Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Working Paper

in Economics and
Development Studies
Department of Economics
Padjadjaran University

No. 201404

Volatility Transmission between


Exchange Rates and Stock Prices in
Indonesia post 1997 Asia Crisis

Anhar Fauzan Priyono


Arief Bustaman

Department of Economics,
Padjadjaran University

February, 2014

Center for Economics and Development Studies,


Department of Economics, Padjadjaran University
Jalan Cimandiri no. 6, Bandung, Indonesia.
Phone/Fax: +62-22-4204510
http://www.ceds.fe.unpad.ac.id
For more titles on this series, visit:
http://econpapers.repec.org/paper/unpwpaper/
Volatility Transmission between Exchange Rates and Stock
Prices in Indonesia post 1997 Asia Crisis

Anhar Fauzan Priyono


Arief Bustaman
Department of Economics, Padjadjaran University

Abstract

Volatility of Indonesia Rupiah and Jakarta Composite Index remain one of main
issues in Indonesia economy after 1997 Asian crisis. The objectives of this research
are (1) determining the volatility of Indonesia Rupiah to US Dollar exchange rates
and Jakarta Composite Index (JCI) and (2) analysing the dynamic volatility
transmission between exchange rates and JCI. Exchange rate and JCI volatility
were measured using GJR-GARCH approach. Estimated using VAR model, this study
found that current volatility of exchange rate (ER) respond significantly to the
change of volatility of Jakarta Composite Index (JCI) in the previous month. On the
other hand, change in previous exchange rate volatility did not affect current JCI
volatility.

Key words: Indonesia financial market, volatility, GJR-GARCH, VAR.

Background

The rapid integration of Indonesia economy with world economy in the last
decade and the adoption of floating exchange rate regime in 1997 has increase
its vulnerability to external shocks.Two notable financial variables that often
exposed to external shocks are exchange rate and Jakarta Composite Index (JCI).
The following graphs depict exchange rate and JCI movements from August 1997
to July 2013. Period of 1997 Asia crisis and 2008 subprime mortgage crisis
marked with Rupiah depreciation and fall of JCI.

Graph 1. Growth of Exchange Rate (IDR/US$) and Jakarta Composite Index


.3

.2

.1

.0

-. 1

-. 2
1998 2000 2002 2004 2006 2008 2010 2012

ER JCI

Source: Authors calculation


Furthermore, despite relationship between the two variables, the graph above
also shows that both exchange rate and JCI have an indication of high volatility
after 1997.Volatility of the two variables is undesirable at least for two reasons:
a) at macro level, it is a sign of country risk; b) at micro level, it isdisruptive to
policymakers as well as investors.

In the past decade, there have been fast growing amount of empirical studies on
the relationship between exchange rate and stock price in Asian emerging
market, for instance Fang and Miller (2002), Bhattarchaya and Mukherjee
(2003), Kumar (2009), AsmanZaini, et al (2006), Baharom, Habibullah, and
Royfaizal (2008) as well as Maryatmo (2010). Most of the existing studies,
however, examining the relation between exchange rates and stock price, not its
volatility. Moreover, our literature review indicates none exists on the
relationship between volatility of the two variables for the case of Indonesia.
Therefore, present study contributes to the literature by considering the role of
volatility of exchange rate and stock index in Indonesia. As previously explained
that volatility of exchange rate and stock price exist in Indonesia and given the
fact that Indonesia is expected to increasingly be more integrated with world
financial market, it is imperative to analyse transmission of volatility between
exchange rates and stock index.

The objectives of this research are twofold: first, determining the volatility of
Indonesia Rupiah to US Dollar exchange rates and Jakarta Composite Index (JCI).
Second, analysing the dynamic volatility transmission between exchange rates
and JCI. The rest of this paper is organised as follow. Section 2 provides a brief
literature review in this area. Section 3 describes data and research
methodology. Section 4presents empirical results. Section 5 provides conclusion
of the study.

