Estimation of potential energy saving and carbon dioxide emission reduction in China based on an extended non-radial DEA approach

Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Contents lists available at ScienceDirect

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

Estimation of potential energy saving and carbon dioxide emission reduction in China based on an extended non-radial DEA approach Yiwen Bian a,n, Ping He b, Hao Xu c a

Sydney Institute of Language & Commerce, Shanghai University, Shanghai 201800, PR China School of Management, University of Science and Technology of China, Hefei 230026, PR China c School of Business, Anhui University, Hefei, Anhui Province 230039, PR China b

H I G H L I G H T S








A non-radial DEA model treating non-fossil energy as a fixed input is developed. A method of measuring potential reductions in energy and CO2 emissions is proposed. Reducing coal consumption helps to identify much more inefficiency in China. Adjusting energy structure is a practical way of reducing CO2 emissions in China. Technology innovation is an important way to improve the performance in China.

art ic l e i nf o

a b s t r a c t

Article history: Received 19 June 2012 Accepted 12 August 2013

In the process of setting operational targets to achieve sustainable development of economy, environment and natural resources, estimation of potential energy saving and potential CO2 emission reduction becomes extremely important. This estimation can be conducted based on the energy efficiency evaluation for different decision-making units (DMUs) by data envelopment analysis (DEA). Non-fossil energy is an important component of energy consumption in China, and it has great impacts on energy efficiency and energy-related carbon dioxide (CO2) emissions. This paper proposes a non-radial DEA model to evaluate regional energy efficiencies in China. In the proposed model, non-fossil energy is treated as a fixed input. Based on the model, a method of measuring potential energy saving and CO2 emission reduction for efficiency improvement is also presented. The proposed approaches are illustrated by using a regional dataset in China. Based on the application results, some implications for improving energy efficiency and reducing CO2 emissions in China are provided. & 2013 Elsevier Ltd. All rights reserved.

Keywords: Data envelopment analysis (DEA) Energy saving Carbon dioxide emission reduction

1. Introduction In past three decades, rapid economic growth has caused great energy consumption and serious environmental and ecological problems in China (Li and Oberheitmann, 2009). China has become the second largest energy-consuming country and the largest emitter of carbon dioxide (CO2) in the world. In 2009, the total energy consumption reached 3.07 billion tons of standard coal equivalent (SCE), which is about 5.37 times that of 1978 (Statistical Year Book of China in 2010). China′s gross domestic product (GDP) accounts for no more than 6.2% of the world′s total GDP, while its carbon emissions account for 20.85% of the world′s total carbon

n

Corresponding author. Tel.: þ 86 21 69980028; fax: þ 86 21 69980017. E-mail addresses: [email protected], [email protected] (Y. Bian), [email protected] (P. He).

emissions (World Bank, 2009). To achieve sustainable developments of economy, environment and natural resources, Chinese government has in recent years implemented various strategies and policies, e.g., closing down backward production facilities, promoting the use of energy-saving technologies and making fiscal and tax policies for energy saving (Jiang et al., 2010; Hou et al., 2011), to save energy consumption and to reduce carbon emissions. Particularly, on the 2009 Copenhagen conference, China announced to reduce the intensity of CO2 emissions per unit of GDP by 40–45% by 2020 compared to the level in 2005. Note that, these strategies and policies should be carried out across all regions in China. It is well known that CO2 emissions are largely attributed to burning fossil energy consumption. Therefore, improving energy efficiency has often been recognized as one of the most cost-effective ways to reduce CO2 emissions and to increase the security of energy supply (Ang et al., 2010; Al-Mansour, 2011).

0301-4215/$ - see front matter & 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.enpol.2013.08.051

Please cite this article as: Bian, Y., et al., Estimation of potential energy saving and carbon dioxide emission reduction in China based on an extended non-radial DEA approach. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.08.051i

2

Y. Bian et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Fig. 1. Energy consumption in China during 1978–2009.

Energy consumption structure in China is composed of fossil energy and non-fossil energy, which has three evident features. First, energy consumption exhibits an overdependence on coal, and little utilization of natural gas and non-fossil energy in China, as shown in Fig. 1. Fig. 1 describes energy consumption structure in China during 1978–2009 (the data is collected from Statistical Year Book of China in 2010). It can be observed in Fig. 1 that fossil energy (including coal, oil and natural gas) is the main source of energy consumption in China, while non-fossil energy (including hydropower, nuclear power, wind power, solar and others) accounts for a very low proportion of total energy consumption. In 2009, coal, oil and natural gas respectively account for approximately 70.4%, 17.9% and 3.9% of the total energy consumption. Second, non-fossil energy has developed significantly in recent years. For example, since the renewable energy law was implemented in 2006, the increasing rate of wind power has been over 100% and China has taken the second position in the newly installed capacity in the world (Jiang et al., 2010). China has stated that by the year of 2020, 15% of primary energy consumption should come from nonfossil energy (Guo et al., 2011; Wang et al., 2011). Third, there exist great disparities in energy structure among different regions in China. For instance, coal consumption in Shanghai accounts for about 50% of its total energy consumption; while in Anhui province, this proportion reaches 90% in 2008. In Fujian province, the percentages of fossil energy and hydropower in its total energy consumption are 84% and 15.8%, respectively (Wang et al., 2011). Fossil energy consumption rather than non-fossil energy consumption is a primary driver of CO2 emissions. Different energy structures (i.e., the percentages of energy sources in total energy consumption) result in different carbon emission structures (i.e., the percentages of CO2 emissions related to coal, oil and natural gas). Therefore, energy structure has significant impacts on regional energy efficiencies and CO2 emissions. This raises three important issues: (1) How to discriminate the effects of fossil energy and non-fossil energy on regional efficiencies and CO2 emissions? (2) How to measure the efficiency of each type of fossil energy? (3) How to measure potential energy saving and CO2 emission reduction? These issues need to be effectively addressed before making appropriate policies for energy saving and CO2 emission reduction in China. In the literature, various approaches have been explored to evaluate energy efficiency or environmental efficiency at macro economy level in recent years. These existing approaches can be generally classified as parametric and non-parametric methods (Sadjadi and Omrani, 2008). Parametric approaches such as stochastic frontier analysis (SFA) measure performance through estimation of a restrictive production or cost function. Therefore, deviations in function forms affect results of such methods. The

