Aug 29, 2013 - (EG) and Energy consumption (EC) for the economy of Pakistan. ... growth Energy consumption Energy resources and Government Policies.
World Applied Sciences Journal 24 (6): 739-745, 2013 ISSN 1818-4952 © IDOSI Publications, 2013 DOI: 10.5829/idosi.wasj.2013.24.06.1107
Nexus Between Energy Consumption and Economic Growth: A Case Study of Pakistan 1
Sulaiman D. Muhammad, 2Muhammad Usman, 3Nooreen Mujahid and 4Ghulam Rasool Lakhan 1 Dean Faculty of Management Sciences, Federal Urdu University, Karachi, Pakistan 2 Department of Economics, Federal Urdu University Karachi, Pakistan 3 Department of Economics, University of Karachi, Pakistan 4 Department of Economics, Federal Urdu University Karachi, Pakistan Submitted: Jul 9, 2013;
Accepted: Aug 17, 2013;
Published: Aug 29, 2013
Abstract: The focus and prime purpose of this study is to find out the association between Economic growth (EG) and Energy consumption (EC) for the economy of Pakistan. In the time series framework, econometric techniques were employed i.e. ADF Unit Root Test (URT), Granger Causality Test (GCT), Johansen cointegration Test (JCT), ARDL and ECM. The findings of this study depict that there is a long term association between EG and EC. Furthermore, study showed there is a unidirectional association from EC to EG. Moreover, the study also recommended and suggested for sustainable EG, the government should take crucial decisions for the revival of Energy sector and make policies to discover other alternative Energy resources. JEL Classification: O49
E21
Key words: Economic growth
L72
L78
Energy consumption
Energy resources and Government Policies
INTRODUCTION
for the last five years is at an average of 3 percent of growth rate. On the other hand, there is a rapid increase in the demand for energy. While, there is a direct association between EG and EC. There is a dire need on the part of the government to take the extreme measures to cope up with the issue of energy shortfall. This is more common that the high growth rate of EC is followed by high growth rate of GDP and on the contrary, lower growth in EC caused lower growth in GDP. Figure 1. Revealed the same observation for Pakistan’s economy. Pakistan power sector consists of hydel, thermal and natural gas. Furthermore, nuclear and coal is also used as means of energy. Whereas, Pakistan’s entire supply was more than 64,727 thousand TOE’s1 instead of 64,522 last year [4]. The Figure 2 illustrates EC from different sources like electricity, coal and LPG; these were mostly equal since 2005-2006. While, the percent of gas consumption is increased compared to fiscal year 2005-2006 because Oil is found as an expensive source. Therefore, the total share of Oil is declined by 3 percent on the contrary; the total share of Gas has been increased 4.7 percent approximately.
Energy is the lifeline of a nation as the whole economy whether it is agriculture, industry or business depends on it. Today, 30% of the population of Pakistan have no access to electricity and about 80% have no access to gas. Pakistan ranks 165th out of 218 countries in per capita access to electricity [1]. Pakistan is currently facing energy crisis due to gap in demand and supply of power. The shortage of energy leads to unrest and agitation in the economy which in turn leads to close down many industrial units. This situation is an outcome of a combination of factors like political bankruptcy, lack of farsightedness, poor planning, an absence of institutional capacity, mismanagement and faulty governance [2]. Developing countries like Pakistan have to keep equilibrium between EC and its present and future demand also. Due to rise in energy demand along with insufficient supply, Pakistan has been facing shortage of energy which adversely affects the growth of the country. Pakistan has to focus on sustained and productive use of energy and think about energy conservation techniques. The growth rate of the economy
Corresponding Author: Muhammad Usman, Department of Economics, Federal Urdu University Karachi, Pakistan. Tel: +92-300-2679529.
