The current issue and full text archive of this journal is available at

www.emeraldinsight.com/2045-4457.htm

RESEARCH ARTICLE Relationship

between stock
The relationship between stock prices and
prices and exchange rates exchange rates

in Asian markets 209

A wavelet based correlation and quantile Received 12 July 2013
Revised 12 August 2013
regression approach 14 October 2013
19 March 2014
Arif Billah Dar 9 May 2014
20 May 2014
Department of Economics, Institute of Management Technology, Ghaziabad, India Accepted 22 May 2014
Aasif Shah
Department of Commerce, School of Management, Pondicherry University,
Pondicherry, India
Niyati Bhanja
Department of Economics and International Business,
University of Petroleum and Energy Studies (UPES), Dehradun, India, and
Amaresh Samantaraya
Department of Economics, Pondicherry University, Pondicherry, India

Abstract
Purpose The purpose of this paper is to estimate the relationship between stock prices and
exchange rates of eight Asian countries. The analysis is based on methodologies that possess the
ability to provide a complete representation of data series from both time and frequency perspectives
simultaneously. In addition, instead of limiting the analysis to focus on the conditional mean of the
response variable y in the regression equation, the authors investigate the extremes of distribution to
reveal a range of hidden relationships between these variables.
Design/methodology/approach Given the limitations of classical methodology of Pearson
correlation and least-squares regression, this study estimates the relationship between stock prices
and exchange rates through wavelet correlation and cross-correlation to serve as a protocol for
different traders who view the market with different time resolutions. In addition, quantile regression
technique robust to heteroscedasticity, skewness and leptokurtosis is used to understand the
relationship between stock prices and a specified quantile of the exchange rates.
Findings In accordance with the portfolio balance effect, it is observed that stock prices and
exchange rates are negatively correlated at all frequencies. In particular, the negative correlation
grows with higher time scales (lower frequency intervals). The findings from quantile regression also
suggest that the coefficients are more inclined to be negative when exchange rates are extremely high.
Originality value The paper contributes to the literature by focussing on the multi-scale
relationship between stock prices and exchange rates. In addition, it also analyzes the relationship
between stock prices and a specified quantile of the exchange rates.
Keywords Quantile regression, Wavelet correlation, Asian
Paper type Research paper
South Asian Journal of Global
Business Research
JEL Classifications G10, G15, C22 Vol. 3 No. 2, 2014
The authors are very thankful to the Editor-in-Chief of this journal for highly useful comments pp. 209-224
r Emerald Group Publishing Limited
and suggestions. The authors also express our gratitude to the reviewers for pointing out some 2045-4457
very critical issues that helped us to improve this paper considerably. Standard caveats apply. DOI 10.1108/SAJGBR-07-2013-0061
SAJGBR 1. Introduction
3,2 The interaction between equity and currency markets has been the subject of much
academic debate over the past three decades (see, e.g. Aggarwal, 1981; Abdalla and
Murinde, 1997; Hau and Rey, 2006; Cumming et al., 2011; Tiwari et al., 2014). This is
understandable given the decisive role that equity and currency markets play in
facilitating economic activity (Akpan, 2008; Choong et al., 2010; Bussiere et al., 2014).
210 Theoretically, the traditional approach (also known as the flow-oriented theory) states
that a depreciation of domestic currency can have a crucial impact on stock prices by
increasing firms competitiveness while in turn raising their profitability (Dornbusch and
Fischer, 1980). In other words, flow-oriented theory proposes that exchange rates would
lead stock prices with a positive relation. On the contrary, the portfolio balance approach
assumes that the stock prices would lead exchange rates on the ground that a rising trend
in stock price induces overseas investors to invest more in domestic stocks. This would
result in more capital inflows due to higher interest rates and eventually an appreciation
in domestic currency (Branson and Henderson, 1985). Hence according to latter approach,
stock prices should lead exchange rates negatively. A third approach, referred to as the
monetary approach or asset market approach argues no linkage between stock price
and exchange rates due to different factors influencing the two variables (MacDonald and
Taylor, 1992; He and Ng, 1998; Griffin and Stulz, 2001).
Empirical studies also provide contradictory evidence. For example, Ibrahim and
Aziz (2003), Kim (2003) and Tian and Shiguang (2010) note that there is essentially
a long-term equilibrium relationship between stock price and exchange rate. On other
hand, Bahmani-Oskooee and Sohrabian (1992), Nieh and Leeb Chung (2001) and Smyth
and Nandha (2003) argue that this relationship is merely short term. Mishra (2004)
finds that there is no causal relationship between the returns on exchange rate and
stock return in India; and Komain contends that stock prices and exchange rates do not
exhibit any long run relationship in Thailand. A more recent study by Tsagkanos and
Siriopoulos (2013) observe that changes in exchange rates drive stock returns in the short
run but the relationship is stable in the long run. Most of these studies however do not
provide any reason for these contradictory relationships. MacDonald and Taylor (1992)
explain these mixed findings (i.e. short run relationship, long run relationship, positive
relationship, negative relationship and no relationship) arguing the use of asset-approach
models responsible for providing inadequate explanations for the behavior of the major
exchange rates.
Literature has also analyzed the theoretical foundation of the relationship between
stock and foreign exchange markets, and two views dominate: the international
trading effect (Dornbusch and Fischer, 1980; Aggarwal, 1981) and the portfolio balance
effect (Branson and Henderson, 1985; Bahmani-Oskooee and Sohrabian, 1992).
Based on the international trading effect, Aggarwal (1981) observes that a change in
exchange rate cannot only directly influence the stock prices of multinational and
export-oriented firms but it can indirectly influence domestic firms. For multinational
firms, exchange rate fluctuations immediately influence the value of its foreign
operations and persistently affect the earnings of the firm. Exchange rate fluctuations
also influence the demand for exports and imports. For example an appreciation of the
dollar will lead to exports becoming more expensive and imports cheaper. Exchange
rate depreciation, on the other hand, increases the competitiveness of exports as well
increases the input cost of import trade. Since Asian countries are reasonably more
export-dominant and currency depreciation usually has a positive effect on the
