See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/327672794

On the Co-movements between Exchange Rate and Stock Price from Japan: A
Multivariate FIGARCH-DCC Approach

Article · January 2018

CITATION READS

1 98

3 authors:

Wajdi Moussa Nidhal Mgadmi

Institut Supérieur de Gestion de Tunis FSEG Mahdia
13 PUBLICATIONS 5 CITATIONS 18 PUBLICATIONS 52 CITATIONS

SEE PROFILE SEE PROFILE

Rym Regaïeg
Institut Supérieur de Gestion de Tunis
7 PUBLICATIONS 37 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

child developement View project

Can we reckon Bitcoin as a hedge for American Dollars? View project

All content following this page was uploaded by Wajdi Moussa on 16 September 2018.

The user has requested enhancement of the downloaded file.

Journal of Statistical and Econometric Methods, vol.7, no.4, 2018, 57-77
ISSN: 2241-0384 (print), 2241-0376 (online)
Scienpress Ltd, 2018

On the Co-movements between Exchange Rate and

Stock Price from Japan: A Multivariate
FIGARCH-DCC Approach

Moussa Wajdi1, Mgadmi Nidhal2 and Regaïeg Rym1

Abstract
This paper aims to investigate the inter-dependence of exchange rate and stock
index from Japan. For this purpose, we use a dynamic conditional correlation
(DCC) model into a multivariate Fractionally Integrated Exponential GARCH
(FIGARCH). Framework takes account long memory and time varying
correlations. Our findings reveal time-varying Co-movements evidence, a high
persistence of conditional correlation and dynamic correlations revolve around a
constant level. The findings support the idea of cross-market hedging and sharing
of common information by investors.

JEL classification numbers: C51, C58, F31, G12.

Keywords: Exchange rate, Stocks, DCC-FIAPARCH, Japan.

1
Higher Institute of Management of Tunis
2
Faculty of Law, Economics and Management of Jendouba, Tunisia

Article Info: Received: June 25, 2018. Revised : July 24, 2018
Published online : October 1 , 2018
58 On the Co-movements between Exchange Rate and Stock Price from Japan

1. Introduction
The past decades have been characterized by financial crises which were caused
such as the Asian financial crisis 1997, the Russian crisis 1998 and the Brazilian
crisis 1999, the global financial crises 2007-2009 initiated from the largest and
most influential economy, the US market, and were spreading over the other
financial markets countries worldwide. Global financial crisis resulted in sharp
declines in asset prices, stock and foreign exchange markets, and skyrocketing of
risk premiums on interbank loans. It also disrupted country's financial system and
threatened real economy with huge contractions, the recent crisis of sovereign debt
in 2011, which led to high volatility in stock markets and foreign exchange
markets. So an investor seeking to develop its heritage, regularly and without
excessive risk (or with a lower risk) is then exposed to the risk of a very sharp rise
or a very sharp drop in the value of his fortune. Therefore, it is interesting for him
and for any long or short term financial decision maker to study this growing
instability of markets, to predict later and based on this study the future volatility
for a clear and encrypted idea on its degree of exposure to different risks, that can
help them in making good decisions and strategies. In the field of modeling the
volatility, the long memory has already been introduced, in particular in the
process integrated fractionally ARFIMA and FIGARCH of Baillie Bollerslev and
Mikkelsen (1996).

Numbers of studies that attempt to examine the exchange rates effect on stock
prices, however, the findings are not uniform (Ibrahim (2000)). Some studies give
evidence of negative exchange rates effects on stock markets (Soenen and
Henningar (1988)), while others found positive effects (Aggarwal (1981)). Other
studies contribute to these results and find that the exchange rate changes have no
significant impact on the stock market (Solnik (1984)). Thus, the existing
literature provides mixed results when analyzing the relationship between stock
prices and exchange rate. For example, Yang and Doong (2004) find that stock
Moussa Wajdi et al. 59

market movements have a significant effect on future exchange rate changes for
the G7 countries over the period 1979-1999. Pan et al. (2007) use a VAR approach
to analyze the interaction between stock markets and exchange markets for seven
East Asian countries, and provide evidence of a significant bidirectional
relationship between these markets before the Asian financial crisis. Akgün and
Sayyan (2005) have examined the asymmetric response of the yields of course
fellows to Istanbul in using the models of the conditional Heteroscedasticity
asymmetric (EGARCH, GJR, APARCH, FIEGARCH, FIAPARCH) for the period
from 04 January 2000 until 25 May 2005. Their results show that the models
APARCH and FIAPARCH provide the most accurate forecasts of volatility.
Chkili et al. (2011) use a Markov-Switching EGARCH model to analyze the
dynamic relationships between exchange rates and stock returns in four emerging
countries (Singapore, Hong Kong, Mexico and Malaysia) during both normal and
turbulent periods. They provide evidence of regime dependent links and
asymmetric responses of stock market volatility to shocks affecting foreign
exchange market. Li and Giles (2015) have examined the volatility of the
exchange rate for the developed countries (United States and Japan) and 6 Asian
countries (China, India, Malaysia, Indonesia, Philippines and Thailand) by means
of the model BEKK-MGARCH. More recently, Owidi and Mugo-Waweru (2016)
have examined the behavior of the volatility performance of Kenya stock market
by the FIEGARCH. These authors found that the model FIEGARCH has the
ability to capture the asymmetry effect in taking into account the characteristics of
the long memory of the volatility.

