conflicting results Shyamal Paul a , Rabindra N. Bhattacharya b, * a Khalisani Mahavidyalaya, Chandannagar-712138, W. B., India b Department of Economics, Kalyani University-741235, W. B., India Available online 25 September 2004 Abstract This note examines the different direction of causal relation between energy consumption and economic growth in India. Applying EngleGranger cointegration approach combined with the standard Granger causality test on Indian data for the period 19501996, we find that bi-directional causality exists between energy consumption and economic growth. Further, we apply Johansen multivariate cointegration technique on the different set of variables. The same direction of causality exists between energy consumption and economic growth. This is different from the results obtained in earlier studies. D 2004 Elsevier B.V. All rights reserved. JEL classification: Q 41 ; Q 43 Keywords: Energy consumption; Economic growth; India; Granger causality; Cointegration 1. Introduction All production and many consumption activities involve energy as an essential input. It is the key source of economic growth, industrialization and urbanization. On the other hand, economic growth, industrialization and urbanization may induce use of more energy, 0140-9883/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.eneco.2004.07.002 * Corresponding author. Tel.: +91 33 2582 8750x288/289; fax: +91 33 2582 8282. E-mail addresses: shyeco@yahoo.co.in (S. Paul)8 rnbeco@yahoo.com (R.N. Bhattacharya). Energy Economics 26 (2004) 977983 www.elsevier.com/locate/eneco particularly commercial energy. India has been passing through an economic reform since 1991, the general aim of which is to quadruple its economic growth and remove the problems of poverty and unemployment. One of the obstacles to achieve these objectives has been the frequent occurrence of energy shortages in the economy. In India, commercial energy consumption over time has grown at a compound annual growth rate (CAGR) of about 6% which is more than the CAGR of GDP during the last two decades. 1 So the relation between energy consumption and economic growth is of great interest to the energy economists. It is not possible to achieve high growth in one, without keeping pace with another. The direction of causation between energy consumption and economic growth has significant implications. If, for example, there exists unidirectional causality running from economic growth to energy consumption, it may be implied that energy conservation policies may be implemented with little adverse or no effects on economic growth. On the other hand, if unidirectional causality runs from energy consumption to income, reducing energy consumption could lead to fall in income. The finding of no causality in either direction, the so-called dneutrality hypothesisT would imply that energy conservation policies do not affect economic growth (Asafu- Adjaye, 2000). However, in India not much attention has been devoted to investigate the causal relationship between economic growth and energy consumption. In some earlier studies Pachuri (1977) and Tyner (1978) using regression approach have found a strong relation between energy consumption and economic growth in India. Cheng (1999) in his quest for causality has established a unidirectional causal relation from economic growth to energy consumption. In contrast, Asafu-Adjaye (2000) has estimated a unidirectional Granger causality from energy consumption to income. The finding of Asafu-Adjaye (2000) does not support the finding of Cheng (1999). Our empirical result identifies bi-directional causality between energy consumption and economic growth. The present analysis attempts to discuss how our finding contrasts with the findings of Cheng (1999) and Asafu-Adjaye (2000). 2. Empirical evidence The empirical findings on the causal issue between energy consumption and economic growth have been mixed or conflicting. The reason for the disparate and often conflicting empirical findings lies in the variety of approaches of time frame considered and testing procedures employed in the analysis. Cheng (1999) has applied JohansenHsiaos version of the Granger causality method on the Indian data for the time period 19521995. This study finds that energy consumption, economic growth, capital and labour are cointegrated and the direction of causality runs from economic growth to energy consumption both in the short-run and 1 It is worthwhile to note that past trend in energy consumption does not fully represent the growth of energy demand. To a large extent it reflects the net availability (Bhattacharya and Paul, 2001). S. Paul, R.N. Bhattacharya / Energy Economics 26 (2004) 977983 978 in the long run. No causal relation is found from energy consumption to economic growth. The modeling strategy adopted in Asafu-Adjaye (2000) is based on the Engle and Granger (1987) methodology. He used three variablescommercial energy use, real income and energy prices proxied by the consumer price index for the period 1973 1995 and found unidirectional Granger causality running from energy consumption