Green development determinants in China: A non-radial quantile outlook

Journal of Cleaner Production 162 (2017) 764e775

Contents lists available at ScienceDirect

Journal of Cleaner Production journal homepage: www.elsevier.com/locate/jclepro

Green development determinants in China: A non-radial quantile outlook Boqiang Lin a, *, Nelson I. Benjamin b a

Collaborative Innovation Center for Energy Economics and Energy Policy, China Institute for Studies in Energy Policy, School of Management, Xiamen University, Fujian, 361005, PR China b China Center for Energy Economics Research, School of Economics, Xiamen University, Xiamen, Fujian, 361005, PR China

a r t i c l e i n f o

a b s t r a c t

Article history: Received 23 November 2016 Accepted 7 June 2017 Available online 9 June 2017

China is at the stage of industrialization and urbanization and because the energy demand and consumption for this process is rigid, coupled with emissions and global warming awareness, China is on a path to cut back emissions as she focuses on alternative and sustainable way to green the economy. This research investigates determining factors in measuring the dynamic changes in green development growth index (GDGI) over a given time frame. By adopting a non-radial directional distance function (NDDF), where more pollutants are added like sulphur dioxide emissions, solid wastes, waste water, and carbon dioxide emissions instead of using only one pollutant in measurement and using a global distance envelop analysis (DEA) to model green performance by considering both desirable and undesirable outputs, the model was decomposed into efficiency change (EC) index, best practice gap change (BPC) index, and technology gap change (TCG) index and these three indexes were employed to measure the green development performance in thirty provinces across china from 2000 to 2012. Results showed that provinces in the eastern region of China are greener than the central and western regions. Analyzing all three calculated dependent variables from a quantile perspective revealed that effects of EC, BPC, and TGC varied across different quantiles of GDGI. The coefficients of BPC were more significant than EC across quantiles, and TGC coefficients only became significant from Q(0.35), and continued on this path until the last observed quantile Q(0.95) however it was less than EC in terms of significance. The values of pseudo R2 also continued to increase from Q(0.20) until the last observed quantile with 86% accuracy in prediction recorded at Q(0.95). Analyzing the 80th percentile revealed that the coefficient of BPC was highest in this percentile which implies that a unit increase in best practice gap change, will influence green development growth by 102.3 percent, and a unit increase in efficiency change, will accelerate green development growth by 99.41 percent while a unit increase in technical gap ratio change will produce 99.38 percent increase in green development growth of China. © 2017 Elsevier Ltd. All rights reserved.

Handling Editor: Yutao Wang Keywords: Green development performance Non-radial directional distance function Quantile analysis

1. Introduction Growth and economy are inseparable when considering the economy of a nation, however not all growth are sustainable meaning that, some growth are actually detrimental to the economy in the long run. Suffice to say that green development is not synonymous with sustainable development. Some consider green

* Corresponding author. Collaborative Innovation Center for Energy Economics and Energy Policy, China Institute for Studies in Energy Policy, School of Management, Xiamen University, Fujian, 361005, PR China. E-mail addresses: [email protected], [email protected] (B. Lin), [email protected] (N.I. Benjamin). http://dx.doi.org/10.1016/j.jclepro.2017.06.062 0959-6526/© 2017 Elsevier Ltd. All rights reserved.

development as a specific and practical way in achieving a sustainable development while others argue that green development is a far more advance concept than sustainable development. However comparing both, green development is proactive in nature which aims to benefit future generations while sustainable development is passive aiming not to harm future generation. Green growth is not a replacement for sustainable development, it rather provides a practical and flexible approach for achieving concrete, measurable progress across its economic and environmental pillars while taking detailed account of the social consequences of greening the growth dynamics of economies (OECD, 2011). In china, the general view of green development is that it is a means required to achieve a sustainable development. Since the

B. Lin, N.I. Benjamin / Journal of Cleaner Production 162 (2017) 764e775

central government of China proposed the concept of green development in the 12th Five Year plan in 2011, theoretical framework and practice of green development have become an indispensable research interest for policy discussions among policy makers. Green development is simply an innovative economic development model that encompass the constraints of environmental and ecological carrying capacity and strives to achieve sustainable development with environmental protection as one of its major focus, or in a nutshell, it can be defined as the pathway of an economic growth where the usage of natural resources are sustainable which provides an alternative outlook to a typical industrial economic growth. The general perspective about economic growth in China is that, Chinese economy is unsustainable, unstable, and unbalanced (Chen and Golly, 2014). However, China has prioritized green development in almost all of its leading economic sectors like energy, transportation, and forestry just to name a few with a potential promise for expanded employment in industries and economic sectors that can reduce the country's environmental impact. The current scale of investment and employment in the following sectors really sheds more light on the greening activities taking place in different sectors of the economy. For instance, the energy sector of China that was coal dependent, has in recent times seen increasing share of renewable energy in the country's energy mix which will significantly reduce emissions and play a vital role in greening the energy supply, this is seen in solar hot water, solar photovoltaic (PV), and wind power. During the 11th Five Year Plan, China's solar PV power sector generated around 2700 direct jobs and 6500 indirect jobs annually on average. This is projected to increase to an average of 6680 for direct jobs and 16,370 for indirect jobs annually between 2011 and 2020. Given the rapid growth in China's solar industry, these estimates for future green jobs could increase immensely in the coming years. China's wind power industry consisting the power generation and turbine manufacturing sectors created an average of 40,000 direct green jobs annually between 2006 and 2010, factoring in increased productivity, China's wind power development between 2011 and 2020 is projected to generate around 34,000 green jobs annually on average (Pan et al., 2011). Formerly regarded as a kingdom of bicycles, china is expected to add around 220 million new vehicles between now and 2020, Chinese market for alternative fueled vehicles is really expanding rapidly despite its newness because by mid 2010, China was home to 5000 such vehicles. Assuming the government continues on the path of prioritizing the development of hybrid and electric vehicles during the 2011e2020 period, cumulative production could reach 16.7 million which is an average of 1.67 million hybrid vehicles annually. This will lead to the creation of about 1.2 million green jobs annually on average. Already at the forefront in high speed rail (HSR) development, china aims to reach 18,000 km of HSR by 2020 which will create more green jobs on an average of 230,000 for direct jobs and 400,000 for indirect jobs annually. Considering Beijing as a case study, as one of the most populated cities across the globe, the municipal government stepped up its urban rail ambitions where 660 km of lines were completed in 2015 at a total investment of $77 billion and construction for another 340 km of lines during 2016-202 at a total investment of $69 billion. This could guarantee more than 437,000 green jobs each year by 2020. In the area of forest resources, China is home to more than 2000 tree species, more than 1800 species of wild animals, and more than 6000 species of bushes, hundreds of which are only found here in China. It is certain that nourishing these forested areas is very important for sustaining China's green transition, and the pattern of economic and employment prospects in forestation, forest management, and forest tourism is really encouraging even though not having abundant forest resources. According to (Pan

