THE DYNAMIC RELATIONSHIP BETWEEN STOCK MARKET RETURNS AND MACROECONOMIC VARIABLES: AN EMPIRICAL STUDY FROM BANGLADESH

In this research paper, attempt has been made to explore the dynamic relationship between stock market and macroeconomic variables i.e. DSE index and three key macro-economic variables (Exchange rate, Industrial production in and Reserve), by using unit root stationary tests and Johansen co-integration test. Monthly data has been used from June, 2003 to June, 2015 for all the variables, like, DSE index, Exchange rate, Industrial production in and Reserve. Results showed that the variables contained a unit root and were integrated of order one. The vector error correction model (VECM) (Johansen (1991)) is utilized to determine the impact of selected macroeconomic variables on stock market. Empirical results show that the stock market and macroeconomics variables have no long-term equilibrium relationship.


Introduction
The dynamic relationship between stock market returns and macroeconomic variables has been widely investigated, especially in developed markets. The early studies on the US stock markets by Lintner (1973), Bodie (1976), Jaffe and Mandelker (1977) and Fama and Schwert (1977) mainly examined whether the financial assets were hedges against inflation. These studies have reported a negative relation between stock returns and changes in the general price level. However, Fama (1981) found a direct positive relationship between stock market returns and real economic activities such as industrial production. Chen et. al. (1986), tested whether a set of macro-economic variables explained unexpected changes in stock market returns. It is recognized that stock markets play a pivotal role in growing industries and commerce of a country that eventually affect the economy. The importance of the stock markets has been well acknowledged in policy makers, portfolio managers, industries and investors perspectives. The stock market avail long-term capital to the listed firms by collecting funds from various potential investors, which allow them to expand in business and also offers investors alternative investment avenues to put their surplus funds in (Naik and Padhi, 2012). It is very interesting to invest in stock market but also a very risky trench of investment. So, potential investors always try to guess the movement of stock market prices to achieve maximum benefits and minimize the future risks. By concerning with the relationship between stock market returns and macroeconomic variables, investors might guess how stock market behaved if macroeconomic indicators such as exchange rate, industrial productions, interest rate, consumer price index and money supply fluctuate (Hussainey and Ngoc, 2009).
Macroeconomic indicators such as compositions of data which frequently used by the policy makers and investors to gathering knowledge for current and upcoming investment priority (Masuduzzaman, 2012).

Objectives of the study
In this study the major of the study are as follows: • To shed light on the nature of dynamic relationship that exists between the stock market and macro-economic variables, i.e., is it unilateral or bilateral.

Limitations of the Study
• For this study major limitation is analysis is mainly based on secondary data which is collected from the published annual reports of different institutions, industries and Bangladesh bank, therefore it may have potential bias from the data source as the limitation outlined.
• Besides to conduct the study only four (04) variables are collected. Small sample size may play a role to create doubt of its representativeness and there might be bias result. Such biasness is unavoidable and could affect the reliability and precision of findings.
She took daily returns, monthly standard deviations of stock returns. She concluded that the volatility measurement indicates upward trend of employment rate and increased economic stability. Chen, Roll and Ross (1986), summarized in their analysis that an equilibrium relationship exists between macroeconomic variables and stock prices. The concluded that the price of asset very sensitive to economic and unanticipated news.

Data and Methodology
The aim of study is to investigate the dynamic relation between stock market and macroeconomic variables. We used panel data in analysis and we also used OLS (ordinary Least Square) method.

Data
The data used in our research is secondary data. Previous studies stated that advantage of using secondary data such as improvement of the clarity of the problem and the situation surrounding the issues and they can also provide additional information (CWBrodeur et al. 2011).
Secondary data means data that are already available i.e., they refer to the data which have already been collected and analyzed by someone else.
Secondary data can be classified into internal and external. Internal secondary data is acquired within the organization where external secondary data is obtained from outside sources such as bank financial statement, annual report, textbooks, journal, articles, past year thesis.

