Dynamics of energy use, technological innovation, economic growth and trade openness in Malaysia

Energy xxx (2015) 1e11

Contents lists available at ScienceDirect

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

Dynamics of energy use, technological innovation, economic growth and trade openness in Malaysia Kazi Sohag a, b, *, Rawshan Ara Begum a, **, Sharifah Mastura Syed Abdullah a, Mokhtar Jaafar b a b

Institute of Climate Change (IPI), Universiti Kebangsaan Malaysia, 43600 UKM, Bangi, Selangor, Malaysia Faculty of Social Science and Humanities, Universiti Kebangsaan Malaysia, 43600 UKM, Bangi, Selangor, Malaysia

a r t i c l e i n f o

a b s t r a c t

Article history: Received 4 September 2014 Received in revised form 16 May 2015 Accepted 27 June 2015 Available online xxx

This study extends the Marshallian demand framework to investigate the effects of TI (technological innovation) on energy use in Malaysia. This extended theoretical frameworks predicts that TI, an exogenous element in the energy demand function, increases energy efficiency and, correspondingly, reduces energy consumption at a given level of economic output. Using an ARDL (autoregressive distributed lag) bounds testing approach for the sample period 1985e2012, this study confirms both short- and log-run theoretical predictions. However, controlling for the effect of TI, this study finds that increasing GDP per capita and trade openness produce a rebound effect of TI on energy use. © 2015 Elsevier Ltd. All rights reserved.

Keywords: Technological innovation Energy use Trade openness ARDL (autoregressive distributed lag) Malaysia

1. Introduction Globally, both economic and population growth continue to be the most important drivers of increases in CO2 emissions from fossil fuel combustion. Despite the UNFCCC (United Nations Framework Convention on Climate Change) and the Kyoto Protocol, GHG emissions grew more rapidly between 2001 and 2010 than during the previous decade. The main contributors to this trend were higher energy demand associated with rapid economic and population growth as well as increased use of fossil fuels (IPCC, WG3, 5AR). Thus, energy efficiency improvements and fugitive emissions reductions during fuel extraction and within energy conversion, transmission, and distribution systems; fossil fuel switching; and low carbon technologies, such as RE (renewable energy) sources, are important to reducing energy sector emissions while addressing the dilemmas of economic growth, energy security, and environmental sustainability. Implementation of energy efficiency measures varies greatly by country as well as by sector and

* Corresponding author. Faculty of Social Science and Humanities, and Institute of Climate Change (IPI), Universiti Kebangsaan Malaysia, 43600 UKM, Bangi, Selangor, Malaysia. ** Corresponding author.

industry, especially if developing countries are taken into account [64]. The scale of economic activity is strongly associated with energy use [9,55] however, the elasticity of economic growth to energy use is heterogeneous among countries because a country's growth pattern depends on its economic activities. Al-Mamun et al. [3] also stressed that energy use is quite sensitive to sectoral domination in the economy because industrialized and sophisticated service-oriented economies consume higher shares of the global energy supply compared to countries with primary sectorebased (agricultural) economies, which consume relatively less energy. Since the early 1980s, the economy of Malaysia has, to a great extent, transformed from an agricultural to a sophisticated industrial and service economy [11] whose industrial and service sectors contribute 40.2% and 50.2% of total GDP, respectively. This change increased per capita energy use from 660.32 units to 2639.43 units between 1980 and 2012. Like economic growth, trade influences domestic energy use through several channels, such as economies of scale, composite effects of production factors, and technological effects [52,57]. For instance, export demand augments the scale of economic activities, which consequently increases energy usage in the domestic country [15]. However, the composite effects of production factors can either promote or hinder the use of energy, depending on three main processes. First, trade may foster specialisation in the local

http://dx.doi.org/10.1016/j.energy.2015.06.101 0360-5442/© 2015 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Sohag K, et al., Dynamics of energy use, technological innovation, economic growth and trade openness in Malaysia, Energy (2015), http://dx.doi.org/10.1016/j.energy.2015.06.101

2

K. Sohag et al. / Energy xxx (2015) 1e11

economy by changing the factors of production (such as capital, labour, and energy) as the economy transitions from agriculture to industry and services. Thus, energy dynamics (patterns of energy use) change through specialization of the economy [25]. Second, trade increases local market competitiveness through international market competition [57] and thus improves the efficiency of production factors. Finally, trade liberalization promotes the diffusion of technology, which is often characterized as energy efficient, from developed countries to less developed countries [60], helping to reduce the energy consumption required to produce a given level of output. Although trade has continuously accounted more than 100% of GDP in Malaysia since the early 1980s, the impact of trade on energy use in Malaysia has not been studied; hence, this article incorporates trade openness into an energy function to explore this nexus. TI (technological innovation) is crucial for improving energy efficiency [20,26,61,28]. Although there are other methods of promoting energy efficiency, such as market-based approaches, policies and controls, the magnitude of the impact of technological innovation is larger due to its direct association with the energy efficiency function. In this case, advanced technologies allow the economy to produce a given level of output using a lower level of energy. Moreover, technological innovation provides opportunities for the economy to switch from depletable sources to renewable sources of energy to meet energy demands. However, technological innovation reduces energy consumption marginally; hence, it may not reduce a great share of the energy used. For instance, if the price of energy falls due to gains in energy efficiency, the reduced price might encourage economic agents to use more energy [1]. Malaysia has experienced considerable technological innovation with its rapid economic growth, moving towards achieving developed nation status by 2020. There is a trend of increasing technological innovation (as measured by the number of patents) in Malaysia from 1980 to 2012, as shown in Fig. 4. Thus, this study investigates the effects of TI (technological innovation) on energy use in Malaysia based on an extended Marshallian demand framework, as it is important to assess the dynamic effects of economic growth and trade openness as well as the effects of TI on energy use in Malaysia. No previous studies have focused on estimating the dynamic impacts of technical innovation on energy use exclusively for the Malaysian economy. This study assesses the impact of technological innovation on energy use by controlling for several important variables, i.e., GDP per capita, trade openness, and price. Employing time series data from the World Development Indicators (1980e2012), this study applies ARDL (autoregressive distributed lag) and dynamic OLS frameworks to explore the short- and longrun impacts of particular variables on energy use. There are few techniques for measuring the long-run relationships among the variables of interest: [65] and Johansen [66,67] are the most extensively applied approaches. The present study applied an ARDL bounds testing approach to the energy, growth, trade and technological nexus due to its favourable characteristics compared to other standard approaches, as highlighted in the methodology section. This study also investigates a possible structural break during the sample years by applying a [63] unit root test and a [24] residual-based test for co-integration in models with regime shifts. Moreover, we address the possibility of an endogeneity problem using simultaneous equations to analyse an energy demand-supply setup. The findings of this study should facilitate the adoption of energy policies to address the dilemmas of economic growth, energy security, and environmental sustainability in Malaysia. The remainder of this article is organized as follows. Section 2 provides the empirical model derived from demand and technological progress theory, the data and the methodology. Section 3

