Relationship between electricity consumption and economic growth of Guangdong Province in China

Abstract

Guangdong is a province with the most electricity consumption (EC) and the fastest economic growth in China. However, there has long been a contradiction between electricity supply and demand in Guangdong and this trend may exist for a long time in the foreseeable future. Therefore, the research on the relationship between EC and economic growth of Guangdong is of very important practical significance to the formulation of relevant policy. In this paper, the econometrics method of granger causality test and co-integration test is used to analyze the relationship between EC and economic growth of Guangdong from 1978 to 2010. The results indicate that there is unidirectional causality between the economic growth and the EC, and the growth of gross domestic product (GDP) and gross industrial output value (GIOV) is the impetus to promote the growth of installedβcapacity (ICAP) and the EC. Therefore, the appropriate restraint of excessive growth of power industry will not necessarily slow down economic growth. There has been a long-term stable equilibrium relationship between the EC and the economic growth. When the GDP and GIOV grows 1 unit respectively, the EC of Guangdong province will increase 0.97 and 0.64 unit respectively. The long-term marginal utility of the EC is more than 1.

Keywords

co-integration / granger causality / electricity consumption (EC) / economic growth

Introduction

Energy is an important material basis and main power source of the economic development of human society. Power is indispensable as a kind of important resource in modern social production and life, and power industry has become a pillar industry of national economic development. In recent years the relationship between electricity consumption (EC) and economic growth has become a hot spot in the energy economy research field. The main motivation of this study is to investigate the effect of the implementation of the power protection policy on national or regional economic growth. If the power consumption has causality with economic growth, then the implementation of power protection policy will have a negative effect on economic growth [1].

With Granger causality test as the core, co-integration analysis method has become the main tool to study the relationship between EC and economic growth in recent years. From the existing research results, it can be seen that different conclusions are often researched in different geographic scope or time period.

Wolde-Rufael [2] studied EC and gross domestic product (GDP) growth of 17 African countries during 1971 to 2001. The result showed that, among all the countries, the per capita GDP of six countries had unidirectional causality with the EC, and that of three countries had bilateral causality. Narayan and Prasad [1] investigated the relationship between EC and economic growth of 30 countries in the Organization for Economic Co-operation and Development (OECD). The result showed that the EC of only eight nations had causality with the GDP. Yoo and Kwak [3] made an analysis on the relationship between EC and economic growth of seven countries in South America from 1975 to 2006. It was found that the unidirectional causality between EC and GDP growth existed in Argentina, Brazil, Chile, Colombia and Ecuador, which meant that the growth of EC in these countries had a direct impact on economic growth. However, in Venezuela there existed a bilateral causality, which meant that the growth of EC directly affected economic growth and economic growth, in turn, stimulated the increase of EC. While in Peru there did not exist such a causality.

Shiu and Lam [4] studied electricity consumption and economic growth in China, whose results showed that there existed unidirectional causality between the EC and economic growth in Chinese mainland, and the economic development had little influence on the promotion of EC. But the power shortage had a huge impact on China’s economy. Zhang [5], Yuan et al. [6] and Zhu et al. [7] reached the similar conclusion. Chen et al. [8] believed that Chinese economic growth could significantly drive the growth of electric power consumption, thermal power consumption. This indicated the economic growth had unidirectional causality with EC.

Guangdong province, whose economic aggregate is the largest and always grows at the fastest speed in China, has the most installed electricity capacity and EC. The problems of environmental pollution brought about by the power industry have become increasingly prominent, and the contradiction between economic development and environmental protection is increasingly significant. Guangdong is experiencing a process of a substitution from non-electric energy to electric energy and expansion of the electric range. The economy and social development are increasingly dependent on electricity, and there is still substantial room for the growth of electricity demand for a long period of time. The study on the relationship between EC and economic growth in Guangdong has an extremely important practical guiding significance for the formulation of regional industry, economic development planning and energy saving policies, and for the exploration of a new road of small cost, effective, low emission and sustainable environmental protection.

Method & model

Unit root test

If the mean of a time series or autocovariance function changes over time, then this sequence is unstable time series. However, only stable single whole sequence of the same order manages to further the co-integration and Granger causality test. The unit root test is a way to test timing stability. Currently common unit root test is Augmented Dickey-Fuller Test.

Assuming that the sequences conform to the AR (p) process, the test equation is

(1)\nabla y_t=\gamma y_{t-\mathrm{1}}+{\xi }_{\mathrm{1}}\nabla y_{t-\mathrm{1}}+{\xi }_{\mathrm{2}}\nabla y_{t-\mathrm{2}}+\cdots +{\xi }_{p-\mathrm{1}}\nabla y_{t-p+\mathrm{1}}+{ϵ}_t.

where ϵ_t is the white noise. The test hypothesis is

Under null hypothesis that the sequence has a unit root, T statistics of significant test on the parameter γ estimates do not conform to the conventional t-distribution. Depending on the different natures of y_t, the ADF test allows the sequence to contain the constant terms, meanwhile to contain both constant and linear time trend terms in two forms.

