Cointegration and Causality Relationship Between Stock Market, Money Market and Foreign Exchange Market in Pakistan

1 Introduction

The dynamic relationship between stock market and money market is always remained an area of great interest for academicians, finance professionals and policy makers in both public and private sectors. Policy makers watch stock market movements very closely and make necessary adjustments in the interest rate to affect borrowing and investing decisions of investors. They use monetary policy as one of the most effective tools to regulate the stock market. The economic rationale of interrelationship between stock prices and monetary variables is primarily based on the efficient market hypothesis (EMH). According to EMH, asset prices reflect the arrival new information in the market about the securities. The reaction of asset prices to the changes in different economic variables in the market is not constant. Therefore, it is important to analyze the true reaction of asset prices to the changes in monetary variables like money supply, interest rate and exchange rate. Essentially, the indirect connection of asset prices to monetary variables is based on the expected cash flows and corporate earnings. Whereas, the corporate earnings reflect the behavior of macroeconomic variables, and any change in the macroeconomic variables will ultimately affect the profitability of the firms. Consequently, the asset prices will reflect the movements in macroeconomic variables that conduits through corporate earnings.

Specifically, the connection between money supply and stock prices can be viewed through inflationary expectations or real activity approach in the market. According to inflationary expectation approach propagated by Keynesian economists, an increase in money supply could have negative impact on stock prices due to high inflation rate in the economy that may lead to increase the interest rate. As a result, discount rates go up that would decrease the present value and expected cash flows of the firms that would further decrease the stock prices. Sellin[2] further argued that economic activities decline as a result of increased interest rate that would further depress the stock prices. On the other hand, as perpetuated by real activity theorists, an increase in money supply means that there is more demand for money in the market due to increased economic activities. High economic activity is positively linked with the profitability of the firms. Consequently, the stock prices increase to high corporate earnings. Hence, the Keynesian economists proposed a negative relationship between money supply and stock prices, while real activity theorists offered a positive connection between these two variables.

Although, the relationship between stock market and monetary policy instruments is interdependent, but it is a very complicated dilemma to deal with. There are quite mixed results on the linkage between stock market and monetary policy variables reported by the previous studies[3 - 5]. They found that central banks policy rate has negative relationship with the asset prices. It is further argued that investors make their rational choices to invest their money by comparing the rate of return that they will receive from both investing opportunities either in stock market or in money market. On the other hand, the studies of [6, 7 ], showed no causal connection between policy rates and asset prices. Similarly, the studies of [8–12] explored the relationship between money supply and asset prices, and majority of them found a positive linkage between money supply and asset prices due to liquidity effect.

There are two schools of thought regarding the relationship between stock market and monetary policy variables. First, the protagonists of proactive monetary policy[13 - 15], and the second are the antagonists of proactive monetary policy[¹⁶, ¹⁷ ]. The protagonists explain the fact that the booms and busts in the stock markets are very costly and create much distortion in the real economic activities, so, the policy makers and authorities should lower (increase) interest rates when assets prices are falling (rising) beyond the fundamental levels. Whereas, the antagonists of proactive monetary policy argue that it is very difficult to decide when to lower or increase the interest rates, and whether interest rate could stabilize the asset prices or not. Therefore, they propose a reactive approach to wait and see whether the asset prices reversal occurs or not, and then to react accordingly. In either case, it is quite possible that monetary policy and stock prices can have a close connection, and in one way or other, monetary policy can affect the asset prices.

In addition, exchange rate is also considered as one of the important macroeconomic factor that can strongly influence the asset prices. However, there are quite contrasting results reported in the previous literature on this nexus. Hasan and Javed[¹¹] reported a long-run equilibrium relationship between stock prices and exchange rate whereas Nieh and Lee[¹⁸] ; Smyth and Nanda[19] found a short term linkage between these two variables^[1]. Moreover, there are quite mixed results in the previous literature on the nature and direction of relationship between stock prices and foreign exchange rate. For example, the studies of [10, 11, 20] found a positive relationship between stock prices and exchange rate while [21, 22] reported a negative connection between these two variables. It is also widely reported in the literature that the causal relationship between exchange rate and stock market is either unidirectional[²³], or bidirectional[²⁴]. Nonetheless, some of the studies[²⁵, ²⁶] found no evidence of causal relationship between stock market and exchange rate.

