Post on 14-Oct-2020
transcript
0
qwertyuiopasdfghjklzxcvbnmqwerty
uiopasdfghjklzxcvbnmqwertyuiopasd
fghjklzxcvbnmqwertyuiopasdfghjklzx
cvbnmqwertyuiopasdfghjklzxcvbnmq
wertyuiopasdfghjklzxcvbnmqwertyui
opasdfghjklzxcvbnmqwertyuiopasdfg
hjklzxcvbnmqwertyuiopasdfghjklzxc
vbnmqwertyuiopasdfghjklzxcvbnmq
wertyuiopasdfghjklzxcvbnmqwertyui
opasdfghjklzxcvbnmqwertyuiopasdfg
hjklzxcvbnmqwertyuiopasdfghjklzxc
vbnmqwertyuiopasdfghjklzxcvbnmq
wertyuiopasdfghjklzxcvbnmqwertyui
opasdfghjklzxcvbnmqwertyuiopasdfg
hjklzxcvbnmrtyuiopasdfghjklzxcvbn
mqwertyuiopasdfghjklzxcvbnmqwert
yuiopasdfghjklzxcvbnmqwertyuiopas
dfghjklzxcvbnmqwertyuiopasdfghjklz
INTERACTION BETWEEN STOCK
PRICES AND
EXCHANGE RATES
Author : Ankit Mital
Student ID: 0950302
Dissertation Supervisor : Alessandro Palandri
Date: August 2010
ABSTRACT
This paper examines the dynamics between the stock prices and
exchange rates for a set of four emerging economies and four advanced
economies using Cointegration & Granger Causality techniques. The
results show feedback relations for India, Brazil, South Korea & Japan.
On the other hand we find unidirectional relationship for the
Philippines, Australia & Canada with stock returns causing exchange
returns. While for the United Kingdom, we fail to detect any significant
relationship.
‘All the work contained within is my own unaided effort and
Conforms with the University’s guidelines on plagiarism.’
‘No substantial part(s) of the work submitted here has also been submitted by me in other
assessments for accredited courses of study, and I acknowledge that if this has been done an
appropriate reduction in the mark I might otherwise have received will be made.’
Word Count : 7252
To,
My Parents
Acknowledgements
I would like to thank my dissertation supervisor, Dr. Alessandro Palandri and our academic director, Dr. Richard Payne for their invaluable advice and guidance. I am grateful to my brother Abhinav Mital and friends Priyanka Khosla, Osman Ansari, Sibtain Masood, Kartik Pental and Hammad Pothiawala for their ready support and giving this paper a patient reading, pointing out its numerous errors.
CONTENTS
SECTION I............................................................................................................................................. 1
INTRODUCTION .............................................................................................................................. 1
SECTION II ........................................................................................................................................... 5
LITERATURE REVIEW ................................................................................................................... 5
SECTION III .......................................................................................................................................... 9
METHODOLOGY ............................................................................................................................. 9
SECTION IV ........................................................................................................................................ 12
DATA ................................................................................................................................................ 12
SECTION V ......................................................................................................................................... 14
EMPIRICAL PROCEDURE AND EVIDENCE ........................................................................... 14
STATIONARITY ......................................................................................................................... 14
COINTEGRATION ..................................................................................................................... 15
GRANGER CAUSALITY ........................................................................................................... 19
VECTOR ERROR CORRECTION MODEL ........................................................................... 29
RECESSIONARY IMPACT ....................................................................................................... 31
SECTION VI ........................................................................................................................................ 34
CONCLUSION ................................................................................................................................ 34
BIBLOGRAPHY ................................................................................................................................. 37
APPENDIX .......................................................................................................................................... 40
1
SECTION I
INTRODUCTION
The first signs of the financial crisis of 2007 came forth in the Equities market
in July 2007 (Mark Taylor & Michael Melvin, 2009). As the Foreign Exchange
market participants watched these developments anxiously, their worst fears
were met on August 16th
2007, when major unwinding of the carry trade
occurred. This was quite opposite to the sequence of events of the Asian
Financial Crisis of 1997 where it was the foreign exchange market that
triggered a tsunami in the equities market. It is believed that the sharp
depreciation of the Thai Baht triggered depreciation of other currencies in the
neighborhood, which eventually led to the collapse of the stock markets as
well. The different experience of the crises throws up a puzzle which needs to
be solved.
This study is undertaken to analyze the dynamics between stock prices and
exchange rates. This link can be exploited to predict the path of the exchange
rates. This will also assist in the management of corporation’s exposure to
foreign exchange rate risk of their earnings. More importantly for Investors,
since currency assets are an important part of investment funds, the
knowledge of the link between currency and other assets in the portfolio is
vital. The widely employed Mean-Variance approach to portfolio analysis
2
suggests that expected utility is maximized by the minimization of the variance
of the portfolio. For the calculation of this variance, the estimation of the
correlation between different assets of the portfolio is essential. Is the
magnitude of correlation different when stock prices are trigger variable or
when the exchange rates are the trigger variable? (Dimitrova, 2005). Also,
stock markets have been found to impact aggregate demand through wealth
and liquidity effects, influencing money demand and exchange rates (Gavin,
1989). Thus while formulating monetary policy or undertaking foreign
exchange intervention, its impact on the stock market needs to be taken into
consideration.
There is no theoretical consensus on the relationship between stock prices and
exchange rates. Basically there exist two hypotheses supporting the causal
relationship between the stock prices and exchange rates. The traditional
Goods Market Approach (or Flow Oriented Models) by Dornbush & Fisher
(1980) and Solnik (1987) suggests that exchange rates affect the value of a firm
via affecting the competitiveness of a firm as it affects the value of earnings
made in foreign currencies and the cost of funds borrowed from abroad. For
instance, a depreciation of the local currency makes the export goods more
attractive, leading to an increase in demand for the local countries’ produce in
the international market and hence the revenue. This leads to an appreciation
3
in the value and hence the stock price of the firm. On the other hand, an
appreciation in the local currency reduces the value of the local firm via
making its exports less competitive and thus causes a fall in its stock price.
However, the ultimate impact upon the stock market would depend upon the
importance of external trade in the economy and the degree of trade
imbalance. A depreciation of the currency, while making exports more
competitive, also makes imports more expensive, thereby it increases the cost
of production.
Alternatively, the Portfolio Balance Approach (or Stock Oriented Models) by
Frankel (1993), focuses on the role of capital account transactions. This
approach stresses the fact that the price of exchange rates is determined by
market mechanisms, supply and demand. According to this theory, a change in
stock prices leads to portfolio adjustments. A booming stock market attracts
Foreign Capital, leading to an increase in demand for the local currency,
leading to its appreciation. A bearish stock market on the other hand,
witnesses investors selling off their positions in the market to avoid further
losses. This would lead to an increase in the supply of the local currency while
increasing the demand for the foreign currency, leading to a depreciation of
the local currency. Moreover, movements in stock prices may influence
exchange rates and money demand because the investor’s wealth and liquidity
4
demand could depend on the performance of the stock market (Gopalan Kutty,
2010). Therefore according to this theory, a rise in the stock market leads to an
appreciation of the exchange rates, while a fall in the stock market leads to
depreciation of the currency.
As we can see, according to the two theories which offer completely opposite
views on both the direction and the lead lag relationships between the
variables, difference forces are at work in determining this relationship. The
traditional goods market approach focusing on the impact of exchange rates
on the competitiveness of exports, predicts that exchange rates lead the
relationship, with depreciation in the exchange rates leading to an
appreciation in the stock prices. On the other hand, the Portfolio balance
approach stressing on the impact of global equity flows on exchange rates
proposes that it’s the stock markets which lead the relationship, with an
appreciation in the stock prices leading to an appreciation in the exchange
rate. Ultimately, this relationship would depend upon the relative strengths of
the aforementioned forces. If the external trade of a country is able to exert a
higher influence on its economy than its capital flows, the Goods Market
approach will prevail. However, if the opposite is the case, the Portfolio
balance approach will prevail. There might even be a case of the two forces
5
cancelling each other out which would result in the absence of any clear
pattern emerging out of the relationship.
To examine this relationship, we employ cointegration tests (after establishing
non-stationarity) followed by Granger Causality tests (1969) or Vector Error
correction Models (VECM), depending upon their applicability to the data.
SECTION II
LITERATURE REVIEW
A number of studies have been undertaken over the decades to analyze the
relationship between stock prices and exchange rates. However, there is a
general lack of consensus on this relationship and the existing literature offers
conflicting evidence on the direction and even the existence of relationship.
One of the earliest notable studies to examine the relationship was undertaken
by Franck and Young (1972) who failed to establish any significant interaction
between the variables. Aggarwal (1981) found a positive correlation between
exchange rates of US dollar and changes in the indices of US stock prices.
Solnik (1987) tried to study the impact of several economic variables on stock
prices, including exchange rates for eight advanced economies. He however
failed to detect any significant impact of exchange rates over stock prices.
6
Ma & Kao (1990) found evidence of significant positive relationship between
the two and supported the traditional Goods Market approach for six major
advanced industrial economies.
Using a Multi County approach, Smith (1992) employing linear regression
techniques examined the relationship in two separate papers, once for US,
Japan and Germany and once for the UK. In all cases, he found that equity
values had a significant effect on the value of exchange rates.
When the above mentioned studies were conducted, techniques of
Cointegration to study relationship between variables had not been developed
adequately and thus were not employed in the analysis. Also, they worked
with either monthly or quarterly data.
Oskooee & Sohrabian (1992) employed cointegration and Granger Causality to
study the relationship between S&P 500 and US Dollar. While failing to
discover any significant long run relationship between the two, they
discovered bidirectional causality in the short run.
Ajayi & Mougoue (1996), using cointegration techniques were able to establish
a significant relationship short run and long run relationship for eight advanced
economies. However, the variables interacted differently over the periods,
7
displaying a negative relationship in the short run, while a positive one in the
long run.
Using similar techniques Abdalla & Murinde (1997) investigated the causal
linkages between leading prices of foreign exchange market and stock markets
for India, Korea, Pakistan and the Philippines. They found unidirectional
causality from exchange rates to stock prices for all the countries except for
the Philippines.
Ajayi, Friedman & Mehdian (1998) studied seven advanced market economies
and eight emerging market economies to explore the relationship between
exchange rates and stock returns. They found Granger Causality between the
stock and currency markets in all the advanced economies, but found no
consistent causal relationship in emerging markets.
Granger, Huang and Yang (2000) tried to determine the appropriate Granger
Causality relation between stock prices and exchange for nine East Asian
economies, with a special focus on the relationship during the Asian Financial
crisis. They were able to establish a significant relationship for all cases, except
two. However, the direction and lead lag relationships were found to be
different for different countries.
8
Nieh & Lee (2001), using cointegration tests did not find any long run
relationship between the stock prices and exchange rates for the G-7
countries. However, they did find bidirectional relationship between the two
variables for the sample countries.
