Energy consumption–economic growth nexus for Pakistan: Taming the untamed

Renewable and Sustainable Energy Reviews 52 (2015) 890–896

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews journal homepage: www.elsevier.com/locate/rser

Energy consumption–economic growth nexus for Pakistan: Taming the untamed Mumtaz Ahmed a,n,,1, Khalid Riaz a, Atif Maqbool Khan b, Salma Bibi c a

Department of Management Sciences, COMSATS Institute of Information Technology, Park Road Chak Shahzad, Islamabad 44000, Pakistan International Institute of Islamic Economics, International Islamic University, Islamabad, Pakistan c Economics Department, Preston University, Islamabad, Pakistan b

art ic l e i nf o

a b s t r a c t

Article history: Received 31 March 2014 Received in revised form 8 June 2015 Accepted 15 July 2015

A recent survey of energy-growth literature has highlighted the potential trade-off between bivariate models that suffer from omitted variable bias, and the danger of over-parameterization of multivariate models in the individual country setting (Narayan and Smyth [2]). This is a serious limitation when the interest is in drawing policy implications for specific countries with short times series of available data. The maximum entropy bootstrap approach was used to re-examine the nature of causal relationship between energy consumption and economic growth for Pakistan where the available time series data was only from 1971 to 2011. Unlike the techniques used in much of the earlier literature, this approach does not rely on asymptotic methods and, therefore, leads to robust inference even in small samples. Moreover, the approach can be applied in the presence of non-stationarity of any type, and structural breaks, without requiring data transformation for to achieving stationarity, and is not sensitive to specification errors such as those in lag length selection. The empirical findings, based on both the bivariate as well as the multivariate frameworks, supported the conservation hypothesis, implying the existence of a unidirectional causality from economic growth to energy consumption. & 2015 Elsevier Ltd. All rights reserved.

Keywords: Entropy Bootstrap Causality

Contents 1. 2. 3. 4.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Energy consumption and economic growth literature for Pakistan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Energy sector in Pakistan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Econometric methodology, data and empirical results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1. Bivariate analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2. Multivariate analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

n Corresponding author. Primary mailing address: Department of Economics, Uris Hall, Cornell University, Ithaca, New York, 14853, USA. Secondary mailing address: Department of Management Sciences, COMSATS Institute of Information Technology, Park Road Chak Shahzad, Islamabad 44000, Pakistan. Tel.: þ 1 607 379 7371, þ 92 314 77 888 66. E-mail addresses: [email protected], [email protected] (M. Ahmed), [email protected], [email protected] (K. Riaz), [email protected] (A. Maqbool Khan), [email protected] (S. Bibi). 1 Current address: Department of Economics, Uris Hall, Cornell University, Ithaca, New York, 14853, USA.

http://dx.doi.org/10.1016/j.rser.2015.07.063 1364-0321/& 2015 Elsevier Ltd. All rights reserved.

890 892 892 894 894 894 895 895 895

1. Introduction The seminal work of Kraft and Kraft [1] inspired a large body of literature that looked at the energy growth relationship (see Narayan and Smyth [2] for a survey). This literature focused on empirically testing four mutually exclusive hypotheses with important policy implications. The hypotheses are: (a) the conservation hypothesis that postulates unidirectional Granger causality running from GDP to energy; (b) the growth hypothesis, that

M. Ahmed et al. / Renewable and Sustainable Energy Reviews 52 (2015) 890–896

suggests the causality runs from energy to GDP; (c) the feedback hypothesis, that posits the existence of bidirectional Granger causality between energy and GDP; and (d) the neutrality hypothesis that postulates energy and GDP as being independent. Mehrar [3], and Coers and Sanders [4] identified several generations of studies that tested the above mentioned hypothesis. While the recent studies adopted more sophisticated methods, consensus has eluded researchers in terms of establishing the existence of long run relationship between energy and growth, and the direction of causality between them (see Ozturk [5]; Payne [6]; both cited in Narayan and Smyth [2]). There are several reasons for this lack of consensus. These include, the omitted variable bias in the bivariate studies leading to wrong conclusions regarding the direction of causality, as cautioned by Lütkepohl [7]; the inclusion in the energy-growth relationship of additional variables selected on an ad hoc basis, the existence of multiple structural breaks that affect the cointegration relationship between energy and other variables of interest (Narayan and Smyth [2]); and the use of asymptotic theory for testing for unit roots and cointegration in small samples, (see Yalta [8] and Yalta [9]). Some studies tried to address these limitations by using multivariate models where the additional non-energy variables were chosen in a theoretically consistent way. For example, Stern ([10,11]) employed the theoretical framework of a production function, where the output (GDP) depended on energy, and other inputs i.e. capital and labor. Other studies employed normalized production function with output and inputs normalized by labor or population (for example, see Liddle [12] and Narayan and Smyth [13]). According to Narayan and Smyth [2], many recent studies used augmented production function framework where the relationship between GDP and energy was augmented by a third or fourth variables such as urbanization (Liddle [12]; Liu and Xie [14]; Mishra et. al. [15], Sadorsky [16]; Wang [17]), indicators of financial development (Coban and Topcu [18]; Jalil and Feridun [19]; Sadorsky [20], Sadorsky [21], Sadorsky [22] and Shahbaz and Lean [23]), and measures of trade (Aissa [24], Farhani et. al. [25]; Lean and Smyth [26]; Narayan and Smyth [13], Sadorsky [21] and Sadorsky [27]).