On the basis of foregoing discussion, the current study focuses on the predominant causal connections between stock market, money market and foreign exchange market in Pakistan. In this study, KSE-100 index represent the stock market, interest rate and money supply represent money market, while exchange rate (PKR/USD) represent the foreign exchange market. The main motivation of this research is to analyze the linkages between stock market, money market and foreign exchange market under the different political regimes in Pakistan^[2]. The political history of Pakistan has always remained very vulnerable, and it is not a positive sign for both local and foreign investors. The results could be misleading if the policy makers ignore the past performance of financial markets during the previous politically instable governance regimes. Therefore, it is quite imperative to analyze the connection between stock market, money market and foreign exchange market under different political regimes (military regime and democratic government). We find quite varying results about the interrelationship between these three markets across different regimes.

This study is aimed to analyze the long term co-movement and causality linkage among money market, stock market and foreign exchange market in Pakistan. This study contributes in two different ways; first, it examines the long-run equilibrium relationship between stock market, money market and foreign exchange market under the different political regimes. Second, we also estimate long run and short run causality relationship between the stock price and other monetary variables included in this study under the three different sub-samples. The study finds significant differences in the relationship between stock prices, money supply, interest rate and exchange rate across the political regimes in Pakistan. Furthermore, as a robust check, we also estimate a multivariate linear regression model that justifies the nature of relationship with partial differences. This study will make a significant contribution in the existing literature on Pakistan and the results will be useful for policy makers and financial analysts in the field of economics and finance.

2 Literature Review

Money market and foreign exchange market are noted as the fundamental factors of stock market movement. The investigation about the relationship between these markets has remained an area of prime interest for the researchers and policy makers. A significant body of literatures has been put forward so for on the long-run equilibrium relationship and causal connection between stock market and macroeconomic variables. Humpe and Macmillan[29] explained that there can be two ways to establish the linkage between macroeconomic variables and stock prices. One way is Arbitrage Pricing Theory (APT) in which multiple risk factors are taken into account in order to explain asset returns. According to this approach, volatility in macroeconomic variables can be reflected in the underlying systematic risk factor that influences the future stock returns. Another approach used by Chen, et al.[30] is Present Value Model (PVM) framework that also explains the impact of macroeconomic variables volatility upon stock returns. The main advantage of this approach in comparison to APT is that it focuses on the long-run relationship between stock prices and macroeconomic variables. But most of the empirical work[31 - 36] are based on APT have modeled a short run relationship between stock prices and macroeconomic variables.

The empirical results about the causal connection between money supply, interest rate, exchange rate and stock market prices/returns are entirely muddled up. For example, Hashemzadeh and Taylor[37] found a feedback causal relationship from S&P-500 to money supply but they mentioned that money supply and interest rate are not the good predictors of stock market while Malliaris and Urrutia found a causal links between money supply and S&P-500, where money supply leads the stock market and stock market leads the economic activity. Mukherjee and Naka[10] also examined the long-run equilibrium relationship between stock market and macroeconomic variables (money supply, industrial production, exchange rate, inflation, long term government bond rates, and call money rate) for Japan. By employing vector error correction model (VECM) on monthly data, they found a significant evidence of positive connection between money supply and stock market. Similarly, many other studies[20, 38, 39] also reported a positive connection between money supply and stock prices. While discussing a causal relationship between money supply and stock prices, Shostak[40] argued that an increase in stock prices also provide an incentive to liquidate the fixed income securities and use that money to buy stocks and other financial assets. Thus, the demand deposits will increase that in turn increases the money supply. This trend can be reversed if stock prices fall. He further argues that causality cannot be achieved only by statistical figures without having a coherent definition of what money is and how it is related to stock prices and other financial assets.

