Research Article: 2021 Vol: 24 Issue: 1S
Dr. Sandeep Vyas, International School of Informatics and Management (IIIM)
The emerging markets owing to their growth potentials have become favored investment destinations for international investors even after considering the risky nature of foreign markets. The research paper examines macroeconomic variables that are assumed to influence stock market returns in India. It attempts to identify whether any causal relationship exists between the stock market returns and macroeconomic indicators by using regression analysis. The indicators that have been taken into account are Index of Industrial Production (IIP) as a proxy for Gross Domestic Product (GDP), Wholesale Price Index (WPI) as a proxy for inflation, Money Supply M1 (MSM1), Rupee Dollar Exchange Rate (REDOLLXR), Foreign Portfolio Investment (FPI in equity only) and Federal Reserve Rates (FRR) on S & P BSE SENSEX index (BSESENX).
Foreign Portfolio Investment, Federal Reserve Rate, BSE Sensex, GDP, IIP, WPI, MSM1.
G10, G14, E2, E44.
After the economic reforms period in India, there has been a notable change in the financial system of India. Though the banking system still dominates the flow of funds, stock markets have acquired an important role in mobilizing funds to the corporate. The research in developing economies like India is drawing attention towards studying the relationship between stock markets performance and macroeconomic variables. Due to the potentials of economic growth in the emerging markets, FPI in India has increased manifold in the last decade. The study uses time-series data in analyzing a causal relationship between the dependent and independent variables. The variables have been tested on all parameters of a good fit regression model. Since the data used in the research is longitudinal, the residual issues have been handled well.
Kuwornu (2011) examines the relationship between macroeconomic variables and stock market returns from 1992 to 2008. He studied the causal relationship between consumer price index, crude oil price, exchange rate, and 91 day Treasury bill rate (as a proxy for interest rate) and stock market returns. Fama (1981) observes a positive correlation between stock market returns and macroeconomic variables.
Under the Arbitrage Pricing Theory (APT) framework, several studies have been conducted between the macroeconomic variables which affect future cash available for investments and returns of a stock. Omran & Pointon (2001) have studied and found a negative relationship between inflation and the stock market of Egypt. Chatrath, Ramchander & Song (1997) also conducted a study on the relationship between inflation and stock prices of Indian companies. The researchers concluded a negative relationship between stock return and inflation. Using the APT framework for research Chen et al. (1986) researched to study the impact of interest rates, inflation rate, exchange rate, bond yield, and industrial production on US stock markets. They observed that these variables significantly influence US stock market returns. Zhao, (1999) finds a strong relationship between inflation and stock prices of China stocks.
Objective of the study
The objective of the study is to examine whether any causal relationship exists between the economic factors such as Index of Industrial Production (IIP) as a proxy for GDP, Wholesale Price Index (WPI) as a proxy for inflation, Money Supply M1 (MSM1), Rupee Dollar Exchange Rate (REDOLLXR), Foreign Portfolio Investment (FPI in equity only) and Federal Reserve Rates (FRR) on S & P BSE SENSEX index (BSESENX).
The methodology used in the study aims at developing a multiple linear regression model based on IIP, WPI, MSM1, REDOLLXR, FPI (in equity only), and FRR as predictor variables and BSESENX as the criterion variable. FRR as the external variable is considered in the study as it is often given weightage in the prediction of the Indian stock market returns for both the short and long term owing to increasing investments of Foreign Portfolio Investors in the Indian equity market. As FPI investment is highly influenced by a change in the FRR of the US, it is used in the regression model along with the FPI.
