the most popular in financial market 1

(iii) Security (also called a financial instrument) is a claim on the issuer’s future income or assets (any financial claim or piece of property that is subject to ownership) 2

(iv) Bond is a debt security that promises to make payments periodically for a specified period of time 3

A portfolio is a collection of investments held by individuals or organizations, typically encompassing stocks, bonds, and mutual funds Mutual funds represent a collective investment, where funds from multiple investors are pooled together and managed by professional fund managers.

(vi) Blue-chip is a nationally recognized, well-established and financially sounds company Blue chips generally sell high-quality, widely accepted products and services.

1 Mishkin, Frederic S, The economics of Money banking and Financial Market, seventh edition, pp.5

2 Mishkin, Frederic S., The economics of Money banking and Financial Market, seventh edition, pp.3

Blue chip companies consistently demonstrate resilience during economic downturns, maintaining profitability even in challenging conditions This ability to thrive amid adversity significantly contributes to their long-standing reputation for stable and reliable growth.

Models for predicting stock market price

Smooth Transition Regression (STR) is a non-linear regression model derived from Quandt's (1958) Conversion Regression Model, which is viewed as a Threshold Regression Model (Tong, 1990) The two-regime version of the Conversion Regression Model is a specific instance of the standard STR model, while models with more than two regimes do not conform to this standard Additionally, the Smooth Transition Autoregressive (STAR) model is classified as a special case of the standard STR model, originating from the general form of the Conversion Regression Model (Bacon and Watts, 1971) This model employs two-way regression to create a smooth transition between states, and Chan and Tong (1986) proposed a univariate STAR model.

The standard STR model is defined as follows:

In this model, the vector of explanatory variables is represented as (wt ’, xt ’), where the parameter vectors consist of p=k+m and the error term ut follows an independent and identically distributed (iid) normal distribution with mean zero and variance σ² The transition function F(γ, c, st) is a bounded function that depends continuously on the transition variable st and remains continuous across the entire parameter space for any st value Here, γ denotes the slope parameter, while c = (c1,…,ck)' represents a vector of location parameters, satisfying the condition c1 ≤ … ≤ ck.

5 http://www.investopedia.com/terms/b/bluechip.asp#ixzz1dhkgespQ

The last expression in (1) indicates that the model can be interpreted as a linear model with stochastic time-varying coefficients

Logistic Smooth Transition Regression (LSTR)

The LSTR can be in the form of logistic function or exponential function:

(2) Where is an identifying restriction Equations (1) and (2) jointly define the logistic STR (LSTR) model The most common choices for K are K=1 and K= 2

For K = 1, the parameters change monotonically as a function of st from to

The LSTR1 model (K=1) with transition function as follows:

The LSTR1 model, with K=1, effectively captures asymmetric behavior in economic cycles For instance, when st represents the business cycle phase, the LSTR1 model illustrates how dynamic properties vary between expansions and recessions, with a smooth transition between these two extreme regimes.

The LSTR2 model (K=2) with transition function as follows:

(4) The LSTR2 model is suitable for the case in which the parameters changes

Chuyên đề thực tập Tốt nghiệp and small values of transition variable), whereas the middle regime is different 6 (For further information in Ocal and Osborn (2000), Van Dijk and Franses (1999)).

(5) The ESTR Model is suitable for the model which the characteristic of models in proportion of absolute value of st are the same.

Trend of transition and intercept

The STRS model enables the level and trend of time series changing gradually The transition happens timely, not immediately.

(6) Where: T is total of observation of yt time series; εt is the stationary process of I(0) and

Ft(γ,c,st) is STR model

This thesis explores the modeling of nonlinear relationships through the STR model, utilizing a specific transition function We outline a comprehensive modeling cycle that encompasses three key stages: specification, estimation, and evaluation The strategic approach to this modeling process is detailed in Terasvirta's work.

(1994) [25] We will now discuss the three stages of the cycle separately, beginning with specification, continuing with estimation, and ending with evaluation.

The model specification includes two steps such testing linear and specifying model.

The STR model incorporates a predefined linear test with a single transition variable, serving two primary objectives The first objective is to evaluate the linear characteristics of the model.

6 Helmut Lỹtkepohl and Markus Krọtzig(2004), “Applied time series econometrics”, Cambridge University Press

The graduation internship focuses on the directions of spatial parameters If the null hypothesis (H0) is not rejected, the linear model is accepted, and the STR model will not be utilized Additionally, the results from the linear test are used to specify the model If H0 is not rejected for at least one of the models, the STR model, which is strongly rejected, will be chosen for estimation.