Co-integration test

If the time series are all d-order single integration, namely I(d), there is a vector named α=(α₁,α₂,…,α_n) satisfying the equation a{y}_t^{\prime }\sim I(d,b), here, y_t = (y_1t,y_2t,…,y_nt ), d≥b≥0, it can be claimed that the sequence y_1t, y_2t , …,y_nt is (d,b) order co-integration, denoted by y_t~CI(d,b), α is the co-integration vector. Only when they are in the same order as a single integer I(d), there may exists co-integration relationship between two time series.

If sequences x_t and y_t are sequences of d-order single integration, one variable can be used to realize the regression of another variable, namely

(2)y_t=\alpha +\beta x_t+{ϵ}_t.

Use\widehat{\alpha }and\widehat{\beta }as the representative of estimates of regression coefficients, the estimated value of model residuals is

(3)\widehat{ϵ}=y_t-\widehat{\alpha }-\widehat{\beta }{x}_t.

If it meets\widehat{ϵ}~I(\mathrm{0}), sequences x_t and y_t have a co-integration relationship with each other, and (1, -β) is the co-integration vector, and Eq. (2) is co-integration regression model.

Granger causality test

The Granger causality between two sequences x_t and y_t can be defined as: under the conditions of containing the past information of variables x_t and y_t, the predictive effect of variable y_t is better than a prediction depending solely on its past information. Variable x_t helps explain the future changes of variable y_t, variable x_t is a Granger causality which caused the change of variable y_t [9].

The regression model of variables x_t and y_t is

(4)y_t={\alpha }_{\mathrm{0}}+{\displaystyle \sum _{i=\mathrm{1}}^m{\alpha }_i{y}_{t-i}}+{\displaystyle \sum _{i=\mathrm{1}}^m{\beta }_i{x}_{t-i}}+{ϵ}_t.

Check β_i (i=1,2,…,m) = 0, assuming that variable x_t is not the variable which brings about the change of variable y_t , if the check results reject the null hypothesis, i.e., to refuse "variable x_t is not the one that causes the change of variable y_t, it can be derived that the Granger causality exists between these two variables. Similarly, the regression model of x_t and y_t is

(5)x_t={\alpha }_{\mathrm{0}}+{\displaystyle \sum _{j=\mathrm{1}}^m{\alpha }_j{y}_{t-j}}+{\displaystyle \sum _{j=\mathrm{1}}^m{\beta }_j{x}_{t-j}}+{ϵ}_t.

Check β_j (j=1,2,…,m)=0, assuming that variable y_t is not the one that brings about the change of variable x_t , if the check results reject the null hypothesis, that is to refuse "variable y_t is not the one that causes the change of variable x_t, it can be derived that the Granger causality exists between these two variables.

There are four cases of Granger causality [10], ① x_t is the cause of the change of y_t, so there exists unidirectional causality from x_t to y_t. ② y_t is the cause of the change of x_t, so there exists unidirectional causality from x_t to y_t. ③ x_t and y_t have mutual causality with each other, then there is unidirectional causality from x_t to y_t, and vice versa. ④ x_t and y_t are independent, and the causality does not exist.

Data selection and processing

Taking the comparability and accessibility of data into account, select two variables, the GDP and the GIOV of Guangdong to represent economic growth. The price is the constant price of 2010, and these data come from Statistical Yearbook of Guangdong Province. Select installed capacity (ICAP) and EC to represent EC, whose data come from Chinese Electrical Power Yearbook. The period of data dates from 1978 to 2010, and the number of samples for the time series of each variable is 33.

Before conducting the co-integration analysis and Granger causality test, in order to eliminate the heteroscedasticity in time series studied, the natural logarithm of the time series of each variable is counted out first, so that the change of natural logarithm of the data will not change the original co-integration and Granger causality of the data and will make its trend linearized. The trend of time series of each variable after logarithm is shown in Fig. 1. The sequences of GDP, GIOV, ICAP and EC after logarithm can be respectively noted as lnGDP, lnGIOV, lnICAP and lnEC.

Results and discussion

Unit root test

A unit root check is taken on time series of economic and power variable to determine the stationarity of the sequence. If the original serial is non-stationary, then count out the first or second order difference of the sequence. The unit root test applies ADF test which is commonly used, and the check results are listed in Table 1.