H0: No significant linear relationship exists between the criterion variable (BSESENX) and the six predictor variables. (IIP, WPI, MSM1, REDOLLXR, FPI and FRR)
H1: There exists a significant linear relationship between the criterion variable (BSESENX) and the six predictor variables. (IIP, WPI, MSM1, REDOLLXR, FPI and FRR)
Sample size and data collection
The study is based on time-series data of monthly observations of the aforementioned variables from April 2010 to March 2017 and includes 83 observations. The sample is taken from the year April 2010 onwards. This period represents the post subprime crisis period that besides having a global impact also negatively affected the Indian Economy and BSE Sensex Index. To remove outliers and to ensure the sanctity of the financial data, the sample is taken from 2010 onwards. For the criterion variable BSESENX, data is taken from the BSE website taking into account the monthly closing values of the S&P BSE SENSEX. The data for predictor variables IIP, WPI, MSM1, REDOLLXR, and FPI is taken from the Reserve Bank of India website. The monthly data for FRR is sourced from the website of the Federal Reserve Bank of the USA.
A multiple linear regression analysis has been conducted by taking S&P BSE SENSEX returns (BSESENX) as the criterion variable and taking IIP, WPI, MSM1, REDOLLXR, FPI, and FRR as the six predictors. Further, stability tests, descriptive statistics, Pearson’s coefficient correlation test for checking multi-collinearity, and other tests have been conducted to test the goodness of fit. The following model will be tested in the study:
BSESENX t = α + β1.IIPt + β2.WPIt + β3.MSM1t + β4.REDOLLXRt + β5.FPIt + β6.FRRt + ?t (1)
Where, BSESENX t= S&P BSE SENSEX index at time‘t’ (criterion variable); α = constant; IIPt= Index of Industrial Production; WPI= Wholesale Price Index; MSM1t= Money Supply M1; REDOLLXR t = Rupee Dollar Exchange Rate; FPIt = Foreign Portfolio Investment; FRR t= Federal Reserve Rates at time‘t’ respectively; and β1, β2, β3, β4, β5and β6 are regression
coefficients of the respective predictor variables and ?t = error term at time‘t’.
Table 1 Description Of Variables |
|||
---|---|---|---|
Variable | Type of Data | Units | Source |
BSESENX | First differenced Raw data of Bombay Stock ExchangeSensitive Index | Monthly closing values | Bombay Stock Exchange India Website |
IIP | First differenced Raw data of Index of IndustrialProduction | Monthly values | Reserve Bank of India Website |
WPI | First differenced data ofWholesale Price Index | Monthly values | Reserve Bank of India Website |
MSM1 | First differenced data ofMoney Stock M1 | Monthly values | Reserve Bank of India Website |
REDOLLXR | First differenced data ofRupee Dollar Exchange Rate | Monthly values | Reserve Bank of India Website |
FPI | First differenced data of Foreign Portfolio Investmentin equity only) | Monthly values | Reserve Bank of India Website |
FRR | First differenced data ofFederal Reserve Rates | Monthly values | Federal Reserve Bank Website |
Source: Generated by the author
The table 1 above describes the variables. The IIP is taken as a proxy for GDP as monthly data on GDP was not available. While WPI is taken as a proxy for inflation as monthly data was not available for inflation. Money stock M1 is taken as an independent variable since the level of money supply affects the level of investment in the stock market. M1 represents narrow money which includes demand deposits and other currency in circulation as this money is used for investment in the stock market. It excludes fixed deposits and other long-term deposits with banks. The rupee-dollar exchange rate is also considered for the study as it also affects the investment decision of foreign portfolio investors which in turn affects BSE Sensex. Federal Reserve Rates, though an exogenous variable, affects investment decisions of FPI in India which in turn affects the BSE Sensex Index. This inference is drawn from the news speculation about the increase or decrease in the FPI in India before Federal Reserve Bank (FRB) decides for change in the FRR from time to time. Hence it is taken as an independent variable. Out of these
predictors, FRR is the only variable that is exogenous and is out of the system, and is not affected by any of these variables.
Data analysis and interpretation
The data were first tested for stationarity of variables using the unit root test. The augmented Dickey-Fuller test (ADF) was used for finding the element of non-stationarity in the variables. The test revealed the presence of unit root in the BSESENX, IIP, WPI, MSM1, REDOLLXR, and FRR. Hence data for these variables were transformed at the first difference to make them stationary and to be fitted in the regression model. The data was then tested for stability of the dependent variable BSESENX using the CUSUM (Cumulative Sum Control Chart) test of recursive residuals at a 5% significance level.