Employing similar methodology, several studies have been undertaken since
for individual countries, like Mansor h. Ibrahim (2000) for Malaysia, Ying Wu
(2000) for Singapore, Nath & Samanta (2003) for India, Aydemir & Demirhan
(2009) for Turkey and Kutty (2010) for Mexico. Mansor, while failing to
establish any long term relationship in the bivariate model, found a long term
relationship when model was extended to include Broad Money and reserves.
He also found a significant short run impact of the macroeconomic variables on
the stock prices. Wu, although establishing a significant long term relationship,
discovered different directions in the relationship for different periods. Nath
and Samanta did not find any strong causal influence from stock market return
to foreign exchange market return. Aydemir & Demirhan found bidirectional
causality between exchange rate and stock market indices. Kutty, although did
not discover any long term relationship for the two markets, found a causal
relationship, with stock prices leading exchange rates.
9
Uddin & Rahman (2009) undertook a similar study for India, Bangladesh and
Pakistan. But in all the three cases, they found neither a long term, nor a short
term relationship between stock prices and exchange rates.
In general, there is very weak evidence for any long term relationship between
stock prices and exchange rates. However, most of the studies did find
evidence for short run relationship between the two, though there is a lack of
consistency in the direction of causality between the two.
Significant work has been done on the theory and application of Granger
Causality tests by Granger (1969), Gupta (1987), MacDonald & Kearney (1987),
Shoesmith (1992) and Reimers & Helmut (1992), which shall be useful in the
application of appropriate tests.
SECTION III
METHODOLOGY
Our objective is to establish whether stock prices causally affect exchange
rates or the other way round. We use Cointegration, Vector Error Correction
Models (VECM) and Granger Causality to test the relationship between the
Stock Prices and Exchange Rates in a bivariate model. First we check the data
for stationarity using Augment Dickey Fuller (ADF) and Phillips-Perron (PP)
Tests. Optimal lags shall be calculated using Schwarz Information Criterion
10
(SIC). If the series are found to be stationary, we will straight away apply the
Granger Causality test. But existing financial research indicates that the spot
prices for the two variables are unlikely to be displaying stationarity and points
towards the presence of unit root. Thus if ADF and PP test find the data to be
non-stationary, we will proceed to testing for cointegration after determining
the number of roots. Determining cointegrating relationship is important to
establish the existence of long run equilibrium relationship. Even though
majority of the existing literature points against the existence of any long term
relationship, some researchers, namely Ajayi & Mougoue (1996) and Mansor
(2000) did find some evidence in favour of a long term relationship. To test for
cointegration we shall employ the Johansen approach. If the series display
cointegration, then an error correction term is required and hence we will
apply VECM to find the causality between the two variables as it can capture
both long term and short term relationships. To this end, we shall estimate the
following equations.
∆�� = �� + Σ �� Δ ���� + Σ �� Δ ���� + ������ + ���
Δ�� = �� + Σ ��� Δ ���� + Σ ��� Δ���� + ������ + ���
Where ���� and ���� are error correction terms, which capture previous
periods disequilibrium between stock prices and exchange rates. The
11
terms, ��� and ��� are stationary random processes included to capture the
information not contained in the lagged terms.
If however, there is no evidence of cointegration, which past works indicate
has been the case for majority of the countries, we shall apply Granger
Causality test to establish the causal relationship. But, before applying the
Granger Causality test, we need to difference the series since they would be
displaying non-stationarity and differencing would induce stationarity into the
data. Failure to do this would lead to spurious regression. Thus we shall
estimate the following equations.
Δ�� = �� + Σ ��Δ���� + Σ ��Δ���� + ��
Δ�� = �� + Σ ���Δ���� + Σ ���Δ���� + ��
The optimal lag length shall be selected by choosing the lag that yields the
lowest SIC.
Then, we perform robustness checks by including more variables in the
specification and work in a multivariate model to see whether they lead to any
significant changes in the results previously obtained. The variables which we
will include to test the robustness are: the US stock market and rate
differential with the US overnight call rate. These variables have been chosen
upon their ability to impact both the stock markets and the exchange rates.
12
SECTION IV
DATA
To undertake our study from the start of the 1999 financial year to the end of
the 2009 financial year, that is from 31st
March 1999 to 31st
March 2010. Our
sample consists of a total of eight countries, four emerging economies and four
advanced economies. The four emerging economies being studied have been
chosen for them having a floating exchange rate and their importance in the
global economy. This obviously leaves out China since the value of its currency
Renminbi is unofficially pegged to the USD. We undertake our analysis for
India, Brazil, the Philippines and South Korea. The advanced economies under
consideration are also the major currency economies, namely, Australia,
Canada, Japan and the United Kingdom. The advanced economy, conspicuous
by its absence is the US. It was dropped since we are using its currency as the
reference currency, i.e. the currency in terms of which all others are being
expressed and also that its stock market and its overnight call rate is being
used as an exogenous variable when we check for the robustness of our
model. Also, none of the advanced economies from the Euro zone have been
included for the obvious reason of them using a shared common currency. Our
analysis starts from 1999, since this was the year when Brazil deregulated its
currency, Brazilian Real. Also the global economy had finally recovered from
the Asian Financial Crisis by this time.
13
We required daily spot rates for exchange rates, stock prices and overnight call
rates for the period under consideration for the relevant countries. The
exchange rates being considered are Rupee (India), the Real (Brazil), Won
(South Korea), Peso (Philippines), AUD (Australia), CAD (Canada), Yen (Japan)
and GBP (United Kingdom). All the currencies are expressed in USD. The stock
markets being studied are BSE 100 (India), Bovespa (Brazil), KOSPI 200 (South
Korea), PSEi (Philippines), S&P/ASX 200 (Australia), S&P/TSX Composite Index
(Canada), NIKKEI 225 (Japan) and FTSE 100 (United Kingdom). It is important to
mention here that the pound is the only currency which instead of being listed
as pounds per dollar, is listed as dollars per pound. Thus to maintain
consistency with other currencies under study, we take inverse of the listed
price to get pounds per dollar. Lastly we use the S&P 500 Composite and the
US Call Money middle rate as exogenous variable to check the robustness of
our model. All the data was readily available and was obtained from
Datastream.
A small note on the interpretation of exchange rate is worth mentioning. An
appreciation of the currency implies a reduction in the exchange rate and
depreciation in the currency implies an increase in the exchange rate.
14
SECTION V
EMPIRICAL PROCEDURE AND EVIDENCE
STATIONARITY
We start our analysis by checking for stationarity and determining the order of
integration. To this end, we employ Augmented Dickey-Fuller (ADF) test and
Phillips-Perron (PP) test. The optimal lag length is determined using Schwarz
Information Criterion (SIC).
The results obtained from stationarity tests are provided in table 1. The results
indicate the presence of non stationarity and the integration of the order of
one (i.e. I(1)) in all cases for both stock prices and exchange rates. The t-
statistics provided in the table indicates towards the failure to reject the null
hypothesis of the presence of unit for both stock prices and exchange rates.
Then we take the first difference of both the variables and check for
stationarity. In this case, we do reject the null hypothesis of the presence of
unit root. This indicates the presence of integration of the order of one.
With the unit root property established, we shall proceed towards testing for
cointegration.
15
Table 1, Unit Root Test Results (1999-2010)
INDIA BRAZIL SOUTH KOREA PHILIPPINES
Variable ADF PP ADF PP ADF PP ADF PP
S -0.38112 -0.33865 -0.0504 0.162903 -1.00659 -0.95915 -0.77231 -0.58529
X -1.81855 -1.89331 -1.48704 -1.39886 -1.6169 -1.8962 -1.92259 -1.91353
∆S -49.3132 -49.2562 -55.7237 -56.0428 -53.1706 -53.2067 -48.1056 -48.0749
∆X -53.8788 -53.9251 -48.8105 -48.6496 -13.9939 -41.8738 -52.6631 -52.6828
AUSTRALIA CANADA JAPAN UNITED KINGDOM
Variable ADF PP ADF PP ADF PP ADF PP
S -1.29333 -1.20457 -1.53271 -1.45107 -1.60513 -1.53505 -1.80706 -1.95537
X -0.87974 -0.79764 -0.69913 -0.72097 -1.59204 -1.47647 -1.47091 -1.40981
∆S -55.9915 -56.2166 -55.0831 -55.214 -54.6449 -54.7419 -35.516 -56.875
∆X -55.1429 -55.2112 -53.566 -53.5724 -56.1819 -56.2353 -50.1101 -50.0086 The critical values are :
1% level = -3.4324, 5% level = -2.8623, 10% level = -2.5672 ∆ = 1
st difference, S = Stock Prices, X = Exchange Rates
COINTEGRATION
Cointegration analysis is used to test whether there exists a long term
relationship between the two variables which contain a unit root. For this, we
apply Johansen approach. The reason for using Johansen approach is that it is
able to accommodate more than two variables, unlike Engle-Granger Two step
method. Though right now we are testing for cointegration between only two
variables, later on, to test the robustness of our results we shall work with a
multivariate model, which would necessitate the use of Johansen approach.
The Johansen approach employs two methods to determine the existence of
cointegrating relationships via the rank of the matrix, namely trace statistic
and maximum eigenvalue statistic. We tested the two series for cointegration
under this approach using different lag lengths, from a lag of one day to five
16
days, i.e., one trading week. The results for the tests based on trace statistic
and maximum eigenvalue statistic are provided in table 2 and table 3
respectively. The results indicate the absence of any cointegrating relationship
between the stock prices and exchange rates for all economies under
consideration, save for the Philippines. In other words, the tests indicate that
there is no long-run equilibrium relationship between stock prices and
exchange rates for all economies except for the Philippines. This is consistent
with existing literature, with majority of previous research providing evidence
against the presence of any long term equilibrium relationship. In the case of
the Philippines, we detect the presence of one cointegrating relationship.
Consequently, we need to include an error correction term while testing for
Granger Causality only in the case of the Philippines and run a Vector Error
Correction Model (VECM). For rest of the cases, where the variables are non-
cointegrated we shall use standard Granger Causality Tests.
17
Table 2, Johansen Cointegration Test Results (Trace statistic) (1999-2010)
INDIA
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
r=0 r>= 0 7.85861 7.5913 7.60785 7.60171 20.2618
r>1 r>=2 1.044673 1.09098 2.07668 2.36960 9.16454
Trace test indicates no cointegration at the 0.05 level
BRAZIL
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
r=0 r>= 0 9.629284 10.0241 10.6006 10.4491 20.2618
r>1 r>=2 3.515322 3.76453 4.47103 4.45527 9.16454
Trace test indicates no cointegration at the 0.05 level
SOUTH
KOREA
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
r=0 r>= 0 4.440368 4.14035 4.07438 4.17672 20.2618
r>1 r>=2 1.790409 1.51235 1.27187 1.25769 9.16454
Trace test indicates no cointegration at the 0.05 level
PHIL.
