Electricity consumption–growth nexus: The case of Malaysia

ARTICLE IN PRESS Energy Policy 38 (2010) 606–612

Contents lists available at ScienceDirect

Energy Policy journal homepage: www.elsevier.com/locate/enpol

Electricity consumption–growth nexus: The case of Malaysia V.G.R. Chandran a,c,, Susan Sharma b, Karunagaran Madhavan c,1 a b c

Department of Economics, Faculty of Economics and Administration, University of Malaya, 50603 Kuala Lumpur, Malaysia School of Accounting, Economics and Finance, Deakin University, Melbourne, Australia Department of Economics, Faculty of Business Management, University Technology MARA, 40540 Shah Alam, Malaysia

a r t i c l e in f o

a b s t r a c t

Article history: Received 13 June 2009 Accepted 2 October 2009 Available online 29 October 2009

The goal of this paper is to model the relationship between electricity consumption and real gross domestic product (GDP) for Malaysia in a bivariate and multivariate framework. We use time series data for the period 1971–2003 and apply the bounds testing approach to search for a long-run relationship. Our results reveal that electricity consumption, real GDP and price share a long-run relationship. The results of the autoregressive distributed lag (ARDL) estimates of long-run elasticity of electricity consumption on GDP are found to be around 0.7 and statistically significant. Finally, in the short-run, the results of the causality test show that there is a unidirectional causal flow from electricity consumption to economic growth in Malaysia. From these findings we conclude that Malaysia is an energy-dependent country, leading us to draw some policy implications. This paper adds support and validity, thus reducing the policy makers concern on the ambiguity of the electricity and growth nexus in Malaysia. & 2009 Elsevier Ltd. All rights reserved.

JEL classification: C22 R41 Keywords: Electricity consumption Economic growth Cointegration

1. Introduction Over the past three decades the relationship between energy consumption and economic growth has been a major issue of debate among economists and policy makers. Currently, in Malaysia, the increasing cost of energy has set the pace for conservation policies. The government is also continuously reviewing its energy policy to ensure sustainability of the energy resources (Mohamed and Lee, 2006). Although Malaysia has been fortunate to be relatively well endowed with fossil sources of energy and managed to meet the energy demand, however, in recent years significant growth in the energy demand is recorded. In 2004, the growth in electricity demand was 9.1% and, it is way above the gross domestic product (GDP) growth of 7.5% (Pusat Tenaga Malaysia, 2004). Projection by Gan and Li (2008) shows that total primary energy consumption would triple by 2030 while the final energy demand is projected to reach 116 mega ton of oil equivalent (Mtoe) by 2020 based on 8.1% annual growth rate (Keong, 2005). In addition, the estimated average annual electricity demand is expected to grow by 8.87% (Keong, 2005) and among the ASEAN countries, Malaysia is one of the countries that recorded high energy per capita and electricity intensity over the years (Pusat Tenaga Malaysia, 2004; Shrestha et al., 2009). Besides, the industrial sectors are projected to absorb 50% of the projected energy

 Corresponding author. Tel.: + 60 17 6843705; fax: +603 80628593.

E-mail addresses: [email protected] (V.G.R. Chandran). [email protected] (S. Sharma), [email protected] (K. Madhavan). 1 Tel.: + 60 16 2925050. 0301-4215/$ - see front matter & 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.enpol.2009.10.013

demand of 116 Mtoe (Keong, 2005). The growing energy demand and high dependency on electricity demand by the manufacturing sectors warrant attention since formulation of wrong energy policies would adversely affect the sectors and consequently the country’s growth. In this literature on applied energy economics, for one to provide reliable results one should be certain that regardless of models and estimation techniques used, the results are robust. In this study, we use a recent dataset and two different model specifications (bivariate and trivarate) that are commonly used in the literature of energy and growth to reduce the gap of ambiguity in research findings on Malaysia. This approach allows us to check the robustness of the empirical outcome of the causal relationships. Owing to the use of single country as a case, this study overcomes the country-specific differences present in studies using pooled data and allows the interpretation of the result to take into account the institutional, structural and policy reforms of that country more precisely. It also leaves more room for the study to draw better policy implications pertaining to the country under study. Another issue that we address is the one that relates to small sample sizes. In this regard, Lee (2005) and Mah (2000) have cautioned researchers on the use of short data spans, which eventually lower the power of the cointegration analysis. Mah (2000) stated that the error correction model (Engle and Granger, 1987), Johansen (1988) and Johansen and Juselius (1990) methods are unreliable for studies that have small sample.2 To remedy these 2 Reinsel and Ahn (1992) and Reimers (1992) suggest the correction of trace statistics to remedy the small sample biasness in Johansen cointegration.