The STR model shares similarities with other non-linear models, as it is defined by an alternative hypothesis rather than a null hypothesis of linear characteristics To test the linear STR model, one can approximate the transition model using a Taylor expansion around the null hypothesis where γ = 0 Typically, we assume K=1 in the model and apply a third-level Taylor approximation The testing results are applicable for both LSTR1 (K = 1) and LSTR2 (K = 2) models.

It assumes that the continuous transition variable st is a component of zt and zt = (1,

), where is a vector of m1 The proximate result after combining and getting back parameter is a supplement regression in below:

(7) Null-hypothesis H0: 1 = 2 = 3 = 0, where j (j = 1, 2, 3) is defined as in which

The hypothesis of linear models is based on the relationship between the function of θ and c Utilizing LM testing ensures that the theory of asymptotic distribution remains unaffected, as indicated by ut However, the asymptotic distribution theory of χ² necessitates an additional condition for its validity.

The freedom degrees of statistical testing of  2 asymptotical distribution is 3m when H0 is accepted However,  2 statistics could be slanted critically in term of dimension

The graduation internship topic involves analyzing small and medium sample sizes In this context, the F statistics will be substituted, with the degrees of freedom for the F approximate distribution being 3m and T – 4m – 1 under the null hypothesis.

The process of specifying and testing the Smooth Transition Regression (STR) model involves selecting potential continuous transition variables, denoted as S = {s1t, …, skt}, and testing each variable individually If the null hypothesis (H0) is rejected for certain continuous transition variables, we then identify those with the smallest p-values In cases where multiple variables exhibit similarly small p-values, we can advance to STR estimation, which will be further defined during the model specification phase.

After dismissing linear characteristics and opting for a single continuous transition variable, the next phase involves model specification, with two options available: K = 1 and K = 2 In the case of LSTR1, the parameters exhibit a monotonic change in relation to the continuous transition variable, though this change does not have to occur on the same side Conversely, LSTR2 features parameters that change symmetrically around the midpoint of (c1+c2)/2.

The specified models LSTR1 and LSTR2 can be derived from supplement regression (2), where the vectors βj (j = 1, 2, 3) are functions dependent on the parameters outlined in (1) In the special case where c = 0, it can be concluded that β2 equals 0 for the LSTR1 model, while β1 and β3 are both 0 for the LSTR2 or ESTR models Even when c is not equal to 0, β2 remains nearly a null vector in comparison.

1 and3when the model is LSTR1, and in contrast to LSTR 2 This stage is orderly testing as follow:

When H03 testing is significantly rejected based on the p-value, we opt for LSTR2 or ESTR, as it is possible for all three hypotheses to be rejected simultaneously at the 0.05 or 0.1 significance levels Conversely, if the hypothesis is not rejected, LSTR1 will be selected This methodology is based on the work of Teräsvirta (1994).

Chuyên đề thực tập Tốt nghiệp process and archived relatively good results Escribano (1999) also suggested other process which requires adding in to (1) and assumes that 1 = 2 = 3 = 4 = 0.

The results of the STR form specification demonstrate effectiveness through the application of the aforementioned testing processes We can modify LSTR1 and LSTR2 (or ESTR) based on the data and select the appropriate model form during the evaluation stage This approach is valid when the p-values for H03, H02, and H04 are similar Notably, rejecting H04 indicates LSTR1 with non-zero thresholds, rejecting H03 signifies LSTR2, and rejecting H02 corresponds to LSTR1 with zero thresholds.

The parameters of the LSTR model are estimated using the Maximum Likelihood Method, with the Newton Raphson method applied to maximize the likelihood Initially, we determine the starting values of γ0 and c0 for an automatic sequential algorithm We then create two dimensions for γ0 and ct to find the parameter values that minimize the remainder of the series, denoted as ut.

PERFORMANCE OF VIETNAM STOCK MARKET

Overview of Vietnam Stock Market

The Vietnamese Stock Market, which officially began operations in 2006 after its inception in 2000, has shown significant growth in response to both domestic and global economic shifts Despite its impressive achievements and rapid expansion, the market still faces inherent challenges typical of emerging and developing economies.

(1)The stock market has rapidly developed in size and gradually played as an important medium and long term capital channel

The Vietnamese Stock Market was once recognized as one of the fastest-growing stock markets globally, particularly in terms of capitalization, proportion, and growth rate Between 2000 and 2005, market capitalization represented only about 1% of GDP However, significant progress was made in 2006, with market capitalization rising to 22.7%, and further increasing to 43% in 2007.