Source: generated by the author
Figure 1 depicts the stability of the criterion variable as it is within the control limits. Hence all the residuals are stable as the cumulative sums are located within the standard deviation band.
The descriptive statistics in table 2 below exhibit that BSESENX, MSM1, and FPI have the highest dispersion of data from their respective mean while deviation is relatively less in IIP, WPI, REDOLLXR, and FRR variables. The mean, minimum and maximum values of BSESENX and MSM1 are the highest. The skewness coefficients reveal some negative distribution of data. The variables BSESENX, IIP, WPI, and MSM1 have long left-tailed negative skewness while REDOLLXR, FPI, and FRR have long right-tailed positive skewness. The Kurtosis values show that the probability density function (PDF) has a fat-tailed distribution for IIP, MSM1, REDOLLXR, and FRR. The p values of the Jarque-Bera test reveal that variables BSESENX, IIP, WPI, and FPI are normally distributed while MSM1, REDOLLXR, and FRR are not normally distributed. But Jarque-Bera (J-B) statistics for the Histogram Normality test gives a J-B p-value of 0.1082 mentioned further in the paper which suggests that the model data is no different from a normal distribution (acceptance of null hypotheses).
Table 2 Descriptive Statistics |
|||||||
---|---|---|---|---|---|---|---|
Statistics | DBSESENX | DIIP | DWPI | DMSM1 | DREDOLLXR | DFPIEQ | DFRR |
Mean | 145.3228 | 0.572934 | 0.562651 | 127.1878 | 0.245766 | 2.977831 | 0.007108 |
Median | 102.8300 | 0.300000 | 0.600000 | 145.3900 | 0.183500 | -0.550000 | 0.000000 |
Maximum | 2339.860 | 25.10000 | 3.500000 | 2934.850 | 5.459200 | 332.3900 | 0.130000 |
Minimum | -2181.330 | -27.70000 | -2.500000 | -3770.580 | -3.797200 | -301.5800 | -0.040000 |
Std. Dev. | 960.7023 | 10.59759 | 1.205648 | 764.3838 | 1.524622 | 115.4207 | 0.031720 |
Skewness | -0.108130 | -0.305013 | -0.329724 | -2.478114 | 0.266043 | 0.086347 | 2.543243 |
Kurtosis | 2.709050 | 3.643320 | 3.088842 | 19.17690 | 4.533726 | 2.927883 | 10.05687 |
Jarque-Bera | 0.454494 | 2.718227 | 1.531232 | 989.9693 | 9.114199 | 0.121126 | 261.6983 |
Probability | 0.796724 | 0.256888 | 0.465047 | 0.000000 | 0.010492 | 0.941235 | 0.000000 |
Sum | 12061.79 | 47.55350 | 46.70000 | 10556.59 | 20.39860 | 247.1600 | 0.590000 |
Sum Sq. Dev. | 75681805 | 9209.324 | 119.1942 | 47911179 | 190.6067 | 1092399 | 0.082506 |
Observations | 83 | 83 | 83 | 83 | 83 | 83 | 83 |
Source: calculated by the author
Multi Collinearity Diagnosis
Further, Pearson's correlation matrix was used to check the problem of multi collinearity as the strong correlation among independent variables can give spurious results in the regression analysis. The correlation matrix in Table 3 below shows the correlation among independent variables for the study period from April 2010 to March 2017. From the table, it can be observed that the independent variables show a correlation value ranging between -0.54 to 0.33 values which means a moderately negative to a moderately positive correlation.