non-parametric approaches, e.g., data envelopment analysis (DEA), evaluate performance based on a linear programming, which relies on construction of a piecewise linear combination of all observed inputs and outputs. A major advantage of the DEA approach is that it does not impose any functional form on the underlying technology (Zhang et al., 2011; Choi et al., 2012). Thus, comparing to parametric approaches, DEA can effectively avoid model misspecification (Wei et al., 2007; Chung, 2011). In addition, DEA can provide sufficient information for improving the efficiency of an inefficient decision making unit by slack and radial adjustments (Shi et al., 2010). With these methodological advantages, DEA has been widely applied to evaluate energy efficiencies or environmental efficiencies in recent years (Zhou et al., 2008). The existing studies on evaluating energy efficiency with CO2 emissions based on DEA approach can be mainly classified into three categories. The first one focuses on investigating the relationships among energy consumption, CO2 emissions and GDP growth for regions or counties, e.g., Ramanathan (2006), Lozano and Gutiérrez (2008) and Li et al. (2011). The second applies DEA approaches to compare efficiencies of energy or energy with carbon emissions for different regions (or countries), or to monitor the efficiency trends for regions or countries, e.g., Zhou et al. (2006), (2010), Chang and Hu (2010), Li (2010), Liou and Wu (2011), Zhang et al. (2011) and Wang et al. (2012). The third not only measures efficiencies for regions or countries, but also explores the potential targets of energy saving or carbon emission reduction by using DEA-based target setting approach, e.g., Hu and Wang (2006), Hu and Kao (2007), Zhou and Ang (2008), Shi et al. (2010), Guo et al. (2011), Lee et al. (2011) and Wei et al. (2012). The above studies have three outstanding features. The first is that, most of them treat energy consumption as an overall input variable in DEA models except Zhou and Ang (2008) who use non-radial measures for all energy inputs. The second is that in efficiency evaluation, the impact of non-fossil energy as an individual input on regional efficiencies is not taken into consideration. The third is that the targets setting for energy and CO2 emissions are obtained completely resting on DEA-based target setting approach, ignoring the effects of changes in energy structure on energy saving and CO2 emission reduction. One special case is Guo et al. (2011), who take energy structure adjustment into account in measuring CO2 emission reduction in China. However, it is not the same case as our problem. Thus it can be concluded that up till now there is no effective approach to simultaneously deal with the efficiency evaluation and estimation of potential energy saving and CO2 emission reduction issue in China. To reasonably evaluate regional energy efficiencies with CO2 emissions in China, the current paper proposes a non-radial DEA model based on environmental DEA technology (Färe and Primont, 1995; Färe and Grosskopf, 2004). Since CO2 emissions are mainly generated from fossil energy consumption rather than non-fossil energy consumption, to improve the energy efficiency, it is better to decrease fossil energy consumption as much as possible but not to reduce the non-fossil energy consumption in real production. As a result, we in the proposed model take non-fossil energy as a fixed input. The rest of the paper is organized as follows. Section 2 introduces a methodology for estimating CO2 emissions in China, constructs a non-radial DEA model for evaluating energy efficiencies of regions, and presents a method based on the proposed model for measuring potential energy saving and CO2 emission reduction. In Section 3, we illustrate the proposed approaches by using regional dataset in China. Conclusions are described in Section 4.

2. Methodology This section firstly presents a method for estimating CO2 emissions of regions in China, and secondly introduces a non-radial DEA

Please cite this article as: Bian, Y., et al., Estimation of potential energy saving and carbon dioxide emission reduction in China based on an extended non-radial DEA approach. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.08.051i

Y. Bian et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

model for performance evaluation incorporating all types of energy inputs and CO2 emissions. Based on the proposed model, this section provides a method to explore the potential targets of energy-saving and CO2 emission reduction for efficiency improvement. We consider that there are n independent regions, denoted by DMURj (j¼ 1, 2,…, n). In the process of regional production, each region employs labor, capital stock, energy consumption as inputs to produce both desirable and undesirable outputs. In the literature, when measuring regional or national energy efficiency or environmental efficiency, labor, capital stock and energy consumption are usually used as three major inputs; gross domestic product (GDP) and CO2 (CO2 emissions) are commonly used as a desirable output and an undesirable output, respectively (Li, 2010; Zhou et al., 2010; Guo et al., 2011; Choi et al., 2012; Wei et al., 2012; Wang et al., 2012). Particularly, Zhou and Ang (2008) analyze economy-wide energy performance of 21 OECD counties by adopting coal, oil, gas and other energy as four energy inputs, labor, capital stock as two non-energy inputs, GDP as a desirable output and CO2 as an undesirable output. To explore the efficiencies of each energy input of regions in China, we in the current study use labor (L) and capital stock (K) as two major non-energy inputs, coal, oil, natural gas and non-fossil energy (namely Ec, Eo, Eg and En, respectively) as four energy inputs, GDP (G) as a desirable output, and CO2 (Ce) as an undesirable output. 2.1. Estimation of CO2 emissions As we know, CO2 is mainly generated from the combustion of fossil energy. Therefore, CO2 emissions from each energy input can be estimated through multiplying the consumption of individual fossil energy input by its carbon emission coefficient (Liu et al., 2010; Guo et al., 2011; Wei et al., 2012; Li et al., 2012). Following the approach of Li et al. (2012), CO2 emissions can be calculated using the following equation: 3

Cej ¼ ∑ ðEij  F i  44=12Þ;

ð1Þ

i¼1

where Cej represents CO2 emissions (10,000 t) of DMUR j (j ¼ 1; 2; :::; n), 44/12 is the conversion coefficient between carbon and carbon dioxide, i is the index of different types of fossil energies (i.e., i¼coal, oil and natural gas), Eij represents the consumption of fossil energy i by DMURj measured by 104 t of standard coal equivalent, and Fi is the carbon emission coefficient of fossil energy i. According to the research results of Energy Research Institute (Dai et al., 2009; Li et al., 2012), Fi in the current study is assumed to be 0.7329, 0.565 and 0.445 for coal, oil and natural gas, respectively.

exhibiting CRS (constant returns to scale) can be expressed as T ¼ fðL; K; Ec; Eo; Eg; En; G; CeÞ : n

n

We know that a regional production is a joint-production process, i.e., CO2 emissions are generated when GDP is produced by consuming labor, capital and energy inputs. In this case, we cannot increase GDP and simultaneously reduce CO2 emissions. In order to reasonably deal with this efficiency evaluation problem, CO2 emissions should be modeled based on the concept of weak disposability of undesirable output (Färe and Primont, 1995; Färe and Grosskopf, 2004; Kuosmanen, 2005; Podinovski and Kuosmanen, 2011). As indicated by Zhou and Ang (2008) and Zhou et al. (2010), the production technology for modeling the joint production has been well defined in concept but it cannot be directly used in empirical efficiency measurement studies, and thus a popular practice is to model it in a DEA framework. Therefore, following the concept of weak disposability of undesirable output, the DEA reference technology, also called environmental DEA technology (Färe and Grosskopf, 2004; Zhou et al., 2008),

n

j¼1

j¼1

∑ λj Lj rL; ∑ λj K j r K; ∑ λj Ecj rEc; j¼1

n

n

n

j¼1

j¼1

j¼1

∑ λj Eoj rEo; ∑ λj Eg j r Eg; ∑ λj Enj r En; n

n

j¼1

j¼1

∑ λj Gj Z G; ∑ λj Cej ¼ Ce; λj Z 0; j ¼ 1; 2; …; ng:

ð2Þ

It is noted that in environmental DEA technology T, Ce is an undesirable output. In energy or environmental efficiency evaluation studies, various DEA models based on the environmental DEA technology have been widely explored and applied in recent years (Zhou et al., 2008). In the literature, there are five ways commonly used to measure the efficiencies of decision making units. The first is undesirable output orientation, which attempts to reduce undesirable outputs as much as possible for given levels of energy inputs, non-energy inputs and desirable outputs (Zhou et al., 2006, 2010; Lozano and Gutiérrez, 2008; Wang et al., 2012). The second is desirable output orientation, which aims at producing more desirable outputs when energy inputs, non-energy inputs and undesirable outputs are given (Lozano and Gutiérrez, 2008; Li, 2010; Wang et al., 2012). The third tries to increase desirable outputs and decrease undesirable outputs simultaneously while keeping energy and non-energy inputs constant (Färe and Grosskopf, 2004; Lozano and Gutiérrez, 2008; Wang et al., 2012). The fourth attempts to minimize the energy consumption with given non-energy inputs, desirable and undesirable outputs (Zhou and Ang, 2008; Wu et al., 2012). The fifth is to reduce energy inputs and undesirable outputs as much as possible while not optimizing other inputs and desirable outputs (Bian and Yang, 2010; Guo et al., 2011). In the case where the primary aim is to explore the efficiency of energy consumption with its related CO2 emissions of regions, we in the current study adopt the fifth way of efficiency measurement to model our problem, i.e., we would like to minimize the use of energy inputs and CO2 emissions in the process of efficiency evaluation for given levels of other inputs and desirable outputs. To investigate the effects of non-fossil energy on regional efficiencies, we first present a DEA model exhibiting constant returns to scale (CRS) for evaluating the efficiency of energy with its related CO2 emissions ignoring non-fossil energy, i.e., 1 Ecp ¼ min ðωc αcp þωo αop þ ωg αg p þ θp Þ 2 n

s:t: 2.2. The proposed efficiency evaluation model

3

∑ λj Lj þsl ¼ Lp ;

j¼1 n

∑ λj K j þ sk ¼ K p ;

j¼1 n

∑ λj Ecj þ sc ¼ αcp Ecp ;

j¼1 n

∑ λj Eoj þ so ¼ αop Eop ;

j¼1 n

∑ λj Eg j þ sg ¼ αg p Eg p ;

j¼1 n

∑ λj Gj sG þ ¼ Gp ;

j¼1 n

∑ λj Cej ¼ θp Cep ;

j¼1

λj Z 0; j ¼ 1; 2; …; n:

ð3Þ

In model (3), the subscript “p” represents the region to be evaluated. Model (3) is a non-radial DEA model, which has

Please cite this article as: Bian, Y., et al., Estimation of potential energy saving and carbon dioxide emission reduction in China based on an extended non-radial DEA approach. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.08.051i

Y. Bian et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

4

a stronger discriminating power in energy sources’ efficiencies comparisons than radial DEA has (Zhou and Ang, 2008). Note that ωc , ωo and ωg are normalized user-specified weights for coal, oil and natural gas, respectively, and ωc þωo þ ωg ¼ 1. Obviously, model (3) can measure efficiencies of energy sources with CO2 emissions by using non-proportional efficiency indices. As a result, model (3) allows energy inputs and CO2 emissions to be reduced with different proportions so as to make the evaluated region reach its ideal benchmarking point in the efficiency frontier of the best practice. Note in model (3) that the index Ecp lies in the interval (0,1]. If a region obtains a larger Ecp , it implies that this region performs better in terms of energy-saving and CO2 emission reduction. Let (αcnp , αonp , αg np , θnp , sln , skn , scn son , sgn , sG þ n , λnj ) (j¼ 1, 2,…,n) be the optimal solution to model (3). A region is said to be efficient, if Ecp ¼ 1 (i.e., αcnp ¼ 1,αonp ¼ 1,αg np ¼ 1 and θnp ¼ 1) and all slacks are equal to zero; otherwise, if Ecp o 1, i.e., there exists at least one efficiency index (αcnp , αonp , αg np or θnp ) less than 1, and (or) some slacks are not zero, then it is said to be DEA inefficient. Evidently, model (3) can measure the efficiencies of fossil energies with CO2 emissions while not taking non-fossil energy into consideration. As we know, fossil energy is non-renewable and it is a major driver of CO2 emissions, while non-fossil energy is renewable and its use does not generate CO2 emissions. As a result, we hope to reduce fossil energy consumption as much as possible while not decreasing non-fossil energy in real production. As indicated by Banker and Morey (1986) and Shi et al. (2010), non-fossil energy in this case can be treated as a fixed input in efficiency evaluation, i.e., a region can consume the greater quantity of the fixed non-fossil energy input but with a smaller quantity of fossil energy inputs as compared to the condition of controlling all energy inputs. Based on this point, we extend model (3) by incorporating non-fossil energy as a fixed input and get the following DEA model: 1 Efp ¼ min ðωc αcp þ ωo αop þ ωg αg p þ θp Þ 2 n

s:t:

∑ λj Lj þ sl ¼ Lp ;

j¼1 n

∑ λj K j þ sk ¼ K p ;

j¼1 n

∑ λj Ecj þ sc ¼ αcp Ecp ;

j¼1 n

∑ λj Eoj þ so ¼ αop Eop ;

j¼1 n

∑ λj Eg j þ sg ¼ αg p Eg p ;

j¼1

Csp ¼ ð1αcnp ÞEcp þ sc ; Osp ¼ ð1αonp ÞEop þso ; Gsp ¼ ð1αg np ÞEg p þ sg ; Cr p ¼ ð1θnp ÞCep :

ð5Þ

Note that in Eq. (5), Csp , Osp , Gsp and Cr p are saving targets for coal, oil, natural gas, and reduction target for CO2 emissions, respectively. Eq. (5) can provide targets of energy saving and CO2 emission abatement for efficiency enhancement. 2.3. A method for measuring energy saving and CO2 emission reduction Generally, the way of improving the performance of energy with its related CO2 emissions is to reduce energy consumption and CO2 emissions in production. As indicated by Guo et al. (2011), CO2 emissions can be reduced by DEA-based target setting approach and energy structure adjustment approach. This subsection proposes a method that combines these two approaches together to measure potential energy saving and CO2 emission abatement for the efficiency improvement in China. Note in Eq. (1) that, CO2 emissions change when energy structure changes, and this provides a potential way of adjusting energy structure to reduce CO2 emissions. The rationality of this is due to the fact that the higher emission coefficient for coal compared to those of oil and natural gas implies a way of abating CO2 emissions. To investigate the impacts of energy structure adjustment on regional efficiencies, energy-savings and CO2 emission reductions, we present the following two assumptions: (1) to meet the requirements of energy consumptions in regional production, total energy consumption of each region is assumed to be constant in the process of energy structure adjustment; (2) since carbon emission coefficient of coal is much higher than those of oil and natural gas, reduction in coal consumption and increase in oil, natural gas and non-fossil energy consumptions are required. Denote the adjustment parameters (i.e., the ratio of reduction or increase in a certain energy to total energy consumption of a region) of coal, oil, natural gas and non-fossil energy by πc, πo, πg and πn, respectively, and we have πc ¼ πo þπg þπn. On the basis of the above assumptions, we extend model (4) to the following programming: 1 Erp ¼ min ðωc αcp þ ωo αop þ ωg αg p þ θp Þ 2

n

∑ λj Enj þ sn ¼ Enp ;

n

j¼1

s:t:

n

∑ λj Lj þsl ¼ Lp ;

j¼1

∑ λj Gj sG þ ¼ Gp ;

n

j¼1

∑ λj K j þ sk ¼ K p ;

n

j¼1

∑ λj Cej ¼ θp Cep ;

n

j¼1

λj Z0; j ¼ 1; 2; :::; n:

Therefore, DEA can be used to survey potential energy-saving and CO2 emission reduction for inefficient regions. Based on the target setting approach, model (4) can determine potential energysaving and CO2 emission reduction of an inefficient region as

ð4Þ

Note in model (4) that non-fossil energy (En) is a fixed input which cannot be reduced by any efficiency measure. Let (αcnp , αonp , αg np , θnp , sln , skn , scn son , sgn , snn , sG þ n , λnj ) (j ¼ 1; 2; :::; n) be the optimal solution to model (4). If all efficiency indices (αcnp ,αonp ,αg np and θnp ) or (Efp ) are equal to unit and all slacks are equal to zero, then the evaluated region is DEA efficient; otherwise, it is inefficient. Based on DEA theory, inefficient regions can become efficient through slack, radial or non-radial adjustments (Shi et al., 2010).

