Impact of macroeconomic variables on economic indicators:
An Empirical Study of India and Sri Lanka (Gurvinder Singh) 1.Introduction
Over thirty years the relationship between macroeconomic variables and stock market prices has been an attractive subject for both financial and macro economists. Although there are a number of studies that investigate the link between macroeconomic variables and stock market, both the academics and the practitioners have not arrived at a consensus on the direction of the causality among these variables, which remained as a source of ambiguity.An efficient capital market is one in which security prices adjust rapidly to the arrival of new information and, therefore, the current prices of securities reflect all information about the security. What this means, in simple terms, is that no investor should be able to employ readily available information in order to predict stock price movements quickly enough so as to make a profit through trading shares.
For the past three decades evidence that key macroeconomic variables help predict the time series of stock returns has accumulated in direct contradiction to the conclusions drawn by the EMH. The onslaught against the conclusions drawn from the EMH includes early studies by Fama and Schwert (1977) and Jaffe and Mandelker (1976), all affirming that macroeconomic variables influence stock returns. Again Chen, Roll and Ross (1986), having first illustrated that economic forces affect discount rates, the ability of firms to generate cash flows, and future dividend payouts, provided the basis for the belief that a long-term equilibrium existed between stock prices and macroeconomic variables. More recently, Granger (1986) proposed to determine the existence of long-term equilibrium among selected variables through cointegration analysis, paving the way for a (by now) preferred approach to examining the economic variables- stock markets relationship. A set of time-series variables are cointegrated if they are integrated of the same order and a linear combination of them is stationary. Such linear combinations would then point to the existence of a long-term relationship between the variables. An advantage of cointegration analysis is that through building an error-correction model (ECM), the dynamic co- movement among variables and the adjustment process toward long-term equilibrium can be examined, Maysami et al (2004).
More recently, Granger (1986) and Johansen and Juselius (1990) proposed to determine the existence of long-term equilibrium among selected variables through cointegration analysis, paving the way for a (by now) preferred approach to examining the economic variables-stock markets relationship. A set of time-series variables are cointegrated if they are integrated of the same order and a linear combination of them is stationary. Such linear combinations would then point to the existence of a long-term relationship between the variables. An advantage of cointegration analysis is that through building an error-correction model (ECM), the dynamic co- movement among variables and the adjustment process toward long-term equilibrium can be examined. 2
Despite the importance of previous studies, until now the majority of research considers developed countries financial markets, which are efficient enough and do not suffer from the inefficiency problems in less developed countries. Considering this matter, the subject of financial markets in developing countries still needs lengthy analysis and more research attention.
2.Objectives The study aims to achieve the following objectives: 1. To study the pattern of CPI, WPI, GDP, GNI and Rate of interest in India and Sri Lanka for the year 2002-2009 2. To study the impact of macroeconomic variable on GDP growth of the Indian and Sri Lankan economy. 3. To comparatively analyze the impact of macro-economic variable on GDP growth in India viz-a-viz Sri Lanka
3.Literature Review For number of years, there has been an extensive debate in the literature assessing the influence of macroeconomics variables on the stock return. The economic theory, in explaining this interrelationship, suggests that stock prices should reflect expectations about futures corporate performance. Corporate profits on the other hand generally may reflect the level of countrys economic activities. Thus, if stock prices accurately reflect the underlying fundamentals, then the stock prices should be employed as leading indicators of future economic activity. However, if economic activities reflect the movement of stock prices, the results then should be the opposite, i.e economic activities should lead stock price. Therefore, the causal relations and dynamics interactions among economics factors and stock prices are important in the formulation of nations macroeconomic policy. According to Oberuc (2004), the economic factors which, usually associated with stock prices movement and being considered greatly by researchers are dividend yield, industrial production, interest rate, term spread, default spread, inflation, exchange rates, money supply, GNP or GDP and previous stock returns, among others.
Emerging stock markets have been identified as being at least partially segmented from global capital markets. In direct contradiction to the conclusions drawn by the EMH, evidence that key macroeconomic variables help predict the time series of stock returns has accumulated for nearly 30 years.
Numerous studies have investigates the relationship between stock returns, interest rates, inflation and real activity (see, inter alia, Fama (1981, 1990), James et al. (1985), Fama and Schwert (1977), Geske and Roll (1983), Mandelker and Tandon (1985), Hendrys (1986), Chen, Roll and Ross(1986), Darrat and Mukherjee (1987), Fama and French (1989) , Asprem (1989),Martinez & Rubio, 1989, Schwert, 1989,Schwert (1990), Ferson and Harvey (1991), Lee (1992), Mukherjee and Naka (1995), Chatrath et al. (1997), Naka A., Mukherjee T. and Tufte D. 3
(1998), Ibrahim (1999) , Gjerde & Saettem, 1999, Wongbangpo and Sharma (2002), , and Vuyyuri (2005). [ Now we divide the whole work of the researchers on macroeconomic variables in different parts on the basis of stock markets, tests and conclusions.