focuses on the empirical findings. Finally, Section 4 concludes and highlights some policy implications. 2. Review of the literature Numerous studies have investigated the energy, economic growth and technology nexus in the context of various countries. Each of these studies is a worthy exploration of these issues. For instance [33], investigated the causal relationship between energy use and economic growth by applying a heterogeneous panel cointegration approach in the context of ASEAN economies over the period 1971e2002. They found a unidirectional causal relationship from energy use to economic growth over the long run but an insignificant causal relationship over the short run. Asafu-Adjaye [8] found unidirectional causality from energy use to economic growth for India and Indonesia, whereas energy use influences economic growth in a bidirectional manner for Thailand and Philippines. Some studies have also found a unidirectional causal relationship between energy use and economic growth in oil exporting economies [37,47]. However, like economic growth, energy prices also help explain the energy demand function. By applying static and dynamic panel techniques [2], revealed that energy prices are negatively associated with natural gas and electricity demand in the US economy, while this relationship is positive for income. In Pakistan, there is a shortage of electricity supply, but electricity demand is elastic with respect to its price and income [27] hence, income taxes and rigid pricing policies would promote energy efficiency. However, in technologically advanced economies, energy becomes an important factor of production with no close substitutes; hence, energy prices and economic growth might not respond to energy demand. For instance [34], revealed that energy is an essential good with no close substitute; hence, economic growth and energy prices are inelastic in OECD economies. Similarly, [62] found that energy is an inelastic good in China due to its massive industrialization. Dahl [16] also revealed that energy sources, particularly gasoline and diesel, are essential products, and hence, the elasticities of price and income are insignificant in explaining energy use in 120 developing and developed countries. Except for domestic energy demand [4], concluded that neither the energy price nor economic growth is responsive to the import demand function for oil in the Turkish economy. Like economic growth, trade helps explain domestic energy dynamics. Sadorsky [46] observed that the volume of imports and exports increases domestic local energy use in Middle Eastern countries. However, trade also has a non-linear impact on energy use. For instance [52], revealed that trade and energy exhibit a Ushaped relationship in high-income countries and an inverted Ushaped relationship in middle- and low-income countries. Yanikkaya [60] highlighted that trade openness facilitates the penetration of technology from developed countries into developing countries. In line with this finding [57], argued that technological diffusion promotes energy efficiency when the diffusion takes place through trade openness in the context of European Union member countries. However, measures of trade openness, economic growth and human development influence energy use in the economies of Thailand, Indonesia and Malaysia [10]. Regarding the nexus between technological innovation and energy efficiency, technological innovation increases the quality of production by augmenting energy efficiency [13]. In fact, OECD countries experience greater energy efficiency gains due to their sizeable technological innovation compared to other developing countries [59]. Outside the OECD found an inverted U-shaped relationship between household final consumption and residential CO2 emissions due to the use of advanced household technologies

Please cite this article in press as: Sohag K, et al., Dynamics of energy use, technological innovation, economic growth and trade openness in Malaysia, Energy (2015), http://dx.doi.org/10.1016/j.energy.2015.06.101

K. Sohag et al. / Energy xxx (2015) 1e11

3

This study measures technological innovation as the number of patent applications following Madsen et al [69] and [6]. Therefore, it could be assumed that the parameter of the partial change in energy consumption due to changes in technology should be t negative, that is, vE vTt ¼ de < 0. Finally, this study incorporates trade openness into the empirical model. As discussed in the introduction section, the impact of trade openness can be either positive or vEt negative on energy use, that is, vTO ¼ trde ±. However, the negative t elasticity implies that trade introduces technologies into the local economy from more technologically advanced economies, which reduces energy use. In contrast, the positive elasticity of trade to energy use implies that energy use increases due to trade. Therefore, the empirical model for the energy demand function of Malaysia can written as follows:

in Malaysia. Technology may reduce energy efficiency marginally; however, in absolute terms, it may produce a rebound in overall energy use [1]. A review identified numerous studies of the factors that affect the adoption of energy efficiency policies or technologies [22,35,36,44,54,58]. These studies showed that differences in endowments, preferences, or technologies create differences in the adoption of energy efficient technologies across countries or individuals over time. Rates of adoption may also be influenced by market failures, such as environmental externalities, information access, and liquidity constraints in capital markets, and behavioural factors. Thus, there is much research on the potential of innovations to reduce production energy or emission, but evidence of the macro effects and the rebound effect remains very limited. Hence, this study conducted empirical analyses to explain the dynamic impacts of energy use by considering the critical factors of economic growth, trade openness and technological innovation, which will fulfil the need for research on energy and macro-economic rebound effects.

LnEt ¼ b0 þ b1 Ln GDPCt þ b2 LnTIt þ b3 Ln TOt þ εt

(3)

where LnEt ¼ logarithmic form of energy use (per capita) at time t; Ln GDPCt ¼ logarithmic form of GDP (per capita) at time; LnTIt ¼ technological innovation at time t; and Ln TOt ¼ logarithmic form of trade openness at time t.

3. Empirical model, data and methodology 3.1. Empirical model

3.2. Data and sources

Theoretically, standard energy demand is associated with income and energy prices (e.g., Refs. [41,49,56]. Assuming the market clearing condition, where energy demand equals energy consumption, the following energy demand function is written within the framework of the standard Marshallian demand [21] function at time t as:

This study employs annual time series data for Malaysia from 1980 to 2012. The data were collected mainly from the WDI (World Development Indicators) dataset. The variables of interest include energy use in kg of oil equivalent per capita (EU) as the dependent variable and the explanatory variables real GDPC (GDP per capita) and technological innovation (total number of patent applications). In addition, this study also considers trade openness (lnTO), which is calculated as the sum of the volume of real exports and imports and normalized by real GDP.

EDt ¼ f ðYt ; Pet Þ

(1)

where EDt ¼ energy demand at time; Yt ¼ income at time t; Pet ¼ energy price at time t Based on the discussion in the introduction section, trade openness and technological innovation can influence energy use in many ways; therefore, this study also considers them

EDt ¼ f ðYt ; Pet ; TIt ; TOt Þ

3.2.1. Economic growth As an emerging country in Asia, Malaysia is experiencing a smooth increasing trend in GDP (per capita). Fig. 1 shows that GDP per capita increased from USD 2318 to USD 6990 from 1980 to 2012, while the economy maintained on an average growth rate of approximately 3.50%. However, while GDP per capita smoothly increased from 1980 to 2012, it decreased drastically in 1998 and 2009. These unusual decreases in GDP per capita might be due to economic recessions.