indigenous stock markets in these countries there may be a positive relationship between
the underlying variables (Ma and Kao, 1990). However, Branson and Henderson (1985); Relationship
Bahmani-Oskooee and Sohrabian (1992) used a portfolio balance approach to analyze the between stock
influence of stock prices on exchange rates. They argue that an upward movement in
stock price increases wealth of domestic investors, thereby leading the demand for the prices and
currency according to the investment portfolio equilibrium theory. The increased demand exchange rates
for money would in turn raise the interest rates, absorb the inflow of foreign capital and
consequently cause the domestic currency to appreciate. Thus, the more optimistic are 211
investors regarding the stock market, the greater will be their investment because of the
speculative demand, which is indirectly responsible for the appreciation of the countrys
currency. Further, they argue that since stock markets and foreign exchange markets in
Asia have also become increasingly attractive to foreign capital in recent decades, it has
also increased the likelihood of portfolio balance effect to subsist in such markets.
The common feature of earlier studies whether based on portfolio balance approach
or international trading effect is that they are based on conventional time-domain
approach without making any attempt to observe these relationships at different
frequencies (Nieh and Leeb Chung, 2001). Since, traders in the financial markets deal at
different time horizons or frequencies (Hacker et al., 2012), the link between stock
market and foreign exchange market can vary across frequencies. For example, consider
the large number, including intraday traders, hedging strategists, international portfolio
managers, commercial banks, large multinational corporations, pension funds and
national central banks. It is obvious that these market participants operate with different
time horizons based on their respective investment requirements and strategies.
Consequently, the true dynamic structure of the relationship between exchange rates and
stock prices is likely to vary across different horizons. Hence, it is important to capture
more precise information on the relationship between exchange rates and stock prices at
different frequencies. Second, using the method of ordinary least square (OLS) for cause
and effect relationship cannot be termed as an approximate standard approach given its
limitation to restrict the precise description of the tails of the distribution of ( y) variable
(Koenker and Bassett, 1978). Our study addresses these two gaps in the literature
and adds to previous studies at least in two ways: First, we use non-orthogonal maximal
overlap discrete wavelet transform analysis (MODWT) to decompose returns of
exchange rate and stock prices and analyze the relationship at different time scales.
The decomposition generates coefficients that are localized in both time and frequency.
Second, we analyze the relationship at different quantiles of exchange rates, rather
than the approximation of the conditional mean. The paper therefore provides a new
explanation based on wavelet correlation and quantile regression for the relationship
between these two markets. Our findings support portfolio balance approach implying
that the relationship between stock indices and exchange rates is essentially negative
across different time scales and the level of weak or strong relationship depends upon the
time horizon of investors and the quantile of exchange rates at which the relationship
is measured. In particular, we observed that when periods of longer than one year are
considered, the existence of intense negative relationship between stock markets and
exchange rates cannot be ruled out. On the other hand, result from quantile regression
suggests that coefficients are more inclined to be negative when exchange rates are
extremely high. This is in accordance with Tsais (2012) work. We believe that our results
are robust due to the use of two different methodologies.
The remainder of the paper is organized as follows. Section 2 describes wavelet
methodology and quantile regression approach. Section 3 presents the data description
and results. Finally, Section 4 draws the conclusions.
SAJGBR 2. Multi-scale decomposition of the exchange rate and stock returns using
3,2 wavelet approach and quantile regression
The notion of multi-scale features is a general issue in macro and financial time series
Ramsey (2002). Hence an observed time series may contain several compositions each
occurring on a different time scale. Wavelet techniques possess an inherent ability to
decompose the time series into several sub-series, which may be associated with a
212 particular time scale Ranta (2013). In particular, wavelet methods present a lens to the
researcher, which can be used to zoom on the details and draw an overall picture of a
time series in the same time.
Until recently, the relationship between exchange rates and stock prices has been
mostly analyzed following conventional time-domain approach where the frequency
based approaches have been ignored[1]. There are number of reasons, which have been
suggested for not using OLS in conventional time domain framework. Noteworthy
reasons include: the spurious regression for non-stationary time series and its inability
to provide information about long run relationship Engle and Granger (1987). Nevertheless,
we focus on the premise that there is a possibility that this relationship may vary at
different frequencies. For example, the heterogeneous market premise, developed by Muller
et al. (1997), contends that stock market consists of a large number of heterogeneous
investors operating at different time scales (from minutes to years). Each market
component has its own reaction time to information, related to its time horizon and
characteristic dealing frequency (Dacorogna et al., 2001) because the market participants
may vary in their beliefs, future outlooks, risk profiles and informational sets, etc. As such
these variations translate their sensitivity to different time scales. Consequently operating
at different frequencies, market participants can influence dynamic of prices and market on
the whole in different ways. Given the multi-horizon nature of different variables, wavelet
transform is a pertinent tool for modeling financial markets heterogeneity and price
dynamics induced by the influence of different types of investors characterized by different
time horizons (Ramsey, 2002). The link between stock market and foreign exchange market
can therefore vary across frequencies, and such relationship may even change over time.
Given its ability to decompose the time series data into different time scales and modeling
financial market heterogeneity wavelet approach is useful. The wavelet and scaling
coefficients at the first level of decomposition are obtained by convolution of the data series
with the father wavelets j(t) and mother wavelet c(t):
Z Z
ctdt 0; jtdt 1 1