Following the March 2011 earthquake, the yen experienced an upward trend and
even reached a new peak (Bloomberg News, June 21, 2011). Since its exporting
character knows the Japanese economy, therefore Japanese companies and
Japanese decision-makers closely supervise the yen's evolution, as the exchange
rate has a great importance in boosting exports (Hashi and Ito (2009)).
60 On the Co-movements between Exchange Rate and Stock Price from Japan

The present study provides a robust analysis of dynamic linkages among exchange
rate and stock index from Japan that goes beyond a simple analysis of correlation
breakdowns. The time-varying DCC are captured from a multivariate FIGARCH-
DCC model which takes into account long memory behavior, market information
speed. The rest of the paper is organized as follows. First part presents the data
and the econometric methodology. Second part displays and discusses the
empirical findings and their interpretation, while final part provides our
conclusion.

2. Data and Econometric methodology

2.1. Data

The data involves daily exchange rate and stock index from Japan. The data are
taken from (http//www.Federalreserves.gov) and (https://finance.yahoo.com). The
sample covers a period from 01/01/2000 until 31/12/2015, leading to a sample size
of 10950 observations. For each exchange rate and stock index in first difference.

2.2. Univariate FIGARCH Model

We assume that stationary process can be described by an AR(1) model as

follows;

𝑟𝑖,𝑡 = 𝜇 + 𝜓𝑖,𝑡−1 + 𝜀𝑖,𝑡 , 𝑡 ∈ 𝑁, 𝑤𝑖𝑡ℎ 𝜀𝑖,𝑡 = 𝑧𝑡 √ℎ𝑖,𝑡 , 𝑧~ 𝑁(0,1) (1)

Where |𝜇| ∈ [0, ∞), |𝜓| < 1, and the innovation, {𝑧𝑡 } are an independently and
identically distributed (i.i.d) process. The conditional variance ℎ𝑖,𝑡 is positive with
probability one and is measurable function of the variance-covariance matrix.

2
ℎ𝑖,𝑡 = 𝜔 + 𝛼(𝐿)𝜀𝑖;𝑡 + 𝛽(𝐿)ℎ𝑖,𝑡 (2)
Moussa Wajdi et al. 61

Where 𝜔 > 0, 𝐿 denotes the lag or backshit operator, 𝛼(𝐿) = 𝛼𝐿 + 𝛼2 𝐿2 + ⋯ +

𝛼𝑞 𝐿𝑞
In case of GARCH(1,1) model, 𝛼 ≥ 0, 𝛽 ≥ 0 and the persistence of conditional
variances is measured by the sum(𝛼 + 𝛽) . A common empirical finding is that
the sum (𝛼 + 𝛽) is quite close to one, implying that shocks are infinitely
persistent, which corresponds to an integrated GARCH (IGARCH) process of
Engle and Bollerslev (1986).
Baillie et al. (1996) developed the fractional IGARCH (FIGARCH) model, which
nests the GARCH and IGARCH models. The FIGARCH model allows for
fractional orders of integration between zero and one, and thus captures the
presence of long-memory behavior in the conditional variance. The FIGARCH
(p,d,q) model can be represented as follows;

2 2
𝜙(𝐿)(1 − 𝐿)𝑑 𝜀𝑖,𝑡 = 𝜔 + [1 − 𝛽(𝐿)](𝜀𝑖,𝑡 − ℎ𝑖,𝑡 ) (3)

Where the long-memory parameter,𝑑 , satisfies the condition 0≤ 𝑑 ≤ 1, 𝜙(𝐿)

and 𝛽(𝐿) are finite order lag polynomials with roots assumed to lie outside the
unit circle, and the fractional differencing operator (1 − 𝐿)𝑑 is defined as
(Hosking, 1981);

1 1 1
(1 − 𝐿)𝑑 = 1 − 𝑑𝐿 − 𝑑(1 − 𝑑)𝐿2 − 𝑑(1 − 𝑑)(2 − 𝑑)𝐿3 − 𝑑(1 − 𝑑)((𝑛 − 1) −
2 3! 𝑛!