It can be observed in the check results listed in Table 1 that the lnGDP, lnGIOV, lnICAP and lnEC are non-stationary sequences. After a second-order difference, all variables are stationary in the confidence level of 1%. D²lnGDP, D²lnGIOV, D²lnICAP and D²lnEC can be regarded as second-order single integration sequences. The Granger causality test and co-integration test can be further conducted.

Causality test

A Granger causality test is taken to analyze the causality between the economy and power variable of Guangdong, namely lnGDP and lnICAP, lnGDP and lnEC, lnGIOV and lnICAP, lnGIOV and lnEC. The check results are presented in Table 2.

It can be seen from the check results that in the confidence level of 5%, lnGDP is the Granger cause of lnICAP, and lnGIOV is the Granger cause of lnICAP. There exist unidirectional causality between GDP, GIOV and ICAP, and the large lag order are respectively 3 and 2, which illustrates the growth of GDP and industrial output value is the cause of promoting the growth of installed electricity capacity in Guangdong, and there exists a response lag in the growth of ICAP for economic growth.

In the confidence level of 5%, lnGDP is the Granger cause of lnEC, and lnGIOV is the Granger cause of lnEC, there exist unidirectional causality between GDP, GIOV and lnEC,. GDP and industrial output growth can promote the growth of the total EC of Guangdong. However, the increase in power consumption will not promote the growth of GDP and industrial output value. This is different from the conclusion made by Wang [11] that there is significant bidirectional causality between economic growth and EC in Guangdong based on the data obtained from 1985 to 2009. The lag order is 1, and the growth of the total EC responses very directly and rapidly to the economic growth.

Co-integration test

A co-integration test is taken with Engle-Granger two-step test method (E-G test method) on the variables of time series, which describes the development of economy and power, namely lnGDP and lnICAP, lnGDP and lnEC, lnGIOV and lnICAP, lnGIOV and lnEC. According to the results of the Granger causality test, economic and power variables are regarded as explanatory variables.

First, an OLS regression analysis of four sets of time series variables is made, and the co-integration equations are listed as follows:

(6)\begin{array}{l}\ln \mathrm{I}\mathrm{C}\mathrm{A}\mathrm{P}=-\mathrm{0.313823}+\mathrm{0.870026}\ln \mathrm{G}\mathrm{D}\mathrm{P}+{ϵ}_{t\mathrm{1}}\\ (0.153472)(0.017419)\end{array},

R^{\mathrm{2}}=\mathrm{0.987726},\mathrm{D}\mathrm{W}=\mathrm{0.217213},F=\mathrm{2494.625},

(7)\begin{array}{l}\ln \mathrm{E}\mathrm{C}=-\mathrm{2.031034}+\mathrm{0.968189}\ln \mathrm{G}\mathrm{D}\mathrm{P}+{ϵ}_{t\mathrm{2}}\\ (0.076054)(0.008632)\end{array},

R^{\mathrm{2}}=\mathrm{0.997542},\mathrm{D}\mathrm{W}=\mathrm{1.124538},F=\mathrm{12579.70},

(8)\begin{array}{l}\ln \mathrm{I}\mathrm{C}\mathrm{A}\mathrm{P}=\mathrm{2.414011}+\mathrm{0.572902}\ln \mathrm{G}\mathrm{I}\mathrm{O}\mathrm{V}+{ϵ}_{t\mathrm{3}}\\ (0.084059)(0.009669)\end{array},

R^{\mathrm{2}}=\mathrm{0.991248},\mathrm{D}\mathrm{W}=\mathrm{0.304805},F=\mathrm{3511.054},

(9)\begin{array}{l}\ln \mathrm{E}\mathrm{C}=\mathrm{1.015123}+\mathrm{0.636298}\ln \mathrm{G}\mathrm{I}\mathrm{O}\mathrm{V}+{ϵ}_{t\mathrm{4}}\hfill \\ (0.052669)(0.006058)\hfill \end{array},

R^{\mathrm{2}}=\mathrm{0.997198},\mathrm{D}\mathrm{W}=\mathrm{0.925168},F=\mathrm{11032.12.}

The relative estimated models of residuals are as follows:

(10){ϵ}_{t\mathrm{1}}=\ln \mathrm{I}\mathrm{C}\mathrm{A}\mathrm{P}-\mathrm{0.870026}\ln \mathrm{G}\mathrm{D}\mathrm{P}+\mathrm{0.313823},

(11){ϵ}_{t\mathrm{2}}=\ln \mathrm{E}\mathrm{C}-\mathrm{0.968189}\ln \mathrm{G}\mathrm{D}\mathrm{P}+\mathrm{2.031034},

(12){ϵ}_{t\mathrm{3}}=\ln \mathrm{I}\mathrm{C}\mathrm{A}\mathrm{P}-\mathrm{0.572902}\ln \mathrm{G}\mathrm{I}\mathrm{O}\mathrm{V}-\mathrm{2.414011},

(13){ϵ}_{t\mathrm{4}}=\ln \mathrm{E}\mathrm{C}-\mathrm{0.636298}\ln \mathrm{G}\mathrm{I}\mathrm{O}\mathrm{V}-\mathrm{1.015123.}

An ADF test on regression residuals is taken, and the check results are demonstrated in Table 3.

