Lan tỏa suất sinh lợi và độ biến thiên giữa các thị trường chứng khoán tt tiếng anh

MINISTRY OF EDUCATION AND TRAINING
HO CHI MINH CITY OPEN UNIVERSITY

LE DINH NGHI
RETURN AND VOLATILITY SPILLOVER
AMONG STOCK MARKETS

A DISSERTATION SUBMITTED FOR THE DEGREE OF
DOCTOR OF PHYLOSOPHY
MAJOR: BUSINESS ADMINISTRATION

Ho Chi Minh city, 2019


i

The dissertation was completed at:

HO CHI MINH CITY OPEN UNIVERSITY

Academic advisors:
Assoc. prof. Nguyen Minh Kieu, PhD.
Reviewer 1: ……………………………………………………………………..
Reviewer 2: ……………………………………………………………………..

This dissertation will be presented at the dissertation committee at Ho Chi Minh
City Open University.
Ho Chi Minh City, …………………………….. , 2019

This dissertation can be found at …………..

1

CHAPTER 1. INTRODUCTION

1.1. Introduction
1.2. Research context and rationale of the study
1.2.1. The Vietnamese Economy and Vietnamese Stock Market

More than thirty years after the reforms called doi moi (renovation) from 1986, the
Vietnamese economy become more integrated into the world economy, by becoming a
member of Free Trade Agreements (FTA), increasing import - export and Foreign Direct
Investment (FDI) activities.
A stock market is a financial market with an important role in the economy. Influencing
by being more integrated economy, Vietnam’s stock market becomes more closely related to
world financial markets. The US is the world's largest economy, and Japan and Korea are
large economies in Asia, so they can affect other countries, including Vietnam.
Hence, Vietnam’s stock market may be affected by the US, Japanese and Korean stock
markets. Because of this, investigating the influence of the US, Japanese and Korean stock
markets on Vietnamese stock market is needed to help investors and policy makers have
information to support their investment decisions
1.2.2. Research context

Investment decision, a part of financial management, is one of the important managerial
decisions. Stock market is an investment channel where company and investors can invest to
gain the benefits. Hence, a stock market is a financial market with an important role in the
economy. Stock returns and risks are the factors that are the most considered by investors in
their investment decisions.
Stock returns can be measured by price growth rate. Risk can be proxied by volatility,
measured by standard deviation of stock returns around the mean and can be estimated by

General Autoregressive Conditional Heteroskedasticity (GARCH) model (Tim Bollerslev
1986)


2

In this era of globalization, the financial systems of countries may be linked. Hence,
research on relationships between stock markets help both investors and policy makers obtain
suitable information for making their decisions. Spillover is a result of the interdependence
among market economies This interdependence means that shocks, whether of global or local
nature, can be transmitted across countries because of their financial linkages (Abou-Zaid,
2011)
Therefore, research on return and volatility spillovers among stock markets plays an
important role to help investors and policy makers have information for their investment
decisions. Although spillover effects among stock markets have been confirmed in many
studies, such as Ng (2000), Miyakoshi (2003), Ergun & Nor (2010), Sakthivel, Bodkhe, &
Kamaiah (2012), Kharchenko & Tzvetkov (2013), Nishimura, Tsutsui, & Hirayama (2015),
Yarovaya, Brzeszczyński, & Lau (2016), Bahadur, Kothari, & Thagurathi (2016), Jebran,
Chen, Ullah, & Mirza (2017), Ishfaq & Rehman (2018), to the best of our knowledge, the
previous literature has not explored return and volatility spillovers from developed markets
to the Vietnamese stock market. Hence, this dissertation aims to test the return and volatility
spillovers from the US, Japanese and Korean markets to the Vietnamese stock market.
In reality, short- and long-term investors may have different considerations. Short-term
investors focus more on the relationship at higher frequencies, that is, short-term fluctuations,
whereas long-term investors focus on the relationship at lower frequencies, i.e., long-term
fluctuations (Gradojevic, 2013). Hence, long- and short-term return and volatility spillovers
should be analyzed separately to reveal more precise information for different investors.
Frequency-domain analysis, i.e., spectral analysis, can be used in this situation.
Despite its usefulness, frequency domain research is relatively scarce in the empirical
economics and finance literatures (Gradojevic, 2013). Few studies used this approach to

investigate the spillover among stock markets. To our best knowledge, only Gradojevic
(2013) investigated return spillover among five regional stock exchanges (Serbia, Croatia,
Slovenia, Hungary, and Germany) in the frequency domain; and volatility spillover analysis
using frequency domain approach was not previously reported in the literature. Therefore,
applying frequency domain approach to analyze spillover effects is needed to fulfill the
theoretical gap and provide more precise information for both long- and short-term investors.


