- Email: [email protected]
Measuring environmental performance in China’s industrial sectors with non-radial DEA
Mathematical and Computer Modelling 58 (2013) 1047–1056
Contents lists available at SciVerse ScienceDirect
Mathematical and Computer Modelling journal homepage: www.elsevier.com/locate/mcm
Measuring environmental performance in China’s industrial sectors with non-radial DEA F.Y. Meng a , L.W. Fan b , P. Zhou a,∗ , D.Q. Zhou a a
College of Economics and Management, Nanjing University of Aeronautics and Astronautics, 29 Jiangjun Avenue, Nanjing 211106, China
b
Business School, Hohai University, 8 Focheng West Road, Nanjing 211100, China
article
info
Article history: Received 11 December 2011 Received in revised form 21 August 2012 Accepted 24 August 2012 Keywords: Data envelopment analysis (DEA) Undesirable outputs Environmental performance Industrial sector China
abstract This paper proposes a non-radial DEA approach consisting of both a static and a dynamic environmental performance index (EPI) for measuring environmental performance. The static EPI is defined as the ratio of a non-radial efficiency measure for reducing undesirable outputs to that for increasing desirable outputs. The proposed non-radial DEA approach has been applied to model the environmental performance of industrial sectors in different provinces of China from 1998 to 2009. It has been found that the static non-radial EPI has a higher discriminating power than the EPI derived from the non-radial undesirable outputs orientation DEA models. Our empirical results also show that the environmental performance of industrial sectors in China improves by 58% in 1998–2009 that is mainly driven by the technological change. © 2012 Elsevier Ltd. All rights reserved.
1. Introduction China’s rapid economic development in the past three decades was mainly attributed to energy and labor-intensive industrial sectors, which not only increased its carbon dioxide (CO2 ) emissions but also caused serious environmental pollution [1]. Since the industrialization process is still far from being completed, China has to take effective measures to control its energy consumption by industrial sectors in order to achieve the national targets of decreasing energy and CO2 emission intensities by 16% and 17% till 2015. Actually, the Ministry of Industry and Information Technology of China has proposed the target of decreasing the energy use and CO2 emissions per unit of industrial value added in 2015 by 18% compared to that in 2010. Meanwhile, the comprehensive utilization rate of industrial solid waste needs to increase to 72%. Clearly, industrial sectors will inevitably play a significant role in promoting China to become a low-carbon, resourceconservation and environment-friendly society. Measuring environmental performance has been widely advocated since it can provide quantitative information for environmental policy analysis and decision making [2]. In China, the industrial sector accounted for more than half of the main pollutants. Measuring China’s industrial environmental performance can help to make environmental policy analysis in China more quantitative and empirically grounded. Since different provinces in China have different industrial structures and development patterns, the environmental performance of industrial sectors in China may vary significantly across different provinces. Assessing the industrial environmental performance of different provinces in China would shed insights on the scientific evaluation of the efforts made by regional governments in energy conservation and environmental protection. The measurement of environmental performance is often in the form of an environmental performance index (EPI), which can be constructed by different weighting and aggregation techniques such as data envelopment analysis (DEA) [3]. DEA,
∗
Corresponding author. Tel.: +86 25 84893751x813; fax: +86 25 84892751. E-mail addresses: [email protected], [email protected] (P. Zhou).
0895-7177/$ – see front matter © 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.mcm.2012.08.009
1048
F.Y. Meng et al. / Mathematical and Computer Modelling 58 (2013) 1047–1056
developed by Charnes et al. [4] is a well-established methodology for evaluating the relative efficiency of a set of comparable entities often called decision making units (DMUs) with multiple inputs and outputs. Since Charnes et al.’s seminal work, DEA has been widely investigated and also gained great popularity in different application areas. The review study by Cook and Seiford [5] provides a sketch of some important methodological developments in DEA. In the context of energy and environmental modeling, [2] conduct a survey study on the use of DEA in which environmental performance measurement is identified as an important application area. The application of DEA to measure environmental performance is built upon the fact that outputs are classified into desirable and undesirable outputs [6]. Traditional DEA models mainly handle desirable outputs that have the property of ‘the more the better’. How to incorporate undesirable outputs with the ‘the less the better’ property is therefore an important topic in DEA and has been investigated by a number of previous studies, e.g. [6–8,2,9]. Among the alternative approaches to modeling undesirable outputs, the concept of environmental DEA technology proposed by Färe et al. [10] has been widely adopted in modeling environmental performance.1 See, for example, [12–17,2,18–25]. Besides, environmental DEA technology has also been used in other application areas, e.g. resource allocation with environmental constraints [26]. Most previous studies using DEA to measure environmental performance often adopt radial efficiency measures for use as they can be directly linked to the Shephard distance functions holding some desirable mathematical properties. However, radial efficiency measures have some limitations such as weak discriminating power in efficiency evaluation [27]. In addition, radial measures may overestimate technical efficiency due to the existence of nonzero slacks [28]. As such, [17] develop a non-radial DEA approach to modeling environmental performance, which can generate several non-radial EPIs by solving undesirable outputs orientation non-radial DEA models. Despite their usefulness, a limitation of the non-radial EPIs proposed by Zhou et al. [17] is that they cannot measure the degree to which an entity succeeds in reducing undesirable outputs while accounting for increasing desirable outputs.2 The study by Färe et al. [13] has demonstrated the advantage of simultaneously contracting undesirable outputs and expanding desirable outputs in modeling environmental performance. Motivated by it, we extend [17] and follow the ideas used in [13] to develop a formal non-radial EPI, which can be calculated by solving several non-radial DEA models. The proposed index has been used to model the environmental performance of industrial sectors in different provinces over time in China. In order to measure the dynamic changes in environmental performance and quantify their contributing factors, we also employ the non-radial Malmquist EPI in our empirical study. The remainder of this paper is organized as follows. Section 2 introduces our methodology for constructing static and dynamic non-radial EPIs. In Section 3, we apply the proposed approach to modeling the environmental performance of industrial sectors in China from 1998 to 2009. Section 4 concludes this study.