891

While the studies employing the augmented production function framework attempted to deal with the problems due to omitted variables, the approach was no panacea for modeling the relationship between energy and growth. The multivariate framework could be a costly modeling choice, leading to overparametrization of the model and the loss of degrees of freedom, if the available time series of data were short. Many researchers tried to use panel data techniques to overcome difficulties posed by short time series. In particular, the studies published from 2007 and 2008 onwards – labeled as ‘fifth generation’ studies (Cooers and Sanders [4]) – employed panel VECM, included other nonenergy inputs, and estimated capital-energy complementarities (for example, see Apergis and Payne [28]; Coers and Sanders [4]; Liddle [12], Narayan and Smyth [13], Sadorsky [21], Sadorsky [27]). The panel data models are not the ideal methodological approach, however, if the interest is in drawing policy implications for the individual countries. The recent survey of literature on the energy-growth relationship by Narayan and Smyth [2] highlighted the trade-off that necessarily arises in these circumstances between using the bivariate model susceptible to omitted variable bias, and employing a multivariate approach with the associated model over-parameterization risk. This is clearly a research gap that needs to be addressed, especially in view of its relevance for formulating energy policy for individual countries. As mentioned above, another important reason for the lack of consensus about the energy-growth relationship is that the traditional studies used asymptotic methods for testing for possible unit root and cointegration in small samples. However, there is no guarantee that this approach would lead to correct inference in small samples (Yalta [8,9], and Zhou [29]). According to Narayan and Smyth [2], “when using data for single countries, a long time span is preferable. Stern and Enflo [30] and Vaona [31] set the gold standard in this regard, although in most cases 150 years of data will not be available.” In fact with the exception of a few countries, the available time series on energy consumption are rather short. This paper attempted to fill the two research gaps identified above. First, it used an approach suited for studying energygrowth relationship when the interest is in drawing policy conclusions for a specific country with short available time series of

Table 1 Studies on energy consumption (EC) and economic growth (Y) for Pakistan. No.

Authors

Time span

Direction of Granger causality

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18.

Javed et al. [36] Muhammad et al. [44] Akhmat and Zaman [37] Abbas and Choudhury [38] Ahmad et al. [49] Chaudhry et al. [43] Shahbaz and Lean [23] Shahbaz et al. [34] Shahbaz et al. [35] Shahbaz and Feridun [48] Bedi-uz-Zaman et al. [47] Kakar and Khilji [42] Jamil and Ahmad [46] Khan and Ahmad [41] Asghar [40] Hye and Riaz [33] Mushtaq et al. [39] Aqeel and Butt [45]

1971–2008 1971–2013 1975–2010 1972–2008 1973–2006 1972–2012 1972–2009 1972–2010 1972–2011 1971–2008 1972–2008 1980–2009 1960–2008 1972–2007 1971–2003 1971–2007 1972–2005 1955–1995

ec2Y EC-Y NEC2Y AEC2AY Y-EC EC-Y ec2Y NGC2Y EC2Y Y-ec Y-EC EC-Y Y-EC ec-Y, CC-Y CC-Y, Y-ec EC2Y ec-Y, Y-OC Y-EC, Y-PC, ec-Y, Y neutral NGC

Note: EC¼ energy consumption; Y ¼economic growth (real GDP); ec ¼ electricity consumption; PC ¼petroleum consumption; NEC ¼ nuclear energy consumption; AEC¼ agricultural electricity consumption; AY ¼ agricultural growth; NGC¼ natural gas consumption; OC ¼oil consumption; CC ¼ coal consumption a. EC2Y indicates a bidirectional causality between energy consumption and economic growth b. Y-EC represents a unidirectional causal relationship that runs from economic growth to energy consumption c. EC-Y shows a unidirectional causal link from energy consumption to economic growth Neutral means no causal relationship.

892

M. Ahmed et al. / Renewable and Sustainable Energy Reviews 52 (2015) 890–896

data. Second, it implemented a methodology that allows robust inference in small samples. To address both of these gaps, the Maximum Entropy Bootstrap method proposed by Vinod [32] was used. The method provides robust inference for unit roots and cointegration in small samples, even in the presence of nonstationarity and structural breaks. This paper aims at drawing energy policy implications for Pakistan by testing the competing hypothesis about energy-growth relationship mentioned above. As elaborated in the next section, Pakistan is chosen because the country continues to face a serious energy crisis. This crisis assumed macroeconomic proportions in 2008 when the severe energy price shock depleted the foreign exchange reserves, almost led to an economic meltdown and to the closure for several weeks of the main bourse of the country. Very few studies exist on Pakistan on this subject and, unfortunately, all are based on asymptotic methods which may not be appropriate in finite samples. The Maximum Entropy Bootstrap is an attractive technique for this purpose because of the limitations on the time span for which the data are available for Pakistan. The structure of rest of the paper is as follows. The next section provides a brief review of studies on energy–growth relationship for Pakistan. Section three gives an overview of the energy sector, and discusses important policy issues. The data, econometric methodology and the empirical results are provided in section four. The conclusions are presented in the final section.

2. Energy consumption and economic growth literature for Pakistan There are only a few studies on this subject for Pakistan. The details of these studies is provided in Table 1 showing mixed results regarding the causal direction between EC and EG. Some of the studies, considering different forms of energy (electricity, natural gas, nuclear energy, etc.), claim bidirectional causality between EC and EG (see for example, Shahbaz and Lean [23], Hye and Riaz [33], Shahbaz et al., [34,35], Javed et al., [36], Akhmat and Zaman [37], Abbas and Choudhury [38]). Some studies suggest one-way causation that goes from EC to EG (e.g., Mushtaq et al., [39], Asghar [40], Khan and Ahmad [41], Kakar and Khilji [42], Chaudhry et al., [43], Muhammad et al. [44] etc.). Other studies claim a one-way causation that runs from EG to EC. (e.g., Aqeel and Butt [45], Jamil and Ahmad [46], Bedi-uz-Zaman et al., [47], Shahbaz and Feridun [48], Ahmad et al. [49], etc.). There exists one study which showed no causal link between gas consumption and economic growth (Aqeel and Butt [45]). Like studies of energy-growth relationship elsewhere, consensus does not exist

among researchers as to the direction of causality in this relationship for Pakistan.

