Journal of Applied Economics and Business Studies, Volume 1, Issue 1 (2017) 1-10 https://doi.org/10.34260/jaebs.111 1 Journal of Applied Economics and Business Studies (JAEBS) Journal homepage: https://jaebs.com ISSN (Print): 2523-2614 ISSN (Online): 2663-693X Do gross domestic product changes have asymmetric effect on India’s energy use? Manzoor Ahmad1*, Zia Ullah Khan2 & Shehzad Khan3 1 PhD Scholar, Department of Industrial Economics, Nanjing University, Nanjing, China 2 School of Business Administration, Department of Industrial Economics, Southwestern University of Finance and economics, China 3 Institute of Business Studies and Leadership, Abdul Wali Khan University Mardan, Khyber Pakhtunkhwa Pakistan ABSTRACT The existing literature on the linkage between Gross Domestic Product (GDP) and energy use in both industrialized and developing economies usually assumes that the impacts of gross domestic product changes are symmetric. In this study, we utilized nonlinear autoregressive distributed lag (NARDL) model and test whether or not the effect of variations in the gross domestic product on energy use is symmetric or asymmetric from the context of India. Using time series data over 1971-2014, the findings depict that the change in the gross domestic product has a symmetric effect on energy use both in short-run and the long-run. Our conclusions infer that there is no asymmetrical association between GDP and energy use, leading to support the symmetric impact of GDP on energy use. Keywords GDP, Energy use, NARDL, India, Asymmetric JEL Classification E10; N75;O13;Q43 Copyright © 2017 SAEBR - All rights reserved 1. Introduction Along with globalization, energy becomes the most important inputs in the process of economic development. Consequently, not only the demand for energy has increased but also a dependency on energy has rapidly risen in both developed and developing countries (Ghali & El-Sakka, 2004). Although the demand for energy in third world economies is even now considerably lower than the global standards, there has been an upturn equivalent to industrial growth and levels of income. Moreover, the elasticity coefficient computed to indicate the association between economic growth and energy use both in strong and weak economies assume values approximately equal to one, which infers that a one percent upsurge in economic growth leads increase energy use by 100 percent in developing economies (Kapusuzoglu & Karan, 2012). Besides, the elasticity coefficient estimated between the demand for energy use and GDP is relatively less than one for developing economies. The difference between developed and developing economies from the 1*Corresponding author: manzoorahmad94@163.com Manzoor Ahmad, Zia Ullah Khan & Shehzad Khan 2 perspective of the association between the demand for energy (energy use) and economic growth generally stems from the continuously escalating need for energy in emerging economies (Dorsman et al., 2012). India as the world second most populous country, while consuming only six percent of the globe’s primary energy and three-quarters of energy use are met by fossil fuels. Since 2000, the demand for primary energy use has nearly increased by two times and also the potential for further fast growth is relatively high. In addition to this, the targeted policy interventions and economic growth in India have lifted more than millions out of tremendous poverty, however, per capita energy use is still close to one-third of the world average and at least 240 million masses have no access to use of electricity. In this situation, even with expanding concentration on subsidy reform and energy efficiency, there are reasons to anticipate continued rapid expansion in energy demand (IEA, 2015). Thus, India being a third-largest big economy; is growing swiftly and policies are in a position to continue with the country’s growing and modernization of its manufacturing. In this sense, the increase in India’s gross domestic product is a prime driver of energy trends (Bakirtas & Akpolat, 2018). The primary energy use of India between 2010 and 2016 are depicted in Figure 1. Data for the figure is obtained from Statista, 2018. Figure 1: Primary energy use in India between 2010 and 2016, by fuel in million metric tons of oil equivalent. In 2016, the demand for renewable energy, hydro, nuclear energy, coal, natural gas, and oil were 16.5, 29.1, 8.8, 411.9, 45.1 and 212.7 million metric tons, respectively. Contrariwise, Figure 2 shows