2. Methodology
2.1. Environmental DEA technology Consider a production process that converts input vector x to desirable output vector y and undesirable output vector q. Then the production technology can be conceptually described as S = {(x, y , q) : x can produce (y , q)}. In order to appropriately characterize the production process with both desirable and undesirable outputs being produced, [10] suggest that the following two assumptions be imposed on the production technology: (a) Outputs are weakly disposable, i.e., if (x, y , q) ∈ S and 0 < θ ≤ 1, then (x, θ y , θ q) ∈ S. The weak disposability assumption implies that the proportional reduction in desirable and undesirable outputs is possible, whereas it may not be feasible to reduce undesirable outputs solely. (b) Desirable and undesirable outputs are null-joint, i.e. if (x, y , q) ∈ S and q = 0, then y = 0. This assumption says that undesirable outputs must be produced in order to produce desirable outputs. The only way to remove all the undesirable outputs is to cease the production process. The production technology S has so far been well defined conceptually, which is usually termed as environmental production technology. In order to put it into practice, we need to use the data on the inputs and outputs of all the DMUs. Suppose there are i = 1, 2, . . . , I decision making units (DMUs) that convert M inputs into N desirable outputs and J undesirable outputs. For DMUi , its input, desirable output and undesirable output vectors are respectively xi = (xi1 , . . . , xiM ), yi = (yi1 , . . . , yiN ) and qi = (qi1 , . . . , qiJ ). In the DEA framework, the environmental production technology can be formulated as
1 In addition to environmental DEA technology, undesirable outputs can also be incorporated by using a data translation approach [7]. The study by Hua et al. [11] provides an example along this line of research. 2 It is worth pointing out that the non-radial directional distance function described in [29] is capable of expanding desirable outputs and contracting undesirable outputs simultaneously, which might also be considered as a latest development in non-radial DEA.-.
F.Y. Meng et al. / Mathematical and Computer Modelling 58 (2013) 1047–1056
(x, y , q) :
S =
I
λi xim ≤ xm ,
1049
m = 1, . . . , M
i=1 I
λi yin ≥ yn ,
n = 1, . . . , N
i =1 I
(1)
λi qij = qj ,
j = 1, . . . , J
i =1
λi ≥ 0,
i = 1, . . . , I
.
In Eq. (1), (λ1 , . . . , λI ) denotes the intensity levels at which the production activities are conducted by the I DMUs, and it provides the weights for constructing the piecewise linear production technology. As Eq. (1) is formulated in a DEA framework, S is also termed as an environmental DEA technology. Clearly, the environmental DEA technology S exhibits constant returns to scale. Other cases, e.g. the environmental DEA technologies exhibiting variant returns to scale, have been investigated by Zhou et al. [2] in the context of environmental performance measurement. 2.2. Static non-radial EPI Within a joint production framework of desirable and undesirable outputs, an EPI may be defined as the degree to which the undesirable outputs can be reduced while not violating the restriction of environmental DEA technology. Mathematically, the DEA model for measuring the environmental performance of DMU0 [12] can be formulated as REI (x0 , y0 , q0 ) = min θ I
s.t.
λi xim ≤ x0m ,
m = 1, . . . , M
i =1 I
λi yin ≥ y0n ,
n = 1, . . . , N
λi qij ≤ θ q0j ,
j = 1, . . . , J
(2)
i =1 I i =1
λi ≥ 0,
i = 1, . . . , I .
The optimal objective value of Eq. (2) may be referred to as a radial EPI since in Eq. (2) undesirable outputs are reduced at the same rate. On the basis of Eq. (2), researchers have developed a number of EPIs or different application contexts. A good example is [13] who derive a formal index number for measuring environmental performance. Since a radial EPI has several limitations, e.g. it has weak discrimination power and cannot incorporate the decision makers’ preference information, [17] extend Eq. (2) and propose the following non-radial DEA model for measuring environmental performance: NREIq (x0 , y0 , q0 ) = min
J
wqj θj
j =1 I
s.t.
λi xim ≤ x0m ,
m = 1, . . . , M
i=1 I
λi yin ≥ y0n ,
n = 1, . . . , N
λi qij = θj q0j ,
j = 1, . . . , J
(3)
i =1 I i =1
λi ≥ 0,
i = 1, . . . , I
where wqj is a user-specified weight and represents the desirability degree of decision makers for adjusting undesirable output j. As [17] discussed, the damage costs or marginal abatement costs of undesirable outputs could be used to determine the weights used in Eq. (3).
1050
F.Y. Meng et al. / Mathematical and Computer Modelling 58 (2013) 1047–1056
Despite the usefulness of Eq. (3), it only accounts for the reduction in undesirable outputs. Färe et al. [13] show that considering the increase in desirable outputs and the decrease in undesirable outputs simultaneously in environmental performance measurement has some advantages. Although the study by Färe et al. [13] is based on the Shephard distance function, the resulting EPI is derived by solving a series of radial DEA models. In this paper, we extend [13] and propose a non-radial EPI by solving non-radial DEA models. In addition to Eq. (3), the derivation of the non-radial EPI still requires to solve the following desirable outputs orientation DEA model: NREIy (x0 , y0 , q0 ) = max
N
wyn βn
n =1 I
s.t.
λi xim ≤ x0m ,
m = 1, . . . , M
i=1 I
λi yin ≥ βn y0n ,
n = 1, . . . , N
(4)
i=1 I
λi qij = q0j ,
j = 1, . . . , J
i=1
λi ≥ 0,
i = 1, . . . , I .