(2)

Here, TIt ¼ technological innovation at time t and TOt ¼ trade openness at time t. According to the standard Marshallian demand function, invE t come elasticity of energy is positive Ett=vY =Yt ¼ εye > 0, whereas price vEt=vPet elasticity is negative Et =Pet ¼ εpe < 0. However, this study does not consider energy prices as independent factors due to their subsidized prices in Malaysia. [25,68] also revealed that technological innovation is a crucial factor in spurring economic growth by accelerating factor productivity and ensuring energy efficiency [6,12,30]. This study focuses on whether technological innovation reduces energy use, ceteris paribus (all other things constant).

3.2.2. Energy use trend Energy use per capita is constantly increasing with the changing of growth pattern of Malaysia. Due to the domination of sophisticated industrial and service sectors in the Malaysia economy, energy use has increased on a large scale. The development of information and communication technologies as well as technology-based household entertainment compounds energy use. Fig. 2 reflects an increase in energy use per capita from 800

8000 7000 6000 USD

5000 4000 3000 2000 1000 2012

2010

2011

2008

2009

2006

2007

2005

2003

2004

2002

2000

2001

1999

1997

1998

1995

1996

1993

1994

1991

1992

1989

1990

1987

1988

1986

1985

1983

1984

1981

1982

1980

0

Fig. 1. GDP per capita (constant 2005 USD).

Please cite this article in press as: Sohag K, et al., Dynamics of energy use, technological innovation, economic growth and trade openness in Malaysia, Energy (2015), http://dx.doi.org/10.1016/j.energy.2015.06.101

4

K. Sohag et al. / Energy xxx (2015) 1e11

Energy use

kg of oil equivalent

3000 2500 2000 1500 1000 500 0

Fig. 2. Trend of energy use per capita (kg of oil equivalent).

Trade (% of GDP) 250

Percentage

200 150 100 50

2011

2012

2010

2009

2008

2007

2006

2005

2004

2003

2001

2002

2000

1999

1998

1997

1996

1995

1994

1993

1992

1990

1991

1989

1988

1987

1986

1985

1983

1984

1981

1982

1980

0

Fig. 3. Trend of trade (share of GDP).

units to 2695 units over the period 1980 to 2012. Therefore, meeting increasing energy demand in this emerging country is of great concern. 3.2.3. Trade (share of GDP) trend Malaysia is a country thriving in trade, which is considered a prime growth engine. Fig. 3 indicates that trade (as a share of GDP) has persistently remained at over 100% since 1998 to 2012. The highest level of trade (as a share of GDP) is 200%, which was observed in 2000. However, the figure shows that trade (as a share of GDP) decreases slightly from 2006 to 2012 due to last global economic crisis. It is worth mentioning that the volume of exports reached 185.05 billion USD from 14.33 billion USD over the period 1980 to 2012. Therefore, it can be argued that local energy demand is largely derived from trade in Malaysia. 3.2.4. Technological innovation [32,70] argued that the number of patents could be considered a proxy for technology innovation because it indicates the interest of industrial and private organizations in exploring a new technology. In addition [50], states that technology innovation can be reflected by a quantitative indicator, such as the number of patents. Hence, following the empirical studies of [5,51,55] this study also considers the number of patents as a proxy for technology innovation. Fig. 4 clearly shows that the number of non-residential patents increased sharply, which implies that significant technological innovation took place in Malaysia during the study period. Therefore, it is worth measuring the dynamic effects of technological innovation on energy use. 3.3. Methodology This study employs ARDL technique by Ref. [43] to estimate the empirical model (equation (3)) for various reasons. With regard to methodology, the standard approaches to co-integration are Engle and Granger [65] and Johansen [66,67]. Because Engle and Granger (E-G) propose a bivariate technique, multivariate analysis is

excluded under this co-integration test. Conversely, the [29] technique is known as a system-based approach to co-integration. This is more efficient than the E-G approach, as it offers multiple cointegrating vectors. Unlike E-G, the Johansen [66,67] approach reduces omitted lagged variables bias by including the lag in the estimation. However, this approach is also criticized, as it is highly sensitive to the number of lags selected [23]. Moreover, interpretation often becomes difficult when more than one co-integrating vector exists in the model [5]. In the case of mixed orders of integration in the regressors, the validity of both the E-G and Johansen [66,67] techniques have been challenged. Thus, these techniques are only valid in cases with the same order of integration. In contrast, the ARDL technique overcomes the above criticisms through some key characteristics: (i) the co-integration relationship is estimated using OLS estimation, which is conducted after choosing the appropriate lag order for the model; (ii) notwithstanding the [29] approach, this technique remains statistically significant irrespective of the nature of the variables' integration i.e., I(0), I(1) or mutual co-integrated; and finally, it is imperative to mention that (iii) the test is necessary and valid for small and finite sample sizes. The ARDL version of the VECM (vector error correction model) can be specified as follows:1

1 Where Dln EUt represents first difference term of logarithm form of energy use per capita. b0 indicates the intercept. lnEUt1 , lnGDPCt1 , lnTPt1 and lnTOt1 indicate one year lag of logarithm form of energy use per capita, GDP per capita technological progress, and trade openness respectively, and b1 ; b2 ; b3 ; and b4 represent coefficient of Pp lnEUt1 ,lnGDPCt1 , lnTPt1 and lnTOt1 respectively. g DlnEUti ; i¼1 i Pq Pq Pq d DlnGDPC ; 4 DlnTP ; and h DlnTO indicate the summatm tj tl m¼1 m j¼1 j l¼1 l tion of the coefficients gi , dj ; 4l , and hm for the values of lag difference (D) lnEUti , lnGDPCtj , lnTPtl and lnTOtm respectively, where optimum lag is from 1 to p for dependent variable and from 1 to q for explanatory variable. Finally, εt indicates the error terms.

Please cite this article in press as: Sohag K, et al., Dynamics of energy use, technological innovation, economic growth and trade openness in Malaysia, Energy (2015), http://dx.doi.org/10.1016/j.energy.2015.06.101

K. Sohag et al. / Energy xxx (2015) 1e11

5

Trend of patent application in Malaysia 8000 6000 4000 2000 0

Patent applications, residents

Patent applications, nonresidents

Fig. 4. Trend of patent application in Malaysia.

respective variables along with their short-run adjustment rates towards the long run.3

DlnEUt ¼ b0 þ b1 lnEUt1 þ b2 lnGDPC t1 þ b3 lnTPt1 þ b4 lnTOt1 þ

p X

gi DlnEUti þ

i¼1

þ

q X

4l DlnTPtl þ

q X

dj DlnGDPCtj

Dln EUt ¼ b0 þ

j¼1 q X

hm DlnTOtm þ εt

þ (4)