Fama (1981) focuses upon the correlation between stock returns and expected and unexpected inflation in the U.S., showing that the observed negative relation is a proxy effect for more fundamental relationships between stock returns and real activity. James et al. (1985) finds strong links between stock returns, real activity and money. Investigating the stock return- inflation relation for the U.S., U.K., Canada and German. Fama and Schwert (1977) estimate the extent to which various assets were hedges against the expected and unexpected components of the inflation rate during the 19531971 period. They finds that U.S. government bonds and bills were a complete hedge against expected inflation, and private residential real estate was a complete hedge against both expected and unexpected inflation. Geske and Roll (1983) offers a supplementary explanation suggesting that stock prices signal changes in expected inflation because money supply responds to changes in expected real activity. Mandelker and Tandon (1985) tests whether the negative relationship between real stock returns and inflation in the United States is in fact proxying for a positive relationship between stock returns and real activity variables in six major industrial countries over 19661979. Hendrys (1986) approach which allows making inferences to the short-run relationship between macroeconomic variables as well as the long-run adjustment to equilibrium, they analysed the influence of interest rate, inflation, money supply, exchange rate and real activity, along with a dummy variable to capture the impact of the 1997 Asian financial crisis. Chen, Roll and Ross (1986), having first illustrated that economic forces affect discount rates, the ability of firms to generate cash flows, and future dividend payouts, provided the basis for the belief that a long-term equilibrium existed between stock prices and macroeconomic variables. Darrat and Mukherjee (1987) finds a significant causal (lagged) relationship between stock returns and some selected macro variables, including money supply, implying market inefficiency in the semi-strong sense on the Indian data over 19481984. Fama and French (1989) finds that expected returns on common stocks and long- term bonds contain a term or maturity premium that has a clear business-cycle pattern (low near peaks, high near troughs). Expected returns also contain a risk premium that is related to longer- term aspects of business conditions. The variation through time in this premium is stronger for low-grade bonds than for high-grade bonds and stronger for stocks than for bonds. The general message is that expected returns are lower when economic conditions are strong and higher when conditions are weak. Asprem (1989) investigates the relationship between stock indices, asset portfolios and macroeconomic variables in ten European countries. It is shown that employment, imports, inflation and interest rates are inversely related to stock prices. Expectations about future real activity, measures for money and the U.S. yield curve are positively related to stock prices. Schwert, 1989 in analyzes the relation of stock volatility with real and nominal macroeconomic volatility, economic activity, financial leverage, and stock trading activity using monthly data from 1857 to 1987. Ferson and Harvey (1991) provides an analysis of the 4
predictable components of monthly common stock and bond portfolio return. Most of the predictability is associated with sensitivity to economic variables in a rational asset pricing model with multiple betas. The stock market risk premium is the most important for capturing predictable variation of the stock portfolios, while premiums associated with interest rate risks capture predictability of the bond returns. Time variation in the premium for beta risk is more important than changes in the betas. Lee (1992) investigates causal relations and dynamic interactions among asset returns, real activity, and inflation in the postwar United States. Major findings are (1) stock returns appear Granger-causally prior and help explain real activity, (2) with interest rates in the VAR, stock returns explain little variation in inflation, although interest rates explain a substantial fraction of the variation in inflation, and (3) inflation explains little variation in real activity. Mukherjee and Naka (1995) investigates whether cointegration exists between the Tokyo Stock Exchange index and six Japanese macroeconomic variables, namely the exchange rate, money supply, inflation, industrial production, long-term government bond rate, and call money rate. They find that a cointegrating relation indeed exists and that stock prices contribute to this relation. Chatrath et al. (1997) investigates a negative relationship between stock market returns and inflationary trends has been widely documented for developed economies in Europe and North America. This study provides similar evidence for India. Naka A., Mukherjee T. and Tufte D. (1998) analyze relationships among selected macroeconomic variables and the Indian stock market. They find that three long-term equilibrium relationships exist among these variables. These results suggest that domestic inflation is the most severe deterrent to Indian stock market performance, and domestic output growth is its predominant driving force. After accounting for macroeconomic factors, the Indian market still appears to be drawn downward by a residual negative trend. Ibrahim (1999) investigates the dynamic interactions between seven macroeconomic variables and the stock prices for an emerging market, Malaysia. The results strongly suggest informational inefficiency in the Malaysian market. The bivariate analysis suggests cointegration between the stock prices and three macroeconomic variables - consumer prices, credit aggregates and official reserves. From bivariate error-correction models, we note the reactions of the stock prices to deviations from the long run equilibrium. Gjerde & Saettem, 1999 investigates to what extent important results on relations among stock returns and macroeconomic factors from major markets are valid in a small, open economy. Wongbangpo and Sharma (2002) investigates the role of select macroeconomic variables, i.e., GNP, the consumer price index, the money supply, the interest rate, and the exchange rate on the stock prices in five ASEAN countries (Indonesia, Malaysia, Philippines, Singapore, and Thailand). They observe long and short term relationships between stock prices and these macroeconomic variables. Vuyyuri (2005) investigates the cointegration relationship and the causality between the financial and the real sectors of the Indian economy using monthly observations from 1992 through December 2002.
Different methods of data analysis have been put into use by the researchers in their studies about the effect of macroeconomic variables on economy in the case of India, Sri Lanka, U.S., U.K., Canada, Germany, Netherland, Switzerland, European countries and ASEAN countries 5
with each other or with country(s) from the other parts of the world. Grangers Causality Model, Cointegration techniques (particularly Johansens Model), VECM and Vector Auto Regression Model are the prominent ones that have been used to analyze the data about the effect of macroeconomic variables on economy. However, a number of researches have used only one or at the most two methods to analyze the data. Grangers causality model has also been used very extensively by the researchers. Darrat and Mukherjee (1987) , Wongbangpo and Sharma (2002), Ibrahim (1999), Vuyyuri (2005) applying Granger-type causality. Ibrahim (1999), Wongbangpo and Sharma (2002) Cheung and Ng (1998), Vuyyuri (2005), Johansen, S. & Juselius, K. 1990 using the Johansen cointegration technique using cointegration. Ferson and Harvey (1991) using rational asset pricing model with multiple betas. James et al. (1985) ,Lee (1992), Gjerde & Saettem, 1999 Using a multivariate vector autoregression (VAR) approach, uses a VARMA approach. Mukherjee and Naka (1995), Naka A., Mukherjee T. and Tufte D. (1998) use Johansen's (1991) vector error correction model (VECM) .Chatrath et al. (1997) using heteroscedasticity and autocorrelation corrected models. The current study contributes to the literature in numerous ways. First, this is the study concentrating on the economy India and Sri Lanka; and studies the linkages within these rather than with the developed world. Secondly, it uses a combination of the various methods used empirically to analyze the data.
4.Research Methodology
In this study monthly data from 2002 onwards to 2009 has been used in case of all the variables like, GDP (Gross Domestic Product), GNI (Gross National Income), wholesale price index (WPI), consumer price index (CPI), exchange rates, bank rates and balance of payments. The major source of data of all the above macro economic variables is International Monetary Fund on-line data source. Index Numbers (2000=100) is used as the base index for the whole research data. We filled the missing values by taking the average of two of the preceding cases and two of the succeeding cases. Data have been analyzed using econometric tools. The analysis of econometrics can be performed on a series of stationary nature. In order to check whether or not the series are stationary, we prepare the line graph for each of the series. Further, we perform the Augmented Dickey-Fuller test under the unit root test to finally confirm whether or not the series are stationary. For the basic understanding of Unit root testing, we may look at the following equation yt = yt1 + xt'o + ct , (1.1) where xt are optional exogenous regressors which may consist of constant, or a constant and trend, and o are parameters to be estimated, and the ct are assumed to be white noise. If || 1 , y is a nonstationary series and the variance of y increases with time and approaches infinity. If ||< 1 , y is a (trend-)stationary series. Thus, we evaluate the hypothesis of (trend-) stationarity by testing whether the absolute value of || is strictly less than one. 6
The Standard Dickey-Fuller test is carried out by estimating equation (1.2) after subtracting yt-1 from both sides of the equation. Ayt = o y t-1 + xt'o + ct, (1.2) where o = - 1. The null and alternative hypotheses may be written as,
H0 : o = 0 H1: o < 0 (1.3) In order to make the series stationary, we take the log of the three series and arrive at the daily return of the three series. All the remaining analysis is performed at the daily return (log of the series) of WPI, CPI and Exchange rates.