the position of India’s energy use with respect to total top 5 emerging economies (BRICS)’ aggregate energy use. 212.7 45.1 411.9 8.8 29.1 16.5 0 100 200 300 400 500 Oil Natural Gas Coal Nuclear energy Hydro Renewable energy Consumption in million metric tons of oil-equivalent P r im a r y e n e r g y C o n su m p ti o 2016 2015 2013 2012 Journal of Applied Economics and Business Studies, Volume 1, Issue 1 (2017) 1-10 https://doi.org/10.34260/jaebs.111 3 Figure 2: Energy use share of each BRICS member nations with respect to the total energy use of BRICS Two approaches that explain the reasons behind the existence of the nexus between economic growth and energy uses are the neoclassical and the ecological approaches. The neoclassical approach primarily considers the structure of the economy as a closed system. Goods are produced by employing labor, capital and are exchanged between sellers and buyers. In this course, the higher gross domestic product is intended to be attained by raising the human capital and labor inputs. Moreover, this approach postulates that rise in the capital, labor quality and technological advancement will also contribute to attaining economic growth. The neoclassical growth model involves three typical models. The first model sort out changes in technology, the second with natural resources and the last model combine the first two models (Barro, 1998; Stern & Cleveland, 2004; Ockwell, 2008). On the other hand, the ecological approach regard energy as an underlying determinant that permitting economic production (Dorsman et al., 2012). There exists a disagreement on the association between energy use and economic growth in the existing literature. Some of this literature include (Longxing et al., 2011; Chen et al., 2012; Mulali et al., 2013; Muhammad et al., 2015; Irwan et al., 2015; Faisal et al., 2015; Bennouna & Hebil, 2016; Bah & Azam, 2017; Shahbaz et al., 2017; Muhammad et al., 2018; Mahalingam & Orman, 2018; Cai et al., 2018). It is noteworthy to mention that very few studies have concentrated on investigating the asymmetrical impact of GDP on energy use in India. Therefore, this is the first study that attempts to inspect the asymmetry arises due to change in gross domestic product and also dissect its effect on energy use from the perspective of India. The remaining paper is divided into three sections. Part two establish a nonlinear ARDL model. Part 3 elaborates the estimated findings, while part 4 ends up with conclusions. Manzoor Ahmad, Zia Ullah Khan & Shehzad Khan 4 2. Research methods Following the methodology of (Shin et al., 2013) the relationships between positive and negative components energy use and GDP are represented by the following long-run regression: EUt = ϑ0 + ϑ1 +GDPt + + ϑ2 −GDPt − + εt (1) where EU is energy use integrating of order one, GDP represents gross domestic product integrating of order one, ϑ = ϑ0, ϑ1 +, ϑ2 −) is a vector of long-run unknown parameters. It is noted that ϑ1 + represent coefficients of the positive component of GDP and ϑ2 − indicated the negative component of GDP. While GDPt = GDP0 + GDPt + + GDPt − .Where GDPt + and GDPt −are partial sum process of positive and negative variation in GDPt follow as; GDPt + = ∑ ∆GDPj + = t j=1 ∑ max (∆GDPj, 0) , GDPt − t j=1 = ∑ ∆GDPj − = t j=1 ∑ min (∆GDPj, 0) t j=1 (2) The Equation 2 is a simple modelling to inspect asymmetrical behaviour among variables included in the model. This modelling was first employed by (Schorderet, 2001) from the perspective of the nonlinear nexus between unemployment and output. Following (Shin et al., 2013), Equation 1 can be fitted in an ARDL setup under the context of (Pesaran et al., 2001) as: ∆EUt = ς0 + ς1EUt−i + ς2 +GDPt−i + + ς3 −EUt−i − + ∑ Θi p i=1 ∆EUt−i + ∑(ϖi +∆ m i=0 GDPt−i + + ϖi −∆GDPt−i − ) + ϵt (3) where 𝑝 and m are lag orders. ∑ ϖi +m i=0 estimates the short-run possible response of GDP increases on the energy use emissions while ∑ ϖi −m i=0 measures the short run impact of GDP reduction on CO2 emissions. Hence, in this setup, along with asymmetric long-term association, the asymmetric short-run impact of variations in GDP on energy use is also captured. The error correction model (ECM) of the Equation 3 is depicted as: ∆ECt = ∑ Αi p i=1 ∆EUt−i + ∑(ηi +∆GDPt−i + m i=1 + ηi −∆GDPt−i − ) + ΥiECTt−i + ρt (4) where Αi , depicts short-run coefficient and ηi + , ηi − indicate short-run adjustment symmetry. While Υiindicates the coefficient of error correction term. Journal of Applied