739
World Appl. Sci. J., 24 (6): 739-745, 2013
Fig. 1: Relationship between growth rate of GDP and energy consumption Source: Hydrocarbon Development Institute of Pakistan, PBS and EA Wing [3]
Fig. 2: Energy consumption by source in %: A comoparison Source: Hydrocarbon Development Institute of Pakistan
As Table 1 also depicts the same trend that the consumption of petroleum products has declined from 2001 to 2004 and 2010 to 2012. Whereas in 2004-2005, 2007-2008 and 2009-2010 there was a positive change, which shows that opus of EC is changing from petroleum to other sources because of volatility in oil prices for the last ten years, the overall average percentage change is 1.1 for petroleum products, 5.1 for gas, 4.8 for electricity and 7.7 for coal. The largest sectors of energy consumers in Pakistan are industrial and transport sectors with 37.6% and 31.4 %
respectively. Nevertheless, domestic sector is also relatively one of the highest consumers of energy with 23.4 percent in 2011-12. Review of the Literature: Over the last half century, numbers of researches have been carried out to investigate and elucidate the association between EG and EC. There are two famous doctrines about the energy conservation. First, the EG is significantly associated with the nation’s wealth and oil consumption. Second, the energy choices depend on market mechanisms [5]. 740
World Appl. Sci. J., 24 (6): 739-745, 2013
Fig. 3: pakistan’s Sectorwise Energy Consumption (2011-2012) EC’s relationship with EG is rapidly changing with respect to time and space for many reasons. Therefore, the relationship is generally discussed [6]. The causal relationship is a prime topic of EG and EC since 1970s [7-8]. Now, this concept is recognized and accepted. While, the correlation between EG and EC shows increase in EC direct to change in EG or on the contrary, EG leads to changes in EC. With the help of causality techniques Sims [9] recognized the strong association between EG and EC. This study found the unidirectional causality from EG towards EC in the context of the USA for the epoch of 1947-1974. Whereas, the contribution of Karft and Karaft [10] confirmed by Akarca and Long [11], who confirmed that the data was taken during the time of the crisis and instability, but in stable condition there was no any causal association between considered variables. Shortly later, the outcomes were as well verified by Yu and Hwang, [12] emphasizing the economy of the USA for the era of 1947-1979. Though, other studies also proved a causal association between EG and EC by adopting many other econometric techniques and countries i.e. Masih et al. [13] studied about half a dozen Asian countries by employing JCT and ECM for the era of 1955 to 1990 and found a variety of answers for different countries. He establishes for India causality runs from EC to EG, whereas, in the case of Indonesia and Pakistan causality runs from EG to EC. He also studied about Malaysia, Philippine and Thailand and discovered there were no causal relationships between both variables. Asafu-Adjaye [14] investigated for the period of 1971 to 1995 and found there is a bidirectional causality between EG and EC in case of Philippine and Thailand and in case of India and Indonesia inducting causality, from EC towards EG. Wolde-Rufael [15] empirically calculated for the nineteen countries around the globe and established diverse results, in the case of Egypt, Ghana, Algeria,
Congo and Ivory Coast and found causality runs from EG towards EC. Whereas, for the Cameroon, Morocco and Nigeria causality was formed EC to EG. But in case of Zambia and Gabon there was bidirectional causality between the EG and EC. While, others were found neither type of causal relationship between the chosen variables i.e. Tunisia, Togo, Sudan, South Africa, Senegal and Kenya etc. Al-Iriani [16] found a causality relationship runs from EC towards EG in the 6 GCC (Globle Co-operation Countries) with the help of Panel cointegration and GMM models. Whereas Mehrara [17] confirmed same results from 11 Oil exporting countries by employing Panel cointegration and GMM models. Whilst, Lee and Chang [18] studied in two major categories of countries i.e. Developing and developed countries by employing Panel VAR(s) and GMM models and explored the unidirectional relationship between the developing countries, causality from EG towards EC whereas, on the contrary the case of developed countries there was a bidirectional causality association between the chosen variables. Akinlo [19] highlighted mixed outcomes for sub-Saharan African countries by employing ARDL bounds test. Whereas same mixed results were also found in these studies by employing Granger causality test for the time series annual data [20-22]. Lee et al. [23] found bidirectional causality between EG and EC from 1960 to 2001 in 22 OECD (Organization for Economic Co-operation and Development) countries by employing Panel cointegration and Panel VEC models. While, Huang et al. [24] studied about 82 countries in three major categories i.e. High, middle and low income countries by using data 1972 to 2002 and employing Panel VAR and GMM models. The outcomes of this research show in high income and middle countries there was a unidirectional causality relationship from EG towards EC. Whereas, in case of low income countries no causality found between EG and EC. Narayan et al. [25] examined the causal association between EG and EC in case of G-7 741
World Appl. Sci. J., 24 (6): 739-745, 2013
countries for the dataset 1972 to 2002. They used Panel cointegration and GCT and found unidirectional causality from EC towards EG. Lots of recent research studies have also conducted on the association of EC and EG [26-30]. All these studies have approximately different outcomes in different scenarios by employing major empirical techniques i.e. Granger causality, Cointegration and VAR tests etc. Our aim here is to see the nexus between EG and EC by using ARDL to cointegration approach and three mentioned tests also respectively.