In 2008, the global financial market changes and Vietnam's struggling economy caused stock market indexes to decline significantly, resulting in a 50% drop in market capitalization, which fell to just 18% of GDP However, with the recovery of both the internal and global economies in the second quarter of 2009, stock market indexes and the number of listed companies surged, leading to a market capitalization of VND 717.2 trillion, approximately 36% of GDP, an increase of VND 100 trillion compared to 2009.

Table 1: Capitalization level, proportion, and growth rate of the stock market

Source: The State Securities Commission, % of GDP.

(2)Increase in the number of listed companies has contributed to increase in market supply and liquidity.

Between 2000 and 2005, Vietnam's stock market experienced a downturn, with only two companies, REE and SAM, listed on the Ho Chi Minh City Stock Trading Center in 2000 However, the landscape changed dramatically after 2006, leading to a rapid increase in listed companies By 2011, the number of companies listed on both Stock Trading Centers reached 765, marking an impressive 18% growth from the previous year and representing the highest figure in a decade.

Figure 1: Size of listing on Vietnam’s Stock Trading Center

Source: The State Securities Commission

The stock market has experienced significant growth in both listing size and market liquidity, with trading sessions increasing from 667,600 shares in 2005 to 2.6 million shares in 2006, marking a 3.93-fold increase This upward trend continued, with trading volumes reaching 9.79 million shares in 2007 and 18.07 million shares in 2008.

(3)System of Intermediate institutions and securities services has been developing in both quantity and quality.

In the last decade, the State Securities Commission has granted licenses to 105 stock companies, 47 asset management firms, 382 foreign investment funds, and 8 custodian banks, a significant increase from just 4 stock companies in 2000 Initially, these stock companies primarily offered brokerage services; however, they have since expanded to include financial consultancy, self-trading, and underwriting services Additionally, the financial strength of these securities companies has improved significantly, with an average charter capital exceeding VND 150 billion per company.

Table 2: The growth rate of the number of stock companies and asset management companies

Number of asset management companies

Source: The State Securities Commission

Between 2005 and 2010, the number of asset management companies surged from 6 to 47, reflecting significant growth in the sector These firms have successfully established and managed stock investment funds, overseeing over 200 portfolios for both local and international clients The total equity mobilized by these asset management companies reached approximately VND 66,000 billion, equivalent to around USD 3.8 billion.

(4) Local and foreign investors have increased in term of quality and quantity

There have been more and more investors joining the stock market, from 3,000 accounts at the end of 2000 to 926,000 accounts in 2010 ( about 300 times within 10 years).

The development of a robust system for professional investors has been essential for the rapid and stable growth of the stock market, enabling it to withstand annual market shocks and maintain national financial stability Additionally, a welcoming investment environment and supportive policies have successfully attracted both local and foreign investors Currently, there are over 10,000 trading accounts held by foreign investors, including more than 1,000 accounts from foreign investment institutions.

Between 2000 and 2005, the VN-index was at 307.5 points, while the HaSTC-index stood at 96.24 The period from 2006 to 2009 saw a rapid increase in the VN-index, peaking at 1,170.76 and the HaSTC-index reaching 459.36 in March 2007 However, the global financial crisis of 2008 severely impacted Vietnam’s securities market, causing the VN-index to plummet to 366.02 points, a 71% drop from the beginning of the year, and the HaSTC-index to decline to 105.12 points, approximately 81% Following the crisis, the stock market continued to decline, hitting a low of 251.44 points for the VN-index and 88.64 points for the HaSTC-index on March 13, 2009 From February 2010 onwards, the stock market exhibited volatility, with the VN-index fluctuating around 500 points, highlighting the characteristic large variations of Vietnam’s stock index, which can change by 25 points in a single trading session, contrasting sharply with the smaller fluctuations seen in international markets.