𝑑)𝐿𝑛 (4)

The conditional variance process or the ARCH (∞) representation of the

FIGARCH model is given by;

ℎ𝑖𝑡 = 𝜔 + 𝛽(𝐿)ℎ𝑖,𝑡 + [1 − 𝛽(𝐿)]𝜀𝑖𝑡2 − 𝜙(𝐿)(1 − 𝐿)𝑑 𝜀𝑖,𝑡

2
= 𝜔[1 − 𝛽(𝐿)]−1 +
2
𝜆(𝐿)𝜀𝑖,𝑡 , (5)
62 On the Co-movements between Exchange Rate and Stock Price from Japan

The FIGARCH model provides greater flexibility for modelling the conditional
variance, because it accommodates the covariance stationary GARCH model
when 𝑑 = 0, and the IGARCH model when 𝑑 = 1, as special cases. For the
FIGARCH model, the persistence of shocks to the conditional variance, or the
degree of long-memory, is measured by the fractional differencing parameter 𝑑.
Thus, the attraction of the FIGARCH model is that, for0 < 𝑑 <, it is sufficiently
flexible to allow for an intermediate range of persistence (Baillie et al., 1996).

2.3. Bivariate model FIEGARCH with dynamic conditional correlations

To evaluate volatility spillovers, we applied a bivariate FIGARCH model to

JPY/USD and NIKKEI225 index in first difference. We decided to model the
structure of conditional correlations using the DCC approach of Engle (2002).
This allows us to not only investigate the time-varying correlations between two
indexes, but also to ensure the positive definiteness of the variance-covariance
matrix (𝐻𝑡 )under simple conditions imposed on specific parameters. The
parameterization of a FIGARCH-DCC model allows for directly inferring the
time-varying correlations between JPY/USD and NIKKEI225 index in first
difference as well as for dealing with a relatively large number of variables in the
system without having a numerical convergence problem at the estimation stage.
In the multivariate case which we use, the variance-covariance matrix of residuals
is defined as follows;
𝐻𝑡 = 𝐷𝑡 𝑅𝑡 𝐷𝑡 (6)

where 𝐷𝑡 is a (2 × 2)diagonal matrix of the time-varying conditional standard

deviation of the residuals, obtained by taking the square root of the conditional
variance, modelled by univariate AR( 1)-GARCH(1,1) and AR( 1)-FI
GARCH(1,1) models, respectively. 𝑅𝑡 is a matrix of time-varying conditional
correlations, which is given by;
Moussa Wajdi et al. 63

−1⁄ −1⁄
2 2
𝑅𝑡 = (𝑑𝑖𝑎𝑔(𝑄𝑡 )) 𝑄𝑡 (𝑑𝑖𝑎𝑔(𝑄𝑡 )) (7)

1 0 1 0
⁄√𝑞 ⁄√𝑞
11 11
𝑞11 𝑞12
That is, 𝑅𝑡 = 1
[𝑞 𝑞22 ] 1
0 ⁄√𝑞
21 0 ⁄√𝑞
22 22
[ ] [ ]

The covariance matrix 𝑄𝑡 = [𝑞𝑖𝑗,𝑡 ] of the standardized residual vector

′
𝑢𝑡 = (𝑢1,𝑡 , 𝑢2,𝑡 , … , 𝑢𝑛,𝑡 ) , 𝑢𝑡 = 𝜀𝑖,𝑡 /√ℎ𝑖𝑖,𝑡 is denoted as;

𝑄𝑡 = (1 − 𝛼𝑑𝑐𝑐 − 𝛽𝑑𝑐𝑐 )𝑄̅ + 𝛼𝑑𝑐𝑐 (𝑢𝑡−1 𝑢𝑡−1

′ )
+ 𝛽𝑑𝑐𝑐 𝑄𝑡−1 , (8)

With 𝛼𝑑𝑐𝑐 > 0, 𝛽𝑑𝑐𝑐 > 0 and 𝛼𝑑𝑐𝑐 + 𝛽𝑑𝑐𝑐 < 1. 𝑄̅𝑡 = {𝑞̅𝑖𝑗,𝑡 } denotes the
unconditional covariance matrix of 𝜀𝑡 . The coefficients 𝛼𝑑𝑐𝑐 𝑎𝑛𝑑 𝛽𝑑𝑐𝑐 are the
estimated parameters depicting the conditional correlation process. The dynamic
correlation coefficient in a bivariate case can be expressed as;

(1−𝛼𝑑𝑐𝑐 −𝛽𝑑𝑐𝑐 )𝑞̅𝑖𝑗+𝛼𝑑𝑐𝑐 𝑢𝑖,𝑡−1 𝑢𝑗,𝑡−1 +𝑞𝑖𝑗,𝑡−1

𝜌𝑖𝑗,𝑡 = (9)
2
√[(1−𝛼𝑑𝑐𝑐 −𝛽𝑑𝑐𝑐 )𝑞̅𝑖𝑖 +𝛼𝑑𝑐𝑐 𝑢𝑖,𝑡−1 2
+𝛽𝑑𝑐𝑐 𝑞𝑖𝑖,𝑡−1 ][(1−𝛼𝑑𝑐𝑐 −𝛽𝑑𝑐𝑐 )𝑞̅𝑗𝑗 +𝛼𝑑𝑐𝑐 𝑢𝑗,𝑡−1 +𝛽𝑑𝑐𝑐 𝑞𝑗𝑗,𝑡−1 ]

