Electrical Power and Energy Systems 53 (2013) 338–347

Contents lists available at SciVerse ScienceDirect

Electrical Power and Energy Systems

journal homepage: www.elsevier .com/locate / i jepes

Electricity consumption and human development level: A comparativeanalysis based on panel data for 50 countries q

0142-0615/$ - see front matter � 2013 Published by Elsevier Ltd.http://dx.doi.org/10.1016/j.ijepes.2013.05.024

q This is an open-access article distributed under the terms of the CreativeCommons Attribution-NonCommercial-No Derivative Works License, which per-mits non-commercial use, distribution, and reproduction in any medium, providedthe original author and source are credited.⇑ Corresponding author at: Cooperative Innovation Center for Arid Environment

and Climate Change, Lanzhou University, Lanzhou 730000, China. Tel./fax: +86 9318914027.

E-mail address: shuwenn@lzu.edu.cn (S. Niu).

Shuwen Niu a,b,⇑, Yanqin Jia b,c, Wendie Wang b, Renfei He b, Lili Hu b, Yan Liu b

a Cooperative Innovation Center for Arid Environment and Climate Change, Lanzhou University, Lanzhou 730000, Chinab College of Earth and Environmental Sciences, Lanzhou University, Lanzhou 730000, Chinac Lanzhou University of Finance and Economics, Lanzhou 730020, China

a r t i c l e i n f o

Article history:Received 20 January 2013Received in revised form 6 May 2013Accepted 10 May 2013

Keywords:Electricity consumptionHuman development levelPanel data model

a b s t r a c t

As a representative of modern energy, the level of electricity consumption can be regarded as an appraisalcriterion of a country’s development level. This study analyses the causality between electricity con-sumption and human development and assesses the changing trend of electricity consumption. The mod-els in this study are established using panel data from 1990–2009 for 50 countries divided into fourgroups according to income. For human development indicators, per-capita GDP, consumption expendi-ture, urbanisation rate, life expectancy at birth and the adult literacy rate were selected. The results showthat long-run bidirectional causality exists between electricity consumption and five indicators. Addi-tionally, the higher the income of a country, the greater is its electricity consumption and the higher isits level of human development. Further, the variables of four income-groupings vary considerably. Spe-cifically, as income increases, the contribution of electricity consumption to GDP and consumptionexpenditure increases, but the urbanisation rate, life expectancy at birth and adult literacy rate presenta weakening trend. This mainly because that the latter indicators in high-income countries are increasingto converge. To improve human development, electricity should be incorporated into the basic publicservices construction to enhance the availability of electricity for low-income residents.

� 2013 Published by Elsevier Ltd.

1. Introduction

As a manmade source of energy [1], electricity can be generatedfrom primary energy [2] and/or convertible into an ultimate formof energy with the help of various technologies [1]. Beyond itsavailability and flexibility, electricity has other advantages. Forexample, it can be supplied continuously and transmitted overlong distances by power lines or grids. In addition, it is clean (theutilisation process does not emit greenhouse gases), convenient,renewable and efficient. Therefore, electricity is thought to be theenergy that has the widest application and plays an important rolein the economic and social development of all countries [3,4]. Leninfamously asserted in 1920 that communism was the result of So-viet regime and the national electrification [5]. Furthermore, alarge number of studies have indicated strong causality betweenelectricity consumption and economic growth [4,6–8].

An electric power service has been the basic requirement forimproving living standards and supporting social developmentfor five main reasons. First, when obtaining electricity, food, vac-cines and drugs can be stored in a refrigerator for a longer time,thereby improving people’s health conditions [9–11]. Second,lighting makes people study longer with the result that the adultliteracy rate increases. In addition, the application of computers,televisions and the Internet improves people’s ability to obtaininformation and knowledge [12]. It can be said that modern societyis highly reliant on network information and communication tech-nologies [6].

Third, the available electricity makes widespread use of varioushousehold appliances. It is more convenient for people in heating,cooling, sanitation, entertainment and so on, which greatly im-proves quality of life [13,14]. Fourth, electricity can take the placeof traditional biomass energy and coal, which is able to reduce in-door pollution and improve the quality of the environment. Mean-while, it can be produced without carbon emissions. Whenrenewable energy (e.g. wind, solar, nuclear and hydro energy) orfossil fuels that can capture and store carbon are used to generateelectricity, it can efficiently reduce carbon emissions and mitigateclimate change [15].

Finally, the utilisation of electricity not only releases peoplefrom hard work, but also saves much time for people. In particular,

S. Niu et al. / Electrical Power and Energy Systems 53 (2013) 338–347 339

for rural women it provides opportunities for self-employment andpotential development [10,11]. Therefore, electricity consumptionis regarded as the reference of well-being [16] and a key measur-able index of life quality [2]. Statistics by the International EnergyAgency (IEA) [17] showed that 1.4 billion people around the worldhave no access to electricity and that 2.7 billion people still lived bytraditional fuel. The common shortage of electricity supply has be-come an obstacle to social development. For instance, in Africa andSouth Asia there is no available electricity for most rural residents[18].

Electricity consumption is an indicator that reflects the level ofsocial development in a country. In the process of modernisation, itis necessary to produce adequate amounts of electric energy to ad-vance sustainable development. As is well known, as a kind of sec-ondary energy, electric energy is transformed from primary energysuch as hydro energy, nuclear energy and coal [2]. The process ofgenerating electricity, from resources development to end use,needs a series of links, such as production and transmission. Inaddition, supporting engineering facilities and a service systemare required to guarantee electricity supplied sustainably andsafely, especially to construct a perfect grid system [1]. In turn, thisrequires corresponding economic strength and techniques for sup-port. Therefore, an increase in electricity consumption cannot beachieved overnight.

Electricity is clean in terms of consumption, but generates someenvironmental problems in terms of production [19]. In particular,when coal is used to generate electricity, it emits large quantities ofgreenhouse gases. To promote socioeconomic development, it isnecessary to increase electricity consumption. However, an in-crease in electricity consumption would have an adverse impacton the environment. Faced with this dilemma, we need to quickenthe development and utilisation of a clean energy source, such ashydroelectric, nuclear, wind, solar and biomass power [20]. How-ever, developing new energy leads to an increase in technologycosts, rising electricity prices and higher consumer spending [21].In particular, it places a larger financial burden on medium- andlow-income earners.