Following [13], we define the non-radial EPI as the ratio of the efficiency measure for reducing undesirable outputs and that for increasing desirable outputs, which can be written as SNREI (x0 , y0 , q0 ) =
NREIq (x0 , y0 , q0 ) NREIy (x0 , y0 , q0 )
.
(5)
Since Eq. (5) is mainly used for cross-sectional comparisons between various DMUs, here we term it as a static non-radial EPI (SNREI). Compared to Eqs. (3) and (5) seems to be more general and flexible in modeling environmental performance. For example, if two DMUs can reduce their undesirable outputs by the same rate while DMU1 can increase its desirable outputs by a higher rate than DMU2 , it implies that DMU2 is more successful in producing desirable outputs with the same undesirable outputs. As such, DMU2 has a higher EPI that is the case of Eq. (5). 2.3. Dynamic non-radial EPI As mentioned in Section 2.2, SNREI is a static EPI that provides a basis for comparing the environmental performance of different DMUs during the same period of time. Since this study also aims to measure the dynamic changes in the environmental performance of industrial sectors in China, we here propose to use the non-radial Malmquist EPI developed in [17] for this purpose. In literature, the non-radial Malmquist productivity index has been developed by Chen [30] without considering undesirable outputs. Recently, [31] propose a Malmquist productivity index based on the double frontier DEA models. In this paper, the non-radial Malmquist EPI is referred to as dynamic non-radial EPI (DNREI) to be distinguished from SNREI. Suppose that t1 and t2 refer to two periods of time (t1 < t2 ). Let NREIqs (xr0 , y0r , qr0 ) (s, r = t1 or t2 ) denote the nonradial environmental performance measure of DMU0 based on its inputs/outputs at period r and for the environmental DEA technology at period s derived from Eq. (3). Then the DNREI is defined as follows:
t DNREIt12
(x0 , y0 , q0 ) =
t
t
t
t
t
t
t
t
t
t
t
t
t
t
t
t
NREIq1 (x02 , y02 , q02 ) NREIq2 (x02 , y02 , q02 ) NREIq1 (x01 , y01 , q01 ) NREIq2 (x01 , y01 , q01 )
1/2 .
(6)
t
DNREIt12 (x0 , y0 , q0 ) can be used to measure the environmental performance change of DMU0 from t1 to t2 . If it is greater than unity, it implies that the environmental performance of DMU0 has improved during the period of time. If it is less than unity, it means that the environmental performance of DMU0 has deteriorated. Furthermore, we follow [17,3] to decompose t DNREIt12 (x0 , y0 , q0 ) into two contributing components: t DNREIt12
(x0 , y0 , q0 ) =
t
t
t
t
t
t
t
t
NREIq2 (x02 , y02 , q02 ) NREIq1 (x01 , y01 , q01 )
t
t
t
t
t
t
t
t
t
t
t
t
t
t
t
t
NREIq1 (x02 , y02 , q02 ) NREIq1 (x01 , y01 , q01 ) NREIq2 (x02 , y02 , q02 ) NREIq2 (x01 , y01 , q01 )
1/2 .
(7)
The first term on the right hand side of Eq. (7) is an efficiency change component that measures the change in the relative environmental performance of DMU0 with regards to the frontier of best practice. The second term measures the shift of environmental DEA technology from t1 to t2 .
F.Y. Meng et al. / Mathematical and Computer Modelling 58 (2013) 1047–1056
1051
Table 1 Descriptive statistics of inputs and outputs for thirty provinces 1998–2009. Variable
Unit
Mean
Std. dev.
Industrial energy consumption Labor force Industrial value added Industrial waste gas Industrial waste water Industrial solid waste Carbon dioxide
Million tons of standard coal equivalent Million workers Million RMB 100 million cubic meter Million tons Million tons Million tons
3,648.87 152.32 2,393.83 7,664.30 72,141.24 3,892.87 11,687.56
2,954.45 108.54 2,914.63 7,229.48 59,077.15 3,489.97 9,957.99
3. Empirical study 3.1. Data We apply the non-radial DEA models described in Section 2 to model the static and dynamic environment performance of the industrial sectors in thirty provinces of China from 1998 to 2009. Two inputs used in this study are industrial labor force (ILF) and industrial energy consumption (IEC). The industrial value added is taken as the single desirable output. In the case of undesirable outputs, we choose industrial carbon dioxide (CO2 ) emissions, industrial waste gas (IWG), industrial waste water (IWW) and industrial solid waste (ISW) for use. CO2 is included as an undesirable output since it contributes to global warming and China government has formally declared the target of reducing CO2 emissions per unit of GDP. The other three variables are included since they are main pollutants of industrial sectors and have significant environmental impacts. The data on ILF, IWG, IWW and ISW are collected from various issues of China Statistic Yearbook [32]. The data on IEC are collected and compiled from various issues of China Energy Statistical Yearbook [33]. The data on industrial CO2 emissions are calculated from the industrial energy consumption only for fossil fuel use. Table 1 shows the descriptive statistics of our collected data for the seven variables. 