In case of Pakistan, Hasan and Javed[11] established a long-run equilibrium relationship between monetary variables (money supply, treasury bill rate, consumer price index, and exchange rate) and equity prices by employing multivariate cointegration approach on monthly data covering from 1998:M6 to 2008:M6. They also employed granger causality test and VAR model (impulse response functions) to analyze the short term causal relationship between the selected variables. The results of multivariate Johansen and Juselius[1] cointegration test indicate a long-run equilibrium relationship between variables, while granger causality test indicates a unidirectional causality moving from monetary variables to equity prices. Moreover, the results of impulse response function are evident of positive relationship between money supply and equity prices while it is negative for interest rate and exchange rate. By using quarterly data, Abbas and McMillan[12] also established a long-run equilibrium between stock market index and monetary variables for Pakistan. Humpe and Macmillan[29] among others reported a negative relationship between stock prices and interest rate for Japan and US. Conversely, Ratanapakorn and Sharma[20] found a positive relationship between short-term interest rate and stock prices for USA. Some of the studies[6, 7, 41] found no causal connection between stock prices and interest rate.

In case of exchange rate, the studies of [10 - 12,20] among others found a long-run equilibrium relationship between exchange rate and stock prices. For feedback causal connection between exchange rate and stock prices in US, Bahmani-Oskooee and Sohrabian[42] found that there is a bidirectional causality between exchange rate and stock prices in short run, whereas, Choi, et al.[43] and Kim[21] claimed a negative and unidirectional causality relationship between exchange rate and stock prices that move from exchange rate to stock prices. Dominguez and Tesar[44] also mentioned that exchange rate fluctuations have a significant impact on the equity prices at firm level and sectoral level for developed economies. Similarly, Chkili and Nguyen[45] examined the relationship between exchange rate and stock returns for BRICS countries by employing regime switching approach. They discovered that exchange rate movement has no effect on stock returns while stock returns have a significant impact of exchange rate changes except South Africa. Lee, et al.[46] examined the relationship between exchange rate and stock prices for Asia-Pacific economies by employing dynamic correlation approach. They found that the correlation between stock market and exchange rate becomes higher as stock market volatility increases. For the same countries, Yang, et al.[47] reported a negative correlation between exchange rate and stock market. Similarly, Moore and Wang[22] also found a negative relationship between exchange rate and stock prices. Hasan and Javed[11] studied the causality relationship between these two variables and found a positive and unidirectional causality moving from exchange rate to stock prices in short run. Based on the literature review and the objectives of this study, we investigate the long-run equilibrium relationship between stock market, money market and stock market in case of all three sample periods. Moreover, we also investigate the short run granger causality relationship between all three markets in full sample period, military regime and democratic regime.

3 Data and Descriptive Summary

Monthly data covering from 2000:M1 to 2015:M12 has been used in this study to investigate the relationship among stock market, money market and foreign exchange market in case of Pakistan. KSE-100 index has been used as a proxy measure for stock prices in the in Pakistan 90 days T-bill rate and M2 are used as a measure of short term interest rate and money supply. By using monthly data, Wong, et al.[48] also studied the cointegration relationship between interest rate, money supply and stock prices for USA and Singapore. The exchange rate (PKR/USD) is used as an explanatory variable in this study in analyze the impact of external shock on local markets[49] among others uses monthly data for the emerging Asian economies to investigate the relationship between stock prices and exchange rate. The data for all these variables has been collected from the CEIC global database^[3]. In order to stabilize the data, all the series are used in the log form except T-bill rate and their time series plots are shown in Figure 2. Maysami, et al.[50] have also used monthly data for Singapore to analyze the relationship between macroeconomic variables and stock market indices, and they found that it gives more robust results than quarterly or daily data.

Figure 1

Time series plot of KSE-100 index, interest rate, money supply and exchange rate

Figure 2

Time series plots of cumulative sum of recursive residuals for vector error correction models (VECM)

The results of descriptive statistics are presented in Table 1. These statistics explain the basic features of the logged variables like mean, standard deviations, minimum and maximum, skewness and kurtosis and Jarque-Bera test-statistic. The results show that the value of standard deviation for KSE-100 index is greater than the other three variables which indicate that Pakistan stock market is quite risky as compared to other macroeconomic variables. The results also indicate that except KSE-100 index, all the other variables are negatively skewed and kurtosis result shows that the distribution of all variables is fatter tailed Jarque-Bera test-statistic for all the variables is significant that indicates that all the variables are not normally distributed.