765

et al., 2011), government led forestation efforts have led to an impressive expansion in nationwide forest cover where the forestation sector has employed more than 1.8 million full time workers in 2010 with an average of 1.6 million workers annually between 2005 and 2010. China's forestation activities could offer more 1.1 million direct and indirect jobs annually during 2011e2020 in order to achieve its 2020 goals, and the management of all newly added forest area during this period would create around a million green jobs. There is also a great expansion potential in forest park tourism because it is home to more than 2000 forest parks, and this green sector alone could provide 392000 green direct jobs and 607000 green indirect jobs which is almost a million green jobs totally. Considering only the three sectors mentioned here - energy, transportation, and forestry, could create around 4.5 million green jobs by the year 2020. If these sectors continue on their path of rapid expansion in addition to other emerging green sectors following same pathway, it is obvious that there is a huge potential for green jobs in the economy. The measurement of green development garnered attention several decades ago among developed and developing countries as we try to cut back on emissions and global warming. By adopting a non-radial directional distance function (NDDF), where more pollutants are added like sulphur dioxide emissions, solid wastes, waste water, and carbon dioxide emissions instead of using only one pollutant in measurement and using a global distance envelop analysis to model green performance by considering both desirable and undesirable outputs respectively, quantile analysis was employed to clearly ascertain the effects of all calculated explanatory variables in the model and visualize their deterministic trend in greening the economy due to the flexibility in quantile for modeling data with heterogeneous distributions. Furthermore, median analysis is more robust to outliers and embodies richer characterization and description of data. The remainder of this research is organized as follows: section 2 reviews some existing literature, section 3 presents methodological issues and data used in the analysis, section 4 provides empirical outcomes and discussion, while section 5 shows conclusion of the analysis, with summary and policy implications. 2. Literature reviews It is obvious that the energy consumption and environmental pollution of china increased significantly, however several research in the past mainly focused on productivity improvement where they analyzed and studied how sustainable was the economic growth of china either through a non parametric framework (Yang et al., 2015) or by a total factor productivity (TFP) approach, with the assumptions that a rising total factor productivity is the vital and only option in attaining a sustainable growth in an economy (Li, 2009; Chen et al., 2008). This approach is misleading in a way because total factor productivity does not recognize the environmental cost of economic growth which gives a false estimate about the true contribution of total factor productivity to economic growth. We believe that this and other factors really gave a false outlook in the past about economic growth in China, where growth was pursued at the detriment of the state of the environment because policy makers overestimated growth trends. In order to efficiently account for undesirable output to an extent in economic growth, green development was introduced as a strategy to enhance production and good environmental performance that ensures green development (Choi, 2015; Ahmed, 2012). This is derived by the integration of both environmental protection and productivity improvement, definitely, this approach will account for undesirable outputs that were left out of the equation by total factor productivity. Some studies on green development

766

B. Lin, N.I. Benjamin / Journal of Cleaner Production 162 (2017) 764e775

performance used the directional distance function in their analysis (Rashidi and Farzipoor Saen, 2015; Sueyoshi and Goto, 2013), while a few others added the total factor carbon dioxide emission performance index and the metafrontier non-radial Malmquist carbon dioxide emission performance index to the model (Zhang et al., 2015; Zhang and Choi, 2013). Also (Yu Ying Lin and Chen, 2013), measured green productivity among 70 countries by the combination of directional distance function model and metafrontier analysis. Considering environmental pollutants as undesirable outputs (Chung et al., 1997), presented a directional distance function (DDF) with the view of increment in goods while decreasing the undesirable output at the same time under the framework of data envelop analysis (DEA). Adopting a non-radial data envelop analysis model (Song et al., 2015) evaluated the green development of Chinese transportation sectors across provinces in China and obtained a poor performance estimate. Acknowledging the shortfall associated with conventional data envelope analysis (Li and Lin, 2015a), combined other indexes so as to improve on their model and found that in terms of development in China, the eastern part comes first, then the western part, and finally central China. Most existing literature in the research of green development performance focused solely on either sulphur dioxide emissions or on carbon dioxide emissions and seldom considered other pollutants of the environment. These approach will give a biased estimates and inconsistent results that can't really explain green growth and like (Li and Song, 2016) argued that measuring green development should include more environmental pollutants and using a non-radial distance directional function model is more efficient. This research agrees with their views, but however, we employed quantile estimate as a variation so as to really see the impact of all determinants, the explanatory variable on green development. A careful target of economic policy towards green development is not feasible if only the effects on averages are analyzed. Policy analysis using the traditional OLS methods are not particularly suitable to target any green conservation policies towards high energy industry in the economy, therefore a more sophisticated tool is needed in order to capture completely the complex dependence between varying energy variables. Since the introduction of quantile regression (QR) by Koenker and Bassett (1978), the popularity has grown tremendously where it has become a tool in modeling dependence because it involves the consideration of set of regression curves that differ across different quantiles (percentiles) on the conditional distribution of the dependent variable (Koenker and Hallock, 2001; Koenker, 2005). Also (Baur, 2013), strongly suggested the use of QR to study the structure and degree of dependence because it can reveal information on the asymmetric and non-linear effects of conditional variables on the dependent variables in economic models. This research uses a new index to evaluate green development in China and uses the perspective of quantile to access salient features in key determinant variables in order to see if all explanatory variables are efficient in greening China.

carbon dioxide emission performance in electricity generation where green performance was modeled with both the desirable and undesirable outputs, non-directional distance function (NDDF) relaxed the assumption for proportional adjustments of inputs and outputs which is vital to the conventional distance directional function. DDDF is define mathematically as:

.

n o ND ðx; y; b; hÞ ¼ sup wT b : ðx; y; bÞ þ h  diagðbÞЄT

(1)

where b ¼ ðbx by bh ÞT  0 is a vector of scaling factors that measures the real distance of inputs or outputs from the optimal state; wT ¼ ðwx wy wb Þ are the inputs or outputs weight, with a directional vector denoted h ¼ ðhx hy hb Þ, and diagonal matrices. The production process (T) can be defined as using x (inputs) to produce y (desirable outputs) while also emitting undesirable environmental pollutants b. Conceptually captured as T ¼ { (x, y, b): x can produce (y,b)}. The technology set T is assumed to be a closed, bounded, and convex set in nature and according to (Fare et al., 2007; Li and Lin, 2015b) must satisfy the following properties: A. If (x, y, b) Є T, and b ¼ 0, then y ¼ 0; B. If (x, y, b) Є T, and x' > x, then T(x)  T(x'); C. If (x, y, b) Є T, and y'