Table 3 Pearson Correlation Matrix of Variables |
||||||
---|---|---|---|---|---|---|
DIIP | DWPI | DMSM1 | DREDOLLXR | DFPIEQ | DFRR | |
DIIP | 1.000000 | - | - | - | - | - |
DWPI | -0.235242 | 1.000000 | - | - | - | - |
DMSM1 | 0.012342 | -0.009571 | 1.000000 | - | - | - |
DREDOLLXR | -0.118294 | 0.007249 | -0.142122 | 1.000000 | - | - |
DFPIEQ | 0.334313 | -0.043924 | 0.137567 | -0.547479 | 1.000000 | - |
DFRR | 0.242510 | -0.233091 | -0.031554 | -0.068332 | 0.060791 | 1.000000 |
Source: calculated by the author
The result of Variance Inflation Factor (VIF) in the Ordinary Least Squares (OLS) regression equation in table 4 below exhibit that the VIF for all the independent variables is between 1 and 5. Therefore, it can be expressed that the predictor variables have very weak multi collinearity and multiple regression analysis is fit to be conducted using all the mentioned independent variables in the study.
Table 4 Variance Inflation Factors |
|||
---|---|---|---|
Variable | Coefficient Variance | Uncentered VIF | Centered VIF |
C | 8155.106 | 1.443738 | NA |
DIIP | 63.53029 | 1.251619 | 1.247928 |
DWPI | 4321.695 | 1.340937 | 1.098728 |
DMSM1 | 0.010067 | 1.057573 | 1.028744 |
DREDOLLXR | 3574.344 | 1.491387 | 1.453166 |
DFPIEQ | 0.691230 | 1.611674 | 1.610589 |
DFRR | 6293851. | 1.163902 | 1.107600 |
Source: calculated by the author
Goodness of Fit
The regression equation was further analyzed for the goodness of fit using the following test parameters:
1.Augment Dickey-Fuller (ADF) unit root
2.Jarque Bera test for
3.Breush-Godfrey Serial correlation LM
4.Breush-Pegan-Godfrey test for
As mentioned earlier, the individual data sets of the variables were found non-stationary by using Augment Dickey-Fuller (ADF) Test. Table 5 below shows the test results of the ADF unit root test for all variables. From the table, it is evident that the p-value of all variables is more than 0.05. Hence, all the variables have unit root at level except FPI whose p-value is more than
0.05. So data for all variables including FPI to ensure uniformity of data was converted at first difference. The observed p values of all variables at first difference were recorded below 0.05. Hence data was found to be stationary at first difference.
Table 5 ADF Unit Root Test |
||||
---|---|---|---|---|
Variable | At level | 1st difference | ||
ADF | t-statistic | p-value | t-statistic | p-value |
BSESENX | -0.547272 | 0.8754 | -9.482785 | 0.0000 |
IIP | -0.419464 | 0.8995 | -11.81768 | 0.0001 |
WPI | -2.166779 | 0.2199 | -4.912811 | 0.0001 |
MSM1 | 0.131635 | 0.9661 | -5.274656 | 0.0000 |
FPI | -5.625940 | 0.0000 | -11.76935 | 0.0001 |
FRR | 4.315501 | 1.0000 | -4.816978 | 0.0001 |
Source: calculated by the author
Figure 2 below shows the Jarque-Bera (J-B) statistics to test whether the residuals are normally distributed. The table observes the p-value of the J-B test is more than 0.05. Hence it can be interpreted that the residuals are normally distributed.
Source: generated by the author
Table 6 below exhibits the results of the Breush-Godfrey serial correlation LM test and Breush- Pegan Godfrey test for heteroskedasticity. The Chi-square value of Breusch – Godfrey (B-G) serial correlation LM test is 0.7899 which means that there is no serial correlation in the residuals. Similarly, Breusch-Pagan-Godfrey (B-P-G) Heteroskedasticity with the probability Chi-square value of 0.9672 validates that there is no heteroskedasticity in the residuals and the model is a good fit model.
