Relationship between Electricity Energy Consumption and GDP : Evidence from India

The paper examines whether electricity energy consumption drives economic growth or vice versa in the Indian context using the annual data covering the period from 1970–1971 to 2011–2012. KPSS tests reveal that both the series, after logarithmic transformation, are non-stationary at level and stationary at first difference. Applying, two step Engle-Granger technique and Granger causality/ Block exogeneity Wald test, the study suggests that it is the electricity energy consumption that fuels economic growth both in short run and long run. It rejects the neo-classical hypothesis and empirically proves that electricity consumption is a limiting factor on economic growth. Using dynamic OLS(DOLS) method, the elasticity of electricity consumption on economic growth is estimated at 0.86 and the elasticity of economic growth on eletrcity consumption is estimated at 1.19. Based upon the elasticity, the energy requirement and energy generation is projected at 1436 BU and 1766 BU at the end of 12th Plan (at end of 2016–2017) period.


Introduction
High levels of economic growth, coupled with growing population and urbanization have resulted in a substantial increase in demand for energy.The relationship between use of energy and economic growth has been a subject of greater interest as energy is considered to be one of the important driving forces of economic growth in all economies (Pokharel, 2006).The dependence on energy by any sector of the economy justifies the link between energy consumption and the overall economic growth rate measured by the Gross Domestic Product in an economy.Therefore, the relationship between energy and economic growth has been a subject of intense research in finding the causal relationship.However, no consensus has arrived from these studies (Soytas & Sari, 2003).The results from research studies can be categorised into three main categories: (1) no causality, (2) unidirectional causality and (3) bi-directional causality between energy consumption and economic growth.Further, the causal relationship between economic growth and energy consumption is summarized into (a) short term causality (b) long term causality.The relationship between economic growth and energy consumption depends on the structure of the economy.Hence, the findings from the studies differ not only because of the structure of the economy but also across countries, but depend also on methodologies on which the studies have been made (Soytas & Sari, 2003).The relationship between these two has a major policy implication in framing the energy policy of the concerned country.Convergence on the relationship and magnitude of impact is very much important for policy formulation and implementation.
Keeping in view that electricity is the major source of energy in India and a vital input for infrastructural and socio-economic development, the main objective of this paper is to examine the relationship between electricity energy consumption and GDP in India for the spanning from 1970-1971 to 2011-2012.The present study examines both short term and long term causal relationship.In addition, one of the major objectives of this paper is to estimate the elasticity between electricity energy consumption and Gross Domestic Product (GDP) of India.The study also attempts to forecast the electricity energy consumption and generation based upon the elasticity.Based upon the findings, this study suggests appropriate energy development policies in India specifically relating to electricity sector.
The remainder of this paper is organized as follows: Section 2 deals with the review of relevant literatures.Section 3 discusses the overview of electricity sector in India.Section 4 describes the data and the methodology followed for the study.Section 5 examines the econometric relationship between electricity consumption and GDP and reports the empirical results.Conclusions and policy implication of the empirical results of the study are presented in section 6.