3

1.2.3. Rationale of the study

From above analysis, this dissertation entitled “Return and volatility spillover among stock
markets”. Using frequency domain analysis, this study examines return and volatility
spillovers from the U.S., Japanese and Korean stock markets to the Vietnamese stock market.
The results will confirm the relationship among stock markets at different frequencies, and
help short- and long-term investors obtain more precise information to support their
investment decisions.
1.3. Research problem

This dissertation investigates the relationship among stock markets. In particular, this
study examines return and volatility spillovers from the US, Japanese and Korean markets to
the Vietnamese stock market.
1.4. Research aim and research question
1.4.1. Research aim

- Testing the return and volatility spillovers from the U.S., Japanese and Korean stock
markets to the Vietnamese stock market.
- Using frequency domain analysis to test the return and volatility spillovers among stock
markets at different frequencies

1.4.2. Research question

This study aims to answer the following research questions:
-

Are there return spillovers from the U.S., Japanese and Korean stock markets to the
Vietnamese stock market?

-

Are there volatility spillovers from the U.S., Japanese and Korean stock markets to the
Vietnamese stock market?

-

Are the statistical test values of return spillovers from the U.S., Japanese and Korean
stock markets to the Vietnamese stock market not the same at different frequencies?

-

Are the statistical test values of volatility spillovers from the U.S., Japanese and
Korean stock markets to the Vietnamese stock market not the same at different
frequencies?


4

1.5. Research data and Research Method

Daily data from the Standard & Poor’s 500 (S&P 500) Composite Index, the Nikkei 225,

KOSPI and the Vietnam Stock Index (VN-Index); a proxy for the US, Japanese, Korean and
Vietnamese stock indices from January 1, 2012, to December 31, 2015, is collected from
Thomson Reuters Datastream.
The quantitative is applied in this study. The GARCH model (Bollerslev, 1986) is used to
estimate volatilities in these stock markets, the Granger Causality Test (Granger, 1969) is
used to examine return and volatility spillovers, and the test for causality in the frequency
domain (Breitung & Candelon, 2006) is used to examine return and volatility spillovers at
different frequencies.
1.6. Contributions of the study
1.6.1. Theoretical contribution

In this era of globalization, research on return and volatility spillovers among stock
markets plays an important role that provide information to investors and policy-makers.
Hence, there are many studies investigating return and volatility spillovers among stock
markets. However, most previous literatures could not analyze spillover effects at different
cycles. Few studies use frequency domain approach to investigate the return spillover among
stock markets. Especially, to our best understanding, volatility spillover analysis using
frequency domain approach were not previously reported in literature. Hence, using
frequency domain approach, this study analyzes return and volatility spillovers at different
cycles. The results offer deep insight into spillover effects among stock markets at different
frequencies. This is the theoretical contribution of this study.
Moreover, although spillover effects among stock markets have been confirmed in many
studies, to our best knowledge, the previous literature has not explored return and volatility
spillovers from developed markets to the Vietnamese stock market. Hence, this study
analyzes the return and volatility spillovers from the developed markets to the Vietnamese
stock market. The results offer the relationship between Vietnamese stock market and
developed markets
1.6.2. Empirical contributions

Research on relationships between stock markets help both investors and policy makers

obtain suitable information for making their decisions. In particular, market index in


5

Vietnamese stock exchange can be determined by the US, Japanese and Korean market
indices if they are fully integrated. In this case, investors and policy makers from Vietnam
should follow information and fluctuations in overseas markets when making their
corresponding decisions. However, if the Vietnamese stock market do not move together with
foreign markets, then foreign investors will benefit from the reduction in the portfolio risk,
by diversification that includes domestic stocks.
Moreover, the results also provide more suitable information for short- and long-term
investors. In particular, short-term investors and long-term investors can make their decisions
based on spillover effects among stock markets at high frequencies and low frequencies,
respectively.
1.7. Dissertation structure

To fulfill above mentioned research objectives, the dissertation is structured as follows:
Chapter 1: Introduction, Chapter 2: Theoretical Frameworks, Chapter 3: Research
methodology, Chapter 4: Data Analysis and Research Results, Chapter 5: Conclusions.


6

CHAPTER 2. THEORETICAL FRAMEWORKS
2.1. Introduction
2.2. Return

The return 𝑟𝑡 is computed using the following equation (Campbell, Lo, & MacKinlay,
1997):

𝑟𝑡 = 𝑙𝑛

𝑃𝑡
𝑃𝑡−1

where 𝑃𝑡 is the market index at time t, 𝑙𝑛(𝑥) is the natural logarithm of 𝑥.
2.3. Volatility
2.3.1. Definition

Volatility is a measure of the dispersion in a probability density (Alexander, 2001). The
most common measure of dispersion is the standard deviation of a random variable.
Accordingly, the higher the volatility, the higher the stock risk.
2.3.2. Constant and time–varying volatility

Constant volatility models only refer to the unconditional volatility of a returns process.
Time–varying volatility models describe a process for the conditional volatility. Generalized
Auto Regressive Conditional Heteroscedastic (GARCH), proposed by Bollerslev (1986), is a
useful model to measure time–varying volatility.
2.3.3. Volatility Model.