Panel Data
Panel data is defined as the data that was generated from a small number of observations which covering a large number of units. In statistics and economics, multidimensional data also known as the panel data which contained elements of both time series and cross-sectional data.
There are several advantages of using panel data, such as (i) they increase the sample considerably, (ii) studying repeated cross-section observation, panel data are better suited to study the dynamics of changes and finally, (iii) panel data enable us to study more complicated behavioral model.

Data Sources
For this analysis, data for period of13years (2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015) have been collected from the web sites of Bangladesh bank and IMF where these data are of secondary in nature. In this study, we extracted the data from the web sites of Bangladesh banks. We used the Microsoft Excel (2013) where we arranging it according to the years and variables. The numbers are easily processed due to the convenience and efficiency provided by the software. After arranging the data from excel, we proceed by using it to Eviews (version 7) in order to examine the relationship between these variables and stock market.

The Unit Root Test
The If t* > ADF critical value, ==> not reject null hypothesis, i.e., unit root exists.
After conducting the unit root test, we will forward towards Johansen test of Co-integration followed by error correction Model.

Johansen Co-integration Test
To investigate the long-run relationship of the DSE index and macroeconomic variables as a system of equations, we employed the Johansen Co-integration test. Johansen developed two likelihood ratio tests for testing the number of Co-integration vectors (r): the trace test and the maximum Eigenvalue test.
The vector error correction model of Johansen (1991) uses the full information maximum likelihood method and the model aims to: 1. Test whether all variables are integrated of the same order by using unit root tests.
2. Find the truncated lag (k) such that the residuals from each equation of the vector error correction model are uncorrelated.
3. Regression DYt against the lagged differences of DYt and DYt-k. Then estimate the cointegrating vectors from the canonical correlations of the set of residuals from the regression equation using the set of variables in the model.
4. Determine the order of Co-integration using the ltrace and lmax test.
5. Test for the presence of a linear trend, test for linear restrictions on the co-integrating vectors.
6. By using the appropriate co-integrating vector, it determines the long run equilibrium relationship.

The Unit Root Test
This study uses DSE index, exchange rate, industrial production in and reserve. All data set, obtained from the Central Bank of Bangladesh, data base is monthly and runs from June2003 to June2015. This study aims to identify the dynamic relationships between stock market and macro-economic variables for Bangladesh. When the unit root test results are examined, it is observed that all four series, including DSE Index figures, are not stationary at their own levels.
ADF test scores show that all variables are integrated from the first order I (1)). Since all variables are not stationary at their own levels, OLS model is not appropriate to test the relations of this study. VAR model is chosen as the basis to test the relationships between selected macro variables and stock market index figures. We know our null hypothesis is series has unit root, we can see in the below table that after taking first difference according to t statistics (which is greater than 1%,5% and 10% level critical value) and p value (which is less than .05), we can reject the null hypothesis and the data become stationary.