definite set of tests discloses a little bit concerningcausality. A variable Yis said to Grangerly-cause Z if it can be revealed, typically via different series tests (such as F and T tests) on previous values of Y, that those Y values give important information concerning upcoming values of Z. Empirical Results and Methodological Framework: This study employs time-series data, from 1971-2013, which has been taken from World Bank. Time-series data always demonstrate trends; therefore, the stationarity properties are essential. Properties of stationarity of the time series indicators of macroeconomic can be examined by utilizing a range of URTs which are accessible in applied Econometrics. We employ one URT; ADF by Fuller and Dickey [33]. The explanation of this test is given below:
Model, Data Sources and Analytical Framework: This study reveals the connection between EG and EC. Therefore, this study anticipates causality and long run association between EG and EC. In the era of industrialization, the use of hi-tech technologies pushes the manufacturing sector in an upward direction, there is a vast increase in the demand for energy but due to numerous reasons Pakistan does not meet this requirement. Poor energy supply condition causes slowdown in EG.
ADF URT: The wide empirical finding is accessible, wherever social scientists have employed OLS method to examine the association between EG and EC. The issue with OLS process is that it is assumed that the error term has zero mean and finite constant and it is normally distributed. This means that without examining unit root properties of variables, OLS gives ambiguous empirical results which are ineffective for policy maker. It is compulsory to examine the stationarity of the variables while examining the long run association among the variables. We begin from the extensively employed URT i.e. ADF formulated by Dickey and Fuller [34]. Which applied if the error term appears to be non-stationary. The familiarity of the unit root properties of the variables is needed to make time series data consistent and efficient. The equation is of the model as follows:
The Model: Because of empirical studies and theoretical literature we use the real GDP as a proxy for EG. Derived from the common form of EC and EG is modeled as follows: EGt = f ( ECt )
(1)
ECt = f ( EGt )
(2)
Where, EG is EG and EC is level of EC. Natural log of both series EG and EC are taking. Cameron [31] pointed out that implementation of log linear condition is a good alternative choice for empirical investigation. The log linear condition gives efficient, coherent and impartial results. ln EGt = 1 + Y ln ECt +
i
(3)
ln ECt = 1 + Y ln EGt +
i
(4)
∆Yt =
1+
2T + Yt −1 +
(3)
m
j
∑ ∆Yt −i + i =1
i
Where is difference operator and following the assumption of normality.
is error term
∆Yt = (Yt − Yt −1 ) , ∆Yt −1 = (Yt −1 − Yt −2 ) , ∆Yt − 2 = (Yt − 2 − Yt −3 ) etc.