The conditional mean and variance of 𝑟𝑡 are represented below (Tsay, 2005)



2
t  Ert | Ft 1  ,  t  Var rt | Ft 1   E rt  t  Ft 1
2



where 𝐹𝑡−1 denotes the information set available at time 𝑡 − 1. Typically, 𝐹𝑡−1 consists of all
linear functions of the past returns.
2.3.4. The Auto Regressive Conditional Heteroscedastic (ARCH) model

The ARCH model can be used for volatility modeling, proposed by Engle (1982), as
below:


7

at   t  t

 t2   0  1at21  ...   m at2m

where  t  is a sequence of independent and identically distributed (iid) random variables
with mean zero and variance 1,  0  0 and  i  0 for i  0 .
2.3.5. The Generalized Auto Regressive Conditional Heteroscedastic (GARCH)

model
Bollerslev (1986) proposes a useful extension of ARCH model, known as the generalized
ARCH (GARCH) model. For a return series 𝑟𝑡 , let at  rt  t be the innovation at time t.
Then 𝑎𝑡 follows a GARCH(m, s) model if

at   t t ,

m

s

i 1

j 1

 t2   0   i at2i    j t2 j

where  t  is a sequence of independent and identically distributed random variables with
mean 0 and variance 1,  0  0 ,  i  0 ,  j  0 , and

max( m ,s )

 
i 1

i

  i   1 . Here it is understood that

i  0 for 𝑖 > 𝑚 and  j  0 for 𝑗 > 𝑠.  t is often assumed to be a standard normal or
standardized Student-t distribution or generalized error distribution. Above equation reduces
to a pure ARCH(m) model if s=0
2.4. Spillover.

Spillover is a result of the interdependence among market economies This
interdependence means that shocks, whether of global or local nature, can be transmitted
across countries because of their financial linkages (Abou-Zaid, 2011). The transmission in
return and volatility are called return spillover and volatility spillover, respectively.
2.5. Literature review about return and volatility spillover

Engle (1982) proposes ARCH model that modified to GARCH model by Bollerslev
(1986). Based on these models, many studies estimate the volatility and test return and

volatility spillover among stock markets, such as Hamao, Masulis, & Ng (1990), Ng (2000),


8

Miyakoshi (2003), Ergun & Nor (2010), Sakthivel, Bodkhe, & Kamaiah (2012), Kharchenko
& Tzvetkov (2013), Nishimura, Tsutsui, & Hirayama (2015), Yarovaya, Brzeszczyński, &
Lau (2016). In Vietnam, Vương Quân Hoàng (2004) is the first study investigating volatility
in the Vietnamese stock market. After that, some research do their studies about volatility in
Vietnamese stock market, such as Nguyễn Thu Hiền & Lê Đình Nghi (2010) and Nghi (2012).
However, to our best knowledge, the previous literature has not explored spillover effects
from developed markets to the Vietnamese stock market.
2.6. Time domain and frequency domain
2.6.1. Introduction

Frequency is the number of occurrences of a repeating event per unit time. In other words,
the number of cycles per unit of time is called the frequency.
Most econometrics techniques, including regression, ARCH, GARCH models, Granger
Causality Test, analyze data in time domain. However, using time domain analysis, it’s hard
to explore different frequency components in financial time series. Therefore, frequency
domain analysis is needed to analyze financial data at different frequencies.
2.6.2. Frequency-domain representation of the data.

Time series can be represented in frequency domain. The frequency domain representation
𝑋(𝑓) of time series data 𝑥(𝑡) is called spectrum of 𝑥(𝑡). Therefore, it’s easy to discover
different frequency components of financial data based on spectrum 𝑋(𝑓) of time series 𝑥(𝑡).
2.6.3. Fourier transform

The Fourier transform is a method that transfer data from time domain to frequency
domain and vice versa.

2.7. Frequency domain analysis

By transferring data between time domain and frequency domain, frequency domain
method can analyze financial data at different frequencies. This method is also useful in
causality analysis. Causal relations in the frequency domain were first proposed by Granger
(1969). Then, some other methods were developed by Geweke (1982), Hosoya (1991), and
Breitung and Candelon (2006). The Breitung and Candelon (2006)’s approach is used in this
study.


9

2.8. Literature review about frequency domain analysis in economics and finance

Some studies investigate methods to decompose time series to different frequency
components, such as Baxter & King (1999), Buss (2010), Larsson & Vasi (2012), Hodrick &
Prescott (1997), Ravn & Uhlig (2002), …
Moreover, some research propose methods to test causality relations between two time
series at different frequency, such as Geweke (1982), Hosoya (1991), and Breitung and
Candelon (2006). Based on these methods, Yanfeng (2013), Chan, Lien, & Weng (2008),
Gradojevic (2013), Ozer & Kamisli (2016) analyze the frequency causality between time
series of different financial and economic data.
Although there are some studies analyzing frequency return spillover among stock
markets, to our best knowledge, volatility spillover analysis using frequency domain approach
was not previously reported in the literature. Hence, based on frequency domain analysis, this
study examines return and volatility spillovers from the US, Japanese and Korean markets to
the Vietnamese stock market at different frequencies. This is the main contribution of this
dissertation