Sources: www.vcbs.com.vn

7 (1) 11/1: Rumours of prime interest rate increasing

(2) 26/1: Prime interest rate rises to 8%

(3) 11/2: Exchange rate (USD - VND) increases by 3,36%

(4) 15/3: ANZ withdraws capital from STB

(5) 01/4: The Government requests the State Bank of Vietnam (SBV) decreases lending interest rate

(6) 07/5: The Government promulgates the Decision No 23/NQ-CP requesting SBV have to bring out suitable solutions for decreasing interest rate

(7) 06/5: The stock market starts to sharply decrease due to worry about public debt in Europe after serous crisis in Greece

(8) 20/5: The Circular No 13 is promulgated

(9) July: Vinashin official announces the debt amount of VND 80,000 billion

(10) 1/8: Rumours of Dragon Capital will be capital withdrawal

(11) 12/8: VEIL and VGF decide to do not capital withdrawal

(12) 17/8: Interest rate in the free market increases and official interest rate rises by 2.1%

(13) 06/9: The government requests SBV review the Circular No13

(14) 28/9: The Circular No19 replaces the Circular No13 and there is no important change.

(15) 30/9: Exchange rate in the free market starts to sharply increase and the price of gold tends to robust increase

(16) 23/10 : CPI increases by1.05% by October

(17) 5/11: Prime interest rate increases by 9%, SBV tightens the money and VND has not depreciated until end of year

(18) 11/11: The price of gold rises by VND38 million per tael of gold.

(19) 24/11: CPI in November increased by 1.86%

(20) 8/12: Techcombank increases lending interest rate by 17%

(21) 14/12: Commercial banks are accepted to increase their charter capital by

VND3000 million and the lending interest rate cap is by 14%.

(22) 15/12: Moody decreases the credit state bonds of Vietnam to B1

(23) 24/12 CPI increases by 1,98% in December

(2)Limitations on size and composition

The Vietnamese stock market features approximately 500 stock and bond tickers, a modest number given the economy's potential In contrast, the government bond market boasts over 500 tickers, which may hinder market liquidity according to international standards While the stock market primarily focuses on share trading, only 60% of listed shares are actively traded, and bond trading remains underdeveloped despite significant potential Overall, the quality of listed companies in Vietnam is moderate, characterized by low corporate management standards, a predominance of small to medium-sized enterprises with limited capital and growth prospects, and reluctance among top companies to enter the public market.

In addition to the two major stock exchanges, HOSE and HASTC, Vietnam features various informal trading markets, including secondary stock market activities, unofficial online trading, and bond mortgages, primarily involving government bonds These trading models have led to increased social costs, diverging from international trends Furthermore, the low proportion of traded shares in public companies and the lack of investor appeal have hindered market growth Some share markets registered at the Custodial Center lack effective trading mechanisms and ownership transfer processes.

Figure 3: Upcom-Index and trading value in Upcom market

The Upcom trading system, which includes a market for unlisted shares, has yet to fulfill its role in market establishment and remains unattractive to investors With only 600,000 shares traded per session, amounting to approximately VND 9 billion, the Upcom index has plummeted over 60% since its inception on June 24, 2009, currently hovering around 40 points This lack of interest can be attributed to investors favoring high liquidity shares on the two official stock exchanges over those on the Upcom exchange.

(4)Shortcomings of intermediate institutions and market development supports

The majority of security companies and custodians are concentrated in Hanoi and Ho Chi Minh City, resulting in a limited presence in other provinces This lack of a widespread network has diminished investor interest, compounded by the insufficient availability of essential services.

Chuyên đề thực tập Tốt nghiệp institutions such as stock investment funds, credit reference companies, and stock transfer agents The staff still lack of qualifications, experience and skills.

In summary, Vietnam's stock market has significantly expanded in both scale and quality, with advancements in various markets, stock exchanges, and bond markets, including UPCoM Improvements in intermediary systems and the quality of information dissemination have enhanced the market's role as a vital capital channel for the economy However, emerging challenges pose serious threats to market transparency and stability Thus, it is essential to implement effective policies for market stabilization and development, while also prioritizing transparency and information disclosure in the future.

Performance of five Blue Chips in therecent time

As of October 31, 2011, the thesis identifies five promising Blue Chip stocks—AGF, BMC, FPT, PVD, and SSI—based on their trading performance in the stock markets, highlighting their potential as valuable investment opportunities.

Table 3: Business performance of 5 Blue Chips

Category Year Unit AGF BMC FPT PVD SSI

8 AGF: An Giang Fisheries Import Export Joint Stock Company

BMC: BinhDinh Minerals Joint Stock Company

FPT: The Corporation for Financing and Promoting Technology

Category Year Unit AGF BMC FPT PVD SSI

Sources: www.bsc.com.vn

Recent analyses indicate that the business results of selected enterprises have been satisfactory, making these blue-chip stocks appealing to investors Figure 4 illustrates the performance of five stock market indices, with AGF and BMC experiencing significant increases in October.