Where, ln ECt and is the natural log of EG, natural log of EC and error term supposed to be normally distributed, identically and independently. The GCT is a statistical or arithmetical hypothesis analysisfor influential whether one variable is helpful in predicting other. [32] Generally, regressions imitate "simple"associations, but the C. Granger, said that a
The key purpose at this time is to test whether =0 or 0. The significant t-values are created by Fuller and Dickey [35] to look at the unit root trouble. The variable Y is estimated value where as, estimated value is greater than the tabulated value of t. 742
World Appl. Sci. J., 24 (6): 739-745, 2013 Table 1: URT (Augmented Dickey fuller)
Table 6: Error Correction Model (ARDL)
Calculated
1% Critical
5% Critical
Variables
value
value
value
EC (0)
-2.551646
-4.211868
-3.529758
0.3033
-5.960373
-4.219126
-3.533083
0.0001
-58070.5
58317.2
-.99577[0.326]
-1.388643
-4.198503
-3.523623
0.8495
ECT(-1)
-0.16351
0.08741
-1.8706[0.070]
-6.560025
-4.205004
-3.526609
0.0000
R2
0.15993
F-stat
2.2210[0.103]
Source: Summarized and Tabulated by Authors “*MacKinnon (1996) one-
Adj R2
0.087921
DW-stat
2.0137
sided p-values. Notes: [Y: The Level form of the variable Y] [ (Y): The first
Source: Summarized and Tabulated by Authors
(EC)(0) EG (0) (EG)(0)
Regressor Probability*
Coefficient
Standard Error
T-Ratio[Prob]
(EC)
48.4722
34.2445
1.4155[0.166]
(Constant)
-688215.3
378633.9
-1.8176[0.078]
(TREND)
difference of the variable Y] [ (Y, 2): The second difference of the variable Y]”
The ARDL to Cointegration Approach: Pesaran et al. [36] develop ARDL bounds testing approach. This approach is better than the traditional long run approaches owing to many aspects. For instance, the ARDL is appropriate to relate for long term association among the variables if the variables are stationary at I (0) or I (1). The ARDL is proper for small sample size. In the existence of a number of the variables who value is determined in the model, the ARDL gives competent long run estimates with suitable t-statistics. Table no. 1 provides the outcomes of ADF URT with the assumption constant or intercept and trend term. The outcomes show that both the selected variables EC and GDP growth are time variant at level but both variables become time invariant at 1st difference. On the basis of above table JCT and ARDL to cointegration test can be applied but before this we check the Granger Causality between these two variables. Pair wise GCT with two lags shows that there is uni-directional causality from EC to EG growth at 10% level and reject the null hypothesis of EC does not Granger Cause EG and also accept the null hypothesis of EG does not Granger Cause EC. Due to unidirectional relationship, further estimation will be incorporated as, Log (EC) dependent variable and LOG (EG) as independent variable. Trace statistics demonstrate that all selected variables are cointegrated significant at the 5% level of and the null hypothesis of no long run cointegration equation is rejected. Johansen Test for Cointegration shows both variables have a long run relationship which also verified by ARDL model. Maximum Eigen statistics also illustrate that all chosen variables are cointegrated significantly at the 5 % level and the null hypothesis of no long run cointegration equation is rejected. This equation explains that the EG is highly affected with EC. There is a direct association between EG and EC.
Table 2: Pair wise GCTs Null Hypothesis:
F-Statistic Probability
LOG(EC) does not Granger Cause LOG(EG)
2.61202
0.08850
LOG(EG) does not Granger Cause LOG(EC)
0.14972
0.86153
Source: Summarized and Tabulated by Authors Table 3: Johansen Test for Cointegration (Trace) Trace
5% Critical
Probability
Statistics
Value
No. of Cointegration Equation
0.0041
27.59235
20.26184
None *
0.2517
5.304113
9.164546
At most 1
Source: Summarized and Tabulated by Authors*significance at the 5% level Table 4: Johansen Test for Cointegration (Max-Eigen) Max-Eigen
5% Critical
Probability
Statistics
Value
No. of Cointegration Equation
0.0043
22.28824
15.89210
None *
0.2517
5.304113
9.164546
At most 1
Source: Summarized and Tabulated by Authors*significance at the 5% level Table 5:
Long
Run
Cointegration
Equation
(Normalized
First
Cointegration Vector) Variables
Coefficient
Standard errors
t values
C
-57.68513
12.8071
-4.5041524
LOG (EC)
5.991101
1.16902
5.1248918
Source: Summarized and Tabulated by Authors Table 6: Autoregressive Distributed Lag (ARDL) Estimates AIC based Regressor
Coefficient
Standard Error
T-Ratio[Prob]
EG (-1)
0.83649
0.08741
9.5697[0.000]
EC
48.4722
34.2445
1.4155[0.166]
Constant
-688215.3
378633.9
-1.8176[0.078]
Trend
-58070.5
58317.2
-.99577[0.326]
R2 0.9531
Adj R2 0.9491
F-stat 237.4[0.00]
DW 2.0137
Serial Correlation
0.0023[0.961]
Functional Form
0.0475[0.829]
