at SciVerse ScienceDirect

Energy 39 (2012) 146e153

Contents lists available

Energy

journal homepage: www.elsevier .com/locate/energy

The dynamics of electricity consumption and economic growth: A revisit study oftheir causality in Pakistan

Muhammad Shahbaz a, Hooi Hooi Lean b,*

aCOMSATS Institute of Information Technology, M. A. Jinnah Campus, Lahore, Pakistanb Economics Program, School of Social Sciences, Universiti Sains Malaysia, 11800 USM, Penang, Malaysia

a r t i c l e i n f o

Article history:Received 8 August 2011Received in revised form20 January 2012Accepted 21 January 2012Available online 22 February 2012

JEL classifications:F15B28

Keywords:Electricity consumptionEconomic growthGranger causality

* Corresponding author. Tel: þ60 4 653 2663; fax:E-mail address: hooilean@usm.my (H.H. Lean).

0360-5442/$ e see front matter � 2012 Elsevier Ltd.doi:10.1016/j.energy.2012.01.048

a b s t r a c t

This study revisits the relationship between electricity consumption and economic growth in Pakistan bycontrolling and investigating the effects of two major production factors e capital and labor. Theempirical evidence confirms the cointegration among the variables and indicates that electricityconsumption has a positive effect on economic growth. Moreover, bi-directional Granger causalitybetween electricity consumption and economic growth has been found. The finding suggests thatadoption of electricity conservation policies to conserve energy resources may unwittingly declineeconomic growth and the lower growth rate will in turn further decrease the demand for electricity.Therefore, government contemplating such conservationist policies should instead explore and developalternate sources of energy as a strategy rather than just increasing electricity production per se in orderto meet the rising demand for electricity in their quest towards sustaining development in the country.

� 2012 Elsevier Ltd. All rights reserved.

1. Introduction

The causal relationship between electricity consumption andeconomic growth has important policy implications based on thefollowing four hypotheses. First, if the causality runs from elec-tricity consumption to economic growth, then electricity conser-vation policies should be discouraged as it would be counter-productive. Second, the electricity conservation policies can onlybe safely adopted if the causality runs instead from economicgrowth to electricity consumption. The third hypothesis states thatif there is an existence of bi-directional causality between the twovariables (feedback effect) which implies that reduction in elec-tricity consumption may adversely affect economic growth andvice-versa, then a different policy approach would be needed.Fourth and lastly, if the neutral hypothesis holds true or if it isproven that there is no causal relationship between the two vari-ables, therefore such a conclusion would reveal the surprisinglyminor role of electricity consumption and its effects in theeconomic growth of a country.

We note that energy is one of the basic production factors.However, Payne [1] argued that electricity plays a crucial role in the

þ60 4 657 0918.

All rights reserved.

production with the evidence of strong correlation between elec-tricity consumption and economic growth in 100 countries foundby Ferguston et al. [2]. Pao [3] also documented that electricity isthe most flexible form of energy and constitutes one of the vitalinfra-structural inputs in socioeconomic development. Here, wefocus on electricity consumption in this study rather than theenergy consumption which has been investigated in the volumi-nous literature.

Electricity is a major source of energy in the industrial andagriculture sectors and together these two sectors contributedalmost 50% of Pakistan’s GDP. As reported in the Pakistan EnergyYear Book 2009, the share of industrial, agriculture and commercialsectors in electricity consumption is 27.5%, 12.5% and 7.5% respec-tively. At the same time, the contribution of industrial, agricultureand services sectors to GDP is 24.26%, 21.55% and 54.18% respec-tively. These statistics show that these sectors contributed signifi-cantly to GDP and electricity consumption. Hence, it is clear thatelectricity consumption does play an important role in GDP ofPakistan through the contribution of these three sectors.

However, production growth and expansion of economicactivities in Pakistan are restrained by its under-developed elec-tricity infrastructure. Intentionally-engineered electrical poweroutages i.e. load-sheddings are frequently used to curb theincreasing demand of electricity in the country. Any shortage inelectricity will be harmful to output. Rehana et al. [4] found that the

1 Albania, Belarus, Bulgaria, Czech Republic, Estonia, Latvia, Lithuania, Mace-donia, Moldova, Poland, Romania, Russian Federation, Serbia, Slovak Republic andUkraine.

2 Aqeel and Butt [6] used oil consumption, natural gas consumption, electricityconsumption and overall energy consumption as indicators of energy consumption.

M. Shahbaz, H.H. Lean / Energy 39 (2012) 146e153 147

recent electricity crises begin from 2007 in Pakistan has caused anoutput loss of at least 12% to the industrial sector in the country.

Khan and Ahmad [5] reported that the current electricityproduction in Pakistan is around 479.16 MWh whereas the elec-tricity demand had jumped from 625 MWh to 833.33 MWh in theyear 2010. This massive increase in demand suggests that thelooming electricity crisis in Pakistan will become even more severein the near future. Hence, investigation of the relationship betweenelectricity consumption and economic growth in the country areextremely important and doubly urgent to the policy makers,market players and consumers both domestic and industrial.

There are already some previous studies conducted in Pakistansuch as Aqeel and Butt [6] and Zahid [7], however, these studiesonly investigated thewider scope of energy consumption instead ofthe narrower scope of electricity consumption. Jamil and Ahmad [8]is the only study that examined the relationship between electricityconsumption and economic growth in Pakistan. However, they didnot consider other potential and vital variables such as capital andlabor in their analysis. Lütkepohl [9] argued that omission ofimportant variables would risk providing potentially biased andinappropriate results. No causal relation is found in the bi-variatesystem due to these neglected variables. Bartleet and Rukmani[10] also recommended incorporating some pertinent variables inthe analysis. Thus, we re-investigate the causality between elec-tricity consumption and economic growth in Pakistan by incorpo-rating in the effects of the capital and labor factors.