2011 compared with the last month and tending to rise The others have varied slightly in recent months.

A: 20-09-2011- Advance payment of share dividend at time 1/2011 by cash, rate VND1000 per share

B: 14-09-2011-Advance payment of share dividend at time 1/2011 by cash, rate VND1000 per share

C: 28-04-2011-Annual meeting of the Board of Director in 2011

D: 15-04-2011-Making share dividend payment at time 2/2010 by cash, rate

F: 29-03-2011-Making share dividend payment at time 2/2010 by cash, rate

H: 29-03-2011- Annual meeting of the Board of Director in 2011

A: 17-08-2011- Making share dividend payment at time1/2011by cash, rate VND1000 per share C: 28-07-2011-Making share dividend payment at time1/2011by cash, rate VND1000 per share

A: 23-09-2011-Advance payment of share dividend in year 2011 by cash, rate VND1000 per share

B: 31-08-2011-Advance payment of share dividend at time 1/2011 by cash, rate VND1000 per share

C: 26-05-2011-Making the rest share dividend payment in 2010by cash, rate

E: 26-03-2011-Annual meeting of the Board of Director in 2011

A: 22-09-2011- Making the rest share dividend payment in 2010by cash, rate VND1000 per share

B: 31-08-2011- Making the rest share dividend payment in 2010by cash, rate VND1000 per share

D: 28-04-2011- Annual meeting of the Board of Director in 2011

Figure 4: Performances of 5 Blue chips on the stock market up to 31 October, 2011

Sources: www.bsc.com.vn

In summary , analyzing the situation of Vietnam’s stock market, particularly VN-

Index and HN-Index in recent years has shown in this chapter The market has made remarkable advances in both size and quality As above mentioned, the stock market

In the second quarter of 2009, both the Chuyên đề thực tập Tốt nghiệp indexes and the number of listed companies experienced rapid growth, driven by the recovery of the internal and global economies The development of intermediate institutions and securities services has improved in both quantity and quality, attracting a diverse range of local and foreign investors However, challenges remain in this emerging market, including limitations in size and composition, inadequate market structure, and fluctuations in Vietnam’s stock index, indicating an unstable growth trajectory Additionally, this chapter provides a brief overview of five blue-chip stocks that will be analyzed and predicted in this thesis.

MODEL SPECIPICATION AND DATA ANALYSIS

Description of data

This thesis analyzes and forecasts the VN-Index, HN-Index, and five emerging blue-chip stocks: AGF, BMC, FPT, PVD, and SSI Data for this study was sourced from the Hanoi Stock Trading Center (HASTC) and the Ho Chi Minh City Stock Exchange (HOSE), covering the period from January 1, 2009, to October 31, 2011 The findings are based on a comprehensive set of observations outlined below.

Estimated results

Before implementing the LSTR model for analyzing and predicting the VN-Index, this study conducts statistical tests to determine whether the data series are linear or non-linear The results of these tests are presented below.

Table 4: Testing linear or non-linear of data series for suggesting model

Trend_HN-Index 4.2855e-14 2.5397e-08 4.9929e-07 1.1102e-14 LSTR1 Trend_AGF 5.3679e-74 3.9183e-03 6.9268e-02 3.1013e-74 LSTR1 Trend_BMC 6.6408e-125 4.5179e-09 3.9968e-14 1.3101e-14 LSTR1

Trend_FPT 3.9513e-129 NaN NaN 3.9524e-14 LSTR1

Trend_PVD 1.8640e-105 1.9207e-14 NaN 4.8352e-11 LSTR1 Trend_SSI 1.1102e-16 2.2867e-06 8.1591e-02 NaN LSTR1

Sources: Estimated results of the model

Note: F statistic of H04; H03; H02Hypothesis are denoted by F4, F3 and F2

The F statistic test regarding linear characteristic of the series data states that H0

Hypothesis is rejected at significance value of 1% Therefore, the suggested model in table

4 will be applied to estimate and forecast stock indexes in next times

To estimate the non-linear STR model, it is essential to first calculate the initial parameters Subsequently, the Newton-Raphson Algorithm can be employed to estimate the remaining unknown parameters, maximizing the conditional likelihood function in the process.

This thesis utilizes the GRID SEARCH method to establish initial values of Gamma and C1 for seven stock indexes, as outlined in Table 4 The estimated values of γ and c obtained from the GRID SEARCH method are illustrated in Figure 5.