Table 6 | |||
---|---|---|---|
Breusch-Godfrey Serial Correlation LM Test and Breusch-Pagan-Godfrey Heteroskedasticity Test | |||
Breusch-Godfrey Serial Correlation LM Test: | |||
F-statistic | 0.211471 | Prob. F(2,74) | 0.8099 |
Obs*R-squared | 0.471685 | Prob. Chi-Square(2) | 0.7899 |
Breusch-Pagan-Godfrey Heteroskedasticity Test | |||
F-statistic | 0.213695 | Prob. F(6,76) | 0.9714 |
Obs*R-squared | 1.377033 | Prob. Chi-Square(6) | 0.9672 |
Scaledexplained SS | 1.414335 | Prob. Chi-Square(6) | 0.965 |
Source: calculated by the author
The following table 7 is the outcome of the regression equation (1) in which BSE Sensex is the target variable and IIP, WPI, MSM1, REDOLLXR, FPI and FRR are independent variables.
Table 7 Model Summary |
||||
---|---|---|---|---|
Dependent Variable: BSESENX Method: Least SquaresSample (adjusted): 2010M05 2017M03 Included observations: 83 after Adjustments | ||||
Variable | Coefficient | Std. Error | t-Statistic | Prob. |
C | 158.7620 | 90.30563 | 1.758053 | 0.0828 |
IIP | 1.682561 | 7.970589 | 0.211096 | 0.8334 |
WPI | 124.1550 | 65.73960 | 1.888588 | 0.0628 |
MSM1 | 0.059404 | 0.100333 | 0.592063 | 0.5556 |
REDOLLXR | -317.0466 | 59.78581 | -5.303041 | 0.0000 |
FPI | 2.338102 | 0.831402 | 2.812239 | 0.0063 |
FRR | -2934.218 | 2508.755 | -1.169591 | 0.2458 |
R-squared | 0.529195 | Mean dependent var | 145.3228 | |
Adjusted R- squared | 0.492026 | S.D. dependent var | 960.7023 | |
S.E. of regression | 684.7146 | Akaike info criterion | 15.97645 | |
Sum squared resid | 35631392 | Schwarz criterion | 16.18045 | |
Log-likelihood | -656.0226 | Hannan-Quinn criteria. | 16.05840 | |
F-statistic | 14.23759 | Durbin-Watson stat | 1.859756 | |
Prob(F-statistic) | 0.000000 |
Source: calculated by the author
The f-statistics of the model in table 7 above has a p-value of less than 0.05 which states that there is a statistically significant linear relationship between the dependent variable BSE Sensex returns and the other six independent variables. Since the p-value of f-statistics is less than 0.05, the null hypothesis is rejected and alternate hypothesis H1 is accepted that there exists a statistically significant linear relationship between the criterion variable (BSESENX) and the six predictor variables (IIP, WPI, MSM1, REDOLLXR, FPI, and FRR).
However, the t-statistics reveal that only the Rupee dollar exchange rate (REDOLLXR) and Foreign Portfolio Investment (FPI) are individually significant in affecting the dependent variable BSESENX. The negative values of the coefficients of REDOLLXR and FRR are in line with the economic theory. As when the Rupee dollar exchange rate increases (depreciation in rupee) the Sensex records a bearish trend and vice versa. While when FRR increases, the FPI divert their funds in the US fixed deposits and bond market as domestic markets are always considered less risky than foreign markets. While when FRR decreases, the influx of FPI in India increases which results in bullish Sensex. The positive values of the coefficients of other independent variables IIP, WPI, MSM1, and FPI are also in line with the theory of intuition. However, these variables have no statistically significant linear relationship with the dependent variable individually. The R square value of the model is 52.9 percent. Though the model cannot be used for forecasting, the study has revealed that IIP, WPI, MSM1, and FRR should not be given much importance in predicting BSE Sensex return behavior. Conclusively it can be argued that the Rupee dollar exchange rate and Foreign Portfolio Investment should be given due importance because they are statistically significant in determining BSE Sensex returns.
The study does not take into account variables that could have explained the movement of BSE SENSEX returns with one hundred percent predictability as the R-squared value of the model is 52.9 percent. There are variables outside the model that are also important in explaining returns of the BSE Sensex. Besides the study does not take into account a comparison between the impact of macroeconomic variables on the stock market indices across countries. This limitation was due to non-access to the panel data of other countries.