to economic growth both in the short run and in the long run. This discrepancy in results between Cheng (1999) and Asafu-Adjaye (2000) may be due either to the choice of the sample period or to the measure of the variables or to the choice of the methodology or to both. Now, based on our exercise, let us indicate the result of causality between energy consumption and economic growth in India using a sample of annual data covering the period 19501996. If a linear combination of the non-stationary variables is non- stationary, then the Standard Granger causality test should be adopted. If a linear combination of the non-stationary is stationary, then error-correction model (ECM) should be adopted. We adopt the standard Grangers causality test (1969) when GDP is treated as a dependent variable and then the error-correction model when energy is considered as a dependent variable. Our combined results of the standard Granger causality test and the EngleGranger error correction approach imply bi-directional causal relation between energy consumption and economic growth. The Johansen multivariate approach (1991) at lag length 3 reveals that the long-run causality exists from economic growth to energy consumption and the short-run causality runs from energy consumption to economic growth. 2 That is, a bi-directional causality exists between energy consumption and economic growth (results are shown in Appendix A). The results of the standard Granger test and the EngleGranger approach show only unidirectional causality between energy consumption and economic growth. But the direction of causality in the approaches is different. In the EngleGranger model, the long-run causality runs from economic growth to energy consumption and in the standard Granger model, the causality runs from energy consumption to economic growth. Use of Johansen multivariate approach reveals a short-run causality from energy consumption to economic growth and the long-run causality from economic growth to energy consumption. The interpretation of this short-run and long-run divergence in causal relation between energy consumption and economic growth can be given as follows. In the short-run, energy, particularly commercial energy acts as an engine of economic growth. It is a very important primary input in the aggregate production function. In the context of a less developed country like India, the substitution of non-commercial energy by commercial energy is likely to be limited in the short-run. But in the long run, with the growth of income and technological progress, it is likely that more and more non- commercial energy will be substituted by commercial energy. 2 The appropriate selection of lag will generate white noise residual process. Small lag misspecifies the model and large lag indicates that the degrees of freedom are wasted (Hamilton, 1994). S. Paul, R.N. Bhattacharya / Energy Economics 26 (2004) 977983 979 3. Discussion and conclusions This note has demonstrated that the causal relation between energy consumption and economic growth in India is mixed or conflicting. The results obtained from one approach are found to be different from another. Our empirical results using the standard Granger causality test reveal a unidirectional causal relation from energy consumption to economic growth. This process does not show any causal effect from economic growth to energy consumption. However, use of EngleGranger cointegra- tion process exhibits a unidirectional long-run causality from GDP along with capital to energy consumption. No short-run relation is found. The results of the Engle Granger approach combined with the standard Granger causality process indicate that there is bi-directional causality between energy consumption and economic growth. The long-run causal relation runs from GDP to energy consumption and the short-run causal relation runs from energy consumption to GDP. The same results are obtained by using Johansen multivariate approach on the different combination of variables. These results are at variance with the results obtained by Cheng (1999) and Asafu- Adjaye (2000). Acknowledgements We are grateful to an anonymous Associate Editor of this journal for some helpful comments on an earlier draft of this paper. However, usual disclaimer applies. Appendix A In the following tables variables used are named as E=commercial energy consumption, G=national income of the Indian economy, F=gross fixed capital formation of the economy and L=population, a proxy for labour. All variables are expressed in terms of natural logarithm. Table 1 Cointegration tests (EngleGranger procedure) Regression equation: Y t X 1t X 2t a b c q (statistics) n G t C t F t 6.08 0.45 0.38 0.35 (2.86) 1 E t G t F t 7.50 0.69 0.43 0.53* (4.04) 1 F t E t G t 0.21 0.53 0.62 0.61* (5.50) 1 (a) Absolute t-statistics are in the parentheses. The 5% critical value for the t-statistic is 3.79 for 50 observations; see Charemza and Deadman (1999), Table 2. (b) The parameter n was chosen in such a way that there was no serial correlation in the residuals in Engle and Granger (1987) two-step procedure of