Figure 5: Defining initial values of γ and c by using the method of GRID SEARCH

The analysis conducted by Gama and C using the GRID SEARCH method reveals that the estimated models for the VN-Index, HN-Index, and five blue-chip stocks are statistically significant.

This thesis evaluates unit roots in the estimated models to confirm their efficiency and stationarity According to the testing results in Table 5, the forecasted error term of the model exhibits characteristics of white noise, as detailed in the attached appendix Additionally, Figure 7 illustrates the ADF test results for the residual error term presented in Table 5.

Table 5: Augmented Dickey-Fuller (ADF) Unit Root and PORTMANTEAU Test results

P-Value (portmanteau test) Value ADF (lag)

Residual of forecasting errors ADF(1)=-14.9965 -2.56 -1.94 -1.62 0.3582

Residual of forecasting errors ADF(1)= -12.1304 -3.96 -3.41 -3.13 0.8900

The results of the ADF test and Portmanteau test indicate that the VN-Index, HN-index, and five blue-chip stocks show significant testing results at the 5% level This suggests that the estimated models are statistically significant concerning the residuals of forecasting errors, demonstrating their relative efficiency and stationarity.

Chuyên đề thực tập Tốt nghiệp AGF

Chuyên đề thực tập Tốt nghiệp PVD

Figure 6: ADF Test for residual of error terms with one lag

The estimation results of the LSTR1 model, as summarized in Table 6, indicate that the models are statistically significant While the estimated outcomes for the VN-Index and HN-Index show relative consistency, notable differences exist between the two models Specifically, the constant term α0 is absent in the HN-Index model, whereas it is a positive value in the VN-Index model Additionally, both β0 values for the HN-Index and VN-Index are positive, suggesting that these indexes are likely to increase prior to the transition period.

Table 6: Estimation results of LSTR1 model 9

In the model, 9α0 represents the constant term of the linear component, while α1 indicates the estimated change in the non-linear aspect The parameter β0 reflects the growth rate of the trend prior to the transition period, and β1 signifies the increasing trend observed during the transition process Additionally, τ denotes the specific time at which the model transition occurs, and γ represents the rate of transition.

Chuyên đề thực tập Tốt nghiệp AGF

Figure 7: The graph of STR model

The short-run prediction process for the VN-index is conducted once the model meets the necessary statistical testing conditions The results demonstrate a strong and significant prediction accuracy, as evidenced by the attached appendix With an average error across 706 observations being relatively small, Figure 9 illustrates the close alignment between the predicted VN-Index and the actual VN-Index from February 1, 2009, to October 31, 2011, indicating that the predicted values closely mirror the real values.

Figure 8: VN-Index, HN-Index, and five blue chips forecast

CONCLUSIONS

In today's financial landscape, investors and organizations are increasingly focused on stock price forecasting to better understand market behavior Analysis reports, particularly those detailing stock price trends, are crucial for selecting optimal investment opportunities and timing market participation However, accurately predicting stock prices remains challenging due to their complex fluctuations and the need to anticipate market players' decisions Despite advancements in forecasting methods since the inception of stock markets, no flawless approach exists for precise predictions, necessitating the acceptance of error margins in the forecasting process This thesis proposes a model designed to provide investors, financial organizations, and newcomers with valuable insights to inform their investment choices, yielding relatively strong and significant results that are deemed acceptable.

This thesis provides a comprehensive overview of the theoretical framework and literature related to the analysis and prediction of stock indexes, specifically focusing on the VN-Index, HN-Index, and five blue-chip stocks It successfully meets its research objectives, answering both central and sub-questions, and demonstrates that the LSTR model is effective for predicting stock prices in these markets The significant results suggest the potential for developing software that can assist investors, financial organizations, and stock market participants by providing timely and updated information on stock price predictions.

The findings of this thesis serve as a valuable reference for investors, financial organizations, and stock market participants However, it is essential for them to consider additional related factors when utilizing this method for making informed investment decisions Due to time constraints and the volatility of the stock market, this thesis does not provide future predicted values for the indexes, indicating that further research is needed for continued development in this area.

[1] A Lendasse, E De Bodt, V Wertz, and M Verleysen (2000), “Non-linear financial time series forecasting – Application to the Bel 20 stock market index”, European

Journal of Economic and Social Systems 14 N° 1 (2000) 81-91.