3.2. Static non-radial environmental performance analysis We first use SNREI to evaluate the environmental performance of industrial sectors in different provinces of China from 1998 to 2009. The weights for different undesirable outputs are assumed to be equal to each other. Table 2 provides the results of NREIq and SNREI in 1999 and 2009 calculated by Eqs. (3) and (5) respectively. It is found that for a number of provinces the NREIq and SNREI values as well as the resulting ranking orders are different. For instance, in 2009, Shandong has a larger NREIq than Hubei, which implies that Shandong performs better in generating undesirable outputs. However, the SNREI value of Hubei is greater than that of Shandong since Hubei is more successful in producing desirable outputs given by Eq. (4). Therefore, SNREI is a more encompassing index since it accounts for both the increase in desirable outputs and the decrease in undesirable outputs simultaneously. Table 3 shows the SNREI value for thirty provinces in 2001–2009. It can be easily found that the SNREI ranking changed significantly during the period of time. The ranks for six provinces, namely Beijing, Hebei, Henan, Inner Mongolia, Jilin and Anhui, have improved, which could be explained by the fact that the economic outputs of these provinces have increased rapidly while their undesirable outputs kept growing slowly. In 2001–2009, the growth rates of industrial value added for Heibei, Inner Mongolia, Anhui and Henan were respectively 67%, 149%, 78% and 77.6%, higher than an average value of 62%. For Beijing and Jilin, though the industrial growth rates are a little lower than the average level, their undesirable outputs are much lower than most provinces. The index of industrial waste water and carbon emission for Beijing even decrease by 59% and 39% during this period. On the contrary, the SNREI ranks for six other provinces, namely Heilongjiang, Zhejiang, Fujian, Yunnan, Shaanxi and Xinjiang, decreased significantly. We find that the growth rates of industrial value added for all of them but Shaanxi are lower than the average value. For Fujian and Yunnan, their carbon emissions grow by 225% and 222%, which are much higher than the average level of 110%. The economy of Yunnan was behind other provinces. In order to boost economic growth, the local government accelerated the development of heavy chemical industries, like the iron and steel industry, chemical industry, coal industry and thermal power during the periods of Tenth and Eleventh Five-Year Plan. Though the industrial development promotes economic growth, it also aggravates undesirable outputs, like industrial waste water and carbon emission which leads to the deterioration of their environmental performance. We further study the static environmental performance of industrial sectors in four economic regions namely East, Central, West and Northeast regions. As shown in Fig. 1, the East region performs the best while the West region is the worst, which could be explained by their discrepancy in the level of economic development. It implies that the Chinese government should strive to improve the environmental performance of the East and Central regions in order to improve the environmental performance of industrial sectors throughout the country. Fig. 2 shows the trend in the average SNREI values of the thirty provinces. Obviously, industrial sector in China as a whole shows a positive shift in environmental performance overtime. It is likely due to the fact that effective measures
1052
F.Y. Meng et al. / Mathematical and Computer Modelling 58 (2013) 1047–1056 Table 2 Comparison between NREIq and SNREI in 1999 and 2009. Provinces
Beijing Tianjin Hebei Shanxi Inner Mongolia Liaoning Jilin Heilongjiang Shanghai Jiang Su Zhejiang Anhui Fujian Jiangxi Shandong Henan Hubei Hunan Guangdong Guangxi Hainan Chongqing Sichuan Guizhou Yunnan Shaanxi Gansu Qinghai Ningxia Xinjiang
1999
2009
NREIq
SNREI
NREIq
SNREI
0.511 (13) 0.890 (3) 0.439 (19) 0.368 (23) 0.306 (25) 0.444 (18) 0.392 (22) 0.625 (8) 0.680 (7) 0.698 (6) 1.000 (1) 0.477 (15) 0.848 (4) 0.561 (10) 0.699 (5) 0.563 (9) 0.479 (14) 0.470 (16) 1.000 (2) 0.287 (28) 0.446 (17) 0.322 (24) 0.512 (12) 0.223 (29) 0.548 (11) 0.395 (21) 0.295 (26) 0.292 (27) 0.184 (30) 0.402 (20)
0.511 (11) 0.890 (3) 0.439 (17) 0.368 (19) 0.306 (23) 0.327 (21) 0.222 (28) 0.576 (8) 0.680 (6) 0.609 (7) 1.000 (1) 0.477 (13) 0.848 (4) 0.561 (9) 0.699 (5) 0.450 (15) 0.479 (12) 0.470 (14) 1.000 (2) 0.287 (25) 0.446 (16) 0.322 (22) 0.367 (20) 0.223 (27) 0.548 (10) 0.395 (18) 0.142 (30) 0.292 (24) 0.184 (29) 0.267 (26)
1.000 (1) 0.826 (2) 1.000 (1) 0.307 (23) 0.607 (6) 0.444 (17) 0.561 (8) 0.611 (5) 1.000 (1) 1.000 (1) 0.757 (3) 0.450 (16) 0.486 (13) 0.461 (15) 0.606 (7)∗ 1.000 (1) 0.524 (11)∗ 0.715 (4) 1.000 (1) 0.410 (19) 0.538 (9) 0.464 (14) 0.532 (10) 0.297 (24) 0.403 (20) 0.499 (12) 0.356 (22) 0.390 (21) 0.190 (25) 0.440 (18)
1.000 (1) 0.826 (2) 1.000 (1) 0.307 (21) 0.607 (6) 0.426 (17) 0.509 (10) 0.611 (5) 1.000 (1) 1.000 (1) 0.757 (3) 0.450 (15) 0.486 (12) 0.461 (14) 0.520 (9)∗ 1.000 (1) 0.524 (8)∗ 0.715 (4) 1.000 (1) 0.410 (18) 0.538 (7) 0.464 (13) 0.486 (12) 0.297 (22) 0.403 (19) 0.499 (11) 0.257 (23) 0.390 (20) 0.190 (24) 0.440 (16)
Note: The values in the brackets are the rank orders of each province.