10

CHAPTER 3. RESEARCH METHODOLOGY
3.1. Introduction
3.2. Research models

This study proposes the model for testing return spillover from the US to Vietnamese stock
markets using Granger Causality Test, based on VAR model, as below
𝑟𝑉𝑁,𝑡 = 𝛼0 + 𝛼1 𝑟𝑉𝑁,𝑡−1 + ⋯ + 𝛼𝑙 𝑟𝑉𝑁,𝑡−𝑙 + 𝛽1 𝑟𝑈𝑆,𝑡−1 + ⋯ + 𝛽𝑙 𝑟𝑈𝑆,𝑡−𝑙 + 𝜀𝑡
𝑟𝑈𝑆,𝑡 = 𝛾0 + 𝛾1 𝑟𝑈𝑆,𝑡−1 + ⋯ + 𝛾𝑙 𝑟𝑈𝑆,𝑡−𝑙 + 𝛿1 𝑟𝑉𝑁,𝑡−1 + ⋯ + 𝛿𝑙 𝑟𝑉𝑁,𝑡−𝑙 + 𝑢𝑡
where 𝑟𝑈𝑆,𝑡 and 𝑟𝑉𝑁,𝑡 are returns of US and Vietnamese stock markets, respectively, at time 𝑡
and 𝛼0 , 𝛼1,… 𝛼𝑙 , 𝛽1 , … , 𝛽𝑙 , 𝛾0 , … , 𝛾𝑙 and 𝛿1 , … , 𝛿𝑙 are regression coefficients. If 𝛽1 , … , 𝛽𝑙 are
statistically significant, return spillover from the US to Vietnamese stock markets exists.
The similar models can be applied to test return spillover from Japanese and Korean to
Vietnamese stock markets.
Similarly, Granger Causality Test could be used to test volatility spillover among stock
markets, where volatilities are used instead of returns. The US, Japanese, Korean and
Vietnamese volatilities σ2t are estimated by GARCH model (Bollerslev, 1986).
Based on theoretical frameworks, literature review and the integration of Vietnam’s
economy, this study proposes the following hypotheses
-

Hypothesis 1 (and 2, 3): There are return spillovers from S&P 500 (and Nikkei 225,
KOSPI) to VN-Index.

-

Hypothesis 4 (and 5, 6): There are volatility spillovers from S&P 500 (and Nikkei 225,
KOSPI) to VN-Index.

Moreover, because some research results, such as Chan et al. (2008), Gradojevic (2013),
Yanfeng (2013), Ozer & Kamisli (2016), confirm that causality results can differ between
frequency bands (Granger and Lin 1995), this study proposes the models for testing return
spillover at different frequencies, as below:
M𝑟𝑈𝑆→𝑟𝑉𝑁 (ω) = 0
M𝑟𝐽𝑃 →𝑟𝑉𝑁 (ω) = 0
M𝑟𝐾𝑅→𝑟𝑉𝑁 (ω) = 0


11

where,
-

𝑟𝑈𝑆 , 𝑟𝐽𝑃 , 𝑟𝐾𝑅 and 𝑟𝑉𝑁 are returns of the US, Japanese, Korean and Vietnamese stock
indices, respectively.

-

My→x (ω) = 0 is the linear restriction to test the hypothesis that y does not cause x at
frequency 𝜔. This approach is proposed by Breitung & Candelon (2006) and included
more details in section 3.6.

The similar approach is applied to test volatility spillovers among stock markets at
different frequencies, where volatilities are used instead of returns.
This study proposes the following hypotheses:
-

Hypothesis 7 (and 8, 9): Return spillovers from S&P 500 (and Nikkei 225, KOSPI) are
not the same at different frequencies.

-

Hypothesis 10 (and 11, 12): Volatility spillovers from S&P 500 (and Nikkei 225,
KOSPI) are not the same at different frequencies.

3.3. Data collection

Daily data from the Standard & Poor’s 500 (S&P 500) Composite Index, the Nikkei 225,
KOSPI and the Vietnam Stock Index (VN-Index); a proxy for the US, Japanese, Korean and
Vietnamese stock indices from January 1, 2012, to December 31, 2015, is collected from
Thomson Reuters Datastream.
3.4. Return spillover among stock markets
3.4.1. Granger Causality Test

A Granger Causality Test (Granger, 1969) is a useful tool for examining return spillover
among financial markets. A Granger-causality test can be performed using vector
autoregression (VAR) as follows (Gujarati 2004):

yt  0  1 yt 1  ...  l yt l  1xt 1  ...  l xt l   t

xt  0  1xt 1  ...  l xt l  1 yt 1  ...  l yt l  ut
and test the null hypothesis:

1  2  ...  l  0
for each equation. The null hypothesis is that x does not Granger-cause y in the first regression
and that y does not Granger-cause in x the second regression.