Economics and Business Studies, Volume 1, Issue 1 (2017) 1-10 https://doi.org/10.34260/jaebs.111 5 We follow the following steps and procedures in order to estimate NARDL model. In step 1, we test each series for an order of integration with the help of Augmented Dickey Fuller (ADF) and Phillips-Perron (PP) Unit Root tests (Dickey & Fuller, 1979; Phillips & Perron , 1988; Elliott, et al., 1996). In the second step, we estimate Equation 3, using the standard OLS procedure. In step 3, the Bound testing technique is carrying out to test the existence of a long-run association between variables (Shin, et al., 2013; Pesaran, et al., 2001). This technique is based on the Wald F test having the null hypothesis of H0 (HYP1): ς1 = ς2 + = ς3 − = 0 and the null hypothesisH0: ς1 ≠ ς2 + ≠ ς3 − ≠ 0. In the fourth step, the possible existence of relationship between GDP and energy use in the long-run and short-run asymmetries in established. Furthermore, we also estimated the symmetric increasing “dynamic multiplier effects” of one percent difference in GDPt−i + and GDPt−i − respectively as: Κb + = ∑ ∂EUt_j ∂GDPt−1 + b j=0 , Κb − = ∑ ∂EUt_j ∂GDPt−1 − b j=0 , b = 1,2,3 … … .. (5) It is noted that as b → ∞, Κb + → ϑ1 + and Κb − → ϑ1 −. In the last step, the following two hypotheses are tested against alternative hypothesis: H0: (HYP2): − ς2 + ς1 ⁄ = − ς3 − ς1 ⁄ (No long Run asymmetrical relationship) H1: − ς2 + ς1 ⁄ ≠ − ς3 − ς1 ⁄ ≠ 0 (Long Run asymmetrical relationship) H0: (HYP3): − ϖi + Θi ⁄ = − ϖi − Θi ⁄ (No short Run asymmetrical relationship) H1: − ϖi + Θi ⁄ ≠ − ϖi − Θi ⁄ (Short Run asymmetrical relationship) The results of these hypotheses and empirical model are presented in the next section. 3. Results and discussion This study used secondary data on energy use in kg oil equivalent and gross domestic product in current US dollar. All the data series have been compiled from the WDI (2018), World Bank Database. The data range used in this study is 1971-2014. As an important condition of time series data, ADF and PP unit root tests are utilized to check the unit root in variables. The estimated results in Table 1 infer that both energy use and GDP are nonstationary at levels but become stationary only at first difference under the 0.001 level of significance. Moreover, the estimated outcomes of PP validated that all the designated variables can be labelled under the I(1) process. Manzoor Ahmad, Zia Ullah Khan & Shehzad Khan 6 Table 1: ADF and PP unit root test Level First difference Variables ADF PP ADF PP EU 3.749478 3.619111 -4.814941** -5.042372** GDP 0.032307 -0.006617 -5.954680** -5.973082** * depicts significance at 5% The main difference between linear ARDL and nonlinear ARDL is that the nonlinear ARDL capture asymmetries arise from the positive and negative shocks of macroeconomic variables. On the other hands, linear ARDL do not account asymmetrical relationship among variables. Thus, we start our analysis from the positive and negative components of GDP. Figure 3 depicted the positive and negative shocks of GDP and its effect on energy use. Our main objective is to compute whether or not the effect of a variation in GDP on India’s energy use is asymmetric or symmetric. Table 2 presents the result of estimated non- linear ADRL coefficients of short-run as well as the long-run. The reported result of short- run coefficients is positive for GDP increases as well as for GDP decreases. As such, the projected elasticities of GDP increase (decrease) relating to energy use is 0.033 (0.0334), infers that a 1% growth (reduction) in GDP is expected to rise (decline) energy use by 0.033% (0.0334%). It is further inferred that the estimated rise in coefficient of GDP is highly significant, nevertheless, the coefficient of GDP decreases is statistically insignificant. Keeping in view, the variations in the significance level and the directions of reported elasticities, the difference in GDP indicates towards an asymmetric effect in the short-run of Indian energy uses. 0.0 0.4 0.8 1.2 1.6 2.0 1975 1980 1985 1990 1995 2000 2005 2010 Positive shocks of GDP -.12 -.10 -.08 -.06 -.04 -.02 .00 1975 1980 1985 1990 1995 2000 2005 2010 Negative Shocks of GDP Figure 3: Positive & negative components of GDP Though, the reported result of Wald test indicates that the null hypothesis of symmetry should be accepted. In contrast, the result of long run divulge that the two estimates are positive. For example, the expansions/reductions in the elasticity of GDP