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
r=0 r>= 0 20.96146 21.8834 22.7168 23.3746 20.2618
r>1 r>=2 4.218562 4.39655 4.62164 4.51412 9.16454
Trace test indicates 1 cointegrating equation(s) at the 0.05 level
AUSTRALIA
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
r=0 r>= 0 5.18152 4.87290 4.78424 4.62984 20.2618
r>1 r>=2 1.38833 1.31234 1.30890 1.32808 9.16454
Trace test indicates no cointegration at the 0.05 level
CANADA
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
r=0 r>= 0 8.618836 8.07410 7.92762 7.74851 20.2618
r>1 r>=2 1.967126 1.92769 1.84467 1.84338 9.16454
Trace test indicates no cointegration at the 0.05 level
JAPAN
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
r=0 r>= 0 5.307801 5.21551 5.23006 5.2284 20.2618
r>1 r>=2 2.196572 2.08611 2.07255 2.07352 9.16454
Trace test indicates no cointegration at the 0.05 level
UNITED
KINGDOM
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
r=0 r>= 0 7.070676 6.77888 6.11631 6.43755 20.2618
r>1 r>=2 2.318772 2.26452 2.03519 1.99598 9.16454
Trace test indicates no cointegration at the 0.05 level
18
Table 3, Johansen Cointegration Test Results (Maximum eigenvalue statistic) (1999-2010)
INDIA
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
r=0 r>= 0 6.81394 6.50031 5.53117 5.23210 15.8921
r>1 r>=2 1.04467 1.09098 2.07668 2.36960 9.16454
Max-eigenvalue test indicates no cointegration at the 0.05 level
BRAZIL
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
r=0 r>= 0 6.11396 6.25958 6.12959 5.99387 15.8921
r>1 r>=2 3.51532 3.76453 4.47103 4.45527 9.16454
Max-eigenvalue test indicates no cointegration at the 0.05 level
SOUTH
KOREA
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
r=0 r>= 0 2.64995 2.628 2.80251 2.91903 15.8921
r>1 r>=2 1.79040 1.51235 1.27187 1.25769 9.16454
Max-eigenvalue test indicates no cointegration at the 0.05 level
PHIL.
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
r=0 r>= 0 17.4868 16.7429 18.0951 18.8605 15.8921
r>1 r>=2 4.39655 4.21856 4.62164 4.51412 9.16454
Max-eigenvalue test indicates 1 cointegrating eqn(s) at the 0.05 level
AUSTRALIA
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
r=0 r>= 0 3.79319 3.56055 3.47533 3.30176 15.8921
r>1 r>=2 1.38833 1.31234 1.30890 1.32808 9.16454
Max-eigenvalue test indicates no cointegration at the 0.05 level
CANADA
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
r=0 r>= 0 6.65171 6.14640 6.08295 5.90513 15.8921
r>1 r>=2 1.96712 1.92769 1.84467 1.84338 9.16454
Max-eigenvalue test indicates no cointegration at the 0.05 level
JAPAN
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
r=0 r>= 0 3.11122 3.12940 3.15750 3.15488 15.8921
r>1 r>=2 2.19657 2.08611 2.07255 2.07352 9.16454
Max-eigenvalue test indicates no cointegration at the 0.05 level
UNITED
KINGDOM
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
r=0 r>= 0 4.75190 4.51436 4.08111 4.44157 15.8921
r>1 r>=2 2.31877 2.26452 2.03519 1.99598 9.16454
Max-eigenvalue test indicates no cointegration at the 0.05 level
19
GRANGER CAUSALITY
Now we run Granger Causality tests for the seven economies not displaying
cointegration to establish a causal relationship between the stock prices and
exchange rates. Granger approaches the question of whether X causes Y by
looking at how much of current Y can be explained by past values of Y and then
by checking whether adding the lagged values of X can improve the results. X is
said to Granger Cause Y if X helps in predicting Y and Y is said to Granger Cause
X if Y helps in predicting X. It is very important to mention that saying that X
Granger Causes Y does not mean that Y is the direct result of X. Rather Granger
Causality simply implies a chronological ordering of movement in the series.
Granger causality test requires the series to be stationary; else we would be
running a spurious regression. We have already established the unit root
property of the series. Thus to induce stationarity, we shall take first difference
of the series. Therefore we shall now be trying to find causality between
changes in stock prices and changes in exchange rate. To do so, we shall
estimate the following equations.
Δ�� = �� + Σ ��Δ���� + Σ ��Δ���� + �� (1)
Δ�� = �� + Σ ���Δ���� + Σ ���Δ���� + �� (2)
20
The lag length is selected by choosing the model which yields the optimal level
of SIC.
For India, South Korea and United Kingdom, the optimal number of lags to test
for Granger Causality came out to be four days. While for Brazil and Canada,
the optimal number of lags turned out to be three days. In case of Australia,
the optimal number of lag came out to be just one day. Lag length determined,
we proceed to estimation of the aforementioned equation.
The next step is to finally test for Granger Causality, i.e. whether Stock Returns
Granger Cause Exchange Returns and whether Exchange Returns Granger
Cause Stock Return. For this we test the joint significance of the coefficients of
the lagged variables of the other series. To do this, we run a Wald Test to test
the joint significance of
�� = �� = �� = … … = �� = 0
in equation (1), and the joint significance of
��� = ��� = ��� = … … = ��� = 0
in equation (2).
21
The f-statistics reported by the Wald Test are presented in table 6. If the f-
statistic is significant, then the coefficients are said to be jointly significant and
one variable is said to Granger Cause the other.
From the results presented in appendix A.1, we find evidence for bidirectional
causality for India, Brazil, Japan and South Korea between changes in stock
prices and changes in exchange rate. On the other hand, for Australia, we find
unidirectional causality from changes in exchange rate to change in stock price.
For Canada, there is evidence for unidirectional causality with changes in stock
price causing changes in exchange rate. Finally in the case of the United
Kingdom, we find evidence neither for bidirectional nor unidirectional causality
between changes in stock price and changes in exchange rate.
Now we move on to the important step of checking the robustness of our
model by including other exogenous variables. We undertake this task by
experimenting with different specifications of the model and observe whether
they have an impact on the results we obtained for causality.
It is well known that interest rates are closely related to exchange rates and
stock prices. The reason is very obvious. Interest rates are extremely important
in determining global capital flows. High interest rates attract capital inflows
which lead to an appreciation in the currency of the host economy and a
22
depreciation in the currency of the source economy. Secondly, interest rates
also impact the stock markets profoundly. Low lending rates create liquidity,
which provides easy money to be invested in the stock market. Also low
interest rates reduce the capital cost of business projects and thus increase the
revenue earned on projects apart from making previously unviable projects,
viable. All this impacts the stock market positively.
Thus, we include the difference between the domestic overnight call money
rate and the overnight call money rate for the US as an exogenous variable
while performing the Granger Causality test. The difference of the rates is
taken since only the difference is able to capture the relative profitability of
investing in either the host or the source economy. We take the difference
from the US rate in appreciation of the size and importance of the US economy
and it being an important source global capital flows. Also, since all the
currencies are expressed in terms of the USD, the case for the inclusion of the
US rate seems a pretty obvious one.
Next we include the S&P 500 Composite Index as another exogenous variable.
This US index is included for the ability of the US stock market in exerting
influence on the stock indices all over the world. The point about importance
of the US economy need not be stressed again.
23
But before including these variables, we must check them for stationarity. The
results are provided in the appendix A.3. As we can see, the S&P 500
Composite index contains a unit root. Thus to include the S&P index in the
Granger Causality, we induce stationarity by taking first difference of the series
since the test requires the variables to be stationary. The interest rate
difference on the other hand throws up different results. We find it to be
stationary in the case of India and Brazil. While we find it to contain a unit root
for South Korea, the Philippines, Australia, Canada, Japan and The United
Kingdom. Since the variables need to be stationary, for the cases where we
find it to be containing a unit root, we take first difference of the series.
Now we proceed to the ‘experimentation’. To do this we run our regression
again, we three different specifications. First we run it by including both the
interest rate difference (∆����) and the S&P index (∆�&���� ).
Δ�� = �� + Σ ��Δ���� + Σ ��Δ���� + ∆�&���� + ∆���� + �� (3)
Δ�� = �� + Σ ���Δ���� + Σ ���Δ���� + ∆�&���� + ∆���� + �� (4)
Then we run it by including only the S&P index.
Δ�� = �� + Σ ��Δ���� + Σ ��Δ���� + ∆�&���� + �� (5)
Δ�� = �� + Σ ���Δ���� + Σ ���Δ���� + ∆�&���� + �� (6)
24
Lastly, we run the regression by including only the interest rate differential.
Δ�� = �� + Σ ��Δ���� + Σ ��Δ���� + ∆���� + �� (7)
Δ�� = �� + Σ ���Δ���� + Σ ���Δ���� + ∆���� + �� (8)
In the cases where the interest rate difference is stationary, regress ����
instead of ∆����.
After obtaining the results for the Granger Causality for all the three
specifications (which are provided in the appendix A.1 along with the case
when no exogenous variables where included), we compare the results with
the previously obtained results where none of the exogenous variables were
included. We check whether the inclusion of the exogenous variables leads to
any significant change in our results. Finally, we choose that specification
which throws up the most appropriate results of the four different
specifications. The results for the VAR estimation are provided in appendix A.2.
The final results for Granger Causality with the appropriate specification are
provided in table 4. In case of India, there is not any significant difference in
the f-statistics provided by the Granger Causality test by running the difference
specifications. We still find evidence of bidirectional causality between changes
in stock price and changes in exchange rate. In the case of Brazil, we do
observe the f-statistics changing significantly by the inclusion of the exogenous
25
variables, but again our conclusion of bidirectional causality stands. Like the
case of India, for South Korea too, the f-statistics hardly change and the
evidence of bidirectional causality is validated. In the case of Australia, while
previously we had detected unidirectional causality only with exchange rate
changes causing changes in stock price, we now get different results under
different specifications. When we include both exogenous variables, we fail to
detect any causality whatsoever. Same holds true when we include only S&P
index. Interestingly, when we include only interest rate differential, we find
evidence of unidirectional causality, but now from changes in stock price to
changes in exchange rate. For Canada, the conclusion over unidirectional
causality running from changes in stock price to changes in exchange rate
stands, even though the f-statistics change considerably. The Japanese
economy had displayed bidirectional causality and the inclusion of both
exogenous variables does not change that conclusion while changing the f-
statistics slightly. However when we included only the interest rate difference,
the evidence for causality running from changes in exchange rate to changes in
stock price becomes tenuous, and leads to its rejection. Finally for the United
Kingdom, the conclusion of the absence of any causality stands, even though
the f-statistics do change a bit.
26
Overall it is safe to say that the inclusion of exogenous variable does not have
much impact upon our conclusion for Granger Causality except for the cases of
Australia and Japan.