Review of Literature
In most of the studies relating to the relationship between energy and economic growth reveal that co-integration exist and, energy consumption granger causes economic growth not vice versa, therefore, limiting the prospectus for further large reductions in energy intensity (Stern & Cleveland, 2004).Akarca and Long (1980), Yu and Hwang (1984), Yu and Choi (1985) and Yu and Jin (1992) observed no relationship between total energy consumption and income for the United States.Whereas, Kraft and Kraft (1978), Stern (1993) and Cheng (1995) have identified a unidirectional causality running from economic growth to energy consumption in USA.Soytas and Sari (2003) investigated the nexus between energy consumption and GDP in France, West Germany, Italy, Japan and Turkey.Their findings support the growth led energy consumption excepting South Korea where the causality runs from energy consumption to GDP.The energy growth nexus is examined by Masih and Masih (1996) in a multivariate framework for economies in Asia such as India; Pakistan; Malaysia; Singapore; Indonesia; Philippines; Korea; and Taiwan.For Malaysia, Singapore and Philippines there neutral nexus is evidenced with consumption led growth for India and Pakistan, and the reverse for Indonesia.
Hence, the empirical research in the energy economic growth nexus can be grouped into Growth-led-Energy, Energy-led-Growth, Growth-led Energy-led-Growth Energy, Energy-led-Growth-led-Energy hypothesis, and the neutrality hypothesis.
The summary of relevant literatures on energy consumption and economic growth nexus is presented in the table give below.It is emerged from the survey of literatures that there is no consensus on the relationship between energy consumption and economic growth mainly because of country specific economic structures, methodology adopted and varying period of study.Country wise findings are presented.Since, the focus of the study is on India, broadly, there is a consensus on Energy-GDP nexus, wherein, energy led growth hypothesis is established.
Besides, the energy-growth nexus, estimation of elasticity of energy consumption on GDP and elasticity of GDP on energy consumption is also vital for policy formulation and implementation.Campo and Sarmiento (2011) in their analysis spanning from 1971 to 2007 on ten Latin American countries have identified the long run relationship between energy consumption and GDP.After establishing the long run relationship, they have estimated the long run elasticity of energy consumption on GDP as well as elasticity of GDP on electricity consumption for all the ten countries.The following table gives the estimated elasticity.It is important to estimate the elasticity of energy consumption on GDP when it is Energy led GDP.If it is GDP led Energy, it is vital to estimate the elasticity of GDP on energy consumption.This estimation helps in projecting the energy consumption and GDP.
The Planning Commission, Govt. of India (2014) has estimated the elasticity of electricity consumption with respect to GDP from first plan to eleventh plan.The elasticity relating to different plan period is tabulated below.

Overview of Electricity Sector in India
Energy is needed for economic growth, for improving the quality of life and for increasing opportunities for development.Some 600 million Indians do not have access to electricity and about 700 million Indians use biomass as their primary energy resource for cooking.Ensuring life line supply of clean energy to all is essential for nurturing inclusive growth, meeting the millennium development goals and raising India's human development index that compares poorly with several countries that are currently below India's level of development (Note 1).GEC has negatively skewed as compared to GEG which is positively skewed.This has resulted into positive skewness of GTD.The average T&D losses relative to total generation are at 24.1% which is quite high as compared to international benchmark of 8-9% (Note 8).High variation is observed in case of GTD.
Both at national and international level, Per-capita Energy Consumption (PEC) PEC and Energy intensity (EI) are the most used policy indicators, both at national and international levels.High energy efficiency indicated by low EI usually refers to less use of energy per unit of output.Gain in energy efficiency directly increases energy uses by other economic activities which further stimulates economic growth.Gain in energy efficiency means may lead reduction in price of certain consumer products which in turn, spurs an increase in the demand for energy indirectly through released purchasing power redirected to energy-using goods and services.
PEC is the total energy consumption during the year relative to the estimated mid-year population of that year.Energy Intensity is defined as energy consumed for producing one unit of Gross Domestic Product (At constant prices).In the absence of data on consumption of non-conventional energy from various sources, particularly in rural areas in the developing countries, including India, these two indicators are generally computed on the basis of consumption of conventional energy (Note 9).The PEC has increased from 1204 unit (KWH) in 1970-1971 to 4816 unit in 2010-2011, a CAGR of 3.44%.The annual increase in PEC from 2009-2010 to 2010-2011 was 3.65%.The PEC of India is one-third of the international average (Note 10) indicating potentially higher energy demand in the long term as the country continues its path of economic development. The

Data a
The relev details of

Testing for Stationary Nature of Data
To examine the relationship between electricity consumption and GDP of India, it is first established whether these time series data are stationary or not.This is done by performing a unit root test on time series data wherein, the unit root test identifies variables that are non-stationary, meaning that they contain stochastic trend that leads them to wander randomly.The presence of unit root is examined using the Kwiatkowski-Phillips-Schmidt-Shin (KPSS) unit root tests (KPSS 1992).To test whether a series, y(t), has a unit root or not, the following model is taken into account.
Both drift and time trend are captured in modeling the series, as time series are observed to have both these components.The n lag terms have been taken to protect against the possibility that y(t) follows a higher order autoregressive process.
The null hypothesis for the KPSS test is that the data series is stationary and, hence, the time series y(t) does not have unit root.This is tested against the alternate hypothesis that that unit root exists and the series y(t) is nonstationary.KPSS test is considered as more powerful tool to test the unit root when there are multiple break points in time series data (Note 13).Since, electricity consumption and GDP time series data have multiple break points( Annex -1), KPSS test is adopted.The test statistic is called LM statistic which based upon asymptotic distribution If the LM stat is less than the asymptotic critical values tabulated by KPSS, then null hypothesis of stationary nature of data is accepted.
The natural logarithmic value of electricity consumption and real GDP is taken to check the unit root.The result of unit root test is given in table 6.Hence, both electricity consumption and GDP are non-stationary at level.They are stationary at first difference level at 1% level or difference stationary.