The father wavelets are used for the low frequency smooth components parts of a
signal and the mother wavelets are used for the high-frequency details components.
In other words father wavelets are used for the trend components and mother wavelets
are used for all the deviation from trend. Hence, a sequence of mother wavelets is
used to represent a function and only one father wavelet is used to represent a function
(see, e.g. Benhmad, 2012; Tiwari et al., 2013; Dar et al., 2014).
A time series, say f(t), can be decomposed by the wavelet transformation, which can
be expressed as follows:
X X X X
f t sJ ;k fJ ;k t dJ ;k cJ ;k t dJ 1;k cJ 1;k t . . . . . . d1;k c1;k t
k k k k
2
where J is the number of multi-resolution levels, and k ranges from 1 to the number Relationship
of coefficients in each level. The wavelet coefficients sJ, k, d J, k, y, d1, k are the wavelet between stock
transform coefficients and f J, k(t) and c J, k(t) represents the approximating wavelets
functions. The wavelets transformations can be expressed as: prices and
Z exchange rates
sJ ;k fJ ; k tf tdt 3
Z 213
dj;k cj; k tf tdt; for j 1; 2; . . . . . . J : 4

where J is the maximum integer such that 2 J takes value less than the number of
observations.
The detail coefficients, d J,k, y, d1,k, represents increasing finer scale deviation from
the smooth trend and s J,k which represent the smooth coefficient capture the trend.
Hence, the wavelet series approximation of the original series f(t) can be expressed
follows:
f t SJ ;k t DJ ;k t DJ 1;k t . . . : : D1 t: 5

where S J,k is the smooth signal and D J,k, D J1,k, D J2,k y D1,k detailed signals. These
smooth and detailed signals are expressed as follows:
X X X
SJ ; k sJ ; k fJ ;k t; DJ ;k dJ ;k cJ ;k t; and D1;k d1;k c1;k t; j 1; 2; . . . ; J  1
k k k

with S J,k, D J,k, D J1,k, D J2,k y D1,k listed in increasing order of the finer scale
components.

2.1 The wavelet correlation and cross-correlation

The wavelet correlation is made up of the wavelet covariance for {xt, y t} and wavelet
variances for {xt} and {y t}. Wavelet variance primarily refers to the substitution of
variability over certain scales for the global measure of variability estimated by sample
variance. The wavelet variance of stochastic process X is estimated using the MODWT
coefficients for scale tj 2 j1 through:
1 NX
1
^j;k 2
^2x tj
s W 7
N^jkLj 1

where W ^j;k the MODWT wavelet coefficient of variable X at scale is tj  N^j N Lj 1

is the number of coefficients unaffected by boundary, and L j (2 j 1)(L1) 1 is the
length of the scale tj wavelet filter.
The wavelet covariance decomposes the covariance between two stochastic processes
on a scale by scale. The wavelet covariance at scale tj can be expressed as follows:

1 NX
1
^x W ^y
gXY tJ covXY tJ W j;k j;k 8
N^j kLj 1
SAJGBR Since covariance does not take into consideration the deviation of univariate time
3,2 series, it is therefore imperative to introduce the concept of wavelet correlation.
Given the wavelet covariance for {xt, y t} and wavelet variances for {xt} and {y t},
the MODWT estimator of wavelet correlation can be expressed as follows:
Covxy tj
^xy tj
r 9
214 ^2x tj ^
s s2y tj

The wavelet correlation is analogous to its Fourier equivalent, the complex coherency
(Gencay et al., 2002, p. 258). It gives the correlation measure at different frequencies.

2.2 Quantile regression

While a great majority of regression models are concerned with analyzing the conditional
mean of a dependent variable, there is increasing interest in methods of modeling other
aspects of the conditional distribution (Koenker and Bassett, 1978). Quantile regression,
that was originally proposed by Koenker and Bassett (1978) attempts to provide
estimates of the linear relationship between regressors and a specified quantile of the
dependent variable. Over the years, several scholars have used this method to analyze
different subjects (see, e.g. Deaton, 1997; Buchinsky, 2001; Bassett and Chen, 2001; Tsai,
2012). The advantage of this approach is that it allows for a more precise description of
the tails of the distribution of y. Moreover, it is robust to heteroskedasticity, skewness and
leptokurtosis which are common features of financial data (Koenker and Bassett, 1978).
The traditional approach uses the OLS to estimate a linear regression model. However,
this method only provides the estimation of median (0.5th quantile) function. Since
quantile regression method can estimate the relationship over different quantiles of
dependent variable, this approach of estimating relationship is useful. Suppose there is a
linear specification for the conditional quantiles of E:
Et Xt b mt 10

where Et is the exchange rate of a country; Xt is K  1 repressors, which represents the

various stock price indices used in this paper; b is the coefficients that has to be
estimated and the purpose of the quantile regression model is to estimate b for different
conditional quantile functions; and mt is error term.
Assume that the conditional mean of E is m(X) X0 b, the approach of OLS is to
estimate the mean:
Xn Xn
min Et  m2 that is; minp Et  X 0t b2 11
m2R b2R
t1 t1

Solving Equation (10) will give the estimation of median (0.5th quantile) function.
For the other quantiles, we let t stand for quantile variable. The conditional quantile
function can be written as:
QE t=X X 0 bt 12

To obtain estimation of the conditional quantile functions, we need to solve:

Xn
minp rt Et  X 0t b2 13
b2R
t1
To minimize the following equation: Relationship
2 3 between stock
X   X   prices and
min4t Et  bX
^ t  1  t Et  bX
^ t 5 14
b^ ^ t
Et XbX ^ t
Et pbX
exchange rates
where X 0t b^t is an approximation to the tth conditional quantile of E. When t is 215
close to 0 (1), X 0t b^t characterizes the behavior of E at the left (right) tail of the
conditional distribution.

3. Data, results and discussion

For empirical estimation, monthly data over the period beginning February 1996 to
September 2013 of stock indices and exchange rates for eight Asian countries was
used. These countries comprised of four SARRC (India, Pakistan, Bangladesh and
Pakistan) and four emerging (Malaysia, Indonesia, Philippines and Thailand: IMF 2012
release). The choice of the SAARC markets was made due to the data availability.
In addition, we have excluded China due to its fixed exchange rate regime. The data
were retrieved from the Bloomberg database[2].
The test of relationship between stock prices and exchange rates forms an
important part of enquiry given that the global portfolio investment industry is going
through rapid change. Thus, understanding and analyzing structural trends of such
markets becomes crucial both for individual as well as institutional investors. For example
recently, the Indian rupee has fallen against the rising dollar demand and narrowing
spread between the USA and Indian bond yields. There was net outflow of $7.2 billion
from the Indian market in June 2013, of which $5.4 billion was from debt market triggered
by global risk aversion and currency weakness (Deutsche Bank India Equity Strategy
Report 2013). The report also highlighted that equity outflows from other emerging Asia
markets like South Korea, Taiwan, Indonesia and Thailand was between $2 billion and
$4.5 billion. This is also true for other South Asian countries (like Pakistan, Bangladesh
and Sri Lanka) included in our group. Within this backdrop, an attempt has been made
to analyze the relationship between equity and currency markets of Asian markets. Prior
to analysis, the data are tested for the stationary property using the Augmented Dickey
Fuller test. The results are shown in Table I. It is shown that both the time series in natural
logarithms, for all the countries, are non-stationary at level. Nevertheless, all these series
attain stationary when their returns are considered by taking the first difference of their