The relationship between electricity consumption and growthmay also differ in the short-run period and the long-run periodeven within the same country. Our study also examines the rela-tionship between electricity consumption and economic growth inboth the short- and long-run periods which was omitted in Jamiland Ahmad [8]. This paper offers two new and fresh contributionsto the existing literature. Firstly, we incorporate the effects of twoimportant production factors, i.e. capital and labor in the analysis.Secondly, we examine the relationship between the two variablesin both the short-run and the long-run periods.

The rest of the paper is organized as follows: Section 2 reviewsthe existing literature. Methodology and data are introduced inSection 3. Section 4 discusses the empirical results and finally theconclusion and major policy implications are presented and dis-cussed in Section 5.

2. Literature review

There is extensive literature on the nexus of energy consump-tion and economic growth; beginning with the paper by Kraft andKraft [11] to the recent studies such as Tsani [12], Shahbaz et al. [13]and Kouakou [14]. Moreover, we note that there was a completesurvey of the literature on energy undertaken by Ozturk [15]. Sincethis paper is on electricity consumption, wewill focus the literaturereview on electricity consumption only.

Empirical studies on the causality between electricityconsumption and economic growth can be divided into crosscountry studies and country specific studies. The empiricalevidence provided mixed results in cross country studies. Forexample, in ASEAN countries, Yoo [16] suggested unidirectionalcausality from real income per capita to electricity consumption inIndonesia and Thailand while in Malaysia and Singapore, bi-directional causality was found. In Africa, Wolde-Rufael [17]found economic growth Granger causes electricity consumptionin Cameroon, Ghana, Nigeria, Senegal, Zambia and Zimbabwewhilethe opposite case was found for Benin, Congo and Tunisia.

Similarly, Chen et al. [18] indicated that electricity consumptionand economic growth are mutually interdependent but unidirec-tional causality was found from GDP to electricity consumption by

utilizing the heterogeneous causality approach. In OPEC economies,Squalli [19] found unidirectional causality from electricityconsumption to economic growth in Indonesia, Nigeria, UAE andVenezuela while economic growth Granger causes electricityconsumption in Algeria, Iraq, Kuwait and Libya.

For OECD countries, Narayan and Prasad [20] showed unidi-rectional causal relation from electricity consumption to economicgrowth for Australia, the Czech Republic, Iceland, Italy, Portugal andthe Slovak Republic, but the reverse case in Finland, Hungry and theNetherlands while the feedback hypothesis was validated in Koreaand the UK. Alinsato [21] assessed the relationship between elec-tricity consumption and economic growth in Togo and Benin.Cointegration was found in Benin with growth-led-electricityconsumption.

For the panel analysis, Narayan and Smyth [22] reported bi-directional causal relationship in the Middle Eastern countries.Yoo and Kwak [23] reported that electricity consumption Grangercauses economic growth in Argentina, Brazil, Chile, Columbia andEcuador; while electricity consumption and economic growthGranger cause each other in Venezuela.

Payne [24] conducted a survey on studies of economic growthand electricity consumption and reported that electricityconsumption has positive impact on economic growth throughsome stylized facts. Acaravci and Ozturk [25] did not indicate anyexistence of cointegration between the variables for 15 transitionaleconomies.1 However, the Granger causality test was not conductedin these two studies.

In some country specific studies, it was found that economicgrowth Granger causes electricity consumption, for instance, Ghosh[26] for India, Jumbe [27] for Malawi, Narayan and Smyth [28] forAustralia, Yoo and Kim [29] for Indonesia, Mozumder and Marathe[30] for Bangladesh, Halicioglu [31] for Turkey, Aktas and Yilmaz[32] for Turkey, Pao [3] for Taiwan, Ciarreta and Zarraga [33] forSpain and Jamil and Ahmad [8] for Pakistan. On the contrary, thereare findings that the causality is running from electricityconsumption to economic growth, for example, Ramcharran [34]for Jamaica, Shiu and Lam [35] and Yuan et al. [36] for China, Alti-nay and Karagol [37] for Turkey, Ho and Sui [38] for Hong Kong,Chandran et al. [39] for Malaysia, Abosedra et al. [40] for Lebanon,Akinlo [41] for Nigeria, and Acaravci [42] for Turkey.

The feedback effect between electricity consumption andeconomic growth is also validated inmany studies, for instance, Yoo[43] for Korea, Zachariadis and Pashouortidou [44] for Cyprus, Tang[45] and Lean and Smyth [46] for Malaysia, Odhiambo [47] forSouth Africa, Ouédraogo [48] for Burkina Faso and Lorde et al. [49]for Barbados. There exists empirical evidence of a neutral hypoth-esis between electricity consumption and economic growth as well,for instance, Murry and Nan [50] for the developing economies,Asafu-Adjaye [51] for Asian countries and, Yusof and Latif [52] forMalaysia.

In the case of Pakistan, only a few studies had investigated thecausal relationship between energy consumption and economicgrowth. Aqeel and Butt [6] used different indicators2 to proxy theenergy consumption. They found unidirectional causality runningfrom economic growth to oil consumption. Although, there is nocausal relation between natural gas consumption and economicgrowth but they found unidirectional causality from total elec-tricity consumption to economic growth. On the other hand, Alam

M. Shahbaz, H.H. Lean / Energy 39 (2012) 146e153148

and Butt [53] and Qazi and Riaz [54] found bi-directional causalrelationship between energy consumption and economic growth.