This paper studies 50 of the main countries around the world,which are divided into four groups (low-income countries, lower-middle income countries, upper-middle income countries andhigh-income countries) according to the latest income-groupingstandards of the World Bank (2010). In 2009, the total populationof the study region was 5.478 billion, the GDP was 32.983199 tril-lion dollars, the area of land was 89.0008 million square kilometresand electricity consumption was 16.058834 trillion kilowatt-hours, accounting for 81.00%, 83.13%, 68.61% and 86.91% of theworld’s totals, respectively. In this study, we analyse the inherentrelation between electricity consumption and human developmentindexes (HDIs) using the methods of the panel Granger causalitytest to the parameter estimation of panel models. The results notonly show that electricity consumption is a big promotion to eco-nomic and social development but also show the changing featuresand regularity of electricity consumption on the time series anddifferent individual sections. Finally, this paper provides sugges-tions on improving public service and promoting electrificationfor undeveloped countries and areas.

For China, electrification reached upper-middle income levels in2010. Nevertheless, regional development is unbalanced and theincome gap between urban and rural residents continues to ex-pand. Specifically, the Beijing-Tianjin region, Yangtze River Deltaand Pearl River Delta have already achieved a high-income level,while other large cities and eastern rural areas belong to theupper-middle income level. The income levels of small- andmedium-sized cities and villages in central China is medium orcomparatively low, whereas the agricultural and pastoral areas inthe western region are still at a low-income level. Using the above

research method of different countries to study the correlation be-tween electricity consumption and human development in differ-ent areas of China can provide reference to launch aid projectsand reduce the development gap between regions. Therefore, ithas great significance in putting forward policy recommendationson how to improve electricity consumption in lower-middle in-come families based on the historical background of building awell-off society.

This paper is structured as follows: We review relevant litera-tures regarding electricity consumption and human developmentin Section 2. Section 3 describes the data source, research ideasand methods. Section 4 analyzes the results calculated by models.Section 5 gives the conclusions and policy proposals.

2. Literature review

Since the correlation between electricity consumption andhuman development is not as director as close as that betweenelectricity consumption and economic growth, the relevant re-search is rarer. Kanagawa and Nakata [22] found that electricityconsumption has significant correlation with GDP as well as HDIfor 120 countries, and the countries which mark high consumptionlevel of per capita electricity, attain upper rank of both economicactivities. Ghali and El-Sakka [23] also noted that per-capita energyand electricity consumption are highly correlated with economicdevelopment and other indicators of modern lifestyle, with theinference that the more energy that is consumed, especially inthe form of electricity, the better life is. Mazur [3] demonstratedthat electricity consumption was essential for people to improvetheir well-being in less-developed countries, especially in popu-lous China and India. However, in industrialised nations, increasingelectricity consumption has little relationship with improving lifequality. Zahnd and Kimber [24] demonstrated that electrical en-ergy can provide for appropriate and sustainable lighting, whichbrings potential health, education, social and economic benefitsto the people who have previously lived in homes with excessiveindoor air pollution. Wu et al. [25] evaluated the inequality of en-ergy consumption using the Lorentz Curve, Gini Coefficient andTheil Index, which reflect differences in the economic developmentlevels of countries that are categorised into high, middle and lowgroups. In Brazil, rural electrification has become an important fac-tor to reduce energy poverty, although it is not the only one [18].The case analysis of Assam state in India suggested that rural elec-trification helped promote social and economic development andachieve the goal of poverty relief [22]. Holtedahl and Joutz [26]gave two possible reasons why higher urbanization might leadto higher energy use. Firstly, urbanization implies greater accessto electricity, since households can be more readily connected tothe grid. Secondly, households who already had access to electric-ity in rural areas are likely to increase their consumption in urbanareas because of increased use of existing appliances and purchaseof new ones. Burney [27] found that higher growth in incomeaccelerates increases in electricity consumption; Increase in elec-tricity consumption is compounded by socioeconomic develop-ment, as reflected by increases in literacy, share of industry andurbanization (see Table 1).

3. Data and methods

3.1. Selection of research subjects

This paper studies 50 of the main countries in the world, includ-ing 15 developed and 35 developing countries. These are distrib-uted across different continents (except Antarctica). The maineconomies are selected and 19 countries of the G20 are included

Table 1Summary of literature review.

Author(s) Countries Main variables Indicators and conclusion(s)

Mazur [3] 21Industrializednations

13 Variables ofwellbeing

Among industrial nations, increases in per capita energy and electricity consumption are notassociated with corresponding improvements in quality of life

Pereira [18] Brazil Energy consumptionHDI, energy poverty

There is stagnation in the HDI with per capita energy consumption below 22 GJ/year

Kanagawa andNakata [22]

120 countriesand India

Energy consumptioncapital, labor

Electricity consumption has significant correlation with GDP as well as HDI for 120 countries.Complete household electrification will be achieved by the year 2012, the literacy rate in Assam mayincrease to 74.4% from 63.3%

Ghali and El-Sakka [23]

Canada Energy consumptioncapital, labor

Per-capita energy and electricity consumption are highly correlated with economic development andother indicators of modern lifestyle

Zahnd [24] Nepal Health; education; socialand economic

Electrical energy can provide for appropriate and sustainable lighting, which brings potential health,education, social and economic benefits to the people

Wu [25] 129 Countries Theil index Ginicoefficient

The international inequality of per capita energy capita consumption is close to the Gini coefficient byHDI equity criterion

Holtedahl andJoutz [26]

Taiwan Urbanization Higher urbanization might lead to higher electrical energy use

Burney [27] 93 Countries Income; literacy;urbanization

Increase in electricity consumption is compounded by socioeconomic development, as reflected byincreases in literacy, share of industry and urbanization

340 S. Niu et al. / Electrical Power and Energy Systems 53 (2013) 338–347

(excluding Saudi Arabia). We divide the 50 countries into fourgroups (low-income countries, lower-middle income countries,upper-middle income countries and high-income countries)according to the latest income-grouping standards of the WorldBank (2010). All selected countries and the four groupings areshown in Table 2.