 1 1 1 1 1 1 and 0; 0; ; ; ; ; ; 3 3 12 12 12 12

3. Methodology and data

wT ¼

3.1. Non-radial directional distance function

h ¼ ð0; 0; E; Y; S; D; W; CÞ

This is an outstanding model in the study of environmental performance because it delves deeper into the concept of green growth and development. Because of the simultaneous production of desirable and undesirable outputs, non-radial directional distance function (NDDF) allows and makes possible the increase of all desirable outputs and also ensures the decrease of undesirable ones. Proposed by (Zhou et al., 2012) in their research of energy and

Also (Lin and Du, 2015) used a similar weight vector assumption in their study, and this implies that in the reduction of energy both the expansion of desirable output and the reduction of carbon dioxide emissions are of equal importance in our model. Based on the above postulations, the value of NDDF for a specific DMU (in this case a province in China) is calculated by solving the following linear programming:

B. Lin, N.I. Benjamin / Journal of Cleaner Production 162 (2017) 764e775

3.2. Decomposition of green development growth index

 1 1 1 1 ND ðK; L; E; Y; S; D; W; C; hÞ ¼ max bE þ bY þ bS þ bD 3 3 12 12  1 1 þ bW þ bC 12 12

.

(2) subject to T X N X

PT

t¼1

PN

i¼1 lIt Kit

K

lit Lit  L

t¼1 i¼1 T X N X

lit Eit  E  bE hE

t¼1 i¼1 T X N X

lit Yit  Y þ bY hY

t¼1 i¼1 T X N X

lit Sit ¼ S  bS hS

t¼1 i¼1 T X N X

lit Dit ¼ D  bD hD

A basic assumption about the index above (GDPI) is that all provinces in China are at par in technological level which means that it fails to acknowledge heterogeneity in technology across different provinces because as we know from facts on ground, that eastern provinces in China have relatively high level of economic development with cutting edge technologies to mitigate undesirable environmental pollutants. The GDPI index above is therefore a static index rather than a dynamic one. To obtain a dynamic GDPI index that can measure changes in green development and recognize regional heterogeneity, a combination of GDPI and meta frontier analysis is necessary. So, dividing the DMUs (provinces in china are 30) into g (g ¼ 3) independent groups which are the eastern, central, and western China and whereby each group has Ng (g ¼ 1,2,3) DMUs with different production technologies. According to (Tulkens and Eechaut, 1995), we can define three technological benchmark sets which are: contemporaneous, intertemporal, and global benchmark technologies respectively. Hence for a given group g, a contemporaneous benchmark technology ðTgC Þ is able to construct a best practice frontier based on the observations for specific g group for a particular time t. The GDPI given by equation (3) above can be denoted as GDPIgC based on bbj ðj ¼ E; Y; S; D; W; CÞ that will be calculated from equation (4) below. Contemporaneous benchmark is therefore represented in a linear programming as follows:

 1 1 1 1 NDg ðK; L; E; Y; S; D; W; C; hÞ ¼ max bE þ bY þ bS þ bD 3 3 12 12  1 1 þ bW þ bC 12 12

.C

t¼1 i¼1 T X N X

767

lit Wit ¼ W  bW hW

t¼1 i¼1

(4) T X N X

subject to

lit Cit ¼ C  bC hC

t¼1 i¼1

Ng X

lit  0ði ¼ 1; 2; …; N; t ¼ 1; 2; …; TÞ

Ng X

We used the whole sample to construct the best practice frontier under the data envelopment analysis (DEA), this implies a single global frontier is used as a benchmark technology for all DMUs, this ensures that estimated values can be compared over periods of time. Assuming the optimal solutions to equation (2) ð0; 0; b*E b*Y b*S b*D b*W b*C ÞT

above is b ¼  0. Under a best practice scenario, the energy inputs, desirable outputs and undesirable environmental pollutants will be given as

b*E;it

b*Y;it

l K i¼1 it it

K

lit Lit  L

i¼1

bj  0ðj ¼ E; Y; S; D; W; CÞ

*

PNg

b*V;it

Eit   Eit ; Yit   Yit and Vit   Vit ðV ¼ S; D; W; CÞ: A DMU on the best practice frontier will occur

when b* ¼ 0: A green development performance index (GDPI) is defined as follows:

. 2 3 * Yit  b*Y;it Yit 1 4 Eit  bE;it Eit 5 GDPIit ¼ 2 Eit =Yit 2 3 * * 1 41 X Jit  bJ;it Jit =Yit  bY;it Yit 5 þ 2 4 J¼S;D;W;C Jit =Yit

lit Eit  E  bE hE

i¼1

Ng X

lit Yit  Y þ bY hY

i¼1

Ng X

lit Sit ¼ S  bS hs

i¼1

Ng X

lit Dit ¼ D  bD hD

i¼1

Ng X

lit Wit ¼ W  bW hW

i¼1

(3)

As usual, a higher GDPI implies a better green development performance and the values should lie between zero and one.

Ng X i¼1

lit Cit ¼ C  bC hC

768

B. Lin, N.I. Benjamin / Journal of Cleaner Production 162 (2017) 764e775



lit  0 i ¼ 1; 2; …; Ng ; t ¼ 1; 2; …; T; g ¼ 1; 2; 3



tþ1 GDGIi;t ¼

 GDPIG Ki;tþ1 ; Li;tþ1 ; Ei;tþ1 ; Yi;tþ1 ; Si;tþ1 Di;tþ1 Wi;tþ1 Ci;tþ1  GDPI G Ki;t ; Li;t ; Ei;t ; Yi;t ; Si;t ; Di;t ; Wi;t Ci;t

bj  0ðj ¼ E; Y; S; D; W; CÞ

(6)

Intertemporal benchmark technology constructs a single best practice frontier for a given group g over a whole time period and is denoted as TgI ¼ Tg1 ∪Tg2 ∪…∪TgT and because we have g ¼ 3 independent groups, there will be a g ¼ 3 different intertemporal technologies distinctively. Using equation (3) above and equation (5) below, GDPI index can be denoted as GDPIgI , and Intertemporal benchmark is therefore represented in a linear programming as follows:

 1 1 1 1 NDg ðK; L; E; Y; S; D; W; C; hÞ ¼ max bE þ bY þ bS þ bD 3 3 12 12  1 1 þ bW þ bC 12 12

.I

(5) subject to Ng T X X

PT

PNg

t¼1

l K i¼1 it it

K

lit Lit  L

t¼1 i¼1

Ng T X X

lit Eit  E  bE hE

t¼1 i¼1

Ng T X X

lit Yit  Y þ bY hY

t¼1 i¼1

Ng T X X

lit Sit ¼ S  bS hs

t¼1 i¼1

Ng T X X

lit Dit ¼ D  bD hD

t¼1 i¼1

Ng T X X

The decomposition of GDGI into different components yields the following: tþ1 GDGIi;t

"  # " 

 # GDPIC :tþ1 GDPII :tþ1 GDPIC :tþ1 ¼  GDPI C ð:t Þ GDPI I ð:t Þ=GDPI C ð:t Þ " 

 # GDPIG :tþ1 GDPII :tþ1  GDPI G ð:t Þ=GDPI i ð:t Þ

(7)

TEtþ1 BPRtþ1 TGRtþ1 b   TEt BPRt TGRt bEC  BPC  TBC The representation of equation (7) above is green development performance at time t, tþ1, Contemporaneous benchmark technology ðTgC Þ is calculated as GDPI C ð:t;tþ1 Þ and abbreviated TEt;tþ1 ; and the term EC measures the closeness of how the DMU of a particular group accelerates towards ðTgC Þ at any given time tþ1 to any given time t (a catch up effect). When a given EC > < 1, it means that a given DMU is moving closer or far away from the best practice frontier at time tþ1 compared to time t. The best practice ratio ðBPRÞ ¼ GDPI I =GDPI C measures the best practice gap ratio between intertemporal benchmark technology and contemporaneous benchmark technology where changes over time period are captured by BPC as seen in equation (7). A BPC > < 1 implies that the contemporaneous technology frontier is moving closer or far away from the intertemporal technology frontier which can also shed more light on innovation effect. Technology gap ratio ðTGR ¼ GDPI G =GDPII Þ measures the technology gap ratio between the global benchmark technology and intertemporal benchmark technology where changes over time period are also captured by TGC as seen in equation (7) above. Also, a TGC > < 1 shows decrease or increase in technology gap between the global technology and the intertemporal technology for a given group which can also shed more light on the technology leadership effect of a particular group.

lit Wit ¼ W  bW hW

t¼1 i¼1

Ng T X X

 GDPIG :tþ1 b GDPIG ð:t Þ

3.3. Quantile outlook

lit Cit ¼ C  bC hC

t¼1 i¼1

 lit  0 i ¼ 1; 2; …; Ng ; t ¼ 1; 2; …; T; g ¼ 1; 2; 3

bj  0ðj ¼ E; Y; S; D; W; CÞ: Thirdly, we define global benchmark technology as T G ¼ T1I ∪T2I ∪…∪TgI . Using the whole observations, it is able to construct only one best practice frontier. Following this substitutions, the corresponding GDPI is represented as GDPI G , and because it is calculated using global DEA model, the estimates can be compared over years. Green development growth index (GDGI) is therefore introduced to estimate the dynamic changes occurring in GDPI based on GDPI G and it is represented as follows:

Let Y be a dependent variable (GDGI) which we assume to be linearly dependent on X ¼ ðEC; BPC; TGCÞ. The tth conditional quantile function of Y is specified as follows:



X Qy ðtjxÞ ¼ inf b Fy ðbjxÞ  t ¼ bk ðtÞxk ¼ x0 bðtÞ

(8)

k

where Fy ðbjxÞ is the conditional distribution function of Y given X, and the QR coefficient b(t) determines the dependence relationship between vector X and the tth conditional quantile of Y. The dependence is unconditional if no exogenous variables are added in X while it is conditional if exogenous variables are added. Also, the values of b(t) for t Є [0, 1] determines the complete dependence structure of Y. The dependence of Y ¼ GDGI based on a specific independent variable (explanatory variable) in vector X ¼ ðEC; BPC; TGCÞ could be any of the following: (i) a constant where the values of b(t) does not change for different values of t;

B. Lin, N.I. Benjamin / Journal of Cleaner Production 162 (2017) 764e775

(ii) a monotonically decreasing (increasing) where b(t) decreases (increases) with the value of t; (iii) a symmetric (asymmetric) where the value of b(t) is similar (dissimilar) for low and high quantiles. Furthermore, the coefficients b(t) for a given t are estimated by the minimization of the weighted absolute deviations between y and x as follows:

bðtÞ ¼ argmin

T  X t¼1





t  1fyt < x0t bðtÞg yt  x0t bðtÞ

(9)

where 1fyt < x0t bðtÞg is an indicator function. The solution to the above problem is solved using the linear programming algorithm suggested by (Koenker and D'Orey, 1987). The QR model in equation (9) above allows us to carefully examine: (a) what kind of dependence structure exist; (b) how the dependence structure is affected by different regressors to mention but a few. We chose quantile approach because of its flexibility for modeling data with heterogeneous distributions, median regression is more robust to outliers and for richer characterization and description of data to ensure proper conceptualization of green development in China.

3.4. Data This study is carried out on 30 provinces in China from the period of 2000e2012 and Tibet was not included due to unavailability of data for the region. In order to ensure a robust estimates for green development performance across these provinces in China, data were collected on inputs, desirable output, and undesirable environmental pollutants and all these are detailed as follows: (i) Inputs: These includes energy consumption (E), capital stock (K), and labor force (L). The data on energy consumption is collected from China Energy Statistical Yearbook and are measured in 104 tons of standard coal equivalent (104 tce), while data on capital stock (constant 2000 prices) and labor force are found in (Li and Lin, 2015a) where capital stock was estimated by the perpetual inventory method with each province having different rate of depreciation of gross fixed capital formation. (ii) Desirable output: Data were also collected from China Energy Statistical Yearbook and is measured by gross regional product (GRP) with a unit of RMB at 100 million Yuan, and with a GDP deflectors, it has been converted into 2000 prices.