Level First difference

Stock price Exchange rate Stock price Exchange rate

India 0.702 (0.84) 0.005 (0.95) 14.00 (0.00) 13.00 (0.00)

Pakistan 0.151 (0.94) 0.734 (0.99) 14.27 (0.00) 11.78 (0.00)
Bangladesh 0.645 (0.85) 1.133 (0.70) 11.59 (0.00) 14.22 (0.00)
Sri Lanka 0.026 (0.95) 0.995 (0.75) 12.82 (0.00) 13.21 (0.00)
Malaysia 1.341 (0.61) 2.315 (0.16) 12.00 (0.00) 13.22 (0.00)
Indonesia 0.283 (0.92) 3.123 (0.02) 12.16 (0.00) 10.82 (0.00)
Philippines 0.281 (0.92) 2.414 (0.13) 12.47 (0.00) 13.46 (0.00)
Thailand 1.705 (0.42) 2.218 (0.20) 14.33 (0.00) 12.61 (0.00) Table I.
Augmented Dickey-Fuller
Notes: Figures represent t-statistic. In parentheses are the MacKinnon p-values unit root test
SAJGBR natural logarithms. The descriptive statistics of the returns on exchange rates and stock
3,2 prices are shown in Tables II and III. It is shown in Table I that the sample mean is positive
for the returns on exchange rates. Similarly, the sample mean is positive except for
Thailand stock returns. The skewness measure indicates that for all the countries, the
returns on exchange rates are positively skewed. Contrary to this, the sock returns of
all the countries except Bangladesh and Sri Lanka are negatively skewed. Both the series
216 for all the countries also show excessive kurtosis indicating that the distributions are
leptokurtic relative to normal distribution. Except for the Sri Lankan stock returns, the
Jarque-Bera test rejects normality of both the returns series in all the countries. Since all
the series attain stationarity after first difference, it is useful to test stock prices and
exchange rates for the cointegration. As the data period consists of important events like
Asian crisis and subprime crisis, we use Gregory-Hansen (1996) cointegation which
considers breaks to test the relationship. The Gregory-Hansen (1996) cointegation test
offers a novel way to test the long-run relationship between the variables of interest.
It involves testing of one unknown shift in the cointegrating vector by testing the null
of no cointegration against the alternative of cointegration with a break. The procedure of
Gregory and Hansen can be applied at different steps to detect the shift in the parameters
of the model. A shift in the intercept of the model is tested by testing the cointegrating
relationship between exchange rate and stock prices (Model C), shift in intercept with the
trend (Model C/T) and finally a more generic formulation involving a regime shift (Model
C/S). The results reported in Table IV suggest that the null hypothesis of no cointegration
cannot be rejected for India, Pakistan and Bangladesh irrespective of the models except for
Pakistan in the regime shift model (Model C/S) at 5 percent level of significance. However,
for Sri Lanka, Malaysia, Indonesia, Philippines and Thailand there is evidence of long-run

India Pakistan Bangladesh Sri Lanka Malaysia Indonesia Philippines Thailand

Mean 0.007 0.0004 0.00103 0.0005 0.0032 0.00185 0.0013 0.0023

Median 0.006 0.0004 6.87E0 0.0001 0.0007 0.00083 6.31E05 0.0005
Maximum 0.155 0.10904 0.05564 0.0505 0.3642 0.0359 0.027 0.0471
Minimum 0.041 0.09691 0.0363 0.0719 0.1489 0.0239 0.0168 0.037
SD 0.021 0.01533 0.01099 0.01084 0.0390 0.00627 0.0049 0.0079
Table II. Skewness 1.778 1.12069 1.52661 0.0271 3.8206 1.08637 2.125 1.5798
Descriptive Kurtosis 14.228 23.6173 9.43625 17.8081 40.5034 10.7612 14.49 13.814
statistics of returns Jarque-Bera 1,225.46 3,799.21 448.269 1,937 12,939.9 573.793 1,326.17 1,121.2
on exchange rates Probability 0 0 0 0 0 0 0 0

India Pakistan Bangladesh Sri Lanka Malaysia Indonesia Philippines Thailand

Mean 0.0038 0.005 0.003 0.0044 0.001 0.004 0.0015 0.000

Median 0.004 0.007 0.000 0.003 0.004 0.008 0.0049 0.005
Maximum 0.108 0.1047 0.247 0.0978 0.127 0.108 0.1440 0.1678
Minimum 0.118 0.194 0.157 0.079 0.123 0.164 0.129 0.159
SD 0.0326 0.0405 0.0421 0.031 0.030 0.0372 0.0325 0.0442
Skewness 0.341 1.035 0.566 0.1826 0.086 1.152 0.423 0.133
Table III. Kurtosis 3.593 7.250 9.991 3.551 6.8925 7.296 6.3683 5.381
Descriptive statistics Jarque-Bera 7.2301 197.48 443.09 3.867 134.10 209.98 106.55 50.73
of stock returns Probability 0.0269 0 0 0.144 0 0 0 0
 Model 2 Model 3 Model 4
Relationship
Break in intercept: Break in intercept: Regime shift between stock
no trend (C) with trend (C/T) (C/S) prices and
Break Lag Test Break Lag Test Break Lag Test
date length statistics date length statistics date length statistics exchange rates
India 1999:2008 4 3.154 2010:2011 0 2.713 2000:2011 1 2.786 217
Pakistan 2003:2008 0 4.346 2003:2008 0 4.651 2005:2009 2 5.29**
Bangladesh 2009:2004 4 4.081 1998:2012 8 4.921 2008:2001 3 4.158
Sri Lanka 1998:2008 4 4.90** 1998:2008 4 5.24** 2001:2011 8 5.023**
Malaysia 1999:2010 1 6.522* 1999:2010 1 6.469* 2001:2006 1 7.412*
Indonesia 2006:2012 1 7.667* 1999:2001 1 8.096* 2001:2008 1 7.885*
Philippines 2009:2009 1 5.775* 2003:2011 1 6.347* 2002:2008 0 8.096*
Thailand 2009:2003 1 6.096* 2004:2002 1 7.275* 2002:2010 0 8.897*
Notes: The critical values for the Gregory-Hansen tests are drawn from Gregory and Hansen (1996).
The approximate asymptotic critical values are 5.13 and 4.61 at 1 and 5 percent, respectively, for
break in intercept and the no trend model; 5.45 and 4.99 at 1 and 5 percent, respectively, for break
in intercept when the trend is included in the model and critical values are 5.47 and 4.95 at 1 and Table IV.
5 percent, respectively, for the full structural break (regime shift) model. *,**Statistical significance at Gregory Hansen
1 and 5 percent level, respectively cointegration test results