Khan and Qayyum [55] examined causal relationship betweenenergy consumption and economic growth in South Asian econ-omies including Pakistan. They suggested that the causality isrunning from energy consumption to economic growth in Paki-stan. Zahid [7] also conducted a similar study in South Asiancountries using different indicators for energy consumption i.e.petroleum, gas, coal, electricity and total energy consumption. Inthe case of Pakistan, unidirectional causal relation is foundrunning from coal consumption to economic growth, also fromeconomic growth to electricity consumption and also towardstotal energy consumption. The existing studies for Pakistan yiel-ded mixed results. Hence, it is difficult to provide appropriatedirection and advice for the policy makers in formulating energypolicies for the country.

Jamil and Ahmad [8] is the only study that investigated therelationship between electricity prices, electricity consumption andreal GDP in Pakistan. They found a long-run relationship andunidirectional causality from economic growth and electricityprices to electricity consumption. Their findings may be biasedbecause they omitted two important variables in their model.Hence, we think it is highly important, if not absolutely necessary torevisit the relationship by incorporating an appropriate investiga-tion of capital and labor in the neoclassical production function tomake the inference more reliable.

3. Methodology and data

We apply Cobb-Douglas production function in investigatingthe relationship between electricity consumption and economicgrowth by incorporating capital and labor as important factors ofproduction. The general and non-linear form of production func-tion is given as follows:

Y ¼ AEa1Ka2La3eu (1)

where Y is the real GDP; E, K and L denote energy consumption, realcapital and labor respectively. The technological parameter isindicated by A and e is the residual term that is assumed to benormally distributed. The a1, a2 and a3 are returns to scale linkedwith energy consumption, capital and labor respectively. We thenconvert the non-linear Cobb-Douglas production function intolinear specification by taking logs for empirical purpose and thenew form of model is given as follows:

lnYt ¼ lnA þ a1lnEt þ a2lnKt þ a3lnLt þ ut (2)

lnYt ¼ a þ a1lnECt þ a2lnKt þ a3lnLt þ ut (3)

where lnYt , lnECt , lnKt and lnLt is the natural logarithm of realGDP per capita, electricity consumption per capita (electricityconsumption per capita may capture the better effect of tech-nology and energy) in KWh, real capital per capita and laborparticipation per capita respectively, mt is the error term. Themodel is a multivariate model where we have four variables in thesystem.3 We investigate the relationship between electricityconsumption and economic growth in UECM and VECM frame-work. Instead of looking at the aggregate effect, we are looking atthe per capita effect which will be more useful when we comparethe effect in Pakistan with other countries. In fact, many studies in

3 The log-linear specification is helpful in attaining consistent and efficientresults as logarithm transformation vanishes the variation in times series data. Thesimple linear specification may provide inappropriate and unreliable results.

the literature (e.g. Aqeel and Butt [6], Ghosh [26], Narayan andSmyth [28]) also employed per capita data for the similar analysis.We observe that electricity is one of major energy consumption inPakistan. Therefore, it is important for us to study the electricityconsumption besides the energy consumption which has beenstudied wisely.

The annual time series data for all variables are obtained fromWorld Development Indicators (WDI, 2010) for the sample periodof 1972e2009. In Pakistan, the data of gross capital formation isavailable in WDI which is a widely use database in the economicliterature. We compute the capital stock using the perpetualinventory method with an annual depreciation of 4%.

There is a need to have comprehensive analysis to find out the“exact” relationship between electricity consumption andeconomic growth. We are interested to know the long-run andshort-run relationship as well as the Granger causality among thefour variables in the system. Each of the tests (unit root, cointe-gration and Granger causality) has their own function and impo-tency. Since we have four variables in the system, the commoneconometric practice is to estimate the model in a VAR/VECMframework.

It is important to understand the interaction between electricityconsumption and income which is the key of successful environ-mental and growth policy. For example, if the causality findingsupports the growth hypothesis, restriction of electricityconsumption may adversely affect the process of economic growth.On the other hand, if there is evidence of unidirectional causalityrunning from income to electricity consumption, environmentalpolicy to conserve electricity consumption may have less or noimpact on the economic growth.

3.1. Saikkonen and Lütkepohl structural break unit root test

The standard unit root tests such as ADF and PP may provideinefficient and biased results when the shift is prevailed in the timeseries. To circumvent this problem, we use the test proposed bySaikkonen and Lutkepohl [56] and Lanne et al. [57]. The equation iswritten as follows:

y ¼ m0 þ m1t þ ftðqÞ0gþ εt (4)

where ftðqÞ0g indicates the shift function while q and g areconsidered as unidentified vectors, εt is generated by anARðpÞprocess. A simple shift dummy variable with shift date TB is used on

the basis of exponential distribution function. This function i.e. f 0t ¼

d1t

�0; t < TB

1; � TB does not seem to entail any parameter q in the

shift term ftðqÞ0g where g is a scalar parameter. We follow Lanneet al. [57] to choose the structural breaks exogenously which allowsus to apply ADF-type test to check the stationarity properties of theseries. Saikkonen and Lutkepohl [52] and Lanne et al. [53] alsosuggested of using large autoregressive in finding break date tominimize the generalized least square error of the objectivefunction.

3.2. ARDL bounds testing approach to cointegration

We apply the ARDL bounds testing approach to cointegrationadvanced by Pesaran et al. [58] to examine the long-run relation-ship between the variables. There are three advantages in using thisapproach. First, the relationship in long-run and short-run can beestimated simultaneously. Second, it can be employed even thoughthe variables have a mixed integration order. Third, it has betterproperties for small sample data sets.