ARTICLE IN PRESS V.G.R. Chandran et al. / Energy Policy 38 (2010) 606–612

607

Table 1 Summary of the literature review. Authors

Time period

Countries

Methods and tests used

Major findings (causality)

Altinay and Karagol (2005) Ang (2008) Chen et al. (2007) Tang (2008)

1950–2000 1971–1999 1971–2001 1972:1–2003:4

Turkey Malaysia 10 developing Asian countries Malaysia

Energy-GDP GDP-energy Energy2GDP Electricity2GDP

Ghosh (2009)

1970–2006

India

Toda and Yamamoto (1995)—MWALD Johansen; Granger causality—VECM Pedroni (1999, 2004) test; ECM Kanioura and Turner’s ECM-based Ftest; Granger and MWALD causality ARDL bounds testing approach

Halicioglu (2009) Lee and Chang(2008) Masih and Masih (1996)

1960–2005 1971–2002 1955–1990

Mehrara (2007) Mozumder and Marathe (2007) Narayan and Smyth (2009)

1971–2002 1971–1999

Turkey 16 Asian countries India, Pakistan, Malaysia, Singapore, Indonesia and the Philippines Iran, Kuwait and Saudi Arabia Bangladesh

1974–2002

6 Middle Eastern countries

Odhiambo (2009)

1971–2006

Tanzania

Payne (2009) Squalli (2006) Yoo (2006)

1949–2006 1980–2003 1971–2002

Yuan et. al (2007)

1978–2004

United States OPEC members Indonesia and Thailand Malaysia and Singapore China

econometric issues in estimation, a relatively recent technique known as the bounds testing approach to a long-run relationship has become popular. Several recent studies have applied the bounds test in sample sizes of around 30 observations (see, for instance, Narayan and Narayan, 2005a, b). The popularity of this test is also due to the availability of small sample size critical values, which are now available in Narayan (2005). Against this background, the goal of this paper is to examine the relationship between real income, electricity consumption and price using annual data for the period 1971–2003. This study makes two contributions to the literature. First, this study uses the bounds testing approach shown to have better statistical properties compared with existing tests. This will add precision and reliability to our results and hence will give credence to policy implications derived from our work. Second, while a number of studies have been done that include Malaysia, these studies use either panel data or bivariate framework that ignores country specific effects and omitted variables biasness, respectively (Sari and Soytas, 2009; Chang et al., 2001; Stern, 2000). Therefore, it is necessary to study countries individually using a multivariate framework. To the best of our knowledge, this will be the first single-country study on Malaysia, modeling electricity consumption, electricity price and GDP in a multivariate framework. The other sections of the paper are organized in the following way. In Section 2, we provide a brief overview of the related literature on the causal relationship between energy consumption and economic growth. The empirical model is described in Section 3. Section 4 reports the empirical evidence and, finally, the policy implications and concluding remarks are provided in Section 5.

2. Brief literature review on the energy–growth nexus It is important for policy makers in Malaysia to understand the relationship between energy consumption, particularly the electricity demand and economic growth. The debate focuses on whether energy causes economic growth or economic growth causes energy consumption or a bi-directional relationship exists.

Bounds testing; Granger causality Panel cointegration techniques Johansen and Juselius (1990); VECM

Toda and Yamamoto (1995); ECM Johansen (1988) and Johansen and Juselius (1990); VECM Panel cointergration test; panel Granger causality test ARDL-bounds testing Toda and Yamamoto (1995) Toda and Yamamoto (1995)—MWALD Johansen and Juselius (1990) Granger—VAR Johansen and Juselius (1990); Hodrick– Prescott filter

Short-run: GDPelectricity Energy-GDP Energy-GDP No relationship

GDP-energy GDP-electricity Electricity2GDP Energy-GDP Electricity-GDP No relationship Energy-GDP GDP-electricity GDP2electricity Energy-GDP