Co-integration Test
After establishing the non-stationary nature of LCE and LGDP at level and stationary nature at first difference, the existence of any long-term equilibrium relationship between these two time series variables is examined.For examining this, the concept of co-integration is applied.Co-integration implies an equilibrium relationship which is a pre-requisite for testing and estimating long run (equilibrium) relationship among selected variables.The co-integration methodology is the two-step process suggested by Engle and Granger (1987).
Two series, z(t) ~ I(1) and x(t) ~ I(1) are said to be co-integrated if there exists a β such that z(t) -βx(t) is I(0) (Maddala 2001).This leads to the following regression equation: Where, u(t) is I(0) variable, z(t) and x(t) do not drift too far apart from each other over time.If z(t) and x(t) are not co-integrated, then u(t) will be I(1), which means that x(t) and y(t) can drift apart more and more over time.
In this case, the relationship obtained by regression z(t) over x(t) is not valid and is of the nature of "spurious regression".
Here, z(t) is the logarithm value of electricity consumption and x(t) the is the logarithm value of GDP.The co-integration regression is carried out on these two variables by the ordinary least squares (OLS) method.
Where z'(t) is the estimated z(t).
Next, the co-integrating residual, u(t) is derived as: Unit root test is applied on these residuals u(t) for unit root test using the KPSS test.A slight modification is made to the normal test (equation 1) as it is based on calculated least square residuals.The model used is: As compared to (1), the drift and the trend part have been removed in (5).In KPSS test,the null hypothesis H 0 is implies that u(t) are stationary.Hence, there is long run relationship between x(t) and z(t),hence they are co-integrated.The alternate hypothesis H 1 is x(t) and z(t) are not co-integrated.Using LM statistic, the residuals from above regression are examined for I(0) or stationary.The results are given in table 7. LGDP 42 0.2150 0.7390 H0 is accepted, Co-integration exists.
From the results given in table 7, it can be inferred that the electricity consumption and GDP are non-stationary at level.However, the co-integration between these two non-stationary series is established implying that that both the have a tendency to converge systematically in the long-run, even if they may drift apart in the short-run.

Error Correction Mechanism
After establishing co-integration and long run equilibrium relationship, the short run equilibrium relationship is examined by using the Granger representation theorem (Gujarati, 1995).The short-term relationship between the two variables is presented in the form of an Error Correction Model (ECM).
In equation ( 6), Δ denotes the first difference operator, ε t is a random error term, and u t-1 = LGDP t-1 -δ 1δ 2 LEC t-1 , that is, the one period lagged value of error from the co -integrating regression Similarly in equation ( 7) Δ denotes the first difference operator, ε' t is a random error term, and u' t-1 = LEC t-1 -δ' 1 -δ' 2 LGDP t-1 , that is, the one period lagged value of error from the co integrating regression.
According to the Granger representation theorem(GRT), negative and statistically significant α 2 and β 2 is a necessary condition for the variables in hand to be co-integrated.In practice, this is regarded as an convincing evidence and confirmation for the existence of cointegration found in the first step.It is also important to note that, in the second step of the EGM, there is no danger of estimating a spurious regression because of the nature stationary of the variables ensured.Combinations of these two steps then provide a model incorporating both the static long-run and the dynamic short-run components.
The short run dynamics of the equation ( 6) & ( 7) are examined through multiple regressions on model given in (8 & 9).The results are given below: In equation ( 8), the negative value of δ 2 shows that 7.95% of the discrepancy between the two variables is eliminated in the next year.The't ratio' of δ 2 is also significant at -2.158 (p value 0.0373) indicating that the impact of electricity consumption on GDP is stable in the long run.As equation ( 9) shows, the negative value of δ' 2 shows that 4.8% of the discrepancy between the two variables is eliminated in the next year.However, the't ratio' is very low at -0.6754 (p value 0.5035) and, not insignificant, indicating that the impact of GDP on electricity consumption is not stable in the long run.From the above results, it is empirically established that electricity consumption has a stable long run impact on the GDP.In other words, the long run equilibrium causal relationship runs from electricity consumption to GDP and not vice versa.