12

3.4.2. Return spillover analysis

The Granger Causality Test is applied to returns of S&P 500 (or Nikkei 225, KOSPI) and
VN-Index to examine return spillovers from the US (or Japanese, Korean) to Vietnamese
stock markets.
3.5. Volatility spillover among stock markets
3.5.1. Volatility estimation using GARCH model

GARCH model is used to estimate volatilities of markets
3.5.2. Volatility spillover analysis

The Granger Causality Test is applied to volatilities of S&P 500 (or Nikkei 225, KOSPI)
and VN-Index to examine volatility spillovers from the US (or Japanese, Korean) to
Vietnamese stock markets.
3.6. Return and volatility spillovers in frequency domain
3.6.1. Testing for Causality: A Frequency-Domain Approach

The frequency-domain causality test developed by Breitung and Candelon (2006) is based
on the framework of Geweke (1982) and Hosoya (1991). Let 𝑧𝑡 = [𝑥𝑡 , 𝑦𝑡 ]′ be a twodimensional time-series vector with 𝑡 = 1, … 𝑇. It is assumed that zt has a finite-order VAR
representation:
𝛩(𝐿)𝑧𝑡 = 𝜀𝑡

where 𝛩(𝐿) = 𝐼 − 𝛩1 𝐿 − ⋯ − 𝛩1 𝐿𝑝 is a 2×2 lag polynomial with 𝐿𝑘 𝑧𝑡 = 𝑧𝑡−𝑘 . It is assumed
that the vector εt is white noise, with E(εt ) = 0 and E(εt ε′t ) = Σ, where Σ is a positive definite
matrix. Next, let G be the lower triangular matrix of the Cholesky decomposition G′ G = Σ −1 ,
such that E(ηt η′t ) = I and ηt = Gεt . The system is assumed to be stationary, implying the
following moving average (MA) representation:
zt = Φ(L)εt = [

Φ11 (L) Φ12 (L) ε1t
][ ]
Φ21 (L) Φ22 (L) ε2t

= Ψ(L)ηt = [

Ψ11 (L) Ψ12 (L) η1t
][ ]
Ψ21 (L) Ψ22 (L) η2t


13

where Φ(L) = Θ(L)−1 and Ψ(L) = Φ(L)G−1 . Using a Fourier transformation on this
representation, the spectral density of 𝑥𝑡 can be expressed as:
2

1

2

fx (ω) = 2π {|Ψ11 (e−iω )| + |Ψ12 (e−iω )| }

The measure of causality suggested by Geweke (1982) and Hosoya (1991) is defined as:
My→x (ω) = log [

2πfx (ω)
|Ψ11 (e−iω )|

2

]

2

= log [1 +

|Ψ12 (e−iω )|

2

|Ψ11 (e−iω )|

]

To test the hypothesis that 𝑦 does not cause 𝑥 at frequency ω, we use the following null
hypothesis:
My→x (ω) = 0

Breitung and Candelon (2006) show that the null hypothesis My→x (ω) = 0 is equivalent
to a linear restriction on the VAR coefficients. First, they use Ψ(L) = Θ(L)−1 G−1 and
Ψ12 (L) = −

g22 Θ12 (L)
|Θ(L)|

(where 𝑔22 is the lower diagonal element of 𝐺 −1 and |Θ(L)| is the

determinant of Θ(L)) to express the null hypothesis as:
|𝛩12 (𝑒 −𝑖𝜔 )| = |∑𝑝𝑘=1 𝜃12,𝑘 𝑐𝑜𝑠(𝑘𝜔) − ∑𝑝𝑘=1 𝜃12,𝑘 𝑠𝑖𝑛(𝑘𝜔)𝑖| = 0

where θ12,k is the (1,2)-element of Θk . Thus, a necessary and sufficient set of conditions for
|Θ12 (e−iω )| = 0 is:
∑𝑝𝑘=1 𝜃12,𝑘 𝑐𝑜𝑠(𝑘𝜔) = 0
∑𝑝𝑘=1 𝜃12,𝑘 𝑠𝑖𝑛(𝑘𝜔) = 0


14

The notation can be simplified by letting 𝑎𝑗 = θ11,j and βj = θ12,j . Then, the VAR
equation for 𝑥𝑡 can be written as:
𝑥𝑡 = 𝑎1 𝑥𝑡−1 + ⋯ + 𝑎𝑝 𝑥𝑡−𝑝 + 𝛽1 𝑦𝑡−1 + ⋯ + 𝛽𝑝 𝑦𝑡−𝑝 + 𝜀1𝑡

The hypothesis My→x (ω) = 0 is equivalent to the linear restriction:
H0 : R(ω)β = 0
′

where β = [β1 , … , βp ] and:
R(ω) = [

cos(ω) cos(2ω) … cos(pω)
]
sin(ω) sin(2ω) … sin(pω)

