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

net/publication/235685933

Measuring Exchange Rate Co-movement: Empirical Investigation in between

Indian Rupees, Japan Yen, Sterling Pound, US Dollar and EURO

Conference Paper · January 2011

CITATIONS READS

0 343

1 author:

Ushan Peiris
Sabaragamuwa University of Sri Lanka
23 PUBLICATIONS 24 CITATIONS

SEE PROFILE

All content following this page was uploaded by Ushan Peiris on 31 May 2014.

The user has requested enhancement of the downloaded file.

 An Empirical Investigation on Exchange Rate Co-Movement in Between
Indian Rupees, Japan Yen, Sterling Pound, US Dollar and EURO.

Abstract
Economic assimilation in between world economies has been advancing rapidly, driven by
growing intra-regional trade, increasing investment and financial integration. Recent discussions
on regional and bilateral free-trade agreements and on emerging financial cooperation
arrangements make this process likely to deepen over time resulting the business cycles across
economies in the world more interrelated. The behavior of Exchange Rate (ER) movements
among economies does reflect how deepen the assimilation is. Thus, this study attempts to
investigate the extent to which international ERs move together. The ER co-movement is
measured between Indian Rupees, Japan Yen, Sterling Pound, US Dollar and EURO by taking
Sri Lankan Rupees as the base currency. Econometric models of Autoregressive Conditional
Heteroskedasticity (ARCH), Generalized ARCH (GARCH), Johansen Cointegration Test,
Vector Autoregression (VAR) and VAR Variance Decompositions are employed. Monthly ERs
over the period from 2007 to 2010 (48 months) were used for the investigation. The empirical
results indicate that ER co-movement is significant since most of the considered ERs depict more
than 46% of their forecast error variance is explained by the other ERs under study while Japan
Yen depicts only 37%. There is a high degree of co-movement between ER against Sterling
Pound and with that of Japan Yen, US Dollar, Indian Rupees and EURO, stood in declining
order of ER co-movement respectively. Japan Yen showed a negative response for all the chocks
given to other ERs while the response of Sterling Pound is persistent to the shocks of Indian
Rupees.

Key Words: Exchange Rate, Co-movement, GARCH and Error Variance

1. Introduction
Economic assimilation in between world economies has been advancing rapidly and
concomitantly in recent past, driven by growing intra-regional trade, increasing investment and
financial integration. Recent discussions on regional and bilateral free-trade agreements and on
emerging financial cooperation arrangements make this process likely to deepen over time
reflecting, increasing trade and capital flows, business cycles across countries in the world are
1
becoming more correlated. A significant indicator of this assimilation of economies is the
exchange rate (ER) volatility. Thus, the behavior of ER movements among economies does
reflect how deepen the assimilation is.
Studies that examined the exchange rate volatility comovement were initiated by Engle and
Hamilton (1990) where they have formulated two possible hypotheses named the „heat waves‟
and the „meteor shower‟. The heat waves refers to exchange rate volatility in one particular
market having only country specific effects while, the meteor shower to volatility being
transmitted to other countries. The authors found supporting evidence of both effects.
Antonakakis (2008) examined the volatility co-movement and spillover among major currencies (Euro,
Franc, Pound and the Yen) before and after the introduction of Euro and documented the existence of
significant volatility spillovers and co-movements across exchange rates, but have declined since
the introduction of the euro. These findings suggest that the launch of the euro itself may have
brought about greater stability in the global financial markets. Moreover, the results show that
euro is the dominant currency in volatility transmission, as its volatility significantly affects the
volatility expectations of other ERs.
On this basis, studies were carried out by Bollerslev (1990), Kearney and Patton (2000), Speight
and McMillan (2001), Melvin and Melvin (2003), Black and McMillan (2004) and by Calvet et
al., (2006) among others. The main feature of these studies is the application of GARCH
modeling to assess volatility dependencies across currencies. The main conclusion among these
studies is the support of volatility comovement between exchange rates. Some of the reasons of
such volatility dependencies across currencies might be attributed to news announcements about
central bank interventions and fundamentals or just to speculation events.
A substantial number of studies have been conducted on ER volatility co-movement among the
major currencies of the world. However, studies on ER volatility co-movements among these
major currencies and currencies of developing countries are still vacuum in the finance literature.
Especially for developing countries like Sri Lanka it is very important to be aware of how and
what extent the other economies influence the behavior of ERs. This phenomenon encouraged
the researcher to investigate rigorously the ER volatility co-movement among the Indian Rupees,
Japan Yen, Sterling Pound, US Dollar and EURO by considering Sri Lanka Rupees as the base
currency.