The significance of 𝛼𝑑𝑐𝑐 𝑎𝑛𝑑 𝛽𝑑𝑐𝑐 implies that the estimators obtained from the
DCC-FIGACH models were dynamic and varied with time. 𝛼𝑑𝑐𝑐 indicates short-
run volatility impact, implying the persistency of the standardized residuals from
the previous period. 𝛽𝑑𝑐𝑐 measures the lingering effect of shock impact on
conditional correlations, indicating the persistency of the conditional correlation
process. 𝜌𝑖𝑗,𝑡 indicates the direction and strength of correlation. If the estimated 𝜌𝑖𝑗,𝑡
is positive, then the correlation between stationary series is moving in the same
direction and vice vers3.

3
See Engle (2002) for further details on the estimation of time-varying conditional correlations.
 64 On the Co-movements between Exchange Rate and Stock Price from Japan

We estimate the DCC model using the quasi-maximum likelihood estimation

method proposed by Bollerslev and Wooldridge (1992) in which the log-
likelihood can be expressed as;

1
𝐿 = − 2 ∑𝑇𝑡=1[𝑘𝑙𝑜𝑔(2𝜋) + 2𝑙𝑜𝑔|𝐷𝑡 | + 𝑙𝑜𝑔|𝑅𝑡 | + 𝜀𝑡′ 𝑅𝑡−1 𝜀𝑡 ] (10)

The DCC model’s design allows for the two-stage estimation procedures of the
conditional covariance matrix 𝐻𝑡 , in the first stage, we fit the univarite GARCH-
type models for each JPY/USD and NIKKEI225 index in first difference; then,
estimates of ℎ𝑖𝑖𝑡 , are obtained. In the second stage, we transform the first
difference series using their estimated standard deviation, which is a result of the
first stage. Then this information is used to estimate the parameters of the
conditional correlation.

3. Empirical Results
3.1. Preliminary analyses and testes
In this section, I report the results of descriptive statistics in table 1, unit root tests
in table 2, Serial correlation and LM-ARCH test in table 3, Long Memory tests in
table 4 and the evolution of exchange rate and stock index in figure 1.

Table I : Descriptive statistics

Mean Maximum Minimum Std.Deviation Skewness Kurtosis Jarque-Bera
Log(JPY/USD) 4.650721 4.903570 4.327042 0.144385 -0.669031 2.397668 524.3077

Log(NIKKEI225) 9.416817 9.945974 8.861489 0.262888 0.187731 1.860414 350.5495

Summary statistics of stock price and exchange rate are displayed in Table 1.
From this table, NIKKEI225 is the most volatile, as measured by the standard
deviation of 26.2888%, while JPY/USD is the most volatile with a standard
 Moussa Wajdi et al. 65

deviation of 14.4385%. Besides, we observe that NIKKEI225 has the highest level
of kurtosis, indicating that extreme changes tend to occur more frequently for the
stock price. In addition, the stock price and exchange rate exhibit high values of
excess kurtosis. To accommodate the existence of “fat tails”, we assume t-student
distributed innovations. Furthermore, the Jarque-Bera statistic rejects normality at
the 1% level for all stock prices and exchange rate. The Ljung-Box test for
correlating series reject the null hypothesis of autocorrelation. This equates that
conditional volatility of the returns series contains long memory.

In Table II, we verify the stationary of the Log(JPY/USD) and Log(NIKKEI225)

index by the tests of Dickey-Fuller (1979-1981) in level and in first difference.
For this, we use the sequential procedure to determine the optimal number of lags
of this index on daily frequencies. Results reported in table 2 indicate that both
series have a unit root since t-statistics are higher in level than the tabulated value
of Mackinnon (1996). La présence d’une racine unitaire en niveau pour l’indice
JPY/USD est détectée à partir d’un modèle sans constant et sans tendance linéaire.
Par contre, l’indice NIKKEI225 contient une racine unitaire en niveau à partir
d’une spécification avec constante et sans tendance linéaire. This unit root
disappeared after a single differencing. Hence, the JPY/USD and NIKKEI225
index are integrated of order 1. The unit root for the JPY/USD and NIKKEI225
index are detected by the Dickey-Fuller test (1979) while the Dickey-Fuller-
Augmented test (1981) is used to identify the existence of a unit root of the
Log(JPY/USD) and Log(NIKKEI225) index.

Table II : ADF Test Statistic

Level First difference Integration
Lags Models
T-Statistic Critical values T-Statistic Critical values order

Log(JPY/USD) 2 M1 0.053 -0.193 -38.37 -1,9409 I(1)

Log(NIKKEI225) 2 M2 0.066 -0.182 -37.612 -1,9409 I(1)

66 On the Co-movements between Exchange Rate and Stock Price from Japan

Furthermore, in Table 3 the Jarque-Bera statistic rejects normality at the 1% first

difference for all exchange rates. Finally, we find that there is a problem of
heteroscedasticity in the JPY/USD variance and NIKKEI225 index variance.
Since the variance of residuals is heteroscedastic. In order to detect long-memory
process in the data, we use the log-periodogram regression (GPH) test of Geweke
and Porter-Hudak (1983) on two proxies of volatility, namely squared returns and
absolute returns.