3.2. Variable selection and data sources

Many indicators reflect the level of human development.Researching human development is thus not as compact as study-ing economic development level, which can be represented by theindex of per-capita income. The United Nations Research Institutefor Social Development put forward nine economic indicators andnine social indicators in 1972 [28]. A report issued by the Organi-zation for Economic Co-operation and Development (OECD) [29]came up with six social variables as the indicators to forecastper-capita GNP. Further, Morris [30] the United States Agency forInternational Development proposed the Physical Quality of LifeIndex in 1979. In addition, the United Nations Development Pro-gram first put forward the concept of the HDI in the Human Devel-opment Report in 1990. It selected three basic indicators, namelyper-capita GDP, life expectancy at birth and the adult literacy rate,to measure the average human development levels of United Na-tions members [31] Since then, the HDI had been widely applied,but at the same time, it has caused hot discussion and has requiredsome modifications [32–36]. In general, the more indexes thereare, the larger the amount of information there is and thus themore inconvenient they are.

This paper selects five indexes to reflect the human develop-ment level of the studied countries: per-capita GDP (constant dol-lars in 2000, GDP), per-capita consumption expenditure (constantdollars in 2000, CON), urbanisation rate (%, URR), life expectancyat birth (year of age, ELB) and the adult literacy rate (%, LIR). Spe-cifically, per-capita GDP reflects economic development level,per-capita consumption expenditure is a sensitive barometer ofpeople’s quality of life, urbanisation rate is able to reflect socialstructure and life expectancy at birth and the adult literacy rate re-flect population quality. In addition, the explanatory variable isper-capita electricity consumption (KW h, ELC). The data on theadult literacy rate is from the International Statistical Yearbookand other indicators come from the World Bank Online Database.The time series of the data is from 1990 to 2009. In order to elim-inate the possible impact of heteroskedasticity, the data are con-

verted into the logarithmic form and respectively termed LGDP,LCON, LURR, LELB, LLIR and LELC. Our research ideas can be gener-alized by Fig. 1.

3.3. Econometric methodology

The panel data model is an econometric model that takesadvantage of parallel numbers to analyse the correlation amongvariables and to predict variables’ changing trends. It can reflectchanges in laws and the individual characteristics of research sub-jects from three dimensions (cross-section, period and variable)and make full use of the information contained in the sample. Inthis study, to research the causality between electricity consump-tion and human development level, the panel unit root test, panelcointegration test and panel causality test were first constructed.Then, we established the panel data model to evaluate the changeregularity of variables on individual sections and time series.

3.3.1. Method of testsPanel unit root test. The panel data unit root test has two main

methods. One is based on the same root, including the LLC testmethod proposed by Levin et al. [37], Breitung test demonstratedby Breitung [38] and Hardi test put forward by Hardi [39]. Theother is the unit root test for different unit roots, including theIPS test proposed by Im et al. [40] as well as the ADF and PP testsput forward by Maddala and Wu [41]. To overcome the deviationcreated by single methods, this paper adopts the tests of LLC,ADF and PP.

Panel cointegration test. There are three major methods of coin-tegration test: the Pedroni test proposed by Pedroni [42], Kao testdemonstrated by Kao [43] and Johansen panel cointegration testput forward by Maddala and Wu [41]. In this study, we selectedthe Kao test because of the comparatively more research variablesand comparatively longer time series as well considering the spacelimit.

Panel Granger causality test. The error correction model basedon panel data put forward by Engle and Granger is constructedto test the causality between variables [44].

3.3.2. Panel data modelThere are three main types of panel data models: individual

fixed-constant coefficient model, individual fixed-varying inter-cept and varying coefficient model. Their forms are as follows:

Table 2List of income-grouped countries.

Grouping Incomestandard

List of countries

(dollar)

Low-income countries <1006 Bangladesh, Tanzania, Ethiopia, Cambodia, Kenya, Mozambique, Tajikistan, KyrgyzstanLower-middle income

countries1006–3975 Ukraine, Egypt, Indonesia, Nigeria, Philippines, Vietnamese, Nicaragua, Sri Lanka, India, Pakistan, Paraguay, Bolivia

Upper-middle incomecountries

3976–12275 South Africa, Argentina, Brazil, Mexico, Malaysia, Venezuela, Turkey, China, Russia, Rumania, Belarus, Bulgaria,Kazakhstan, Thailand

High-income countries >12775 Israel, Italy, Japanese, Canada, Britain, America, France, Germany, Netherlands, New Zealand, Australian, Spain, Korea,Singapore, Czech, Poland

Notes: the source from World Bank database. http://data.worldbank.org.cn/indicator.

Fig. 1. Flow chart of the research methodology.

S. Niu et al. / Electrical Power and Energy Systems 53 (2013) 338–347 341

yit ¼ aþ bxit þ uit ði ¼ 1;2; . . . ;N; t ¼ 1;2; . . . ; TÞ ð1Þ

yit ¼ ai þ bxit þ uit ði ¼ 1;2; . . . ;N; t ¼ 1;2; . . . ; TÞ ð2Þ

yit ¼ ai þ bixit þ uit ði ¼ 1;2; . . . ;N; t ¼ 1;2; . . . ; TÞ ð3Þ

Whether a model is selected is usually determined by covari-ance analysis. If the latter two models are used, it is necessary toutilise the Hausman statistic to determine whether the random ef-fects model or the fixed effects model should be established.

4. Results

4.1. Results of the panel unit root tests

The study tested the stationarity of the panel data of 50 coun-tries. The results show that all tests of the first difference rejectthe joint null hypothesis at 1% significance (Table 3). Therefore,they are integrated to the first-order of a unit root series, signedas I(1). Therefore, the cointegration analysis on the panel datacan be further conducted.

4.2. Results of the panel cointegration tests

Based on a stationary time series, the ADF test proposed by Kaois applied to determine if a long-run equilibrium relationship ex-ists among the correlated variables of the four income-groupings.The null hypothesis of this method is of no cointegration. Table 4displays the results of the panel cointegration tests.

Table 4 shows that only the relationship between electricityconsumption and LCON in low-income countries exists at the10% significance level. Relations between electricity consumption

and other variables of all groups are shown at a 5% or 10% signifi-cance level. Therefore, we can conclude that there are long-runequilibrium relationships between electricity consumption andall other variables for every country, only with different signifi-cance levels. In particular, the relationship for upper-middleincome-countries is more significant.