Normality ( 2)
967[0.00]
Heteroscedasticity
0.28387[0.597]
Source: Summarized and Tabulated by Authors
Table 6: Long Run Coefficients (ARDL) Regressor
Coefficient
EC
296.444
Standard Error 165.85
T-Ratio[Prob] 1.7874[0.083]
Constant
-4208956
1592318
-2.6433[0.012]
TREND
-355145.1
311575.9
-1.1398[0.262]
Source: Summarized and Tabulated by Authors
743
World Appl. Sci. J., 24 (6): 739-745, 2013
ARDL method of estimation is also used to calculate the long run association among variables. According to F-stat defines in the above table there is exist long run association between EC and EG and this association is significant at the 1 % level. Table no. 6 shows the long run coefficients generated by the ARDL to cointegration method. EC coefficient is positively related with EG and significant at 10% level. Coefficient of constant term is significant at 5% level. Although trend term is insignificant but we have trend in the data so trend term should not removed from the model. Table no. 6 shows the results of ECM through ARDL to cointegration approach. The table illustrates that the ECM model exists at the 11 % level of significance. The error correction term is significant at the 7 % level and minus sign of ECT shows the convergence of equilibrium from short to long run.
10-12 billion US dollars in a year on the imports of crude oil and different petroleum products. REFERENCES 1.
2.
3.
4.
CONCLUSION
5.
Through seminal literature and empirical results it is suggested that there is a significant direct association between EC and EG. ARDL and other econometric techniques also verified this relationship. The ADF tests showed selected variables EC and EG are stationary at 1st difference. Furthermore, Johansen Test for Cointegration shows both variables have a long run relationship which also verified by ARDL model later. Whereas, Granger Causality test depicts that there is a unidirectional relationship between EC and EG which runs from EC to EG.
6.
7.
8.
9.
Policy Recommendations: EG is directly linked with an energy requirement of a country. Pakistan’s economy is adversely affected by mismanagement, incompetency and negligence. Where else, EG can be attained with sustainable energy supply and effective government policies. The Government may form a National Energy Authority and must focus on the indigenous resources to expand its energy base and seek new emerging energy sources. Furthermore government should show his political will to improve and expedite the sector and take strong actions to complete major projects i.e. Pak-Iran gas pipeline, Pak-Qatar gas pipeline, Turk-Afghan-Pak pipeline and construction of dams at the run of the rivers. It is also recommended to pursue technologies that maximize the efficiency especially in industry and domestic use. There is a dire need to reduce our dependence on imported oil as a country spends around
10.
11.
12. 13.
14.
744
World Fact Book., 2012. Central Intelligence Agency, US Government Printing Office Available at : https:// www.cia.gov/library/publications/the-worldfactbook/index.html Annual Report, 2012. National Electric Power Regulatory Authority, Pakistan Available at : http:// www.nepra.org.pk/annualreports.htm Energy Yearbook, 2012. Hydrocarbon Development Institute of Pakistan, Pakistan Journal of Hydrocarbon Research, available at: http:// hdip.com.pk/contents/Energy-Yearbook-2012 Economic Survey of Pakistan, 2012-13. Ministry of Finance, Government of Pakistan: available at: http://www.finance.gov.pk/survey_1112.html, 188190. Percebois, J., 1978. Energie, croissance et calcul e´conomique. Revue e´conomique , 29: 464-493. Diandy, I.Y., 2009. Consommation d’e´lectricite ´ et croissance dans l’UEMOA: une analyse en termes de causalite´.Me´moire, Universite´ Cheikh Anta Diop de Dakar. Kraft, J. and A. Kraft, 1978. On the relationship between Energy and GNP. Journal of Energy and Development, 3: 401-403. Akarca et al., 1980. On the relationship between Energy and GNP: a re-examination Journal of Energy and Development, 5: 326-331. Sims, 1972. Money, income and causality American Economic Review, 62: 540-552. Kraft, J. and A. Kraft, 1978. On the relationship between Energy and GNP. Journal of Energy and Development, 3: 401-403. Akarca, et al., 1980. On the relationship between Energy and GNP: a re-examination Journal of Energy and Development, 5: 326-331. Yu, et al., 1984. The relationship between energy and GNP” Energy Economics, 6: 186-199. Masih, A., 1996. Energy consumption and real income temporal Causality, results of a multi-country study based on cointegration and Error-correction techniques. Energy Economics, 18: 165-183. Asafu-Adjaye, J., 2000. The relationship between Energy consumption, energy prices and Economic Growth: time series evidence from Asian developing countries. Energy Economics, 22: 615-625.