2
The rest of the sections of this study are structured as follows. Section 2 explains the
methodology adopted in this study. Section 3 outlines the data analysis and discussion. Finally,
the conclusion of the study is explained by Section 5.

2. Methodology
This important section primarily deals with the description of data of this study. This section also
gives through explanation about statistical models for ER volatility. The methodology adopted
here is mostly similar to the prominent studies discussed in the early section. However, some
diverse methodological procedures will apply when ever appropriate for the current study.

2.1 Data collection

This study attempts to investigate the volatility co-movement among the average ERs of major
currencies of Indian Rupees, Japanese Yen, Sterling Pound, US Dollars and Euro by taking Sri
Lanka Rupees as the base currency1. That indicates the ERs are calculated as number of Sri
Lanka Rupees per unit of foreign currency. Monthly average ERs over the period from 2007 to
2010 (48 months) were obtained from annual reports of the Central Bank, Sri Lanka.

2.2 Developing the Mean Equation

The Autoregressive Moving Average (ARMA) models are considered when developing the
mean equation. In this study the researcher considered only up to the first lag both in AR and
MA components in the ARMA model. Hence, three models of ARMA (1, 0), ARMA (0, 1) and
ARMA (1, 1) were tested for the best fitted mean model. Log likelihood estimate, Akaike Info
Criterion (AIC) and Schwarz Criterion (SIC) are used to select the most fitted model out of the
three models considered above.

2.3 Developing the Volatility Model

As Tsay (2005) suggests, the fitted mean equation should be tested for Autoregressive
Conditional Heteroskedasticity (ARCH) effects before developing a volatility model. The test
applied for conditional heteroscedasticity effects in this study is the Lagrange multiplier test
(ARCH LM) of Engle (1982). If ARCH effects are statistically significant Generalized ARCH

1
According to the Central Bank annual reports of Sri Lanka, Indian Rupees (InRs/SLRs), Japanese Yen
(Yen/SLRs), Sterling Pound (StPound/SLRs), US Dollars (Dollars/SLRs ) and Euro(Euro/SLRs) are been
considered as the Major Currencies.

3
(GARCH) model is fitted for the data series. Hence, GARCH (1,1) model is fitted for the
purpose of this study. The GARCH (1,1) can be characterized as follows:
 t2    1 t21  1 t21 (1)

 t2 is a function of lagged values of  t2 ,  .  1 and 1 are nonnegative constants.  1 is the

coefficient of ARCH term and  1 is the coefficient of GARCH term.

2.4 Johansen Cointegration Test (JCT) and Vector Autoregression (VAR)

This study aims to understand how shocks and volatility in one ER is transmitted to other ER in
objectively recognizable fashion. This requires looking at the Cointegration and the extent to
which multi-lateral interaction exists between these ERs. For this purpose, JCT and the VAR
model is applied.
The mathematical representation of a VAR is:
m
Yt  c   AsYt s   t (2)
s 1

where Yt is a 5 x 1 column vector of rates of return of five ERs, C is the deterministic

component comprised of a constant, A(s) are respectively , 5 x 1 and 5 x 5 matrices of

coefficients, m is the lag length, and  t is the 5 x 1 innovation vector. By construction,  t is

uncorrelated with all the past Yt . A shock to the i-th variable not only directly affects the i-th
variable but is also transmitted to all of the other endogenous variables through the dynamic (lag)
structure of the VAR.

3. Results and Discussion

This chapter describes the data analysis and discussion of the results thereof. GARCH, JCT and
VAR are the main econometric models used to derive at the results of this study.

3.1 ARMA models and ARCH LM test

The following table 3.1 demonstrates the test results for the ARMA models and the ARCH LM
test thereof.