M. Shahbaz, H.H. Lean / Energy 39 (2012) 146e153 149

The unrestricted error correction model (UECM) below is usedfor the estimation.

DlnYt ¼ fy0 þ 2yT þ py1lnYt�1 þ py2lnECt�1 þ py3lnKt�1 þ py4lnLt�1

þ Ppi¼1

lyiDlnYt�i þ Pqj¼0

gyiDlnECt�j þ Pri¼0

ayiDlnKt�i þPsi¼0

byiDlnLt�i þ ε1t(5)

DlnECt ¼ fec0 þ 2ecT þ pec1lnYt�1 þ pec2lnECt�1 þ pec3lnKt�1 þ pec4lnLt�1

þ Ppi¼1

leciDlnECt�i þ Pqj¼0

geciDlnYt�j þ Pri¼0

aeciDlnKt�i þPsi¼0

beciDlnLt�i þ ε2t(6)

DlnKt ¼ fk0 þ 2kT þ pk1lnYt�1 þ pk2lnECt�1 þ pk3lnKt�1 þ pk4lnLt�1

þ Ppi¼1

lkiDlnKt�i þ Pqj¼0

gkiDlnYt�j þ Pri¼0

akiDlnECt�i þPsi¼0

bkiDlnLt�i þ ε3t(7)

DlnLt ¼ fl0 þ 2lT þ pl1lnECt�1 þ pl2lnYt�1 þ pl3lnKt�1 þ pl4lnLt�1

þ Ppi¼1

lliDlnLt�i þ Pqj¼0

gliDlnYt�j þPri¼0

aliDlnECt�i þPsi¼0

bliDlnKt�i þ ε4t(8)

Table 1Saikkonen and Lütkepohl unit root test results.

where D is the first difference operator, 4 is the constant, ps are thelong-run coefficients; l;g;a; b represent the short-run dynamicsand εt is the error term which is assumed to be white noise. Thetime trend is indicated by T. Akaike Information Criteria (AIC) isused to select the optimal lag length.

The null hypothesis of no cointegration isH0 : py1 ¼ py2 ¼ py3 ¼ py4 ¼ 0 against the alternative hypoth-esis of cointegration Ha : py1spy2spy3spy4s0 for Equation (5).The same hypotheses can be derived for Equations (6)e(8). Thecomputed F-statistics are FY ðY=EC;K; LÞ, FECðEC=Y ;K; LÞ,FKðK=Y; EC; LÞ and FLðL=Y; EC;KÞ for Equations (5)e(8) respectively.If the computed F-statistic is more than the upper critical bound,then the null hypothesis is rejected and there is cointegrationbetween the variables. The hypothesis of no cointegration can beconcluded if the computed F-statistic is less than the lower criticalbound. Nevertheless, the results will be inconclusive if the calcu-lated F-statistic is between the lower and upper critical bounds.4

We use critical bounds generated by Turner [59] which are moresuitable for small samples as compared to the Pesaran et al. [58]study. In addition, to examine the stability of the ARDL boundstesting approach to cointegration, stability tests namely CUSUMand CUSUMSQ are applied.

The long-run relationship between economic growth and theother variables can be shown as follows, given that the variables arecointegrated:

lnYt ¼ F0 þ F1lnECt þ F2lnKt þ F3lnLt þ mt (9)

whereF0 ¼ fy0=py1;F1 ¼ �py2=py1;F2 ¼ �py3=py1;F3 ¼ �py4=py1and mt is the iid error term.

Variables Break date Statistics (k)

lnYt 1992 �0.5223(0)DlnYt 1992 �3.8819*(0)lnECt 1992 �1.3284(0)DlnECt 1992 �4.6414***(0)lnLt 1997 0.5178 (1)DLt 1997 �3.2234**(0)lnKt 2004 �0.9813(0)

3.3. VECM Granger causality

Morley [61] pointed out that if there is long-run relationshipbetween the variables, there must be Granger causality. Weemploy VECM Granger causality to detect the direction ofcausality between electricity consumption and economic growth

DlnKt 2004 �3.6862***(

Note: (1) ***, ** and * denote the significance at 1%, 5% and 10% level respectivelyk indicates lag order, the lag selection is based on AIC. (3) Critical vaare �3.55, �3.03 and �2.76 at 1%, 5%, and 10% respectively from Lanne et al. [5

4 Error correction method is appropriate and reliable to investigate the cointe-gration (Bannerjee et al. [60]).

in

the presence of capital and labor. The direction of causalitybetween the variables provides a clearer picture to the policymakers in formulating the electricity efficient economic policies.Given the existence of long-run relationship among variables, anerror correction term is added in the framework of VECM asfollows:

ð1� LÞ

2664lnYtlnECtlnKtlnLt

3775 ¼

2664a1a2a3a4

3775þ

Xpi¼1

ð1� LÞ

2664b11ib12ib13ib14ic21ic22ic23ic24id31id32id33id34ie41ie42ie43ie44i

37752664lnYtlnECtlnKtlnLt

3775

þ

2664qw

fr

3775½ECTt�1�þ

2664ε1tε2tε3tε4t

3775 (10)

where ð1� LÞ is the difference operator, ECTt�1 is the lagged errorcorrection term and ε1t ;ε2t ;ε3t and ε4t are the iid error terms. Ifthere is significant relationship in the first p difference of thevariables, it will show the short-run causal relationship through thesignificance of F-statistics. A significant coefficient of ECTt�1 via itst-statistic shows the long-run causality. For example, b12;is0ciindicates that electricity consumption Granger causes economicgrowth in the short-run. The joint short- and long-run Grangercausality is investigated by the significance of joint c2-statistic on

1)

. (2)lues7].