The direction of causality between these variables is important to policy makers in order to implement the energy conservation policy (Jumbe, 2004). For example, if we find evidence of a positive unidirectional causality running from income to energy consumption, it implies that the country does not depend on energy for economic development. Therefore, the country can adopt energy conservation policies without any detrimental effect on economic growth. If, however, there is a unidirectional causality running from energy consumption to income, it indicates that the country is dependent on energy consumption for economic growth. As a result, energy conservation policies may harm economic growth (Narayan and Singh, 2007). However, the results pertaining to the causal relationship between economic growth and energy consumption have been mixed and remain ambiguous. A summary of the literature review is provided in Table 1. Some recent studies have found that causality runs from economic growth to energy consumption (Ghosh, 2009; Mozumder and Marathe, 2007; Mehrara, 2007) while other studies have found that causality runs from energy consumption to economic growth (Odhiambo, 2009; Halicioglu, 2009; Yuan et al., 2007; Squalli, 2006; Altinay and Karagol, 2005). There are also some studies that found no causal (Payne, 2009; Masih and Masih, 1996) and bilateral causality (Chen et al., 2007; Narayan and Smyth, 2009) relationship between economic growth and energy consumption. According to Masih and Masih (1998) and Hondroyiannis et al. (2002) the main reason for these conflicting empirical results is due to differences in institutions, structural reforms and policies adopted by different countries. In addition, the use of different econometric estimation techniques and sample periods also influences the results. It is evident that in case of Malaysia, two of the studies established a bi-directional causality (Tang, 2008; Yoo, 2006), two found unidirectional (Ang, 2008; Lee, 2005), while the other (Masih and Masih, 1996) found no causality. On the whole, the literature seems to show mixed evidence of energy–income causality in Malaysia depending on data, period of study and methodology employed. This has motivated the need to reinvestigate the notion of causality between energy consumption and

ARTICLE IN PRESS 608

V.G.R. Chandran et al. / Energy Policy 38 (2010) 606–612

economic growth for Malaysia. A second feature of the literature on Malaysia is that none of the studies consider the electricity market in a multivariate framework. Hence, our study is novel in that we consider for the first time Malaysia’s electricity market in both bivariate and multivariate frameworks. The objective here is to check the robustness of the results.

3. Model specification In this study, consistent with the literature, we use two different specifications to model the GDP–electricity nexus. Model 1 takes the usual bivariate specification by only including the real GDP and electricity consumption variables. In Model 2, a price variable is used as a third variable, in addition to real GDP and electricity consumption. Model 2 follows the demand side approach (energy demand function) where aggregated data on energy,3 GDP and price proxied by consumer price index are used (Oh and Lee, 2004). The inclusion of price series was due to its importance in influencing income and energy consumptions (Dunkerley, 1982; Hoa, 1993). Since the price of energy was unavailable, we follow the normal practice of using consumer price index (CPI) as the proxy. The information-intensive difficulty of this exercise is compounded by the fact that the use of energy sources (such as use of coal, oil, etc.) varies in different economies and different prices exist for residents and industries. Furthermore, industries that are energy-intensive may well be subsidised by the government and therefore face different prices (Mahadevan and Asafu-Adjaye, 2007). Hence the CPI is used instead as energy prices are expected to be sufficiently reflected in this index. A number of studies have used the CPI as a proxy for electricity and or energy prices (see, for instance, Mahadevan and Asafu-Adjaye, 2007; Oh and Lee, 2004; Hondroyiannis et al., 2002; Asafu-Adjaye, 2000; Masih and Masih, 1998). While this can be considered as a weakness, in the absence of any energy price indices, the CPI seems to be an attractive alternative. We included three dummy variables to account for two oil crises and the Asian financial crisis. Our initial estimation results showed that the two oil crises do not seem to be significant. However, we observe that the 1997–1998 Asian financial crisis had a significant influence on the real GDP—a finding well established in the economic growth literature. Hence, we estimate both the models with one dummy variable to account for the Asian financial crisis. 3.1. Estimation procedures To determine the order of integration of the three variables, first we conducted a unit root test using both the augmented Dickey Fuller (ADF) and Phillips and Perron (PP) test. Both tests were used to check the robustness of the results. The PP test is used because it allows for milder assumptions on the distribution of errors. Further, the test controls for higher order serial correlation in the series and is robust against heteroscedasticity. Once the order of integration is determined, we test if there is a long-run relationship between the variables. Here, we use the autoregressive distributive lag (ARDL) method proposed by Pesaran et al. (2001). The ARDL procedure is popular in applications involving small sizes because, as Pesaran et al. (2001) show, it performs statistically better compared with traditional cointegration tests, such as the Johansen (1988) and the Engle and Granger (1987) tests. The popularity of the ARDL 3 The production side approach follows the aggregate production function (see Lee and Chang, 2008). In this study following other studies (e.g. Payne, 2009; Sari, and Soytas, 2009; Lee and Chang, 2008) we used aggregated values of GDP and EC, and not per capita data.

procedure has been boosted by its relatively more flexible statistical pre-requisites: that is, it is applicable irrespective of whether the underlying variables are integrated of order one or zero. The order of integration is always a central issue in time series models, and competing cointegration tests require all variables to be integrated of order one. Essentially, the ARDL procedure involves estimating an unrestricted error correction model (UECM) in first difference form, augmented with one period lagged of all variables in the model. It follows that the UECM has the following form: Model 1:

Drgdpt ¼ a0M þ a1M DUM97 þ

k1 X

biM Drgdpti þ

i¼1

k2 X

ziM Decti

i¼1

þ y1M rgdpt1 þ y2M ect1 þ eMt

Dect ¼ a0E þa1E DUM97 þ

k1 X

biE Decti þ

i¼1

ð1Þ

k2 X

ziE Drgdpti

i¼1

þ y1E rgdpt1 þ y2E ect1 þ eEt

ð2Þ

Model 2:

Drgdpt ¼ a0M þ a1M DUM97 þ

k1 X

biM Drgdpti þ

i¼1

þ

k3 X

k2 X

ziM Decti

i¼1

liM Dpt1 þ y1M rgdpt1 þ y2M ect1 þ y3M pt1 þ eMt

i¼1

ð3Þ

Dect ¼ a0E þa1E DUM97 þ

k1 X i¼1

þ

k3 X

biE Decti þ

k2 X

ziE Drgdpti

i¼1

liE Dpti þ y1E rgdpt1 þ y2E ect1 þ y3E pt1 þ eEt

ð4Þ

i¼1

where the rgdp, ec and p are the natural logarithm of real GDP, electricity consumption and price index, respectively. The first difference operator is denoted as D. DUM97 is the dummy variable for Asian financial crisis. The F-test can be used to test the longrun relationship by testing the lagged levels of the variables. The null hypothesis of no cointegration in Eqs. (1) (2) (3) and (4) are y1M = y2M = 0, y1E = y2E = 0, y1M = y2M = y3M = 0 and y1E = y2E = y3E = 0, respectively. The computed F-statistics under the null hypothesis no long-run relationship are denoted as FM(rgdp|ec), FE(ec|rgdp), FM(rgdp|ec,p) and FE(ec|rgdp,p), respectively. The asymptotic distributions of the test statistics are non-standard regardless of whether the variables are I(0) or I(1). For this purpose, Pesaran et al. (2001) computed two sets of asymptotic critical values where the first set assumes variables to be I(0) and the other as I(1) which are known as lower bounds (LCB) and upper bounds critical values (UCB), respectively. Given that Pesaran et al.’s (2001) critical values are computed for a large sample size (T=500), Narayan (2005) estimated a new set of critical values for sample sizes ranging from T= 30 to 80. Narayan and Narayan (2005a, b) and Narayan and Smyth (2006a, b) explain clearly the advantages of using this small sample size critical values; we follow their suggestion since our sample size is only 33 observations, and use the critical values from Narayan (2005).4 A decision on whether cointegration exists between the dependent variable and its regressors is then made as follows. If the computed F-statistics is 4 For other studies that have used small sample sizes applied to the bounds test, see Narayan (2004), and Narayan and Smyth (2006a, b).

ARTICLE IN PRESS V.G.R. Chandran et al. / Energy Policy 38 (2010) 606–612

greater than UCB, then we reject H0 and conclude that the dependent variable and the regressors are cointegrated. On the other hand, if the computed F-statistics is lesser than the LCB, we fail to reject H0, which signifies that there is no cointegration. However, if the computed F-statistics falls between the UCB and LCB, then the results are inconclusive. 3.2. Granger causality If the series are not cointegrated, we perform the Granger causality as a vector autoregressive (VAR) in first differences form. Alternatively, if the specifications of the series are cointegrated, we conduct the Granger causality with inclusion of the lagged error correction term (ECT) (which is obtained from the long-run cointegration relationship) as an additional independent variable in the equation. This is to avoid the misleading use of VAR estimation as proposed by Engle and Granger (1987). Hence, with the presences of cointegration, the Granger causality relationships are written as vector error correction models (VECM) given below: Model 1:

Drgdpt ¼ a0M þ

k X

biM Drgdpti þ

i¼1

k X

ziM Decti þ aM ECTt1 þ eMt

i¼1

ð5Þ

Dect ¼ a0E þ

k X

biE Drgdpti þ

i¼1

k X

ziE Decti þ aE ECTt1 þ eEt

ð6Þ

i¼1

Model 2:

Drgdpt ¼ a0M þ

k X

biM Drgdpti þ

i¼1

k X

ziM Decti þ

i¼1

k X

liM Dpt1

i¼1

þ aM ECTt1 þ eMt

Dect ¼ a0E þ

k X

biE Decti þ

i¼1

þ aE ECTt1 þ eEt

ð7Þ

k X i¼1

ziE Drgdpti þ

k X

609

Table 2 Unit root test results. Variables

Level

First differences

PANEL A: Phillip–Perron tests rgdp  1.37 p  3.42** ec  1.72

 4.60***  3.31**  6.93***

PANEL B: ADF tests rgdp p ec

 4.64***  3.01**  7.13***

 1.40  1.27  1.69

The optimal lags for the ADF tests were selected based on optimizing Akaike’s information criteria (AIC) while for Philip–Perron tests, the Newey–West Bandwidth were used. *** and ** denote significant level at 1% and 5% level, respectively.