relating to energy use is 0.375 (0.379), recommending that increase/decrease of one percent change in GDP is likely to increase (decrease) energy use by 0.375% (0.379%). Journal of Applied Economics and Business Studies, Volume 1, Issue 1 (2017) 1-10 https://doi.org/10.34260/jaebs.111 7 Table 2: Results of short-run and long-run NARDL equation Variable Short run coefficient Std. error t-statistic C 0.210570* 0.122684 1.716355 𝐿𝐸𝑈(−1) -0.088130* 0.051533 -1.710170 𝐺𝐷𝑃+ 0.032951*** 0.011565 2.849271 𝐺𝐷𝑃− 0.033371 0.035813 0.3572 ECT(-1) -0.088130 0.007487 -11.77051 Variable Long run coefficient Std. error t-statistic 𝐺𝐷𝑃+ 0.373890*** 0.103710 3.605138 𝐺𝐷𝑃− 0.378662 0.528090 0.717039 C 2.389308*** 0.021840 4.598309 NARDL bound test F − statistic (HYP1) = 32.16222 [LB = 2.63, UB = 3.35 at 10%] [LB = 3.1, UB = 3.87 at 5%] [LB = 3.55, UB = 4.38 at 1%] *** and * represent 1% and 10% significance level, respectively. The estimations provided in equation (3) are statistically significant, which indicates a “co-integration” relationship between the explained variables in the model. Thus, it is important to conduct a “co-integration” for the association between GDP and energy use. The most common test for this is to evaluate and compare the upper critical value to the estimated F-statistic value as suggested by the Pesaran, Shin, and Smith (2011). When we follow this procedure, the assessed F-statistic is 32.16 and the one percent and five percent upper critical value is 3.55 (4.38). Henceforth, our study strongly accepts that there is a long-run cointegration association between variables. Table 3: Testing hypothesis of asymmetrical effect Null hypothesis F-statistic Probability Long run symmetries (HYP2) 1.0156 0.2455 Short run symmetries (HYP3) 1.9761 0.4912 *** and * indicate that the null hypothesis is accepted at 1% and 10% level of significant Besides, the previous estimation results we got in Equation 3 do not suffer the spurious regression issue. Another form of validating long-run relationship is that the equilibrium takes when the error-correction coefficient should not only negative but also significant. In reality, we get the negative coefficient of ECT (-0.088), and this is also statistically significant at 1% levels. Thus, our estimated results provide an evidence of supporting co- integration. Manzoor Ahmad, Zia Ullah Khan & Shehzad Khan 8 Table 4: Results of short and long-run ARDL equations Variable Short run coefficient Std. error t-statistic C -0.145726*** 0.030242 -4.818702 𝐿𝐸𝑈(−1) -0.088478** 0.039997 -2.212100 GDP 0.032999*** 0.010578 3.119425 ECT(-1) -0.088478*** 0.007517 -11.77049 Variable Long run coefficient Std. error t-statistic GDP 0.372958*** 0.057021 6.540666 C -1.647040** 0.619384 -2.659140 NARDL bound test F − statistic = 43.98235 [LB = 3.02, UB = 3.51 at 10%] [LB = 3.62, UB = 4.16 at 5%] [LB = 4.18, UB = 4.79 at 1%] *** and ** represent 1% and 5% significance level After validating the symmetric relationship between GDP and energy use, we estimate a linear ARDL model. The estimated outcomes of linear ARDL are reported in Table 4. The findings show that there exist short-run and long-run positive relationships between GDP and energy use. Ultimately, to validate and confirm our estimated model, we carry out several diagnostic tests for the residuals of our data. Table 5 signifies the estimated values of different diagnostic tests. All the results of diagnostic tests divulge that our model is free of any statistical problem. Table 5: Validation tests Test Test-statistic Probability White test 2.891395 0.7167 LM test 0.860972 0.6502 Ramsay reset test 0.495172 0.4858 CUSUM Stable 0.05 CUSUM square Stable 0.05 4. Conclusion Changes in the Gross Domestic Product are expected to have asymmetric effects on energy use. However, the previous literature on the nexus between GDP and energy used not directly test the asymmetry hypothesis in their works of research. Thus, the current study is conducted to test whether or not the effect of changes in the gross domestic product on energy use is asymmetric from the perspective of India, currently one of the top emerging economies among developing countries and third largest end user of energy in all over the world. The findings of nonlinear ARDL validate that variations in GDP have an asymmetric effect on energy use in both short-run and the long-run. In other words, the asymmetry of changes in GDP is not observed in both short-run and long-run. Journal of Applied Economics and Business