769

(iii) Undesirable environmental pollutants: Pollutants in this research are simply sulphur dioxide emission (S), solid wastes (D), waste water (W), and carbon dioxide emission (C). All data on sulphur dioxide, solid waste, and waste water were also collected from China Statistical Yearbook, and CO2 was estimated by using the fuel based carbon calculation model by IPCC (2006) on eight fuel types (coal, crude oil, natural gas, diesel oil, fuel oil, gasoline, kerosene, and coke) because fossil fuels are considered main emitters of carbon dioxide and their consumptions are all from China Energy Statistical Yearbook. Tables 1 and 2 shown below gives the total representation of all the descriptive statistics for inputs and outputs variables (N ¼ 30), and the average values and growth rates for all variables are divided into three groups. The existence of very high maximum values as compared to minimum values clearly shows the presence of heterogeneity across provinces in China. As a result of this, provinces can be divided into three groups using geographical nearness and national development strategies as dividing factors for group determination, which implies our thirty province are grouped into eastern, central, and western groups respectively. As can be seen from the table below, the eastern province average GRP is much higher and also used greater inputs than all other province in the remaining groups (central and western provinces). The differences in the patterns of inputs and outputs across groups can also signify varied production technology across different regions in the country, and the undesirable environmental pollutants aren't left out in this processes either. As can be seen also that the average value of waste water for the eastern provinces are much higher than the central and western provinces and the average values sulphur dioxide emission, solid wastes and carbon dioxide emission, even though they were very close to central provinces were still much higher to those in the western province. The presence of heterogeneity really justifies the decomposition of green growth using green development growth index (GDGI) and allowing quantile to capture every minute differences that could be observed.

4. Empirical outcomes and discussion 4.1. The green development performance index Using linear programming to solve equations (2) and (3), we obtained the results for the green development performance index

Table 1 Descriptive statistics of variables. Variable (units)

Mean (Std. deviation)

Maximum

Minimum

Capital stock (100 million RMB) Labor (ten thousand workers) Energy consumption (104tce) Gross Regional Product (100 million RMB) Sulphur dioxide emission (104 ton) Solid wastes (104 ton) Waste water (104 ton) Carbon dioxide emission (104 ton)

20067.40(17952.30) 2321.60(1540.90) 9570.20(7005.90) 7211.30(6938.50) 62.70(39.20) 5725.90(5876.10) 75186.60(63500.80) 26934.80(21364.20)

108951.70 6242.50 38899.00 42865.90 171.50 45576 296318 122379.40

1390.70 239.50 479.10 263.70 1.93 75 3453 445.30

Table 2 Average values and growth rates for variables in three groups. Group

No.

K

L

E

Y

S

D

W

C

Eastern Central Western

156 117 117

28701.90 (14.60%) 17641.10 (16.70%) 10980.40 (14.30%)

2536.30 (2.90%) 2602.41 (1.50%) 1754.70 (1.70%)

12057.60 (9.20%) 9880.50 (9.80%) 5943.60 (10.70%)

11031.50 (12.20%) 6192.40 (12.60%) 3136.50 (12.10%)

67.70 (0.30%) 67 (4.30%) 51.85 (6.20%)

6054.20 (11.60%) 6925.60 (11.90%) 4088.44 (19.30%)

110001 (1.70%) 67196.30 (1.90%) 36746.20 (1.90%)

32726.50 (10.50%) 30775.60 (9.30%) 15371.90 (11.80%)

770

B. Lin, N.I. Benjamin / Journal of Cleaner Production 162 (2017) 764e775

(GDPI, which is a static analysis) across the thirty provinces of China under the global benchmark technology, which is presented in Table 3 below. Green development performance value was highest in Guangzhou with a unity score all through the years except for 2002 and 2005 but averagely, the value of green development performance index in China for the sample period 2000e2012 was 0.310, and very few scores of GDPI were unity and corresponded mainly to developed provinces like Guangzhou, Beijing, and Tianjin. Among the 390 estimated values, only eighty five (85) were higher than 0.5 which signified that China lags behind in the level of green development. Fifty eight (58) estimates were higher than 0.5 during the time periods of 2000e2009 and twenty seven (27) for 2010e2012. These estimates however showed a positive progress when compared to previous time periods for green development performance index like in the year 2000, it is found that twenty six provinces actually made good progress in 2012. Some provinces in the western group regressed signifying a low green development, these are provinces like Shaanxi, Yunnan, Xinjiang, and Ningxia. During the time period of 2008 and 2009, which was a time of global crises, it was observed that the green development price indexes for eight provinces recorded a lower GDPI in the year 2009 than the previous year (2008). Fig. 1 plots the green development performance index (GDPI) across provinces from which it is clearly seen that provinces from the east have a better green development performance than other provinces (central and western). Fig. 2 on the other hand plots the average values of GDPI in all provinces and the highest GDPI occurred in the eastern provinces, an upward trend was observed among provinces also with the eastern provinces having the highest growth rate, followed by the central provinces and finally the western provinces. In a nutshell, the rate of growth of GDPI is really slow in the central and western provinces when compared to eastern provinces. The acceleration of green development performance in the central provinces and the western provinces is a major decision in ensuring

sustainable green economic development in China. Clearly seen from Fig. 1, Guangdong from the eastern group had best performance of green development, followed by Heilongjiang from the central group, and then Sichuan from the western group.

4.2. The green development growth index Green development growth index (GDGI) is able to capture the dynamism in green development performance change in China and is calculated using equation (6) or (7) by linear programming. The presentation in Table 4 below is the average values of GDGI decomposed during the period 2000e2012. The presence of heterogeneity was confirmed among groups from different regions and innovation effect (BPC) was the main driving factor in green development. The TGC values was unity for the eastern region implying the availability of best technological level in China while other regions (central and western) were less than unity which indicates a deterioration or decrease in technology leadership effect (there is a huge gap in technology between eastern provinces and other provinces). This is also confirmed in Fig. 3 where the average annual green growth rate is higher in the eastern region than other regions (central and western) in green development. There were more fluctuations in green development growth index experienced in the eastern region which could be attributed to the occurrence of economic crises at a given period of time and also EC values in the western region was the lowest. Finally Table 5 below presents all empirical results for average green development growth index (GDGI), efficiency change (EC), best practice gap change (BPC), and technical gap ratio change (TGC) of provinces in China. Referencing GDGI, Tianjin from the eastern province had the highest value with an 11.1% increase, all other provinces had greater than unity values also only for Yunnan, Xinjiang, and Ningxia with a less than unity value. Regarding EC, only Tianjin, Anhui, Hubei, Jilin, Liaoning, Gansu, and Guizhou had