Table 3 The results of cointegration test. The bounds test for cointegration Model 1 FM(rgdp|ec) FE(ec|rgdp) Model 2 FM(rgdp|ec, p) FE(ec|rgdp, p)

F-statistics

Engle–Granger test (Model 1) (Residual based cointegration test)

t-statistics  3.570**

5%  3.00

1%  3.70

Johansen cointegration test (Model 2) None At most 1 At most 2

Trace statistic 37.923*** 13.798 3.098

5% 29.68 15.41 3.76

1% 35.65 20.04 6.65

5.36** 0.531 8.152*** 0.617

Critical value 5% LCB: 4.090 UCB: 4.663 1% LCB: 5.155 UCB: 6.265

*** and ** denote significant level at 1% and 5%, respectively. Lag selection is based on AIC. The critical values for bounds test is obtained from Narayan (2005). Following, Reinsel and Ahn (1992) and Reimers (1992), we correct the Johansen trace test statistics for finite-sample bias by the scale factor of ((T pk)/T).

liE Dpti

i¼1

ð8Þ

In (5) and (7), to test whether lagged first differences of ec Granger cause rgdp, we impose restrictions on all the lagged ec using the F-test. This is equivalent to testing the short run causality of ec on rgdp. To test the significance of the lagged error correction term (ECT), we used the t-statistics. If the ECT is significant then we have evidence of long-run causality.

4. Empirical results Table 2 shows the results of the unit root test. Based on the ADF test, all the variables are found to be non-stationary (integrated of order one, I(1)). On the other hand, when we use the PP test, we find that only real GDP and electricity consumption are non-stationary; the price variable turns out to be stationary, since we are able to reject the null hypothesis of a unit root at the 5% of significance. In sum, there seems to be some uncertainty regarding the integrational properties of the data series. Therefore, the use of bounds test is warranted since the bounds test is applicable for variables that are either I(1) or I(0). The results of the cointegration test based on bounds test are summarized in Table 3. For robustness of results, we also include the results of the Engle–Granger cointegration (for Model 1) and Johansen multivariate cointegration (for Model 2) with modified trace statistics. The computed F-statistics for Model 1 when real GDP is the dependent variable turns out to be 5.36 which is above the upper bound critical value of 4.663. Likewise in Model 2, the

result when real GDP is the dependent variable (F-statistics of 8.15) points to the same conclusion that there is cointegration between the variables under study but this time at a higher level of significance (at the 1% level). This indicates that there exists a long-run relationship between GDP and electricity consumption. Similarly, the Engle–Granger residual based cointegration indicates cointegration, and the Johansen multivariate cointegration indicates at least one cointegration. In the case of ec and p serving as the dependent variable, there is no evidence of any long-run relationship between the regressors and its determinants (results not reported to conserve space). The ec and p appear to be the long-run forcing variable for rgdp based on the bounds test. However, in the short run, the direction of Granger causality may be different. The direction of causality in the short run is given in Table 4. The results show that in the short run, ec to have a significant impact on economic performance at 1% significant level for both models. The price is only found to be significant at 10% level. The short run adjustment process is measured by the error correction term (ECT). If the ECT value is between 0 and  1, the correction to rgdp in period t is a fraction of the error in period t  1. In this case, the ECT tends to cause the rgdp to converge monotonically to its long-run equilibrium path in relation to changes in the exogenous variables. From Table 4, we see that the ECT is between 0 and  1 and is statistically significant at the 1% significance level. This implies that, the error correction process converges monotonically to the equilibrium path relatively quickly. The statistical significance of the ECT also

ARTICLE IN PRESS 610

V.G.R. Chandran et al. / Energy Policy 38 (2010) 606–612

Table 4 Granger causality test. Dependent variable

F-statistics (prob.)

[t-statistics]

P

P

P

ECTt  1

Model 1 Drgdpt

–

17.15 (0.002)***

–

 0.34 [  3.58]***

Dect

2.25 (0.127)

–

–

–

16.68 (0.003)***

2.91 (0.088)*

 0.33 [  3.66]***

–

0.31 (0.736)

–

Drgdpt  1

Model 2 Drgdpt

2.18 (0.136)

Dect

Dect  1

Dpt  1

Diagnostic testa

A: (0.019)[0.892] B: (2.324)[0.126] C: (3.712)[0.156] D: (2.355)[0.309] A: (0.358)[0.549] B: (0.405)[0.524] C: (0.871)[0.351] D: (2.691)[0.237] A: (0.487)[0485] B: (2.312)[0.128] C: (2.071)[0.163] D: (0.623)[0.732] A: (0.319)[0.572] B: (0.117)[0.738] C: (0.542)[0.462] D: (2.152)[0.450]

*** and * donates significant level at the 1% and 10%, respectively. Lag selection is based on AIC. The CUSUM plot and CUSUMSQ plot showed that the line stays within the 5% significant level, hence, indicating no evidence of any structural instability (not reported to conserve space and available upon request). a

Diagnostic test for Eqs. (5)–(8). A, B, C and D is the serial correlation (LM test), functional form (RESET), heteroscedasticity (ARCH) and normality test, respectively.