4.3. Results of the Granger causality tests

The cointegration analysis only indicates causality betweenvariables; it cannot identify the direction of the interacting rela-tionships. Therefore, this study further applies the error correctionmodel based on panel data to test the direction of the causality be-tween variables. The results are listed in Table 5.

In the short run, there are no unidirectional Granger causalitiesfrom electricity consumption to the five HDIs in low-income coun-tries. Unidirectional causality only exists from LURR to LELC,namely urbanisation advances electricity consumption. This find-ing indicates that it is difficult to generate an interaction effect be-tween electricity consumption and HDI in the low-income stage. Inlower-middle income countries, a bidirectional causal linkage isobserved between LELC and LGDP and unidirectional causality isfound from LCON to LELC. Moreover, bidirectional causalities existbetween LELC with LGDP and LCON in upper-middle income andhigh-income countries, while no causality is found between theother variables. This result suggests that, in the short run, the inter-actions between electricity consumption with GDP and consump-tion expenditure are represented only when the income level isrelatively high. In addition, electricity consumption has no relationwith urbanisation rate, life expectancy at birth or the adult literacyrate. This means that it will take a long time to accelerate humandevelopment through increasing electricity consumption.

Table 3The results of panel unit root tests.

Variables Level First difference

LLC Fisher-ADF Fisher-PP LLC Fisher-ADF Fisher-PP

LELC 9.919 22.992 14.325 �7.843*** 238.495*** 305.752***

LGDP 12.876 10.058 4.709 �7.078*** 232.593*** 242.964***

LCON 12.500 9.412 3.936 �9.210*** 276.312*** 295.404***

LURR 2.632 19.538 54.342 �6.846*** 186.430*** 179.279⁄⁄⁄

LELB 9.748 26.046 22.579 �5.319*** 206.319*** 274.950***

LLIR 6.995 3.774 4.100 �17.358*** 379.389*** 382.520***

Notes: Null hypothesis is that there is a unit root in the sequence. ⁄⁄ and ⁄Indicate respectively the rejection of the null hypothesis at a significance level of 5%, 10%.*** Indicate respectively the rejection of the null hypothesis at a significance level of 1%.

Table 4Panel cointegration tests on electricity consumption and the dependent variables.

Countries ADF value

LGDP LCON LURR LELB LLIR

Low-income countries �2.43*** �1.394* �2.425*** �2.040** 1.887**

Lower-middle income countries �3.733*** �1.737** 2.017** 1.807** 2.723***

Upper-middle income countries �2.183*** �2.769*** �1.715** �4.078*** �4.110***

High-income countries �2.370*** �1.585** �2.179** �1.955** �2.428***

Notes: Null hypothesis is that there is no cointegration relationship between two variables.*** Indicate respectively the rejection of the null hypothesis at a significance level of 1%.** Indicate respectively the rejection of the null hypothesis at a significance level of 5%.* Indicate respectively the rejection of the null hypothesis at a significance level of 10%.

342 S. Niu et al. / Electrical Power and Energy Systems 53 (2013) 338–347

In the long run, there exist bidirectional Granger causalities be-tween electricity consumption and the five indicators among all 50countries. In detail, the causalities in low-income countries are at10% or 5% significance levels, while in the other three groups thesignificance level is 5% or 1%. This demonstrates that long-run bidi-rectional causalities exist between electricity consumption and thefive HDIs. An increase in electricity consumption promotes eco-nomic and social development, which, in turn, stimulates an in-crease in electricity consumption. In particular, the interactionamong the variables is more evident when the income level is high.

4.4. Parameter estimation of the panel models

The results of the panel Granger causality tests show that bidi-rectional causality exists between electricity consumption and allfive HDIs in the long run. Based on these results, we established pa-nel regression models for the variables of electricity consumptionand the five HDIs to estimate the parameters related to these vari-ables. According to the Hausman test, fixed effects models were re-quired. To facilitate the interpretation of parameters, raw datawere used to analyse the panel models.

4.4.1. Individual fixed-varying intercept modelThe fixed effects-varying intercept model (Model 1) for the rela-

tionship between the electricity consumption and GDP of the 50countries can be established as follows:

GDPit ¼ 904:17þ ai þ 1:89ELCit ði ¼ 1;2; . . . ;50; t

¼ 1;2 . . . ;20Þ ð4Þ

Eq. (4) indicates that per-capita GDP increases with an increase inper capita electricity consumption and that the marginal contribu-tion ratio of per-capita electricity is 1.89. In Eq. (4), C is a constantand ai reflects the fixed effects of the individual country. Similarly,the models reflecting the relationship between electricity consump-tion and the other four dependent variables can be also established.The relevant parameters are listed in Table 6 and the correspondingvalues of ai are shown in Fig. 2. In particular, the values of ai in Mod-

els 1 and 2 are measured in per-capita one thousand dollars becauseof the larger absolute values of GDP and consumer expenditure.Fig. 1 shows that the value of ai gradually increases from low-in-come to high-income countries. In addition, most values of ai ofhigh-income (low-income) countries are positive (negative), indi-cating differences in the development bases of these countries. Itshould be noted that in Model 4 the values of ai are negative forcountries with higher life expectancies, such as the United Statesand Canada. This is mainly because C and b are constant and thereis excessive electricity consumption in these countries, so the val-ues of ai become negative.

This study established another five individual fixed-varyingintercept models for the variables of electricity consumption andthe five HDIs of the 50 income-grouped countries. Table 7 showsthat intercept C increases with a rise in income in all five models,which means that the basic level of the five HDIs is higher in high-er-income countries than it is in other countries. Moreover, the val-ues of C of high-income countries differ greatly from those of low-income countries in Models 1, 2 and 3. Specifically, the values of Cof high-income countries are 14 times, 20 times and 3.6 times thatof low-income countries in Models 1, 2 and 3, respectively. For thecoefficient b, the variation trend of each b in the five models hasindividual characteristics. The b shows an upward tendency as in-comes rise in Models 1 and 2, which indicates that not only theamount of electricity consumption is great but also the marginaleffect of unit electricity consumption is large in higher-incomecountries. Both aspects cause the values of the dependent variablesof high-income countries to be much larger than those of low-in-come countries. However, in Models 3, 4 and 5, b shows a down-ward trend as incomes increase, while all b values are less than1%. This is because the level of urbanisation rate, life expectancyat birth and the adult literacy rate has reached a comparativelyhigh level in relatively high income countries. Although electricityconsumption increased during 1990–2009, the three indicatorschanged little. Instead, the values of the three indicators in low-er-middle income countries increased with an increase in electric-ity consumption over the same period, which indicated a highermarginal effect.