Table 3 Estimates of green development performance index. Provinces

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

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

1.000 0.409 0.351 0.394 0.532 0.580 0.499 0.477 0.294 0.288 0.371 0.435 0.298 0.371 0.367 0.285 0.324 0.277 0.221 0.303 0.314 0.300 0.236 0.261 0.207 0.238 0.206 0.178 0.190 0.170

1.000 0.466 0.388 0.411 1.000 0.637 0.530 0.497 0.308 0.321 0.417 0.408 0.310 0.429 0.393 0.322 0.338 0.299 0.246 0.287 0.304 0.321 0.261 0.269 0.226 0.242 0.218 0.191 0.172 0.210

0.955 0.498 0.434 0.436 0.504 0.561 0.493 0.514 0.334 0.349 0.378 0.393 0.328 0.363 0.386 0.323 0.340 0.298 0.275 0.325 0.299 0.306 0.238 0.273 0.229 0.231 0.246 0.186 0.165 0.190

1.000 0.530 0.464 0.469 0.504 0.536 0.491 0.515 0.366 0.350 0.377 0.375 0.345 0.360 0.369 0.316 0.336 0.302 0.292 0.333 0.299 0.300 0.238 0.278 0.228 0.235 0.229 0.176 0.175 0.150

1.000 0.630 0.493 0.518 0.520 0.524 0.490 0.486 0.382 0.364 0.378 0.354 0.361 0.353 0.345 0.308 0.311 0.310 0.292 0.321 0.293 0.286 0.231 0.271 0.241 0.235 0.220 0.172 0.176 0.160

0.924 0.632 0.533 0.538 0.543 0.497 0.491 0.444 0.382 0.373 0.385 0.321 0.372 0.331 0.340 0.318 0.313 0.327 0.300 0.281 0.285 0.284 0.235 0.266 0.238 0.239 0.204 0.198 0.183 0.140

1.000 0.732 0.601 0.538 0.522 0.517 0.501 0.460 0.412 0.391 0.358 0.344 0.357 0.341 0.353 0.331 0.323 0.346 0.320 0.289 0.300 0.290 0.249 0.263 0.252 0.249 0.204 0.198 0.176 0.139

1.000 1.000 0.711 0.667 0.512 0.534 0.526 0.493 0.461 0.425 0.369 0.363 0.371 0.354 0.360 0.354 0.340 0.298 0.340 0.302 0.311 0.306 0.265 0.264 0.271 0.257 0.211 0.212 0.196 0.132

1.000 0.601 0.836 0.737 0.516 0.552 0.556 0.528 0.537 0.446 0.396 0.391 0.387 0.373 0.381 0.325 0.306 0.295 0.339 0.317 0.336 0.332 0.288 0.268 0.280 0.264 0.205 0.226 0.212 0.131

1.000 0.717 0.903 0.793 0.547 0.555 0.581 0.557 0.582 0.473 0.412 0.410 0.402 0.387 0.339 0.343 0.302 0.317 0.363 0.290 0.274 0.249 0.314 0.255 0.297 0.189 0.214 0.230 0.224 0.129

1.000 0.879 1.000 1.000 0.579 0.576 0.602 0.571 0.610 0.486 0.423 0.426 0.427 0.390 0.343 0.351 0.311 0.334 0.384 0.310 0.284 0.262 0.330 0.243 0.219 0.213 0.226 0.239 0.228 0.130

1.000 1.000 1.000 0.913 0.519 0.597 0.624 0.577 0.714 0.494 0.428 0.426 0.442 0.394 0.356 0.352 0.320 0.342 0.367 0.329 0.298 0.262 0.246 0.215 0.299 0.206 0.210 0.241 0.228 0.127

1.000 1.000 1.000 1.000 0.544 0.622 0.661 0.607 1.000 0.484 0.451 0.455 0.461 0.405 0.371 0.363 0.350 0.375 0.381 0.365 0.313 0.277 0.250 0.191 0.230 0.252 0.212 0.245 0.232 0.140

B. Lin, N.I. Benjamin / Journal of Cleaner Production 162 (2017) 764e775

771

Fig. 1. Green development performance index (GDPI) across provinces (E ¼ Eastern province; C¼ Central province; W¼ Western province).

values greater than unity (presence of “catch-up” effects) while Hainan, Fujian, Zhejiang, Jiangsu, Shandong, Guangxi, Henan, Yunnan, Hebei, Qinghai, Shanxi, and Ningxia had values less than unity (absence of “catch-up” effects). For BPC, all provinces had greater than unity values which signifies the presence of innovation

effects. All eastern provinces including Guangxi and Qinghai had values greater than unity in TGC implying that these provinces has high level of technology across China while Anhui, Hubei, Jilin, Chongqing, Shaanxi, Yunnan, Gansu, Guizhou, Shanxi, and Ningxia had values less than unity implying the opposite.

Fig. 2. Average green development performance index (GDPI) across provinces (E ¼ Eastern province; C¼ Central province; W¼ Western province).

772

B. Lin, N.I. Benjamin / Journal of Cleaner Production 162 (2017) 764e775

Table 4 The average values of GDGI from 2000 to 2012. Group

GDGI

EC

BPC

TGC

Eastern Central Western China

1.0375 1.0214 1.0146 1.0258

0.9868 0.9995 0.9897 0.9915

1.0535 1.0322 1.0570 1.0482

1.0000 0.9992 0.9902 0.9968

Generally TGC indicates technology leadership but it is unable to show which provinces are top innovators without using the three benchmark technology sets described here. When its value is greater one, it implies that innovative province should be among technologically leading province and BPC >1 implies an innovation effect among group innovators. Guangdong from the eastern group had best performance of green development, followed by Heilongjiang from the central group, and then Sichuan from the western group. It is therefore pertinent to analyze all three factors used in this research to see their extent of influence and how robust they were in capturing information and determining the green development growth across provinces in China. 4.3. Quantile estimates Reported in Table 6 are the quantile estimates for efficiency change, best practice gap change, and technical gap ratio change of provinces in China, shown are numerical results for nineteen quantiles from 5 percentile to 95 percentile with all associated pvalues. Estimates showed how the effects of independent variables (EC, BPC, and TGC) varied across different levels of the dependent variable (GDGI). Results shows that these effects are not constant across different spectrum of the Green development growth index, coefficients of efficiency change (EC) and best practice gap change (BPC) was significant at Q(0.05) and became insignificant as we progress across quantiles until Q(0.35) when they became significant. These significance in coefficients continued to the last observed quantile Q(0.95), and it was also observed that the coefficients of BPC were more significant than EC across quantiles. Considering TGC, it's coefficients only became significant from

Table 5 The average green development growth index, efficiency change, best practice gap change, and technical gap ratio change of provinces in China. Province