4
Table 3.1 ARMA Model fitting & ARCH LM test
Sector Mean Model ARCH LM test ARCH LM test
F-Statistic P-value

Indian RS ARMA(1,1) 0.118292* 0.032533

Japanese Yen ARMA(1,1) 0.307528* 0.0282009
Sterling Pound ARMA(1,1) 0.325983* 0.0070939
US Dollars ARMA(1,1) 1.006187* 0.021303
EURO ARMA(1,1) 1.552774* 0.019320
Note: * denotes statistical significance at 5% level.

In this study three models of ARMA (1,0), ARMA (0,1) and ARMA (1,1) were tested for the
best fitted mean model. ARMA (1,1) model depicted the highest log likelihood estimate and the
lowest AIC and SIC among the three models considered for all the ERs. Hence, ARMA (1,1)
model is fitted as the mean model and the residuals were tested for the ARCH effects. ARCH LM
test results indicate that ARCH effects are significant (P-value<0.05) at 5% confidence level.
Hence, further modeling of GARCH can be applied for all the ERs.

3.2 Volatility models

When deriving the volatility of these ERs GARCH (1, 1) model is employed. The following
Table 3.2 depicts the results of the volatility models fitted for the ERs considered with this study.
Table 3.2 GARCH (1, 1) model
Index Fitted model Mean Model Volatility Model
AR(1) MA(1)
ARMA(1,1)-
Indian RS
GARCH(1,1) 0.894551* 0.95614* 0.040616* -0.112871* -0.614615*
Japanese ARMA(1,1)-
Yen GARCH(1,1) 1.007253* 0.200095* 0.000133* 0.109196* 0.708945*
Sterling ARMA(1,1)-
Pound GARCH(1,1) 0.999656* 0.385907* 4.88E-05 -0.019035* 0.961792*
US ARMA(1,1)-
Dollars GARCH(1,1) 0.995512* 0.086428 3.654846 -0.328145* 0.496586*
ARMA(1,1)-
Euro
GARCH(1,1) 0.920125* 0.963136* 2.481137* -0.29407* 0.056265
Note: * denotes statistical significance at 5% level.

The reported results in the Table 3.2 indicate that all most all the coefficients except MA(1) term
in US Dollars and GARCH term in the Euro are statistically significant (p-value<0.05)
indicating that GARCH(1,1) model is adequate in deriving the volatility of the considered ERs.

5
3.3 Testing the Cointegration
The volatility series developed for the five ERs from the above econometric model is then tested
for the Cointegration. Following Table 3.3 indicates the results for the JCT.
Table 3.3 Unrestricted Cointegration Rank Test (Trace)
No. of CE(s) Eigen value Statistic P-value
None 0.429895 59.80534 0.2414
At most 1 0.311503 33.95634 0.5042
At most 2 0.255304 16.78708 0.6557
At most 3 0.05581 3.227236 0.9556
At most 4 0.012649 0.585567 0.4441

JCT results indicate that there is no any Cointegration at 5% significance level among these ERs.
Hence, the results confirm that VAR model is adequate to depict the volatility co-movement
among these ERs.

3.4 VAR Variance Decomposition Analysis

When using a VAR model, the selection of number of lags is very essential. Friedman and
Shachmurove (1997) suggest that higher lag orders ensures that all the dynamics in the data is
captured. However, in order to keep the model parsimonious, the Information Criteria governed
the lag length in this study. According to the Information Criteria it was found that lag order of
two is adequate to proceed with VAR analysis. After estimating a VAR model, the next step is
obtaining the variance decompositions. Variance decomposition separates the variation in an
endogenous variable into the component shocks to the VAR. Thus, the variance decomposition
provides information about the relative importance of each random innovation in affecting the
variables in the VAR. In other words, decomposition of variance of the forecast errors of the
volatility of a given ER indicates the relative importance of the various other ERs in causing the
fluctuations of that ER. An influencing ER is one which explains a large percentage of the error
variance of other ERs while its own forecast error is not explained by innovations in other ERs.
Table 3.6 reports the results obtained for VAR variance decompositions for one-day horizon for
the selected ERs. Entries show the percentage forecast error variance of the ER in the first
column explained by the ER in the first row.