At the stationary log series of the three stock exchanges, we perform the Grangers causality model in order to observe (i) whether the LOG of WPI granger causes the LOG of Exchange rate and/or at CPI; (ii) whether the LOG of Exchange rate granger causes the return at LOG of CPI and/or at LOG of WPI; and (iii) whether the LOG of CPI granger causes the LOG of WPI and/or at Exchange Rate.
The Granger (1969) approach to the question of whether x causes y is to see how much of the current y can be explained by past values of y and then to see whether adding lagged values of x can improve the explanation. y is said to be Granger-caused by x if x helps in the prediction of y , or equivalently if the coefficients on the lagged x s are statistically significant. It is pertinent to note that two-way causation is frequently the case; x Granger causes y and y Granger causes x. It is important to note that the statement x Granger causes y does not imply that y is the effect or the result of x. Granger causality measures precedence and information content but does not by itself indicate causality in the more common use of the term. In Grangers Causality, there are bivariate regressions of the under-mentioned form yt = o0 + o1 yt-1 + + ol yt-l + |1 xt-1 + + |l xt-l + ct xt = o0 + o1 xt-1 + + ol xt-l + |1 yt-1 + + |l yt-l + t (1.4)
for all possible pairs of (x, y) series in the group. In equation (1.4), we take lags ranging from 1 to l. In Grangers model, one can pick a lag length, l that corresponds to reasonable beliefs about the longest time over which one of the variables could help predict the other. The reported F- statistics are the Wald statistics for the joint hypothesis: |1 = |2 = = |t = 0 (1.5)
for each equation. The null hypothesis is that x does not Granger-cause y in the first regression and that y does not Granger-cause x in the second regression.
We follow the application of Grangers causality with the Vector Auto Regression (VAR) Model. The Vector Auto Regression (VAR) is commonly used for forecasting systems of interrelated time series and for analyzing the dynamic impact of random disturbances on the system of variables. The VAR approach sidesteps the need for structural modeling by treating 7
every endogenous variable in the system as a function of the lagged values of all of the endogenous variables in the system. The mathematical representation of a VAR is: yt = A1 y t-1 + + Ap y t-p + Bxt + ct (1.6) where yt is a k vector of endogenous variables, xt is a d vector of exogenous variables, A1, , Ap and B are matrices of coefficients to be estimated, and ct is a vector of innovations that may be contemporaneously correlated but are uncorrelated with their own lagged values and uncorrelated with all of the right-hand side variables.
Finally, we apply the Variance Decomposition Analysis in order to finally quantify the extent upto which the three indices are influenced by each other. While impulse response functions trace the effects of a shock to one endogenous variable on to the other variables in the VAR, variance decomposition separates the variation in an endogenous variable into the component shocks to the VAR. Thus, the variance decomposition provides information about the relative importance of each random innovation in affecting the variables in the VAR.
5.Results of the Study 5.1 Descriptive Statistics and Correlation matrix of Indian yearly data Table 5.1 Mean Median Min. Max. Variance Std.Dev. Coef. Var. Skewness Kurtosis Exchange Rates 45.40 45.31 41.35 48.61 6 2.46 5.4240 -0.168645 -0.47373 Bank Rates 6.03 6.00 6.00 6.25 0 0.09 1.4655 2.828427 8.00000
From the above table in which descriptive values of all the variables have been calculated shows that standard deviation is very high in case of GNI comparative to others which portrays nothing but that it is dispersed around its mean value by 13346.42 i.e., there is high volatility in its values. From the skewness measure we found that exchange rates and balance of payment is negatively skewed while bank rates are more positively skewed compared to other variables. In 8
case of kurtosis, all variables are negatively skewed except bank rates and balance of payments. Next step is to check out the correlation between the variables in consideration in this study. .Table 5.2
In the table 5.2 there is a positive correlation between Exchange rates - Bank rates, Exchange rates-Balance of payments, Bank rates - Balance of payments, W.P.I. - C.P.I., WPI - G.D.P., WPI-GNI, CPI - GDP and CPI-GNI. In the same table there is a negative correlation between Exchange rates - WPI, Exchange rates - CPI, Exchange rates-GDP, Exchange rates GNI, Bank Rates -WPI, Bank Rates -CPI, Bank Rates GDP, Bank Rates GNI, WPI - Balance of Payments, CPI - Balance of Payments. Balance of payments - GDP and Balance of payments - GNI. In table highlighted values are significant at 0.05 level of significance. When GDP is Dependent Variable Table 5.3 Coefficients a
Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) -5371.526 22912.473 -.234 .853 Exchange Rates -7.066 82.443 -.001 -.086 .946 Bank Rates 568.306 1987.636 .004 .286 .823 WPI 23.592 105.386 .031 .224 .860 CPI -16.876 102.537 -.025 -.165 .896 GNI 1.079 .254 1.097 4.244 .147 Balance of Payments 1.298 .263 .110 4.928 .127 a. Dependent Variable: GDP
From table 5.3, we can formulate the regression equation Y= a + bX, where in Y is the dependent variable (GDP) and X is the independent variable (Exchange rates, Bank rates, WPI, 9
CPI, GNI and Balance of Payments). Hence, we arrive at the regression equation GDP = - 5371.526 + (-7.066) Exchange Rates + (568.306) Bank Rates + (23.592) WPI + (-16.876)CPI + (1.079) GNI + (1.298) Balance of Payments. Using this regression equation to the entire series we find out the predicted values of GDP for the given values of independent variables. These values shall be presented later in this section. Table 5.4 ANOVA b
Model Sum of Squares df Mean Square F Sig. 1 Regression 1.207E9 6 2.012E8 5542.540 .010 a
Residual 36302.156 1 36302.156
Total 1.207E9 7
a. Predictors: (Constant), Balance of Payments, Exchange rates, Bank Rates, CPI, WPI, GNI b. Dependent Variable: GDP
Table 5.4 shows the anova table in which we find the sum of squares, mean square, f statistic and level of significance for regression equation as also for the residuals. The first important value that we can look at is the level of significance the value of which is found to be 0.010. This value is significant at 5% level of significance. Looking at the sum of squares, we find that the regression equation accounts for a major proportion of the values of the dependent variable (GDP). The detailed values of GDP at every level of independent variables are presented in the table 5.5 below. Table 5.5 Predicted & Residual Values