Table 3: Serial correlation and LM-ARCH Test

Serial Correlation LM-ARCH

LB(20) LB2(20) ARCH(10)
33.5368 337.717 13.661
 Log (JPY/USD)
[0.029434]** [0.0000]*** [0.0000]***
41.4636 3729.67 159.20
 Log (NIKKEI225)
[0.00324]*** [0.0000]** [0.0000]***
Notes: The superscripts ***, ** and * denote the statistical significance at 1%, 5% and 10% levels,
respectively

In table 3 the Ljung-Box statistic was used to check for the presence of serial
correlation in the standardized and squared standardized residuals up to the 20th
order. ***denotes the rejections of the null hypothesis of no serial correlation at
the 1% significance level. Thus, the series in first difference seemed to follow
ARCH-type dependencies.The LM-ARCH test for conditional heteroscedasticity
represents statistical significance at the 1% level. This further encourages the use
of ARCH-type models in order to describe this feature of the data.
Moussa Wajdi et al. 67

Table 4 : Long Memory tests

Hurst-Mandelbrot R/S test Squared Absolute

 Log (JPY/USD) 6.551*** 12.108***

 Log NIKKEI225) 6.756*** 12.985***

Squared Absolute
Lo’s modified R/S test 𝑞=1 𝑞=5 𝑞=1 𝑞=5

 Log (JPY/USD) 5. 427*** 3.954*** 10.662*** 7.725***

 Log (NIKKEI225) 5. 443*** 4.457*** 10.751*** 8.021***

Squared Absolute

GSP test 𝑚 = 𝑇⁄4 𝑚 = 𝑇⁄16 𝑚 = 𝑇⁄4 𝑚 = 𝑇⁄16

 Log (JPY/USD) 0.222*** 0.417*** 0.326*** 0.445***

 Log (NIKKEI225) 0.258*** 0.444*** 0.395*** 0.478***

Notes: The superscripts ***, ** and * denote the statistical significance at 1%, 5% and 10% levels,
respectively.

The analysis in table 4 was done via the Hurst-Mandelbrot R/S, Lo’s modified
R/S and GSP tests conducted on absolute and squared of the JPY/USD and
NIKKEI225 index in first difference, as a proxy of variance. The results suggest
that, at very high probability levels, the null hypothesis of no long-range memory
should be rejected for all asset series. This is an indication of the persistence of
volatility shocks for periods of time longer than the usual exponential decay for
which standard GARCH models are appropriate. That suggests that FIGARCH
approach should successfully measure the long memory presence in the volatility
of the observed asset series.

Below, figure 1 illustrates the evolution of exchange rate and stock index during
the period from January 1, 2000 until December 31, 2015. The figure shows
significant variations in the levels during the turmoil, especially at the time of
Lehman Brothers failure (September 15, 2008). Specifically, when the global
financial crisis triggered, there was a decline for all prices. Moreover, Fig. 1 plots
the evolution of exchange rate behavior and stock price behavior over time.
68 On the Co-movements between Exchange Rate and Stock Price from Japan

JPY/USD

130

120

110

100

90

80

2000 2002 2004 2006 2008 2010 2012 2014 2016

NIKKEI225
20000

18000

16000

14000

12000

10000

8000

2000 2002 2004 2006 2008 2010 2012 2014 2016

Figure: 1.1. Exchange rate and stock index behavior over time.

0.03 DL(JPY/USD)

0.02

0.01

0.00

-0.01

-0.02

-0.03

-0.04

-0.05

2000 2002 2004 2006 2008 2010 2012 2014 2016

DL(NIKKEI225)

0.10

0.05

0.00

-0.05

-0.10

2000 2002 2004 2006 2008 2010 2012 2014 2016

Figure.1.2: Exchange rate and stock index in first difference behavior over time.
Moussa Wajdi et al. 69

The figure shows that all stock indexes and exchange rate trembled since 2008
with different intensity during the global financial and European sovereign debt
crises. Moreover, the plot shows a clustering of larger return volatility around and
after 2008. This means that stock price is characterized by volatility clustering,
i.e., large (small) volatility tends to be followed by large (small) volatility,
revealing the presence of heteroscedasticity. This market phenomenon has been
widely recognized and successfully captured by ARCH/GARCH family models to
adequately describe stock market returns dynamics. This is important because the
econometric model will be based on the interdependence of the stock markets in
the form of second moments by modeling the time varying variance-covariance
matrix for the sample.