[2] Bachelier, Louis (1900) Théorie de la Spéculation, Annales Scientifique de l'École Normale Supérieure, 3 e série, tome 17, 21-86 [English translation in Cootner; original French with a more recent English translation in Davis and Etheridge.]

[3] Bacon D.W and Watts D.G (1971), Estimating the transition between two intersecting straight lines, Biometrika 58: 525-534.

[4] Bali, Turan G., K Ozgur Demirtas, and Haim Levy (2008), “Nonlinear mean reversion in stock prices”, Journal of Banking & Finance, Vol 32, pp 767-782.

[5] Chan K.S and Tong H (1986), On estimating thresholds in autoregressive models, Journal of Time Series Analysis 7: 178-190

[6] David G McMillan (2002), Non-Linear Predictability of UK Stock Market Returns, Department of Accountancy and Finance, University of Aberdeen.

[7] F Black and M Scholes (1973), ‘The pricing of options and corporate liabilities’, Journal of Political Economy, Vol 81, pp 637-659, 1973.

[8] F Chan and M McAleer, “Evaluating the forecast performance of smooth transition volatility models”, University of Western Australia

[9] Franses P.H and D.van DiJk (2000), “Nonlinear time series models in empirical finance”, Cambridge: Cambridge university press.

[10] Granger C.W.J and T.Terasvirta (1993), “Modeling non-linear economic relationships”, Oxford: Oxford University Press.

[11] Gropp, Jeffrey, (2004), “Mean reversion of industry stock returns in the U.S., 1926-

1998”, Journal of Empirical Finance, Vol 11, pp 537-551.

[12] Gujarati, Damodar N (2004), “Basic Econometrics”, fourth edition

[13] Hamilton J (1989), “A new approach to the economic analysis of non-stationary time series and the business cycle”, Econometrica 57, 357-384

[14] Laopodis, Nikiforos T (2009), “Fiscal policy and stock market efficiency: Evidence for the United States”, The quarterly Review of Economics and Finance, Vol 49, pp 633-

[15] Marcella Niglio (2002), “Nonlinear time series models with switching structure: A comparison of their forecast performances”, Quaderni di Statistica, Vol.4.2002.

[16] McMillan D.G (2001), “Non-linear predictability of stock market returns: evidence from non-parametric and threshold models”, Department of Economics, St.

Salvator’s College, University of St Andrews, Discussion paper series No 0102

[17] Milani, Fabio (2010), “The impact of foreign stock markets on macroeconomic dynamics in open economies: A structural estimation”, Journal of International

[18] Nektarios Aslanidis (2002), “Smooth Transition Regression Models in UK Stock

Returns”, School of Economic Studies, University of Manchester.

[19] Quandt, R E (1958), The estimation of parameters of a linear regression system obeying two separate regimes, Journal of the American Statistical Association 53: 873-880

[20] R.C Merton (1973), ‘Theory of rational option pricing’, Bell Journal of Economics and Management Science, Vol 4, pp 141-183, 1973.

[21] Rodrigo Aranda and Patricio Jaramillo (2008), “Nonlinear dynamic in the Chilean stock market: Evidence from returns and trading volume”, working paper, Central Bank of Chile.

[22] Samuelson, Paul A (1965), “Rational Theory of Warrant Pricing” Indus Management rev 6 (Spring 1965): 13-31.

[23] Sarantis N (2001), “Nonlinearities, cyclical behavior and predictability in stock markets: international evidence”, International Journal of Forecasting 17, 459-482

[24] Terasvirta T and H.M Anderson (1992), “Characterizing nonlinearities in business cycles using smooth transition autoregressive models”, Journal of Applied

[25] Terasvirta, Timo (1994), “Specification, estimation and evaluation of smooth transition autoregressive models”, Journal of the American Statistical Association

[26] Tong H (1990), Non-Linear Time Series, A Dynamical System Approach, Oxford University Press, Oxford.

In his 2008 paper presented at the National Economic University workshop on financial economics and applied mathematics, Hoang Dinh Tuan explores the application of random processes to analyze stock price movements in Vietnam's stock market.

[28] Nguyen Khac Minh (2009), “Smooth Transition and GDP growth in Vietnamese industrial sector”, Journal of Economics & Development

[29] Nguyen Viet Hung (2010), “Short term prediction of VN INDEX and HNX-INDEX”,

Journal of Economics & Development, N0 152, pp 36-42

The VN-Index analysis in the autoregressive (AR) model reveals key variables including the constant trend at time t and the lagged residuals at t-1 The restrictions applied involve setting theta to zero and phi to negative theta, with the transition variable being the trend at time t The sample range spans from 3 to 706, resulting in a total of 704 observations The transition function utilized is LSTR1, and the analysis was conducted over seven iterations, providing insights into the model's estimates, standard deviations, t-statistics, and p-values.