Table 2Results of bounds testing to cointegration.

Estimated model FY ðY=EC;K; LÞ FECðEC=Y ;K; LÞ FK ðK=Y; EC; LÞ FLðL=Y;EC;KÞOptimal lag

length(2, 1, 1, 1) (2, 2, 1, 2) (2, 2, 2, 2) (2, 2, 2, 2)

F-statistics 6.4095* 12.2451*** 8.1691** 4.3891Critical values# 1 percent

level5 percentlevel

10 percentlevel

Lower bounds 7.397 5.296 4.401Upper bounds 8.926 6.504 5.462Diagnostic testsR2 0.6249 0.8707 0.7271 0.7426Adjusted-R2 0.3486 0.7686 0.4371 0.4691F-statistics 2.2617** 8.5305* 2.5078** 2.7157**Durbin-Watson 2.3541 1.2854 2.1518 2.2343

Note: *, ** and *** denote the significance at 10%, 5% and 1% level respectively. Theoptimal lag structure is determined by AIC. #Critical values bounds are from Turner[59].

Table 3Results of Johansen cointegration Test.

Hypothesis Trace statistic Maximum eigen value

R ¼ 0 125.1119*** 75.4751***R � 1 49.6367** 27.8837**R � 2 21.7530 12.1759R � 3 9.5770 9.5617R � 4 0.0153 0.0153

Note: *** and ** show the significance at 1% and 5% level respectively.

1.4

1.5

1.6

1.7

1.8

1.9

2.0

1975 1980 1985 1990 1995 2000 2005 2010

Trend in Energy Intensity

Year

Fig. 1. Trends in energy intensity.

-15

-10

-5

0

5

10

15

92 94 96 98 00 02 04 06 08

CUSUM 5% Significance

Fig. 2. Plot of cumulative sum of recursive residuals. The straight lines representcritical bounds at 5% significance level.

M. Shahbaz, H.H. Lean / Energy 39 (2012) 146e153150

the lagged error correction term and first difference lagged con-cerned independent variable.

4. Empirical results

The results of Saikkonen and Lutkepohl [56] unit root test arereported in Table 1. We use an impulse dummy to detect thestructural break for all variables. We find that real GDP per capita isstationary at first, with a difference and the presence of a structuralbreak in 1992. That date is consistent with the implementation ofa structural adjustment program in the country. The same inferencecan be drawn for electricity consumption. Nevertheless, labor and

Table 4Long and short-runs results.

Dependent variable: lnYt

Long-run resultsVariable Coefficient Std. error t-statisticConstant 6.1316 0.2881 21.2768***lnECt 0.3142 0.0272 11.5364***lnKt 0.1191 0.0322 3.6999***lnLt 0.2984 0.0453 6.5809***Short-run resultsVariable Coefficient Std. error t-statisticConstant 0.0074 0.0056 1.3061lnECt 0.2575 0.0690 3.7281***lnKt 0.1279 0.0446 2.8635***lnLt 0.1108 0.1023 1.0827ECMt�1 �0.5699 0.1943 �2.9331***R2 0.4241Adj� R2 0.3498F-statistic 5.7082***Diagnostic Test F-statistic Prob. valuec2NORMAL 1.4342 0.4881c2SERIAL 0.6601 0.5248c2ARCH 0.3716 0.5464c2WHITE 1.5649 0.2092c2REMSAY 1.1272 0.2971

Note: *** shows the significance at 1% level.

capital are also stationary at first, with differences and structuralbreaks occurring in 1997 and 2004 respectively.

Selection of the appropriate lag length is necessary for the ARDLbounds testing approach to cointegration because the calculation ofthe F-statistic is sensitive to any lag order. Several selection criteriahave been considered and the appropriate lag length is 2 based onAIC. Lütkepohl [62] pointed out that AIC is superior for smallsample. Results of ARDL bounds test in Table 2 suggest that thehypothesis of cointegration is accepted when Yt , ECt and Kt asdependent variables respectively. This empirical evidence confirmsthe long-run relationship between the variables in Pakistan.

We also conduct Johansen and Juselius [63] cointegrationapproach to check the robustness of a long-run relationship. Resultsin Table 3 confirm that the long-run relationship between thevariables is valid and robust.

The long-run coefficients in Table 4 reveal the significant posi-tive effect of electricity consumption on economic growth. It isnoted that a 1 percent increase in electricity consumption willstimulate economic growth by 0.3 percent. There is also a positiveeffect of capital and labor on economic growth and they arestatistically significant at 1% level of significance. These resultsimply that capital and labor together with electricity are theimportant factors of production in Pakistan. Our findings are

-0.4

0.0

0.4

0.8

1.2

1.6

92 94 96 98 00 02 04 06 08CUSUM of Squares 5% Significance

Fig. 3. Plot of cumulative sum of squares of recursive residuals. The straight linesrepresent critical bounds at 5% significance level.

Table 5Results of Granger causality.

lnYt lnECt lnKt lnLt ECTt�1 Joint (short- and long-run)

DlnYt ;ECTt�1 DlnECt ;ECTt�1 DlnKt ;ECTt�1 DlnLt ;ECTt�1

lnYt e 4.0832**[0.0282]

5.4844**[0.0100]

0.3234[0.7264]

�0.7668***[�3.4667]

e 5.8877***[0.0032]

6.0810***[0.0027]

4.0940**[0.0162]

lnECt 5.9449***[0.0073]

e 2.0233[0.1518]

0.8659[0.4320]

�0.5086**[�2.6608]

8.6887***[0.0003]

e 2.7546*[0.0619]

3.5786**[0.0268]

lnKt 3.3018*[0.0521]

2.0848[0.1439]

e 0.0327[0.9679]

�0.3501**[�2.7571]

4.6249***[0.0098]

3.4314**[0.0310]

e 2.7911*[0.0596]

lnLt 1.2149[0.3124]

1.4189[0.2594]

2.5211*[0.0991]

e 0.0952[0.8586]

1.1879[0.3203]

1.2688[0.3049]

1.9239[0.1495]

e

Note: The asterisks ***, ** and * denote the significance at 1, 5 and 10 per cent respectively.