3.2. The univariate FIGARCH estimates

Table 5 reports the Maximum Likelihood Estimates (MLE) for the student-t-
FIGARCH (1,d,1) model. LB2(20) indicate the Ljung-Box tests for serial
correlation in the squared standardized residuals. Student-df denotes the t-student
degrees of freedom parameter***,**and * denote statistical significance at 1%,
5% and 10% levels, respectively. ARCH (10) means the test of autoregressive
conditional hetroskedasticity up to lag 10.The null hypothesis of ARCH test is no
ARCH effects up to lag k. Diagnostic tests LB2(20) and ARCH (10) showed that
the FIGARCH model with Student’s t-distribution was well specified because
standardized residuals are not subject to either serial correlation or ARCH effects.
This result confirms our preliminary analysis and, subsequently, by the choice of
the t-student as an appropriate distribution. The estimation results for the student-t-
FIGARCH model suggest that the FIGARCH model captures the long-memory
property in the volatility processes of the JPY/USD and NIKKEI225 index in first
difference. In fact, the long-memory parameters (d-Figarch) are significant at the
l% level, suggesting that all index in first difference volatility processes were
persistent over time. Overall, the FIGARCH model most accurately represents the
70 On the Co-movements between Exchange Rate and Stock Price from Japan

long-memory property in the conditional variance of the JPY/USD and

NIKKEI225 index in first difference. The presence of long-memory in volatility
implies dependencies between distant observations over time, which can be used
to predict future volatility values. The ARCH and GARCH parameters are
statistically significant and not negative at the 1% which justifying the
appropriateness of the FIGARCH model and implying that the volatility is highly
persistent in all JPY/USD and NIKKEI225 index in first difference4.

Table 5 : Estimation results from the univariate FIGARCH (1, d, 1) model

 Log (NIKKEI225)  Log (JPY/USD)

Estimates Coefficients t-prob Coefficients t-prob
0.0002 0.0050 0.0001 0.0015
CST(𝑚)
AR(1) -0.0295 0.0006 -0.0216 0.0155

CST(𝜐) 0.1105 0.0004 1.4669 0.0021

d-Figarch 0.6619 0.0000 0.6821 0.0000

ARCH 0.2871 0.0000 0.3689 0.0000

GARCH 0.8694 0.0000 0.9286 0.0000

Student-df 2.3811 0.0000 2.3346 0.0000

Log-likelihood 18540.624 23370.182

Diagnostic tests

ARCH (10) 0.3012 0.9820 1.5084 0.1305

LB2(20) 286.634 0.3407 128.700 0.11007

4
The traditional efficient market hypothesis of Fama (1970) suggested that stock returns showed a
random walk, making it impossible to make predictions from past patterns. Strictly speaking, it is
to some extent inappropriate that the definition of the efficient market hypothesis is applied to
stock price volatility. However, the long-memory property in stock returns or volatility has
improved the definition of market efficiency, where predictable components occur in asset pricing.
Moussa Wajdi et al. 71

3.3. The bivariate FIGARCH (1, d, 1)-DCC estimates

This subsection considers volatility spillover effects between the JPY/USD and
NIKKEI225 index in first difference using the bivariate FIGARCH (1,d,1)-DCC
models with Student’s t-distribution. Table 5 presents the estimation results of the
bivariate FIGARCH-DCC model between the JPY/USD and NIKKEI225 index in
first difference.

Table 6: Estimation results from the DCC-FIGARCH (1, d, 1) model

 Log (NIKKEI225)  Log (JPY/USD)

Estimates Coefficients t-prob Coefficients t-prob
CST(𝑚) 0.0003 0.0151 0.0005 0.0386

AR(1) -0.0249 0.0377 -0.0117 0.0451

CST(𝜐) 0.0260 0.0411

0.9418 0.0006
d-Figarch 0.5742 0.0000 0.5450 0.0000

ARCH 0.2616 0.0000 0.5374 0.0000

GARCH 0.7744 0.0000

0.7895 0.0000
Student-df 2.3964 0.0000 2.3745 0.0000

Log-likelihood 17924.318 22554.245

Dynamic conditional correlation

Coefficients t-prob

𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝐶𝑂𝑅𝑖𝑗 0.5211 0.0397

𝛼𝑑𝑐𝑐 0.0055 0.0307

𝛽𝑑𝑐𝑐 0.9920 0.0000

Student-df 2.6008 0.0000

2 Log-likelihood 42196.035

Diagnostic tests
72 On the Co-movements between Exchange Rate and Stock Price from Japan