3. Energy sector in Pakistan The patterns of total energy consumption in Pakistan, as well as its breakdown by the fuel type during 2005–2006 to 2012–2013, are presented in Table 2. During this period, the energy demand increased from 33.9 million tons of oil equivalent (Mtoe) to 40.2 Mtoe. In 2012–2013, the industrial sector had the largest share in total energy consumption (35.5%), followed by transport (31.6%), and domestic sector (25.2%) (see Table 3). The main energy sources were oil and gas, which supplied 30.4% and 43.6%, respectively, of the total energy consumption. The share of electricity, the third major source of energy, was 15.6%. Pakistan has significant coal deposits but coal provided only9.1% of the total consumed energy. Over the period 2005–2013, the share of gas in the total energy consumption increased from 39.3% to over 43.6%, while the share of oil, most of which is largely imported, declined from 32% to 30.4%. This was largely due to conversion of many industrial units and power plants to gas, and of transportation vehicles to compressed natural gas (CNG), following the increases in the international prices of oil. The substitution of imported oil for locally available gas resulted in some cost saving but led to accelerated depletion of country's gas reserves. In absence of priority for development of the country's considerable hydel resources and coal reserves, this policy of demand management did not offer a long term solution to Pakistan's energy problems. Pakistan has not fully exploited its potential for renewable energy (Zaigham and Nayyer [50] provide a survey of renewable energy option for Pakistan; Malik and Sukhera [51] discuss constraints in development of renewable energy, and suggest ways of dealing with them). For instance, hydropower potential is estimated at 50,000 MW (Kazi [52]) but installed capacity for hydroelectric generation is only 6858 MW (Government of Pakistan [53]). According to estimates, micro hydropower can contribute 300 MW (Hassan [54]), which may be an attractive option for local power generation in the mountainous northern regions. Low-head, high discharge power generation is a viable option in the Punjab plains (Zaigham and Nayyer [50]). Moreover, Pakistan has high potential for solar energy. With an average solar radiation of 5–7 kWh/m2/day and 7.6 h per day of sunshine on average, the country's potential for solar energy has been estimated at 19 MJ m  2 (Shaikh et. al. [55]). Wind can be another potentially good source of energy in the country, especially in the coastal regions. Along the Sind coast, for instance, the recorded wind

Table 2 Energy consumption by source (Tons of oil equivalent). Sources: Table 1.2 in Pakistan Energy Year Book, 2011 [73]. Source\Year

FY 05–06

FY 06–07

FY 07–08

FY 08–09

FY 09–10

FY 10–11

FY 11–12a

FY 12–13a

Oilb

10,877,601 32.04% 13,325,251 39.25% 3,611,490 10.64% 5,505,555 16.22% 625,792 1.84% 33,945,689

10,575,330 29.37% 14,701,024 40.83% 4,149,041 11.52% 5,921,635 16.45% 658,225 1.83% 36,005,255

11,528,722 29.25% 15,881,990 40.30% 5,404,715 13.71% 5,977,697 15.17% 619,944 1.57% 39,410,068

10,842,614 29.03% 16,307,898 43.67% 3,893,001 10.42% 5,731,032 15.35% 569,995 1.53% 37,344,540

10,829,455 27.93% 17,024,933 43.91% 4,282,061 11.05% 6,054,921 15.62% 576,631 1.49% 38,768,001

11,252,938 28.97% 16,781,247 43.20% 4,025,380 10.36% 6,278,947 16.17% 503,272 1.30% 38,841,784

11,617,788 29.0% 17,618,199 44.0% 4,057,678 10.1% 6,251,421 15.6% 481,064 1.2% 40,026,149

12,219,941 30.4% 17,521,615 43.6% 3,661,193 9.1% 6,253,675 15.6% 528,417 1.3% 40,184,842

Gas

c

Coalc Electricityd LPG Total a

Denotes that figures are taken from Table 1.2 in Pakistan Energy Year Book, 2013 [74]. Excluding consumption for power generation. c Excluding consumption for power generation and feedstock. d @ 3412 Btu/kW h being the actual energy content of electricity. b

M. Ahmed et al. / Renewable and Sustainable Energy Reviews 52 (2015) 890–896

893

Table 3 Energy Consumption by Sector (Tons of Oil Equivalent). Sources: Table 1.5 in Pakistan Energy Year Book, 2011 [73]. Sector\year

FY 05–06

FY 06–07

FY 07–08

FY 08–09

FY 09–10

FY 10–11

FY 11–12a

FY 12–13a

Domestic

7,054,587 20.78% 1,247,992 3.68% 14,654,360 43.17% 732,699 2.16% 9,493,667 27.97% 762,384 2.25% 33,945,689

7,605,145 21.12% 1,377,247 3.83% 15,792,049 43.86% 767,266 2.13% 9,721,184 27.00% 742,364 2.06% 36,005,255

8,046,294 20.42% 1,455,527 3.69% 16,804,303 42.64% 803,837 2.04% 11,567,395 29.35% 735,712 1.87% 39,413,068

8,092,132 21.67% 1,459,817 3.91% 14,845,670 39.75% 789,008 2.11% 11,371,868 30.45% 786,045 2.10% 37,344,540