From table 5.5, we find that the predicted values in most of the cases are quite near to the 1% of the observed values except one case of year 2004, which indicates that there is a significant impact of the independent variables on the GDP. Besides presenting the predicted values of the series, table 5.5 also presents the residual value, standardized predicted value, standard error of the predicted value, Mahalanobis distance, deleted residual, and Cooks distance. When GNI is Dependent Variable Table 5.6 Coefficients a
Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 526.488 21214.056
.025 .984 Exchange Rates -5.809 74.386 -.001 -.078 .950 Bank Rates -198.429 1853.503 -.001 -.107 .932 WPI -1.903 97.361 -.002 -.020 .988 CPI 34.711 87.036 .050 .399 .758 Balance of Payments -1.149 .325 -.096 - 3.540 .175 GDP .878 .207 .864 4.244 .147 a. Dependent Variable: GNI From table 5.6, we can formulate the regression equation Y= a + bX, where in Y is the dependent variable (GNI) and X is the independent variable (Exchange rates, Bank rates, WPI, CPI, GDP and Balance of Payments). Hence, we arrive at the regression equation GDP = 526.488 + (-5.809) Exchange Rates + (-198.429) Bank Rates + (-1.903) WPI + (34.711) CPI + (-1.149) Balance of Payments + (.878) GDP. Using this regression equation to the entire series we find out the predicted values of GDP for the given values of independent variables. These values shall be presented later in this section in the table 5.7 Table 5.7 ANOVA Model Sum of Squares df Mean Square F Sig. 1 Regression 1.247E9 6 2.078E8 7040.431 .009 a
Residual 29516.652 1 29516.652
Total 1.247E9 7
a. Predictors: (Constant), GDP, Exchange Rates, Bank Rates, Balance of Payments, WPI, CPI b. Dependent Variable: GNI
Table 5.7 shows the anova table in which we find the sum of squares, mean square, f statistic and level of significance for regression equation as also for the residuals. The first important value that we can look at is the level of significance the value of which is found to be 0.009. This value is significant at 5% level of significance. Looking at the sum of squares, we find that the 11
regression equation accounts for a major proportion of the values of the dependent variable (GNI). The detailed values of GNI at every level of independent variables are presented in the table 5.8 below. Table 5.8
From table 5.8, we find that the predicted values in all cases are quite near to the 1% of the observed values from year 2002 to 2009, which indicates that there is a significant impact of the independent variables on the GNI. Besides presenting the predicted values of the series, table 5.8 also presents the residual value, standardized predicted value, standard error of the predicted value, Mahalanobis distance, deleted residual, and Cooks distance. For performing the econometric analysis, it is very essential for the researcher to make sure that the series under reference are stationary. In order to make the series stationary, we take log of the three series on which the further analysis shall be performed. In this way, three new variables are created and we assign those, names LOGExchange, LOGWPI and LOGCPI which denote the LOG of Exchange rate, WPI and CPI respectively. Going further in the paper, we shall discuss the linkages between the logs of exchange rate, WPI and CPI
Table 5.9 presents the descriptive statistics of the series of LOGExchange, LOGWPI and LOGCPI
12
Table 5.9 Descriptive statistics of the Exchange rates, WPI and CPI
Table 5.9 shows that the mean at the Exchange rates, WPI and CPI happens to be 45.4384, 130.7451 and 133.9192 respectively. Since there are a total of 96 observations for a period of 8 years. The value of median is highest in the case of WPI than the CPI and Exchange rates which are 128.9000, 127.5890 and 45.5355 respectively. The variance and standard deviation in the case of CPI is higher than the WPI and Exchange rate which shows that the volatility is more in CPI than the others. In Figure 1 to 4 present the line graphs of the LOG of WPI, LOG of CPI and LOG of Exchange rate of India. While the return on WPI, CPI and Exchange Rates are individually presented in figures 1 to 3, figure 4 presents common line graphs for the three macro economic variables under study.
5.2 Stationarity and Causality Analysis of Indian Monthly data
After all these statistics stationarity tests are carried out on the variables because to apply Granger causality, first the series have to be made stationary. Augmented Dickey Fuller (ADF)) test have been done and after the application of these tests all the series have been found stationary at various significance levels.
LOGExchange LOGWPI LOGCPI Valid N 96 96 96 Mean 45.4384 130.7451 133.9192 Median 45.5355 128.9000 127.5890 Frequency 1 2 5 Minimum 39.3740 104.7800 113.8000 Maximum 51.2290 161.8180 184.6630 Variance 7.5766 266.5260 311.2582 Std.Dev. 2.75256 16.32562 17.64251 Coef.Var. 6.05778 12.48660 13.17400 Skewness -0.456286 0.217605 1.302944 Kurtosis -0.29634 -1.07108 0.94365 13
Figures 1 to 4 demonstrate the value of the three macro economic variables. It is indicated from the figures that values at all the three macro economic variables are stationary in nature. In order to further check the stationarity of the three series, we perform the Unit Root Test in order to further confirm the same. 15
The unit-root test is performed on the three series in order to test the null hypothesis that the series has a unit root. The findings of the unit-root test and the augmented Dickey- Fuller test are shown below in the following tables.
Table 5.10 Unit root test on LOG exchange
Variable Coefficient Std. Error t-Statistic Prob. LOGXCHNG_NA(-1) -0.991078 0.104187 -9.512455 0.0000 C -0.000457 0.002292 -0.199305 0.8425 R-squared 0.495854 Mean dependent var -6.41E-05 Adjusted R-squared 0.490374 S.D. dependent var 0.031124 S.E. of regression 0.022219 Akaike info criterion -4.754684 Sum squared resid 0.045419 Schwarz criterion -4.700571 Log likelihood 225.4701 Hannan-Quinn criter. -4.732826 F-statistic 90.48679 Durbin-Watson stat 1.999856 Prob(F-statistic) 0.000000
By the way of unit-root test, the null hypothesis that series Log of Exchange Rates has a unit-root is tested. Probability value of less than 0.05 in above table shows that the Null hypothesis is rejected and the variable does not have a unit-root, which confirms that the series is stationary. Hence, the econometric models can now be applied on the series.