The next step would be to check the model for econometric problems, mainly
for autocorrelation and structural breaks. To test for autocorrelation, we
perform Autocorrelation Lagrange Multiplier (LM) test upto the 4th
order. The
results are provided in the appendix A.4. India is the only economy for which
we find evidence of autocorrelation of all orders upto four. For Brazil,
Australia, Canada and Japan we don’t find any evidence for any order of
autocorrelation. For South Korea and The United Kingdom, we do find some
evidence of autocorrelation for some orders.
We test for structural break arising out of the current financial crisis using the
Chow Breakpoint test. To do this we test for structural breaks around the date
August 16th
2007 when the crisis had formally hit the foreign exchange market.
The Chow Breakpoint test divides the data into subsamples around the
specified date and checks for the stability of parameters. The results for the
Chow Breakpoint tests are provided in appendix A.5. The tests throw up mixed
results about the stability of the parameters. The only economies to observe
stable parameters before and after the recession for both directions of
causality are Australia and Japan (when interest rate difference is included).
27
While at the other end of the spectrum, Brazil, Canada and The United
Kingdom observe unstable parameters for both directions of causality. India
and South Korea throw up mixed results with either of the causal relationships
displaying instability in parameters. This result is very crucial. It necessitates
the carrying out of our exercise of establishing the relationship between stock
prices and exchange rates specifically for the period of the recession, that is,
from 16th
August 2007 onwards.
28
Table 4, Granger Causality Conclusion (1999-2010)
COUNTRY NULL HYPOTHESIS H0 f-stat critical f-value RESULT
INDIA a ∆St does not Granger Cause ∆Xt 3.88011 F(4,2862) = 2.37 Reject H0
∆Xt does not Granger Cause ∆St 110.630 F(4,2862) = 2.37 Reject H0
BRAZIL b ∆St does not Granger Cause ∆Xt 4.7576 F(3,2863) = 2.6 Reject H0
∆Xt does not Granger Cause ∆St 2.8421 F(3,2863) = 2.6 Reject H0
SOUTH KOREA a ∆St does not Granger Cause ∆Xt 78.4755 F(4,2862) = 2.37 Reject H0
∆Xt does not Granger Cause ∆St 2.96389 F(4,2862) = 2.37 Reject H0
AUSTRALIA b ∆St does not Granger Cause ∆Xt 3.1557 F(1,2865) = 3.84 Fail to Reject H0
∆Xt does not Granger Cause ∆St 2.6694 F(1,2865) = 3.84 Fail to Reject H0
AUSTRALIA d ∆St does not Granger Cause ∆Xt 9.1239 F(1,2865) = 3.84 Reject H0
∆Xt does not Granger Cause ∆St 0.8113 F(1,2865) = 3.84 Fail to Reject H0
CANADA b ∆St does not Granger Cause ∆Xt 3.9972 F(3,2863) = 2.6 Reject H0
∆Xt does not Granger Cause ∆St 1.4508 F(3,2863) = 2.6 Fail to Reject H0
JAPAN b ∆St does not Granger Cause ∆Xt 7.0535 F(2,2864) = 3 Reject H0
∆Xt does not Granger Cause ∆St 3.0791 F(2,2864) = 3 Reject H0
JAPAN d ∆St does not Granger Cause ∆Xt 4.5035 F(2,2864) = 3 Reject H0
∆Xt does not Granger Cause ∆St 2.9795 F(2,2864) = 3 Fail to Reject H0
UNITED KINGDOM b ∆St does not Granger Cause ∆Xt 1.5255 F(4,2862) = 2.37 Fail to Reject H0
∆Xt does not Granger Cause ∆St 0.2431 F(4,2862) = 2.37 Fail to Reject H0
a - no exogenous variable
b - including both S&P 500 and interest rate difference
c - including S&P 500
d - including interest rate difference
29
VECTOR ERROR CORRECTION MODEL
As noted above, since stock prices and exchange rates are found to be
following I(1) processes and there exists a cointegrating relationship between
the two for the Philippines, we need to include an error correction term and
run a VECM to test for Granger Causality. Thus we run the following equation.
∆����. �� = �� + Σ �� Δ ����. ���� + Σ �� Δ ����. ���� + ������ + ��� (9)
ΔPHIL. �� = �� + Σ ��� Δ ����. ���� + Σ ��� Δ����. ���� + ������ + ���(10)
The equation is specified in a manner such that it restricts the long run
behavior of variables to converge to their cointegrating relationships while
leaving room for short term adjustment dynamics. Here ���� and ���� are error
correction terms, where ���� = (����. ���� − & ����. ����) and ���� =
(����. ���� − (����. ����). Provided that ����. �� and ����. �� are
cointegrated with cointegration coefficient δ and η, then ���� and ���� will be
I(0), even though the constituents are I(1). The coefficients of error correction
terms, �� and ��, capture the adjustment of Δ ����. �� and Δ ����. ��
towards their long run equilibriums. While �� and ��� , the coefficients of
Δ����. ���� and Δ ����. ���� capture their short term movements. The
terms, ��� and ��� are stationary random processes included to capture the
information which is not provided by the variables included in the model.
30
We choose the optimal lag length by running the model with different lag
length and choosing the model which throws up the lowest SIC, which was four
days. Then as we did in the previous cases, we test with four different
specifications. The results for the estimation of the VECM are provided in
appendix A.6. Finally we test for Granger Causality by running the Granger
Causality/Block Exogeneity Wald Test for all the four cases, whose results are
provided in table 5. In all the cases, we find evidence of unidirectional
causality, with changes in stock exchange return causing changes in exchange
rate. The result for autocorrelation test is provided in appendix A.3.
Table 5, Philippines VEC Granger Causality/Block Exogeneity Wald Test Results (1999-2010)
Specification NULL HYPOTHESIS H0 Chi2
stat Results
No exogenous variables included
∆St does not Granger Cause ∆Xt 290.2977 Reject H0
∆Xt does not Granger Cause ∆St 2.224357 Fail to reject H0
Including S&P 500
and Interest difference.
∆St does not Granger Cause ∆Xt 291.1111 Reject H0
∆Xt does not Granger Cause ∆St 2.320225 Fail to reject H0
Including S&P 500.
∆St does not Granger Cause ∆Xt 289.9951 Reject H0
∆Xt does not Granger Cause ∆St 2.20864 Fail to reject H0
Including Interest Difference.
∆St does not Granger Cause ∆Xt 291.9105 Reject H0
∆Xt does not Granger Cause ∆St 2.34364 Fail to reject H0
critical value at 5% = 9.488 (4 d.f.)
31
RECESSIONARY IMPACT
The Chow test for structural breaks revealed the instability of parameters
when we broke down the data into pre and post recession. This necessitated
the testing of the causal relationship during the recession, which is from 16th
August 2007 onwards. To do this we undertook our whole exercise again, but
now on the subset of the whole data from 16th
August 2007 to 31st
March
2010. First we checked the variables for cointegration and the results are
provided in appendix A.7 and A.8. The results were same except for the
Philippines, where the previous evidence of cointegration broke down. This
breakdown in the long run relationship between the two prices might be
because of the instability experienced in the stock markets and currency
markets during the recession.
So now, we were unable to detect cointegration for neither of the economies
under consideration. Thus we proceeded to the estimation of the VAR whose
results are provided in appendix A.9. Then we tested for Granger Causality
under different specifications, as done previously, using four different
specifications. The results of that exercise are reported in appendix A.10.
The results for Granger Causality during the recession with appropriate
specification are provided in table 6. For India during the recession, we detect
unidirectional causality from changes in exchange rate to changes in stock
32
price. In the case of Brazil, when both exogenous variables are considered, we
fail to detect any causality whatsoever. While when we include only interest
difference, we find changes in stock price Granger Causing changes in
exchange price. South Korea and the Philippines display unidirectional causality
running from stock prices changes to changes in exchange rates. We fail to
detect any causality for Australia. The same is the case for Canada when we
include both exogenous variables. But when we include only interest
difference, we find evidence for stock price changes Granger Causing changes
in exchange rate. When we include both exogenous variables for Japan, we
find strong evidence for bidirectional causality. However when only interest
difference is considered, we find evidence only for changes in exchange rate
Granger Causing changes in stock price. Finally, for the United Kingdom, we
find evidence for stock price changes Granger Causing changes in exchange
rate when we include both the exogenous variables. However, when we
consider only interest rate difference we fail to detect any causality.
33
Table 6 , Granger Causality test results (2007-2010)
COUNTRY NULL HYPOTHESIS H0 f-stat critical f-value Result
INDIAb ∆St does not Granger Cause ∆Xt 0.885105 F(4,676) =2.37 Fail to Reject H0
∆Xt does not Granger Cause ∆St 30.77113 F(4,676) =2.37 Reject H0
BRAZILb ∆St does not Granger Cause ∆Xt 0.697377 F(3,675) = 2.6 Fail to Reject H0
∆Xt does not Granger Cause ∆St 2.187855 F(3,675) = 2.6 Fail to Reject H0
BRAZILd ∆St does not Granger Cause ∆Xt 15.9489 F(3,675) = 2.6 Reject H0
∆Xt does not Granger Cause ∆St 2.187855 F(3,675) = 2.6 Fail to Reject H0
S. KOREAa ∆St does not Granger Cause ∆Xt 35.27671 F(4,676) =2.37 Reject H0
∆Xt does not Granger Cause ∆St 1.939398 F(4,676) =2.37 Fail to Reject H0
PHIL.a ∆St does not Granger Cause ∆Xt 79.08529 F(2,679) = 3 Reject H0
∆Xt does not Granger Cause ∆St 0.691745 F(2,679) = 3 Fail to Reject H0
AUSTRALIAb ∆St does not Granger Cause ∆Xt 0.239587 F(1,681) =3.84 Fail to Reject H0
∆Xt does not Granger Cause ∆St 2.669466 F(1,681) =3.84 Fail to Reject H0
CANADAb ∆St does not Granger Cause ∆Xt 1.947463 F(3,675) = 2.6 Fail to Reject H0
∆Xt does not Granger Cause ∆St 1.38201 F(3,675) = 2.6 Fail to Reject H0
CANADAd ∆St does not Granger Cause ∆Xt 1.947463 F(3,675) = 2.6 Reject H0
∆Xt does not Granger Cause ∆St 1.38201 F(3,675) = 2.6 Fail to Reject H0
JAPANb ∆St does not Granger Cause ∆Xt 3.147687 F(2,679) = 3 Reject H0
∆Xt does not Granger Cause ∆St 7.11893 F(2,679) = 3 Reject H0
JAPANd ∆St does not Granger Cause ∆Xt 1.258694 F(2,679) = 3 Fail to Reject H0
∆Xt does not Granger Cause ∆St 4.327407 F(2,679) = 3 Reject H0
UKb ∆St does not Granger Cause ∆Xt 5.964761 F(4,676) =2.37 Reject H0
∆Xt does not Granger Cause ∆St 0.353515 F(4,676) =2.37 Fail to Reject H0
UKd ∆St does not Granger Cause ∆Xt 1.518974 F(4,676) =2.37 Fail to Reject H0
∆Xt does not Granger Cause ∆St 0.305808 F(4,676) = .37 Fail to Reject H0
a – no exogenous variables.
b - Including S&P and Interest difference.
c – Including S&P.
d- Including Interest difference
34
SECTION VI
CONCLUSION
In this study we examined the interaction between stock prices and exchange
rates for India, Brazil, South Korea, the Philippines, Australia, Canada, Japan
and the United Kingdom. We used Cointegration, Granger Causality tests and
Error Correction model techniques to establish the relationship. The first step
was to test our data for stationarity using ADF and PP tests. All the data came
out to be containing a unit root. The next step was to apply Johansen
Cointegration techniques to check for the existence of any long run
relationship between the two variables. Save for the Philippines, we were
unable to detect any evidence for any long run relationship between the two
variables for none of the economies. The absence of any cointegrating
relationship led us to apply Granger Causality tests for all the economies,
except for the Philippines. In the case of the Philippines, we constructed a
Vector Error Correction model and then tested for Granger Causality.