8,360,016 21.56% 1,530,154 3.95% 15,604,871 40.25% 849,595 2.19% 11,654,834 30.06% 768,531 1.98% 38,768,001

8,724,790 22.46% 1,521,171 3.92% 14,956,785 38.51% 772,930 1.99% 12,019,251 30.94% 846,857 2.18% 38,841,784

9,360,514 23.39% 1,585,498 3.96% 15,034,116 37.56% 720,393 1.80% 12,562,448 31.39% 763,183 1.91% 40,026,151

10,119,014 25.18% 1,644,845 4.09% 14,256,099 35.48% 659,986 1.64% 12,713,300 31.64% 791,598 1.97% 40,184,842

Commercial Industrial Agriculture Transport Other Govt. Total a

Denotes that figures are taken from Table 1.5 in Pakistan Energy Year Book, 2013 [74].

speed is 7–8 m/s at 80 m height. At Gharo Keti Bandar corridor alone, there is potential to generate 50,000 MW of wind energy (Qamar, [56]). This figure is estimated to be around 300,000 MW for the entire country (Mirza et al. [57]). Pakistan is facing a crippling energy crisis. The gap between supply and demand of electricity stands at 4500 to 5500 MW (MW). This has resulted in ‘load-shedding’ or power outages [58]. The industrial sector and domestic consumers are very badly hit. Among the domestic consumers, the rural households face a longer duration of power outages compared to their urban counterparts. Part of the problem lies in the inefficient power transmission and distribution system. The recorded system losses range from 23% to 25%, which are not only due to poor infrastructure but also due to mismanagement and outright theft. The cost of electricity generation is Pakistani rupee (PKR) 12.0 per unit but the cost of electricity charged to the final consumer is PKR 15.60 per unit. The extra cost over and above the generation costs includes transmission losses as well as the cost of inefficiencies and mismanagement mentioned above. There is also a financial dimension to the energy crisis. The electric power in Pakistan is purchased by Central Power Purchasing Agency (CPPA) from the generating companies. Often, there is a cash shortfall at CPPA because: (i) the realized revenues by the electricity distribution companies are less than the cost of power supplied, or (ii) the distribution companies prefer to first meet their own cash flow needs before paying CPPA. This creates circular debt within the entire energy supply chain that results in shortage of fuel supply to public sector thermal generating companies and Independent Power Producers (IPPs). The estimated circular debt in June 2011 was PKR 537 billion. By June 2012, the circular debt rose to PKR 872 billion, which was nearly 4% of GDP [59]. The impact of circular debt is by no means limited to electric power sector alone. Many power plants run on furnace oil, which is imported by the state run entity, Pakistan State Oil (PSO). In January 2015, PSO's receivables from power companies increased so much that it was not able to import oil. The supply of gasoline to fuel stations was severely disrupted, resulting in long lines of vehicles at fuel stations (Shirani [60]). There appears to be a close relationship between the growth performance of Pakistan's economy and the evolution of energy use (see Fig. 1). During 2001–2007 the rate of GDP growth remained fairly strong at 5.4%, reaching 8.9% in the year 2005. However, in the aftermath of a macroeconomic crisis, the rate of GDP growth fell drastically to 0.36% in 2009. The total energy use, which was 39.4 Mtoe in 2008 came down sharply to 37.3 Mtoe – a

Y (left)

160000

EC (right)

90000

140000

80000

120000

70000 60000

100000

50000

80000

40000

60000

30000

40000

20000

20000

10000

0 1985

1990

1995

2000

2005

2010

0

2015

Year Fig. 1. Plot of real GDP (Y) and energy consumption (EC).

9.5% drop – in the following year. Clearly, the slowdown in economic activity caused energy consumption to decline. The energy-growth relationship in Pakistan is more complex, however. Apart from the direct linkages between energy and growth, the international oil price dynamics and their complex inter-relationships with the key macroeconomic variables, add new dimensions to this complexity. Pakistan is a manpower exporting country with large numbers of overseas Pakistanis employed in the oil-rich economies of Middle East. Higher oil prices increase employment opportunities in oil exporting countries, and increase the flow of remittances from Pakistanis working in the region. But severe oil price shocks have the potential to cause serious deterioration in the balance of payments, and precipitate a macroeconomic crisis. During the period (2001– 2007) high growth in Pakistan was accompanied by build-up of foreign exchange reserves on the back of higher remittances and capital inflows, despite rising oil prices. In 2008, when the international oil prices rose very sharply to unprecedented levels, the foreign exchange reserves declined drastically from 14 billion dollars to 7.2 billion dollars. The following year, the growth rate of GDP also fell sharply, as the central bank aggressively raised interest rates in an effort to control inflationary spiral. Given these complicated dynamics involving energy prices, key macroeconomic variables, and growth, judging the direction of causality between energy and growth is not a trivial exercise. The nature of energy-growth relationship, and particularly the question regarding the direction of causality, has important policy implications. If the direction of causality ran from energy to growth, for instance, energy supply shocks can impact growth

894

M. Ahmed et al. / Renewable and Sustainable Energy Reviews 52 (2015) 890–896

adversely. In those circumstances, the policies for ensuring uninterrupted energy supplies would be required. This has specific policy implications for Pakistan where in recent years, poor governance, inefficient power generation and distribution infrastructure, and a tariff differential subsidy (TDS) on electricity, has led to creation of huge circular debt,2 and choking up of energy supply chain. The resulting revenue inadequacy led many power producers to shut down temporarily, thus disrupting supplies to the commercial, industrial and domestic sectors (see Kessides [61] for an in-depth analysis of Pakistan's electricity crisis). Alternatively, if the causality ran from growth to energy, periods of high growth would create higher demand for energy, and balance of payments vulnerability, especially if the growth took place in period of rising oil prices. This would underscore the importance of policies for insulating the economy from energy price shocks, for example by reducing the dependence on imported fuels.