cointegrating relationship across the models at 5 percent level of significance or better.

The results, thus, suggest evidence of long run relationship between exchange rate and
stock prices with structural break. Moreover, since the estimated break points correspond
to the minimum values of test statistics, these break points may be treated as time points
at which the relationship has strongest tendency to shift. The break points are observed at
different time period for different countries and vary across the models.
Since the null hypothesis of no cointegration cannot be rejected for India, Pakistan
and Bangladesh we test for their correlation in their return form. In order to get the
rough idea of the correlation between two variables we calculate the Pearson
correlation between returns on exchange rates and stock prices for the sample of
countries under analysis. Our results presented in Table V indicate the negative
correlation for all countries.
As it is evident from Table V, among all the countries, the underlying variables
are highly negatively related for Indian market followed by Malaysia and Philippines.
The least negative correlation is observed for Bangladesh. However, given the market
heterogeneity in the financial markets and the limitation of studying the correlation
within the time domain framework (see Section 2), we decompose returns on exchange

Country Correlation

India 0.466
Pakistan 0.221
Bangladesh 0.112
Sri Lanka 0.167
Malaysia 0.427 Table V.
Indonesia 0.123 Coefficients of Persons
Philippines 0.411 correlation for sample
Thailand 0.174 countries
SAJGBR rates and stock prices into four different time scales (D1, D2, D3 and D4) using the
3,2 methodology of wavelets. The time dynamics of each time scale is 2-4, 4-8, 8-16 and
16-32 months, respectively. Especially, we use MODWT to decompose the returns
on exchange rates and stock prices because MODWT comes with several advantages.
For example, contrary to DWT, the MODWT does not decimate the coefficients
hence the number of scaling and wavelet coefficients at every level of transform is
218 the same as number of sample observations. Our results (Figure 1) demonstrate that
the relationship between stock indices and exchanges rates is heterogeneous
across different time scales. We find weak relationship at lower time scales (higher
frequencies) but stronger relationship at higher time scales (lower frequencies).
In particular, we find the negative relationship grows as the scale increase from D1 to
D4 for all countries implies that the consistent gain of stock markets in the long run
will essentially contribute toward the appreciation of currencies. Our results provide
ample evidence at all (frequencies) that increase in stock prices have negative impact
on exchange rates and hence follow a portfolio balance approach of Branson and
Henderson (1985). However, the support for the negative relationship is more evident
at lower frequencies. In other words the portfolio balance approach is more apparent
in the long run.
Next we proceed to quantile regression estimates for several reasons as discussed
by Koenker and Bassett (1978), Buchinsky (1998) and Hao and Naiman (2007). But
before running the proposed model, we use simple regression to estimate the relationship
between returns on exchange rates following Aggarwal (1981) and Bahmani-Oskooee
and Sohrabian (1992). Table VI confirms the negative relationship between exchange
rates and stock prices. This means the increase (decrease) in returns of stock prices
will decrease (increase) the exchange rate causing the domestic currency to appreciate
(depreciate) and hence support the portfolio balance effect. However, the traditional
OLS approach only provides the estimation of median (0.5th quantile) function. We
therefore re-estimate the relationship between exchange rates and stock prices using the
quantile regression technique. Recently using quantile regression Tsai (2012) showed an
interesting pattern in the relation of these two markets in Asia. Our results are shown in
Tables VI and VII. Results based on quantile regression show some asymmetries in the
relationship between exchange rates and stock prices. By and large, all eight countries
have similar pattern in the coefficients obtained by different quantile functions. The
coefficients are more inclined to be negative when exchange rates are extremely high and
hence support the findings drawn by Tsai (2012) who observed that the negative relation
between stock and foreign exchange markets is more obvious when exchange rates are
extremely high or extremely low. Our estimated results demonstrate that the coefficients
using data of India and Pakistan are observed to be significant for all quantiles.
Further the estimated coefficients using the slope equality Wald test at both lower and
higher quantiles are found significantly different from the median coefficient for India
(Table VII) unlike Pakistan and Bangladesh where the significance is associated merely
for higher quantiles. On the other hand, the findings from Sri Lankan data set suggest
no significant difference between the results of OLS and quantile coefficients. Further, the
estimated coefficients for countries like Malaysia, Indonesia, Philippines and Thailand
are generally high at higher quantiles and even significant from median coefficients.
Hence, it can be concluded that the increase in stock markets will lead the appreciation of
a domestic currency based on its level of strength. For example, when the domestic
currency is least strong (excessive weak) the increase in stock markets will contribute
little toward its appreciation but when it is strong (less weak), the appreciation of
 INDIA PAKISTAN
Relationship
U 0.4 U between stock

Wavelet Correlation
Wavelet Correlation

U U
0.2 U U
0.2 prices and
0.0 U * *
0.4 * *
exchange rates
* * L U

0.6 L 0.4 L L
L
L
* 219
0.8 *
0.8
L L

1 2 4 8 1 2 4 8
Wavelet Scale Wavelet Scale

BANGLADESH SRI LANKA

U U U
0.2

Wavelet Correlation
Wavelet Correlation

U U
U
U
0.0 0.0
U
* * *
0.2 * * *
L L
L L
0.4 * 0.4 *
L
L
0.6
0.8 L
L
1 2 4 8
1 2 4 8
Wavelet Scale
Wavelet Scale