Studies, Volume 1, Issue 1 (2017) 1-10 https://doi.org/10.34260/jaebs.111 9 References Bah, M. M., & Azam, M., (2017). "Investigating the relationship between electricity consumption and economic growth: Evidence from South Africa." Renewable and Sustainable Energy Reviews, 80 (2017) 531–537 Bakirtas, T., & Akpolat, A. G. (2018). The relationship between energy use, urbanization, and economic growth in new emerging-market countries. Energy , 147, 110-121. Barro, R. J. (1998). "Government spending in a simple model of endogenous growth." NBERWorking Paper, No. 2588. Washington, DC: NBER. Bennouna, A., & Hebil, C. E. (2016). "Energy needs for Morocco 2030, as obtained from GDP-energy and GDP-energy intensity correlations." Energy Policy , 88: 45-55. Cai, Y., Sam, C. Y., & Chang, T. (2018). "Nexus between clean energy use, economic growth and CO2 emissions." Journal of Cleaner Production, 182: 1001-1011. Chen, P. Y., Chen, S. T., & Chen, C. C. (2012). "Energy use and economic growth—New evidence from meta analysis." Energy Policy , 44: 245-255. Dickey, D. A. & Fuller, W. A. (1979) Distribution of the estimators for autoregressive time series with a unit root. Journal of American Statistical Association, 74, 427- 3 Dorsman, A., Simpson, J. L., & Westerman, W. (2012). "Energy Economics and Financial Markets." Springer Science & Business Media. Elliott, G., (1996). Efficient Tests for an Autoregressive Unit Root. Econometrica, 64(4), 813-36 Faisal, Tursoy, T., & Resatog, N. G. (2015). "Energy use, electricity, and GDP causality; the case of Russia, 1990-2011. 3rd Global Conference on Business, Economics, Management And Tourism,26-28 November 2015, Rome, Italy. Procedia Economics and Finance. Ghali, K. H., & El-Sakka, M. I. (2004). "Energy use and output in Canada: A multivariate cointegration analysis." Energy Economics, 26, 225–238. Irwan S., Haseeb, M., Azam, M. & Islam, R., (2015). "Foreign direct investment, financial development, international trade and energy consumption: Panel data evidence from selected ASEAN countries." International Journal of Energy Economics and Policy 5(3), 841-850. IEA. (2015). Mexico energy outlook, world energy outlook special report. OECD/IEA. Kapusuzoglu, A., & Karan, M. B. (2012). The Drivers of Energy use in Developing Countries. In Energy Economics and Financial Markets (pp. 49-69). Springer Science & Business Media. Longxing, J., Meizhen, L., Xiuqing, L., & Zhen, Z. (2011). "Energy use per GDP in Various regions of China and its Mode. IACEED2010, (pp. 2335–2339). Mahalingam, B., & Orman, W. H. (2018). "GDP and energy use: A panel analysis of the US." Applied Economics , 213: 208-218. Muhammad, A., Qayyum, K. Bakhtayar, B. & Emirullah, C (2015). "The causal relationship between energy consumption and Economic Growth in the ASEAN-5 countries." Renewable & Sustainable Energy Reviews, 47: 732–745 Muhammad, A., Khan, A.,Q., & Ihan, O., (2018). "The effects of energy on investment, human health, environment and economic growth: Fresh empirical evidence from China." Environmental Science and Pollution Research DOI : 10.1007/s11356-019- 04497-4 Manzoor Ahmad, Zia Ullah Khan & Shehzad Khan 10 Mulali, U. A., Fereidouni, H. G., Lee, J. Y., & Sab, C. N. (2013). "Examining the bi- directional long run relationship between renewable energy use and GDP growth." Renewable and Sustainable Energy Reviews , 209-222. Ockwell, D. (2008). "Energy and economic growth: Grounding our understanding in physical reality." Energy Policy , 36, 4600–4604. Pesaran, M. H., Shin, Y., & Smith, R. J. (2001). "Bounds testing approaches to the analysis of level relationships." Journal of Applied Econometrics, 16: 289–326. Phillips, P. C. & Perron, P. (1988). "Testing for a unit root in time series regression. " Biometrika, 75, 335-46. Schorderet, Y. (2001). Discussion Paper 2001-13; Revisiting Okun’s Law:An Hysteretic Perspective. University of Califonia . Shahbaz, M., Hoang, T. H., Mahalik, M. K., & Roubaud, D. (2017). "Energy use, financial development and economic growth in India: New evidence from a nonlinear and asymmetric analysis." Energy Economics, 63: 199-212. Shin, Y., Yu, B., & Greenwoo, M. (2013). Modelling Asymmetric Cointegration and Dynamic Multipliers in a Nonlinear ARDL Framework. New York: Springer. Stern, D. I., & Cleveland, C. J. (2004). "Energy and economic growth." Rensselaer Working Papers in Economics, No. 0410). Troy, NY: Rensselaer Polytechnic Institute . World Development Indicators (WDI) (2018), the World Bank