3.3.1. Estimation procedure The estimation procedure begins with equation (4), which can be estimated using OLS to examine the likelihood of a long-run relationship among the variables where the Wald Test or F-test shows the overall significance of the coefficient of the lagged variables. In this case, the null hypothesis is no existence of cointegrating relationships among the regressors and regress (H0: b1 ¼ b2 ¼ b3 ¼ b4 ¼ 0Þ while the alternate hypothesis states the opposite (Ha: b1 sb2 sb3 sb4 s0). Then, the F-statistic is compared against the critical values of the upper and lower bounds [43]. On the one hand, if the F-statistic is higher than the upper critical value, then the null hypothesis shows no cointegrating relationship, which means that the hypothesis is rejected. On the other hand, if the F-statistic is less than the lower critical value, this indicates acceptance of the null hypothesis due to the absence of co-integrating relationship among respective variables. However, if the F-statistic is observed to be between the lower and upper critical values, then the test is inconclusive. When this procedure is completed, the next step is to proceed with the estimation of the long-run coefficient of the ARDL model according to equation (5).2

ln EUt ¼ b0 þ

p X

gi ln EUti þ

i¼1

þ

q2 X l¼0

4l lnTPtl þ

q1 X

dj ln tedR Square lnGDPCtj

j¼0 q3 X

gi Dln EUti

i¼1

m¼1

l¼1

p X

hm TOtm þ εt

(5)

m¼0

Finally, this study estimates the error correction model as presented in equation (6) to investigate the short-run dynamics of the

2 Where ln EUt indicates the logarithmic form energy use per capita, and b0 inPp Pq Pq dicates the intercept. g DlnEUti ; d DlnGDPCtj ; 4 DlnTPtl ; and i¼1 i j¼0 j l¼0 l Pq m¼0 hm DlnTOtm the values of lag difference(D) lnEUti ,lnGDPCtj , lnTPtl and lnTOtm respectively, where optimum lag is from 1 to p for dependent variable and from 0 to q for explanatory variable. Finally εt indicates the error terms indicate the summation of the coefficients gi , dj ; 4l , and hm for.

q X

4l DlnTPtl þ

l¼1

q X

dj DlnGDPCtj

j¼1 q X

hm DTOtm þ wemct1 þ εt

m¼1

(6) Finally, following [42] this study employs cumulative sum of recursive residuals (CUSUM) and cumulative sum of squares of recursive residuals (CUSUMSQ) tests to examine the stability of the coefficients.

4. Results and discussion 4.1. Order of integration of the respective variables Prior to testing for co-integration, this study conducts a test of the order of integration for each variable using ADF (Augmented Dickey-Fuller), DFGLS (Dickey-Fuller Generalized Least Squared), and PhillipsePerron (PeP) tests (Table 1). The tests are conducted to make sure that no variable surpasses the order of integration I(1) and to justify the appropriateness of the ARDL approach. Table 1 shows that the logarithmic form of GDP per capita is non-stationary at level but becomes stationary after considering the first difference, which is confirmed by all three approaches to unit root tests. Similarly, per capita energy use and trade openness are non-stationary at level but stationary after taking the first difference at the 1 per cent significance level. However, DF-GLS and PeP unit root tests reveal that logarithmic form of technological progress is stationary at I(0), while the ADF test rejects it. However, both the ADF and PeP tests show that technological innovation is stationary after the first difference, while the DF-GLS test rejects it. Therefore, the presence of such mixed orders of integration, as reported in Table 1, supports the application of the ARDL approach as opposed to standard approaches.

3 Where Dln EUt indicates the difference values of logarithmic form of energy use Pp P per capita, and b0 indicates the intercept. g ln EUti ; q1 d i¼1 i j¼1 j P Pq3 lnGDPCtj ; q2 4 lnTP ; and h TO represent the summation of the long tm tl m¼1 m l¼1 l run coefficients gi , dj ; 4l , and hm for the values of lnEUti ,lnGDPCtj , lnTPtl and lnTOtm respectively, where optimum lag is from 1 to p for dependent variable and from 1 to q for explanatory variable. Finally εt indicates the error terms. Finally, w is the coefficient of the one-year lag values of error correction mechanism (ecmt1).

Please cite this article in press as: Sohag K, et al., Dynamics of energy use, technological innovation, economic growth and trade openness in Malaysia, Energy (2015), http://dx.doi.org/10.1016/j.energy.2015.06.101

6

K. Sohag et al. / Energy xxx (2015) 1e11

Table 1 Results of Augmented Dickey-Fuller (ADF), Dickey-Fuller Generalized Least Squared (DFGLS), and PhillipsePerron (PeP) unit root tests. Log levels (Zt)

Log 1st difference (Zt)

Variable

ADF

DFGLS

PeP

Variable

ADF

DFGLS

PeP

LGPDC LNEU LNTO LTI

0.597 1.129 1.789 2.472

0.363 0.387 1.321 2.475**

0.600 1.260 1.390 4.439**

DLNGDPC DEU DTO DLP

4.674*** 5.797*** 3.613** 10.434***

4.075*** 4.816*** 3.466*** 1.063

4.674*** 6.084*** 3.609*** 17.307***

(**) and (***) represent 5% and 10% level of significance, respectively.

Table 2 Selection criteria of VAR lag order. Lag

LogL

LR

FPE

AIC

SC

HQ

0 1 2 3 4

59.613 139.136 151.417 164.062 198.556

NA 125.911* 15.351 11.590 20.121

1.140 5.880* 9.120 1.740 9.010

4.634 9.928 9.618 9.338 10.879*

4.438 8.946* 7.851 6.786 7.541

4.582 9.667 9.149 8.661 9.997*

*indicates lag order selected by the criterion, LR: sequential modified LR test statistic (each test at the 5% level), FPE: Final prediction error, AIC: Akaike information criterion, SC: Schwarz information criterion, HQ: Hannan-Quinn information criterion.

4.2. Co-integration and long-run impact of technological progress and control variables Given the mixed order of integration of respective variable, the application of the bounds testing approach to co-integration is suitable. However, estimating equation (4) requires identifying the optimum lag. Estimating the VAR model indicates that the optimum lag is 1, according to the Schwarz information criterion (Table 2). To examine the co-integrating relationship, a Wald joint significance test (F-statistic) is applied on the coefficients of the lagged variables. In this regard, the null hypothesis (b1 ¼ b2 ¼ b3 ¼ b4 ¼ 0) of no co-integration is examined against the alternative hypothesis of co-integration in the model. To determine the total number of co-integrating vectors, this study calculates four F-statistics by normalizing each variable as the dependent variable. Finally, all calculated F-statistics are compared with [43] critical value (Table 3). The results in Table 3 suggests that the calculated F-statistic is 3.974** {FlnEUlnEUjlnGDPC, lnTI, lnTO)}, which is larger than the upper bound critical value of [43] at the 5 per cent significance level. It indicates that the null hypothesis of no co-integration is rejected. Thus, the long-run co-integrating relationship among the respective variables is recognized when lnEU is normalized as regressive. Nevertheless, when GDP per capita is considered the dependent variable, the calculated F-statistic falls below the lower bound critical value, implying the non-existence of a co-integration relationship. Conversely, when total patent applications (LTP) is