Co-integration test
The VAR model is an effective means of characterizing the dynamic interactions among economic variables since it introduces very few restrictions (Lastrapes and Koray, 1990;McMillin, 1991). The use of the VAR model also allows inclusion of the appropriate lag lengths. This is important because of the time delays in the production of information concerning the macroeconomic variables. In particular, the transmission and incorporation of information into stock returns are not always instantaneous. This may be the case because reporting delays may create a lag between the observation of data concerning a macroeconomic variable and the incorporation of that information into stock returns (Abugri, 2006, p. 5). In order to decide what type of VAR model will be used in this study, after determination of unit roots and integration at first order, Johansen Co-integration tests are applied to control whether Co-integration exists among these four variables. Co-integration analysis is important, since if the error term coming from the linear combination of two variables is stationary, then there is Co-integration between the two variables. When there is no Co-integration between the two variables, then there is no long term relationship between two variables. Co-integration analyses have been used to test long run relationships between macroeconomic variables and stock market. This study uses Co-integration analysis not only to test whether there is a long-term relationship between macro variables and stock market, but also to decide specific VAR model to use in adjustment and short-term coefficient estimations. Johansen test is used to test Co-integration among DSE index, Exchange rate, Industrial production in and Reserve by using up to four lags length. The lag length is decided by using Akaike IC. It is seen from Johansen Co-integration test, both Maximum Eigenvalue and Trace tests result in the same decision: there are at most one Co-integration relationships among four variables we study. This means that there is one long-term stable relationship among these four variables. In other words, looking at the information coming from the past changes in DSE index figures and three macroeconomic indicators, it may be concluded that all four variables move together in the long run.
As the four variables are co-integrated we can run the VECM model. We want to know what p value for each variable is. So to know the p value we use the system equation.   In the above we can derive the residual of the co-integrating equation when DSE index is the dependent variables. Here C(1) = Speed of adjustment towards long run equilibrium but it must be significant and the sign must be negative. In above we can see it is significant that's mean the p value is less than .05 but the coefficient sign is not negative. So we can say there are no long run causality from the three independent variables such as Exchange rate, Industrial production in and Reserve. Exchange rate, Industrial production in and Reserve have no influence on the dependent variables such as DSE index in the long run. In other word, there is no long run causality running from Exchange rate, Industrial production in, Reserve and DSE index. And there is also no short run relation going from the three independent variables such as Exchange rate, Industrial production in and Reserve.   In the above we can derive the residual of the co-integrating equation when Exchange rate is the dependent variables. Here C(19) = Speed of adjustment towards long run equilibrium but it must be significant and the sign must be negative. In above we can see it is neither significant that's mean the p value is greater than .05 nor the sign of coefficient is negative. So we can say there are no long run causality from the three independent variables such as DSE index, Industrial production in and Reserve. DSE index, Industrial production in and Reserve have no influence on the dependent variables such as Exchange rate in the long run. In other word, there is no long run causality running from DSE index, Industrial production in, Reserve and Exchange rate. And there is no short run relation going from DSE index, Industrial production in and Reserve to Exchange rate except C (25), C (30), C (31), C (32) and C (33).   In the above we can derive the residual of the co-integrating equation when Industrial production in is the dependent variables. Here C (37) = Speed of adjustment towards long run equilibrium but it must be significant and the sign must be negative. In above we can see it is significant that's mean the p value is less than .05 but the coefficient sign is not negative. So we can say there are no long run causality from the three independent variables such as DSE index, Exchange rate, and Reserve. Meaning that DSE index, Exchange rate, and Reserve have no influence on the dependent variables such as Industrial production in the long run. In other word, there is no long run causality running from DSE index, Exchange rate, Reserve and Industrial production in. And there is no short run going from DSE index, Exchange rate and Reserve to Industrial production in except C (40), C (49) and C (50). Sample (  In the above we can derive the residual of the co-integrating equation when Reserve is the dependent variables. Here C (55) = Speed of adjustment towards long run equilibrium but it must be significant and the sign must be negative. In above we can see it is not significant that's mean the p value is greater than .05 but the coefficient sign is negative. So we can say there are no long run causality from the three independent variables such as DSE index, Exchange rate and Industrial production in. Meaning that DSE index, Exchange rate and Industrial production in have no influence on the dependent variables such as Reserve in the long run. In other word, there is no long run causality running from DSE index, Exchange rate, Industrial production in and Reserve. And there is no short run going from DSE index, Exchange rate and Industrial production in to Reserve except C (68).