GDGI

EC

BPC

TGC

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

1.001 1.096 1.093 1.083 1.039 1.007 1.024 1.021 1.111 1.045 1.018 1.005 1.037 1.010 1.002 1.022 1.008 1.027 1.047 1.018 1.002 0.998 1.010 0.975 1.021 1.012 1.004 1.028 1.019 0.990

1.000 1.000 1.000 1.000 0.956 0.964 0.989 0.987 1.033 1.000 1.000 1.000 1.005 0.969 0.973 1.011 0.998 1.002 1.004 1.000 1.000 0.973 0.967 1.000 1.008 1.000 0.939 1.032 0.988 0.956

1.001 1.063 1.093 1.083 1.098 1.050 1.037 1.035 1.084 1.039 1.021 1.002 1.036 1.042 1.031 1.040 1.042 1.042 1.044 1.045 1.047 1.053 1.046 1.022 1.063 1.024 1.080 1.086 1.058 1.053

1.000 1.039 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.015 1.002 1.005 0.998 1.000 1.000 0.976 1.000 0.990 1.000 0.975 0.960 0.982 1.000 1.006 0.982 1.034 1.000 0.979 0.978 0.988

Q(0.35), and continued on this path until the last observed quantile Q(0.95). The values of pseudo R2 also continued to increase from Q(0.20) until the last observed quantile with 86 percent accuracy in prediction recorded at Q(0.95). Analyzing the 80th percentile, coefficient of BPC was highest here implying that a unit increase in best practice gap change, will influence green development growth of China by 102.3 percent, and a unit increase in efficiency change,

Fig. 3. The trends in green development growth index (GDGI).

B. Lin, N.I. Benjamin / Journal of Cleaner Production 162 (2017) 764e775

773

Table 6 Quantile estimates for efficiency change, best practice gap change, and technical gap ratio change of provinces in China. Quantile

LN(EC)

LN(BPC)

LN(TGC)

Pseudo R2

Q(0.05) Q(0.10) Q(0.15) Q(0.20) Q(0.25) Q(0.30) Q(0.35) Q(0.40) Q(0.45) Q(0.50) Q(0.55) Q(0.60) Q(0.65) Q(0.70) Q(0.75) Q(0.80) Q(0.85) Q(0.90) Q(0.95)

0.1855368 (0.000)* 0.3292314 (0.081) 0.3012532 (0.500) 0.4735512 (0.428) 0.5191498 (0.088) 0.9299654 (0.049)** 0.9347934 (0.000)* 0.966953 (0.000)* 0.9880735 (0.000)* 0.9936793 (0.000)* 0.9962976 (0.000)* 0.9752924 (0.000)* 0.9554544 (0.000)* 0.9730504 (0.000)* 1.008255 (0.000)* 0.994104 (0.000)* 1.040777 (0.000)* 0.9905048 (0.000)* 0.9905048 (0.000)*

0.7320853 (0.000)* 0.5515069 (0.120) 0.4184479 (0.156) 0.6400361 (0.152) 0.4695482 (0.048)** 0.8388696 (0.025)** 0.8575549 (0.000)* 0.8893071 (0.000)* 0.9174258 (0.000)* 0.9250042 (0.000)* 0.9155248 (0.000)* 0.911341 (0.000)* 0.9194266 (0.000)* 0.9460903 (0.000)* 0.9823337 (0.000)* 1.022628 (0.000)* 1.000000 (0.000)* 1.000000 (0.000)* 1.000000 (0.000)*

-0.0963811 (0.234) 0.4958759 (0.118) 0.3366491 (0.367) 0.365181 (0.437) 0.6921065 (0.091) 0.86973 (0.107) 0.8735728 (0.000)* 0.9189461 (0.000)* 0.9450647 (0.000)* 0.9523324 (0.000)* 0.9583145 (0.000)* 0.9383498 (0.000)* 0.9692449 (0.000)* 1.016826 (0.000)* 0.9876279 (0.000)* 0.9937955 (0.000)* 1.033934 (0.000)* 1.033934 (0.000)* 1.033934 (0.000)*

0.3574 0.2382 0.2360 0.2381 0.2671 0.3490 0.4287 0.4924 0.5477 0.5858 0.6235 0.6597 0.6962 0.7303 0.7630 0.7948 0.8279 0.8488 0.8596

The numbers in parenthesis are p-values and the asterisk (* and **) denotes statistical significance at 1% and 5% level respectively.

will accelerate green development growth by 99.41 percent while a unit increase in technical gap ratio change will produce 99.38 percent increase in green development growth and the model goodness of fit was 79.5 percent. This illustrates that all independent variables had strong explanation for the model. Also illustrated in Fig. 4 is the graphical results for all the quantiles conditioning on GDGI, results shows an incremental trend as we progressed across quantiles and the trend was really steep after Q(0.80) which implied that GDGI below Q(0.35) had a lower green development growth index and above it, green development growth index started increasing. QR provided a richer description of the dynamics of the response of green development growth index to each of the three influencing factors and considering the fact that this model is robust because instead of modeling the mean, we used the median which is less affected by skewness and outliers because the data was skewed to the right. As can be seen also in Fig. 5, where the horizontal lines are ordinary least square coefficient which did not vary across different quantiles (stable) while the changing lines are the coefficients for different quantiles and the gray areas are the confidence interval for quantiles estimates. In ascending order of influence we have the following scenario: Technical gap ratio change (TGC) / Efficiency change (EC) / Best practice gap change (BPC).

Fig. 4. Quantile plot of green development growth index (GDGI).