As in the conventional causality test, the Wald test statistic based on the linear restriction
in above equation is asymptotically distributed as χ2 (2) for ω ∈ (0, π) (Yanfeng, 2013).
Therefore, the distribution χ2 (2) can be used to test the hypothesis My→x (ω) = 0 at frequency
ω.
3.6.2. Return spillover analysis in frequency domain

The frequency domain causality test (Breitung & Candelon, 2006) is applied to returns of
S&P 500, Nikkei 225, KOSPI and VN-Index to examine the return spillovers from the US,
Japanese, Korean to Vietnamese stock markets at different frequencies.
3.6.1. Volatility spillover analysis in frequency domain

The frequency domain causality test (Breitung & Candelon, 2006) is applied to volatilities
of S&P 500, Nikkei 225, KOSPI and VN-Index to examine the volatility spillovers from the
US, Japanese, Korean to Vietnamese stock markets at different frequencies.


15

CHAPTER 4. DATA ANALYSIS AND RESEARCH
RESULTS
4.1. Introduction
4.2. Market indices in research period

This section represents S&P 500, Nikkei 225, KOSPI and VN-Index indices from January
1, 2012, to December 31, 2015. All indices have tended to increase during research period.
4.3. Descriptive statistics

Table 4.1 lists some descriptive statistical properties of daily market returns in the four
countries
Table 4.1: Descriptive Statistics of Daily Returns on the US, Japanese, Korean and
Vietnamese Stock Indices
S&P500

Nikkei 225

KOSPI

VN-Index

Mean

0.000466

0.000778

0.000068

0.000478

Median

0.000195

0.000278

0.000000

0.000101

Standard Deviation

0.007916

0.013154

0.007830

0.011053

Skewness

-0.256127

-0.356637

0.010041 -0.609907

Kurtosis

5.054130

6.255896

4.601093

5.787362

Source: author’s calculation
4.4. Return spillover among stock markets
4.4.1. Stationarity tests on time series

The ADF (Augmented Dickey-Fuller) test results indicate that all the US, Japanese,
Korean and Vietnamese market index return time series are stationary.
4.4.2. Return spillovers tests

The results in Table 4.2 show significant return spillovers from the US to the Vietnamese

stock markets at the 1% significance level, from the Japanese to the Vietnamese stock markets
at the 10% significance level and from the Korean to the Vietnamese stock markets at the 5%
significance level. These results are consistent with reality because the US is the world's
largest economy, and Japan and Korea are large economies in Asia, so they can affect other
countries, including Vietnam. This also indicates integration of the Vietnamese economy in


16

the world economy. These results support the conclusion of Tsutsui & Hirayama (2005) that
most, if not all, the literature offers evidence on the existence of stock market linkage.
Table 4.2: Return spillover from the US, Japanese and Korean stock markets to the
Vietnamese stock market
Granger-Causality Test
Hypothesis
H0

S&P 500 returns do not Nikkei 225 returns do KOSPI returns do not
Granger-cause

VN- not Granger-cause VN- Granger-cause

Index returns

Index returns

VN-

Index returns

F-Statistic

34.3253

2.51135

2.55941

Conclusion

Rejected at the 1%

Rejected at the 10%

Rejected at the 5%

significance level

significance level

significance level

(Not rejected at the 5%
significance level)
Source: author’s calculation
Moreover, the test for return spillover from Japanese to Vietnamese stock markets shows
that H0 is rejected at the 10% significance level, but not at the 5% significance level, which
indicates that the evidence on return spillover between these markets is not clear. Thus, US
and Korean markets influence on the Vietnamese market tends to be greater than that of the
Japanese market.

4.5. Volatility spillover among stock markets.
4.5.1. GARCH estimation

The ADF test results indicate that all the US, Japanese, Korean and Vietnamese market
index return time series are stationary. The volatilities of the four markets returns are
estimated by applying GARCH models to market returns. The estimated volatility data of the
U.S, Japanese, Korean and Vietnamese stock indices are tested for stationarity with the ADF
test. The results show that all volatility time series are stationary
4.5.2. Volatility spillovers tests

Volatility spillover from the US, Japanese and Korean markets to Vietnamese market are
in Table 4.3


17

Table 4.3: Volatility spillover from the US, Japanese and Korean markets to
Vietnamese market
Granger-Causality Test
Hypothesis

S&P 500 volatility does Nikkei 225 volatility KOSPI volatility does

H0

not Granger-cause VN- does not Granger-cause not Granger-cause VNIndex volatility

VN-Index volatility

Index volatility

F-Statistic

4.41116

0.54608

3.72983

Conclusion

Rejected at the 1%

Not rejected at the 10%

Rejected at the 5%

significance level

significance level

significance level

Source: author’s calculation
The results in Table 4.3 show significant volatility spillovers from the US and Korean
markets to the Vietnamese stock market. However, volatility spillover from Japanese to
Vietnamese markets is not found at the 10% significance level, i.e., the transmission of shocks
from Japanese to Vietnamese markets is insignificant
4.6. Return spillovers in frequency domain

