The main purpose of the dissertation is to analyse the relationship between exchange rates and stock prices in selected Asian markets, namely Japan, Singapore, Hong Kong, Malaysia, Indonesia, Philippines and Thailand over the period from 2001 to 2011, incorporating the impacts of the 2007 financial crisis.
With the significant growth of international trade and the increased integration in financial markets, exchange rate has been considered one of the main factors influencing the business profitability and stock prices (Kim, 2003). Besides, studying the relationship between exchange rates and stock prices is necessary as it has both theoretical and practical significance (Hatemi-J and Roca, 2005). Theoretically, this relationship is an essential input to open macro-economy models as well as hedging models; practically, when understanding the linkage between stock and currency prices, investors are able to diversify the portfolios and hedge their investments (Hatemi-J and Roca, 2005). Given this, the exchange rate-stock price relation has received much attention in the area of empirical finance for the past few decades.
Empirical studies to verify the relationship between exchange rates and stock prices have been carried out since 1970s; yet, the conclusions have been mixed (Alagidede et al., 2011). Moreover, more attention has been paid to explore the exchange rate-stock price relation under the crisis period because it is the time when policymakers may intervene to change exchange rates; thus, it is critical to understand how changes in foreign exchange market will affect other markets (Hatemi-J and Roca, 2005).
Therefore, this dissertation will attempt to contribute to the literature by examining the interaction between stock prices and exchange rates in selected Asian countries under both the normal conditions and the crisis period. The dissertation aims to answer the following research questions:
First, is there any long-run relationship between exchange rates and stock prices in selected Asian markets?
Second, is there any causal relationship between exchange rates and stock prices? If yes, what is the direction, the sign and the magnitude of this causal relationship?
Third, how has the 2007 financial crisis affected the exchange rate-stock price relation in chosen markets?
And fourth, is there any difference in the interaction between stock and currency markets in Asian developed and emerging markets?
To answer these above questions, we examine the long-run and short-run association between exchange rates and stock prices. First, the long-run relationship is explored using the cointegration tests, in which both the conventional and the more improved test taking account of a structural break are employed. Second, the short-run relationship and the direction of causality are tested using the Granger causality tests in the vector autoregressive (VAR) model or the vector error correction model (VECM). Besides, we also use the impulse response functions (IRFs) and forecast error variance decompositions (FEVDs) to explore the sign and magnitude of the relationship between exchange rates and stock prices.
The structure of the dissertation is provided in the next section.
Structure of the study
The rest of the dissertation is structured as follows:
Chapter 2 presents the theoretical background for the dissertation by discussing possible explanations for the exchange rate-stock price relation and briefly reviewing the methodologies used in previous studies. This chapter also highlights the related empirical studies concerning both the international and the Asian markets.
Chapter 3 explains the data set used in this study; specifically, how the data are collected and the preliminary statistics of the data series.
Chapter 4 discusses the detailed methodology, including the concept and mathematical specification of the models employed. The methods to estimate the parameters and to give conclusions on these tests are also described.
Chapter 5 generalizes the empirical results of the various tests mentioned in chapter 4 and discusses the findings on the relationship between exchange rates and stock prices in selected markets.
Chapter 6 draws conclusions on this study by summarizing the objectives and general findings. Moreover, limitations, suggestions and recommendations from this study are also provided.
Chapter 2: Literature Review
This chapter will review the theoretical background and empirical studies on the interaction between exchange rates and stock prices, which has received much attention for the past 40 years. Specifically, the chapter is organized as follows:
Section 2.1 discusses theoretical approaches explaining the existence of the linkage between exchange rates and stock prices. Besides, common models used in the literature to evaluate this relationship are presented in the second part of this section.
Section 2.2 reviews related empirical studies conducted in both international markets and Asian markets. These studies have reported that there is a significant correlation between exchange rates and stock prices but the conclusions have not yet reached a consensus. These studies will be used as references to compare with the results of this study in the subsequent chapters.
Finally, section 2.3 summarizes the chapter.
Theoretical Background
Theoretical links between exchange rates and stock prices
This section concentrates on providing possible theoretical explanations for the presence of the relation between stock prices and exchange rates. The current literature suggests a number of approaches to explain the link between exchange rates and stock prices; yet, no agreement has been made on which approach is superior. Therefore, it is necessary as the first step of our analyses to present the two common approaches, the traditional approach and the portfolio balance approach, to interpret the relationship between exchange rates and stock prices.