Literature Review

There has been no consensus either on the direction or the positive/negative


relationship between exchange rate and stock prices, both on theoretical basis as
well as on the empirical studies. There are two views on the direction of
relationship: (1) flow oriented model suggests direction of causality from
exchange rate to stock price, while (2) stock oriented model argued in opposite
direction. (Ali, et al. 2013). The flow-oriented model, suggests by classical
economists argued change of exchange rate will affect firms competitiveness
and trade balance, which furthermore will affect firms stock prices (Dornbusch
and Fischer, 1980). On the other hand, stock oriented theory (Branson, 1983 and
Frankel, 1983) suggests that stock prices dynamic will affect exchange rate
movements.
Similar to theories attempting to explain the direction of relationship, there
aretwo theoretical views on the relationship between exchange rate and stock
prices: one arguing a positive relationship between exchange rates and stock
prices, and the other one suggesting a negative relationship.The goods market
hypothesis argued that movement in exchange rate will affects firms profit and
stock prices. For exporting firms, exchange rate depreciation might increase its
revenue, profit, and hence stock prices, vice versa. On the other hand, for
importing firms, currency depreciation will depress its profit, as importing cost
is higher. This will be followed by contraction of its stock prices.Another theory,
namely portfolio balance model, considers a negative relationship between stock
prices and the exchange rate. An increase in domestic stock prices will
appreciate exchange rate as a result of higher foreign capital inflow.

Several previous empirical studies concerning the relationship between


exchange rate and stock price movement have been conducted. Chiang, Yang &
Wang (2000) presented a bivariate GARCH(1,1) conditional variance in
analyzing stock returns in Asian countries and their interaction to the foreign
exchange rate changes. They found that the value of national currency are
positively related to national stock returns. On the other hand, they conclude
that the regional and world factors, proxied by Japanese and US stock returns,
respectively, have a positive effect on the Asian stock returns.

Caporale, Pittis&Spagnolo (2002) employed a GARCH BEKK to test for causality


in-variance of the relationship between stock prices and exchange rates volatility
in four East Asian countries. They found that stock prices lead exchange rates
negatively in Japan and South Korea, in the pre-crisis, while in Indonesia and
Thailand, they were positively related. Yang &Doong (2004) explore
transmission mechanism of mean and volatility between stock and foreign
exchange markets for the G-7 countries. They found evidence supporting the
asymmetric volatility spillover effect, and shows that movements of stock prices
will affect future exchange rate movements. On the other hand, future changes of
stock prices are less influenced by the changes in exchange rates.

In contrast to the above studies, Morales & ODonnell (2006) found lack volatility
spillovers between stock markets and exchange rates in all East Asian markets,
except Taiwan. They conclude that starting from the Asian financial crisis,
investors and portfolio managers can diversify their assets between stocks and
currencies, significantly
Data and Research Methodology
Data used in this study are exchange rate of Indonesia Rupiah to US$ and Jakarta
Composite Index on monthly basis for period of September 1999 to July 2013.
This period is chosen based on the consideration that prior September 1999
Indonesia was still suffer from 1997 crisis. The following depicts descriptive
statistics for the growth of exchange rate (growth of IDR/US$) and the growth of
Jakarta Composite Index (return of JCI):

Table 1
Descriptive Statistics for Growth of IDR/US$

50
Series: ER
Sample 1999M09 2013M07
40 Observations 167

Mean 0.000798
30 Median 0.000649
Maximum 0.066505
Minimum -0.082189
20 Std. Dev. 0.014866
Skewness -0.315628
Kurtosis 11.41058
10
Jarque-Bera 494.9903
Probability 0.000000
0
-0.075 -0.050 -0.025 0.000 0.025 0.050

Table 2
Descriptive Statistics for Return of JCI

30
Series: JCI
Sample 1999M09 2013M07
25
Observations 167

20 Mean 0.005450
Median 0.009309
Maximum 0.079657
15 Minimum -0.163815
Std. Dev. 0.031212
10 Skewness -1.060605
Kurtosis 7.158341

5 Jarque-Bera 151.6314
Probability 0.000000
0
-0.15 -0.10 -0.05 0.00 0.05

Descriptive statistics above reports that the level of kurtosis exceed its normal
value (normal value is 3), namely leptokurtic. This is an indication of volatility
clustering exists in both data. This finding can be more clearly illustrated by
examining the following two graphs.