MALAYSIA INDONESIA
0.4 U 0.4 U
Wavelet Correlation

Wavelet Correlation

0.2 U
0.2 U U
0.0 U 0.0 *
0.2 U *
U * 0.2 L * *
*
0.4 * L
L L 0.4
0.6 * L
L 0.6
0.8 L L

1 2 4 8 1 2 4 8
Wavelet Scale Wavelet Scale

PHILIPPINES THAILAND
0.2 U
Wavelet Correlation

U 0.4
Wavelet Correlation

0.0 0.2 U
U U U
0.0 U
0.2 U *
* *
0.2 L
*
0.4 L *
* * *
0.4 L
0.6 L
L 0.6 L
L
0.8 L

1 2 4 8 1 2 4 8
Wavelet Scale Wavelet Scale

Notes: L refers to the Lower bound of 95 percent confidence Interval and U the Upper
bound of 95 percent confidence Interval. Estimations have been done using R-statistical Figure 1.
package Wavelet correlation
for selected Asian
stock markets
SAJGBR Quantile Coefficient t-statistics Quantile Coefficient t-statistics
3,2
India 0.100 0.155 6.154* Pakistan 0.100 0.019 3.057*
0.200 0.084 4.019* 0.200 0.012 3.173*
0.300 0.069 4.146* 0.300 0.011 2.936*
0.400 0.072 4.139* 0.400 0.009 2.181**
220 0.500 0.062 3.685* 0.500 0.012 2.453*
0.600 0.075 3.996* 0.600 0.027 2.487*
0.700 0.107 4.337* 0.700 0.035 2.576*
0.800 0.131 6.401* 0.800 0.044 2.355*
0.900 0.228 4.960* 0.900 0.094 6.321*
OLS Mean 0.131 7.633* Mean 0.113 5.98*
Bangladesh 0.100 0.001 0.030 Sri Lanka 0.100 0.052 1.302
0.200 0.006 1.441 0.200 0.020 1.472
0.300 0.002 0.304 0.300 0.007 0.764
0.400 0.000 0.196 0.400 0.008 0.798
0.500 0.006 1.462 0.500 0.013 1.319
0.600 0.009 1.908 0.600 0.025 2.241**
0.700 0.022 3.8697* 0.700 0.034 2.859*
0.800 0.026 5.101 0.800 0.019 1.290
0.900 0.037 4.816* 0.900 0.005 0.373
OLS Mean 0.016 1.158 Mean 0.033 2.460*
Malaysia 0.100 0.142 3.696* Indonesia 0.100 0.161 2.856*
0.200 0.093 3.403* 0.200 0.159 2.993*
0.300 0.050 2.009** 0.300 0.179 4.113*
0.400 0.018 2.380* 0.400 0.212 4.646*
0.500 0.003 0.633 0.500 0.214 4.643*
0.600 0.011 1.969* 0.600 0.211 4.637*
0.700 0.033 1.735 0.700 0.257 8.414*
0.800 0.086 4.383* 0.800 0.265 9.525*
0.900 0.159 4.274* 0.900 0.341 4.524*
OLS Mean 0.151 6.858* Mean 0.129 1.802
Philippines 0.100 0.115 3.913* Thailand 0.100 0.115 3.913*
0.200 0.135 4.296* 0.200 0.135 4.296*
0.300 0.128 4.646* 0.300 0.128 4.646*
0.400 0.140 6.980* 0.400 0.140 6.980*
0.500 0.134 6.930* 0.500 0.134 6.930*
0.600 0.123 6.952* 0.600 0.123 6.952*
0.700 0.132 7.347* 0.700 0.132 7.347*
0.800 0.133 5.488* 0.800 0.133 5.488*
0.900 0.181 4.891* 0.900 0.181 4.891*
Table VI. OLS Mean 0.139 6.542* Mean 0.060 2.574*
Estimates of quantile
regression coefficients Note: *,**Statistical significance at 1 and 5 percent level, respectively

currency will be contributed significantly by stock markets increase. The premise is true
for all countries except for the Sri Lanka.
Overall, our results support the portfolio balance approach, which argues for the
negative relationship between stock prices and exchange rates. It may be argued
that the results may get distorted by the effects of 1997 currency crisis. Nevertheless,
the two methodologies are robust to the outliers. The results are also robust to the two
alternative methodologies employed.
Quantiles Variable Restr. value SE Prob.
Relationship
between stock
0.1, 0.5 India 0.092295 0.026210 0.0002 prices and
0.5, 0.9 0.165838 0.043910 0.0001
0.1, 0.5 Pakistan 0.007143 0.006863 0.2980 exchange rates
0.5, 0.9 0.081863 0.014085 0.0000
0.1, 0.5 Bangladesh 0.004575 0.021500 0.8315 221
0.5, 0.9 0.016386 0.004588 0.0004
0.1, 0.5 Sri Lanka 0.038861 0.038482 0.3126
0.5, 0.9 0.008498 0.013335 0.5240
0.1, 0.5 Malaysia 0.138436 0.037179 0.0002
0.5, 0.9 0.156270 0.036055 0.0000
0.1, 0.5 Indonesia 0.053386 0.059907 0.3728
0.5, 0.9 0.126262 0.074439 0.0499
0.1, 0.5 Philippines 0.018884 0.029282 0.5190
0.5, 0.9 0.047406 0.035904 0.0867
0.1, 0.5 Thailand 0.022688 0.025024 0.3646
0.5, 0.9 0.019969 0.022804 0.3812
Table VII.
Notes: Restriction detail: b(th)b(tk) 0. The significance of coefficients from median quantile has Estimates of slope
been carried out by slope equality Wald test equality test