Short Term Causality
After establishing, the long run association between electricity consumption, the short run association is examined by applying Granger causality/ Block exogeneity Wald test (Enders, 2003, p. 284).Granger causality indicates that lagged values of a variable provide statistically significant information to predict another variable.Essentially, Granger causality tests the presence of correlation between the current value of one variable and the lagged values of other variables in the system.In addition, Granger causality tests decide about the exogeneity of a variable.This test detects whether the lags of block variables can Granger-cause any other variables in the VAR system.For example, rejection of the null hypothesis implies that if all lags of electricity consumption cannot be excluded in explaining GDP, then GDP is an endogenous variable and there is causality of electricity consumption on GDP.Therefore, in order to determine which variables are exogenous in the VAR model, the Granger causality/block exogeneity Wald tests are undertaken.The lag order of 1 is selected based upon schwarz information criterion (SIC).The output of the test is given below.The null hypothesis of GDP does not cause electricity consumption is accepted at 95% level of significance with low chi square value.The null hypothesis of electricity consumption does not cause GDP is rejected at 95% level of significance and high chi square value.The results show that in the short run, causality runs from electricity consumption to GDP but not vice versa.This corroborates the findings of the equation (8).

Estimation of Long Run Elasticity
The elasticity can be estimated by transforming both the variables into logarithmic form (double log model).
LGDP = a 0 + a 1 LEC + u t (10) Where: 'a' is the degree of responsiveness of real GDP for one percentage change in electricity consumption.
'b 1 ' is the degree of responsiveness of electricity consumption for one percentage change in real GDP.
The DOLS technique is applied for calculating the long-run elasticity.The Dynamic OLS procedure introduced by Stock and Watson (1993) involves estimation of long-run equilibrium via dynamic OLS (DOLS).DOLS involves in regressing one of the I(1) variables on other I(1) variable by augmenting the co-integrating equation with lags and lead of these first difference of the regressor.The essence of incorporating the first difference variables and the associated lags and leads is to make the resulting co-integrating equation error term is orthogonal and to correct for regressor endogeneity.In addition it has the same asymptotic optimality properties as the Johansen distribution.HAC (Newey-West) covariance matrix estimator is adopted in executing DOLS.
The regression output is given in the table 8 & 9. From the above output in Table 9, it can be interpreted that about 97% of variation in growth in growth in real GDP is explained by variations in growth in electricity consumption.The LEC coefficient is statistically significant and an increase of 10% in electricity consumption is likely to increase the real GDP by 8.6%.Since, India is poised to grow at a rate of 8% in real terms; the electricity consumption should grow by 10% annually.At 99% confidence level, the growth in electricity consumption is observed to be a statistically significant impact growth in GDP in India.Zero p-value corroborates this observation.In Table 10, 96% of variation in growth in electricity consumption is explained by variations in growth in real GDP.The LGDP coefficient is statistically significant and an annual growth of real GDP by 8% will lead to annual increase of electricity consumtion by 9.5%.At 99% confidence level, the LGDP is observed to be a statistically significant with zero p-value corroborates this observation.
The elasticity of 1.19 is in tune with the average elasticity from 1969-2012 (Fourth Plan to Eleventh Plan Period) is calculated at 1.27 calculated by the Planning Commission.
However, the estimate of elasticity of GDP w.r.t electricity consumption by the Ministry of Power at 0.90 for the period 2012-2017 is much lower than long run elasticity of 1.19.