A frequency-domain causality test (Breitung & Candelon 2006) is applied to examine
return spillover from the US market to Vietnamese market at different frequencies. The results
are in Table 4.4
The results in Table 4.4 show a significant return spillover from the US to the Vietnamese
stock markets at all frequencies. These results are consistent with the results of return spillover
testing using a traditional Granger-causality test in Table 4.2. They also show that the
statistical test values (χ2 distribution values) are not the same at different frequencies. The
results support the hypothesis that causality is not the same at different frequencies (Granger
and Lin 1995). However, these differences are small, and the spillover results are the same at
all frequencies.
Next, this study examines return spillover from the Japanese to the Vietnamese stock
market in the frequency domain. Because evidence of return spillover from Japan to Vietnam
in the time domain is not clear—that is, the null hypothesis is rejected at the 10% significance
level, but not at the 5% significance level—this test is performed at both 5% and 10%
significance level to obtain deeper insight on this relation. The results are in Table 4.5.


18

Table 4.4: Return Spillover from the US to Vietnamese Stock Markets in the
Frequency Domain
Hypothesis H0
S&P 500 returns do not Granger cause VN-Index returns
Frequency ω

Cycles

𝑻=

𝟐𝝅

𝝎

Test Statistic

Conclusion

𝛘𝟐

(days)

0.0100

628

31.17

H0 rejected

0.3567

18

30.69

H0 rejected

0.7033

9

29.04

H0 rejected

1.0500

6

25.40

H0 rejected

1.3967

5

19.97

H0 rejected

1.7433

4

18.66

H0 rejected

2.0900

3

22.73

H0 rejected

2.4367

3

26.23

H0 rejected

2.7833

2

27.98

H0 rejected

3.1300

2

28.52

H0 rejected

Source: author’s calculation
The results in Table 4.5 show that, at the 5% significance level, the return spillover effect
is not supported at all frequencies. These results are consistent with results of return spillover
testing using a traditional Granger-causality test in Table 4.2. However, at the 10%
significance level, the conclusions changed.


19

Table 4.5: Return Spillover from the Japanese to the Vietnamese Stock Markets in the
Frequency Domain
Hypothesis H0
Nikkei 225 returns do not Granger cause VN-Index returns
Frequency

Cycles

ω
𝑻=

𝟐𝝅

Test
Statistic

𝝎

(days)

𝟐

Conclusion
10% significance

5% significance

level

level

(χ2 = 4.61)

(χ2 = 5.99)

𝛘

0.0100

628

2.24

H0 not rejected

H0 not rejected

0.3567

18

2.20

H0 not rejected

H0 not rejected

0.7033

9

2.13

H0 not rejected

H0 not rejected

1.0500

6

2.16

H0 not rejected

H0 not rejected

1.3967

5

2.76

H0 not rejected

H0 not rejected

1.7433

4

4.23

H0 not rejected

H0 not rejected

2.0900

3

5.26

H0 rejected

H0 not rejected

2.4367

3

5.49

H0 rejected

H0 not rejected

2.7833

2

5.47

H0 rejected

H0 not rejected

3.1300

2

5.44

H0 rejected

H0 not rejected

Source: author’s calculation
The results in Table 4.5 show that the statistical test values (χ2 distribution values) are not
the same at different frequencies. Thus, the frequency-domain approach provides more
information than a traditional Granger-causality test. In this case, although the results from a

time-domain analysis indicate that, at the 10% significance level, there is significant return
spillover from the Japanese to the Vietnamese stock market, the frequency-domain approach
shows that this conclusion is not true at any frequency. Example, for ω=0.35667,
corresponding to the approximately eighteen-day cycle, the statistical test value is 2.20; but
for ω=2.7833; corresponding to the approximately two-day cycle, the statistical test value is
5.47. Therefore, at the 10% significance level, the null hypothesis is not rejected for


20

ω=0.35667, but rejected for ω=2.7833, that is, there is significant return spillover from the
Japanese to the Vietnamese stock markets at ω=2.7833, but not at ω=0.35667. Table 4.5 also
indicates that short-term investors (cycles less than or equal to three days) should take note
of the Nikkei 225 returns to obtain more information for their investment decisions, but it is
not necessary for long-term investors (cycles longer than three days) investors. These results
support the hypothesis that causality is not the same at different frequencies (Granger and Lin
1995). This implies that investors should make different decisions based on investment
cycles.
Table 4.6: Return Spillover from the Korean to the Vietnamese Stock Markets in the
Frequency Domain
Hypothesis H0
KOSPI returns do not Granger cause VN-Index returns
Frequency ω

Cycles

𝑻=

𝟐𝝅
𝝎

(days)

Test Statistic

Conclusion

𝛘𝟐

5% significance level
(χ2 = 5.99)