The Traditional Goods & Market approach to the relationship predicts that
changes in exchange rates lead changes in stock prices. On the other hand,
according to the Portfolio Balance Approach, changes in stock prices cause
changes in exchange rates. The results obtained present a mixed picture on the
validation of either of the approaches to the relationship. The presence of
35
unidirectional causality, with changes in Stock prices causing changes in
Exchange rates for the Philippines, Australia, Canada and Japan, present
support for the Portfolio Balance approach. However, the evidence of
bidirectional causality or feedback relationship for India, Brazil and South
Korea and the absence of any relationship for the United Kingdom contradicts
that conclusion and offers no resolution over the debate over the validity of
either approach to the relationship. These results are consistent with the ones
obtained by Solnik (1987) in the sense we have failed to detect any significant
impact of exchange markets on stock markets. One conclusion we can make is
that the economies which display unidirectional causality with changes in stock
prices leading changes in exchange rates are all advanced economies, except
for the Philippines. As noted before, the Portfolio Balance approach stresses
on the flow of global capital flows. Thus, we may say that the Portfolio Balance
approach is at work in economies which experience high Capital inflows and
outflows.
The degree of freedom from capital control and liberalization of external trade
along with the relative dominance of global capital flows over foreign trade in
the different economies may account for the varying results we have obtained
for different countries.
36
Finally, we tested whether the conclusion reached above held during the
period of the recession. For the recession, we again find conflicting results.
India and Japan display unidirectional causality, with changes in exchange rates
leading stock prices and thus lending support to the traditional argument. On
the other hand, we find support for the Portfolio Balance approach, with
changes in stock prices causing changes in exchange rates for Brazil, South
Korea, the Philippines, Canada and the United Kingdom. We fail to detect any
relationship for Australia. But one must not read too much into the results
obtained for the period of the recession, due to the high volatility, instability
and government intervention in these markets during the recession.
In conclusion, although we find some evidence for the Portfolio Balance
approach for advanced economies, it is far from unequivocal. This is in line
with the existing literature, with no conclusive evidence in support for either of
the approach. This necessitates the undertaking of further deeper and
widespread studies into the debate.
37
BIBLOGRAPHY
Aggarwal, Exchange rates and Stock price: A study of the US Capital markets
under floating exchange rates, Akron Business and Economic Review, 1981.
Basabi Bhattacharya & Jaydeep Mukherjee, 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.
B Solnik, Using financial prices to test exchange rate models: A Note, Journal of
Finance, 1987.
CC Nieh & CF Lee, Dynamic relationship between stock prices and exchange
rates for G-7 countries, The Quarterly Review of Economics and Finance, 2001.
CE Smith, Stock markets and the exchange rate: A multi-country approach,
Journal of Macroeconomics, 1992.
CE Smith, Equities and the UK exchange rate, Applied Economics, 1992.
Chris Brooks, Introductory Econometrics for Finance, Second Edition,
Cambridge University Press, 2008.
CK Ma, GW Kao, On exchange rate changes and stock price reactions, Journal
of Business Finance & Accounting, 1990.
CWJ Granger, B Huang & C Yang, A Bivariate causality between stock prices and
exchange rates: evidence from recent Asian flu, The Quarterly Review of
Economics and Finance, 2000.
CWJ Granger, Investigating causal relations by econometric models and cross-
spectral methods , Econometrica: Journal of the Econometric Society, 1969.
Dimitrios Asteriou & Stephen G. Hall, Applied Econometrics: A Modern
Approach Using Eviews and Microfit, Palgrave Macmillan, 2007.
Desislava Dimitrova, The Relationship between Exchange Rates and Stock
Prices: Studied in a Multivariate Model, Issues in Political Economy Vol. 14,
August 2005.
38
F. Beer & F. Hebein, An Assessment Of The Stock Market And Exchange Rate
Dynamics In Industrialized And Emerging Markets, International Business &
Economic Research Journal, August 2008.
GC Nath & GP Samanta, Dynamic Relation between Exchange Rate and Stock
Prices – A Case for India, 39th
Annual Conference Paper of Indian Econometric
Society, 2003.
GL Shoesmith, Non-cointegration and causality: Implications for VAR
modelling, International Journal of Forecasting, 1992.
Gopalan Kutty, The Relationship between Exchange rates and Stock prices: The
case of Mexico, North American Journal of Finance and Banking Research Vol.
4 No. 4, 2010.
H Lutkepohl & HE Reimers, Granger-causality in cointegrated VAR processes -
The case of the term structure, Economics Letters, 1992.
ISA Abdalla & V Murinde, Exchange rate and stock price interactions in
emerging financial markets: Evidence on India, Korea, Pakistan and the
Philippines, Applied financial economics, 1997.
JA Frankel & KM Dominguez, Does foreign Exchange Intervention Matter? The
Portfolio Effect, The American Economic Review Vol. 83 No.5, 1993
Mansor H Ibrahim, Cointegration and Granger causality tests of stock price and
exchange rate interactions in Malaysia, ASEAN Economic Bulletin, 2000.
M Melvin, MP Taylor, The crisis in the foreign exchange market, Journal of
International Money and Finance, 2009.
M Gavin, The stock market and exchange rate dynamics, Journal of
International Money and Finance, 1989.
Md. Lutfur Rahman & Jashim Uddin, Dynamic relationship between stock prices
and exchange rates: evidence from three South Asian countries, International
Business Research Vol. 2 No. 2, 2009.
39
M Bahmani-Oskooee & A Sohrabian, Stock prices and the effective exchange
rate of the dollar, Applied economics, 1992.
Naeem Muhammad & Abdul Rasheed, Stock Prices and Exchange Rates: Are
they Related? Evidence from South Asian Countries, The Pakistan Development
Review, Pakistan Institute of Development Economics, Vol. 41(4), 2002, pages
535-550.
O Aydemir & E Demirhan, The relationship between stock prices and exchange
rates: evidence from Turkey, International Research Journal of Finance and
Economics, 2009.
P Franck & A Young, Stock price reaction of multinational firms to exchange
realignments, Financial Management, 1972.
RA Ajayi & M Mougoue, On the dynamic relation between stock prices and
exchange rates, Journal of Financial Research, 1996.
RA Ajayi, J Friedman & SM Mehdian, On the relationship between stock returns
and exchange rates: tests of Granger causality, Global Finance Journal, 1998.
Ronald MacDonald & Colm Kearney, On the specification of granger-causality
tests using the cointegration methodology, Economics Letters Vol. 25 Issue 2,
1987, Pages 149-153.
R Dornbusch & S Fisher, Exchange Rates and the Current Account, The
American Economic Review Vol. 70 No.5, 1980
S Gupta, Testing causality: Some caveats and a suggestion, International
Journal of Forecasting, 1987.
Y Wu, Stock prices and exchange rates in VEC model - The case of Singapore in
the 1990s, Journal of Economics and Finance, 2000.
40
APPENDIX
A.1 , Granger Causality test results (1999-2010)
COUNTRY NULL HYPOTHESIS H0 f-stata
f-statb
f-statc
f-statd
critical f-value
INDIA ∆St does not Granger Cause ∆Xt 3.88011 3.8516 3.8269 3.9029 F(4,2862) =2.37
∆Xt does not Granger Cause ∆St 110.630 99.2359 99.3872 110.4665 F(4,2862) =2.37
BRAZIL ∆St does not Granger Cause ∆Xt 2.77639 4.7576 4.7291 22.8552 F(3,2863) = 2.6
∆Xt does not Granger Cause ∆St 22.8114 2.8421 2.7878 2.8421 F(3,2863) = 2.6
S. KOREA ∆St does not Granger Cause ∆Xt 78.4755 74.9773 75.0605 78.4077 F(4,2862) =2.37
∆Xt does not Granger Cause ∆St 2.96389 2.6873 2.3013 2.8506 F(4,2862) =2.37
AUSTRALIA ∆St does not Granger Cause ∆Xt 0.81785 3.1557 3.159 9.1239 F(1,2865) =3.84
∆Xt does not Granger Cause ∆St 9.13060 2.6694 2.6824 0.8113 F(1,2865) =3.84
CANADA ∆St does not Granger Cause ∆Xt 8.19099 3.9972 2.0722 8.1182 F(3,2863) = 2.6
∆Xt does not Granger Cause ∆St 2.25731 1.4508 1.4666 2.2488 F(3,2863) = 2.6
JAPAN ∆St does not Granger Cause ∆Xt 4.49667 7.0535 7.0576 4.5035 F(2,2864) = 3
∆Xt does not Granger Cause ∆St 3.00 3.0791 3.0898 2.9795 F(2,2864) = 3
UK ∆St does not Granger Cause ∆Xt 0.31162 1.5255 1.5254 0.7654 F(4,2862) =2.37
∆Xt does not Granger Cause ∆St 0.76000 0.2431 0.2465 0.3103 F(4,2862) =2.37
a – no exogenous variables.
b - Including S&P and Interest difference.
c – Including S&P.
d- Including Interest difference.