M. Shahbaz, H.H. Lean / Energy 39 (2012) 146e153 151

comparable with Yuan et al. [64] for China and Erbaykal [65] forTurkey.

The short-run coefficients also show that electricity consump-tion and capital have a significant positive impact on economicgrowth. Although labor participation has positive and significantimpact on economic growth in the long-run, but in the short-run,the impact of labor participation rate is positive but statisticallyinsignificant. This result implies that labor takes more time to showits contribution to the economic growth may be because of thelearning process. We also note that the coefficient of lagged errorcorrection term ðECMt�1Þ is negative and statistically significant at1% level of significance. The significance of coefficient of ECMt�1corroborates the established relationship among the variables ofinterest. Moreover, the negative sign implies that deviation inshort-run towards long-run is corrected by 57% from the previousperiod to the current period.

The CUSUM and CUSUMSQ tests are applied to examine thestability of long-run parameters. Both graphs of CUSUM andCUSUMSQ statistics are plotted in Fig. 2 and Fig. 3 respectively. Theplotted data points are within the critical bounds implying that thelong-run coefficients are stable.

4.1. VECM Granger causality analysis

The results of Granger causality are reported in Table 5. Since thevariables are cointegrated, the direction of causality can be dividedinto short- and long-run causations. Beginning with the long-runcausality, the coefficient of ECTt�1 is having a negative sign andstatistically significant in all equations except when labor acts asthe dependent variable. This implies that bi-directional causalitybetween economic growth, electricity consumption and capital isfound in the long-run period. In addition, if the system is exposed toa shock, it will converge to the long-run equilibrium at a relativelyhigh speed for economic growth (77%) and electricity consumption(51%).

In the short-run, there is bi-directional causality between elec-tricity consumption and economic growth. This empirical evidenceprovides support to the findings of Yoo [43] for Korea, Odhiambo[47] for South Africa and Lorde et al. [49] for Barbados. However,our results show contradictions with the evidences by Aqeel andButt [6], Zahid [7] and Jamil and Ahmad [8] for Pakistan. Thiscontradiction can be argued upon with a plausible view that thosefindings are biased or have been skewed due to their exclusion ofpotentially important variables such as capital and labor. There isalso bi-directional causality between capital and economic growthwhich implies that both capital and economic growth are impor-tant if not vital complements for effective consideration. Thisfinding is consistent with Arbay and Batool [66].

Results of joint short- and long-run causalities (Nasir and Reh-man [67]) are also pasted in Table 5. We find significant combined

short- and long-run effects in all equations except the labor equa-tion. These results imply that there exists “strong bi-directionalGranger causality” (Oh and Lee [68]) which is to be foundbetween electricity consumption and capital, and again once morebetween electricity consumption and economic growth in Pakistan.

5. Conclusion

This paper revisits the dynamics relationship between elec-tricity consumption and economic growth in Pakistan. The empir-ical evidence indicates that electricity consumption, economicgrowth, capital and labor are in the long-run equilibrium. We alsofind that electricity consumption, capital and labor have positiveand significant impact on economic growth. Bi-directional causalrelationship between electricity consumption and economicgrowth or feedback hypothesis exist in Pakistan for both the short-run and long-run periods. Bi-directional causal relation is alsofound between capital and economic growth.

Fig. 1 shows the decline trend of energy intensity over thesample period. This decline in energy intensity is due to theadoption of energy efficient technology in various sources of energyin Pakistan and shift of economic activity. Furthermore, adoption ofautonomous energy efficient techniques also plays an importantrole to decline energy intensity.

Thereforewe conclude that electricity conservation policies mayinversely affect the rate of economic growth and in turn, causea decline in economic growth andwill in turn lower the demand forelectricity. This fact suggests that the Government of Pakistan mustchange their policy focus to support research and developmentexpenditures to explore new sources of energy in order to meet therising demand for electricity and power; and adopt more advancedtechnology to produce and save energy. The adoption of advancedtechnology will not only prevent environmental degradation butalso sustain economic development in the country. Additionally,alternative energies such as solar power, hydro power, and windpower should be seriously considered because these alternativeenergy production methods are environmentally friendlycompared to the current fossil fuel powered productioninfrastructure.

Our model has the potential to investigate the relationshipbetween electricity consumption and economic growth byincluding other variables such as: electricity prices and exports asindicated by Lean and Smyth [46]; financial development andexchange rate mentioned by Karanfil [69] and also internationaltrade which was suggested by Halicioglu [70] and Narayan andSmyth [22]. The relationship between electricity consumption atdisaggregated levels and economic growth could be explored suchas in the case of Pakistan and South Asian Association for RegionalCorporation countries which had been conducted by Payne [71] inthe US. Analysis on disaggregated electricity consumption and

M. Shahbaz, H.H. Lean / Energy 39 (2012) 146e153152

economic growth will be more useful for policy makers to formu-late a comprehensive policy with a view towards saving energy andreducing environmental degradation.