Table 5The results of Panel granger causality tests.

Variables Short-run Long-run

Low-incomecountries

Lower-middleincome countries

Upper-middleincome countries

High-incomecountries

Low-incomecounties

Lower-middleincome countries

Upper-middleincome countries

High-incomecountries

LELC ? LGDP – 2.67*** 6.80*** 6.44*** �1.94* �2.86** �2.54** �2.78***

LELC ? LCON – – 5.77*** 2.52** �2.23** �2.04** �2.49** �2.49**

LELC ? LURR – – – – �1.60* �14.99** �7.62*** �9.44***

LELC ? LELB – – – – �1.76* �2.06** �3.02*** �4.08***

LELC?LLIR – – – – �1.78* �2.17** �3.08*** �5.24***

LGDP ? LELC – 4.45*** 4.93*** 8.02*** �2.08** �2.91*** �2.63*** �2.11**

LCON ? LELC – 1.82⁄ 4.10*** 6.41*** �1.86* �3.38*** �2.60*** �2.75***

LURR ? LELC 5.28*** – – – �2.47** �3.51*** �4.92*** �2.71**

LELB ? LELC – – – – �2.77** �2.61*** �3.62*** �2.86***

LLIR ? LELC – – – – �1.96** �2.03** �4.49*** �5.59***

*** Indicate respectively the rejection of the null hypothesis at a significance level of 1%.** Indicate respectively the rejection of the null hypothesis at a significance level of 5%.* Indicate respectively the rejection of the null hypothesis at a significance level of 10%.

Table 6The parameter estimation of varying intercept panel data models for electricity consumption and five HDIs.

Parameter GDP (M1) CON (M2) URR (M3) ELB (M4) LIR (M5)

C 904.17 (4.69) 1000.89 (7.56) 52.12 (105.24) 63.61 (191.57) 83.66 (32.16)b 1.89 (34.29) 0.98 (25.70) 0.002 (11.18) 0.002 (17.26) 0.001 (11.52)AR2 0.990 0.987 0.989 0.964 0.609F 2017.581 1473.513 1791.277 533.570 32.063

Notes: AR2, F respectively represents mean square deviation and F statistics. The values in the brackets are the corresponding t statistics. Mi is the model number.

Fig. 2. The changes of intercept for electricity consumption and five HDIs in 50 countries.

S. Niu et al. / Electrical Power and Energy Systems 53 (2013) 338–347 343

4.4.2. Individual fixed-varying coefficient modelThe regression equation of formula (3) shows that the indica-

tors are influenced not only by the individual but also by the struc-ture of electricity consumption in every country. In other words,the intercept ai and coefficient bi are variable. The varying coeffi-cient models of electricity consumption in 50 countries and fiveindexes (Yit) are thus expressed by the following formula:

Yit ¼ C þ ai þ biELCit ði ¼ 1;2; . . . ;50; t ¼ 1;2; . . . ;20Þ ð5Þ

Because ai, bi, the explanatory variables and the dependent vari-ables are all variable, bi is difficult to reflect the differences in differ-ent income-grouped countries. Therefore, it is necessary to combine

bi with the corresponding explanatory variable to explain the differ-ences in the 50 countries. We build a three-dimensional scatter dia-gram of bi, the average electricity consumption of every country in20 years and the corresponding average values of dependent vari-ables. In doing so, the differences in the coefficients of different in-come-grouped countries are clear. As can be seen in Fig. 3, in Model1 electricity consumption in high-income countries contributesmost to GDP, followed by upper-middle income countries. For low-er-middle income countries as well as low-income countries,although the bi values of some of these countries are larger, GDPis still low because of low electricity consumption. In Model 2,the change in coefficients is similar to that in Model 1 (Fig. 4). For

Table 7The parameters of electricity consumption and five HDIs in four income-groupings.

Para-meter Grouping GDP (M1) CON (M2) URR (M3) ELB (M4) LIR (M5)

C Low-income countries 175.3(10.0) 122.9(8.8) 19.4(30.2) 56.8(85.1) 64.6(6.9)Lower-middle income countries 485.9(24.7) 509.7(23.9) 39.3(50.2) 62.6(152.2) 73.0(69.7)Upper-middle income countries 1023.1(7.1) 796.5(6.4) 58.4(71.4) 64.8(109.9) 87.9(133.3)High-income countries 2389.2(3.1) 2453.1 (4.6) 71.0(120.5) 66.4(155.4) 95.4(426.8)

b Low-income counties 0.20(7.11) 0.18(8.25) 0.008(7.37) 0.001(0.85) 0.008(0.52)Lower-middle income countries 0.51(18.96) 0.11(3.89) 0.002(2.26) 0.004(7.67) 0.005(3.44)Upper-middle income countries 0.80(15.67) 0.44(10.02) 0.003(10.5) 0.002(7.97) 0.002(8.11)High-income countries 2.30(22.88) 1.18(16.67) 0.001(13.1) 0.001(26.8) 0.0004(12.3)

AR2 Low-income counties 0.700 0.696 0.904 0.901 0.282Lower-middle income countries 0.939 0.893 0.934 0.923 0.921Upper-middle income countries 0.956 0.932 0.976 0.883 0.903High-income countries 0.969 0.962 0.985 0.873 0.943

F Low-income counties 47.412 46.575 187.186 181.075 8.807Lower-middle income countries 308.264 167.878 286.644 240.102 232.132Upper-middle income countries 435.092 274.503 833.246 151.943 187.764High-income countries 637.233 507.605 1364.244 138.882 334.781

Notes: AR2, F respectively represents mean square deviation and F statistics. The values in the brackets are the corresponding t statistics. Mi is the model number.