considered the dependent variable, the calculated F-statistic falls between the upper and lower bound critical values; hence, whether a co-integration relation exists is inconclusive. Finally, when technological innovation is normalized, the F-statistic (FlnTI (lnTIjlnGDPC, lnEU, lnTO) ¼ 7.728) surpasses the critical value of [43] at the 1 per cent significance level, which confirms the existence of co-integration in the model. The main finding of this analysis implies that energy use, GDP per capita, trade openness and technological progress are co-integrated in Malaysia. Because a co-integration relationship is identified, this study proceeds to estimate equation (6) following the lag specification 1, 0, 1, 0 (Table 4). Interestingly, the long-run estimation of the ARDL framework reveals that the coefficient of GDP (per capita) is 0.913, which is also significant at the 10% level. This finding implies that a 1-unit increase in GDP (per capita) is associated with an increase in energy (per capita) of 0.913 units, all other things constant. This finding is consistent from both theoretical (Marshallian demand theory) and empirical points of view. A number of empirical studies find similar results, e.g., Refs. [9,18,39,40,55]. The positive but highly significant coefficient of trade openness indicates that trade openness is also responsible for increased energy use in Malaysia. This result does not support [60] argument that trade openness makes technologies readily available to a country from trading partners. Instead, trade openness positively and significantly influences long-run energy demand by augmenting local energy demand. As expected, the sign of technological progress is negative, but the validity of such a finding largely depends on the probability; this study indicates that technological progress reduces energy consumption at the 12 per cent level of significance. It is worth performing a robustness check using a DOLS framework to observe whether these findings change. 4.3. Impact of technological innovation long control variable over the short run This study estimates the short-run impact of the respective variables on energy use along with an error correction mechanism (ecm) using equation (6). Similar to the long-run phenomena, a 1-unit increase in GDP per capita leads an approximately 0.53-unit increase in energy use

Table 3 Result of ARDL bounds test (equation (5)). Dep. var.

SIC Lag

F-stat.

Probability

Outcome

FlnEU(lnEUjlnGDPC, lnTP, lnTO) FlnGDPC(lnGDPCjlnEU, lnTP, lnTO) FlnTO(lnTOjlnGDPC, lnEU,lnTI) FlnTI(lnTIjlnGDPC, lnEU, lnTO) Critical value, Pesaran et al. [43]

1 1 1 1 I(0)

3.974** 1.110 3.465 7.728*** I(1)

0.020 0.386 0.032 0.001

Cointegration No Cointegration Inconclusive Cointegration

1 percent significance level 5 percent significance level 10 percent significance level

3.29 2.56 2.20

4.37 3.49 3.09

(**) and (***) represent 5% and 10% level of significance, respectively.

Please cite this article in press as: Sohag K, et al., Dynamics of energy use, technological innovation, economic growth and trade openness in Malaysia, Energy (2015), http://dx.doi.org/10.1016/j.energy.2015.06.101

K. Sohag et al. / Energy xxx (2015) 1e11 Table 4 Long run coefficients using the ARDL approach, (1,0,1,0) selected based on SIC, dependent variable is LnEU. Regressor

Coefficient

Standard error

T-ratio [Prob]

lnGDPC lnTO lnTP C

0.913*** 0.265** 0.053 1.068*

0.087 0.112 0.033 0.535

10.418[0.000] 2.349[0.029] 1.619[0.120] 1.996[0.059]

(**) and (***) represent 5% and 10% level of significance, respectively.

Table 5 Error Correction Representation of ARDL Model (1,0,1,0) selected based on SIC: Dependent variable is D LNEU. Regressor

Coefficient

Standard error

T-ratio [Prob]

D lnGDPC D lnTO D lnTI DC

0.533*** 0.142 0.031* 0.623* 0.583***

0.130 0.143 0.018 0.355 0.115

4.082[0.000] 0.988[0.334] 1.718[0.100] 1.752[0.094] 5.036[0.000]

ecm(-1)

(**) and (***) represent 5% and 10% level of significance, respectively.

(Table 5). Conversely, the coefficient of trade openness is negative and statistically insignificant, implies that no effect has taken place in energy use due to changes in trade openness. Likewise, the longand short-run coefficients of LTP are also negative and significant at the 10 per cent level, which implies that technological progress promotes the efficiency of energy use. The negative and highly significant coefficient of the error correction mechanism implies that short-run disequilibrium adjusts by 58% per cent towards the long-run equilibrium. 4.4. Diagnostic test results of the model Equation (5) has passed several diagnostic tests that endorse the validity of the model. The values of R2 and Adjusted R2 are 0.987 and 0.985, respectively, which indicates that the model is well fitted. Table 6 reveals the model does not have serial correlation, normality, functional error or heteroscedasticity problems.

7

The CUSUM (cumulative sum) and CUSUMQ (cumulative sum of squares) Fig. 5 from a recursive estimation of the model also indicate the model is stable, as the residuals are within the critical bounds at the 5% significance level.

4.5. Robustness check 4.5.1. Energy-technological innovation nexus: application of dynamic ordinary least squares The robustness of the coefficients in Table 4 obtained from the long-run ARDL estimator is evaluated by applying an alternative single equation estimator, a DOLS (dynamic OLS) procedure. The primary benefit of the DOLS approach is that it also considers the presence of a mix order of integration of the respective variables in the co-integrated framework. The estimation of DOLS involves regressing one of the I(1) variable against other I(1) and I(0) variables by including leads (p) and lags (-p) in the framework [7]. Thus, this estimator solves possible endogeneity and small sample bias problems. Moreover, the obtained co-integrating vectors from DOLS estimators are asymptotically efficient. The re-estimated equation (4) using a dynamic OLS approach is presented in Table 7. It indicates that the magnitudes of each variable coefficient change slightly from the ARDL estimators. Specifically, the signs and significance levels of the coefficients of lnGDPC and trade openness remain the same as in the ARDL model. However, as expected, the coefficient of technological innovation becomes negative and statistically significant at the 1 per cent level. This result indicates the technological innovation promotes energy efficiency. As a further robustness check, this study sketches the impact of each regressor on the regressive variable using a quadratic prediction plot with CIs (confidence intervals). Fig. 6 highlights the consistency of the findings produced by the ARDL and DOLS estimators, that is, an increase in lnGDPC leads to an increase in energy use. Similarly, Fig. 7 indicates trade openness fosters energy use in the Malaysian economy. Finally, the graphical representation shows that technological innovation reduces energy use (Fig. 8).