∑ λj ðEcj πcn ET j Þ þ sc ¼ αcp ðEcp πcn ET p Þ;

j¼1 n

∑ λj ðEoj þ πon ET j Þ þ so ¼ αop ðEop þ πon ET p Þ;

j¼1 n

∑ λj ðEg j þ πg n ET j Þ þ sg ¼ αg p ðEg p þ πg n ET p Þ;

j¼1 n

∑ λj ðEnj þ πnn ET j Þ þ sn ¼ Enp þ πnn ET p ;

j¼1 n

∑ λj Gj sG þ ¼ Gp ;

j¼1

Please cite this article as: Bian, Y., et al., Estimation of potential energy saving and carbon dioxide emission reduction in China based on an extended non-radial DEA approach. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.08.051i

Y. Bian et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎ n

5

3. The application

∑ λj Ceaj ¼ θp Ceap ;

j¼1

πc ¼ πo þ πg þ πn; λj Z0; j ¼ 1; 2; :::; n:

ð6Þ

In model (6), ET j (j¼1, 2,…,n) is the total energy consumption of the jth region; CO2 emissions Ceaj are re-estimated by using Eq. (1) after energy structure changed. Let (αcnp , αonp , αg np , θnp , sln , skn , scn son , sgn , snn , sG þ n , λnj ) (j¼1, 2,…,n) be the optimal solution to model (6). If DMURp is inefficient (i.e., Erpn o 1), the potential reductions in fossil energy consumption and CO2 emissions can be obtained by the following equations: Csap ¼ ð1αcnp ÞðEcp πcn ET p Þ þ sc ; Osap ¼ ð1αonp ÞðEop þ πon ET p Þ þ so ; Gsap ¼ ð1αg np ÞðEg p þ πg n ET p Þ þ sg ; Cr ap ¼ ð1θnp ÞCeap þ Cep Ceap

ð7Þ

Note in Eq. (7) that CO2 emissions of an inefficient region can be abated through energy structure adjustment and DEA-based target setting approaches. As for an efficient region (i.e., Erpn ¼ 1), its CO2 emission reduction can be achieved by using energy structure adjustment, i.e., Cr ap ¼ Cep Ceap :

ð8Þ

Eqs. (7) and (8) provide the information of potential reductions in fossil energies and CO2 emissions for the efficiency improvement of efficient or inefficient regions in China. The implementation procedures of estimating energy saving and CO2 emission reduction for the efficiency improvement are summarized as follows. Step 1: Initialize the adjustment parameters πc, πo, πg and πn. πc, πo, πg and πn can also be regarded as user-specified parameters. In this way, we can perform sensitivity analysis on these parameters. Step 2: Re-estimate CO2 emissions Ceaj (j ¼ 1; 2; :::; n) of DMURj by using Eq. (1). Step 3: For given user-specified weights ωc , ωo and ωg , calculate regions’ efficiencies Erj n (j¼1, 2,…,n) by solving model (6). Step 4: Calculate energy-savings and CO2 emission reductions through using Eq. (7) for inefficient regions and Eq. (8) for efficient regions. After the calculation of energy-savings and CO2 emission reductions of all regions, the potential reductions in energy consumption and CO2 emissions of China can be obtained. It is noted that this approach can measure potential CO2 emission reductions for efficient and inefficient regions simultaneously.

In this section, through using the energy consumption, CO2 emission and economic production datasets of regions in mainland-China, we firstly compare efficiency results obtained from model (4) with those obtained from model (3) to illustrate the rationality of the proposed efficiency evaluation model; and then apply the method of measuring potential energy-saving and CO2 emission reduction to identify the targets for the efficiency enhancement in China. 3.1. Data source There are 31 regions (provinces, autonomous region and municipalities) in mainland China and one can refer to Hu and Wang (2006) for details. Each region utilizes multiple inputs to produce multiple outputs. In the current study, we take GDP as one desirable output, CO2 as one undesirable output, labor and capital stock (namely capital) as two non-energy inputs, coal, oil, natural gas and non-fossil energy as four inputs of end-use energy consumptions. The labor input is calculated as employee in all organizations and individual workers at the end of the current year, which is expressed in units of 10-thousand persons. GDP and capital are both stated in units of 100-million RMB Yuan. The units of coal, oil, natural gas and non-fossil energy are all 10thousand tons of standard coal equivalent. CO2 is described in units of 10-thousand tons. The data set of labor and GDP is collected from Statistical Yearbook of China 2010, and the data on energy inputs are obtained from China Energy Statistical Yearbook 2010. The data on capital stock in 1952 prices are obtained based on the results of Shan (2008), which are also used by Guo et al. (2011) and Wang et al. (2012). The data on CO2 emissions is estimated by using Eq. (1). Table 1 presents the summary statistics of all inputs and outputs. Tibet is excluded in this study, since its data on energy inputs is not available. In addition, because the data of capital stock of Chongqing is combined with Sichuan province, they together are regarded as one region in this research, namely Sichuan. Therefore, totally 29 regions are taken into account. 3.2. Efficiency results Evidently, the choice of preferred weights is a key issue in the applications of model (3) and model (4). Due to the fact that coal′s carbon emission coefficient (0.7329) is much larger than those of oil and natural gas (0.565 and 0.445, respectively), we should set the largest weight attached to coal, the second largest to oil, and the smallest to natural gas, respectively. For simplicity, we here can calculate the user-specified weights for coal, oil and natural gas based on their carbon emission coefficients, i.e., ωc ¼ 0:7329=ð0:7329 þ 0:565 þ 0:445Þ ¼ 0:4205; ωo ¼ 0:565=ð0:7329 þ 0:565 þ 0:445Þ ¼ 0:3242; ωg ¼ 1ωc ωo ¼ 0:2553:

Table 1 Summary statistics of inputs and outputs. Indicators Non-energy inputs Energy inputs

Desirable output Undesirable output

Labor Capital Coal Oil Natural gas Non-fossil energy GDP CO2

Max

Min

Mean

Std. Dev

6823.71 19757.71 24854.19 8390.24 2124.72 814.47 39482.56 82808.81

285.54 413.93 383.50 0.00 14.69 0.72 1081.27 3243.03

2579.04 6447.58 8649.44 2070.05 405.42 229.98 12581.46 27897.84

1815.19 5590.38 6232.74 2120.70 452.66 226.18 9890.97 18888.80

Please cite this article as: Bian, Y., et al., Estimation of potential energy saving and carbon dioxide emission reduction in China based on an extended non-radial DEA approach. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.08.051i

Y. Bian et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

6

Table 2 Efficiency results of regions in China. Region

East area

Central area

West area

a

Model (3)