344 S. Niu et al. / Electrical Power and Energy Systems 53 (2013) 338–347

the indicators of urbanisation rate, life expectancy at birth and theadult literacy rate, the three indicators of high-income countries areclose to the upper limit. Therefore, there is little room to improvethese indicators by increasing electricity consumption. In otherwords, the bi values of high-income countries change little, whichmeans that increasing electricity consumption seems to be of noremarkable significance for improving the three indices. However,the three indicators have larger room to increase in all other coun-tries. Specifically, the bi values of part countries increase faster aselectricity consumption increases. It indicates that unit electricityconsumption has a greater positive impact on the three indicators(Figs. 5–7). Obviously, the results are found to be the same as thatof the varying intercept model. In a word, when the dependent vari-ables are close to the upper limit or converge, increasing the explan-atory variables contributes little to them. On the contrary, when thedependent variables still have larger room to be increased, increas-ing the explanatory variables has a great influence on them. The re-search of Mazur [3] showed that an increase in electricityconsumption is difficult to improve life quality in industrialised na-tions. This conclusion is similar to the results of this study.

The coefficients change not only at the individual level, but alsoin time series. The time-varying coefficient model is as follows:

Yit ¼ C þ ai þ btELCit ði ¼ 1;2; . . . ;50; t ¼ 1;2; . . . ;20Þ ð6Þ

The time-varying coefficient bt is much smaller than the indi-vidual varying coefficient bi. The bt values of every income-groupedcountry are on the rise during 1990–2009 but vary from year toyear. In the time-varying coefficient model for electricity consump-tion and GDP, the bt’s basic value of high-income countries is larg-est and increases quicker than other countries, from 0.82 in 1990to 1.39 in 2007, although there is a slight decrease in the subse-quent two years. For the upper-middle income and lower-middleincome countries, the bt values increase from 0.51 to 0.77 and from0.49 to 0.65, respectively. In addition, the bt values of low-incomecountries increase slowly but the basic value is small. In short, thehigher the country’s income, the more electricity consumptioncontributes to GDP (Fig. 8). In the time-varying coefficient modelfor electricity consumption and CON, the basic values of the fourgroups of bt have no apparent sequence. However, the trend of bt

shows that the bt values of higher-income countries increase fasterthan those of lower-income countries, which indicates a positiveconnection between consumer expenditure and electricity con-sumption (Fig. 9). The coefficients of electricity consumption tourbanisation rate, life expectancy at birth and the adult literacyrate are all below 0.06, which means the effect of electricity con-

sumption on urbanisation rate, life expectancy at birth and theadult literacy rate is very small. In addition, the bt values of low-er-income countries are larger than those of higher-income coun-tries. In other words, a negative correlation is found between bt

and income (Fig. 10). This implies that the time effect of electricityconsumption on the three indicators is clear in lower-middle in-come countries but small in high-income countries.

5. Conclusion and recommendations

5.1. Conclusion

Panel cointegration tests show a long-run equilibrium relation-ship between electricity consumption and the five HDIs of income-grouped countries. In particular, in upper-middle and high-incomecountries the relationship is most significant. Granger causalitytests further prove that a long-run bidirectional causal linkageexists between the variables. This result indicates that electricityconsumption not only promotes human development, but also thathuman development expands the social demand of electricity. Inother words, the dependent and explanatory variables are cause-and-effect. However, such a good interaction relationship is notconsistent in the short-term. Only when income is relatively highcan the influence of electricity consumption on GDP and consump-tion expenditure be observed. In addition, electricity consumptionhas no effect on urbanisation rate, life expectancy at birth and theadult literacy rate. Further, the reverse effect is also inexistent.Thus, any improvement in the HDI needs a comparatively longperiod.

In the individual fixed-varying intercept models of the 50 coun-tries, the intercept ai reflects the differences in the developmentbases of different income countries. ai Increases gradually fromlow-income to high-income countries. Specifically, the ai valuesof most high-income countries are positive, whereas they arenegative for low-income countries. In the five varying interceptmodels established according to income groups, the constant C in-creases as income increases and the base of the five HDIs in higher-income countries is larger. For the coefficient b, its changing trendis of two situations. In the models of electricity consumption withGDP and consumption expenditure separately, b is on the rise asincome increases. This reflects that the higher the income, thelarger is electricity consumption, while the marginal effect is alsohigher. In the models of electricity consumption with urbanisation,life expectancy and the adult literacy rate, b decreases with an

Fig. 3. The changes of coefficients for ELC to GDP in 50 countries.

Fig. 4. The changes of coefficients for ELC to CON in 50 countries.

Fig. 5. The changes of coefficients for ELC to URR in 50 countries.

Fig. 6. The changes of coefficients for ELC to ELB in 50 countries.

Fig. 7. The changes of coefficients for ELC to LIR in 50 countries.

Fig. 8. The time-varying coefficients for ELC to GDP of four income-groups.

S. Niu et al. / Electrical Power and Energy Systems 53 (2013) 338–347 345

increase in income. This result is because the three indicators inhigher-income countries have reached a higher level and areincreasingly converging.

In individual fixed-varying coefficient models, the changingtrend of bi is consistent with that of b in the varying intercept mod-els built according to income groups. Similarly, with regard to GDPand consumption expenditure, the bi values of high-income coun-tries are larger, whereas bi changes little in the other indicators of

high-income countries. The changing trend of bi of low-incomecountries is the opposite. The results indicate that the interactionof electricity consumption with GDP and consumption expenditureshows a clear Matthew effect. Because the urbanisation rate, lifeexpectancy at birth and adult literacy rate of high-income coun-tries are close to the upper limit, increasing electricity consump-tion contributes little to these three indicators. For the othercountries, an increase in electricity consumption shows a positive

Fig. 9. The time-varying coefficients for ELC to CON of four income-groups.

Fig. 10. The time-varying coefficients for ELC to URR of four income-groupings.

346 S. Niu et al. / Electrical Power and Energy Systems 53 (2013) 338–347

effect on the three indicators for which the room for growth iscomparatively large.

The effect of bt in the time-varying coefficient models is muchsmaller than that of bi. All the bt values of every income-groupedcountry in 1990–2009 show an upward trend. However, the fourgroups’ bt values vary from each other. In the time-varying coeffi-cient models for electricity consumption with GDP and consump-tion expenditure, the lower the income level, the smaller is thebt. Further, the bt values of high-income countries increase rapidly.However, the changing trend of bt is opposite in the time-varyingcoefficient models for electricity consumption with urbanisationrate, life expectancy at birth and the adult literacy rate. In particu-lar, high-income countries’ bt values change very little.