Traditional approach
The traditional approach (Dornbusch and Fisher, 1980) represents that there is a causal relationship between exchange rates and stock prices which runs from the former to the latter. Based on the efficient market hypothesis, the traditional approach regards stock prices as the present values of a firm's expected future cash flows. Besides, from the macroeconomic perspective, it is assumed that a country's current account is an important determinant of exchange rates. Movements in exchange rates will affect the country's real economic variables such as income and output; then in turn, influence the firm's competitiveness, earning and net worth (Stavarek (2005)). Consequently, exchange rate fluctuations affect the firm's cash flows and will be reflected in the firm's stock prices (Richards et al., 2009).
Moreover, the way exchange rates influence stock prices depends on the characteristics of the companies. When the direct quotation is employed and the firm is a net exporter, there is a positive causal relationship which can be interpreted as follows: depreciation in the domestic currency will make local exporters more competitiveness as their exports become cheaper. Higher exports lead to higher incomes and thus, increase stock prices (Granger et al. (2000), Stavarek (2005)).
Portfolio Balance Approach
The portfolio balance approach, on the other hand, suggests a negative relationship between exchange rates and stock prices and the causality runs from stock prices to exchange rates. According to this approach, exchange rates' role is to balance the supply and demand of domestic as well as foreign assets. When there is an increase in domestic stock prices, investors tend to buy more domestic assets while simultaneously selling foreign assets, leading to the local currency appreciation or a decrease in exchange rates (Stavarek (2005), Pan et al. (2007)).
Furthermore, another channel that stock price movements can affect changes in exchange rates is through investors' wealth and money demand, which rely on stock markets' performance (Climent et al., 2004). During hard times, stock prices decrease and cause a reduction in the domestic investors' wealth. Demand for money is lowered and thereby, interest rates decrease. Hence, capital outflows may rise, resulting in the currency depreciation or a growth of exchange rates (Climent et al. (2004), Pan et al. (2007)). Again, this shows the negative linkage between stock prices and exchange rates.
It is also noted that how exchange rates react to changes in stock prices depends on the segmentation and liquidity of stock markets. If stock markets are highly segmented and illiquid, high transaction costs will make it difficult to buy or sell stocks and large foreign currency exposure will hinder foreign investments (Richards et al., 2009).
In short, from theoretical view, stock markets and foreign exchange markets can interact in various ways. The lack of agreement between these above approaches gives rise to empirical studies to investigate the interaction between exchange rates and stock prices. In this dissertation, the approaches to explain the relation between stock prices and exchange rates will be examined using various models discussed in the subsequent parts.
Review of models used to investigate the relationship between exchange rates and stock prices
In order to investigate the relationship between exchange rates and stock prices, this dissertation focuses on the directions of causality between exchange rates and stock prices using tests in the VAR framework, including the cointegration tests and the Granger causality tests. Yet, before describing these tests in details, it is necessary to get the overview of the common models used in the literature to examine the interaction between stock markets and foreign exchange markets.
In this section, general descriptions of the common models employed by previous studies are reviewed in the sequence of time: first, the early regression models and second, the tests in the VAR framework. It is noted that this section just briefly introduces the models while the detailed specification of the models used in the dissertation will be discussed in the following chapters.
Early regression models
Early empirical works concentrate on the contemporaneous relation between stock prices and exchange rates using the correlation and regression models (Stavarek (2005), Rahman and Uddin (2009)). In these regression models, either stock returns or exchange rate movements are employed as the dependent variable; the other is considered the independent variable.
According to the traditional approach, exchange rates lead stock prices; hence, the relationship is as follows:
(2.1)
Where: is the rate of return on firm i's common stock, is the rate of variation of exchange rate at time t (t =1,…,T) and is the error term.
Source: Jorion (1990)
On the other hand, following the portfolio balance approach, stock prices cause exchange rates with a negative correlation:
Where is the exchange rate changes, is the stock return differential (domestic minus foreign), and is the variation of interest rate differential.
Source: Solnik (1987)
Although the correlation and regression analysis are easy to adopt, they have some limitations. First, regression analysis strictly requires the assumption of stationary in order to avoid spurious results (Stavarek, 2005). Moreover, due to the stationary requirement, the first-differenced data are usually used instead of the levels. However, by differencing the variables, some information about the correlation between the levels of variables may be lost, which then changes the relationship between them (Stavarek, 2005). Besides, Richards et al. (2009) quoted Brooks (2002) that standard regression framework only assumes the contemporaneous relationship between dependent and independent variables. Hence, standard regression analysis is not adequate to imply the causal relationship (Richards et al., 2009).