The growth pattern of four provinces of Pakistan is different. Forfuture, study can be focus on the provincial level to detect thedirectional of causal relationship between electricity consumptionand economic growth which will provide a help to policy makers informulating comprehensive energy policy to sustain economicgrowth at provincial level.

Acknowledgment

The authors would like to thank Esmond Byrne for his excellentproofreading.

References

[1] Payne JE. Survey of the international evidence on the causal relationshipbetween energy consumption and growth. Journal of Economic Studies 2010;37:53e95.

[2] Ferguson R, Wilkinson W, Hill R. Electricity use and economic development.Energy Policy 2000;28:923e34.

[3] Pao H. Forecast of electricity consumption and economic growth in Taiwan bystate space modeling. Energy 2009;34:1779e91.

[4] Siddiqui R, Jalil HH, Nasir M, Khalid M, Malik WS. The cost of unserved energy:evidence from the industrial sector of Pakistan; 2011. PIDEWorking Paper No.75.

[5] Khan AM, Ahmed U. Energy demand in Pakistan: a disaggregate analysis. ThePakistan Development Review 2009;47:347e455.

[6] Aqeel A, ButtMS. The Relationship between energy consumption and economicgrowth in Pakistan. Asia-Pacific Development Journal 2001;8:101e10.

[7] ZahidA.EnergyeGDPrelationship:acausalanalysis for thefivecountriesofSouthAsia. Applied Econometrics and International Development 2008;1:167e80.

[8] Jamil F, Ahmad E. The relationship between electricity consumption, elec-tricity prices and GDP in Pakistan. Energy Policy 2010;38:6016e25.

[9] Lütkepohl H. Non-causality due to omitted variables. Journal of Econometrics1982;19:367e78.

[10] Bartleet M, Rukmani G. Energy consumption and economic growth in NewZealand: results of trivariate and multivariate models. Energy Policy 2010;38:3508e17.

[11] Kraft J, Kraft A. On the relationship between energy and GDP. Journal ofEnergy Development 1978;3:401e3.

[12] Tsani SZ. Energy consumption and economic growth: a causality analysis forGreece. Energy Economics 2010;32:582e90.

[13] Shahbaz M, Tang CF, Shabbir MS. Electricity consumption and economicgrowth nexus in Portugal using cointegration and causality approaches.Energy Policy 2011;39:3529e36.

[14] Kouakou AK. Economic growth and electricity consumption in Cote d’Ivoire:evidence from time series analysis. Energy Policy 2011;39:3638e44.

[15] Ozturk A. A literature survey on energy-growth nexus. Energy Policy 2010;38:340e9.

[16] Yoo S. The causal relationship between electricity consumption and economicgrowth in ASEAN countries. Energy Policy 2006;34:3573e82.

[17] Wolde-Rufael Y. Electricity consumption and economic growth: a time seriesexperience for 17 African countries. Energy Policy 2006;34:1106e14.

[18] Chen ST, Kuo HI, Chen CC. The relationship between GDP and electricityconsumption in 10 Asian countries. Energy Policy 2007;35:2611e21.

[19] Squalli J. Electricity consumption and economic growth: bounds and causalityanalyses for OPEC members. Energy Economics 2007;29:1192e205.

[20] Narayan PK, Prasad A. Electricity consumption-real GDP causality nexus:evidence from a bootstrapped causality test for 30 OECD countries. EnergyPolicy 2008;36:910e8.

[21] Alinsato AS. Electricity consumption and GDP in an electricity community:evidence from bound testing cointegration and granger-causality tests.Munich Personal Research Archive; 2009. Working Paper No.20816.

[22] Narayan PK, Smyth R. Multivariate granger-causality between electricityconsumption, exports, and GDP: evidence from a panel of middle Easterncountries. Energy Policy 2009;37:229e36.

[23] Yoo S, Kwak S. Electricity consumption and economic growth in seven SouthAmerican countries. Energy Policy 2010;38:180e8.

[24] Payne J. A survey of the electricity consumption-growth literature. AppliedEnergy 2010;87:3723e31.

[25] Acaravci A, Ozturk I. Electricity consumption-growth nexus: evidence frompanel data for transition countries. Energy Economic 2010;32:604e8.

[26] Ghosh S. Electricity consumption and economic growth in India. Energy Policy2002;30:125e9.

[27] Jumbe CBL. Cointegration and causality between electricity consumption andGDP: empirical evidence from Malawi. Energy Economics 2004;26:61e8.

[28] Narayan PK, Smyth R. Electricity consumption, employment and real incomein Australia: evidence from multivariate Granger causality tests. Energy Policy2005;33:1109e16.

[29] Yoo SH, Kim Y. Electricity generation and economic growth in Indonesia.Energy 2006;31:2890e9.

[30] Mozumder P, Marathe A. Causality relationship between electricityconsumption and GDP in Bangladesh. Energy Policy 2007;35:395e402.

[31] Halicioglu F. Residential electricity demand dynamics in Turkey. EnergyEconomics 2007;29:199e210.

[32] Aktas C, Yilmaz V. Causality between electricity consumption and economicgrowth in Turkey. ZKÜ Sosyal Bilimler Dergisi 2008;4:45e54.

[33] Ciarreta A, Zarraga A. Electricity consumption and economic growth in Spain.Applied Economics Letters 2010;14:1417e21.

[34] Ramcharran H. Electricity consumption and economic growth in Jamaica.Energy Economics 1990;12:65e70.

[35] Shiu A, Lam PL. Electricity consumption and economic growth in China.Energy Policy 2004;32:47e54.

[36] Yuan J, Zhao C, Yu S, Hu Z. Electricity consumption and economic growth inChina: cointegration and co-feature analysis. Energy Economics 2007;29:1179e91.