41
A.2, results for VAR estimation (1999-2010). Standard errors in ( ) & t-statistics in [ ]
COUNTRY INDIAa
BRAZILb
SOUTH KOREAa
DEPENDENT VARIABLE ∆St ∆Xt ∆St ∆Xt ∆St ∆Xt
C 2.904839 0.000565 58.39044 0.000803 0.047508 0.050350
(1.62161) (0.00292) (36.6210) (0.00135) (0.14002) (0.04568)
[ 1.79133] [ 0.19317] [ 1.59445] [ 0.59429] [ 0.339 28] [ 1.10221]
∆St-1 0.039106 -8.95E-05 -0.09409 -2.96E-06 0.191065 -0.0022
(0.01869) (3.4E-05) (0.02370) (8.7E-07) (0.01874) (0.00611)
[ 2.09263] [-2.65693] [-3.97007] [-3.38146] [ 10.19 72] [-0.3590]
∆St-2 0.018283 -6.02E-06 -0.03467 9.27E-07 -0.0834 0.020456
(0.01751) (3.2E-05) (0.01987) (7.3E-07) (0.01898) (0.00619)
[ 1.04400] [-0.19056] [-1.74496] [ 1.26439] [-4.395 0] [ 3.30443]
∆St-3 -0.02477 7.03E-05 -0.08212 -6.61E-07 -0.10971 -0.00501
(0.01750) (3.2E-05) (0.01976) (7.3E-07) (0.01900) (0.00620)
[-1.41579] [ 2.22891] [-4.15631] [-0.90694] [-5.773 4] [-0.8076]
∆St-4 -0.01518 5.27E-05
0.047471 -0.00351
(0.01746) (3.1E-05)
(0.01775) (0.00579)
[-0.86936] [ 1.67328]
[ 2.67450] [-0.6066]
∆Xt-1 -6.99105 -0.00911 -226.09 0.053326 -0.98154 0.007054
(10.3772) (0.01871) (529.553) (0.01954) (0.05757) (0.01878)
[-0.67370] [-0.48680] [-0.42695] [ 2.72938] [-17.04 8] [ 0.37554]
∆Xt-2 -44.6091 -0.02508 -1240.35 -0.0372 -0.20102 -0.02066
(10.3674) (0.01869) (529.176) (0.01952) (0.06044) (0.01972)
[-4.30283] [-1.34169] [-2.34393] [-1.90550] [-3.325 8] [-1.0477]
∆Xt-3 -213.778 0.007535 -812.55 -0.02935 -0.1764 0.021794
(10.3961) (0.01874) (519.811) (0.01918) (0.06051) (0.01974)
[-20.5633] [ 0.40202] [-1.56316] [-1.53060] [-2.915 3] [ 1.10412]
∆Xt-4 -27.8882 0.021506
-0.17502 -0.00688
(11.1066) (0.02002)
(0.06054) (0.01975)
[-2.51097] [ 1.07403]
[-2.8908] [-0.3482]
∆S&Pt-1
3.710908 -0.00022
(1.05227) (3.9E-05)
[ 3.52657] [-5.56103]
It-1
-2.44006 -5.42E-05
(2.52929) (9.3E-05)
[-0.96472] [-0.58104] a – no exogenous variables.
b – Including S&P and Interest difference.
c – Including S&P.
d – Including Interest difference.
42
A.2 cont. Standard errors in ( ) & t-statistics in [ ]
COUNTRY AUSTRALIAb
AUSTRALIAd
CANADAb
DEPENDENT VARIABLE ∆St ∆Xt ∆St ∆Xt ∆St ∆Xt
C 0.812537 -0.00018 0.720468 -0.00017 -0.00017 2.127057
(0.71545) (0.00022) (0.84775) (0.00023) (0.00013) (2.27844)
[ 1.13571] [-0.8067] [ 0.8496] [-0.7079] [-1.3302] [ 0.93356]
∆St-1 -0.0962 -9.14E-06 -0.05022 -1.62E-05 -0.03749 -627.602
(0.01659) (5.1E-06) (0.01960) (5.4E-06) (0.01969) (348.309)
[-5.7975] [-1.7764] [-2.5628] [-3.0205] [-1.9041] [ -1.8018]
∆St-2
0.025159 -359.736
(0.01964) (347.370)
[ 1.28116] [-1.0356]
∆St-3
-0.01626 137.4635
(0.01954) (345.563)
[-0.8324] [ 0.39780]
∆St-4
∆Xt-1 -98.3392 -0.04202 -64.2306 -0.04729 1.38E-06 -0.16168
(60.1886) (0.01866) (71.3098) (0.01957) (1.5E-06) (0.02579)
[-1.6338] [-2.2517] [-0.9007] [-2.4164] [ 0.94445] [-6.2696]
∆Xt-2
-1.22E-06 -0.04016
(1.1E-06) (0.01985)
[-1.0862] [-2.0228]
∆Xt-3
-3.55E-06 0.011981
(1.1E-06) (0.01966)
[-3.1897] [ 0.60932]
∆Xt-4
∆S&Pt-1 1.647430 -0.00026
-7.40E-05 1.472058
(0.04840) (1.5E-05)
(1.2E-05) (0.21289)
[ 34.0392] [-16.968]
[-6.1499] [ 6.91470]
∆It-1 1.562180 -0.00077 1.735522 -0.0008 -0.00102 -9.86579
(2.77886) (0.00086) (3.29277) (0.00090) (0.00091) (16.1015)
[ 0.56216] [-0.8927] [ 0.5270] [-0.8807] [-1.1202] [-0.6127]
43
A.2 cont. Standard errors in ( ) & t-statistics in [ ]
COUNTRY JAPANb
JAPANd
UNITED KINGDOMb
DEPENDENT VARIABLE ∆St ∆Xt ∆St ∆Xt ∆St ∆Xt
C -0.01025 -1.54952 -0.01057 -1.71908 -0.24701 9.93E-06
(0.01303) (3.09539) (0.01339) (3.49721) (1.13112) (6.9E-05)
[-0.7867] [-0.5005] [-0.7895] [-0.4915] [-0.2183] [ 0.14464]
∆St-1 -0.03783 10.05808 -0.03464 11.72662 -0.25311 2.64E-06
(0.01896) (4.50491) (0.01948) (5.08928) (0.01996) (1.2E-06)
[-1.9953] [ 2.23269] [-1.7780] [ 2.30418] [-12.683] [ 2.17587]
∆St-2 -0.01866 5.341871 -0.01998 4.647578 -0.05041 -1.77E-07
(0.01892) (4.49608) (0.01945) (5.07966) (0.01751) (1.1E-06)
[-0.9862] [ 1.18812] [-1.0275] [ 0.91494] [-2.8792] [-0.1664]
∆St-3
-0.06595 1.23E-06
(0.01752) (1.1E-06)
[-3.7642] [ 1.15501]
∆St-4
0.034558 3.05E-08
(0.01754) (1.1E-06)
[ 1.96992] [ 0.02865]
∆Xt-1 -0.00026 -0.07216 -0.00018 -0.03309 235.4590 0.060695
(7.3E-05) (0.01733) (7.5E-05) (0.01952) (306.661) (0.01862)
[-3.5386] [-4.1636] [-2.4562] [-1.6952] [ 0.76781] [ 3.26035]
∆Xt-2 -9.54E-05 -0.00699 -0.00013 -0.02619 136.8260 -0.00915
(7.3E-05) (0.01726) (7.5E-05) (0.01948) (306.923) (0.01863)
[-1.3143] [-0.4052] [-1.7706] [-1.3442] [ 0.44580] [-0.4912]
∆Xt-3
-3.82606 -0.03766
(306.926) (0.01863)
[-0.0124] [-2.0210]
∆Xt-4
-100.815 -0.01413
(306.415) (0.01860)
[-0.3290] [-0.7597]
∆S&Pt-1 0.011263 5.904487
1.744453 -3.28E-05
(0.00088) (0.20977)
(0.08739) (5.3E-06)
[ 12.7601] [ 28.1477]
[ 19.9616] [-6.1759]
∆It-1 0.012898 -10.4511 -0.02348 -29.5182 -4.24236 -1.57E-05
(0.09548) (22.6905) (0.09810) (25.6246) (3.51302) (0.00021)
[ 0.13508] [-0.4605] [-0.2392] [-1.1519] [-1.2076] [-0.0737]
44
A.3, unit root results (1999-2010)
Interest Difference ∆Interest Difference
Country ADF Country ADF
India -5.40373
South Korea -25.43416
Brazil -3.84683
Philippines -29.43486
South Korea -1.14404
Australia* -24.98059
Philippines -2.72894
Canada* -25.26396
Australia* -0.24682
Japan* -25.63841
Canada* -1.45844
United Kingdom* -28.13589
Japan* -1.47009
United Kingdom* -1.60152
ADF ADF
S&P 500 -1.81503 ∆S&P 500 -41.8996
The Critical Values are:
series containing an intercept but no trend
1% level = -3.4324 5% level = -2.8623, 10% level = -2.5672 *series without an intercept of trend
1% level = -2.56577 5% level = -1.940934 10% level = -1.61663
45
A.4, Autocorrelation LM Test Results (1999-2010) Null hypothesis, H0 : No Autocorrelation
Critical value for chi-square with 4 d.f.
at 1% = 13.277, 5% = 9.488 ,
*significant at 1%
a – No exogenous variables
b – Including S&P and interest diff.
c – Including S&P.
d – Including interest diff.