Due to these limitations, later empirical studies have adopted more sophisticated econometric methodologies, which are the cointegration tests and the causality tests to investigate the causal relationship between exchange rates and stock prices.
Vector autoregressive (VAR) models
This dissertation use models in the VAR framework to inspect the causal relationship between exchange rates and stock prices. The VAR models have been known to be helpful in capturing and forecasting the dynamic behavior of the time series (Lütkepohl, 2011). As there are different forms of the VAR models, this section concentrates on the VAR model in reduced form and introduces the concept of the cointegration and causality tests.
According to Richards et al. (2009), a VAR model is required to investigate the causal relationship between variables because the causal analysis relies on the assumption that the current value of one variable may be affected by the lagged values of other variables. The VAR model incorporates the lagged values of variables as follows:
Where is a k-dimensional vector of a time series (; are matrices of coefficients; is a k-dimensional vector of the white noise process and .
Source: Lütkepohl, 2011
This model is known as the VAR model in reduced form because all the variables in the right-hand side are predetermined (lagged values) (Lütkepohl, 2011). The VAR models provide interpretations about the dynamic relationship between involved variables, one of which is the causal relationship proposed by Granger in 1969 (Lütkepohl, 2011). However, before going to the Granger causality test, it is important to note that one requirement of the VAR model is the time series are stationary. As a result, empirical studies have adopted the cointegration analysis to deal with non-stationary time series before applying the Granger causality test.
The idea of cointegration is postulated by Engle and Granger (1987) and tested by a two-step cointegration procedure as follows:
Step 1: Suppose two time series are integrated of order one (), which are non-stationary in levels but stationary in first differences. We run the following regression:
Where is the cointegration parameter and is the residual obtained from the cointegrating regression.
Step 2: Apply the augmented Dickey-Fuller (ADF) unit root test on the residual obtained from (2.4).
If is found to be stationary, then are said to be cointegrated. The presence of cointegration suggests a long-run relationship between the variables and any deviations from this long-run equilibrium relationship will be adjusted.
Although the Engle and Granger cointegration test is easily conducted, it has some shortcomings (Shirvani and Wilbratte, 1997). Firstly, the results depend on which variable, , is chosen as the dependent variable. If the sample size is large, it is suggested that the unit root results could be similar on the residuals obtained from regressing and from regressing . However in many cases, the sample size is smaller than the required size and thereby, the unit root results are different. Secondly, this test relies on a two-step procedure; thus, errors in the first step are brought to the second step, making the results untrustworthy. Furthermore, when several cointegrating vectors exist, the Engle-Granger approach may produce a linear combination of these vectors (Shirvani and Wilbratte, 1997).
Another typical method to test for cointegration is the Johansen-Juselius cointegration test, which overcomes the drawbacks of the Engle-Granger method. The Johansen-Juselius method is a one-step estimation allowing for multiple cointegrating vectors. This study will adopt the Johansen-Juselius test and another advanced cointegration test, known as the Gregory-Hansen cointegration test to examine the long-run relationship between exchange rates and stock prices.
After examining the long-run relationship by using the cointegration analysis, the Granger causality test is performed to explore the short-run linkage and the causal relation between variables. The Granger causality test indicates a time series Granger-causes another time series if is forecasted by using lagged values of not only but also (Pan et al., 2007). When there is no cointegration, Granger (1969) suggested the standard causality test for two variables:
Where are two stationary time series; p is the number of lags; are two uncorrelated white-noise series. The conclusion on the causal direction depends on whether the coefficients and/or are significant or not.
Nevertheless, Engle and Granger (1987) argued that the standard Granger causality test is suitable only when there is no long-run equilibrium relationship. Otherwise, the Granger test should include the error-correction terms obtained from the cointegration test to capture the short-run dynamics needed to get back into the long-run equilibrium. This is known as the vector error correction model (VECM).
This dissertation will adopt both the Granger causality test in VAR model and the VECM to explore the causal relationship between exchange rates and stock prices.
Empirical studies
This section aims to review the empirical studies on the relationship between exchange rates and stock prices. Although there is a significant literature on the association between foreign exchange markets and stock markets, the conclusions and findings are still mixed, even for the same markets. Besides, as this dissertation chooses Asian markets as target markets, this section divides empirical studies into two groups: studies consider the international markets and studies concern the Asian markets. It makes clearer to compare and contrast the results of this study with past research in the next chapters.
Studies on international markets
Empirical research on the international markets will be reviewed based on the approaches mentioned in section 2.1.1, which are the traditional approach and the portfolio balance approach. In fact, there is no consensus on which approach is favoured as there are some studies confirming either the traditional or portfolio balance approach; some studies propose both approaches while others find no interaction between stock markets and foreign exchange markets. Besides, it is also clear that early studies on the exchange rate-stock price relation focused more on mature markets and then, the interest has switched to other new developed and emerging markets.