Graph 2
Movement of Growth of IDR/US$ (Sept 1999 July 2013)

ER
.08

.06

.04

.02

.00

-.02

-.04

-.06

-.08

-.10
2000 2002 2004 2006 2008 2010 2012

Graph 3
Movement of Return of JCI (Sept 1999 July 2013)

JCI
.10

.05

.00

-.05

-.10

-.15

-.20
2000 2002 2004 2006 2008 2010 2012
This research will attempt to model the relationships between volatilities of
exchange rate (Volat_ER) and Jakarta Composite Index (Volat_JCI). Those
variables will be estimated in a Vector Autoregression (VAR) system.

The procedures required before estimating a VAR model in this research are as
follow:
1. Testing for the existence of unit root using Augmented Dickey Fuller
(ADF) test.
2. Modeling each variable of interest (growth of exchange rate, growth of
Jakarta Composite Index) in an Autoregressive (AR) model. The proper
lag (autoregressive order) selection will be based onAkaike and Schwarz
criterions.
3. Utilizing the outcomes from the second procedure, we could determine
whether the generated residuals have property of heteroscedasticity
(time variant of variance). Most financial data has a unique characteristic,
namely volatility clustering (i.e., its variances are not constant over time).
The presence of heteroscedasticitymight lead us to employ Generalised
Autoregressive Conditional Heteroscedasticity (GARCH) class model, to
estimate the conditional volatility of a time series data. The unconditional
variance, however, will be constant over time. ARCH-effect test will be
conducted to determine the presence of volatility clustering.
4. Estimating the conditional variance of variables of interest over the
selected time horizon. This procedure is required to identify the volatility
of those variables in every single period. The volatility estimation will be
conducted in two ways. Firstly, in a GARCH(p,q) model, and secondly, in
Glosten, Jaganathan and Runkle (GJR)-GARCH. The first model is a natural
GARCH model, where variance is a function of its past squared residuals,
and its past variance. On the other hand, GJR-GARCH adds a dummy
variable taking 1 if the past residual is below zero, and 0 otherwise. The
GJR-GARCH has an advantage of determining different magnitude of bad
and good news in influencing the variance. If residual is below 0, then we
can say that bad news has occured, vice versa. Finally, Root Mean Squared
Error (RMSE), Mean Absolute Percentage Error (MAPE) and Mean
Absolute Error (MAE) procedures will be employed to evaluate forecast
performance of GARCH and GJR-GARCH.

Next, variances generated from the best volatility model will be used as variables
in VAR system. The main advantage of VAR model is its ability to produce
information from impulse response and variance decomposition. Impulse
response will show the reaction of any dynamic system in response to some
external change, while variance decomposition indicates the amount of
information each variable contributes to the other variables in VAR system.
Empirical Results

Augmented Dickey Fuller Test

We employed ADF approach to test for the presence of unit root in each data
series. The detailed results from ADF test are as below (t-stats in parentheses):

ADF test for growth of exchange rate (ER):


( ) = 0.0008 0.92 + 0.16( ) ...............................................................(1)
(0.75) (-9.67)*** (2.29)**
ADF stat = -9.67 ***

ADF test for return of JCI (JCI):


( ) = 0.004 0.79 ...................................................................................................(2)
(1.82)* (-10.4)***
ADF stat = -10.4 ***

***, **, * : significant at 1%, 5%, and 10% significance level

Equation (1) and (2) showed that the null hypotheses of each variables have a
unit root rejected, meaning that those series are stationary at level.