0.0100

628

4.78457935003835

H0 not rejected

0.3567

18

4.10883449393248

H0 not rejected

0.7033

9

1.9652536359483

H0 not rejected

1.0500

6

0.184566517075998

H0 not rejected

1.3967

5

3.32678406536558

H0 not rejected

1.7433

4

6.38316809610236

H0 rejected

2.0900

3

7.96589208064529

H0 rejected

2.4367

3

8.88393503396178

H0 rejected

2.7833

2

9.24762577310448

H0 rejected

3.1300

2

9.33987363447699

H0 rejected

Source: author’s calculation


21

Table 4.6 show the return spillover from Korean to Vietnamese stock markets in
frequency domain. The results also indicate that the statistical test values (χ2 distribution
values) are not the same at different frequencies. Thus, the frequency-domain approach
provides more information than a traditional Granger-causality test. Although the results from
a time-domain analysis indicate that, at the 5% significance level, there is significant return
spillover from Korean stock market to the Vietnamese stock market, the frequency-domain
approach shows that this conclusion is only true at high frequencies. In particular, at cycles
less than or equal to four days, the null hypothesis is rejected, but at cycles longer than four
days, the null hypothesis is not rejected. Therefore, short-term investors (cycles less than or
equal to four days) should take note of the KOSPI returns to obtain more information for their
investment decisions, but it is not necessary for long-term investors (cycles longer than four
days) investors.
4.7. Volatility spillovers in frequency domain

A frequency-domain causality test (Breitung & Candelon 2006) is applied to examine
volatility spillover from the US market to Vietnamese market at different frequencies. The
results are in Table 4.7
Table 4.6: Volatility Spillover from the US to Vietnamese Stock Markets in the
Frequency Domain
Hypothesis H0
S&P 500 volatility do not Granger cause VN-Index
volatility
Frequency

Cycles

Test Statistic

Conclusion

𝛘𝟐

5% significance level

ω
𝑻=

𝟐𝝅
𝝎

(days)
0.0100

628

11.8668

H0 rejected

0.3567

18

11.7296

H0 rejected

0.7033

9

11.8028

H0 rejected

1.0500

6

12.4992

H0 rejected

1.3967

5

12.9171

H0 rejected


22

Hypothesis H0
S&P 500 volatility do not Granger cause VN-Index
volatility
Frequency

Cycles

Test Statistic

Conclusion

𝛘𝟐

5% significance level

ω
𝑻=

𝟐𝝅
𝝎

(days)
1.7433

4

13.0733

H0 rejected

2.0900

3

13.1357

H0 rejected

2.4367

3

13.1633

H0 rejected

2.7833

2

13.1756

H0 rejected

3.1300

2

13.1792

H0 rejected

Source: author’s calculation
The results in Table 4.7 show that the statistical test values (χ2 distribution values) are not
the same at different frequencies. However, these differences are small, and null hypothesis
is rejected at all frequencies.
Table 4.8: Volatility Spillover from Japanese to Vietnamese Stock Markets in the
Frequency Domain
Hypothesis H0
Nikkei 225 volatility do not Granger cause VN-Index
volatility
Frequency

Cycles

Test Statistic

ω

Conclusion
10% significance level

𝑻=

𝟐𝝅
𝝎

𝛘𝟐

(days)
0.0100

628

1.4046

H0 not rejected

0.3567

18

1.3011

H0 not rejected

0.7033

9

1.278

H0 not rejected

1.0500

6

1.3974

H0 not rejected


23

Hypothesis H0
Nikkei 225 volatility do not Granger cause VN-Index
volatility
Frequency

Cycles

Test Statistic

ω

Conclusion
10% significance level

𝑻=

𝟐𝝅
𝝎

𝛘𝟐

(days)
1.3967

5

1.4597

H0 not rejected

1.7433

4

1.4885

H0 not rejected

2.0900

3

1.5031

H0 not rejected

2.4367

3

1.5109

H0 not rejected

2.7833

2

1.5148

H0 not rejected

3.1300

2

1.516

H0 not rejected

Source: author’s calculation
Table 4.8 shows the volatility spillover from Japanese stock market to Vietnamese stock
market. The results show that the statistical test values (χ2 distribution values) are not the
same at different frequencies and support the hypothesis that causality is not the same at
different frequencies (Granger and Lin 1995). However, these differences are small, and null
hypothesis is not rejected at all frequencies. Therefore, volatility spillovers from Japanese to
Vietnamese stock markets are not found in both short- and long-term.
Table 4.9 shows the volatility spillover from Korean stock market to Vietnamese stock
market. The frequency-domain approach provides more information than a traditional
Granger-causality test. Although the results from a time-domain analysis indicate that, at the
5% significance level, there is significant volatility spillover from Korean stock market to the
Vietnamese stock market, the frequency-domain approach shows that this conclusion is only
true at high frequencies. In particular, at cycles less than or equal to nine days, the null
hypothesis is rejected, but at cycles longer than nine days, the null hypothesis is not rejected.

Therefore, short-term investors (cycles less than or equal to nine days) should take note of the
KOSPI volatility to obtain more information for their investment decisions, but it is not
necessary for long-term investors (cycles longer than nine days) investors.