Table 6 Diagnostic test of ARDL model. Test-statistic R-squared Serial correlation c2 (1) Functional form c2 (1)

P-value

0.987 0.200 0.029

R-bar-squared Normality c2 (2) Heteroscedasticity c2 (1)

0.654 0.864

Plot of Cumulative Sum of Recursive Residuals

Test-statistic

P-value

0.985 1.083 0.196

0.582 0.657

Plot of Cumulative Sum of Squares of Recursive Residuals

15

1.5

10

1.0

5 0

0.5

-5

0.0

-10 -15 1987

1992

1997

2002

2007

The straight lines represent critical bounds at 5% significance level

2012

-0.5 1987

1992

1997

2002

2007

2012

The straight lines represent critical bounds at 5% significance level Fig. 5. Stability test.

Please cite this article in press as: Sohag K, et al., Dynamics of energy use, technological innovation, economic growth and trade openness in Malaysia, Energy (2015), http://dx.doi.org/10.1016/j.energy.2015.06.101

8

K. Sohag et al. / Energy xxx (2015) 1e11

Table 7 Dynamic OLS: dependent variable LNEU. Variables

Coefficient

Robust standard error

Z

P-Value

lnGDPC lnTO lnTI lnGDPC FD D1 LD lnTO FD D1 LD lnTI FD D1 LD Constant R-squared

0.997*** 0.236*** 0.095***

0.065 0.037 0.027

15.33 6.36 3.51

0.000 0.000 0.000

0.093 0.345*** 0.037

0.060 0.089 0.051

1.53 3.88 0.73

0.126 0.000 0.462

0.031 0.294*** 0.348

0.128 0.084 0.145

0.25 3.48 2.40

0.806 0.000 0.017

0.034* 0.018 0.001 1.278 0.982

0.019 0.022 0.008 0.619

1.82 0.82 0.21 2.06

0.069 0.414 0.833 0.039

Robust standard errors in parentheses ***p < 0.01, **p < 0.05, *p < 0.1, FD, D1 and LD indicates lag difference, lead difference, first difference and lag difference.

7.5 7 6.5

Energy Consumption

8

4.5.2. Energy-technological innovation nexus: assumption of structural breaks The study period includes changes in energy policies in Malaysia. For example, the government of Malaysia implemented

the Four Fuel Diversification Policy in 1981, Five Fuel Policy in 2001 and Biofuel Policy in 2006 (NC2, 2010). This study contextualizes these changes using Zivot-Andrews unit root test [63]. Table 8 shows that the variables of interest present unit root properties at level but are stationary with one endogenous structural break at first difference. The results also provide evidence about the timing of the structural break. For instance, most notable structural breaks in GDP per capita, energy use, trade openness and technological innovation took place in 1998, 1989, 1987 and 2007, respectively. Because we find the strong evidence of structural breaks for all variables, we further evaluate whether these variables are cointegrated under the assumption of endogenous structural breaks. This study employs a GregoryeHansen co-integration approach [24]. Table 9 presents the result, which is consistent with regard to long-run relationships under the assumption of a level change. The table reports an ADF statistic that confirms the existence of cointegration under the assumption of a change in level. However, no co-integration is found when the GregoryeHansen cointegration approach considers the assumption of a change in regime or of a change in regime and trend. This implies that structural breaks took place as an intercept change while the slope of the model remains same, which further implies that the equilibrium equation shifts in parallel manner [24]. Table 6 also detects the year of break, which occurred around 2000. A closer look at the energy literature indicates that energy generating capacity increased by approximately 20% since 2000 [38].

7.8

8

8.2

GDP Per Capita

8.4

8.6

8.8

7.5 7 6.5

Energy Consumption

8

Fig. 6. Energy consumption and GDP per capita.

4.6

4.8

5 Trade Openness

5.2

5.4

Fig. 7. Energy consumption and trade openness.

Please cite this article in press as: Sohag K, et al., Dynamics of energy use, technological innovation, economic growth and trade openness in Malaysia, Energy (2015), http://dx.doi.org/10.1016/j.energy.2015.06.101

9

.02 -.02

0

Energy Consumption

.04

.06

K. Sohag et al. / Energy xxx (2015) 1e11

-2

-1

0

Technological Progress

1

2

3

Fig. 8. Energy consumption and technological progress.

4.5.3. Energy-technological innovation nexus: assumption of endogeneity It should be mentioned that ARDL bounds testing co-integration and DOLS can provide unbiased coefficients in the presence of possible endogeneity problems by using sufficiently long lags [7,48]. However, this study applies a simultaneous estimation Table 8 ZivoteAndrews structural break unit root test. Variable

Z&A test for level

lnGDPC lnEU lnTO lnTI

Z&A test for 1st difference

T-statistic

TB

Outcome

T-statistic

TB

Outcome

3.157 4.685 1.779 3.630

1991 1991 1988 1987

Unit Unit Unit Unit

5.993*** 6.653*** 5.129*** 7.299***

1998 1989 1987 2007

Stationary Stationary Stationary Stationary

root root root root

Note ***, **, & * indicate 1%, 5%, & 10% significance level respectively.

Table 9 GregoryeHansen test of cointegration with regime shifts: model: change in level. Test

Statistic

Breakpoint

Date

1%

5%

10%

ADF 5.46** 21 2000 5.77 Zt 4.31 22 2001 5.77 Za 24.03 22 2001 63.64 GregoryeHansen test for cointegration with regime shifts, regime and trend

5.28 5.02 5.28 5.02 53.58 48.65 model: change in

ADF 3.39 21 2000 6.51 Zt 4.91 14 1993 6.51 Za 29.25 14 1993 80.15 GregoryeHansen test for cointegration with regime shifts, regime and trend

6.00 5.75 6.00 5.75 68.94 63.42 model: change in

ADF Zt Za

5.92 5.19 30.97

15 14 14

1994 1993 1993

6.89 6.89 90.84

6.32 6.32 78.87

6.16 6.16 72.75

Note ***, **, & * indicate 1%, 5%, & 10% significance level respectively.

Table 10 Result from simultaneous estimation.

Qd¼a1 þb1 priceþb2 price of competeing productþb2 incomeþb3 technology þb4 tradeþε1 Qs¼a2 þb3 priceþb4price

of

raw

material

þb3 technology þb4 tradeþε1

Equilibrium condition Qd ¼ Qs : where Qd is quantity demanded and Qs is quantity supplied. In this framework, price is determined simultaneously with demand and supply. The statistical implication is that price is not a predetermined variable and that it is correlated with the disturbances of both equations. However, it is rare that quantity is related to two disturbances. This phenomenon poses no problem, as the errors in the regression are specified in the behavioural demand and supply equations. Therefore, this study places price on the left-hand side by making this endogeneity explicit in the specification. However, energy prices are subsidized in the Malaysia; hence, this study uses the CPI (consumer price index) and PPI (producer price index) as proxies for the energy and raw material prices, respectively. This study estimates simultaneous equations by taking CPI as instrumented and lnGDPC, lnTO, lnTI and PPI as instruments. The solution to the simultaneous equations is perfectly consistent with the results of the ARDL and DOLS estimators. Table 10 indicates that energy use is positively associated with income and trade. As expected, the table also shows that the coefficient of technological innovation is negative and significant, which implies that it promotes energy efficiency in the Malaysian economy. 5. Conclusion

DV: lnEU

Coef.