Beijing Tianjin Shanghai Liaoning Hebei Shandong Jiangsu Zhejiang Fujian Guangdong Hainan Heilongjiang Jilin Inner M a Henan Shanxi Anhui Hubei Hunan Jiangxi Guangxi Sichuan Shaanxi Yunnan Gansu Xinjiang Guizhou Qinghai Ningxia

Model (4)

αcp

αop

αg p

θp

Ecp

αcp

αop

αg p

θp

Efp

1.0000 1.0000 1.0000 1.0000 0.3925 0.5343 1.0000 0.9689 1.0000 1.0000 1.0000 0.4585 0.4681 1.0000 0.5104 1.0000 0.8572 0.7089 0.8108 1.0000 1.0000 0.6760 0.2974 1.0000 0.5068 0.1543 0.5164 0.4123 0.1739

1.0000 1.0000 1.0000 1.0000 0.6038 0.4471 1.0000 0.5907 1.0000 1.0000 0.1261 0.2246 0.5691 1.0000 0.8224 1.0000 0.3898 0.6430 1.0000 1.0000 1.0000 1.0000 0.4104 0.9999 0.0494 0.2034 0.0571 0.6587 0.3346

1.0000 1.0000 1.0000 1.0000 0.4149 0.7687 1.0000 0.9629 1.0000 1.0000 0.0698 0.1831 0.3872 1.0000 0.1569 1.0000 0.6168 0.4208 0.6977 1.0000 1.0000 0.0777 0.4670 1.0000 0.0425 0.2908 0.6860 0.0988 0.0576

1.0000 1.0000 1.0000 1.0000 0.4082 0.5205 1.0000 0.8851 1.0000 1.0000 0.4411 0.4014 0.4799 1.0000 0.5191 1.0000 0.8311 0.6975 0.8238 1.0000 1.0000 0.6521 0.3294 1.0000 0.3493 0.1764 0.5171 0.3837 0.1799

1.0000 1.0000 1.0000 1.0000 0.4374 0.5432 1.0000 0.8649 1.0000 1.0000 0.4601 0.3569 0.4800 1.0000 0.5202 1.0000 0.7377 0.6557 0.8335 1.0000 1.0000 0.6402 0.3534 0.9999 0.2946 0.1907 0.4639 0.3979 0.1881

1.0000 1.0000 1.0000 1.0000 0.4039 0.5343 1.0000 1.0000 1.0000 1.0000 1.0000 0.4585 0.3322 1.0000 1.0000 1.0000 1.0000 0.7089 0.8108 1.0000 1.0000 0.6760 0.2974 1.0000 0.4897 0.1543 1.0000 0.4123 0.1706

1.0000 1.0000 1.0000 1.0000 1.0000 0.4471 1.0000 1.0000 1.0000 1.0000 0.1261 0.2246 0.8729 1.0000 1.0000 1.0000 1.0000 0.6430 1.0000 1.0000 1.0000 1.0000 0.4104 0.9999 0.0714 0.2034 1.0000 0.6587 0.3701

1.0000 1.0000 1.0000 1.0000 1.0000 0.7687 1.0000 1.0000 1.0000 1.0000 0.0698 0.1831 1.0000 1.0000 1.0000 1.0000 1.0000 0.4208 0.6977 1.0000 1.0000 0.0777 0.4670 1.0000 0.0798 0.2908 1.0000 0.0988 0.0656

1.0000 1.0000 1.0000 1.0000 0.4526 0.5205 1.0000 1.0000 1.0000 1.0000 0.4411 0.4014 0.4141 1.0000 1.0000 1.0000 1.0000 0.6975 0.8238 1.0000 1.0000 0.6521 0.3294 1.0000 0.3459 0.1764 1.0000 0.3837 0.1790

1.0000 1.0000 1.0000 1.0000 0.6010 0.5432 1.0000 1.0000 1.0000 1.0000 0.4601 0.3569 0.5460 1.0000 1.0000 1.0000 1.0000 0.6557 0.8335 1.0000 1.0000 0.6402 0.3534 0.9999 0.2977 0.1907 1.0000 0.3979 0.1937

Inner M indicates Inner Mongolia.

By using these weights, we calculate efficiencies for all regions under model (3) and model (4). The results are shown in Table 2. It should be stated that, four regions, i.e., Shanxi, Hainan, Sichuan and Guizhou, encounter infeasible solutions for their efficiencies, which may be due to their special energy structures. Shanxi and Guizhou are two important provinces in coal productions and consumptions in China. Coal consumption accounts for about 96% and 99% of their total energy consumptions respectively, while their end-use crude oil consumptions are both zero. In Sichuan, coal consumption dominates its energy consumption, and crude oil consumption only accounts for 2.84% of its total energy consumption. Different from these provinces, Hainan′s dominant energy consumption is crude oil, while coal consumption accounts for only 20.18% of its total energy consumption, which is much lower than that (70.4%) of the whole country. To deal with these infeasible issues, following Adler and Golany (2001), special energy inputs and CO2 emissions are slightly adjusted until the infeasibilities are eliminated. Note in Table 2 that, when non-fossil energy is ignored, 11 regions are deemed as efficient under model (3). However, we have 15 efficient regions under model (4) when non-fossil energy is taken into consideration, namely Beijing, Tianjin, Shanxi, Inner Mongolia, Liaoning, Shanghai, Jiangsu, Zhejiang, Anhui, Fujian, Jiangxi, Henan, Guangdong, Guangxi and Guizhou. These results show that, if non-fossil energy is not dealt with properly in performance evaluation, the ranking of regions' efficiencies may be severely distorted. This conclusion can be further verified by mean efficiencies obtained from the two models, i.e., mean efficiency score (0.7610) under model (4) is larger than that (0.7041) under model (3). Efficiencies reported in Table 3 show obvious geographic characters. The east area owns the largest portion of efficient regions (72.73%), which is much larger than those of the central

and the west areas (60% and 12.5%, respectively). Similarly, mean efficiency of the east area (0.8731) is higher than that of the central area (0.8392), and both of them are much higher than that of the west area (0.5092). These findings match the real conditions of regional developments in China. The east area is the most developed area in China, the GDP of which is around half of that in China, and the west area is the least developed area in China. In general, a more developed region can launch more upgraded production technologies and more advanced production processes to improve energy consumption performance and to abate carbon dioxide emissions than other regions can do. The proposed non-radial measure has a notable advantage over those radial ones (Hu and Wang, 2006; Hu and Kao, 2007; Shi et al., 2010 and Guo et al., 2011) in evaluating energy efficiency, i.e., more accurate information provided for efficiency improvement by using non-radial measure. It can be obviously observed from Table 2 that, there exist great disparities of efficiency scores of different energy inputs in different inefficient regions. For example, Hebei is evaluated as inefficient and its inefficiencies are mainly resulted from coal consumption and CO2 emissions. Thus, to improve Hebei′s efficiency, the main attention should be paid to making policies or adopting new production technologies to enhance the performance of coal consumption. As for Qinghai province, it is rated as inefficient in all types of energies and CO2 emissions. This indicates that, in order to improve its energy efficiency, policy making should take all energy inputs into consideration. Particularly the efficiency of its natural gas is 0.0988, which is much lower than those of coal and oil (0.4123 and 0.6587, respectively). In this case, Qinghai should pay more attention to enhancing the performance of natural gas than those to improving efficiencies of coal and oil. These results provide some important practical information for inefficient provinces

Please cite this article as: Bian, Y., et al., Estimation of potential energy saving and carbon dioxide emission reduction in China based on an extended non-radial DEA approach. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.08.051i

Y. Bian et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

with inequalities in energy inputs' efficiencies to improve their efficiencies and to reduce their CO2 emissions.