4. Econometric methodology, data and empirical results Most of the existing studies relied on the popular Granger Causality test proposed by Granger [62]. Though, this test provides some useful insights of the relation between the candidate variables regarding the time structure, but, the conclusions can be misleading if the candidate variables were either integrated or cointegrated. The condition becomes sever in finite samples due to its unusual asymptotic properties (Toda and Phillips [63]). Thus it is very difficult to get reliable inferences in small samples like the one used in this study. This inspired us to use the recently developed maximum entropy bootstrap approach due to Vinod [32]. According to Vinod [32,64] and Vinod and De-Lacalle [65], this approach promises to provide reliable and robust empirical analysis regarding causal nexus between the candidate variables. 4.1. Bivariate analysis To set the stage, the causality between energy consumption (EC) and economic growth (EG) is first examined in a bivariate framework using a standard bivariate vector autoregressive (VAR) model presented as a system of two Eqs. (1) and (2) below ln Y t ¼ c1 þ

p X

α1i ln Y t  i þ

i¼1

ln ECt ¼ c2 þ

p X i¼1

p X

β1i ln ECt  i þ ε1t

measured over time and no information is tempered. In addition, this approach satisfies the conditions of ergodic as well as central limit theorem (Vinod and De-Lacalle [65], Yalta and Caker [66]). For each ensemble, the VAR models in (1) and (2) were estimated, and ‘m’ estimates of each parameter were obtained. Next the (1  α)100% confidence intervals were constructed (α being the significance level) from these estimates, using high density regions (HDR) proposed by Hyndman [67]. Since the confidence intervals were constructed using the estimated values of the coefficients, so in order to test the null hypothesis that lnEC does not Granger cause lnY (lnY does not Granger cause lnEC), a check was made to see if the zero (value under the null hypothesis) fell inside the confidence limits for the parameter β1j (α1i). If this was the case, it was concluded that the null hypotheses could not be rejected. The results for the bivariate case using VAR (1) and VAR (2) models are presented in Table 4 along with respective Akaike Information Criterion (AIC) and Schwarz Bayesian Information Criterion (SBIC) values. The empirical results suggested that there existed a unidirectional causality from lnY to lnEC. The findings were the same regardless of the choice of VAR model (VAR (1) and VAR (2)). The reported findings are consistent with Jamil and Ahmad [46], Shahbaz and Feridun [48] for electricity consumption-growth nexus, Ahmad et al. [49], Mushtaq et al. [39] for oil consumption and Aqeel and Butt [45], Mehrara [3], Mehrara and Muasi [68] among many others. 4.2. Multivariate analysis The bivariate analysis was extended by using the theoretically consistent framework production function framework (Stern [10,11]) where capital (K) and labor (L) were included as two additional inputs in the production process. Gross Fixed Capital Formation (measured in US dollars at 2005 prices) was used as a proxy for the capital stock (K). The basic idea of introducing the multivariate framework was to avoid any bias due to omitted Table 4 Bivariate analysis. Variable

ð1Þ

i¼1

α2i ln Y t  i þ

p X

β2i ln ECt  i þ ε2t

ð2Þ

i¼1

where, the ci's (i¼ 1, 2) are the constant terms, and theεi 's (i¼1, 2) are the white noise error terms; ‘p’ is the maximum lag length used; lnY is the natural logarithm of real GDP (measured in US dollars at 2005 prices) as a proxy for economic growth; and lnEC is the natural logarithm of energy use (measured in kilo ton of oil equivalent) as a proxy for energy consumption. The annual time series data were obtained from the World Development Indicators (WDI) produced by World Bank. The plot of both series is provided in Fig. 1, showing an increasing trend and hence the nonstationary behavior of both series. The maximum entropy bootstrap approach resamples both data series (EC and EG) a fixed number of times (m ¼ 999) providing us ‘m’ different ensembles. In this approach each ensemble retains all the properties of the original data series

Intercept lnY(  1) lnEC(  1) lnY(  2) lnEC(  2) AIC SBIC

VAR(1): 95% confidence limits

VAR(2): 90% confidence limits

lnY

lnY

lnEC

lnEC

LCL

UCL

LCL

UCL

LCL

UCL

LCL

UCL

 2.549 0.342  0.087 – –  4.875  4.748

8.055 1.170 0.313 – –

 3.087 0.678  0.064 – –  4.998  4.871

1.051 1.046 0.263 – –

 0.374 0.603  0.274  0.333  0.235  4.973  4.760

3.797 1.201 0.388 0.253 0.373

 2.784 0.505  0.207  0.285  0.217  5.013  4.800

0.786 1.170 0.320 0.328 0.325

Note: LCL and UCL stand for lower and upper confidence limits, respectively. K (left)

30000

L (right)

70 60

25000

50

20000

40

15000

30

10000

20

5000 2

TDS refers to the cost of electricity generation and the tariff charged to the consumers. In the past, the Government of Pakistan has been reluctant to pass on the higher generation costs to the consumer. But the weak fiscal position of the government often resulted delays in paying TDS. This caused revenue inadequacy for power producing companies which could not buy fuel to operate their plants.

10

0 1985

1990

1995

2000

2005

2010

Year Fig. 2. Plot of total capital stock (K) and total labor force (L).