Table 5.11 Unit root test on LOG CPI
Variable Coefficient Std. Error t-Statistic Prob. LOGCPI_NA(-1) -0.954503 0.104062 -9.172398 0.0000 C 0.004433 0.001739 2.549087 0.0125 R-squared 0.477667 Mean dependent var 8.68E-05 Adjusted R-squared 0.471990 S.D. dependent var 0.022325 S.E. of regression 0.016223 Akaike info criterion -5.383769 Sum squared resid 0.024212 Schwarz criterion -5.329657 Log likelihood 255.0372 Hannan-Quinn criter. -5.361912 F-statistic 84.13288 Durbin-Watson stat 1.999738 Prob(F-statistic) 0.000000 The probability value of unit-root test in table 5.11 points towards the fact that the null hypothesis can be rejected at 0.05 level of significance. It implies that the LOG of CPI of India is also a stationary one. Hence, the econometric models can now be applied on the series. 16
Table 5.12 Unit root test on LOG WPI
Variable Coefficient Std. Error t-Statistic Prob. LOGWPI_NA(-1) -0.931885 0.104279 -8.936453 0.0000 C 0.004203 0.001063 3.954774 0.0002 R-squared 0.464681 Mean dependent var -0.000105 Adjusted R- squared 0.458862 S.D. dependent var 0.012484 S.E. of regression 0.009184 Akaike info criterion -6.521737 Sum squared resid 0.007759 Schwarz criterion -6.467624 Log likelihood 308.5216 Hannan-Quinn criter. -6.499879 F-statistic 79.86019 Durbin-Watson stat 2.015654 Prob(F-statistic) 0.000000 The probability value of unit-root test in table 5.12 points towards the fact that the null hypothesis can be rejected at 0.05 level of significance. It implies that the LOG of WPI of India is also a stationary one.
Table 5.13 Granger Causality test on India monthly data
LOGXCHNG_NA does not Granger Cause LOGCPI_NA 93 0.01602 0.9841 LOGCPI_NA does not Granger Cause LOGXCHNG_NA 1.42415 0.2462
LOGWPI_NA does not Granger Cause LOGCPI_NA 93 0.19801 0.8207 LOGCPI_NA does not Granger Cause LOGWPI_NA 0.53034 0.5903
LOGWPI_NA does not Granger Cause LOGXCHNG_NA 93 0.66470 0.5170 LOGXCHNG_NA does not Granger Cause LOGWPI_NA 1.11568 0.3323
Table 5.13 presents the results about the application of Grangers Causality model to the WPI, CPI and Exchange Rates of India. Null hypothesis in the case of Grangers causality model is that A does not granger cause B. On those lines, table 6 tests the hypotheses about the three variables in pairs. The results show that the probability value for the hypotheses Exchange rate does not Granger Cause LOGCPI and LOGCPI does not Granger Cause LOGEXCHNG is more than 0.05 which means that in both the cases null hypotheses can be accepted. And the same results are observed in the case of LOGWPI & LOGCPI and LOGWPI & LOGEXCHNG. 17
Now we apply the Vector Auto Regression (VAR) model on the series under reference in order to further confirm the results produced by the Grangers Causality model.
In table 5.14, we present the application of Vector Auto Regression (VAR) Model at the three stock exchanges.
Table 5.14 Vector Auto Regression test on Indias monthly data
By the application of VAR Model, we observe that the integration of macroeconomic variables with the other can be established if the p-value is more than 1.96. Table 5.14 shows that the LOGCPI at the lag of 1 and 2, does not have any influence on LOGCPI, LOGWPI and LOGEXCHNG.. Similarly, LOGWPI at a lag of 1 and 2 does not have any influences on the LOGCPI, LOGWPI and LOGXHNG. In LOGXCHNG, the table reveals that LOGXCHNG at a lag of 1 and 2 does not have any effect on the LOGCPI, LOGWPI and LOGXCHNG. Table 5.15 Variance Decomposition on Indias monthly data
Finally, the Variance Decomposition Analysis of the three macro economic variables is presented in the table 5.15. The table decomposes the values at the three macro economic variables for a period ranging from 1 to 10. The Variance Decomposition Analysis as presented in table 5.15. It implies that on LOGCPI, the impact of other two macro economic variables is negligible. Rather the LOGCPI itself with the lag of 1 through 10 impacts the LOGCPI in the current period. However, the table reveals that in the case of LOGWPI, there is visible impact of LOGCPI for periods 1 to 10 and LOGEXCHNG for the periods 2 to 10. In LOG WPI the impact on LOGCPI is more than the LOGEXCHNG. In the case of LOGEXCHNG, there is also visible impact of LOGCPI and LOGWPI for the periods of 2 to 10. The impact is more in the case of LOGCP than the LOGWPI. Variance Decomposition Analysis shows that the macro economic variables under study are not much influenced by each other. 5.3 Descriptive Statistics and Correlation matrix of Sri Lanka yearly data Table 5.16 Mean Median Minimu m Maximu m Variance Std. Dev. Coef. Var. Skewnes s Kurtosis Exchange Rates 104 103 96 115 4.685914E+01 7 6.5846 0.39081 -1.01195 Bank Rates 15 15 15 18 1.125000E+00 1 6.8986 2.82843 8.00000
From the above table in which descriptive values of all the variables have been calculated shows that standard deviation is very high in case of Balance of Payments comparative to others which portrays nothing but that it is dispersed around its mean value by 163888 i.e., there is high volatility in its values. From the skewness measure we found that only balance of payment is negatively skewed while bank rates are more positively skewed compared to other variables. In case of kurtosis, all variables are negatively skewed except bank rates and balance of payments. Next step is to check out the correlation between the variables in consideration in this study.