Conclusions and Policy Implications
It is to be noted that the previous studies tried to relate the aggregate energy consumption with economic growth in India but there may be a practical difficulty in aggregating the various forms of real energy consumption as their units of measurement differ.The conversion depends upon the quality or productivity of energy.Therefore, the present study makes a departure from the earlier studies by trying to relate only electricity as energy consumption with economic growth.This will help to have different policy strategies in devising the demand for electricity.The previous studies have either taken aggregate energy consumption or if there is a disaggregation, they have considered some forms of energy and leaving the most important component of energy i.e. electricity.Probably this is the reason why the studies have employed the traditional co-integration technique.
This paper has examined the existence and direction of causality between electricity consumption and economic growth in India using the annual data covering the period 1950-1951 to 1996-1997.The two step procedures of Engel Granger approach and VAR Granger Causality/Block Exogeneity Wald tests are applied to establish both long term and shot term causality.Empirical results have established the existence of long run as well as short run causality running from electricity consumption to economic growth without any feedback effect.Thus, a growth in electricity consumption is responsible for a higher economic growth.The findings of this empirical study is in consensus with the earlier findings (Table 1) excepting one (Ghosh, 2002) in the context of India.This result can be interpreted as follows.
The results of this study reject the neo classical theory of neutrality of energy consumption.Since, the causality runs from electricity consumption to GDP in India, electricity consumption is a limiting factor on GDP growth.
The policy of ''energy must lead economic growth'' should be emphasized for a long period which is contextually more important as a subdued growth rate of 4.5% and 4.9% has been witnessed during 2012-2013and 2013-2014 respectively (Note 14) respectively (Note 14).
The findings of this study are relevant to policymakers.Since, the Indian Economy is energy dependent, and as consequence, a conservation policy may counterproductive in slowing down the economy with adverse socio-economic effects.
The expansion of industrial and commercial sectors where electricity has been used as basic energy input because of its clean and efficient nature stimulates economic growth.The share of GDP by industrial sector and services sector are the highest and, they consume maximum electricity as compared to others category of consumers.Electricity consumption in agricultural and transport sector has also accelerated to keep pace with country's economic growth.The household sectors use the electricity at the cheapest form of energy.This helps in to add to their financial savings.Since, the household sectors are the major contributor of the total saving of the Indian economy; these savings are used to finance capital formation which leads to higher economic growth.
Since 1970-1971, the growth rate in electricity consumption in household sector, commercial sector, Agricultural sector and Railways is peaking up.It can be safely deduced that their involvement in economic activities has been growing up.The component wise of National Account Statistics corroborate this.
The result has important policy implications.India being, the fourth largest consumer in the world, energy deficit is persistent since 1980-1981.The energy efficiency of India is very low.Low efficiency and high T&D loss are the limiting factors on the economic growth of India as energy consumption causes economic growth.Since, the economic growth needs more energy and the economic growth is contributed by construction, steel, metallurgy, equipment electro-analysis aluminum, glass and infra sectors which are which are high consumers of energy.This is concerned with the socio-economic development of the economy.The most important way to have efficiency gain is to reduce high T&D loss in the electricity distribution sector.
Because of electricity consumption led growth is established, conservation of electricity will inhibits the economic growth.So, there is little scope for energy conservation policy.Besides, the long run elasticity of electricity consumption w.r.t real GDP is inelastic (0.86), more electricity consumption is required to induce higher growth in GDP.This is also corroborated by the mean buoyancy of GDP relative to electricity consumption which is estimated at 3.
Nevertheless, if the high T&D loss in the distribution sector can be reduced to 15% level as envisaged in Accelerated Power Development Restructuring Program (APDRP, 2002), then energy conservation is possible along with efficiency gain without affecting the end consumers.Facilitation of cleaner and renewable forms of higher quality from hydro and thermal based electricity would help in efficiency gain.
The outlook for real GDP growth in India for the remaining year of 12th Plan period (2012-2017) is 8%.With elasticity of 1.19, the requirement of energy will grow at 9.52%.The Energy requirement was 998.11BU in 2012-2013.At the end of 2016-2017, the energy requirement will be 1436 BU.This compares fairly well with the projection of energy requirement of 1403 BU at the end of 2016-2017 by Ministry of Power, Govt. of India.
If the average T&D losses in the remaining four years are pegged at 23%, the projected net generation of electricity will be at 1766 BU at the end of 12th Plan Period.
In order to ensure sustainable economic growth, a sufficient amount of energy supply must be ensures.The formulate and implement energy policy that will take care of energy security, prevent excessive energy consumption and improve energy efficiency, reducing the energy intensity and also to encourage to create new energy sources are the challenging task for policy makers.
From sustainability point of view in the long run, the growth and development in India, policy intervention is required to change its economic structure towards a more efficiency-oriented and less resource-depleting one and to rely more on renewable energy sources.Renewable energy technologies have an enormous potential to solve energy problems in India.