Fig. 1. Mean values of SNREI for four regions in China, 1998–2009.
of environmental protection have been taken by Chinese government. For example, industrial environment performance has been considered as an important criterion in the assessment of regional governments which force them to deal with environmental issues more seriously. Besides, the adjustment of economic and energy structures has also made a contribution to the positive shift of industrial environmental performance. 3.3. Dynamic non-radial environmental performance analysis To measure the dynamic changes in the environmental performance of industrial sectors in China, we further employ Eqs. (3) and (6) to compute DNREI. In the calculation process, there are some infeasible mix-period linear programming problems since not all of the input and output observations in a certain period can be enveloped by the production frontier constructed from the observations in another period. To solve this problem, [14] proposed a three-year window method
F.Y. Meng et al. / Mathematical and Computer Modelling 58 (2013) 1047–1056
1053
Table 3 SNREI values of thirty provinces, 2001–2009. Province
2001
2002
2003
2004
2005
2006
2007
2008
2009
Mean
Beijing Tianjin Hebei Shanxi Inner Mongolia Liaoning Jilin Heilongjiang Shanghai Jiang Su Zhejiang Anhui Fujian Jiangxi Shandong Henan Hubei Hunan Guangdong Guangxi Hainan Chongqing Sichuan Guizhou Yunnan Shaanxi Gansu Qinghai Ningxia Xinjiang
0.598 0.680 0.426 0.306 0.316 0.365 0.286 0.803 0.658 0.662 1.000 0.284 1.000 0.410 0.627 0.455 0.477 0.519 1.000 0.307 0.441 0.394 0.442 0.246 0.492 0.303 0.306 0.285 0.156 0.436
0.738 0.689 0.455 0.346 0.334 0.428 0.340 0.939 0.742 0.579 1.000 0.457 0.809 0.413 0.753 0.481 0.459 0.554 1.000 0.325 0.750 0.416 0.347 0.309 0.502 0.425 0.212 0.361 0.189 0.590
0.912 0.739 0.443 0.337 0.332 0.380 0.326 1.000 0.837 0.675 1.000 0.315 1.000 0.368 0.740 0.574 0.385 0.487 1.000 0.286 0.458 0.460 0.311 0.309 0.466 0.449 0.320 0.373 0.179 0.608
0.901 0.876 0.392 0.327 0.274 0.276 0.305 1.000 1.000 0.741 1.000 0.323 1.000 0.321 0.718 0.472 0.334 0.462 1.000 0.242 0.494 0.431 0.297 0.283 0.486 0.281 0.191 0.340 0.225 0.442
1.000 0.584 0.343 0.338 0.275 0.315 0.303 0.932 1.000 0.678 0.832 0.298 0.537 0.462 0.671 0.469 0.238 0.475 1.000 0.295 0.541 0.415 0.477 0.316 0.423 0.567 0.337 0.277 0.182 0.545
1.000 0.776 0.335 0.261 0.289 0.267 0.357 0.915 1.000 0.672 0.791 0.318 0.540 0.492 0.652 0.486 0.260 0.544 1.000 0.331 0.697 0.413 0.529 0.275 0.434 0.725 0.392 0.278 0.205 0.606
1.000 0.814 0.330 0.317 0.322 0.308 0.461 0.838 1.000 0.696 0.783 0.338 0.533 0.478 0.514 0.483 0.283 0.519 1.000 0.315 0.654 0.439 0.418 0.272 0.430 0.611 0.390 0.289 0.199 0.549
1.000 0.869 1.000 0.339 0.337 0.398 0.471 0.850 1.000 0.704 0.796 0.414 0.534 0.451 0.500 0.637 0.395 0.561 1.000 0.338 0.580 0.290 0.452 0.339 0.437 0.529 0.268 0.294 0.209 0.568
1.000 0.826 1.000 0.307 0.607 0.426 0.509 0.611 1.000 1.000 0.757 0.450 0.486 0.461 0.520 1.000 0.524 0.715 1.000 0.410 0.538 0.464 0.486 0.297 0.403 0.499 0.257 0.390 0.190 0.440
0.807* 0.772* 0.503 0.322 0.333 0.336 0.334 0.812* 0.853* 0.695 0.913* 0.368 0.774* 0.439 0.644 0.539 0.379 0.522 1.000* 0.312 0.524 0.386 0.411 0.277 0.473 0.516 0.262 0.312 0.186 0.486
Note: The mean values are from 1988 to 2009. * The mean value is larger than 0.7.
Fig. 2. Trend in SNREI mean values over time.
by constructing the environmental DEA technology in period t using the observations in period t, t − 1 and t − 2. In our study, we follow the procedure used by Färe et al. [14] to compute the DNREI values for each province from 1998/1999 to 2008/2009. Table 4 shows the results obtained. It can be observed from Table 4 that most provinces experienced a change for each consecutive two-year period. Some provinces, such as Zhejiang and Guangdong, remain DNREI values greater than 1 for each consecutive two-year period. It indicates that the environmental performance of these provinces keeps improving year by year. But for some other provinces, taking Anhui and Henan as examples, their DNREI values become less than 1 since 2007. Since the year 2005, Anhui province has taken out several polices to promote energy conservation and emission reduction, like strengthening the supervision and phasing out low capacity enterprises. In 2006–2009, per unit energy consumption of industrial added value fell 31.9% accumulatively. The reduction of energy consumption in the industrial sector contributed greatly to the improvement of environmental performance. To examine the driving forces behind the change in environmental performance of different provinces, we use Eq. (7) to decompose DNREI into its two contributing components, namely efficiency change (ECH) and technological change (TCH). We further calculate the geometric mean of DNREI and its components for each of the four regions. Fig. 3 shows the result
1054
F.Y. Meng et al. / Mathematical and Computer Modelling 58 (2013) 1047–1056
Table 4 DNREI estimates from 1998/1999 to 2008/2009. Province
98/99
99/00
00/01
01/02
02/03
03/04
04/05
05/06
06/07
07/08
08/09
Beijing Tianjin Hebei Shanxi Inner Mongolia Liaoning Jilin Heilongjiang Shanghai Jiang Su Zhejiang Anhui Fujian Jiangxi Shandong Henan Hubei Hunan Guangdong Guangxi Hainan Chongqing Sichuan Guizhou Yunnan Shaanxi Gansu Qinghai Ningxia Xinjiang
0.985 0.908 1.048 0.947 1.087 0.900 1.007 1.026 1.095 1.125 1.000 1.136 1.348 0.920 1.074 1.158 1.121 1.063 1.126 1.151 0.698 1.479 1.044 1.055 1.151 2.851 1.025 1.080 1.094 0.995
0.920 1.139 0.987 1.210 1.050 1.118 0.969 0.867 1.034 1.067 1.000 1.178 1.010 1.533 1.102 1.081 0.989 0.970 1.000 0.992 1.146 1.057 1.092 0.921 1.157 1.027 1.067 1.074 1.374 0.845