Country Order LM-Stat Result
Indiaa 1 19.2572 Reject H0
2 22.4787 Reject H0
3 11.71892 Reject H0
4 16.53189 Reject H0
Brazilb 1 10.93675 Fail to Reject H0
*
2 5.913935 Fail to Reject H0
3 6.584866 Fail to Reject H0
4 4.732997 Fail to Reject H0
South 1 1.999343 Fail to Reject H0 Cntry. Order LM-stat Result
Koreaa 2 2.951343 Fail to Reject H0
Phil. 1
2
3
4
14.108
6.052
2.519
5.339
Reject
Fail to Reject
Fail to Reject
Fail to reject
3 13.4508 Reject H0
4 11.26928 Fail to Reject H0*
Australiab 1 55.52432 Reject H0
2 14.46430 Reject H0
3 9.471459 Fail to Reject H0
4 0.912460 Fail to Reject H0
Australiad 1 13.28761 Fail to Reject H0
*
2 11.28267 Fail to Reject H0*
3 5.074659 Fail to Reject H0
4 0.708953 Fail to Reject H0
Canadab 1 8.199857 Fail to Reject H0
2 3.471770 Fail to Reject H0
3 5.526818 Fail to Reject H0
4 7.787062 Fail to Reject H0*
Japanb 1 47.98933 Reject H0
2 3.180137 Fail to Reject H0
3 6.700516 Fail to Reject H0
4 5.829503 Fail to Reject H0
Japand 1 3.769007 Fail to Reject H0
2 5.663514 Fail to Reject H0
3 5.468736 Fail to Reject H0
4 6.702414 Fail to Reject H0
United 1 29.09026 Reject H0
Kingdomb 2 17.22139 Reject H0
3 0.502647 Fail to Reject H0
4 28.24224 Reject H0
46
A.5, Chow Breakpoint Test Results (1999-2010)
Country Causality f-Stat Critical Values(5%) Result
Indiaa ∆St — ∆Xt 1.732977 F(9,2848)= 1.88 Fail to Reject H0
∆Xt — ∆St 4.92715 F(9,2848)= 1.88 Reject H0
Brazilb ∆St — ∆Xt 14.68836 F(7,2853)= 2.01 Reject H0
∆Xt — ∆St 4.713201 F(7,2853)= 2.01 Reject H0
South Koreaa ∆St — ∆Xt 20.30255 F(9,2848)= 1.88 Reject H0
∆Xt — ∆St 0.805434 F(9,2848)= 1.88 Fail to Reject H0
Australiab ∆St — ∆Xt 34.8627 F(3,2863)= 2.6 Reject H0
∆Xt — ∆St 31.19437 F(3,2863)= 2.6 Reject H0
Australiad ∆St — ∆Xt 0.935721 F(3,2863)= 2.6 Fail to Reject H0
∆Xt — ∆St 1.700828 F(3,2863)= 2.6 Fail to Reject H0
Canadab ∆St — ∆Xt 2.803662 F(7,2853)= 2.01 Reject H0
∆Xt — ∆St 2.566381 F(7,2853)= 2.01 Reject H0
Japanb ∆St — ∆Xt 10.65451 F(5,2858)= 2.21 Reject H0
∆Xt — ∆St 3.364789 F(5,2858)= 2.21 Reject H0
Japand ∆St — ∆Xt 0.469108 F(5,2858)= 2.21 Fail to Reject H0
∆Xt — ∆St 1.889155 F(5,2858)= 2.21 Fail to Reject H0
United Kingdomb ∆St — ∆Xt 8.626655 F(9,2848)= 1.88 Reject H0
∆Xt — ∆St 2.549293 F(9,2848)= 1.88 Reject H0
Null Hypothesis, H0: No breaks at specified breakpoints (16/8/2007)
a - no exogenous variable
b - including both S&P 500 and interest rate difference
c - including S&P 500
d - including interest rate difference
47
A.6, VECM Estimates for the Philippines (1999-2010)
Sample (adjusted): 4/07/1999 3/31/2010
Included observations: 2866 after adjustments
Standard errors in ( ) & t-statistics in [ ]
Cointegrating Eq: CointEq1
PESOt-1 1.000000
PSEIt-1 0.007749
(0.00172)
[ 4.50349]
C -65.8153
(3.69334)
[-17.8200]
Error Correction: ∆PESO ∆PSEI
CointEq1 -0.00294 0.119231
(0.00070) (0.10579)
[-4.19330] [ 1.12702]
∆PESOt-1 0.006194 -2.268547
(0.01866) (2.81874)
[ 0.33189] [-0.80481]
∆PESOt-2 -0.02779 -2.214883
(0.01865) (2.81737)
[-1.48998] [-0.78615]
∆PESOt-3 -0.03229 -1.552042
(0.01866) (2.81779)
[-1.73080] [-0.55080]
∆PESOt-4 -0.00687 2.001873
(0.01779) (2.68767)
[-0.38580] [ 0.74484]
∆PSEIt-1 -0.00211 0.105512
(0.00012) (0.01875)
[-16.9622] [ 5.62819]
∆PSEIt-2 9.04E-05 -0.010846
(0.00013) (0.01977)
[ 0.69051] [-0.54866]
∆PSEIt-3 -7.59E-05 -0.009365
(0.00013) (0.01977)
[-0.58027] [-0.47383]
∆PSEIt-4 0.000140 0.008397
(0.00013) (0.01967)
[ 1.07604] [ 0.42695]
48
A.7, Johansen Cointegration Test Results (Trace statistic) (2007-2010)
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
INDIA r=0 r>= 0 7.28426 6.75659 6.8655 6.86050 20.2618
r>1 r>=2 0.94010 0.96865 0.9602 0.99852 9.16455
Trace test indicates no cointegration at the 0.05 level
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
BRAZIL r=0 r>= 0 10.6961 12.6339 12.557 11.0663 20.2618
r>1 r>=2 2.28847 2.32161 2.2445 2.14689 9.16455
Trace test indicates no cointegration at the 0.05 level
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
S. KOREA r=0 r>= 0 12.4659 11.7323 10.811 9.78403 20.2618
r>1 r>=2 1.71961 1.68362 1.8350 2.01463 9.16455
Trace test indicates no cointegration at the 0.05 level
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
PHIL. r=0 r>= 0 11.2114 12.3652 12.671 13.8722 20.2618
r>1 r>=2 3.80755 3.81330 4.3331 4.30695 9.16455
Trace test indicates no cointegration at the 0.05 level
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
AUST. r=0 r>= 0 8.46029 7.08010 7.0783 7.34520 20.2618
r>1 r>=2 1.66411 1.55727 1.6182 1.57860 9.16455
Trace test indicates no cointegration at the 0.05 level
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
CANADA r=0 r>= 0 11.5497 12.7451 12.267 13.9021 20.2618
r>1 r>=2 1.56667 1.58648 1.6246 2.07003 9.16455
Trace test indicates no cointegration at the 0.05 level
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
JAPAN r=0 r>= 0 8.21395 9.41419 9.3368 9.86616 20.2618
r>1 r>=2 2.22022 2.19047 2.3441 2.58167 9.16455
Trace test indicates no cointegration at the 0.05 level
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
UK r=0 r>= 0 7.26898 6.67626 6.9720 6.74144 20.2618
r>1 r>=2 1.96511 1.89057 1.6826 2.14580 9.16455
Trace test indicates no cointegration at the 0.05 level
49
A.8, Johansen Cointegration Test Results (Maximum eigenvalue) (2007-2010)
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
INDIA r=0 r>= 0 6.34415 5.78794 5.90530 5.86197 15.8921
r>1 r>=2 0.94010 0.96865 0.96027 0.99852 9.164546
Max-eigenvalue test indicates no cointegration at the 0.05 level
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
BRAZIL r=0 r>= 0 8.40765 10.3123 10.3132 8.91945 15.8921
r>1 r>=2 2.28847 2.32161 2.24458 2.14689 9.164546
Max-eigenvalue test indicates no cointegration at the 0.05 level
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
S. KOREA r=0 r>= 0 10.7463 10.0487 8.97597 7.76939 15.8921
r>1 r>=2 1.71961 1.68362 1.83506 2.01463 9.164546
Max-eigenvalue test indicates no cointegration at the 0.05 level
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
PHIL. r=0 r>= 0 7.40387 8.55189 8.33786 9.56530 15.8921
r>1 r>=2 3.80755 3.81330 4.33313 4.30695 9.164546
Max-eigenvalue test indicates no cointegration at the 0.05 level
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
AUSTRALIA r=0 r>= 0 6.79618 5.52283 5.46009 5.76659 15.8921
r>1 r>=2 1.66411 1.55727 1.61822 1.57860 9.164546
Max-eigenvalue test indicates no cointegration at the 0.05 level
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
CANADA r=0 r>= 0 9.98302 11.1586 10.6431 11.8320 15.8921
r>1 r>=2 1.56667 1.58648 1.62462 2.07003 9.164546
Max-eigenvalue test indicates no cointegration at the 0.05 level
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
JAPAN r=0 r>= 0 5.99372 7.22371 6.99269 7.28449 15.8921
r>1 r>=2 2.22022 2.19047 2.34416 2.58167 9.164546
Max-eigenvalue test indicates no cointegration at the 0.05 level
null
hypothesis
alternate
hypothesis L=1 L=2 L=3 L=4
critical
value
UK r=0 r>= 0 5.30387 4.78569 5.2894 4.59564 15.8921
r>1 r>=2 1.96511 1.89057 1.68260 2.14580 9.164546
Max-eigenvalue test indicates no cointegration at the 0.05 level
50
A.9, Results for VAR estimation (2007-2010). Standard errors in ( ) & t-statistics in [ ]
COUNTRY INDIAb BRAZIL
b BRAZIL
d
DEPENDENT VARIABLE ∆St ∆Xt ∆St ∆Xt ∆St ∆Xt
C 21.81043 -0.04339 191.0293 -0.007457 460.4141 -0.01037
-11.9402 -0.02171 -151.535 -0.00442 -216.011 -0.00472
[ 1.8266] [-1.99888] [ 1.26063] [-1.68576] [ 2.1314 ] [-2.19880]
∆St-1 0.044058 -8.03E-05 0.030966 -6.79E-06 -0.0861 -5.52E-06
-0.03811 -6.90E-05 -0.03014 -8.80E-07 -0.0426 -9.30E-07
[ 1.1561] [-1.15938] [ 1.02737] [-7.71495] [-2.0212 ] [-5.93964]
∆St-2 0.030583 3.15E-05 -0.04232 1.69E-06 -0.08259 2.13E-06
-0.03487 -6.30E-05 -0.03114 -9.10E-07 -0.04444 -9.70E-07
[ 0.8770] [ 0.49748] [-1.35905] [ 1.86210] [-1.8585 ] [ 2.19361]
∆St-3 -0.0397 0.000112 -0.0766 -7.15E-07 -0.13667 -6.56E-08
-0.0348 -6.30E-05 -0.0309 -9.00E-07 -0.04402 -9.60E-07
[-1.1406] [ 1.76964] [-2.47935] [-0.79249] [-3.1046 ] [-0.06831]
∆St-4 -0.03312 7.80E-05
-0.03471 -6.30E-05
[-0.9542] [ 1.23679]
∆Xt-1 -14.6969 0.001171 265.7224 -0.189281 -1276.57 -0.17261
-21.1618 -0.03847 -1375.18 -0.04014 -1962.97 -0.04285
[-0.6945] [ 0.03045] [ 0.19323] [-4.71494] [-0.6503 ] [-4.02827]
∆Xt-2 -31.0139 -0.04796 -3053.27 0.068429 -4513.1 0.084204
-21.6194 -0.0393 -1390.87 -0.0406 -1985.57 -0.04334
[-1.4345] [-1.22026] [-2.19522] [ 1.68532] [-2.2729 ] [ 1.94268]
∆Xt-3 -227.006 0.013785 82.57036 -0.086128 -3124.37 -0.05147
-21.8683 -0.03975 -1326.59 -0.03873 -1887.41 -0.0412
[-10.380] [ 0.34677] [ 0.06224] [-2.22402] [-1.6553 ] [-1.24932]
∆Xt-4 -20.2587 0.053361
-23.2145 -0.0422
[-0.8726] [ 1.26446]
∆S&Pt-1 1.23104 0.000354 42.85917 -0.000463
-0.2995 -0.00054 -1.61674 -4.70E-05
[ 4.109] [ 0.6506] [ 26.509] [-9.8131]
It-1 -4.0253 0.011696 -14.6611 0.000729 -43.926 0.001045
-2.5516 -0.00464 -15.3555 -0.00045 -21.881 -0.00048
[-1.577] [ 2.5215] [-0.9547] [ 1.6262] [-2.007] [ 2.1881]
a – no exogenous variables.
b – Including S&P and Interest difference.
c – Including S&P.