Autoregression modelling (OLS based)

Based on autocorrelation function (ACF) in correlogram and AIC, SC, HQBC


criterions, the estimated equations for ER and JCI obey an AR(3) process (t stat
in parentheses):

= 0.0009 + 0.22 0.14 0.01 ..................................................(3)


(0.83) (2.83)*** (-1.84)** (-0.19)
R2 = 0.06

= 0.004 + 0.23 0.07 0.13 ....................................................(4)


(1.56) (2.91)*** (-0.94) (1.73)*
R2 = 0.06

***, **, * : significant at 1%, 5%, and 10% significance level

Residuals generated from the above equations, then, tested for the presence of
autocorrelation and heteroscedasticity. We employed Breusch-Godfrey LM-test
for autocorrelation test, and ARCH-effect test for testing the presence of volatility
clustering (heteroscedasticity). Residuals estimated on equation (3) and (4)
shows no autocorrelation exist, but found that they exhibit volatility clustering
(heteroscedasticity).
Volatility Modeling

In order to obtain estimated volatility (variance) of each variables and given they
exhibit volatility clustering, we employed two variant of Generalized
Autoregressive Conditional Heteroscedasticity (GARCH) model. These are AR(3)-
GARCH(1,1) and AR(3)-GJR-GARCH model.

The detailed estimated volatility with AR(3)-GARCH(1,1) models for ER and JCI
are represented below (conditional variance equations; z-stat in parentheses):

Estimated volatility of ER:


( ) = 1.19 10 + 1.14 + 0.23( ) ................................................................(5)
(3.06)*** (7.18)*** (3.73)***
LL = 527.93

Estimated volatility of JCI:


( ) = 2.89 10 + 0.006 + 0.95( ) ............................................................(6)
(0.99) (0.37) (27.21)***
LL = 350.79

***, **, * : significant at 1%, 5%, and 10% significance level

Furthermore, the detailed estimated volatility with AR(3) GJR-GARCH models for
ER and JCI are represented below (conditional variance equations; z-stat in
parentheses & D is dummy variable taking 1 if previous residual is less than one
and 0 otherwise):

Estimated volatility of ER:


( ) = 9.79 10 ! + 1.60 + 0.27( ) 0.93"
(2.92)*** (5.18)*** (3.78)*** (-2.71)***
LL = 530.67

Estimated volatility of JCI:


( ) = 14.41 10 0.12 + 0.98( ) + 0.14"
(2.22)** (-1.87)* (37.23)*** (1.88)*
LL = 351.86

***, **, * : significant at 1%, 5%, and 10% significance level


Comparison of forecast evaluation between GARCH(1,1) and GJR-GARCH:
Table 3
Forecast Evaluation between GARCH(1,1) and GJR-GARCH

GARCH(1,1) RMSE MAE MAPE


ER 0.01374 0.0084 217.0738
JCI 0.03049 0.02281 148.7307
GJR-GARCH RMSE MAE MAPE
ER 0.0137 0.00841 208.2875
JCI 0.03035 0.02289 148.1839
Note: Numbers in bold indicates more efficient in forecast performance

Table 3 above implies that we better use GJR-GARCHto obtain the predicted
volatility of ER and JCIdue to its efficiency relative to GARCH(1,1). The volatility
graph obtained from GJR-GARCH of the two variables are as follow:

Graph 4
Conditional Variance (Volatility) of ER
VOLAT_ER
.006

.005

.004

.003

.002

.001

.000
2000 2002 2004 2006 2008 2010 2012

Graph 5
Conditional Variance (Volatility) of JCI

VOLAT_JCI
.0024

.0020

.0016

.0012

.0008

.0004

.0000
2000 2002 2004 2006 2008 2010 2012
Vector Autoregression (VAR) Modelling

The table below shows the Granger Causality test between the volatility of ER
and JCI:

Table 4
Granger Causality Test

Null Hypothesis Obs F-stat P-value


Volat_JCI does not Granger Cause Volat_ER 9.99*** 8*10-5
162
Volat_ER does not Granger Cause Volat_JCI 0.58 0.56
*** : significant at 1% significance level

Table 4 shows that volatility of JCI affect volatility of ER, and not vice versa. The
null hypothesis of JCI volatility not causing ER volatility is rejected.