Major study findings
Our regression model (table 4.4) where we take DSE index as dependent variables presents the output of the Ordinary Least Square (OLS) method to show the dynamic relation between stock market and macroeconomic variables. We can see that R-square is .140202 or 14.02% which is less than 60%. So it is not a good sign for this model. It indicates that the three independent variables can explain about 14.02% variability of dependent variable i.e. DSE Index. The adjusted R-square is also below 60% which is not a good sign at all. We know C (1)) is the speed of adjustment towards long run equilibrium and C (2) to C (18) are short run equilibrium. And we also know if the coefficient of C (1) is negative and the p value is less than .05 then it calls significant that means there is a long run relationship. Here we found the coefficient of C (1) is positive but the p value is less than .05, it doesn't fulfill the two criteria of significance.
So that we can say there is no long run relation going from Exchange rate, Industrial production in and Reserve to DSE index. And there is no short run going from Exchange rate, Industrial production in and Reserve to DSE index. It is not a good sign because we know at least 50% of the independent variables should be statistically significant with dependent variable.
Our regression model ( Industrial production in. The adjusted R-square is also below 60% which is not a good sign at all. We know (C (37)) is the speed of adjustment towards long run equilibrium and C (38) to C(54) are short run equilibrium. And we also know if the coefficient of C (38) is negative and the p value is less than .05 then it calls significant that means there is a long run relationship.
Here we found the coefficient of C (38) is positive but the p value is less than .05, it doesn't fulfill the two criteria of significance.
So that we can say there is no long run relation going from DSE index, Exchange rate and Reserve to Industrial production in. And there is no short run going from DSE index, Exchange rate and Reserve to Industrial production in except (C(40) which is a lag of DSE index), (C(49) which is a lag of Industrial production in), (C(50) which is a lag of Reserve). It is not a good sign, because we know at least 50% of the independent variables should be statistically significant with dependent variable. Our regression model (table 4.7) where we take Reserve as dependent variables presents the output of the Ordinary Least Square (OLS) method to show the dynamic relation between stock market and macroeconomic variables. We can see that Rsquare is .707092 or 70.71% which is greater than 60%. So it is a good sign for this model. It indicates that the three independent variables can explain about 70.71% variability of dependent variable i.e. Reserve. The adjusted R-square is also above 60% which is a good sign.
We know C (55)) is the speed of adjustment towards long run equilibrium and C (56) to C(72) are short run equilibrium. And we also know if the coefficient of C (55) is negative and the p value is less than .05 then it calls significant that means there is a long run relationship. Here we found the coefficient of C (55) is negative but the p value is greater than .05, it doesn't fulfill the two criteria of significance. So that we can say there is no long run relation going from DSE index, Exchange rate and Industrial production in to Reserve. And there is no short run going from DSE index, Exchange rate and Industrial production in to Reserve except C (68) which is a lag of Reserve). It is not a good sign because we know at least 50% of the independent variables should be statistically significant with dependent variable.

Recommendation
We should take the followings measures to overcome the limitation of this model • We use monthly data of this model; we should use quarterly or yearly data for significant result of this model.
• We only use four variables (DSE index, Exchange rate, Industrial production in and Reserve), we should use more variables for significant result of this model.

Conclusions
This paper analyzes long-term equilibrium relationships between a group of macroeconomic variables and the stock market. The macroeconomic variables are represented by DSE index, Exchange rate, Industrial production in and Reserve, model is employed to avoid potential misspecification biases that might result from the use of a more conventional vector auto regression modeling technique. All of the new research in this area has focused on industrial countries, with relatively little attention paid to developing countries. Accordingly, I believe that this paper will add to our understanding as to whether similar empirical results are observed in developing countries. In addition, these findings may have important policy implications because they could be crucial in areas such as the design of stabilization and adjustment programs.
The Johansen Co-integration test indicates that there exists a long run relationship between stock market and the macroeconomic variables tested. The empirical evidence shows that these macroeconomic variables are co-integrated i.e. there exists a co-integrating relation among the variables. After conducting VECM, we find there is no long run relationship among those variables and stock market. Analysis of the results indicates that this Co-integration relationship is consistent with the earlier findings, and the signs of the variables are also consistent with the earlier studies.