d – Including Interest difference
51
A.9 cont. Standard errors in ( ) & t-statistics in [ ]
COUNTRY SOUTH KOREAa PHILIPPINES
a AUSTRALIA
b
DEPENDENT VAR. ∆St ∆Xt ∆St ∆Xt ∆St ∆Xt
C -0.00201 0.282759 -67.6272 0.454496 -1.197628 -0.00025
-0.13198 -0.49058 -46.8576 -0.18753 -2.90012 -0.00064
[-0.01525] [ 0.57638] [-1.4432] [ 2.4235] [-0.41296 ] [-0.39145]
∆St-1 -0.00955 -1.68E+00 1.10161 -0.00184 -0.040507 -5.90E-06
-0.03899 -0.14494 -0.0385 -0.00015 -0.04448 -9.80E-06
[-0.2449] [-11.5872] [ 28.613] [-11.951] [-0.91075] [-0.59954]
∆St-2 0.01513 -0.22418 -0.11527 0.002357
-0.0425 -0.15802 -0.05972 -0.00024
[ 0.356] [-1.4186] [-1.9301] [ 9.8622]
∆St-3 0.05015 -0.29336 0.01426 -0.00056
-0.0423 -0.15753 -0.04217 -0.00017
[ 1.183] [-1.862] [ 0.3383] [-3.3149]
∆St-4 -0.0126 -0.2770
-0.0417 -0.15518
[-0.302] [-1.785]
∆Xt-1 -0.0037 0.204129 -7.87E+00 1.109771 -179.118 -0.01695
-0.01045 -0.03882 -9.53307 -0.03815 -201.348 -0.04454
[-0.3546] [ 5.2576] [-0.825] [ 29.087] [-0.88959] [ -0.38042]
∆Xt-2 0.02781 -0.096 2.44E+00 -0.08899
-0.0105 -0.0392 -13.3328 -0.05336
[ 2.635] [-2.447] [ 0.18334] [-1.667]
∆Xt-3 -0.0041 -0.15237 6.87E+00 -0.0281
-0.0106 -0.03941 -8.66771 -0.03469
[-0.386] [-3.866] [ 0.792] [-0.8101]
∆Xt-4 -0.0088 0.06301
-0.0094 -0.03497
[-0.937] [ 1.802]
∆S&Pt-1
0.34233 -4.34E-05
-0.14538 -3.20E-05
[ 2.35469] [-1.349]
∆It-1
2.738288 0.000385
-6.5309 -0.00144
[ 0.41928] [ 0.2665]
52
A.9 cont. Standard errors in ( ) & t-statistics in [ ]
COUNTRY CANADAb CANADA
d JAPAN
b
DEPENDENT VAR. ∆St ∆Xt ∆St ∆Xt ∆St ∆Xt
C 1.237397 -0.00017 -1.987482 -5.64E-05 -6.82009 -0.03526
-5.03869 -0.00032 -7.37227 -0.00037 -8.35806 -0.0334
[ 0.24558] [-0.5099] [-0.2695] [-0.1516] [-0.8159] [-1.0556]
∆St-1 0.034091 -8.06E-06 -0.128091 -2.57E-06 -0.11796 -0.00029
-0.03125 -2.00E-06 -0.04492 -2.30E-06 -0.05015 -0.0002
[ 1.09106] [-4.0061] [-2.85135] [-1.13650] [-2.3522] [-1.4251]
∆St-2 -0.050722 1.87E-06 -0.093589 3.32E-06 -0.05964 -0.00013
-0.03107 -2.00E-06 -0.04542 -2.30E-06 -0.0502 -0.0002
[-1.63239] [ 0.9357] [-2.06059] [ 1.4502] [-1.1881] [-0.6285]
∆St-3 -0.004301 -3.09E-06 0.01902 -3.88E-06 -0.02104 -1.70E-05
-0.03063 -2.00E-06 -0.04481 -2.30E-06 -0.04963 -0.0002
[-0.14041] [-1.5690] [ 0.42444] [-1.71830] [-0.4239] [-0.0859]
∆St-4
38.1978 -0.02087
-12.5621 -0.0502
[ 3.0407] [-0.41570]
∆Xt-1 -1481.71 0.008086 -1789.158 0.018484 5.29282 -0.03367
-607.312 -0.0391 -888.667 -0.04482 -12.6867 -0.0507
[-2.43978] [ 0.2067] [-2.01330] [ 0.41242] [ 0.4171] [-0.66421]
∆Xt-2 -219.174 0.068429 -1117.504 0.098811 -11.6406 -0.02831
-604.944 -0.03895 -884.082 -0.04459 -12.5614 -0.0502
[-0.36230] [ 1.7567] [-1.26403] [ 2.21619] [-0.9266] [-0.56395]
∆Xt-3 44.04605 -0.04343 253.5908 -0.05052
-601.848 -0.03875 -880.749 -0.04442
[ 0.07318] [-1.1206] [ 0.28793] [-1.13726]
∆Xt-4
∆S&Pt-1 7.159053 -0.00024
1.108753 0.002821
-0.2578 -1.70E-05
-0.41935 -0.00168
[ 27.7604] [-14.5815]
[ 2.6439] [ 1.68363]
∆It-1 -31.77951 -0.0005 -33.56512 -0.00049 2.458713 -0.22732
-22.9543 -0.0014 -33.594 -0.00169 -38.5466 -0.15403
[-1.38447] [-0.369] [-0.99914] [-0.28692] [ 0.0637] [-1.47576]
53
A.9 cont. Standard errors in ( ) & t-statistics in [ ]
COUNTRY JAPANd UNITED KINGDOM
b UNITED KINGDOM
d
DEPENDENT VAR. ∆St ∆Xt ∆St ∆Xt ∆St ∆Xt
C -7.29372 -0.0364 0.332348 0.000211 -0.78965 0.000236
-8.3932 -0.0334 -2.68954 -0.0002 -3.2454 -0.00021
[-0.86900] [-1.090] [ 0.123] [ 1.0411] [-0.2433] [ 1.14461]
∆St-1 -0.12335 -0.0003 -0.03373 -2.54E-08 -0.07547 9.23E-07
-0.05033 -0.0002 -0.0337 -2.50E-06 -0.04058 -2.60E-06
[-2.45083] [-1.492] [-1.0010] [-0.01000] [-1.8598] [ 0.35784]
∆St-2 -0.06209 -0.0001 0.009414 -2.59E-06 -0.0392 -1.49E-06
-0.05041 -0.0002 -0.03389 -2.60E-06 -0.04077 -2.60E-06
[-1.23154] [-0.658] [ 0.2777] [-1.01579] [-0.9614] [-0.57293]
∆St-3 -0.01138 7.54E-06 -0.09210 3.48E-06 -0.07925 3.18E-06
-0.04972 -0.0002 -0.03349 -2.50E-06 -0.04041 -2.60E-06
[-0.22885] [ 0.038] [-2.7505] [ 1.37915] [-1.9611] [ 1.23873]
∆St-4
0.066199 -4.29E-06 0.083751 -4.69E-06
-0.0333 -2.50E-06 -0.04018 -2.60E-06
[ 1.9876] [-1.71124] [ 2.0843] [-1.83466]
∆Xt-1 38.05273 -0.0212 36.49851 0.122962 371.7634 0.11534
-12.6177 -0.0502 -529.939 -0.03988 -639.227 -0.04065
[ 3.01582] [-0.422] [ 0.0688] [ 3.08352] [ 0.5815] [ 2.83759]
∆Xt-2 4.990476 -0.0344 158.6251 -0.03386 204.9122 -0.03491
-12.7425 -0.0507 -533.329 -0.04013 -643.728 -0.04093
[ 0.39164] [-0.678] [ 0.2974] [-0.84362] [ 0.3183] [-0.85282]
∆Xt-3 -13.5009 -0.0330 555.0297 -0.03153 535.6102 -0.03109
-12.5973 -0.0501 -533.017 -0.04011 -643.358 -0.04091
[-1.07172] [-0.658] [ 1.0413] [-0.78604] [ 0.8325] [-0.75986]
∆Xt-4
-577.95 -0.08292 -449.606 -0.08584
-528.928 -0.0398 -638.363 -0.04059
[-1.0926] [-2.08333] [-0.7043] [-2.11461]
∆S&Pt-1
2.372273 -5.39E-05
-0.13527 -1.00E-05
[ 17.537] [-5.29847]
∆It-1 -2.06523 -0.2388 -4.78716 -0.0004 -12.1784 -0.00023
-38.6794 -0.1540 -10.9191 -0.00082 -13.1698 -0.00084
[-0.05339] [-1.549] [-0.4384] [-0.48983] [-0.9247] [-0.27995]
54
A.10, Granger Causality test results (2007-2010)
COUNTRY NULL HYPOTHESIS H0 f-stata f-stat
b f-stat
c f-stat
d critical f-value
INDIA ∆St does not Granger Cause ∆Xt 1.339273 0.885105 1.39795 1.752966 F(4,676) =2.37
∆Xt does not Granger Cause ∆St 36.24194 30.77113 31.3117 35.48433 F(4,676) =2.37
BRAZIL ∆St does not Granger Cause ∆Xt 15.86576 0.697377 0.650185 15.9489 F(3,675) = 2.6
∆Xt does not Granger Cause ∆St 2.560147 2.187855 2.423323 2.187855 F(3,675) = 2.6
S. KOREA ∆St does not Granger Cause ∆Xt 35.27671 32.21524 76.95274 35.11324 F(4,676) =2.37
∆Xt does not Granger Cause ∆St 1.939398 1.501437 1.885885 1.830983 F(4,676) =2.37
PHIL. ∆St does not Granger Cause ∆Xt 79.08529 79.34672 78.59098 79.80409 F(2,679) = 3
∆Xt does not Granger Cause ∆St 0.691745 0.721546 0.740637 0.682294 F(2,679) = 3
AUSTRALIA ∆St does not Granger Cause ∆Xt 0.538087 0.239587 0.215099 0.515350 F(1,681) =3.84
∆Xt does not Granger Cause ∆St 0.545830 2.669466 0.669773 0.542610 F(1,681) =3.84
CANADA ∆St does not Granger Cause ∆Xt 2.564655 1.947463 2.552379 2.603058 F(3,675) = 2.6
∆Xt does not Granger Cause ∆St 2.139583 1.38201 1.466687 2.125238 F(3,675) = 2.6
JAPAN ∆St does not Granger Cause ∆Xt 1.275765 3.147687 3.128548 1.258694 F(2,679) = 3
∆Xt does not Granger Cause ∆St 4.488609 7.11893 7.093467 4.327407 F(2,679) = 3
UK ∆St does not Granger Cause ∆Xt 1.509938 5.964761 5.590323 1.518974 F(4,676) =2.37
∆Xt does not Granger Cause ∆St 11.69298 0.353515 0.361415 0.305808 F(4,676) =2.37
a – no exogenous variables.
b - Including S&P and Interest difference.
c – Including S&P.
d- Including Interest difference