4. Conclusion
We provided two alternative and more relevant assessments of the relationships
between exchange rates and stock prices for selected Asian countries. In the beginning
of this paper, we decomposed the exchange rates and stock prices into different time
scales by using wavelet methodology in accordance with the market heterogeneity
theory. We then calculated the correlation between exchange rate and stock prices at
different time scales. Our results showed that the relationship is negative in accordance
with the portfolio balance approach and this relationship is heterogeneous across
different time scales. In particular, we observed that when periods of longer than
one year are considered, the existence of intense negative relationship between
stock markets and exchange rates cannot be ruled out. Next, we proceeded with the
estimation of relationship using OLS. It was found that the relationship is significantly
negative for all the countries. However, the traditional OLS approach only provided the
estimation of median (0.5th quantile) function. We therefore estimated the relationship
at different quantiles of exchange rates by using quantile regression approach. The
findings from quantile regressions estimates provided a comprehensive representation
of the relationship between exchange rate and stock indices, which according to it was
found to be more obvious at higher quantiles for most of countries. These results are
the unique contribution of this study. We argue that the relationship between exchange
rate and stock prices are rather heterogeneous across different time scales and quantiles
and emphasize that the methodology applied can be a valuable tool for obtaining
additional insights of the relationship between exchange rates and stock prices. Our
results therefore confirm the portfolio balance approach using the two alternative
methodologies.
The main implications are that, governments in these countries should frame
exchange rate polices while considering developments in their stock markets.
The significant effect of stock prices on exchange rates could be attributed to the
slowdown of an economy which affects stock prices. This prompts withdrawal of capital
SAJGBR by international investors to induce downward pressure on the currency. The stimulation
3,2 of economic growth and stock markets could serve a useful vehicle to attract capital
flows. South Asian countries have often been the victims of sudden capital withdrawals
and the attendant currency crisis. The development of stock markets in these countries
by respective governments should therefore be seen as an opportunity to restrict the
currency crises.
222 While this note applies a bivariate approach to test the relationship between returns
on exchange rates and stock prices, multivariate approach may provide new insights
about this relationship. Inclusion of variables like oil prices, interest rates and money
supply in multivariate context may provide useful results of this relationship. Further
research could also be extended to test the two market segments for predicting each other.
Notes
1. Frequency is the rate of change with respect to time. Change in a short span of time means
high frequency. Change overlong span of time means low frequency. If a signal does not change
at all, its frequency is zero. If a signal changes instantaneously, its frequency is infinite.
2. Bloomberg codes: SENSEX:IND for India, KSE-100:IND for Pakistan, DHAKA:IND for
Bangladesh, CSEALL:IND for Sri Lanka, FBMKLCI:IND for Malaysia, JCI:IND for Indonesia,
PASHR:IND for Philippines and SET:IND for Thailand.