3.3. Estimation of energy saving and CO2 emission reduction In this sub-section, scenario analysis is undertaken to estimate potential energy saving and CO2 emission reduction. Here the scenarios are differentiated by different energy structures. The original energy structure is denoted as scenario 0, and other three scenarios are described as shown in Table 3. The scenarios assumed here are mainly based on the following three reasons. First, there is enormous potential for China to lessen its coal consumption at present, because the proportion of coal consumption in China as a primary energy source exceeds the global average level by 42% (National Development and Reform Commission, 2007). Second, the government has announced recently a prospective plan that non-fossil energy will account for approximately 15% of total energy consumption in 2020. Third, coal is still a major supply of energy in most of regions currently and the supply of oil and natural gas is relatively stable in China. In scenario 0, potential reductions in energies and CO2 emissions of all inefficient regions can be obtained by using Eq. (5). Then, we have total energy saving and CO2 emission reduction of three areas and the whole country. Fig. 2 shows the proportions of energy savings and CO2 emission reductions of three areas as well as the country. Note that in Fig. 2 coal, oil and natural gas can be saved about 21.27%, 24.52% and 36.36% of their total consumptions in China, i.e., 53394.60, 13995.18 and 4274.49 10-thousand tons of standard coal equivalent, respectively. This means that the total energy saving is about 71664.27 10-thousand tons of standard coal equivalent, which accounts for about 22.12% of the total energy consumption of the whole country. We can see that about 22.10% (179464.82 10-thousand tons) of the total CO2 emissions of the whole country can be abated. It can also be observed from Fig. 2 that, regional disparity in energy savings and CO2 emission reductions is evident. The east area has the highest potential to reduce its coal consumption (9.11%) and CO2 emissions (8.95%), and the lowest potential to save its natural gas consumption (3.36%). The west area can respectively reduce about 10.45% and 29.16% of its oil and natural gas Table 3 Parameters of the three scenarios.

πc πo πg πn

Scenario 1 (%)

Scenario 2 (%)

Scenario 3 (%)

10 0 5 5

15 2 8 5

20 3 10 7

7

consumptions, which are much larger than those in the east and central areas. Relative to other two areas, the central area has the lower potential percentages of its energy saving and CO2 emission abatement. It is obvious that natural gas consumption has the greatest potential to be reduced in production, especially in the west area. These disparities should be simultaneously taken into consideration when making policies to improve the performance of energy with CO2 emission reduction in China. The results shown in Fig. 2 are obtained only from those inefficient regions. This does not mean that there is no space of energy savings and CO2 emission reductions for those efficient regions, e.g., Beijing and Shanghai, but means that these efficient regions cannot further improve their performance compared with other regions under the current energy structure. However, when the energy structure changes, the results will also change. When scenario 1 is undertaken, efficiencies of some regions change greatly. Table 4 lists the efficiency results of those regions, where the number of efficient energy inputs are decreased in this scenario compared to scenario 0. We can find in Table 4 that, four regions, namely Zhejiang, Henan, Jiangxi and Guizhou, rated as efficient in scenario 0 are deemed as inefficient. It should be stated that Yunnan is evaluated as efficient in this scenario. These results mean that there are 12 efficient regions in scenario 1. Moreover, Hebei and Jilin measured as efficient in one or two energy inputs in scenario 0, are evaluated as inefficient in all energy inputs in scenario 1. Similar results can also be found in scenarios 2 and 3. Particularly, Shanxi becomes inefficient in scenarios 2 and 3. These results indicate that with current production technologies, more inefficiency hided by coal dominated energy structure can be identified when energy structure changes. This conclusion can be further verified by mean efficiencies of all regions in the two cases, i.e., mean efficiency (0.7067) in scenario 1 is lower than that (0.7610) in scenario 0. The changes in regional efficiencies lead to direct changes in regional potential energy savings and CO2 emission reductions via Eqs. (7) and (8). The proportions of potential energy savings and CO2 emission reductions of China and its three areas in scenario 1 are depicted in Fig. 3. In Fig. 3, T-Carbon dioxide and A-Carbon dioxide denote potential proportions of CO2 emission reduction to its total emissions in scenario 0 via target setting and energy structure adjustment, respectively; while coal, oil and natural gas indicate their potential saving proportions to their total consumptions in scenario 1, respectively. Note in Fig. 3 that, through reducing coal consumption by 10% and increasing both consumptions of natural gas and non-fossil energy by 5%, more potential proportion of coal saving (9.4%) of the whole country can be obtained in scenario 1 than that in scenario 0; while less potential proportion of natural gas saving (8.30%) is also detected. As for total potential energy saving, its proportion achieves 27.24% in this scenario, which is larger than Table 4 Efficiency results in scenario 1n.

Fig. 2. Proportions of energy savings and CO2 emission reductions in scenario 0.

Regions

αcp

αop

αgp

θp

Efp

Hebei Jilin Zhejiangn Jiangxin Henann Sichuan Guizhoun Mean

0.38804 0.27665 0.87426 0.83612 0.49233 0.6832 0.51449 0.6995

0.60383 0.70799 0.6707 0.91845 0.981 1.0000 0.057075 0.7466

0.4247 0.81309 1.0000 1.0000 0.42619 0.23506 0.5507 0.7315

0.40672 0.36578 0.83159 0.85275 0.51338 0.65399 0.51578 0.6904

0.4370 0.4596 0.8360 0.8787 0.5736 0.6627 0.4456 0.7067

n The sterisk indicates the regions rated as efficient under model (4). “Mean” here represents the average efficiency of all regions.

Please cite this article as: Bian, Y., et al., Estimation of potential energy saving and carbon dioxide emission reduction in China based on an extended non-radial DEA approach. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.08.051i

8

Y. Bian et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Fig. 3. Energy-savings and CO2 emission reductions in scenario 1.