5. Conclusion and recommendation Research on green development performance or green growth continues to be on the forefront of economic research and study because policy makers and environmentalist are inventing measures and other feasible applications or changes to curb climate change as low as possible, though there exist several research with different factors and models considered in the past, non has actually studied the determinants of green development performance using a non-radial approach by adopting a data envelopment analysis. This research incorporates four major pollutants deemed environmentally harmful such as sulphur dioxide emission, solid wastes, waste water, and carbon dioxide emission to measure the factors that determine green development performance in China which were calculated by data envelopment analysis (DEA) models. Adopting a total-factor production efficiency framework coupled with nonradial slacks, we obtained a static analysis for green development performance index under a global benchmark technology, after which a dynamic analysis of green development growth index was obtained. The green development growth index was decomposed into three indexes which were: efficiency change, best practice change, and technology gap change. Because these three indexes were used in the study of green development growth, quantile regression was implored to carefully study these determinants of green development performance in order to capture the impact of all three indexes on GDGI on different percentiles because quantile is considered more flexible for modeling data with heterogeneous distributions and median regression is more robust to outliers and for richer characterization and description of data to ensure proper analysis of green development in China. This is helpful because we really want to see the magnitude of influence these decomposed indexes has on green development and which one among them can be improved upon as China strives to be a global leader in emissions reduction in line with government set goals. Results shows that the effects of efficiency change (EC), best practice gap change (BPC), and technical gap ratio change (TGC) were different across all quantiles of green development growth index where the coefficients of BPC were more significant than EC across all quantiles, and the coefficients of EC were more significant than the coefficients of TGC. Because there is often a big gap between the knowledge of something and translating that knowledge into action, from the analysis it is confirmed that best-practice gap

774

B. Lin, N.I. Benjamin / Journal of Cleaner Production 162 (2017) 764e775

Fig. 5. Changes in the quantile regression coefficients.

change is the major determinant in Chinese green development performance (fulcrum) as was confirmed also by Fig. 3, followed by efficiency change and technical gap ratio been the least among all three chosen determinants, and these determinants were also able to explain 86% of data in one of the percentiles. Based on these findings, the following transformation are central to green development in China: (i) Transformation from a resource dependent economic development model to an innovative driven economic model. (ii) Transformation from an industrial dominant development economy to a service based economy. (iii) Transformation from an investment driven growth to a consumption based growth, and (iv) Drastic transformation from a high carbon emissions economy to an increasingly lower carbon emissions economy. Though policy makers in China have establish long term green economic growth, attention needs to paid to all steps taken to achieve these goals because building a sustainable future requires adopting measures and approaches that must be feasible and sustainable in practice. The government must ensure it enhances good administrative efficiency and incorporate a market based approach in supporting and creating a stable and sustainable green economy that flourishes. Policies to enhance the technological development of the central and western provinces should be enacted and these provinces must be technological empowered like their eastern counterpart. Acknowledgements The paper is supported by the Grant for Collaborative Innovation Center for Energy Economics and Energy Policy (No: 1260Z0210011), Xiamen University Flourish Plan Special Funding (No:1260-Y07200) and China National Social Science Fund (No. 15ZD058).

References Ahmed, E.M., 2012. Green TFP intensity impact on sustainable east asian productivity growth. Econ. Analysis Policy 42 (1), 67e68. Baur, D.G., 2013. The structure and degree of dependence: a quantile regression approach. J. Bank. Finance 37, 786e798. Chen, S., Golly, 2014. Green productivity growth in China's industrial economy. Energy Econ. 44, 89e98. Chen, P.C., Yu, M., Chang, C.C., Hsu, S.H., 2008. Total factor productivity growth in China's agricultural sector. China Econ. Rev. 19 (4), 580e593. Choi, Y., 2015. Intermediary propositions for green growth with sustainable governance. Sustainability 7 (11), 14785e14801. Chung, Y.H., Fare, R., Grosskopf, S., 1997. Productivity and undesirable outputs: a directional distance function approach. J. Environ. Manag. 51, 229e240. Fare, R., Grosskopf, S., Pasurka Jr., C.A., 2007. Environmental production functions and environmental directional distance functions. Energy 32, 1055e1066. Koenker, R., 2005. Quantile Regression. Econometric Society Monograph Series. Cambridge University Press, New York. Koenker, R., D'Orey, V., 1987. Algorithm AS 229: computing regression quantiles. J. R. Stat. Soc. Ser. C. Appl. Stat. 36 (3), 383e393. Koenker, R., Hallock, K.F., 2001. Quantile regression. J. Econ. Perspect. 15, 143e156. Li, K.W., 2009. China's total factor productivity estimates by region, investment sources and ownership. Econ. Syst. 33 (3), 213e230. Li, K., Lin, B., 2015. Metafrontier energy efficiency with CO2 emissions and its convergence analysis for China. Energy Econ. 48, 230e241. Li, K., Lin, B., 2015. Measuring green productivity growth of Chinese industrial sectors during 1998-2011. China Econ. Rev. 36, 279e295. Li, Ke, Song, Malin, 2016. Green development performance in China: a metafrontier non-radial approach. Sustainability 8, (8), 3390e30219. Lin, B., Du, K., 2015. Energy and CO2 emissions performance in China's regional economies: do market-oriented reforms matter? Energy Policy 78, 113e124. OECD, 2011. What Is Green Growth and How Can it Help Deliver Sustainable Development?. Pan, Jihua, Ma, Haibing, Zhang, Ying, 2011. Green Economy and Green Jobs in China: Current Status and Potential for 2020. World watch Institute. ISBN 978-09835437-0-1. Rashidi, K., Farzipoor Saen, R., 2015. Measuring eco-efficiency based on green indicators and potentials in energy saving and undesirable output abatement. Energy Econ. 50, 18e26. Song, X., Hao, Y., Zhu, X., 2015. Analysis of the environmental efficiency of the Chinese transportation sector using an undesirable output slacks-based

B. Lin, N.I. Benjamin / Journal of Cleaner Production 162 (2017) 764e775 measure data envelopment analysis model. Sustainability 7 (7), 9187e9206. Sueyoshi, T., Goto, M., 2013. DEA environmental assessment in a time horizon: Malmquist index on fuel mix, electricity and CO2 of industrial nations. Energy Econ. 40, 370e382. Tulkens, H., Eechaut, P.V., 1995. Non-parametric efficiency, progress and regress measures for panel data: methodological aspects. Eur. J. Operational Res. 80, 474e499. Yang, Q., Wan, X., Ma, H., 2015. Assessing green development efficiency of municipalities and provinces in China. Integrating models of super efficiency DEA and malmquist index. Sustainability 7 (4), 4492e4510. Yu Ying Lin, E., Chen, C.C., 2013. Measuring green productivity of country: a

775

generalized metafrontier Malquist productivity index approach. Energy 55, 340e353. Zhang, N., Choi, Y., 2013. Total factor carbon emissions performance of fossil fuel power plants in China: a metafrontier non-radial Malmquist index analysis. Energy Econ. 40, 549e559. Zhang, N., Zhou, P., Kung, C.C., 2015. Total factor carbon emission performance of the Chinese transportation industry: a bootstrapped non-radial Malmquist index analysis. Renew. ans Sustain. Energy Rev. 41, 584e593. Zhou, P., Ang, B.W., Wang, H., 2012. Energy and CO2 emission performance in electricity generation: a non-radial directional distance function approach. Eur. J. Operational Res. 221, 625e635.