5.2. Policy recommendations

The use of electricity relies on electrical equipment. People notonly have to pay for the electricity but also must purchase electri-cal appliances, which are often more expensive than the electricitybill. Therefore, electric energy is also considered to be a kind ofcostly energy [45]. For low-income families, although they havethe willingness to increase electricity consumption, they do nothave the necessary ability to pay for electricity. Therefore, theyhave no choice but to save electricity [46]. Even though increasingelectricity can improve the level of human development, it is moreimportant to improve the level of income.

It should be noted that a country’s HDI is the comprehensive re-sult of multiple factors. Electricity consumption is just one of thesefactors. This study analysed the relationship between electricityconsumption and five HDIs and concluded that there is a long-run causality link between them. However, it cannot be regardedthat as soon as electricity consumption increases, HDIs completelyimprove. In addition, we find that (a) electricity consumption isalso influenced by a country’s geographic position, energy struc-ture and other factors, (b) the use of electricity in countries at highlatitudes is greater than that in other countries, and (c) electricityconsumption is less in countries where non-renewable energy isused more. Limited by the length of the study, the above issuesare not discussed here.

It can be expected that the status of electricity consumption andhuman development in upper-middle and high-income countrieswill be the future of that in lower-middle and low-income coun-tries. Low-income countries or regions can learn from the experi-ences of higher-income countries. They can plan electricitydevelopment according to the requirements of human develop-ment. In turn, they can set goals for human development basedon the supply capacity of electricity.

The results in this paper show that a good interactive relation-ship exists between electricity consumption and human develop-ment in the 50 countries studied herein. The following fourpolicy implications can be drawn:

(1) Lower-middle and low-income countries should regard elec-tricity supply as a basic public service in modernisation inorder to improve the availability and accessibility of electric-ity [47].To build a sound electricity grid system, they shoulddevelop and utilise multi-energy resources. In particular, theimplementation of electrical engineering in remote ruralareas can provide the necessary conditions for human devel-opment. Moreover, to improve the electric power-supplyingcapacity of recipient regions in the South–South cooperationand the process of international aid, countries should fostertechnical and economic cooperation between nations.

(2) To reduce the environmental problems brought about by anincrease in electricity consumption, we should aim todevelop new energy that is clean, renewable and low cost.This should vigorously apply to new technology when utilis-ing water power, solar, wind, nuclear and biomass energy toproduce electricity. We should also promote low-costenergy-saving technology to improve the efficiency ofhousehold appliances [48] such as lighting equipment, airconditioners, refrigerators and television sets. Previous stud-ies have shown that average household electricity would becut by 50.9% if energy-saving lamps replaced incandescentbulbs in England [49]. Further, in China, if the efficiency ofhousehold appliances is effectively improved, the electricityconsumption of urban households could reduce by 28% by2020 [50].

(3) Electricity construction plays an important role in improvinghuman development, but it is not the only approach to do so.In relatively low income countries or regions, as well asaccelerating the construction of electrification, they needmake efforts to eliminate poverty, make education universaland improve health conditions to create conditions that arecompatible with human development [24].

(4) In a large country such as China, there are huge variations inhuman development from east to west and from cities torural areas. Similarly, regions can be categorised into fourgroups according to income levels. To promote humandevelopment in middle- and low-income regions, we shouldspeed up the construction of a power grid. At the same time,we should try to achieve a power grid and electricity price in

S. Niu et al. / Electrical Power and Energy Systems 53 (2013) 338–347 347

the countryside that is the same as that in cities. In addition,the activity ‘‘home appliances to the countryside’’ shouldcontinue [51]. Based on these measures, a farmer’s electric-ity cost could be reduced.

We consider that in order to improve HDIs, electric power con-struction should be incorporated into basic public services con-struction to improve the availability of electricity for low-incomeresidents. In the future work, we will obtain a large number ofaccurate electricity consumption information through surveys, sta-tistics and fixed-site observation. Case studies are conducted, dif-ferences of HDI index among regions with different degree ofelectrification are compared. The impact of electricity pricechanges on residential electricity is analyzed. The relationship be-tween the electrification degree gaining ground and the improve-ment of other public services is discussed.

Acknowledgements

The authors would like to thank two anonymous reviewers forreferees and the editors for their valuable suggestions and helpfulcomments. We also thank Qi Jinghui for the modification of figures.This study is supported by National Natural Science Foundation ofChina and Ministry of Education of China (Grant Nos. 41171437,20100211110018).

References

[1] Bhargava N, Singh B, Gupta S. Consumption of electricity in Punjab: structureand growth. Energy Policy 2009;37(6):2385–94.

[2] Petterssona F, Soderholmb P, Lundmark R. Fuel switching and climate andenergy policies in the European power generation sector: a generalizedLeontief model. Energy Econ 2012;34(1):183–98.

[3] Mazur A. Does increasing energy or electricity consumption improve quality oflife in industrial nations? Energy Policy 2011;39(5):2568–72.

[4] Apergis N, Payne JE. A dynamic panel study of economic development and theelectricity consumption–growth nexus. Energy Econ 2011;33(5):770–81.

[5] Lenin V. Collected works vol. 31; 1920. p. 516. <http://www.marxists.org/archive/lenin/works/cw/index.htmS>.

[6] Gurgul H, Lach Ł. The electricity consumption versus economic growth of thepolish economy. Energy Econ 2012;34(2):500–10.

[7] Ferguson R, Wilkinson W, Hill R. Electricity use and economic development.Energy Policy 2000;28(13):923–34.

[8] Payne JE. A survey of the electricity consumption–growth literature. ApplEnergy 2010;87(3):723–31.

[9] Joyeux R, Ripple RD. Household energy consumption versus income andrelative standard of living: a panel approach. Energy Policy 2007;35(1):50–60.

[10] Bhide A, Monroy CR. Energy poverty: a special focus on energy poverty in Indiaand renewable energy technologies. Renew Sustain Energy Rev2011;15(2):1057–66.

[11] Kaygusuz K. Energy services and energy poverty for sustainable ruraldevelopment. Renew Sustain Energy Rev 2011;15(2):936–47.