0 2015

M. Ahmed et al. / Renewable and Sustainable Energy Reviews 52 (2015) 890–896

895

where, lnK and lnL, respectively, are the natural logarithms of total capital stock (K) and total labor force (L). The results were similar to those obtained in the case of bivariate models. The results of multivariate case are provided in Table 5 along with respective Akaike Information Criterion (AIC) and Schwarz Bayesian Information Criterion (SBIC) values. The results suggested that there existed unidirectional causation from real GDP to energy consumption. Compared to the bivariate case, the results obtained from the multivariate model were reliable because they did not suffer from omitted variable bias.

energy-growth relationship worldwide, where the consensus is similarly lacking. The conflicting results of earlier research are partly attributable to the use of asymptotic methods that are known to perform poorly in small samples, and partly due to the problems associated with structural breaks, omitted variables, and ad hoc specifications. In this study, the Maximum Entropy Bootstrap technique was used, which is known to yield reliable and robust results in small samples in the presence of non-stationarity and possible structural breaks. The empirical results suggested that in the case of Pakistan and for the time period studied, unidirectional causality ran from economic growth to energy consumption. Thus the findings of this paper lent support to the conservation hypothesis. This has several policy implications. Note that the four hypotheses that were investigated in this paper are mutually exclusive. Therefore, the failure to reject the conservation hypothesis implied the rejection of the growth hypothesis. Hence it was concluded that for the period under study, the energy conservation efforts in Pakistan would not have led to a reduction in economic growth. This underscored the relevance of the paper's findings for formulating future energy policy. Accordingly, the policy recommendation for the government is to more proactively devise and implement the energy conservation programs in the future. Secondly, the results indicated that economic growth was the key driver of energy demand in Pakistan. This implies that if during periods of high growth in future, energy prices also increase sharply, the balance of payments could deteriorate significantly. For hedging against this vulnerability, the policy-makers in Pakistan should aim at reducing the growing dependence of the country on imported fuels. This could be done in a number of ways, including switching towards indigenous energy sources by more fully exploiting the considerable hydropower potential of the country, and by speeding up the development of proven local coal reserves. It must be noted that for the achievement of this objective, it would be counter-productive to distort relative prices between imported and domestic fuels in an attempt to shift demand to the latter. Such a distortion may provide some respite from the balance of payments difficulties in the short run, but the accelerated depletion of domestic energy resources would make the country even more dependent on imported fuels in the future, as the depletion of the sizeable gas reserves clearly demonstrates. The appropriate long-term policy response would be the development of local energy resources while continuing to align prices of various fuels to their respective international prices. Thirdly, the country should attach greater priority for developing non-traditional renewable resources. There is good potential for development of solar power in Sind and Baluchistan provinces. Wind energy can be harnessed along the coastal belt. In addition, there is potential for generating energy from biomass and solid waste, as indicated by Zaigham and Nayyer [50].

5. Conclusion

Acknowledgments

This paper investigated the energy-growth relationship with the objective of deriving policy implications for Pakistan, which is facing severe energy shortages for the past decade. The paper tested four competing energy-growth hypothesis. For the purpose of drawing policy implications for single countries from this relationship, the literature has clearly recognized the trade-off between bivariate models that could suffer from omitted variable bias, and the danger of over-parameterization of multivariate models due to short time series typically available for such countries (Narayan and Smyth [2]). With only short time series of data available, earlier studies on Pakistan reported conflicting results. This mirrors the findings of the larger body of literature on

We are thankful to two anonymous referees whose comments helped us in improving the quality of the paper. Any errors left are our own.

Table 5 Multivariate analysis. Variable

Intercept lnY(  1) lnEC(  1) lnK(  1) lnL(  1) lnY(  2) lnEC(  2) lnK(  2) lnL(  2) AIC SBIC

VAR(1): 95% confidence limits

VAR(2): 90% confidence limits

lnY

lnY

lnEC

lnEC

LCL

UCL

LCL

UCL

LCL

UCL

LCL

UCL

 3.544 0.567  0.202  0.146  0.247 – – – –  4.816  4.605

4.812 1.043 0.366 0.223 0.587 – – – –

 5.991 0.526  0.157  0.134  0.266 – – – –  4.942  4.731

1.541 1.059 0.320 0.178 0.496 – – – –

 3.532 0.458  0.315  0.135  0.489  0.358  0.311  0.164  0.449  4.957  4.573

4.928 1.141 0.411 0.203 0.604 0.262 0.393 0.193 0.675

 5.984 0.397  0.256  0.133  0.440  0.315  0.246  0.153  0.453  4.866  4.482

1.366 1.117 0.336 0.167 0.540 0.338 0.351 0.158 0.591

Note: LCL and UCL stand for lower and upper confidence limits, respectively.

variables that might arise in the bivariate case (Cheng [69]). This multivariate set up has been used extensively in the existing literature (see for example, Chien and Hu [70], Apergis and Payne [28], Wei and Hu [71], Apergis and Danuletiu [72], etc.). The annual data on these two additional inputs variables was also obtained from the World Development Indicators (WDI). The plot of both series is provided in Fig. 2. The multivariate specifications are provided below ln Y t ¼ c1 þ

q X

α1j ln Y t  j þ

j¼1

þ

q X

λ1j ln K t  j þ

q X

q X

q X

π 1j ln Lt  j þ u1t

ð3Þ

j¼1

λ2j ln K t  j þ

j¼1

q X

α2j ln Y t  j þ

j¼1

þ

β1j ln ECt  j

j¼1

j¼1

ln ECt ¼ c2 þ

q X

β2j ln ECt  j

j¼1 q X

π 2j ln Lt  j þ u2t

ð4Þ

j¼1

References [1] Kraft J, Kraft A. Relationship between energy and GNP. J Energy Dev 1978;3(2):401–3. [2] Narayan P, Smyth R. Applied econometrics and a decade of energy economics research. Monash University; 2014 Discussion Paper, 21/14, Unpublished Manuscript. [3] Mehrara M. Energy consumption and economic growth: the case of oil exporting countries. Energy Policy 2007;35(5):2939–45.