Firstly, a number of early studies reported no relationship existing between exchange rates and stock prices. The first research on the exchange rate-stock price relation was performed by Franck and Young in 1972. Employing the correlation regression analyses, they found no interaction between stock prices and movements in exchange rates. Similarly, Ang and Ghallab (1976) examined the effect of the US dollar devaluation on the multinational companies' stock prices and found no relationship.
Solnik (1987) regarded stock returns as changes in economic activity and test the relation between economic activity and exchange rates. Monthly and quarterly data from July 1973 to December 1983 were employed on eight selected developed countries. The multivariate regression and cross correlation analysis were adopted. The study indicated changes in exchange rates had a low power in explaining stock returns. Jorion (1990) followed the study by Solnik (1987), testing the relation between the US dollar effective exchange rate and the US multinational companies' stock returns from 1971 to 1987 and reported a modest relationship.
Stavarek's study in 2005 collected monthly data of exchange rates and stock indices in the USA as well as in four old and four new EU-member countries to examine the relationship. The author employed the cointegration and the Granger causality tests and found it not possible to register any long-run or short-run relation between stock prices and exchange rates.
On the other hand, some studies have backed up the traditional approach. For example, Aggarwal (1981) studied the contemporaneous relationship between exchange rates and stock prices in the USA, using regression models and monthly data for the period from 1974 to 1978. This study reported a positive correlation and was consistent with the traditional approach.
Ma and Kao (1990) employed monthly data from January 1973 to December 1983 and investigated how stock prices interact to exchange rate movements in six industrialized countries: the UK, France, Canada, West Germany, Italy and Japan. The findings supported the traditional approach that the relationship depended on whether the economy was strong at exports or imports. A local currency appreciation negatively influenced stock prices if the country had advantages at exports and vice versa if it was a net-importer.
Besides, the portfolio balance approach is also reinforced by the study of Najang and Seifert in 1992 when they investigated the relationship between interest rate differential, exchange rate volatility and stock return. They employed daily data from the USA, Canada, the UK, Germany and Japan to conduct the GARCH models and indicated that absolute changes in stock returns and interest rate differentials positively influenced exchange rate movements.
Nevertheless, the majority of previous studies have confirmed both the traditional and portfolio balance approach. For instance, Ajayi and Mougoue (1996) applied the causality test in VECM to examine the relation between stock prices and exchange rates in eight industrial economies from 1985 to 1991. The results showed that an increase in stock prices had a negative short-run but a positive long-run influence on currency values. On the other hand, currency depreciation affected stock prices in a negative way in both the short run and long run.
Tabak (2006) studied the dynamic linkage between stock prices and exchange rates in Brazil. The data set contained the daily IBOVESPA index price and the exchange rate from August 1, 1994 to May 14, 2001 (Tabak, 2006). To test the connection between the variables, the study employed the improved Gregory-Hansen cointegration test, which allows for a structural break in the time series, the linear and nonlinear Granger causality tests. No long- run relationship was detected but a linear causality from stock prices to exchange rates and a nonlinear causality from exchange rates to stock prices were found, which supported both the traditional and portfolio balance approach.
Recently, Alagidede (2011) investigated the causal relationship between stock prices and exchange rates in Australia, Canada, Japan, Switzerland, and the UK from January 1992 to December 2005. The study detected no long-run relationship between exchange rates and stock prices. Considering the short-run relation, there was a causal linkage from exchange rates to stock prices in Canada, Switzerland, and UK while the causal relation from stock prices to exchange rates was only found for Switzerland. Moreover, the nonlinear causality test indicated causality from stock prices to exchange rates in Japan.
Parsva and Lean (2011) carried out the research on six Middle Eastern countries before and during the 2007 financial crisis. Monthly time series data were used and the research period was from January 2004 to September 2010. In line with most previous studies, the paper employed the cointegration and Granger causality tests. The result showed that the interaction between the financial markets grew during the financial turmoil. Before the crisis, bi-directional causality was found only for 3 countries while it was observed for all countries except for Iran during the crisis.
In short, the results of the previous studies on the international markets are mixed even when similar models or markets are employed. This evidence is also shown for the previous studies on the Asian markets, which are reviewed in the following part.
Studies on Asian markets
The relationship between exchange rates and stock prices in the Asian markets has received more attention since the 1997 crisis when both the stock and currency markets experienced severe deterioration. Just like the international markets, studies on the Asian markets also provide conflicting conclusions.