4 Methodology

In the time series data, presence of long-run equilibrium relationship between non-stationary variables is one of the core concerns of economists. Granger[51], and Engle and Granger[52] introduced the concept of “cointegration” to capture the relationship of such nature. Stock[53] investigated that if two series say Y_t and X_t are non-stationary and highly cointegrated, then they would produce highly efficient and consistent estimates of the parameters. If in the time series data two variables have common trend, then they are more likely to have long-run relationship between them. So in order to check the nature of trend in variables, cointegration tests are important. A significant body of literature [48,54 - 56] discussed the details of this concept of cointegration. In order to analyze the long term co-movement of KSE-100 index, money supply (M2), Interest rate (6 months T-bill rate) and exchange rate (PKR/USD), a variety of cointegration tests may be used. In this study, we use a multivariate approach developed by Johansen and Juselius[1, 57] to capture the equilibrium relationship between stock market index and financial economic variables.

The first step in the empirical analysis is to check the stationarity of variables by using unit root test procedure developed by Dickey-Fuller[58, 59 ]. Any of the series, say Z_t is said to be integrated of order “d” if it has stationarity and stochastic ARMA at differencing “d” times and it is denoted by Y_t = I(d). Most of the non-stationary financial variables are integrated of order I(1), but if they are stationary and integrated at zero difference then they are denoted by I(0). Suppose there are two variables say, Y_t and X_t are integrated at I(1) then the regression equation (1) is used to estimate the long-run equilibrium relationship between variables, and further stationarity tests are used to test the stationarity of estimated residuals. If the variables are not cointegrated, then the estimated residuals will be integrated at order I(1), otherwise the residuals will be stationary and integrated at order I(0). Moreover, the regression results of non-stationary series will be spurious and misleading and the residuals will not be I(0) in such conditions.

here in Equation (1)e_t is the estimated residuals of these time series variables and two most common tests of stationarity Augmented Dickey-Fuller (ADF) and Phillips-Perron (PP) are used to check the stationarity of these estimated residuals

As we have more than two variables in the model, so, there is a possibility of having more than one integrating vector. It means that the variables in the model might form several equilibrium relationships governing the joint evolution of all the variables. In general, for n number of variables we can have only up to n–1 cointegrating vectors. Therefore, when n = 4, then this is very serious problem which cannot be resolved by simple single equation approach. Therefore, an alternative approach is needed and this is multiple equations approach developed by Johansen and Juselius[1].

This approach can be presented by extending the single equation error correction model to a multivariate one. Suppose, there are three endogenous variables, say W_t, X_t and Z_t then we can have (using matrix notation for Y_t = [W_t, X_t, Z_t])

Equation (2) is comparable to the single equation dynamic model (1) of two variables of Y_t and X_t. Further, Equation (2) can be reformulated in a vector error correction model (VECM) as follows:

where Γ_i = (I – A₁– A₂ – ··· – A_k) (i = 1, 2, ···, k – 1) and Π = – (I – A₁– A₂ – ··· – A_k). Here u_t’s are assumed to be identically distributed and independent as a multi-normal distribution with zero mean and δ variance. As we have assumed that there are three variables in Y_t = [W_t, X_t, Z_t], so, we need to examine 3 × 3 Π matrix. In Equation (3), Π matrix contains information about the long-run equilibrium relationship. The information matrix (Π) can be decomposed into Π = αβ where α includes the speed of adjustment to achieve the long-run equilibrium level while β is the long-run matrix of coefficients. Therefore, the βY_t-1 term is equivalent to error correction term (Y_t-1 - β - β₁X_t-1)in case of two variables model, except the fact that here βY_t-1 contains up to (n – 1) vectors in a multivariate framework. For simplification reasons, we assume that there are only two lags (k = 2), the model can be presented as follows:

Equation (4) can also be written as follows:

In Equation (5) the error correction part (i.e., for ΔW_t on LHS) can be analyzed by the following equation:

where Π₁ is the first row of the Π matrix, and the above Equation (6) can also be written as follows:

Equation (7) clearly shows the cointegrating vectors with α₁₁ and α₁₂ terms of their respective speed of adjustment.