[12] Kanagawa M, Nakata T. Analysis of the energy access improvement and itssocio-economic impacts in rural areas of developing countries. Ecol Econ2007;62(2):319–29.

[13] Dilaver Z, Hunt LC. Modelling and forecasting Turkish residential electricitydemand. Energy Policy 2011;39(6):3117–27.

[14] Niu SW, Zhang X, Zhao CS, Ding YX, Niu YZ, Christensen TH. Household energyuse and emission reduction effects of energy conversion in Lanzhou city. ChinaRenew Energy 2011;36(5):1431–6.

[15] Sugiyama M. Climate change mitigation and electrification. Energy Policy2012;44:464–8.

[16] Pinheiro G, Rendeiro G, Pinhoc J, Macedo E. Rural electrification for isolatedconsumers: sustainable management model based on residue biomass. EnergyPolicy 2011;39(10):6211–9.

[17] International energy agency (IEA), energy and poverty. World energy outlook2010. International Energy Agency, Paris; 2010. p. 2016–29.

[18] Pereira MG, Freitas MAV, Silva NF. Rural electrification and energy poverty:empirical evidences from Brazil. Renew Sustain Energy Rev 2010;14(4):1229–40.

[19] Loisel R. Power system flexibility with electricity storage technologies: atechnical-economic assessment of a large-scale storage facility. Electr PowerEnergy Syst 2012;42:542–52.

[20] Sim T, Castro R, Sim J. Wind power pricing: from feed-in tariffs to theintegration in a competitive electricity market. Electr Power Energy Syst2012;43:1155–61.

[21] He YX, Xia T, Xiong W, Wang B, Liu YY. Risk transmission assessment ofelectricity price chain in China based on ISM and ECM. Electr Power EnergySyst 2013;46:274–82.

[22] Kanagawa M, Nakata T. Assessment of access to electricity and the socio-economic impacts in rural areas of developing countries. Energy Policy2008;36(6):2016–29.

[23] Ghali KH, El-Sakka MIT. Energy use and output growth in Canada: amultivariate cointegration analysis. Energy Econ 2004;26:225–38.

[24] Zahnd A, Kimber HM. Benefits from a renewable energy village electrificationsystem. Renew Energy 2009;34:362–8.

[25] Wu QS, Clulow V, Maslyuk S. Energy consumption inequality and humandevelopment. In: 2010 International conference on management science &engineering (17th), Melbourne, Australia; 2010. p. 1398–409.

[26] Holtedahl P, Joutz FL. Residential electricity demand in Taiwan. Energy Econ2004;26(2):201–24.

[27] Burney NA. Socioeconomic development and electricity consumption: a cross-country analysis using the random coefficient method. Energy Econ1995;17(3):185–95.

[28] Granaham M. Measurement and analysis of socio-economicdevelopment. Geneva: UNRISD; 1972.

[29] OECD/DAC. Socioeconomic typologies or criteria for 82 developing countries.Organization for economic co-operation and development, Paris; 1977.

[30] Morris MD. Measuring the world’s poor: the physical quality of life index.Pergamon Press, 1979; New York.

[31] UNDP (United Nations Development Programme)-Human DevelopmentReport Office, 1990. Human development reports 1990. New York: OxfordUniversity Press through 2005; and Palgrave Macmillan since 2006.

[32] Noorbakhsh F. A modified human development index. World Dev1998;26(3):517–28.

[33] Grimm M, Harttgen K, Klasen S, Misselhorn M. A human development index byincome groups. World Dev 2008;36(12):2527–46.

[34] Engineer MH, Roy N, Fink S. Healthy human development indices. Soc IndicRes 2010;99(1):61–80.

[35] Harttgen K, Klasen S. A household-based human development index. WorldDev 2012;40(5):878–99.

[36] Sager AD, Najam A. The human development index: a critical review. Ecol Econ1998;25(3):249–64.

[37] Levin A, Lin CF, Chu CS. Unit root tests in panel data: asymptotic and finite-sample properties. J Econ 2002;108(1):1–24.

[38] Breitung J. Nonstationary panels, panel cointegration, and dynamic panels.Adv Econ 2000;15:161–78.

[39] Hardi K. Testing for stationarity in heterogeneous panel data. Econ J2000;3(2):148–61.

[40] Im KS, Pesaran MH, Shin Y. Testing for unit roots in heterogeneous panels. JEcon 2003;115(1):53–74.

[41] Maddala GS, Wu SW. A comparative study of unit root tests with panel dataand a new simple test. Oxford Bull Econ Stat 1999;61(s1):631–52.

[41] Pedroni P. Critical value for cointegration tests in heterogeneous panels withmultiple regressors. Oxford Bull Econ Stat 1999;61(s1):653–70.

[43] Kao C. Spurious regression and residual-based tests for cointegration in paneldata. J Econ 1999;90(1):1–44.

[44] Engle RF, Granger CWJ. Cointegration and error–correction: representation,estimation and testing. Econometrica 1987;55:251–76.

[45] Arthur MFSR, Zahran S, Bucini G. On the adoption of electricity as a domesticsource by Mozambican households. Energy Policy 2010;38(11):7235–49.

[46] Niu SW, Zhang X, Zhao CS, Niu YZ. Variations in energy consumption andsurvival status between rural and urban households: a case study of theWestern Loess Plateau. China Energy Policy 2012;49:515–27.

[47] Sayin C, Mencet MN, Ozkan B. Assessing of energy policies based on Turkishagriculture: current status and some implications. Energy Policy2005;33(18):2361–73.

[48] Lee YC, Hu JL, Kao CH. Efficient saving targets of electricity and energy forregions in China. Electr Power Energy Syst 2011;33:1211–9.

[49] Wall R, Crosbie T. Potential for reducing electricity demand for lighting inhouseholds: an exploratory socio-technical study. Energy Policy2009;37(3):1021–31.

[50] Akinobu M, Yasuhiko K, Mu HL, Zhou WS. Electricity demand in the Chineseurban household-sector. Appl Energy 2008;85(12):1113–25.

[51] Wu RJ, Li DJ, Wu B. The study on Chinese rural consumers’ purchase intentionfor home appliances going to the countryside. China Soft Sci 2010;1:40–52.