It is observed from the unit root check results of regression residuals that when the t values of ϵ_t1 and ϵ_t3 ​​are greater than the threshold of test under any confidence level, the sequence is not stationary. That is, there does not exist a co-integration relationship between GDP, GIOV and ICAP, and there does not exist a long-term stable equilibrium relationship between ICAP and economy. When the t values of ϵ_t1 and ϵ_t3 ​​are smaller than the threshold of test under the confidence level of 1%, the situation is opposite. Equations (7) and (9) are respectively the co-integration relationship model of GDP with EC, GIOV with EC.

It can be seen from the numerical relationship of co-integration model that, when GDP and GIOV grows 1 unit respectively, there will be an increase of 0.97 and 0.64 unit in the EC of Guangdong respectively. The long-term marginal utility of EC is more than 1.

Conclusions

From the check results of Granger causality, it can be concluded that from 1978 to 2010, the economic growth of Guangdong province has unidirectional causality with EC, and the economical development is the driving force for the growth of installed electricity capacity and EC. However, the total ICAP of electricity and EC growth do not necessarily promote the economic growth of Guangdong. Therefore, the appropriate restraint of the excessive growth of power industry will not necessarily slow down the economic growth. However, a slowdown of the economic growth rate will lead to a slowdown of the pace of growth of the power industry in Guangdong. In other words, the growth of EC and ICAP, which is brought by the sustainable and rapid economic growth of Guangdong, cannot be avoided.

It can be concluded from the results of the co-integration test that from 1978 to 2010, there doesn’t exist a co-integration relationship between the installed electricity capacity and economic growth of Guangdong. However, there is a long-term stable equilibrium relationship between EC and economic growth. When the GDP and GIOV grows 1 unit respectively, there will be an increase of 0.97 and 0.64 unit respectively of the EC in Guangdong. The long-term marginal utility of EC is more than 1, which indicates that the power efficiency of Guangdong is relatively high.

References

[1] Narayan P K, Prasad A. Electricity consumption—real GDP causality nexus: Evidence from a bootstrapped causality test for 30 OECD countries. Energy Policy, 2008, 36(2): 910-918

[2] Wolde-Rufael Y. Electricity consumption and economic growth: a time series experience for 17 African countries. Energy Policy, 2006, 34(10): 1106-1114

[3] Yoo S H, Kwak S Y. Electricity consumption and economic growth in seven South American countries. Energy Policy, 2010, 38(1): 181-188

[4] Shiu A, Lam P L. Electricity consumption and economic growth in China. Energy Policy, 2004, 32(1): 47-54

[5] Zhang J. Co-integration Analysis and Suggestions on electricity industry and economic growth. North China Electric Power University (Social Science Edition), 2004, (4): 35-37 (in Chinese)

[6] Yuan J H, Ding W, Hu Z G. Co-integration and volatility analysis of Electricity consumption and economic development. Power System Technology, 2006, 30(9): 10-14 (in Chinese)

[7] Zhu T X, Xu M S, Li Q, Li D. Study on relations between China's electricity consumption and economic growth. Price Theory and Practice, 2010, (10): 69-70 (in Chinese)

[8] Chen H L, Ma C Q, Qin T. Analysis on relationship between electricity consumption and economic growth in China. Systems Engineering, 2007, 25(8): 68-73 (in Chinese)

[9] Han Z Y, Wei Y M, Jiao J L, Fan Y, Zhang J T. Analysis on Chinese energy consumption and economic growth Co-integration and causality. Systems Engineering, 2004, 22(12): 17-21 (in Chinese)

[10] Dong H Z, Qi Z F, Shi C D. Co-integration relationship between total industrial output value and energy consumption in Shandong Province. Chinese Population Resources and Environment, 2011, 21(11): 56-60 (in Chinese)

[11] Wang R G. Study on the relationship between economic growth and electricity consumption in Guangdong Province. Prices Monthly, 2010, (7): 64-67 (in Chinese)

RIGHTS & PERMISSIONS

Higher Education Press and Springer-Verlag Berlin Heidelberg