Coefficients t-prob

𝐻𝑜𝑠𝑘𝑖𝑛𝑔(20) 164.055 0.0000

𝐻𝑜𝑠𝑘𝑖𝑛𝑔2(20) 26.5222 1.0000

𝐿𝑖 - 𝑀𝑐𝐿𝑒𝑜𝑑(20) 167.302 0.0000

𝐿𝑖 – 𝑀𝑐𝐿𝑒𝑜𝑑2(20) 26.5994 1.0000

The ARCH and GARCH term is positive and significant, implying volatility
persistence or volatility clustering in the JPY/USD and NIKKEI225 index in first
difference. The fractional integrated coefficient (d) is highly significant,
indicating a high level of persistence in conditional variances. The JPY/USD and
NIKKEI225 index addressed the higher persistence regardless of time
intervals.The ARCH effect 𝛼𝑑𝑐𝑐 is positive and significant, underlying the
importance of shocks between the JPY/USD and NIKKEI225 index. For the
GARCH effects 𝛽𝑑𝑐𝑐 , the parameters are significant and very close to one for all
JPY/USD and NIKKEI225 index in first difference, confirming the higher
persistence of volatility between the JPY/USD and NIKKEI225 index. The sum
of these parameters (𝛼𝑑𝑐𝑐 +𝛽𝑑𝑐𝑐 ) in each model is less than unity and this shows
that conditional pair -wise correlations are mean-reverting. In sum, the
significance of the parameters, especially (d), 𝛼𝑑𝑐𝑐 and 𝛽𝑑𝑐𝑐 indicates the existence
of the dynamic and time-varying long-memory features.Similarly, the average
conditional correlation 𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝐶𝑂𝑅𝑖𝑗 were all positive and significant at the 5%

level, indicating bi-directional causality between JPY/USD and NIKKEI225

index. According to the diagnostic tests, Kroner and Ng (1998) have confirmed
the fact that only few diagnostic tests are kept to the multivariate GARCH-class
models compared to the diverse diagnostic tests devoted to univariate
counterparts. Furthermore, Bauwens et al. (2006) have noted that the existing
literature on multivariate diagnostics is sparse compared to the univariate case. In
our study, we refer to the most broadly used diagnostic tests, namely the Hosking's
Moussa Wajdi et al. 73

and Li and McLeod's Multivariate Portmanteau statistics on both standardized and

squared standardized residuals. According to Hosking (1980), Li and McLeod
(1981) and McLeod and Li (1983) autocorrelation test results reported in Table 5
the multivariate diagnostic tests allow accepting the null hypothesis of no serial
correlation on squared standardized residuals and thus there is no evidence of
statistical misspecification.

CORR_DL(JPY/USD)_DL(NIKKEI225)

0.4

0.3

0.2

0.1

0.0

-0.1

2000 2002 2004 2006 2008 2010 2012 2014 2016

Figure 2: The DCC behavior over time.

Figure 2 illustrates the evolution of the estimated dynamic conditional correlations

dynamics among exchange rate and stock index in first difference . Compared to
the pre-crises period, the estimated DCC show a decline during the post-crises
period. Such evidence is in contrast with the findings of previous research on
exchange rate and stock price, which show increases in correlations during periods
of financial turmoil (Kenourgios & al. (2011), Dimitriou & al.(2013), Dimitriou
and Kenourgios (2013)). Nevertheless, the different path of the estimated DCC
displays fluctuations for all pairs of exchange rate and stock price during the
phases of the global financial and European sovereign debt crises, suggesting that
the assumption of constant correlation is not appropriate. The above findings
motivate a more extensive analysis of DCC, in order to capture contagion
dynamics during different phases of the two crises.
74 On the Co-movements between Exchange Rate and Stock Price from Japan

4. Conclusion
In this study, we tried to analyse the presence of long memory in volatility of the
JPY/USD and NIKKEI225 index in first difference. We employ a univariate AR-
FIGARCH and a multivariate FIGARCH-DCC model, during the period from
01/01/ 2000 to 31/12/ 2015, for the purpose of the analysis. In the first part, we
tested for long memory in the absolute first difference series, and the squared first
difference series, the latter two serving as a measure of conditional volatility.
Next, we analysed the presence of long memory in conditional volatility by
employing a univariate AR-FIGARCH and a multivariate AR-DCC-FIGARCH
model with Student’s t-distribution. Our results document strong evidence of time-
varying comovement, a high persistence of the conditional correlation (the
volatility displays a highly persistent manner) and the dynamic correlations
revolve around a constant level and the dynamic process appears to be mean
reverting. In addition, we was found that the JPY/USD and NIKKEI225 index in
first difference has an underlying fractal structure, as the presence of long memory
was confirmed by the analysis. The presence of long memory can have multiple
implications. From a theoretical point of view the presence of self-similar patterns
refutes the notion of the efficient market hypothesis in its weak form. Further, the
presence of long memory in volatility series indicates that it would be better to
develop and employ long-memory models as opposed to the traditional GARCH
models to forecast market returns and volatility. Kumar, D. (2013) examines
asymmetry and long memory properties in the volatility of Portugal, Italy, Greece,
Ireland, and Spain, over the period 2003 to 2011 using the AR-GARCH,
IGARCH, FIGARCH, FIGARCH, EGARCH and FIEGARCH for comparative
purpose. The results show that the AR-FIGARCH model specification is better
able to capture the long memory property of conditional volatility than the
conventional GARCH and IGARCH models. From an investors’ perspective, the
presence of patterns in the market in volatility structure implies that it will be
Moussa Wajdi et al. 75

possible to predict trends in volatility. Hence the possibility of gaining extra-

normal profit and diversifying risk exists.