0.987 1.214 1.130 1.039 0.965 1.006 0.950 0.982 1.036 1.111 1.000 1.060 1.131 1.019 1.090 1.069 0.893 1.046 1.028 1.098 0.987 0.888 1.128 1.043 1.060 1.025 0.905 1.008 0.907 1.064
0.895 1.087 1.031 0.958 1.042 0.966 0.959 0.938 0.935 1.067 1.000 1.012 1.306 1.074 0.916 1.001 1.094 1.051 1.056 1.026 0.588 1.039 1.136 0.877 1.078 1.066 1.024 0.871 0.908 0.881
0.877 1.006 1.056 0.955 0.950 1.081 1.095 0.929 0.915 0.963 1.000 1.045 1.050 1.037 1.070 1.016 1.176 1.209 1.027 1.140 1.756 0.973 1.076 0.953 1.055 0.926 1.096 1.016 1.139 1.005
0.956 0.868 1.078 0.922 1.132 1.130 0.988 0.947 0.845 1.020 1.024 1.004 1.297 1.087 1.017 1.017 1.077 1.054 1.029 1.129 0.966 1.087 1.067 1.013 0.896 1.061 0.911 1.006 0.796 1.050
0.883 1.217 1.116 0.925 0.953 1.073 1.093 1.159 0.961 1.100 1.229 1.052 1.953 0.786 1.072 1.024 1.221 1.014 1.038 0.888 0.961 1.093 0.902 0.888 1.157 0.782 0.978 1.226 1.264 1.008
1.035 0.984 1.058 1.128 0.977 1.075 1.030 1.143 1.003 1.061 1.113 1.123 1.081 1.114 1.007 1.042 1.063 0.923 1.034 0.990 0.826 1.095 1.076 1.171 1.031 0.916 0.901 1.038 0.920 0.927
0.973 0.976 1.036 0.953 0.876 0.931 0.961 1.149 1.017 1.009 1.073 1.109 1.065 1.080 1.146 1.082 1.008 1.098 1.063 1.100 1.112 0.983 1.302 0.947 1.036 1.258 0.989 0.971 1.054 1.129
0.960 0.918 0.531 0.901 0.938 0.985 0.991 0.971 1.012 1.013 1.008 0.991 0.995 1.045 1.007 0.918 0.961 0.938 1.036 0.938 1.137 1.050 0.811 0.766 0.977 1.115 0.988 0.971 0.921 0.964
1.038 1.109 1.081 1.162 0.758 0.947 1.036 1.458 1.052 0.872 1.079 0.963 1.136 1.019 1.075 0.861 0.955 0.937 1.020 0.971 1.121 0.927 0.982 1.204 1.145 1.104 1.104 0.929 1.146 1.363
Fig. 3. Mean values of DNREI, ECH and TCH for four regions, 1998–2009.
obtained. It can be found that the average values of DNREI and TCH for all the regions are larger than unity, which indicates that the improvement in environmental performance of industrial sectors in China is mainly attributable to technological change. In addition, we can also observe that the Central region showed the most obvious increase in environmental performance as well as technological change. To examine the overall environmental performance changes from 1998 to 2009, we compute the cumulative DNREI values in 2009 with 1998 as the base year. Both the DNREI values and their contributing components for the thirty provinces are provided in Table 5. As a whole, we find that the environmental performance of industrial sectors in China improves by 58% in 1998–2009. Not surprisingly, technological change contributes to the improvement of environmental performance positively. On the other hand, the efficiency change component is less than unity and plays a negative role in improving the environmental performance. A possible explanation is that the discrepancy between the industrial sectors in different provinces becomes larger so that most of them become further from the frontier of best practice.
F.Y. Meng et al. / Mathematical and Computer Modelling 58 (2013) 1047–1056
1055
Table 5 Cumulative DNREI and its components in 2009. Provinces
Cumulative DNREI
Efficiency change
Technological change
Beijing Tianjin Hebei Shanxi Inner Mongolia Liaoning Jilin Heilongjiang Shanghai Jiang Su Zhejiang Anhui Fujian Jiangxi Shandong Henan Hubei Hunan Guangdong Guangxi Hainan Chongqing Sichuan Guizhou Yunnan Shaanxi Gansu Qinghai Ningxia Xinjiang Mean
0.595 1.420 0.961 1.046 0.715 1.194 1.068 1.553 0.884 1.451 1.634 1.881 6.964 1.752 1.719 1.260 1.646 1.304 1.555 1.458 0.885 1.747 1.689 0.779 1.987 3.485 0.962 1.159 1.452 1.152 1.579
0.451 0.871 0.411 1.009 0.497 0.795 0.624 0.939 0.659 0.687 1.321 1.068 2.058 0.996 1.106 0.575 0.905 0.618 1.000 0.716 0.514 0.891 0.886 0.712 1.381 2.004 0.753 0.727 0.951 0.816 0.898
1.318 1.631 2.339 1.036 1.437 1.503 1.713 1.655 1.342 2.112 1.237 1.762 3.383 1.759 1.554 2.191 1.819 2.112 1.555 2.038 1.723 1.960 1.907 1.095 1.439 1.739 1.277 1.594 1.528 1.413 1.706
4. Conclusion Global awareness on environmental problems has created much interest in analyzing environmental performance. Most of the previous studies only account for the reduction in undesirable outputs in modeling environmental performance. In this paper, we develop a static non-radial environmental performance index called SNREI, which considers both the reduction in undesirable outputs and the expansion of desirable outputs simultaneously. A dynamic non-radial environmental performance index is also introduced to measure the dynamic changes in environmental performance over time. The two environmental performance indexes can be derived by solving a series of non-radial DEA models. The proposed non-radial approach has also been applied to model the environmental performance of industrial sectors in different provinces from 1998 to 2009. It is found that SNREI has a higher discriminating power in modeling environmental performance compared to the undesirable outputs orientation non-radial DEA models. Our empirical results also show that the environmental performance of industrial sectors in China as a whole improves by 58% from 1998 to 2009, which is mainly attributed to technological change. Our empirical results have some policy implications. First, the proposed indexes, i.e. SNREI and DNREI, are useful for the Chinese government to build an evaluation system of environmental performance of the industrial sectors at province or even lower levels. Second, the Chinese government should pay more attention to the western region and promote interregional exchanges in order to reduce the gap of environment performance between different regions. Third, given that there exists an imbalance between technological change and efficiency change in improving the dynamic environmental performance, the local government may strengthen the management and supervision of industrial sectors, improve regulation rules and encourage citizens to adopt green and low energy consumption lifestyles. Meanwhile, the whole country should continue to promote technological progress. Since this study takes equal weights for use in calculating the EPIs, further research may be carried out to use different weight settings to examine how SNREI and DNREI values are affected by the weights. Given data availability, it is also possible to employ the methodology described in this paper to model China’s industrial environmental performance at lower levels, which may provide more insights for China to build an environment-friendly society. Acknowledgments The authors are grateful for the financial support provided by the National Natural Science Foundation of China (no. 70903031), the Jiangsu Qing Lan Project, the Program for New Century Excellent Talents in University (no. NCET-10-0073),