[37] Altinay G, Karagol E. Electricity consumption and economic growth: evidencefrom Turkey. Energy Economics 2005;27:849e56.

[38] Ho C, Sui K. A dynamic equilibrium of electricity consumption and GDP inHong Kong: an empirical investigation. Energy Policy 2007;35:2507e13.

[39] Chandran VGR, Sharma S, Madhavan K. Electricity consumptionegrowthnexus: the case of Malaysia. Energy Policy 2009;38:606e12.

[40] Abosedra S, Dah A, Ghosh S. Electricity consumption and economic growth:the case of Lebanon. Applied Energy 2009;86:429e32.

[41] Akinlo AE. Electricity consumption and economic growth in Nigeria: evidencefrom cointegration and co-feature analysis. Journal of Policy Modeling 2009;31:681e93.

[42] Acaravici A. Structural breaks, electricity consumption and economic growth:evidence from Turkey. Journal for Economic Forecasting 2010;2:140e54.

[43] Yoo SH. Electricity consumption and economic growth: evidence from Korea.Energy Policy 2005;33:1627e32.

[44] Zachariadis T, Pashourtidou N. An empirical analysis of electricity consump-tion in Cyprus. Energy Economics 2007;29:183e98.

[45] Tang CFA. Re-examination of the relationship between electricityconsumption and economic growth in Malaysia. Energy Policy 2008;36:3077e85.

[46] Lean HH, Smyth R. Multivariate granger causality between electricity gener-ation, exports, prices and GDP in Malaysia. Energy 2010;35:3640e8.

[47] Odhiambo NM. Electricity consumption and economic growth in South Africa:a trivariate causality test. Energy Economics 2009;31:635e40.

[48] Ouédraogo M. Electricity consumption and economic growth in Burkina Faso:a cointegration analysis. Energy Economics 2010;3:524e31.

[49] Lorde T, Waithe K, Francis B. The importance of electrical energy for economicgrowth in Barbados. Energy Economics 2010;32:1411e20.

[50] Murry DA, Nan GD. A definition of the gross domestic product-electrificationinterrelationship. Journal of Energy and Development 1996;19:275e83.

[51] Asafu-ADjaye J. The relationship between energy consumption, energy pricesand economic growth: time series evidence from Asian developing countries.Energy Economics 2000;22:615e25.

[52] Yusof NYM, Latif NWA. Causality between electricity consumption andeconomic growth in Malaysia. Working paper, Department of Economics &Entrepreneur Development, College of Business Management & Accounting,Universiti Tenaga Nasional, Malaysia; 2007.

[53] Alam S, Butt MS. Causality between energy consumption and economicgrowth in Pakistan: an application of cointegration and error correctionmodeling techniques. Pacific and Asian Journal of Energy 2002;12:151e65.

[54] Qazi MAH, Riaz S. Causality between energy consumption and economicgrowth: the case of Pakistan. The Lahore Journal of Economics 2008;13:45e58.

[55] Khan AM, Qayyum A. Dynamic modeling of energy and growth in South Asia.The Pakistan Development Review 2007;46:481e98.

[56] Saikkonen P, Lütkepohl H. Testing for a unit root in a time series with a levelshift at unknown time. Econometric Theory 2002;18:313e48.

[57] Lanne M, Lutkepohl H, Saikkonen P. Comparison of unit root tests for timeseries with level shifts. Journal of Time Series Analysis 2002;27:663e85.

[58] Pesaran MH, Shin Y, Smith RJ. Bounds testing approaches to the analysis oflevel relationships. Journal of Applied Econometrics 2001;16:289e326.

[59] Turner P. Response surfaces for an F-test for cointegration. Applied EconomicsLetters 2006;13:479e82.

[60] Banerjee A, Dolado JJ, Mestre R. Error-correction mechanism tests for coin-tegration in a single-equation framework. Journal of Time Series Analysis1998;19:267e83.

[61] Morley B. Causality between economic growth and migration: an ARDLbounds testing approach. Economics Letters 2006;90:72e6.

[62] Lütkepohl H. New introduction to multiple time series analysis. Germany:Springer-Verlag; 2005.

[63] Johansen S, Juselius K. Maximum likelihood estimation and inference oncointegration with applications to the demand for money. Oxford Bulletin ofEconomics and Statistics 1990;52:169e210.

[64] Yuan J, Kang J, Zhao C, Hu Z. Energy consumption and economic growth:evidence from China at both aggregated and disaggregated levels. EnergyEconomics 2008;30:3077e94.

[65] Erbaykal E. Disaggregate energy consumption and economic growth:evidence from Turkey. International Research Journal of Finance andEconomics 2008;20:172e9.

M. Shahbaz, H.H. Lean / Energy 39 (2012) 146e153 153

[66] Arby MF, Batool I. Estimating quarterly gross fixed capital formation; 2007.SBP Working Paper Series No. 17, State Bank of Pakistan.

[67] Nasir M, Rehamn F. Environmental Kuznets curve for carbon emissions inPakistan: an empirical investigation energy policy, vol. 39; 2011. pp.1857e1864.

[68] Oh W, Lee K. Causal relationship between energy consumption and GDP: thecase of Korea 1970e1999. Energy Economics 2004;26:51e9.

[69] Karanfil F. How many times again will we examine the energyeincome nexususing a limited range of traditional econometric tools? Energy Policy 2009;37:1191e4.

[70] Halicioglu F. An econometric study of CO2 emissions, energy consumption,income and foreign trade in Turkey. Energy Policy 2009;37:1156e64.

[71] Payne J. On the dynamics of energy consumption and output in the US. EnergyPolicy 2009;86:575e7.