cointegration test. (c) * denotes statistically significant. S. Paul, R.N. Bhattacharya / Energy Economics 26 (2004) 977983 980 Table 3 Test statistics of temporal causality using ECM based on cointegrating regression of EngleGranger procedure Controlled variable 1st manipulated variable 2nd manipulated variable FPE F-value Error-correction (Ec t 1 ) 1st Regression equation E(i=1) 2.100329103 E(i=1) F( j=1) G(m=1) 2.162463403 0.82 E(i=1) F( j=1) G(m=1) 2.234479603 0.48 0.36* (3.11) 2nd Regression equation F(i=4) 1.569449103 F(i=4) G( j=1) E(m=1) 1.647142803 0.44 F(i=4) G( j=1) E(m=1) 1.728100003 0.08 0.55* (3.86) (a) The figures in parentheses behind the variables are the optimal lag length. Other figures in parentheses are the absolute t-statistics. (b) * denotes statistically significant. Table 4 Johansen ML Procedure Null Alternative Statistics 95% Critical value r=0 r=1 27.33 22.00 rV2 r=2 13.90 15.67 rV3 r=3 6.49 9.23 Cointegration LR test Statistics: three variables VAR. I(1) Variables included: E, G, F; Lag ( P)=3. (a) r denotes the number of cointegrating vectors. (b) Statistics on the above table based on the Maximal Eigen value of the Stochastic matrix. Table 2 Granger standard causality test Controlled variable 1st manipulated variable 2nd manipulated variable FPE F-value Economic growth equation G(i=1) 6.05904904 G(i=1) E( j=2) 5.40954304 4.77 * G(i=1) E( j=2) F(m=1) 5.66098404 0.31 Energy consumption equation E(i=1) 2.56107403 E(i=1) G( j=1) 2.65269803 0.39 E(i=1) G( j=1) F(m=1) 2.74081303 0.002 (a) The figures in parentheses behind the variables are the optimal lag length. Other figures in parentheses are the absolute t-statistics. (b) * denotes statistically significant. S. Paul, R.N. Bhattacharya / Energy Economics 26 (2004) 977983 981 Table 6 Vector Error-Correction (VEC) model Equation no. Dependent variable Independent variable ( F-statistics) Error-correction term (t-statistics) DE t i DG tj DF tm Ec t 1 DE t ... 0.31 (0.74) 1.65 (0.21) 3.28* (0.00) 2 DG t 4.75* (0.02) ... 1.03 (0.37) 0.01 (0.19) (a) P-values are in the parentheses. b) * denotes statistically significant. Table 7 Johansen ML Procedure Null Alternative Statistics 95% Critical value r=0 r=1 33.02 28.14 rV2 r=2 19.16 22.00 rV3 r=3 9.39 15.67 rV4 r=4 5.23 9.24 Cointegration LR test Statistics: four variables VAR. I(1) Variables included: E, G, F, L; Lag ( P)=3. (a) r denotes the number of cointegrating vectors. (b) Statistics on the above table based on the Maximal Eigen value of the Stochastic matrix. Table 8 Johansen ML Procedure Null Alternative Statistics 95% Critical value r=0 r=1 66.79 53.12 rV2 r=2 33.78 34.91 rV3 r=3 14.62 19.96 rV4 r=4 5.23 9.24 Cointegration LR test Statistics: four variables VAR. I(1) Variables included: E, G, F, L; Lag ( P)=3. (a) r denotes the number of cointegrating vectors. (b) Statistics on the above table based on the Trace of the Stochastic matrix. Table 5 Johansen ML Procedure Null Alternative Statistics 95% Critical value r=0 r=1 47.72 34.91 rV2 r=2 20.39 19.96 rV3 r=3 6.49 9.23 Cointegration LR test Statistics: three variables VAR. I(1) Variables included: E, G, F; Lag ( P)=3. (a) r denotes the number of cointegrating vectors. (b) Statistics on the above table based on the Trace of the Stochastic matrix. S. Paul, R.N. Bhattacharya / Energy Economics 26 (2004) 977983 982 References Asafu-Adjaye, J., 2000. The relationship between energy consumption, energy prices and economic growth: time series evidence from Asian developing countries. Energy Economics 22, 615625. Bhattacharya, R.N., Paul, S., 2001. Sectoral changes in consumption and intensity of energy in India. Indian Economic Review xxxvi (2), 381392. Charemza, W.W., Deadman, D.F., 1999. New Directions in Econometric Practice. Edward Elger Publishing, UK. Cheng, B.S., 1999. Causality between energy consumption and economic growth in India: an application of cointegration and error-correction modeling. Indian Economic Review xxxiv (1), 3949. Engle, R.F., Granger, C.W.J., 1987. Cointegration and error correction: representation, estimation and testing. Econometrica 55, 251276. Granger, C.W.J., 1969. Investigating causal relations by econometric and cross-spectral methods. Econometrica 37, 424438. Hamilton, J.D., 1994. Time Series Analysis. Princeton University Press, Princeton, NJ. Johansen, S., 1991. Estimation and hypothesis testing of cointegration vector in Gaussian vector autoregressive models. Econometrica 59, 15511580. Pachuri, R.K., 1977. Energy and Economic Development in India. Praeger Publishers, New York. Tyner, W.A., 1978. Energy Resources and Economic Development in India. Boston, Martines Nijhoff Social Sciences Division, Netherlands. Table 9 Vector Error-Correction (VEC) model Equation no. Dependent variable Independent variable ( F-statistics) Error-correction term (t-statistics) DE ti DG tj DF tm DL tn Ec t 1 DE t ... 1.19 (0.32) 0.67 (0.52) 0.61 (0.55) 2.80* (0.00) 2 DG t 4.50* (0.02) ... 1.14 (0.33) 0.21 (0.81) 1.35 (0.19) (a) P-values are in the parentheses. (b) * denotes statistically significant. S. Paul, R.N. Bhattacharya / Energy Economics 26 (2004) 977983 983