1056
F.Y. Meng et al. / Mathematical and Computer Modelling 58 (2013) 1047–1056
the Fundamental Research Funds for the Central Universities (no. 2012B04314), and the Humanity and Social Science Foundation of the Ministry of Education (nos. 10YJC630020 and 12YJCZH039). References [1] M. Song, S. Wang, H. Yu, L. Yang, J. Wu, To reduce energy consumption and to maintain rapid economic growth: analysis of the condition in China based on expended IPAT model, Renewable and Sustainable Energy Reviews 15 (2011) 5129–5134. [2] P. Zhou, B.W. Ang, K. Poh, Measuring environmental performance under different environmental DEA technologies, Energy Economics 30 (2008) 1–14. [3] P. Zhou, B.W. Ang, J.Y. Han, Total factor carbon emission performance: a Malmquist index analysis, Energy Economics 32 (2010) 194–201. [4] A. Charnes, W.W. Cooper, E. Rhodes, Measuring the efficiency of decision making units, European Journal of Operational Research 2 (1978) 429–444. [5] W.D. Cook, L.M. Seiford, Data envelopment analysis (DEA)—thirty years on, European Journal of Operational Research 192 (2009) 1–17. [6] H. Scheel, Undesirable outputs in efficiency valuations, European Journal of Operational Research 132 (2001) 400–410. [7] L.M. Seiford, J. Zhu, Modeling undesirable factors in efficiency evaluation, European Journal of Operational Research 142 (2002) 16–20. [8] T. Kuosmanen, Weak disposability in nonparametric production analysis with undesirable outputs, American Journal of Agricultural Economics 87 (2005) 1077–1082. [9] V.V. Podinovski, T. Kuosmanen, Modeling weak disposability in data envelopment analysis under relaxed convexity assumptions, European Journal of Operational Research 211 (2011) 577–585. [10] R. Färe, S. Grosskopf, C.A.K. Lovell, C. Pasurka, Multilateral productivity comparisons when some outputs are undesirable: a nonparametric approach, The Review of Economics and Statistics 71 (1989) 90–98. [11] Z. Hua, Y. Bian, L. Liang, Eco-efficiency analysis of paper mills along the Huai River: an extended DEA approach, Omega 35 (2007) 578–587. [12] D. Tyteca, Linear programming models for the measurement of environmental performance of firms—concepts and empirical results, Journal of Productivity Analysis 8 (1997) 183–197. [13] R. Färe, S. Grosskopf, F Hernandez-Sancho, Environmental performance: an index number approach, Resource and Energy Economics 26 (2004) 343–352. [14] R. Färe, S. Grosskopf, C.A. Pasurka Jr., Pollution abatement activities and traditional productivity, Ecological Economics 62 (2007) 673–682. [15] R. Färe, S. Grosskopf, C.A. Pasurka Jr., Toxic releases: an environmental performance index for coal-fired power plants, Energy Economics 32 (2010) 158–165. [16] P. Zhou, B.W. Ang, K.L. Poh, Slacks-based efficiency measures for modeling environmental performance, Ecological Economics 60 (2006) 111–118. [17] P. Zhou, K.L. Poh, B.W. Ang, A non-radial DEA approach to measuring environmental performance, European Journal of Operational Research 178 (2007) 1–9. [18] P. Zhou, B.W. Ang, D.Q. Zhou, Weighting and aggregation in composite indicator construction: a multiplicative optimization approach, Social Indicators Research 96 (2010) 169–181. [19] G.E. Halkos, G. Nickolaos, Tzeremes, Exploring the existence of Kuznets curve in countries’ environmental efficiency using DEA window analysis, Ecological Economics 68 (2009) 2168–2176. [20] Y. Bian, F. Yang, Resource and environmental efficiency analysis of provinces in China: a DEA approach based on Shannon’s entropy, Energy Policy 38 (2010) 1909–1917. [21] S.K. Mandal, S. Madheswaran, Environmental efficiency of the Indian cement industry: an interstate analysis, Energy Policy 38 (2010) 1108–1118. [22] T. Sueyoshi, M. Goto, Methodological comparison between two unified (operational and environmental) efficiency measurements for environmental assessment, European Journal of Operational Research 210 (2011) 684–693. [23] T. Sueyoshi, M. Goto, DEA approach for unified efficiency measurement: assessment of Japanese fossil fuel power generation, Energy Economics 33 (2011) 202–303. [24] K. Wang, Y.M. Wei, X. Zhang, A comparative analysis of China’s regional energy and emission performance: Which is the better way to deal with undesirable outputs? Energy Policy 46 (2012) 574–584. [25] K. Wang, S. Yu, W. Zhang, China’s regional energy and environment efficiency: a DEA window analysis based dynamic evaluation, Mathematical and Computer Modelling 58 (2013) 1117–1127. [26] J. Wu, Q. An, S. Ali, L. Liang, DEA based resource allocation considering environmental factors, Mathematical and Computer Modelling 58 (2013) 1128–1137. [27] J. Zhu, Data envelopment analysis with preference structure, Journal of the Operational Research Society 47 (1996) 136–150. [28] H. Fukuyama, W.L. Weber, A directional slacks-based measure of technical efficiency, Socio–Economic Planning Sciences 43 (2009) 274–287. [29] P. Zhou, B.W. Ang, H. Wang, Energy and CO2 emission performance in electricity generation: a non-radial directional distance function approach, European Journal of Operational Research 221 (2012) 625–635. [30] Y. Chen, A non-radial Malmquist productivity index with an illustrative application to Chinese major industries, International Journal of Production Economics 83 (2003) 27–35. [31] Y.M. Wang, Y.X. Lan, Measuring Malmquist productivity index: a new approach based on double frontiers data envelopment analysis, Mathematical and Computer Modelling 54 (2011) 2760–2771. [32] National Bureau of Statistics of China (NBSC), China Statistical Yearbook, China Statistical Press, 1998–2010. [33] Department of Energy Statistics, National Bureau of Statistics of China (DES-NBSC), China Energy Statistical Yearbook, China Statistical Press, 1998–2010.