Fig. 4. Energy-savings and CO2 emission reductions in scenarios 2 and 3.

that (22.12%) in scenario 0. It is remarkably observed that the sum of T-Carbon dioxide (27.36%) and A-Carbon dioxide (7.47%) is much larger than that (22.10%) in scenario 0, which means that CO2 emissions can be significantly reduced in scenario 1. It is clear that more potential of CO2 emission reduction of ACarbon dioxide can be identified by adjusting energy structure, which cannot be obtained by using those traditional DEA-based target setting approaches, e.g., Shi et al. (2010), Guo et al. (2011) and Wei et al. (2012). The reason is that the proposed approach can measure potential CO2 emission reductions not only for inefficient regions but also for efficient regions (e.g., Beijing, Shanghai and Tianjin). For example, when energy structure is changed to scenario 1, potential reduction of 824.99 10-thousand tons of CO2 emissions can be measured in Beijing. There are notable changes in energy savings and CO2 emission reductions in three areas when energy structure is adjusted. The east, central and west areas can decrease their coal consumptions by 11.86%, 9.53% and 9.28% in scenario 1, which are all larger than those (i.e., 9.11%, 4.83% and 7.33%) in scenario 0, respectively. Potential CO2 emission reductions of T-Carbon dioxide in these three areas, especially in the east and central areas, are higher than those in scenario 0, respectively; and the east area has the highest potential to reduce its CO2 emissions shown by T-Carbon dioxide in the three areas. Similar results can also be found in A-Carbon dioxide with scenario changes. The east area obtains the highest A-Carbon dioxide (3.37%), and the west area has the lowest A-Carbon dioxide (1.44%). All these results indicate that balancing energy structure by lessening coal consumption can help to identify more potential coal saving and CO2 emission reduction in China. These results may also partially reflect the reality, that the coal dominated energy structure causes much inefficiency in energy consumption, especially in those more developed regions. However, potential natural gas saving proportion of the west area declines remarkably, which is reduced from 29.16% to 16.83% with the change of energy structure. This may imply that the performance of natural gas in west regions can be improved when energy structure changes. These findings shown in Fig. 3 may also provide some important practical implications for the efficiency improvement in China, i.e., to effectively decrease energy consumption and CO2 emissions in China, reduction in coal consumption should be combined with advancements in production technologies simultaneously. The above indications for regional efficiency improvements can also be explored in scenarios 2 and 3. Significant changes of the potential energy saving and CO2 emission reduction for the whole country can be obviously found in these two scenarios. The results are illustrated in Fig. 4. Comparing Fig. 4 to Fig. 3, we find that potential savings of coal, oil and natural gas are all sharply decreased in scenarios 2 and 3 compared to scenario 1, while potentials of T-Carbon dioxide and

A-Carbon dioxide are all slightly increased. For instance, in scenario 2, coal, oil and natural gas can be lessened by 24.24%, 4.62% and 3.33% in production, which are all much lower than those in scenario 1, i.e., 30.66%, 24.70% and 28.05%, respectively. However, the potentials of T-Carbon dioxide (32.31%) and A-Carbon dioxide (9.21%) in scenario 2 are both larger than those (27.36% and 7.47%) in scenario 1, respectively. These results imply that energy structure adjustment is an effective way to identify more inefficiency information for improving the efficiency of China in practice. Particularly, if reduction targets of CO2 emissions in scenario 2 can be realized, needless to say, scenario 3, 41.53% of its emissions can be cut in the whole country. In this case, the goal of reducing carbon intensity by 40–50% can be reached in the future. Complete achievement of this goal will mainly depend on the development of non-fossil energy, improvements in oil and natural gas supplies, and reduction in coal supply by more than 15%. Moreover, we find that there are no dramatic differences in results between scenarios 2 and 3. This may indicate that when coal consumption decreased by about 20%, i.e., coal consumption accounts for about 50% of total energy consumption in China, the efficiencies become relatively stable. In this case, to improve the national efficiency, the more potential way may be the innovations of production and CO2 emission abatement technologies.

4. Conclusions The current paper proposes a non-radial DEA approach combining energy structure adjustment and DEA-based target setting together to measure energy saving and energy related carbon dioxide emission reduction in China. In the proposed approach, non-fossil energy incorporated as a fixed factor cannot be decreased in the efficiency optimization process. Based on the proposed model, a method for measuring potential energy-saving and CO2 emission reduction is developed. An application to regions in China is used to illustrate the proposed approaches. The application results demonstrate that the proposed approaches can be effectively applied to estimate potential reductions in energy consumption and its related CO2 emissions in China. Based on the application results, some conclusions can be drawn. First, since there are great disparities in potential energy savings and CO2 emission reductions in regions, policy making for the efficiency improvement should take these differences into consideration. Second, energy structure adjustment by decreasing coal consumption while increasing other energies’ supplies is an effective way to measure inefficiencies of coal dominated energy structure in China. Third, to effectively decrease CO2 emissions under the current production situation, reduction in coal consumption, the development of non-fossil energy and the slightly

Please cite this article as: Bian, Y., et al., Estimation of potential energy saving and carbon dioxide emission reduction in China based on an extended non-radial DEA approach. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.08.051i

Y. Bian et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

increasing supplies of oil and natural gas are required. Finally, energy structure adjustment combined with production technologies’ innovations (e.g., cleaning production) is a practical way to improve the performance of regions in China, especially in the case of coal′s reduction up to 20%. Note that, the above conclusions theoretically provide important information for China to draw up policies for reducing energy consumption and its related CO2 emissions. However, due to great unbalances of energy supplies, energy structures and economic structures of regions, it would be better for local governments to realize their targets for energy-savings and carbon reductions with their own policies. Therefore, considering all the above factors, some policy implications for energy-saving and CO2 emission reduction can be achieved as follows.


Energy structure adjustment by restricting coal supply and







promoting other energies’ supplies is an effective way to reduce CO2 emissions for most of Chinese regions, except some regions like Shanghai and Hainan. The reason is that, coal consumptions of Shanghai and Hainan account for about 53% and 20.18% of their total energy consumptions, which cannot be decreased in the current energy supply situations. As for Hainan province, it is better to reduce oil supply while increasing non-fossil energy (e.g., nuclear power, wind power) supply to optimize its energy structure. It is practical for the whole country to enhance its energy use efficiency to reduce energy consumption and its related CO2 emissions, especially for the regions such as Shanghai, the coal consumption proportions of which are about 50% of their total energy consumptions. This can be done by encouraging energy-saving technology innovations, closing down backward production companies, improving industrial structure, etc. It is noteworthy that policy making should consider the fact that the efficiencies of different energy inputs are different among regions. For example, the inefficiency of energy consumption in Hubei is mainly caused by coal consumption, while in Qinghai all energy inputs are inefficient. Particularly, in the east regions, more efforts should be made to improve coal and oil efficiencies, while in the west regions, it is important for them to focus on gas efficiency optimization. Developing non-fossil energy is an important way to adjust the existing energy structure. However, this should be done according to local conditions. For example, nuclear power is the main supply of all non-fossil energies in China in the current state due to its huge capacity and lower generating cost (Jiang et al., 2010). The nuclear power plants are all located in the coastal regions (i.e., the east regions), which indicates that these regions should expand this energy supply efficiently. The biomass power has developed significantly in recent years, and the plants are mainly located in 14 regions, e.g., Henan, Inner Mongolia, Sichuan and Hebei. Wind power plants are mainly located in the north and east China regions; hydropower plants are situated in the regions along the bigger rivers in China (e.g., Hubei along the Yangtze River). These conditions imply that these regions may promote their local non-fossil energies to fulfill their adjustments of energy structure.

Scenario analysis is used to estimate the potential reductions in energy consumption and its related CO2 emissions. Three specific scenarios are undertaken in this study, which can be further extended to other scenarios by changing parameters of energy structure adjustment. Furthermore, energy structure adjustment can also set different adjusting parameters for different areas or regions by considering their particular energy consumption situations. These issues may provide some interesting topics in future research.

9

Acknowledgements This research was supported by grants from the national Natural Science Foundation of China (no. 71001094, 71101085, 71371010), NSFC major program (no. 71090401/71090400), Shanghai University 085 Project-Social Development of Metropolis and Construction of Smart City (no. 085SHDX001) and innovation program of Shanghai Municipal Education Commission (no. 12ZS099).