References

[1] Aggarwal, R, Exchange Rate and Stock Prices: A study of U.S. Capital
Market under Floating Exchange Rates, Akron Business and Economic
Review, 12, (1981), 7-12.
[2] Akgül, I., Sayyan, H., Forecasting volatility in ISE-30 stock returns with
asymmetric conditional heteroskedasticity models, Asian Journal of Finance
& Accounting, 8, (2005), 48-61.
[3] Baillie, R.T., T. Bollerslev and H.O. Mikkelsen, Fractionally integrated
generalized autoregressive conditional heteroskedasticity, Journal of
Econometrics, 74, (1996), 3-30.
[4] Bollerslev, T., and Mikkelsen, H.O, Modeling and pricing long memory in
stock market volatility, Journal of Econometrics, 73, (1996), 151-184.
[5] Chkili, W., Aloui, C., Nguyen, K., Asymmetric effects and long memory in
dynamic volatility relationships between stock returns and exchange rates,
Journal of International Financial Markets, 22, (2011), 738-757.
[6] Conrad, C., Karanasos, M., and Zeng, N, Multivariate fractionally integrated
APARCH modeling of stock market volatility : A multi-country study,
Journal of Empirical Finance, 18, (2011), 147-159.
[7] Dickey, D.A., Fuller, W, Distribution of the estimators for autoregressive
time series with a unit root, Journal of the American Statistical Association,
74, (1979), 427-431.
[8] Dickey, D.A, Fuller, W, Likelihood ratio statistics for autoregressive time
series with a unit root, Econometrica, 49, (1981), 1057-72.
[9] Dimitriou, D., and Kenourgios, D., Financial crises and dynamic linkages
among international currencies, Journal of International Financial Markets,
Institutions and Money, 26, (2013), 319-332.
76 On the Co-movements between Exchange Rate and Stock Price from Japan

[10] Dimitriou, D., Kenourgios, D., and Simos, T, Global financial crisis and
emerging stock market contagion: a multivariate FIAPARCH-DCC approach,
International Review of Financial Analysis 30, (2013), 46–56.
[11] Engle, R., Dynamic conditional correlation: a simple class of multivariate
generalized autoregressive conditional heteroskedasticity models, Journal of
Business and Economic Statistics, 20, (2002), 339-350.
[12] Fama, E. F., Efficient Capital Markets: A Review of Theory and Empirical
Work Journal of Finance, 25, (1970), 383–417.
[13] Geweke, J., and Porter-Hudak, S, The estimation and application of long-
memory time series models, Journal of Time Series Analysis, 4, (1983), 221–
238.
[14] Hosking, J.R.M, The multivariate portmanteau statistic, Journal of American
Statistical Association, 75, (1980), 602-608.
[15] Ibrahim, M., Co-integration and Granger Causality Tests of Stock Price and
Exchange Rates Interaction in Malaysia. ASEAN, Economics Bulletin, Vol
17, n°1, (2000).
[16] Kenourgios, D., and Padhi, P, Emerging markets and financial crises :
Regional, global or isolated shocks?, Journal of Multinational Financial
Management, 22, (2012), 24-38.
[17] Kumar, D, Asymmetric long memory volatility in the PIIGS economies,
Review of Accounting and Finance, 12, (2013), 23-43.
[18] Li, K., McLeod, I., Distribution of the residual autocorrelations in
multivariate ARMA time series models, Journal of the Royal Statistical
Society, series B (Methodological), 43, (1981), 231-239.
[19] Li, Y., Giles, E., Modelling volatility spillover effects between developed
stock markets and Asian emrging stock markets, International Journal of
Finance and Economics, 20, (2015) ,155-177.
[20] Mackinnon, J. G, Numerical distribution functions for unit root and
cointegration tests, Journal of Applied Econometrics, 11, (1996), 601-618.
 Moussa Wajdi et al. 77

[21] McLeod, A.I., and Li, W.K, Diagnostic checking ARMA time series models
using squared residual autocorrelations, Journal of Time Series Analysis, 4,
(1983), 269-273.
[22] Owidi, O., Mugo-Waweru, F., Analysis of Asymmetric and Persistence in
Stock Return Volatility in the Nairobi Securities Exchange Market Phases,
Journal of Finance and Economics, 4, (2016), 63-73.
[23] Pan, S., Fok, W., Liu, A., Dynamic linkages between exchange rates and
stock prices: evidence from East Asian markets, International Review of
Economics and Finance, 16, (2007), 503–520.
[24] Soenen, A., Hennigar, S., An analysis of exchange rates and stock prices: The
US experience between 1980 and 1986 , Akron Business and Economic
Review, (1988), 7-16.
[25] Solnik, H., Stock and Money variables: The international evidence. Financial
Analyse Journal, (1984), 69-73.

View publication stats