R2: 9.9224e-01 adjusted R2: 0.9923 variance of transition variable:

SD of transition variable: 203.3716 variance of residuals: 0.0003

2 HN-Index variables in AR part: CONST HNX_trend(t) str_resids(t-

1) restriction theta=0: restriction phi=0: CONST restriction phi=-theta: transition variable: HNX_trend(t) sample range: [3, 386], T = 384

Chuyên đề thực tập Tốt nghiệp transition function: LSTR1 number of iterations: 17 variable start estimate SD t-stat p-value

The AGF variables in the AR component include the constant AGF_Trend(t) and the lagged residuals str_resids(t-1), with specific restrictions applied: theta equals zero, phi equals zero, and phi equals negative theta The transition variable is AGF_Trend(t), analyzed over a sample range of [3, 336] with a total of T = 334 observations The transition function utilized is LSTR1, and the analysis was conducted over six iterations, providing essential statistical metrics such as variable start estimates, standard deviation, t-statistics, and p-values.

In the AR part of the BMC model, key variables include CONST, BMC_trend(t), and str_resids(t-1), with restrictions applied to theta and phi The transition variable is identified as BMC_trend(t), and the sample range spans from 3 to 506, yielding a total of T = 504 observations The transition function utilized is LSTR1, with the analysis conducted over five iterations Results include variable start estimates, standard deviations, t-statistics, and p-values, providing crucial insights into the model's performance.

Chuyên đề thực tập Tốt nghiệp variance of residuals: 0.0009

The FPT variables in the AR part include CONST, FPT_trend(t), and str_resids(t-1), with specific restrictions applied: theta=0, phi=0, and phi=-theta The transition variable is identified as FPT_trend(t), with a sample range of [3, 463] and a total of T = 461 The transition function utilized is LSTR1, and the analysis involved five iterations, providing estimates for standard deviation, t-statistics, and p-values for the variables.

6 PVD variables in AR part: CONST PVD_trend(t) str_resids(t-1) restriction theta=0: str_resids(t-1) restriction phi=0: restriction phi=-theta:

Chuyên đề thực tập Tốt nghiệp transition variable: PVD_trend(t) sample range: [3, 400], T = 398 transition function: LSTR1 number of iterations: 4 variable start estimate SD t-stat p-value

The analysis of SSI variables in the AR component includes the transition variable TREND(t) and utilizes a sample range of [3, 433], resulting in T = 431 The transition function applied is LSTR1, with two iterations conducted Key restrictions include setting theta to 0 and phi to both 0 and -theta The output details the variable start estimates, standard deviations, t-statistics, and p-values for the resulting model.

Chuyên đề thực tập Tốt nghiệp adjusted R2: 0.9933 variance of transition variable:

II ADF Test for reidual of data series

ADF Test for series: u_resids sample range: [1961 Q2, 2136 Q2], T = 701 lagged differences: 2 no intercept, no time trend asymptotic critical values reference: Davidson, R and MacKinnon, J (1993),

"Estimation and Inference in Econometrics" p 708, table 20.1,

-2.56 -1.94 -1.62 value of test statistic: -14.9965 regression results:

OPTIMAL ENDOGENOUS LAGS FROM INFORMATION CRITERIA sample range: [1963 Q2, 2136 Q2], T = 693 optimal number of lags (searched up to 10 lags of 1 differences):

ADF Test for series: str_resids sample range: [1961 Q2, 2056 Q2], T = 381 lagged differences: 2 intercept, time trend asymptotic critical values reference: Davidson, R and MacKinnon, J (1993),

ADF Test for series: str_resids sample range: [1961 Q2, 2059 Q4], T = 395 lagged differences: 2 no intercept, no time trend asymptotic critical values reference: Davidson, R and MacKinnon, J (1993),

Tiêu đề	Non-linear Model Predictability of Vietnam Stock Market Price
Tác giả	Dinh Van Tuan
Người hướng dẫn	Ph.D. Nguyen Viet Hung
Trường học	National Economics University
Chuyên ngành	Development Economics
Thể loại	thesis
Năm xuất bản	2012
Thành phố	Hanoi

Định dạng
Số trang	65
Dung lượng	1,1 MB