Yu (1997) chose leading East Asian markets: Singapore, Hong Kong and Tokyo to conduct the research. The data were daily data of spot exchange rates and stock indices over the period from January 3, 1983 to June 15, 1984. The author also applied the Granger causality test and found bi-directional causal relationship in Tokyo. In Hong Kong, only changes in exchange rates Granger caused changes in stock prices while in Singapore, no causality was detected. Additionally, a stable long-run correlation between the levels of variables was affirmed for all markets.
Using the same methodology as Yu (1997), Ajayi et al. (1998) tried to examine the difference of the association between exchange rates and stock prices in both developed and developing countries. The study took daily data in seven advanced markets from 1985 to 1991 and eight Asian developing markets from 1987 to 1991. Regarding advanced markets, the relationship was found consistent with the portfolio balance approach. Regarding emerging markets, non-significant causal connection was reported in Hong Kong, Singapore, Thailand and Malaysia. According to the research, the differences of the exchange rate-stock price relation in these countries were due to the structural differences in each country's stock and currency markets. Specifically, the relationship in emerging countries appeared to be weaker because emerging countries had less stable political environment and therefore, they were smaller, less attractive than advanced countries (Richards, 2009).
Granger et al. (2000) was one of the studies concentrating on the Asian turmoil in 1997. The data set were daily data from January 3, 1986 to June 16, 1998 for nine Asian countries suffering from the financial crisis. To better dissect the relationship, three sub-periods were chosen as there were clear evidences of structural breaks in the time series: the 1987-crash period, the after-crash period and the Asian crisis period (Granger et al., 2000). The main methods used were the Granger causality test and the impulse reaction analysis. During the first period, all countries except for Singapore had a weak stock price-exchange rate relation. The second period again showed exchange rates caused stock prices in Singapore while in Taiwan and Hong Kong, the result was opposite. Given the impact of the 1997 crisis, the result in South Korea was consistent with the traditional approach whereas the result from Philippines market was as per the portfolio balance approach (Granger et al., 2000). Except for Indonesia and Japan revealing no relation, the rest markets reported significant feedbacks.
Muhammad and Rasheed (2002) used monthly data of stock prices and exchange rates in India, Pakistan, Siri Lanka and Bangladesh from 1994 to 2000. Bangladesh and Siri Lanka's markets appeared to have long-run bi-directional causality while no relation was detected for the other two countries. Employing the same countries for the period 1995-2001, however, Smyth and Nandha (2003) indicated no long-run relation between these variables. In short-run, changes in exchange rates caused that in stock prices in India and Siri Lanka.
Beer and Hebein (2008) explored the interaction between stock prices and exchange rates using EGARCH model for two groups: four developed countries (USA, Canada, UK, Japan) and five emerging ones (Hong Kong, Singapore, South Korea, India, Philippines). Using weekly closing prices and exchange rates, the results indicated positive price spillovers from currency markets to stock markets in Japan, Canada, USA, India and South Korea. Furthermore, a significant persistence of volatility in both markets was affirmed for the emerging markets but not for the industrialized countries.
From the review of empirical studies, it can be seen that neither the traditional nor the portfolio balance approach is confirmed to best explain the exchange rate-stock price relation. The direction of causality and the degree of interdependence between these two markets are also controversial. This fact, therefore, gives motivations to conduct further investigation on the relationship between exchange rates and stock prices.
Summary
Chapter 2 provided the theoretical background for the dissertation and reviewed the literature in order to build the foundation for empirical analysis in the following chapters. Also, the concept, advantages and issues of the common models used in the literature were discussed.
From the literature review, we can draw some points: Firstly, although past researches and studies have greatly added to the understanding of the exchange rate-stock price linkage, the results are diversified given the different choices of markets, study periods and models. Secondly, regarding the Asian markets, while there are a number of studies interested in exploring the influence of the 1997 Asian crisis on the interaction between stock and currency prices, the impact of the recent global crisis in 2007 has not been comprehensively examined. Thirdly, there are still some issues with the conventional tests that need to be solved so as to give reliable results.
Therefore, choosing seven Asian markets, the dissertation will contribute to the literature by exploring the causal relationship between exchange rates and stock prices using updated data from 2001 to 2011, combining the impact of the 2007 global financial crisis. Furthermore, the dissertation will employ more advanced tests to solve some of the issues mentioned in the conventional tests discussed in section 2.1.2. The following chapters will discuss in details the data set and the methodology employed in this dissertation.