In the table 5.17 there is a positive correlation between Exchange rates -WPI, Exchange rates- CPI, Exchange rates-GDP, Exchange rates- GNI, Bank rates - Balance of payments, W.P.I. - C.P.I., WPI - G.D.P. and WPI-GNI In the same table there is a negative correlation between Exchange rates Balance of Payments, Exchange rates Bank Rates, Bank Rates -WPI, Bank Rates -CPI, Bank Rates GDP, Bank Rates GNI, WPI - Balance of Payments, CPI - Balance of Payments, GDP-Balance of payments and GNI-Balance of payments. In table highlighted values are significant at 0.05 level of significance. GDP as a Dependent variable Table 5.18 Coefficients a
Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) -199957.986 227360.801 -.879 .541 xchnge -203.156 1830.257 -.001 -.111 .930 Bank Rates 4508.881 1747.712 .004 2.580 .235 WPI -854.678 599.947 -.043 -1.425 .390 CPI 3164.948 1370.841 .136 2.309 .260 GNI .911 .046 .893 19.712 .032 Balance of Payments -.151 .040 -.021 -3.750 .166 a. Dependent Variable: GDP 21
From table 5.18, we can formulate the regression equation Y= a + bX, where in Y is the dependent variable (GDP) and X is the independent variable (Exchange rates, Bank rates, WPI, CPI, GNI and Balance of Payments). Hence, we arrive at the regression equation GDP = -199957.986 + (-203.156) Exchange Rates + (4508.881) Bank Rates + (-854.678) WPI + (3164.948) CPI + (0.911) GNI + (-0.151) Balance of Payments. Using this regression equation to the entire series we find out the predicted values of GDP for the given values of independent variables. These values shall be presented later in this section in the table 5.19 Table 4.19 ANOVA b
Model Sum of Squares df Mean Square F Sig. 1 Regression 1.003E13 6 1.671E12 185498.790 .002 a
Residual 9007333.846 1 9007333.846 Total 1.003E13 7 a. Predictors: (Constant), Balance of Payments, Bank Rates, Exchange Rate, CPI, WPI, GNI b. Dependent Variable: GDP Table 5.19 shows the anova table in which we find the sum of squares, mean square, f statistic and level of significance for regression equation as also for the residuals. The first important value that we can look at is the level of significance the value of which is found to be 0.002. This value is significant at 5% level of significance. Looking at the sum of squares, we find that the regression equation accounts for a major proportion of the values of the dependent variable (GDP). The detailed values of GDP at every level of independent variables are presented in the table 5.20 below. Table 5.20 Predicted & Residual Values
Medi an 2695731 2696781 249.00 -0.22781 0.082966 2972.344 5.991052 12845.3 5.13798
From table 5.20, we find that the predicted values in all cases are quite near to the 1% of the observed values from year 2002 to 2009, which indicates that there is a significant impact of the independent variables on the GDP. Besides presenting the predicted values of the series, table 5.20 also presents the residual value, standardized predicted value, standard error of the predicted value, Mahalanobis distance, deleted residual, and Cooks distance. GNI as a dependent variable Table 4.21 Coefficients a
Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 206911.824 259697.553 .797 .572 Exchange Rates 318.889 1994.159 .002 .160 .899 Bank Rates -4890.749 2050.417 -.004 -2.385 .253 WPI 910.217 694.910 .047 1.310 .415 CPI -3392.804 1672.082 -.149 -2.029 .292 GDP 1.095 .056 1.117 19.712 .032 Balance of Payments .164 .050 .023 3.299 .187 a. Dependent Variable: GNI
From table 5.21, we can formulate the regression equation Y= a + bX, where in Y is the dependent variable (GNI) and X is the independent variable (Exchange rates, Bank rates, WPI, CPI, GDP and Balance of Payments). Hence, we arrive at the regression equation GDP = 206911.82 + (318.889) Exchange Rates+ (-4890.749)Bank Rates + (910.217)WPI + (-3392.804) CPI+ (0.164) Balance of Payments + (1.095) GDP. Using this regression equation to the entire series we find out the predicted values of GDP for the given values of independent variables. These values shall be presented later in this section in the table 5.22 Table 5.22 ANOVA b
Model Sum of Squares df Mean Square F Sig. 1 Regression 9.645E12 6 1.608E12 148397.650 .002 a
Residual 1.083E7 1 1.083E7 Total 9.645E12 7 a. Predictors: (Constant), Balance of Payments, Bank Rates, Exchange Rates, CPI, WPI, GDP b. Dependent Variable: GNI
Table 5.22 shows the anova table in which we find the sum of squares, mean square, f statistic and level of significance for regression equation as also for the residuals. The first important 23
value that we can look at is the level of significance the value of which is found to be 0.002. This value is significant at 5% level of significance. Looking at the sum of squares, we find that the regression equation accounts for a major proportion of the values of the dependent variable (GNI). The detailed values of GNI at every level of independent variables are presented in the table 5.23 below. Table 5.23 Predicted & Residual Values
From table 5.23, we find that the predicted values in all cases are quite near to the 1% of the observed values from year 2002 to 2009, which indicates that there is a significant impact of the independent variables on the GNI. Besides presenting the predicted values of the series, table 5.23 also presents the residual value, standardized predicted value, standard error of the predicted value, Mahalanobis distance, deleted residual, and Cooks distance. 5.4 Stationarity and Causality Analysis of Indian Monthly data Table 5.24 Exchange Rates WPI CPI Valid N 96 96 96 Mean 103.9292 184.4419 186.0590 Median 102.7625 168.3500 167.2500 Mode 114.2580 233.5400 Multiple Frequency 2 4 2 Minimum 93.3830 120.3540 129.6000 Maximum 116.9210 294.7000 277.6590 Variance 43.077 3036.495 2224.671 24
Std. Deviation 6.56329 55.10440 47.16642 Coef. Of variance 6.31516 29.87630 25.35025 Skewness 0.263307 0.529218 0.525176 Kurtosis -1.18682 -1.10259 -1.27886 As we did in the case of Indias data same as that Figure 5 to 8 present the line graphs of the values of WPI, CPI and Exchange rates of Sri Lanka. While the return on WPI, CPI and Exchange Rates are individually presented in figures 5 to 7, figure 8 presents common line graphs for the three macro economic variables under study. Figure 5
Figures 5 to 8 demonstrate the value of the three macro economic variables. It is indicated from the figures that values at all the three macro economic variables are stationary in nature. In order to further check the stationarity of the three series, we perform the Unit Root Test in order to further confirm the same.
The unit-root test is performed on the three series in order to test the null hypothesis that the series has a unit root. The findings of the unit-root test and the augmented Dickey-Fuller test are shown below in Table 5.25
Table 5.25 26
Unit Root test on CPI
Augmented Dickey-Fuller Test Equation Dependent Variable: D(LOGCPI_NA) Method: Least Squares Date: 03/23/11 Time: 10:55 Sample (adjusted): 3 95 Included observations: 93 after adjustments
R-squared 0.432031 Mean dependent var 7.74E-05 Adjusted R-squared 0.419410 S.D. dependent var 0.030323 S.E. of regression 0.023105 Akaike info criterion -4.665803 Sum squared resid 0.048046 Schwarz criterion -4.584106 Log likelihood 219.9598 Hannan-Quinn criter. -4.632816 F-statistic 34.22974 Durbin-Watson stat 2.100575 Prob(F-statistic) 0.000000
The probability value of unit-root test in table 5.25 points towards the fact that the null hypothesis can be rejected at 0.05 level of significance. It implies that LOG of CPI of Sri Lanka is also a stationary one. Hence, the econometric models can now be applied on the series..
Table 5.26 Unit root test on WPI
Variable Coefficient Std. Error t-Statistic Prob.