6
Table 3.4: VAR Variance Decompositions
Period Indian Rs Japan yen St. Pound Dollars Euro
Indian Rs 10 36.20131 15.27519 5.400109 27.91876 15.20463
Japan yen 10 20.22833 62.53825 7.551895 5.26316 4.418369
St. pound 10 15.29482 31.08781 25.62515 23.1676 4.824609
Dollars 10 11.61968 7.965344 21.14826 52.0259 7.240825
Euro 10 2.542191 6.212693 34.99278 11.76891 44.48343

Table 3.4 indicates that ER co-movement is very much momentous since most of the considered
ERs depict more than 46% of their forecast error variance is explained by the other ERs under
study while Yen/SLRs depicts only 37%. The ER between InRs/SLRs accounts approximately
20.23%, 15.29%, 11.62% and 2.52%, of the forecast error variances of ERs of Yen/SLRs,
StPound/SLRs, Dollars/SLRs and Euro/SLRs respectively. This indicates that Yen/SLRs ER is
the most sensitive ER to the InRs/SLRs. Yen/SLRs accounts approximately 15.28%, 31.09%,
7.97% and 6.21% of the forecast error variances of InRs/SLRs, StPound/SLRs, Dollars/SLRs
and Euro/SLRs ERs respectively. StPound/SLR found to be the most sensitive ER to the
Yen/SLRs. StPound/SLRs accounts approximately 5.40%, 7.55%, 21.15% and 34.99% of the
forecast error variances of InRs/SLRs, Yen/SLRs, Dollars/SLRs and Euro/SLRs ERs
respectively. Euro/SLRs found to be the most sensitive ER to the StPound/SLRs. Dollars/SLRs
accounts approximately 27.92%, 5.26%, 23.17% and 11.77% of the forecast error variances of
InRs/SLRs, Yen/SLRs, StPound/SLRs and Euro/SLRs ERs respectively. InRs/SLRs found to be
the most sensitive ER to the Dollars/SLRs. Euro/SLRs accounts approximately 15.20%, 4.42%,
4.82% and 7.24% of the forecast error variances of InRs/SLRs, Yen/SLRs, StPound/SLRs and
Dollars/SLRs ERs respectively. InRs/SLRs found to be the most sensitive ER to the Euro/SLRs.
Yen/SLRs and Dollars/SLRs are found to be the most independent ERs since they account
62.54% and 52.03% respectively of their own forecast error variances.
Error! Reference source not found. (see appendix 1) depicts the recorded results of generalized
impulses as described by Pesaran and Shin (1998) constructs an orthogonal set of innovations
that does not depend on the VAR ordering. Panel I depicts that the ER of InRs/SLRs is highly
sensitive to the shocks given to Dollars/SLRs. According to the Panel II Yen/SLRs shows a
negative response for all the chocks given to other ERs while the response of StPound/SLRs is

7
persistent to the shocks of Indian Rupees in Panel III. According to the Panel IV the responses of
Dollars/SLRs are negative to the chocks given to Yen/SLRs and StPound/SLRs. Panel V depicts
that response of Euro/SLRs looks persistence to the shocks given to Dollars/SLRs.

4. Conclusions
This study investigated the volatility co-movement among the ERs of InRs/SLRs, Yen/SLRs,
StPound/SLRs, Dollars/SLRs and Euro/SLRs. VAR Variance Decompositions and Impulse
Response analysis were the two main econometric models used in deriving at final results of the
study. The results indicate as in the most previous studies of Bollerslev (1990), Kearney and
Patton (2000), Speight and McMillan (2001), Melvin and Melvin (2003), Black and McMillan
(2004) and by Calvet et al., (2006) that the co-movement is significant since most of the
considered ERs depict that very high percentage (46%) of their forecast error variance is
explained by the other ERs. Yen/SLRs and Dollars/SLRs are found to be the most independent
ERs since they are the least sensitive ERs to the movements of other ERs considered with this
study. There is a high degree of co-movement between StPound/SLRs and with other ERs of
Yen/SLRs, Dollars/SLRs, InRs/SLRs and Euro/SLRs, stood in declining order of ER co-
movement respectively. Interestingly StPound/SLRs ER is found to be the most sensitive
meanwhile, it is the most influential ER for the movements of other ERs considered with this
study. The second most influential ER is Dollars/SLRs while Yen/SLRs is the next most
influential ER. Antonakakis (2008) reported that Euro is the most dominant currency. However,
this study does not have enough evidences to confirm the above finding of Antonakakis. Rather,
this indicates that Euro/SLRs is the least influential ER.