Table 5 ARDL estimates of long-run elasticities. Variables

Model 1

Model 2

Constant ec p

5.206 (36.18)*** 0.683 (56.41)*** –

5.535(26.73)*** 0.787(11.35)***  0.297(  1.64)

Lag selection was based on AIC. *** denotes 1% significant level.

confirms the presence of a long-run equilibrium between the rgdp and the regressors (ec and p) and suggests that in the long-run both ec and p Granger cause rgdp. The magnitude of the ECT term suggests that a deviation from the equilibrium level of rgdp during the current period will be corrected by 33–34% in the next period. Having found a long-run relationship between real GDP and electricity consumption in both the models when GDP serves as the dependent variable, we proceed to estimate the long-run elasticities. In other words, we investigate the impact of electricity consumption on GDP for Malaysia. We used the autoregressive distributed lag model (ARDL) to estimate the elasticities. Based on Table 5, in both the models, it is found that electricity consumption has a positive impact on real GDP in Malaysia in the long run. A 1% increase in ec will result in a 0.68–0.79% increase in rgdp. The estimated elasticities are slightly higher compared to the results (0.548) reported by Ang (2008) between commercial energy consumption and economic growth in Malaysia. It should be noted that these studies use a completely different measure of energy, so the different result is expected.

5. Conclusions and policy implications This study reassessed the relationship between electricity consumption and economic growth for Malaysia using time series data for the period 1971–2003. The modeling approach was based on the popular small sample size based estimation technique proposed by Pesaran et al. (2001). Our model included real GDP,

electricity consumption and price. We found a long-run relationship among these variables. We find fairly robust results on the impact of electricity consumption on GDP. In both our models, electricity consumption has a positive impact on GDP for Malaysia. The magnitude of the impact ranges from 0.68 to 0.79, implying that a 1% increase in electricity consumption leads to between 0.68% and 0.79% increase in GDP. This finding implies that in the case of Malaysia, energy serves as an important source of economic growth. The long-run result suggests that Malaysia is an energydependent country. Hence, it implies that any conservation policies or a shock to energy supply will have an adverse effect on economic growth. In the Ninth Malaysia Plan, the government initiates a series of energy saving programs to ensure efficient utilization of energy resources and for the promotion of environmental measures. These programs mainly focus on the industrial, transportation, as well as other commercial sectors. The industrial sector is encouraged to implement energy saving measures by improving the existing plant and equipment as well as processes. If these measures are made compulsory, it is likely that these conservation policies will have significant negative impacts on the growth of these industries.5 Additionally, it is important for the government to ensure a secure, reliable and cost-effective supply of energy for economic growth. Generally, for electricity supply, it is reported that transmission and distribution (T&D) losses in developing countries to be two to four times higher than in the OECD countries (Karki et al., 2005). Therefore, the government-formulated strategies should focus on efficient production and utilization of energy. To achieve efficiency, a market-based approach needs to be adopted. Policy should emphasize in improving productivity and efficiency of energy suppliers as well as discouraging wasteful competition. For future economic growth, investment in energy sector needs to be further developed. Realizing that this sector is

5 However, Smulders and Michiel (2003) showed that the tradeoffs between energy use reduction and economic growth become less severe when new efficient technologies are implemented.

ARTICLE IN PRESS V.G.R. Chandran et al. / Energy Policy 38 (2010) 606–612

capital-intensive, there is a need for global cooperation in developing the appropriate mechanism and institution for a more sustainable energy supply. This should be subsequently followed up with investments in energy related infrastructure. Subsequently moving to alternative energy as a source is important since increasing dependence on energy will burden the economy. Recently, as an alternative, Malaysia has also embarked in an ambitious project to develop alternative energy supply. Malaysia has approved 75 biodiesel manufacturing projects. However, turning to renewable energy supply (biofuel), as an alternative it needs ample resources for research and development especially to allow technologies such as biodiesel to move from laboratory to the market. Malaysia’s effort in moving towards this direction is still uncertain. Owing to the energydependent nature of the economy, the study, therefore suggest that Malaysia seriously pursue effort in developing the alternative source of energy. We suggest that more government support in terms of research and development to be offered with other mechanism such as institutions, incentives and the like to be further developed to encourage the progress of the alternative energy sectors. As a whole, although this study sheds some light on the direction of causality between electricity consumption and growth, the limitation of this study warrants some attention. This study is limited in that it has only used different models and estimation techniques on electricity consumption and growth. Future research should focus on the impact of different types of energy consumption (i.e., electricity, petroleum, coal and gas) on growth so that sector specific policy implications can be drawn.