References
Abdalla, I.S.A. and Murinde, V. (1997), Exchange rate and stock price interactions in emerging
financial markets: evidence on India, Korea, Pakistan and the Philippines, Applied
Financial Economics, Vol. 7 No. 1, pp. 25-35.
Aggarwal, R. (1981), Exchange rates and stock prices: a study of the United States capital
markets under floating exchange rates, Akron Business and Economic Review, Vol. 12
No. 1, pp. 7-12.
Akpan, P.L. (2008), Foreign exchange market and economic growth in an emerging petroleum
based economy: evidence from Nigeria (1970-2003), African Economic and Business
Review, Vol. 6 No. 2, pp. 46-58.
Bahmani-Oskooee, M. and Sohrabian, A. (1992), Stock prices and the effective exchange rate of
the dollar, Applied Economics, Vol. 24 No. 4, pp. 459-464.
Bassett, G. and Chen, H.L. (2001), Quantile style: return-based attribution using regression
quantiles, Empirical Economics, Vol. 26 No. 1, pp. 293-305.
Benhmad, F. (2012), Modeling nonliner Granger causality between the oil price and US dollar:
a wavelet based approach, Economic Modelling, Vol. 29 No. 1, pp. 1505-1514.
Branson, W.H. and Henderson, D.W. (1985), The specification and influence of asset markets,
Handbook of International Economics, Vol. 2 No. 1, pp. 749-805.
Buchinsky, M. (1998), Recent advances in quantile regression models: a practical guideline for
empirical research, Journal of Human Resources, Vol. 33 No. 1, pp. 88-126.
Buchinsky, M. (2001), Quantile regression with sample selection: estimating womens return to
education in the US, Empirical Economics, Vol. 26 No. 1, pp. 87-113.
Bussiere, M., Chiaie, D.S. and Peltonen, T.A. (2014), Exchange rate pass-through in the global
economy: the role of emerging market economies, IMF Economic Review, Vol. 62 No. 1,
pp. 146-178.
Choong, C.K., Baharumshah, A.Z., Yusop, Z. and Habibullah, M.S. (2010), Private capital flows,
stock market and economic growth in developed and developing countries: a comparative
analysis, Japan and World Economy, Vol. 22 No. 2, pp. 107-117.
Cumming, D., Johan, S. and Li, D. (2011), Exchange trading rules and stock market liquidity, Relationship
Journal of Financial Economics, Vol. 99 No. 1, pp. 651-671.
between stock
Dacorogna, M., Gencay, R., Muller, U., Olsen, R. and Pictet, V. (2001), An Introduction to
High-Frequency Finance, Academic Press, San Diego, CA. prices and
Dar, A.B., Samantaraya, A. and Shah, F.A. (2014), The predictive power of yield spread: evidence exchange rates
from wavelet analysis, Empirical Economics, Vol. 46 No. 1, pp. 887-901.
Deaton, A. (1997), The Analysis of Household Surveys, Johns Hopkins, Baltimore, MD. 223
Dornbusch, R. and Fischer, S. (1980), Exchange rates and the current account, American
Economic Review, Vol. 70 No. 5, pp. 960-971.
Engle, R.F. and Granger, C.W.J. (1987), Co-integration and error correction: representation,
estimation and testing, Econometrica, Vol. 55 No. 2, pp. 251-276.
Genacay, R., Selcuk, F. and Whitcher, B. (2002), An Introduction to Wavelets and Other Filtering
Methods in Finance and Economics, Academic Press, New York, NY.
Gregory, A.W. and Hansen, B.E. (1996), Residual-based tests for cointegration in models with
regime shifts, Journal of Econometrics, Vol. 70 No. 1, pp. 99-126.
Griffin, J.M. and Stulz, R.M. (2001), International competition and exchange rate shocks:
a cross-country industry analysis of stock returns, The Review of Financial Studies,
Vol. 14 No. 1, pp. 215-241.
Hacker, R.S., Karlsson, H.K. and Mansson, K. (2012), The relationship between exchange rates
and interest rate differentials: a wavelet approach, The World Economy, Vol. 35 No. 9,
pp. 1162-1185.
Hao, L. and Naiman, D.Q. (2007), Quantile Regression, Sage Publications, Thousand Oaks, CA.
Hau, H. and Rey, H. (2006), Exchange rates, equity prices and capital flows, Review of Financial
Studies, Review Of Financial Studies, Vol. 19 No. 1, pp. 273-317.
He, J. and Ng, L.K. (1998), Foreign exchange exposure of Japanese multinational corporations,
Journal of Finance, Vol. 53 No. 1, pp. 733-753.
Ibrahim, H. and Aziz, H. (2003), Macroeconomic variables and the Malaysian equity market:
a view through rolling subsamples, Journal of Economic Studies, Vol. 30 No. 1, pp. 6-27.
Kim, K. (2003), Dollar exchange rate and stock price: evidence from multivariate cointegration
and error correction model, Review of Financial Economics, Vol. 12 No. 1, pp. 301-313.
Koenker, R. and Bassett, G. (1978), Regression quantiles, Econometrica, Vol. 46 No. 1, pp. 33-50.
MacDonald, R. and Taylor, M. (1992), Exchange rate economics: a survey, IMF Staff Papers,
Vol. 39 No. 1, pp. 1-57.
Ma, C.K. and Kao, G.W. (1990), On exchange rate changes and stock price reactions, Journal of
Business Finance and Accounting, Vol. 17 No. 3, pp. 441-449.
Mishra, A.K. (2004), Stock market and foreign exchange market in India: are they related,
South Asia Economic Journal, Vol. 5 No. 2, pp. 209-232.
Muller, U., Dacorogna, M., Dav, R., Olsen, R., Pictet, O. and Weizsacker, J. (1997), Volatilities
of different time resolutions analyzing the dynamics of market components, Journal of
Empirical Finance, Vol. 1 No. 4, pp. 213-239.
Nieh, C.C. and Leeb Chung, C.F. (2001), Dynamic relationship between stock prices and
exchange rates for G-7 countries, The Quarterly Review of Economics & Finance, Vol. 41
No. 4, pp. 477-490.
Ramsey, J.B. (2002), Wavelets in economics and finance: past and future, Studies in Nonlinear
Dynamics and Econometrics, Vol. 6 No. 3, pp. 1-27.
Ranta, M. (2013), Contagion among major world markets: a wavelet approach, International
Journal of Managerial Finance, Vol. 9 No. 2, pp. 133-149.
SAJGBR Smyth, R. and Nandha, M. (2003), Bivariate causality between exchange rates and stock prices
in South Asia, Applied. Economics. Letters, Vol. 10 No. 1, pp. 699-704.
3,2
Tian, G.G. and Shiguang, M. (2010), The relationship between stock returns and the foreign
exchange rate: the ARDL approach, Journal of the Asia Pacific Economy, Vol. 15 No. 4,
pp. 490-508.
Tiwari, A.K., Dar, A.B. and Bhanja, N. (2013), Oil price and exchange rates: a wavelet based
224 analysis for India, Economic Modelling, Vol. 31 No. 1, pp. 414-422.
Tiwari, A.K., Dar, A.B. and Islam, F. (2014), Time-frequency relationship between share prices
and exchange rates in India: evidence from continuous wavelets, Empirical Economics,
doi:10.1007/s00181-014-0800-3, available at: http://link.springer.com/article/10.1007/
s00181-014-0800-3#page-1
Tsagkanos, A. and Siriopoulos, C. (2013), A long-run relationship between stock price
index and exchange rate: a structural nonparametric co-integrating regression approach,
Journal of International Financial Markets, Institutions and Money, Vol. 25 No. 1,
pp. 106-118.
Tsai, C.I. (2012), The relationship between stock price index and exchange rate in Asian
markets: a quantile regression approach, International Financial Markets, Institutions
and Money, Vol. 22 No. 1, pp. 609-621.

Further reading
Fernandez, R. (1997), Returns to regionalism: an evaluation of non traditional gains from regional
trade agreements, World Bank Policy Research, Working Paper No. 1816, Department of
Economics, New York University.
Jiranyakul, K. (2012), Linkages between Thai stock and foreign exchange markets under the
floating regime, Journal of Financial Economic Policy, Vol. 4 No. 4, pp. 305-319.
Ozair, A. (2006), Causality between stock prices and exchange rates: a case of the United States,
Master of science thesis, Florida Atlantic University, Boca Raton, FL.
Soenen, L.A. and Hennigar, E.S. (1988), An analysis of exchange rates and stock prices the US
experience between 1980 and 1986, Akron Business and Economic. Review, Vol. 19 No. 1,
pp. 7-16.
Solnik, B. (1984), Why not diversity internationally rather than domestically, Financial Analysts
Journal, Vol. 30 No. 1, pp. 48-54.

About the authors

Arif Billah Dar presently works as an Assistant Professor at the IMT, Ghaziabad, India. He has
to his credit number of research papers published in reputed journals of Springer and Elsevier.
Arif Billah Dar is the corresponding author and can be contacted at: billaharif0@gmail.com
Aasif Shah is a Research Scholar at the Pondicherry Central University, India. He has
published number of papers in international journals.
Dr Niyati Bhanja works as an Assistant Professor at the University of Petroleum and Energy
Studies, Dehradun, India. Her publications have appeared in number of reputed international
journals.
Dr Amaresh Samantaraya is an Associate Professor at the Pondicherry Central University,
India. Previously he worked as an Economist with the Reserve Bank of India.

To purchase reprints of this article please e-mail: reprints@emeraldinsight.com

Or visit our web site for further details: www.emeraldinsight.com/reprints
Reproduced with permission of the copyright owner. Further reproduction prohibited without
permission.