896

M. Ahmed et al. / Renewable and Sustainable Energy Reviews 52 (2015) 890–896

[4] Coers R, Sanders M. The energy GDP nexus: addressing an old question with new methods. Energy Econ 2013;36:708–15. [5] Ozturk I. A literature survey on energy–growth nexus. Energy Policy 2010;38(1): 340–9. [6] Payne JE. A survey of the electricity consumption-growth literature. Appl Energy 2010;87(3):723–31. [7] Lütkepohl H. Non-causality due to omitted variables. J Econom 1982;19:267–378. [8] Yalta TA. Analyzing energy consumption and GDP nexus using maximum entropy bootstrap: the case of Turkey. Energy Econ 2011;33(3):453–60. [9] Yalta AY. Revisiting the FDI-led growth hypothesis: the case of China. Econ Model 2013;31:335–43. [10] Stern DI. Energy use and economic growth in the USA: a multivariate approach. Energy Econ 1993;15:137–50. [11] Stern DI. A multivariate cointegration analysis of the role of energy in the U.S. macroeconomy. Energy Econ 2000;22:267–83. [12] Liddle B. The energy, economic growth, and urbanization nexus across development: evidence from heterogeneous panel estimates robust to crosssectional dependence. Energy J 2013;34:223–44. [13] Narayan PK, Smyth R. Energy consumption and real GDP in G7 countries: new evidence from panel cointegration with structural breaks. Energy Econ 2008;30(5):2331–41. [14] Liu Y, Xie Y. Asymmetric adjustment of the dynamic relationship between energy intensity and urbanization in China. Energy Econ 2013;36:43–54. [15] Mishra V, Sharma S, Smyth R. Are fluctuations in energy consumption per capita transitory? Evidence from a panel of Pacific Island countries Energy Policy 2009;37:2318–26. [16] Sadorsky P. Do urbanization and industrialization affect energy intensity in developing countries? Energy Econ 2013;37:52–9. [17] Wang Q. Effects of urbanization on energy consumption in China. Energy Policy 2014;65:332–9. [18] Coban S, Topcu M. The nexus between financial development and energy consumption in the EU: a dynamic panel data analysis. Energy Econ 2013;39:81–8. [19] Jalil A, Feridun M. The impact of growth, energy and financial development on the environment in China: a cointegration analysis. Energy Econ 2011;33:284–91. [20] Sadorsky P. The impact of financial development on energy consumption in emerging economies. Energy Policy 2010;38:25–8. [21] Sadorsky P. Financial development and energy consumption in Central and Eastern European frontier economies. Energy Policy 2011;39:999–1006. [22] Sadorsky P. Trade and energy consumption in the Middle East. Energy Econ 2011;33:739–49. [23] Shahbaz M, Lean HH. The dynamics of electricity consumption and economic growth: a revisit study of their causality in Pakistan. Energy 2012;39(1):146–53. [24] Aissa MSB, Jebli MB, Yousef SB. Output, renewable energy consumption and trade in Africa. Energy Policy 2014;66:11–8. [25] Farhani S, Shahbaz M, Arouri M, Teulon F. The role of natural gas consumption and trade in Tunisia's output. Energy Policy 2014;66:677–84. [26] Lean HH, Smyth R. On the dynamics of aggregate output, electricity consumption and exports in Malaysia: evidence from multivariate Granger causality tests. Appl Energy 2010;87:1963–71. [27] Sadorsky P. Energy consumption, output and trade in South America. Energy Econ 2012;34:475–88. [28] Apergis N, Payne JE. A global perspective on the renewable energy consumption-growth nexus. Energy Sources Part B Econ Plan Policy 2012;7(3):314–22. [29] Zhou S. Testing structural hypothesis on cointegration relations with small samples. Econ Inq 2007;38:629–40. [30] Stern DI, Enflo K. Causality between energy and output in the long-run. Energy Econ 2013;39:135–46. [31] Vaona A. The sclerosis of regional electricity intensities in Italy: an aggregate and sectoral analysis. Appl Energy 2013;104:880–9. [32] Vinod HD. Ranking mutual funds using unconventional utility theory and stochastic dominance. J Empir Financ 2004;11(3):353–77. [33] Hye QMA, Riaz S. Causality between energy consumption and economic growth: the case of Pakistan. Lahore J Econ 2008;13(2):45–58. [34] Shahbaz M, Zeshan M, Afza T. Is energy consumption effective to spur economic growth in Pakistan? New evidence from bounds test to level relationships and Granger causality tests Econ Model 2012;29(6):2310–9. [35] Shahbaz M, Lean HH, Farooq A. Natural gas consumption and economic growth in Pakistan. Renew Sustain Energy Rev 2013;18:87–94. [36] Javed AY, Javid M, Awan ZA. Electricity consumption and economic growth: evidence from Pakistan. Econ Bus Lett 2013;2(1):21–32. [37] Akhmat G, Zaman K. Nuclear energy consumption, commercial energy consumption and economic growth in South Asia: bootstrap panel causality test. Renew Sustain Energy Rev 2013;25:552–9. [38] Abbas F, Choudhury N. Electricity consumption-economic growth nexus: an aggregated and disaggregated causality analysis in India and Pakistan. J Policy Model 2013;35:538–53. [39] Mushtaq K, Abbas F, Abedullah M, Ghafoor A. Energy use for economic growth: cointegration and causality analysis from the agriculture sector of Pakistan. Pak Dev Rev 2007;46(4-II):1065.