Acknowledgements We would like to thank the anonymous referees and the editor of Energy Policy, Nicky France, for their insightful comments and suggestions. The usual disclaimer applies.

Appendix A. Data sources

Data series

Source

Gross domestic product (GDP) Electricity consumption (kWh)–(ec) [GDP was deflated using GDP deflator to obtain the real values]

Asian Development Bank (key indicators 1999–2009 available at http://www.adb. org/Documents/Books/ Key_Indicators/default.asp and Malaysian Energy Balance Report, Malaysia, various years) International Financial Statistics (IFS) Database

GDP deflator (2000 as base year) Consumer price index (2000 as base year)–(p)

References Altinay, G., Karagol, E., 2005. Electricity consumption and economic growth: evidence from Turkey. Energy Economics 27, 849–856. Ang, J.B., 2008. Economic development, pollutant emissions and energy consumption in Malaysia. Journal of Policy Modeling 30, 271–278. 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. Chang, T., Fang, W., Wen, L.F., 2001. Energy consumption, employment, output and temporal causality: evidence from Taiwan based on cointe-

611

gration and error-correction modeling techniques. Applied Economics 33, 1045–1056. Chen, S.T., Kou, H.I., Chen, C.C., 2007. The relationship between GDP and electricity consumption in 10 Asian countries. Energy Policy 35 (4), 2611–2621. Dunkerley, J., 1982. Estimation of energy demand: the developing countries. Energy Journal 3, 79–99. Engle, R.F., Granger, C.W.J., 1987. Cointegration and error correction: representation, estimation and testing. Econometrica 55, 251–276. Gan, P.Y., Li, Z., 2008. An econometric study on long-term energy outlook and the implications of renewable energy utilization in Malaysia. Energy Policy 36, 890–899. Ghosh, S., 2009. Electricity supply, employment and real GDP in India: evidence from cointegration and Granger-causality tests. Energy Policy 37, 2926–2929. Halicioglu, F., 2009. An econometric study of CO2 emissions, energy consumption, income and foreign trade in Turkey. Energy Policy 37, 1156–1164. Hoa, T.V., 1993. Effects of oil on output growth and inflation in developing countries: the case of Thailand. International Journal of Energy Research 17, 29–33. Hondroyiannis, G., Lolos, S., Papapetou, E., 2002. Energy consumption and economic growth: assessing the evidence from Greece. Energy Economics 24, 319–336. Johansen, S., Juselius, K., 1990. Maximum likelihood estimation and inference and cointegration with applications to the demand for money. Oxford Bulletin of Economics and Statistics 52, 169–210. Johansen, S., 1988. Statistical analysis of cointegration vectors. Journal of Economic Dynamics and Control 12, 231–254. Jumbe, C.B.L., 2004. Cointegration and causality between electricity consumption and GDP: empirical evidence from Malawi. Energy Economics 26, 61–68. Karki, S.K., Michael, D.M., Hossein, S., 2005. Energy and environment in the ASEAN: challenges and opportunities. Energy Policy 33, 499–509. Keong, C.Y., 2005. Energy demand, economic growth, and energy efficiency—the Bakun dam-induced sustainable energy policy revisited. Energy Policy 33, 679–689. Lee, C.C., 2005. Energy consumption and GDP in developing countries: a cointegrated panel analysis. Energy Economics 27, 415–427. Lee, C.C., Chang, C.P., 2008. Energy consumption and economic growth in Asian economies: a more comprehensive analysis using panel data. Resource and Energy Economics 30, 50–65. Mah, J.S., 2000. An empirical examination of the disaggregate import demand of Korea-—the case information technology products. Journal of Asian Economics 11, 237–244. Mahadevan, R., Asafu, A.J., 2007. Energy consumption, economic growth and prices: a reassessment using panel VECM for developed and developing countries. Energy Policy 35, 2481–2490. Masih, A.M.M., Masih, R.A., 1998. Multivariate cointegrated modeling approach in testing temporal causality between energy consumption, real income and prices with an application to two Asian LDC’s. Applied Economics 30, 1287– 1298. Masih, A.M.M., Masih, R.A., 1996. Energy consumption, real income and temporal causality: results from a multi-country study based on cointegration and errorcorrection modeling techniques. Energy Economics 18, 165–183. Mehrara, M., 2007. Energy consumption and economic growth: the case of oil exporting countries. Energy Policy 35, 2939–2945. Mohamed, A.B., Lee, K.T., 2006. Energy for sustainable development in Malaysia: energy policy and alternative energy. Energy Policy 34, 2388–2397. Mozumder, P., Marathe, A., 2007. Causality relationship between electricity consumption and GDP in Bangladesh. Energy Policy 35, 395–402. Narayan, P.K., 2004. Do public investments crowd out private investments? Fresh evidence from Fiji. Journal of Policy Modelling 26, 747–753. Narayan, P.K., 2005. The saving and investment nexus for China: evidence from unit roots, cointegration and Granger causality. Applied Economics 37, 1979–1990. Narayan, P.K., Narayan, S., 2005a. An empirical analysis of Fiji’s aggregate import demand function. Journal of Economic Studies 32, 158–168. Narayan, P.K., Narayan, S., 2005b. Estimating income and price elasticities of imports for Fiji in a cointegration framework. Economic Modelling 22, 423–438. Narayan, P.K., Singh, B., 2007. The electricity consumption and GDP nexus for the Fiji Islands. Energy Economics 29, 1141–1150. Narayan, P.K., Smyth, R., 2006a. What determines migration flows from low income to high income countries? An empirical investigation of Fiji-–US migration 1972–2001. Contemporary Economic Policy 24, 332–342. Narayan, P.K., Smyth, R., 2006b. Higher education, real income and real investment in China. Education Economics 14, 107–125. Narayan, P.K., Smyth, R., 2009. Multivariate Granger causality between electricity consumption, exports and GDP: evidence from a panel of Middle Eastern countries. Energy Policy 37, 229–236. Odhiambo, N.M., 2009. Energy consumption and economic growth nexus in Tanzania: an ARDL bounds testing approach. Energy Policy 37, 617–622. Oh, W., Lee, K., 2004. Causal relationship between energy consumption and GDP revisited: the case of Korea 1970–1999. Energy Economics 26, 51–59. Payne, J.E., 2009. On the dynamics of energy consumption and output in the US. Applied Energy 86, 575–577. Pedroni, P., 1999. Critical values for cointegration tests in heterogeneous panels with multiple regressors. Oxford Bulletin of Economics and Statistics 61, 653–670.