LOGWPI_NA(-1) -1.212027 0.101879 -11.89667 0.0000 C 0.008339 0.003974 2.098294 0.0386
R-squared 0.606048 Mean dependent var 1.35E-05 Adjusted R-squared 0.601766 S.D. dependent var 0.060103 S.E. of regression 0.037929 Akaike info criterion -3.685174 Sum squared resid 0.132349 Schwarz criterion -3.631061 Log likelihood 175.2032 Hannan-Quinn criter. -3.663316 F-statistic 141.5308 Durbin-Watson stat 1.991122 Prob(F-statistic) 0.000000
The probability value of unit-root test in table 5.26 points towards the fact that the null hypothesis can be rejected at 0.05 level of significance. It implies that the LOG of WPI of Sri Lanka is also a stationary one. Hence, the econometric models can now be applied on the series..
Table 5.27 27
Unit Root Test on Exchange Rates
Variable Coefficient Std. Error t-Statistic Prob.
LOGEXCHNG_NA(-1) -0.792874 0.102070 -7.767954 0.0000 C 0.001687 0.001031 1.637207 0.1050
R-squared 0.396092 Mean dependent var -3.30E-05 Adjusted R-squared 0.389528 S.D. dependent var 0.012489 S.E. of regression 0.009758 Akaike info criterion -6.400354 Sum squared resid 0.008761 Schwarz criterion -6.346242 Log likelihood 302.8167 Hannan-Quinn criter. -6.378497 F-statistic 60.34111 Durbin-Watson stat 1.969786 Prob(F-statistic) 0.000000
The probability value of unit-root test in table 5.27, points towards the fact that the null hypothesis can be rejected at 0.05 level of significance. It implies that the LOG of Exchange Rate of Sri Lanka is also a stationary one. Hence, the econometric models can now be applied on the series. Table 5.28 Granger Causality test on Sri Lanka monthly data
Lags: 2
Null Hypothesis: Obs F-Statistic Prob.
LOGEXCHNG_NA does not Granger Cause LOGCPI_NA 93 0.86114 0.4262 LOGCPI_NA does not Granger Cause LOGEXCHNG_NA 0.08428 0.9193
LOGWPI_NA does not Granger Cause LOGCPI_NA 93 2.11268 0.1270 LOGCPI_NA does not Granger Cause LOGWPI_NA 0.49521 0.6111
LOGWPI_NA does not Granger Cause LOGEXCHNG_NA 93 1.12065 0.3307 LOGEXCHNG_NA does not Granger Cause LOGWPI_NA 1.07596 0.3454
Table 5.28 presents the results about the application of Grangers Causality model to the WPI, CPI and Exchange Rates of India. Null hypothesis in the case of Grangers causality model is that A does not granger cause B. On those lines, table 5.28 tests the hypotheses about the three variables in pairs. The results show that the probability value for the hypotheses Exchange rate does not Granger Cause LOGCPI and LOGCPI does not Granger Cause LOGEXCHNG is more than 0.05 which means that in both the cases null hypotheses can be accepted. And the same results are observed in the case of LOGWPI & LOGCPI and LOGWPI & LOGEXCHNG.
Now we apply the Vector Auto Regression (VAR) model on the series under reference in order to further confirm the results produced by the Grangers Causality model.
In table 5.29, we present the application of Vector Auto Regression (VAR) Model at the three stock exchanges. Table 5.29 28
Vector Auto Regression test on Sri Lankas monthly data
Vector Autoregression Estimates Date: 03/11/11 Time: 16:30 Sample (adjusted): 3 95 Included observations: 93 after adjustments Standard errors in ( ) & t-statistics in [ ]
By the application of VAR Model, we observe that the integration of macroeconomic variables with the other can be established if the p-value is more than 1.96. Table 5.29 shows that the LOGCPI at the lag of 1 and 2, does not have any influence on LOGWPI and LOGEXCHNG. However, it influences the returns at LOGCPI in period 0. Similarly, LOGEXCHNG at a lag of 1 and 2 does not have any influences on the LOGCPI, LOGWPI and LOGEXHNG. In LOGWPI at the lag 1 does not have any influence on LOGCPI and LOGEXCHNG but it influences the return at LOGWPI in period 0. In LOGWPI at lag 2 LOGWPI have influence on LOGCPI but it does not influence LOGEXCHNG and LOGWPI. Table 5.30 Variance Decomposition on Sri Lankas monthly data
Finally, the Variance Decomposition Analysis of the three macro economic variables is presented in the table 5.30. The table decomposes the values at the three macro economic variables for a period ranging from 1 to 10.
The Variance Decomposition Analysis as presented in table 5.30. It implies that on LOGCPI, the impact of other two macro economic variables is visible. The impact is near about constant in the LOG of Exchange rate but in LOG of WPI impact increases step by step than the previous one. However, the table reveals that in the case of LOGEXCHNG, there is visible impact of LOGWPI for periods 2 to 10 and no impact on LOGCPI. In the case of LOGWPI, there is also visible impact of LOGEXCHNG for the periods of 3 to 10. Variance Decomposition Analysis shows that the macro economic variables under study are not much influenced by each other.
Conclusion The aim of this research is to find out the impact of macroeconomic variables on GDP growth and study the pattern of CPI, WPI, GDP, GNI and Rate of interest in India and Sri Lanka. To solve this basic purpose monthly data was used from 2002 to 2009 of both of the countries and the basic and believed to be indicator variables were used and studied and analyzed by first applying the basic statistical tools like correlation and descriptive statistical tools and finally regression, unit root test, Granger causality, VAR and Variance decomposition models. The application of Unit-root test (Augmented Dickey-Fuller test) reveals that the null hypothesis can be rejected at 0.05 level of significance. It implies that the series of WPI, CPI and Exchange rates of India and Sri Lanka are stationary. Grangers Causality Model when applied to the three series indicates that probability value for the hypotheses Exchange rate does not Granger Cause LOGCPI and LOGCPI does not Granger Cause LOGEXCHNG in Indian data and the probability value for the hypotheses Exchange rate does not Granger Cause LOGCPI and LOGCPI does not Granger Cause LOGEXCHNG is more than 0.05 which means that in both the cases null hypotheses can be accepted. And the same results are observed in the case of Sri Lankas data. None of the other variables happen to Granger cause any of the other variables under study. The application of the VAR model implies that the LOGCPI at the lag of 1 and 2, does not have any influence on LOGCPI, LOGWPI and LOGEXCHNG. Similarly, LOGWPI at a lag of 1 and 2 does not have any influence on the LOGCPI, LOGWPI and LOGXHNG. In LOGXCHNG, the table reveals that LOGXCHNG at a lag of 1 and 2 does not have any effect on the LOGCPI, LOGWPI and LOGXCHNG in Indian data and in the case of Sri Lankas data Vector Auto regression (VAR) model implies that LOGCPI at the lag of 1 and 2, does not have any influence on LOGWPI and LOGEXCHNG.. However, it influences the returns at LOGCPI in period 0.Similarly, LOGEXCHNG at a lag of 1 and 2 does not have any influences on the LOGCPI, LOGWPI and LOGEXHNG. In LOGWPI at the lag 1 does not have any influence on 31
LOGCPI and LOGEXCHNG but it influences the return at LOGWPI in period 0. In LOGWPI at lag 2 LOGWPI have influence on LOGCPI but it does not influence LOGEXCHNG and LOGWPI. The Variance Decomposition Analysis implies that on LOGCPI, the impact of other two macro economic variables is negligible. Rather the LOGCPI itself with the lag of 1 through 10 impacts the LOGCPI in the current period. However, in the case of LOGWPI, there is visible impact of LOGCPI for periods 1 to 10 and LOGEXCHNG for the periods 2 to 10. In LOG WPI the impact on LOGCPI is more than the LOGEXCHNG. In the case of LOGEXCHNG, there is also visible impact of LOGCPI and LOGWPI for the periods of 2 to 10. The impact is more in the case of LOGCP than the LOGWPI. In case of Sri Lanka data it implies that on LOGCPI, the impact of other two macro economic variables is visible. The impact is near about constant in the LOG of Exchange rate but in LOG of WPI impact increases step by step than the previous one. However, the table reveals that in the case of LOGEXCHNG, there is visible impact of LOGWPI for periods 2 to 10 and no impact on LOGCPI. In the case of LOGWPI, there is also visible impact of LOGEXCHNG for the periods of 3 to 10.