5. References
1. Antonakakis, N., 2008. “Exchange Rate Volatility Comovements and Spillovers before
and after the Introduction of Euro: A Multivariate GARCH Approach.” Department of
Economics, Business School, University of Strathclyde.
2. Black, A. J. and McMillan D. G., 2004. “Long Run Trends and Volatility Spillovers in
Daily Exchange Rates.” Applied Financial Economics, 14, pp. 895-907.
3. Bollerslev, T. 1990. “Modelling the coherence in short-run nominal exchange rates: a
multivariate generalized ARCH model”. Review of Economics and Statistics, 72. pp.
498-505.

8
 4. Calvet, L., Fisher A. J. and Thompson S. B. 2006. “Volatility Comovements: A
Multifrequency Approach.” Journal of Econometrics, 131, pp. 179-215.
5. Engle, C. and Hamilton, J.D., 1990. "Long Swings in the Dollar: Are They in the Data
and Do
6. Engle, R. F., 1982. “Autoregressive Conditional Heteroscedasticity with Estimates of the
Variance of United Kingdom Inflation.” Econometrica. 50(4). pp. 987-1007.
7. Friedman, J. and Shachmurove, Y., 1997 “Co-Movements of Major European
Community Stock Markets: A Vector Autoregression Analysis”. Global Finance Journal.
8. pp. 257-278.
8. Kearney C. and Patton, A. J. 2000. “Multivariate GARCH Modeling of Exchange Rate
Volatility Transmission in the European Monetary System.” The Financial Review, 41,
pp. 29-48.
9. Markets Know It?," American Economic Review, 80, 689-713.
10. Melvin, M., and Melvin, B. P. 2003 “The Global Transmission of Volatility in the
Foreign Exchange Market.” Review of Economics and Statistics, 85, pp. 670-679.
11. Pesaran, H. H. and Shin Y., 1998 “Generalized Impulse Response Analysis in Linear
Multivariate Models.” Economics letters. 58(1). pp. 17-29.
12. Speight A. E. H., and McMillan, D. G., 2001 “Volatility spillovers in East European
Black-Market Exchange Rates.” Journal of International Money and Finance, 20, pp.
367-378.
13. Tsay, R. S., 2005. Analysis of Financial Time Series. Wiley-Interscience.

6. Appendix
Appendix 1: Impulse response analysis

Panel: Response Indian Rupees

To JAPAN YEN To STERLINGPOUND To US DOLLARS To EURO

.012 .012 .012 .012

.008 .008 .008 .008

.004 .004 .004 .004

.000 .000 .000 .000

-.004 -.004 -.004 -.004

-.008 -.008 -.008 -.008

-.012 -.012 -.012 -.012

1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

Panel II: Response of Japan Yen

To INDIAN Rs To STERLING POUND To US DOLLARS To EURO

.004 .004 .004 .004

.000 .000 .000 .000

-.004 -.004 -.004 -.004

-.008 -.008 -.008 -.008

-.012 -.012 -.012 -.012

-.016 -.016 -.016 -.016

1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

9
 Panel III: Response of Sterling Pound

To INDIAN Rs To JAPAN YEN To US DOLLARS To EURO

.02 .02 .02 .02

.01 .01 .01 .01

.00 .00 .00 .00

-.01 -.01 -.01 -.01

-.02 -.02 -.02 -.02

-.03 -.03 -.03 -.03

1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

Panel IV: Response of US Dollars

To INDIAN Rs To JAPAN YEN To STERLING POUND To EURO

.004 .004 .004 .004
.003 .003 .003 .003
.002 .002 .002 .002
.001 .001 .001 .001
.000 .000 .000 .000
-.001 -.001 -.001 -.001
-.002 -.002 -.002 -.002
-.003 -.003 -.003 -.003
-.004 -.004 -.004 -.004
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

Panel V: Response EURO

T o INDIAN Rs To JAPAN YEN T o ST ERLING POUND T o US DOLLARS

.02 .02 .02 .02

.01 .01 .01 .01

.00 .00 .00 .00

1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

10

View publication stats