Notes
Figure 3 Figure

Table 1 .
Summary of findings from selected literatures

Table 3 .
Elasticity of electricity consumption w.r.t.GDP

Table 4 .
1980contrito 2011-2012.In 1980or, Central Sector and Private Sector are 39.37%, 28.73% and 31.88%respectively to the total installed capacity in India (Note 3).The following Figure shows the share of sources of energy in total installed capacity at the end of 2013-2014.The thermal and hydroelectricity constitutes 66% of the installed capacity.Descriptive statistics As shown in Figure2, the co-movement between growth rate in electricity consumption and GDP exists from 1980-1981 onwards.The descriptive statistics of annual growth rate in electricity Consumption (GEC), GDP (GDP), electricity generation (GEG) and T&D losses (GTD) is presented below.Some observations are emerging from the table 4. The minimum and maximum of GEC is more than GEG which implies the energy deficit.In fact, energy deficit has been witnessed in India since 1980-1981.The energy deficit relative to energy requirement is recorded at 8% annually from1980 -1981to 2011-2012.In 1980 -1981, the energy deficit was at 16,384 MU and increased by at 5.8% on annual CAGR basis to reach at a deficit level of 86905 MU at the end of 2011-2012.High levels of economic growth, coupled with growing population and urbanization have resulted in a substantial increase in demand for power.However, power supply has been lagging behind; in 2011-2012, the country had a power deficit of nearly 8.7% per cent of the total requirement.
www.ccse ntral Electrical Au ed energy bas energy is prod owth of the In r in the world a result of the major source 010-2011.Co generation (N AGR) basis (N y grew at a ra on.Since, elec al generation, tion.In fact, T at current pric tion of electri Fig MIE, *GEC & GG In Fig uthority, India.sket of India duced from co ndian economy d after China, e country's dy of energy con al and Lignite Note 6) of the Note 7) from ate of 6.62% o ce acity tinue in futur ources.wer.India is th need for ener tion over the p about 51% of eum (20%) w on annual com e same perio e lower than th mission & Dis ason for relat me time perio e growth rate and nominal GDP.Vol. 7, N re.Electricity he fourth-larg rgy supply co past several ye f the total con was third (Note mpound avera od, the consum he CAGR of stribution (T& tively lesser od.However, t in both gene

Table 5 .
D As a part of second step of EG procedures, Error Correction Mechanism (ECM) is applied to test the long run equilibrium relationship.Granger causality/ Block exogeneity Wald test is applied to investigate the short run causality.Dynamic OLS method is used to estimate the elasticity.

Table 6 .
Testing presence of unit roots using KPSS test *LEC and LGDP are logarithmic value of Electricity Consumption and GDP respectively.

Table 7 .
Results of KPSS test for cointegration

Table 9 .
Elasticity of electricity consumption on Real GDP

Table 10 .
Elasticity of real GDP on electricity consumption Note 1. Ministry of Power, GOI.Note 2. Ministry of Power: http://powermin.nic.in/indian_electricity_scenario/introduction.htmNote 3. Ministry of Power, GOI (2013).Note 4. US Energy Information Administration.http://www.eia.gov/countries/analysisbriefs/India/india.pdfNote 5. Energy Statistics, 2012.MOSPI, GOI.Note 6.Total generation is sum of net generation, energy received from captive plan and energy imported.The net generation is gross generation net of auxiliary consumtion.Note 7. CAGR is calculated taking the OLS regression model: log(y) = c + r*t, where r is the CAGR and y is the variable for which r is calculated.