After applying all the models on the data of both the countries the results do not lead us to any clear-cut conclusion because the results from all the models are different. Granger model and VAR model indicates that LOGCPI, LOGWPI and LOGEXCHNG does not have any influence on each other in the case of both of the countries but the Variance decomposition model shows visible impact of macroeconomic variables on each other in some of the cases in Indian and Sri Lankan data which is mention above.
References 32
1. Asprem, M. 1989, Stock prices, asset portfolios and macroeconomic variables in ten European countries, Journal of Banking and Finance, 13, 589-612. 2. Chatrath, A., S. Ramchander and F. Song (1997), Stock prices, inflation and output: evidence from India. Applied Financial Economics, Volume 7(4), 439-445. 3. Chen, N. F., Roll, R. & Ross, S. 1986. Economic forces and the stock market. Journal of Business 59(3): 83-403. (no abstract) 4. Chong, C. S. & Goh, K. L. 2003. Linkages of economic activity, stock prices and monetary policy: the case of Malaysia. 5. Darrat, A. F. and T. K. Mukherjee, 1987, The Behavior of the Stock Market in a Developing Economy, Economics Letters 22, 273-78. 6. Fama E. F. & Schwert, W.G. 1977. Asset returns and inflation. Journal of Financial Economics 5: 115-146. 7. Fama, E.F. 1981, Stock returns, real activity, inflation and money, American Economic Review, 71, 545-565. 8. Fama, E.F. and K.R. French 1989, Business conditions and expected returns on stocks and bonds, Journal of Financial Economics, 25, 23-49. 9. Fama, E.F. 1990, Stock returns, expected returns and real activity, Journal of Finance, 45, 1089-1108. 10. Ferson, W. and C. Harvey 1991, The variation of economic risk premiums, Journal of Political Economy, 99, 385-415. 11. Geske, R. and R. Roll 1983, The fiscal and monetary linkage between stock returns and inflation, Journal of Finance, 38, 1-33. 12. Gjerde, ., & Saettem, F. (1999). Causal relations among stock returns and macroeconomic variables in a small, open economy. Journal of International Finance Markets Institutions and Money, 9, 61-74. 13. Granger, C. W. J. 1986. Developments in the study of cointegrated economic variables.Oxford Bulletin of Economics and Statistics 48: 213-228 14. Hendry, D. F. 1986. Econometric modeling with cointegrated variables: An overview. Oxford Bulletin of Economics and Statistics 48(3) 201-212. 15. Ibrahim, M.H. (1999), "Macroeconomic variables and stock prices in Malaysia: an empirical analysis", Asian Economic Journal, Vol. 13 No.2, pp.21931. 16. Jaffe, J. & Mandelkar, G. 1976. The Fisher effect for risky assets: An empirical investigation. Journal of Finance 31: 447-456. (no abstract) 17. James, C., S. Koreisha and M. Partch 1985, A VARMA analysis of the causal relations among stock returns real output and nominal interest rates, Journal of Finance, 40, 1375- 1384. 18. Johansen, S. & Juselius, K. 1990. Maximum likelihood estimation and inference on cointegration with application to the demand for money. Oxford Bulletin of Economics and Statistics 52: 169-210. 33
19. Lee, B. 1992, Causal relations among stock returns, interest rates, real activity and inflation, Journal of Finance, 47, 1591-1603. 20. Mandelker, G. and K. Tandon 1985, Common stock returns, real activity, money and inflation: Some international evidence Journal of International Money and Finance, 4, 267-268. 21. Martinez, M. A., Rubio, G. (1989). Arbitrage pricing with macroeconomic variables: an empirical investigation using Spanish data. Working Paper, Universidad del Pais Vasco. 22. Maysami, R. C. & Koh, T. S. 2000. A vector error correction model of the Singapore stock market. International Review of Economics and Finance 9: 79-96. 23. Mukherjee, T. K. & Naka, A. 1995. Dynamic relations between macroeconomic variables and the Japanese stock market: an application of a vector error correction model. The Journal of Financial Research 18(2): 223-237. 24. Naka A., Mukherjee, T. and Tufte D., 1998, Macroeconomic Variables and the Performance of the Indian Stock Market, Working Papers, Department of Economics & Finance, University of New Orleans. 25. Nelson, C. R. 1976. Inflation and rates of return on common stocks. Journal of Finance 31(2): 471-483 26. Oberuc, R. E. 2004. Dynamic Portfolio Theory and Management: Using Active Asset Allocation to Improve Profits and Reduce Risk, Mc-Graw Hills, U.S. 27. Schwert, W. G. (1989). Why does stock market volatility change over time? Journal of Finance, 44, 1115-1153. 28. Schwert, G.W. 1990, Stock returns and real activity: A century of evidence,Journal of Finance, 45, 1237-1257. 29. Vuyuri, S. 2005. Relationship between real and financial variables in India: A cointegration analysis. Available at http://ssrn.com/abstract=711541 30. Wongbanpo, P. & Sharma, S. C. (2002). Stock market and macroeconomic fundamental dynamic interactions: ASEAN-5 countries. Journal of Asian Economics, 13, 27-51.