World Appl. Sci. J., 24 (6): 739-745, 2013
15. Wolde-Rufael, 2005. Energy demand and Economic Growth: the African Experience. Journal of Policy Modeling, 27(8): 891-903. 16. Al-Iriani, M.A., 2006. Energy-GDP relationship revisited: an example from GCC Countries using panel causality. Energy Policy, 34(17): 3342-3350. 17. Mehrara, M., 2007. Energy consumption and Economic Growth: the case of oil Exporting countries. Energy Policy, 35(5): 2939-2945. 18. Lee, C.C. and C.P. Chang, 2007. Energy consumption and GDP revisited: a panel analysis of developed and developing countries. Energy Economics, 29: 1206-1223. 19. Akinlo, A.E., 2008. Energy consumption and economic growth: evidence from 11 Sub-Sahara African countries. Energy Economics, 30(5): 2391-2400. 20. Chiou-Wei, S.Z. C.hen, Ching-Fu and Z. Zhu, 2008. Economic growth and energy consumption revisited—evidence of linear and nonlinear Granger causality. Energy Economics, 30(6): 3063-3076. 21. Imran, K., 2010. Energy Consumption and Economic Growth: A Case Study of Three SAARC Countries. European Journal of Social Sciences, 16: 2. 22. Ghosh, S., 2009. Import demand for crude oil and Economic Growth: evidence from India. Energy Policy, 37: 699-702. 23. Lee, C.C. and C.P. Chang, 2008. Energy consumption and Economic Growth in Asian Economies: a more comprehensive analysis using panel data” Resource and Energy Economics, 30(1): 62-65. 24. Huang, B.N. et al., 2008. The causal relationship between energy Consumption and GDP growth revisited: a dynamic panel data approach Ecological Economics, 67: 41-54.
25. Narayan, P.K. and R. Smyth, 2008. Energy consumption and real GDP in G7 countries: New evidence from panel cointegration with structural breaks. Energy Economics, 30: 2331-2341. 26. Raymond, Li, 2012. Coal consumption and Economic Growth in China, Energy Policy, 40: 438-44. 27. Eggoh, C.J., 2011. Energy consumption and Economic Growth revisited in African countries. Energy Policy 39: 7408-7421. 28. Tamba, et al., 2012. Causality analysis of diesel consumption and Economic Growth in Cameroon Energy Policy, 45: 567-575. 29. Zhang, 2011. Interpreting the dynamic nexus between Energy consumption and Economic Growth: Empirical evidence from Russia Energy Policy, 39: 2265-2272. 30. Ozturk, et al., 2010 Energy consumption and Economic Growth relationship: Evidence from panel data for low and middle income countries Energy Policy, 37: 617-622. 31. Cameron, S., 1994. A review of the econometric evidence on the effects of capital punishment. Journal of Socio-economics, pp: 197-214. 32. Granger, C.W.J., 1969. Investigating Causal Relations by Econometric Models and Cross-spectral Methods. Econometrica 37(3): 424-438. do: 10.2307/ 1912791.JSTOR1912791. 33. Dickey, D. and W.A. Fuller, 1979. Distribution of the Estimates for Autoregressive Time Series with a Unit Root, Journal of the American Statistical Association, 74: 427-431. 34. Ibid 35. Ibid 36. Pesaran, et al., 2001. Bounds testing approaches to the analysis of level relationships Journal of Applied Econometrics, 16: 289-326.
745