Std. err.

Z

P-value

CP lnGDPC lnTO lnTI Constant

0.262 0.879 0.141 0.037 1.444

0.293 0.239 0.061 0.018 0.626

0.90 3.68 2.32 1.99 2.30

0.371 0.000 0.020 0.047 0.021

Instrumented: CP. Instruments: LNGDPC LNTO LTI PP. R square 0.84.

technique to check the robustness of the previous estimator. In doing so, we consider some other important determinants of energy use, e.g., income, trade, and price. However, energy use is determined by the market mechanism, while the quantities demanded and supplied of energy are at equilibrium given the demand price and supply price. Therefore, it is expected that this model faces an endogeneity problem. However, this study also addresses endogeneity by applying simultaneous equations.

This article demonstrates that rapid economic growth (e.g., GDP per capita) and trade openness are significant factors in increasing energy use in Malaysia during the study period. However, technological innovation helps to reduce energy use by increasing the energy efficiency of production processes, which ultimately reduces emissions. The empirical analysis of this study has produced several interesting findings. First, an increase in GDP per capita augments energy use over both the short and long run in the

Please cite this article in press as: Sohag K, et al., Dynamics of energy use, technological innovation, economic growth and trade openness in Malaysia, Energy (2015), http://dx.doi.org/10.1016/j.energy.2015.06.101

10

K. Sohag et al. / Energy xxx (2015) 1e11

Malaysian economy. However, the magnitude of the impact of GDP per capita on energy use is higher over the long run than over the short run. Second, trade openness also increases domestic energy use over the long run in the Malaysian economy. The empirical findings indicate that technological innovation plays an important role in reducing energy use and improving energy efficiency. The study findings are very much in line with the IPCC's latest report (IPCC, wg3), which indicates that technological innovation and diffusion support overall economic growth and determine both the energy intensity of economic output and the carbon intensity of energy. However, regardless of the energy efficiency gains produced by technological innovation, economic growth and trade openness produce rebound effects in energy use. The application of a structural break co-integration framework also found cointegrating relationships among energy use, innovation and other controls variables. Finally, the study provided robust and consistent empirical findings by employing various methods, e.g., ARDL, DOLS, GregoryeHansen and simultaneous estimation techniques. Although this study did not consider CO2 emission as a variable, the negative coefficient of technological innovation indicates that meeting Malaysia's goals of reducing CO2 emissions and accelerating economic growth would be feasible through larger-scale substitution of older technologies with energy efficient technologies. This shift may be enhanced by implementing joint public-private investments and initiatives to promote research and development for innovation in renewable and energy efficient technologies. According to the IPCC Fifth Assessment Report (IPCC, WG3, AR5, 2014), the largest emerging economies have all built effective systems for the development and deployment of new technologies, including low emissions technologies. Thus, policy makers can take serious action on climate change mitigation by improving energy efficiency and increasing the share of renewable energy use in the energy sector. However, this shift can also be accomplished through the implementation of the Malaysian government's policies, such as the National Green Technology Policy, National Policy on Climate Change and National Renewable Energy Policy and Action Plan. Acknowledgement The authors are thankful for the research grants ‘Fundamental Research Grant Scheme (FRGS)’ under the Ministry of Education, Malaysia (Project Code: FRGS/1/2013/TK07/UKM/02/4) and ‘Research Development Fund/Dana Pembangunan Penyelidikan for research group TJ’ (DPP-2014-164). References [1] Greening AL, Greene DL, Difiglio C. Energy efficiency and consumptiondthe rebound effectda survey. Energy Policy 2000;28(6):389e401. [2] Alberini A, Filippini M. Response of residential electricity demand to price: the effect of measurement error. Energy Econ 2011;33(5):889e95. [3] Al-Mamun M, Sohag K, Mia MAH, Uddin GS, Ozturk I. .Regional differences in the dynamic linkage between CO2 emissions, sectoral output and economic growth. Renew Sustain Energy Rev 2014;38:1e11. [4] Altinay G. Short-run and long-run elasticities of import demand for crude oil in Turkey. Energy Policy 2007;35(11):5829e35. [5] Ang JB. CO2 emissions, research and technology transfer in China. Ecol Econ 2009;68(10):2658e65. [6] Ang JB. Financial development, liberalization and technological deepening. Eur Econ Rev 2011;55(5):688e701. [7] Ang JB. Financial development and economic growth in Malaysia. Routledge Stud Growth Econ Asia 2008. [8] Asafu-Adjaye J. The relationship between energy consumption, energy prices and economic growth: time series evidence from Asian developing countries. Energy Econ 2000;22(6):615e25. [9] Athukorala PPA, Wilson C. Estimating short and long-term residential demand for electricity: new evidence from Sri Lanka. Energy Econ 2010;32:S34e40. [10] Azam M, Khan AQ, Zaman K, Ahmad M. Factors determining energy consumption: evidence from Indonesia, Malaysia and Thailand. Renew Sustain Energy Rev 2015;42:1123e31.