EstimationThe VAR model for those 2 variables are shown below:

Table 5
Vector Autoregression Estimation Result
Volat_ER Volat_JCI
Volat_ER(t-1) 0.53 -0.01
[6.81]*** [-1.08]
Volat_ER(t-2) -0.16 0.006
[-2.01] [0.5]
Volat_JCI(t-1) 1.62 1.04
[3.44]*** [13.57]***
Volat_JCI(t-2) -1.09 -0.08
[-2.39]** [-1.06]
Constant -0.0002 3.54*10-5
[-2.15]** [1.71]*
R2 0.40 0.93
[] = t stat
***, **, * : significant at 1%, 5%, and 10% significance level

VAR estimation results on table 5 above shows current volatility of exchange rate
is significantly affected by volatility of JCI in previous 2 periods, despite its own
lag 1 contribution. Furthermore, the magnitude of JCI volatility contribution on
exchange rate volatility is positive on lag 1 and negative on lag 2. On the other
hand, volatility of exchange rate in previous period is not statistically significant
contributes to current volatility of JCI. These results suggest that the direction of
relationship between exchange rate and stock price in Indonesia is stock price to
exchange rate (i.e., supporting portfolio balance model). We offer the following
explanation that might help understand the findings. Firstly, as an emerging
market, after 1997 crisis Indonesia financial market has been very attractive for
foreign investors. At the same time, however, investors are very sensitive to
perceived economic and political risks that lead to a situation where foreign
capitals are easily come and out of the financial market. This situation more
likely contibute to volatility in exchange rate. Secondly, it can be argued that flow
oriented model is not empirically found in this study is due to characteristics of
export firms in Indonesia. In the case of export firms, while the flow oriented
model argued that exchange rate depreciation will increase exporters profit
hence its stock price, the fact that majority of Indonesia export products are
mostly using imported materials, exchange rate depreciation would not means
increase competitiveness and profit (due to higher production cost as a result of
higher importing costs).

The impulse response and variance decomposition are presented below:

Graph 6
Response of Volat_JCI to Volat_ER (Response of Cholesky 1 SD Innovation)

Response of VOLAT_JCI to Cholesky


One S.D. VOLAT_ER Innovation
.00004

.00003

.00002

.00001

.00000

-.00001

-.00002

-.00003
1 2 3 4 5 6 7 8 9 10

Graph 7
Response of Volat_ER to Volat_JCI (Response of Cholesky 1 SD Innovation)

Response of VOLAT_ER to Cholesky


One S.D. VOLAT_JCI Innovation
.00025

.00020

.00015

.00010

.00005

.00000
1 2 3 4 5 6 7 8 9 10
Table 6
Variance Decomposition of Volat_ER

Period S.E Volat_ER Volat_JCI


1 0.000592 100 0
2 0.000701 95.21 4.79
3 0.00072 91.79 8.21
4 0.000726 90.31 9.69
5 0.000729 89.51 10.49
6 0.000732 88.88 11.12
7 0.000734 88.29 11.71
8 0.000737 87.75 12.25
9 0.000739 87.26 12.74
10 0.000741 86.82 13.18
Average 90.582 9.418

Table 7
Variance Decomposition of Volat_JCI

Period S.E Volat_ER Volat_JCI


1 0.00009 3.07 96.93
2 0.000138 1.96 98.04
3 0.000166 1.55 98.45
4 0.000187 1.39 98.61
5 0.000205 1.33 98.67
6 0.00022 1.29 98.71
7 0.000232 1.27 98.73
8 0.000243 1.25 98.75
9 0.000252 1.23 98.77
10 0.00026 1.22 98.78
Average 1.556 98.444