[40] Asghar Z. Energy–GDP relationship: a causal analysis for the five countries of South Asia. Appl Econom Int Dev 2008;8(1):167–80. [41] Khan MA, Ahmad U. Energy demand in Pakistan: a disaggregate analysis. Pak Dev Rev 2008;47(4-II):437–55. [42] Kakar ZK, Khilji BA. Energy consumption and economic growth in Pakistan. J Int Acad Res 2011;11(1):33–6. [43] Chaudhry IS, Safdar N, Farooq F. Energy Consumption and economic growth: empirical evidence from Pakistan. Pak J Soc Sci 2012;32(2):371–82. [44] Muhammad SD, Usman M, Mujahid N, Lakhan GR. Nexus between energy consumption and economic growth: a case study of Pakistan. World Appl Sci J 2013;24(6):739–45. [45] Aqeel A, Butt MS. The relationship between energy consumption and economic growth in Pakistan. Asia–Pac Dev J 2001;8(2):101–10. [46] Jamil F, Ahmad E. The relationship between electricity consumption, electricity prices and GDP in Pakistan. Energy Policy 2010;38(10):6016–25. [47] Bedi-uz-Zaman MF, Ullah S. Sectoral oil consumption and economic growth in Pakistan: an ECM approach. Am J Sci Ind Res 2011;2(2):149–56. [48] Shahbaz M, Feridun M. Electricity consumption and economic growth empirical evidence from Pakistan. Qual Quant 2012;46(5):1583–99. [49] Ahmad N, Hayat MF, Hamad N. Energy consumption and economic growth: evidence from Pakistan. Aust J Bus Manag Res 2012;2(6):09–14. [50] Nayyer Alam Zaigham N.A. and Nayyar Z.A. Prospects of Renewable Energy Sources in Pakistan. In: renewable-energy technologies and sustainable development. Khan H. A., Qurashi M. M., Hussain T., Hayee I. (Eds.), COMSATS' series 4, 2005, 65–86. [51] Malik SN, Sukhera OR. Management of natural gas resources and search for alternative renewable energy resources: a case study of Pakistan. Renew Sustain Energy Rev 2012;16(2):1282–90. [52] Kazi AH. The Indus river; water, power & environment: in the Indus river. In: Meadows Azra, Meadows Peter S, editors. Karachi, Pakistan: Oxford University Press; 1999. p. 141–50. [53] Government of Pakistan. Pakistan economic survey 2013–14. Islamabad: Ministry of Finance; 2014. [54] Hassan MU. Development of renewable energy technologies in Pakistan – an overview: Pakistan council of renewable energy technologies. Minist Sci Technol 2002:1–3. [55] Shaikh PH, Shaikh F, Mirani M. Solar energy: topographical asset for Pakistan. Appl Sol Energy 2013;49(1):49–53. [56] Qamar ZC. An investigation on wind power potential of Gharo-Sindh, Pakistan. Pak J Meteorol 2009;6(11):1–11. [57] Mirza AI, Khan AN, Memon N. Development of benchmark wind speed for Gharo and Jhimpir, Pakistan. Renew Energy 2010;35(3):576–82. [58] Government of Pakistan. National power policy. Islamabad: Ministry of Water and Power; 2013. [59] Government of Pakistan. The Causes and Impacts of Power Sector Circular Debt in Pakistan. A study commissioned by Planning Commission of Pakistan; 2013b. [60] Shirani, S. Anatomy of the petrol crisis. The Dawn, 23 January 2015. 〈http:// www.dawn.com/news/1158756/anatomy-of-the-petrol-crisis〉; 2015 [accessed 24.01.15]. [61] Kessides IN. Chaos in power: Pakistan's electricity crisis. Energy Policy 2013;55:271–85. [62] Granger CWJ. Investigating causal relations by econometric models and crossspectral methods. Econometrica 1969;37(3):424–38. [63] Toda HY, Phillips PC. Vector autoregressions and causality. Econometrica 1993;61(6):1367–93. [64] Vinod HD. Hands-on intermediate econometrics using R. 1st ed. Singapore: World Scientific Publishing Ltd; 2008. [65] Vinod HD, de Lacalle JL. Maximum entropy bootstrap for time series: the meboot R package. J Stat Softw 2009;29(5):1–19. [66] Yalta TA, Cakar H. Energy consumption and economic growth in China: a reconciliation. Energy Policy 2012;41(C):666–75. [67] Hyndman RJ. Computing and graphing highest density regions. Am Stat 1996;50(2):120–6. [68] Mehrara M, Musai M. Granger causality between electricity consumption and economic growth in oil-dependent countries. Univ J Manag Soci Sci 2012;2 (6):134–9. [69] Cheng BS. Causality between energy consumption and economic growth in India: an application of co-integration and error-correction modeling. Indian Econ Rev 1999;34:39–49. [70] Chien T, Hu JL. Renewable energy: an efficient mechanism to improve GDP. Energy Policy 2008;36:3045–52. [71] Wei F, Hu C. Empirical analysis of regional energy consumptions and economic growth. J Theor Appl Inform Technol 2013;48(1):311–8. [72] Apergis N, Danuletiu D. Energy consumption and growth in Romania: evidence from a panel error correction model. Int J Energy Econ Policy 2012;2(4):348–56. [73] Pakistan Energy Yearbook. Hydrocarbon Development Institute of Pakistan. Ministry of Petroleum and Natural Resources; 2011. [74] Pakistan Energy Yearbook. Hydrocarbon Development Institute of Pakistan. Ministry of Petroleum and Natural Resources; 2013.