[11] Begum RA, Sohag K, Abdullah SMS, Jaafar M. CO2 emissions, energy consumption, economic and population growth in Malaysia. Renew Sustain Energy Rev 2015;41:594e601. [12] Bottazzi L, Peri G. The international dynamics of R&D and innovation in the long run and in the short run. Econ J 2007;117(518):486e511. [13] Brock WA, Taylor MS. Economic growth and the environment: a review of theory and empirics. Handb Econ Growth 2005;1:1749e821. [15] Cole MA. Does trade liberalization increase national energy use? Econ Lett 2006;92(1):108e12. [16] Dahl CA. Measuring global gasoline and diesel price and income elasticities. Energy Policy 2012;41:2e13. [18] Dergiades T, Tsoulfidis L. .Estimating residential demand for electricity in the United States, 1965e2006. Energy Econ 2008;30(5):2722e30. [20] Fisher-Vanden K, Jefferson GH, Liu H, Tao Q. What is driving China's decline in energy intensity? Resour Energy Econ 2004;26(1):77e97. [21] Friedman M. The Marshallian demand curve. J Political Econ 1949;57(6):463. [22] Gillingham K, Newell RG, Palmer K. Energy efficiency economics and policy (No. w15031). National Bureau of Economic Research; 2009. [23] Gonzalo J. Five alternative methods of estimating long-run equilibrium relationships. J Econ 1994;60(1):203e33. [24] Gregory AW, Hansen BE. Residual-based tests for cointegration in models with regime shifts. J Econ 1996;70(1):99e126. [25] Grossman GM, Helpman E. Trade, knowledge spillovers, and growth. Eur Econ Rev 1991;35(2):517e26. [26] Hang L, Tu M. The impacts of energy prices on energy intensity: evidence from China. Energy Policy 2007;35(5):2978e88. [27] Jamil F, Ahmad E. Income and price elasticities of electricity demand: aggregate and sector-wise analyses. Energy Policy 2011;39(9):5519e27. [28] Jiahua P, Guiyang Z, Yan Z, ShouXian Z, Qianyi X. Clarification of the concept of low-carbon economy and analysis of its core elements. Int Econ Rev 2010;4: 88e102. [29] Johansen S, Juselius K. Maximum likelihood estimation and inference on cointegrationdwith applications to the demand for money. Oxf Bull Econ Stat 1990;52(2):169e210. [30] Jones A. An environmental assessment of food supply chains: a case study on dessert apples. Environ Manag 2002;30(4):560e76. [32] Kumaresan N, Miyazaki K. An integrated network approach to systems of innovationdthe case of robotics in Japan. Res Policy 1999;28(6):563e85. [33] Lee CC, Chang CP. Energy consumption and economic growth in Asian economies: a more comprehensive analysis using panel data. Resour Energy Econ 2008;30(1):50e65. [34] Lee CC, Chiu YB. Modeling OECD energy demand: an international panel smooth transition error-correction model. Int Rev Econ Finance 2013;25:372e83. [35] Linares P, Labandeira X. Energy efficiency: economics and policy. J Econ Surv 2010;24(3):573e92. [36] Matisoff DC. The adoption of state climate change policies and renewable portfolio standards: regional diffusion or internal determinants? Rev Policy Res 2008;25(6):527e46. [37] Mehrara M. Energy consumption and economic growth: the case of oil exporting countries. Energy Policy 2007;35(5):2939e45. [38] Mohamed AR, Lee KT. Energy policy for sustainable development in Malaysia. In: The Joint International Conference on “Sustainable Energy and Environment (SEE)” 1-3 December 2004, Hua Hin, Thailand; 2004. [39] Nakajima T, Hamori S. Change in consumer sensitivity to electricity prices in response to retail deregulation: a panel empirical analysis of the residential demand for electricity in the United States. Energy Policy 2010;38(5):2470e6. [40] Narayan PK, Prasad A. Electricity consumptionereal GDP causality nexus: evidence from a bootstrapped causality test for 30 OECD countries. Energy Policy 2008;36(2):910e8. [41] Nicholson W. Intermediate microeconomics and application. 2000. [42] Pesaran MH, Pesaran B. Working with Microfit 4.0: interactive econometric analysis; [Windows version]. Oxford University Press; 1997. [43] Pesaran MH, Shin Y, Smith RJ. .Bounds testing approaches to the analysis of level relationships. J Appl Econ 2001;16(3):289e326. [44] Popp D. International technology transfer, climate change, and the clean development mechanism. Rev Environ Econ Policy 2011;5(1):131e52. [46] Sadorsky P. Trade and energy consumption in the Middle East. Energy Econ 2011;33(5):739e49. [47] Salahuddin M, Gow J. .Economic growth, energy consumption and CO2emissions in Gulf Cooperation Council countries. Energy 2014;73:44e58. [48] Samargandi N, Fidrmuc J, Ghosh S. Financial development and economic growth in an oil-rich economy: the case of Saudi Arabia. Econ Model 2014;43:267e78. [49] Samuelson L. Microeconomic theory. Kluwer-Nijhoff; 1986. [50] Schmoch U. Indicators and the relations between science and technology. Scientometrics 1997;38(1):103e16. [51] Schmoch U. Double-boom cycles and the comeback of science-push and market-pull. Res Policy 2007;36(7):1000e15. [52] Shahbaz M, Nasreen S, Ling CH, Sbia R. Causality between trade openness and energy consumption: what causes what in high, middle and low income countries. Energy Policy 2014;70:126e43. [54] Stern DI. The role of energy in economic growth. Ann N Y Acad Sci 2011;1219(1):26e51. [55] Tang CF, Tan EC. Exploring the nexus of electricity consumption, economic growth, energy prices and technology innovation in Malaysia. Appl Energy 2013;104:297e305.

Please cite this article in press as: Sohag K, et al., Dynamics of energy use, technological innovation, economic growth and trade openness in Malaysia, Energy (2015), http://dx.doi.org/10.1016/j.energy.2015.06.101

K. Sohag et al. / Energy xxx (2015) 1e11 [56] Varian HR, Repcheck J. Intermediate microeconomics: a modern approach, vol. 6. New York, NY: WW Norton & Company; 2010. [57] Wan J, Baylis K, Mulder P. Trade-facilitated technology spillovers in energy productivity convergence processes across EU countries. In: Energy Econ; 2015. [58] Wei M, Patadia S, Kammen DM. Putting renewables and energy efficiency to work: how many jobs can the clean energy industry generate in the US? Energy Policy 2010;38(2):919e31. [59] Wong SL, Chang Y, Chia WM. Energy consumption, energy R&D and real GDP in OECD countries with and without oil reserves. Energy Econ 2013;40: 51e60. [60] Yanikkaya H. Trade openness and economic growth: a cross-country empirical investigation. J Dev Econ 2003;72(1):57e89. [61] Zhou N, Levine MD, Price L. Overview of current energy-efficiency policies in China. Energy Policy 2010;38(11):6439e52. [62] Zhou S, Teng F. Estimation of urban residential electricity demand in China using household survey data. Energy Policy 2013;61:394e402.

11

[63] Zivot E, Andrews DWK. Further evidence on the great crash, the oil-price shock, and the unit-root hypothesis. J Bus Econ Stat 2002;20(1):25e44. [64] Sanstad AH, Roy J, Sathaye JA. Estimating energy-augmenting technological change in developing country industries. Energy Econ 2006;28(5):720e9. [65] Engle RF, Granger CW. Co-integration and error correction: representation, estimation, and testing. Econometrica: J Econometric Soc 1987:251e76. [66] Johansen S. Statistical analysis of cointegration vectors. J Econ Dyn Control 1988;12(2):231e54. [67] Johansen S. Estimation and hypothesis testing of cointegration vectors in Gaussian vector autoregressive models. Econometrica: J Econometric Soc 1991:1551e80. [68] Rivera-Batiz LA, Romer PM. Economic integration and endogenous growth (No. w3528). National Bureau of Economic Research; 1990. [69] Madsen JB, Ang JB, Banerjee R. Four centuries of British economic growth: the roles of technology and population. J. Econ. Growth 2010;15(4):263e90. [70] Kortum S. Equilibrium r&d and the patenter&d ratio: Us evidence. Am Econ Rev 1993:450e7.

Please cite this article in press as: Sohag K, et al., Dynamics of energy use, technological innovation, economic growth and trade openness in Malaysia, Energy (2015), http://dx.doi.org/10.1016/j.energy.2015.06.101