ARTICLE IN PRESS 612

V.G.R. Chandran et al. / Energy Policy 38 (2010) 606–612

Pedroni, P., 2004. Panel cointegration: asymptotic and finite sample properties of pooled time series tests with an application to the PPP hypothesis. Econometric Theory 20 (3), 597–625. Pesaran, M.H., Shin, Y., Smith, R.J., 2001. Bounds testing approach to the level relationship. Journal of Applied Econometrics 16, 289–326. Pusat Tenaga Malaysia, 2004. National Energy Balance 2004 Malaysia. Pusat Tenaga Malaysia:, Selangor. Reimers, H.E., 1992. Comparisons of tests for multivariate cointegration. Statistical Papers 33, 335–359. Reinsel, G.C., Ahn, S.K., 1992. Vector autoregressive models with unit roots and reduced rank structure: estimation, likelihood ratio test, and forecasting. Journal of Time Series Analysis 13, 353–375. Sari, R., Soytas, U., 2009. Are global warming and economic growth compatible? Evidence from five OPEC countries. Applied Energy 86, 1887–1893. Shrestha, R.M., Anandarajah, G., Liyanage, M.H., 2009. Factors affecting CO2 emission from the power sector of selected countries in Asia and the Pacific. Energy Policy 37, 2375–2384.

Smulders, S., Michiel, N., 2003. The impact of energy conservation on technology and economic growth. Resource and Energy Economics 25, 59–79. Squalli, J., 2006. Electricity consumption and economic growth: bounds and causality analyses of OPEC members. Energy Economics 29 (6), 1192–1205. Stern, D.I., 2000. A multivariate cointegration analysis of the role of energy in the US macroeconomy. Energy Economics 22, 267–283. Tang, C.F., 2008. A re-examination of the relationship between electricity consumption and economic growth in Malaysia. Energy Policy 36, 3077–3085. Toda, H.Y., Yamamoto, T., 1995. Statistical inference in vector autoregressions with possibly integrated process. Journal of Econometrics 66 (1–2), 225–250. Yoo, S.H., 2006. The causal relationship between electricity consumption and economic growth in the ASEAN countries. Energy Policy 34, 3573–3582. Yuan, J., Zhao, C., Yu, S., Hu, Z., 2007. Electricity consumption and economic growth in China: cointegration and co-feature analysis. Energy Economics 29 (6), 1179–1191.