Looking at variance decomposition results above, on 10 months average, the


contribution of JCI volatility to ER volatility is about 9.4%, while volatility of ER
contribution to JCI volatility is 1.55%. These findings suggests that volatility of
JCI contributes more to the volatility of ER than volatility of ER to the JCI.
Conclusion

In this study we analysed the transmission between volatility of exchange rate


and stock prices. The AR(3) GJR-GARCH model chosen relative to AR(3)
GARCH(1,1), and employed to measure volatility, thus VAR model was used to
analyse the transmission direction and relationship of the two variables under
consideration. The resultsshow current volatility of exchange rate is significantly
affected by volatility of JCI in previous 2 periods, despite its own lag 1
contribution. On the other hand, volatility of exchange rate in previous period is
not statistically significant contributes to current volatility of JCI. In this case, one
might argue that the results support portfolio balance model. Results from
variance decomposition analysis present the magnitude of relationship between
volatility of exchange rate and JCI.

References

Ali G., Ahmad Anwar, M.B., and Ziaei, S.M. 2013. A Bivariate Causality Test
Between Exchange Rates and Equity Markets in BRIC Countries.Middle-
East Journal of Scientific Research 13: 213-219

AzmanZaini, WMH.,Habibullah, M.S., Law, S.K., and Dayang-Afizzah, A.M. 2006.


Stock Prices, Exchange Rates and Causality in Malaysia: A Note. MPRA
Paper, No 656

Baharom, A.H., Habibullah, M.S., and Royfaizal, RC.2008. Pre and Post Crisis
Analysis of Stock Price and Exchange Rate: Evidence from Malaysia, MPRA
Paper, No 12445.

Bhattacharya, B. and J. Mukherjee. 2003. Causal Relationship between Stock


Market and Exchange Rate. Foreign Exchange Reserves and Value of Trade
Balance: A Case Study for India, Paper presented at the Fifth Annual
Conference on Money and Finance in the Indian Economy, January 2003.

Branson, W. 1983. Macroeconomic Determinants Of Real Exchange Risk,


Managing Foreign Exchange Risk (Cambridge University, Cambridge).

Caporale, G.M., Pittis, N &Spagnolo, N. 2002. Testing for Causality-in-Variance:


An Application to the East Asian Markets. International Journal of Finance
and Economics.7: 235-245.

Chiang, T.C., Yang, S.Y & Wang, T.S. 2000. Stock Return and Exchange Rate Risk:
Evidence from Asian Stock Markets Based on a Bivariate GARCH Model.
International Journal of Business, 5(2).
Dornbusch, Rudiger and Stanley Fischer (1980), Exchange Rates and the Current
Account, American Economic Review, Vol. 70, pp. 960971.

Fang, W.S and Miller, S.M. 2002. Currency Depreciation and Korean Stock
Market Performance during the Asian Financial Crisis, Economics Working
Papers. Paper 200230

Frankel, J. A. 1983, Monetary and Portfolio-Balance Models of Exchange Rate


Determination, in J. S. Bhandari and B. H. Putnampp (eds.), Economic
Interdependence and Flexible Exchange Rates, Cambridge: MIT Press, pp.
84115.

Kumar, M. 2009. "A Bivariate Linear and Nonlinear Causality between Stock
Prices and Exchange Rates, Economics Bulletin, Vol. 29(4)

Maryatmo, R. 2010. The Short Run and Long Run Impact of Changes in Interest
Rate and Exchange Rate on Stock Prices: Empirical Evidence in Indonesia
(2000:12010:4). MPRA Paper, No 25532

Morales, L & ODonnell, M. 2006. Volatility Spillovers between Stock Prices and
Excnge Rates: Empirical Evidence from Six APEC Economies. Conference
Papers. School of Accounting and Finance. Dublin Institute of Technology

Yang, S-Y.,Doong, S-C. 2004. Price and Volatility Spillovers between Stock Prices
and Exchange Rates: Empirical Evidence from the G-7 Countries.
International Journal of Business and Economics, vol.3, no.2, 139-153.

You might also like