References Adler, N., Golany, B., 2001. Evaluation of deregulated airline networks using data envelopment analysis combined with principal component analysis with an application to Western Europe. European Journal of Operational Research 132, 260–273. Al-Mansour, F., 2011. Energy efficiency trends and policy in Slovenia. Energy 36, 1868–1877. Ang, B.W., Mu, A.R., Zhou, P., 2010. Accounting frameworks for tracking energy efficiency trends. Energy Economics 32, 1209–1219. Banker, R.D., Morey, R.C., 1986. Efficiency analysis for exogenously fixed inputs and outputs. Operations Research 34, 513–521. Bian, Y., Yang, F., 2010. Resource and environment efficiency analysis of provinces in China: a DEA approach based on Shannon’s entropy. Energy Policy 38, 1909–1917. Chang, T.-P., Hu, J.-L., 2010. Total-factor energy productivity growth, technical progress, and efficiency change: an empirical study of China. Applied Energy 87, 3262–3270. Choi, Y., Zhang, N., Zhou, P., 2012. Efficiency and abatement costs of energy-related CO2 emissions in China: a slacks-based efficiency measure. Applied Energy 98, 198–208. Chung, W., 2011. Review of building energy-use performance benchmarking methodologies. Applied Energy 88, 1470–1479. Dai, Y., Zhu, Y., Bai, Q., Hu, X., Yu, S., 2009. Chin’s Low Carbon Development Pathways by 2050. Science Press, Beijing. Färe, R., Grosskopf, S., 2004. Modeling undesirable factors in efficiency evaluation: comment. European Journal of Operational Research 157, 242–245. Färe, R., Primont, D., 1995. Multi-output Production and Duality: Theory and Applications. Kluwer Academic Publishers, Boston. Guo, X., Zhu, L., Fan, Y., Xie, B., 2011. Evaluation of potential reductions in carbon emissions in Chinese provinces based on environmental DEA. Energy Policy 39, 2352–2360. Hou, J., Zhang, P., Tian, Y., Yuan, X., Yang, Y., 2011. Developing low-carbon economy: actions, challenges and solutions for energy saving in China. Renewable Energy 36, 3037–3042. Hu, J.L., Kao, C.H., 2007. Efficient energy-savings targets for APEC economics. Energy Policy 35, 373–382. Hu, J.L., Wang, S.C., 2006. Total-factor energy efficiency of regions in China. Energy Policy 34, 3206–3217. Jiang, B., Sun, Z., Liu, M., 2010. China′s energy development strategy under the lowcarbon economy. Energy 35, 4257–4264. Kuosmanen, T., 2005. Weak disposability in nonparametric productivity analysis with undesirable outputs. American Journal of Agricultural Economics 87, 1077–1082. Lee, Y., Hu, J., Kao, C., 2011. Efficient saving targets of electricity and energy for regions in China. Electrical Power and Energy Systems 33, 1211–1219. Li, F., Dong, S., Li, X., Liang, Q., Yang, W., 2011. Energy consumption-economic growth relationship and carbon dioxide emissions in China. Energy Policy 39, 568–574. Li, M., 2010. Decomposing the change of CO2 in China: a distance function approach. Ecological Economics 70, 77–85. Li, H., Mu, H., Zhang, M., Gui, S., 2012. Analysis of regional difference on impact factors of China′s energy-related CO2 emissions. Energy 39, 319–326. Li, Y., Oberheitmann, A., 2009. Challenges of rapid economic growth in China: reconciling sustainable energy use, environmental stewardship and social development. Energy Policy 37, 1412–1422. Liou, J-L., Wu, P.I., 2011. Will economic development enhance the energy use efficiency and CO2 emission control efficiency? Expert Systems with Applications 38, 12379–12387. Liu, L., Wang, J., Wu, G., Wei, Y., 2010. China′s regional carbon emissions change over 1997–2007. International Journal of Energy and Environment 1, 161–176. Lozano, S., Gutiérrez, E., 2008. Non-parametric frontier approach to modelling the relationships among population, GDP, energy consumption and CO2 emissions. Ecological Economics 66, 687–699. National Development and Reform Commission, 2007. The 11th 5-year Plan for Energy Development. National Development and Reform Commission of China, Beijing (in Chinese). Ramanathan, R., 2006. A multi-factor efficiency perspective to the relationships among world GDP, energy consumption and carbon dioxide emissions. Technological Forecasting & Social Change 73, 483–494.

Please cite this article as: Bian, Y., et al., Estimation of potential energy saving and carbon dioxide emission reduction in China based on an extended non-radial DEA approach. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.08.051i

10

Y. Bian et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Podinovski, V.V., Kuosmanen, T., 2011. Modelling weak disposability in data envelopment analysis under relaxed convexity assumptions. European Journal of Operational Research 211, 577–585. Sadjadi, S.J., Omrani, H., 2008. Data envelopment analysis with uncertain data: an application for Iranian electricity distribution companies. Energy Policy 36, 4247–4254. Shan, H., 2008. Re-estimating the capital stock of China: 1952–2006. Journal of Quantitative and Technical Economics 10, 17–31. (in Chinese). Shi, G., Bi, J., Wang, J., 2010. Chinese regional industrial energy efficiency evaluation based on a DEA model of fixing non-energy inputs. Energy Policy 38, 6172–6179. Wang, R., Liu, W., Xiao, L., Liu, J., Kao, W., 2011. Path towards achieving of China′s 2020 carbon emission reduction target—a discussion of low-carbon energy policies at province level. Energy Policy 39, 2740–2747. Wang, Q., Zhou, P., Zhou, D., 2012. Efficiency measurement with carbon dioxide emissions: the case of China. Applied Energy 90, 161–166. Wei, Y., Jiao, H., Fan, Y., 2007. An empirical analysis of energy efficiency in China′s iron and steel sector. Energy 32, 2262–2270.

Wei, C., Ni, J., Du, L., 2012. Regional allocation of carbon dioxide abatement in China. China Economic Review 23, 552–565. Wu, F., Fan, L., Zhou, P., Zhou, D., 2012. Industrial energy efficiency with CO2 emissions in China: a nonparametric analysis. Energy Policy 49, 164–172. World Bank, 2009. World Development Indicators. World Bank, Washington. Zhang, X., Cheng, X., Yuan, J., Gao, X., 2011. Total-factor energy efficiency in developing countries. Energy Policy 39, 644–650. Zhou, P., Ang, B.W., 2008. Linear programming models for measuring economywide energy efficiency performance. Energy Policy 36, 2911–2916. Zhou, P., Ang, B.W., Poh, K.L., 2008. A survey of data envelopment analysis in energy and environment studies. European Journal of Operational Research 189, 1–18. Zhou, P., Ang, B.W., Han, J.Y., 2010. Total factor carbon emission performance: a Malmquist index analysis. Energy Economics 32, 194–201. Zhou, P., Ang, B.W., Poh, K.L., 2006. Slacks-based efficiency measure for modeling environmental performance. Ecological Economics 60, 111–118.

Please cite this article as: Bian, Y., et al., Estimation of potential energy saving and carbon dioxide emission reduction in China based on an extended non-radial DEA approach. Energy Policy (2013), http://dx.doi.org/10.1016/j.enpol.2013.08.051i