In order to test the rank of Π matrix, Johansen[57] developed a reduced-rank regression procedure. Johansen and Juselius[1] explained two methods that could determine the number of cointegration equations and both involve estimation of Π matrix. The first method proposed by Johansen is maximum eigenvalue statistics denoted by λ_max. The test consists of ordering the largest eigenvalue in descending order and considering whether they are significantly different from zero. To test how many of the numbers of characteristics roots are significantly different from zero this test uses the following statistics:

The second method proposed by Johansen is based on likelihood ration test about the trace of the matrix and that is why, it is known as trace test. The trace test considers whether the trace is increased by adding more eigenvalue beyond the rth eigenvalue. The null hypothesis in this case is that “the number of cointegrating vectors is less than or equal to r”. This test-statistic value is calculated by

Critical values for both statistics are provided by Johansen and Juselius[1]. Once it is decided that the long-run relationship between stock prices index and financial economics variables exists, then we adopt the bivariate VAR model to estimate granger causality relationship. For further investigation in this pairwise granger causality, two-way causation of variables, say X and Y can be tested. The following bivariate regression is used where X granger causes Y and Y granger causes X

The calculated F-statistics is used to accept or reject the hypothesis. The first regression equation (10) is used to test the hypothesis “macroeconomic economic variables, say X does not granger cause the stock prices index say Y” and the second regression equation (11) is used to test the hypothesis “stock prices index does not granger cause the macroeconomic economic variables”.

5 Empirical Estimation

6 Concluding Remarks

The cointegration and causality relationship between stock market, money market and foreign exchange market has been established in this study. Unlike the other relevant studies[11, 12, 64] in Pakistan perspective, we have decomposed the sample period into two sub-samples in order to analyze the significance of good governance in the integration of financial markets in Pakistan. The first sub-sample represents the military regime, whereas the second sub-sample represents the democratic era. Our empirical analysis is based on the multivariate Johansen and Juselius[1] cointegration test, vector error correction model (VECM) and pairwise granger causality tests by using monthly data from 2000:M1 to 2015:M12.

A long-run equilibrium relationship between KSE-100 index and other financial variables is identified by trace test-statistic under the multivariate Johansen-Juselius cointegration analysis for the full sample time period. The coefficients of normalized cointegrating vector show that KSE-100 has positive relationship with M2, while negative with TBR and ER during the full sample and two sub-sample periods only with one exception of ER during the military regime that has a positive connection with the KSE-100. Furthermore, the results of vector error correction model (VECM) for KSE-100 index and other financial variables indicate that there is a significant evidence of long-run causality moving from independent variables to dependent variable in the system of VAR during the full sample periods. Finally, the results of pairwise granger causality test ascertain that M2 does not granger cause KSE-100 and vice versa during the full sample period while there is a unidirectional causality from M2 to KSE-100 during the military regime, and bidirectional causality during the democratic regime. Similarly, there is a unidirectional causality moving from TBR to KSE-100 index during the full sample period and military regime while no causal linkages are identified during the democratic regime. In case of exchange rate, there is a unidirectional causality moving from ER to KSE-100 during the two sub-sample periods while it has no causal connections during the full sample period. Moreover, a strong causality relationship among the other financial variables (M2, TBR and ER) is also witnessed by the results of granger causality tests during all the sample periods.

The overall results are evident that all the three markets are strongly integrated during the full sample period while the results for the two sub-sample periods are more favorable for the democratic regime. The dependence of stock market is more evident during the both military and democratic regime and for the remaining financial variables.

The results of long term causal relationship are more favorable during the democratic regime. It means that markets during the democratic regime are more interdependent as compared to military regime. Only stock market reflects the movement in other financial variables during both regimes otherwise the markets show more close interaction during the democratic regime as compared to military regime. The stock market is strongly driven by the money market and foreign exchange market and vice versa. Therefore, it is important for the investors, portfolio managers and financial analysts to make appropriate decisions by keeping in mind the behavior of money market and foreign exchange market during the different political governance systems. In the same way, the monetary policy makers and regulatory authorities should also set the monetary policy and market regulations accordingly. Based on the results of granger causality, it is important to consider the behavior of exchange rate in order to devise a well-directed internal policy regarding stock market and money market that could have the flexibility to absorb these external shocks. As a matter of fact, when markets are strongly cointegrated they can be much sensitive to the behavior of each other and such sensitivity can leave some adverse effects if the dynamics of these markets are not dealt properly.