Contents lists available at SciVerse ScienceDirect
Mathematical and Computer Modelling journal homepage: www.elsevier.com/locate/mcm
Measuring environmental performance in China’s industrial sectors with non-radial DEA F.Y. Meng a , L.W. Fan b , P. Zhou a,∗ , D.Q. Zhou a a
College of Economics and Management, Nanjing University of Aeronautics and Astronautics, 29 Jiangjun Avenue, Nanjing 211106, China
b
Business School, Hohai University, 8 Focheng West Road, Nanjing 211100, China
article
info
Article history: Received 11 December 2011 Received in revised form 21 August 2012 Accepted 24 August 2012 Keywords: Data envelopment analysis (DEA) Undesirable outputs Environmental performance Industrial sector China
abstract This paper proposes a non-radial DEA approach consisting of both a static and a dynamic environmental performance index (EPI) for measuring environmental performance. The static EPI is defined as the ratio of a non-radial efficiency measure for reducing undesirable outputs to that for increasing desirable outputs. The proposed non-radial DEA approach has been applied to model the environmental performance of industrial sectors in different provinces of China from 1998 to 2009. It has been found that the static non-radial EPI has a higher discriminating power than the EPI derived from the non-radial undesirable outputs orientation DEA models. Our empirical results also show that the environmental performance of industrial sectors in China improves by 58% in 1998–2009 that is mainly driven by the technological change. © 2012 Elsevier Ltd. All rights reserved.
1. Introduction China’s rapid economic development in the past three decades was mainly attributed to energy and labor-intensive industrial sectors, which not only increased its carbon dioxide (CO2 ) emissions but also caused serious environmental pollution [1]. Since the industrialization process is still far from being completed, China has to take effective measures to control its energy consumption by industrial sectors in order to achieve the national targets of decreasing energy and CO2 emission intensities by 16% and 17% till 2015. Actually, the Ministry of Industry and Information Technology of China has proposed the target of decreasing the energy use and CO2 emissions per unit of industrial value added in 2015 by 18% compared to that in 2010. Meanwhile, the comprehensive utilization rate of industrial solid waste needs to increase to 72%. Clearly, industrial sectors will inevitably play a significant role in promoting China to become a low-carbon, resourceconservation and environment-friendly society. Measuring environmental performance has been widely advocated since it can provide quantitative information for environmental policy analysis and decision making [2]. In China, the industrial sector accounted for more than half of the main pollutants. Measuring China’s industrial environmental performance can help to make environmental policy analysis in China more quantitative and empirically grounded. Since different provinces in China have different industrial structures and development patterns, the environmental performance of industrial sectors in China may vary significantly across different provinces. Assessing the industrial environmental performance of different provinces in China would shed insights on the scientific evaluation of the efforts made by regional governments in energy conservation and environmental protection. The measurement of environmental performance is often in the form of an environmental performance index (EPI), which can be constructed by different weighting and aggregation techniques such as data envelopment analysis (DEA) [3]. DEA,
∗
Corresponding author. Tel.: +86 25 84893751x813; fax: +86 25 84892751. E-mail addresses: [email protected], [email protected] (P. Zhou).
0895-7177/$ – see front matter © 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.mcm.2012.08.009
1048
F.Y. Meng et al. / Mathematical and Computer Modelling 58 (2013) 1047–1056
developed by Charnes et al. [4] is a well-established methodology for evaluating the relative efficiency of a set of comparable entities often called decision making units (DMUs) with multiple inputs and outputs. Since Charnes et al.’s seminal work, DEA has been widely investigated and also gained great popularity in different application areas. The review study by Cook and Seiford [5] provides a sketch of some important methodological developments in DEA. In the context of energy and environmental modeling, [2] conduct a survey study on the use of DEA in which environmental performance measurement is identified as an important application area. The application of DEA to measure environmental performance is built upon the fact that outputs are classified into desirable and undesirable outputs [6]. Traditional DEA models mainly handle desirable outputs that have the property of ‘the more the better’. How to incorporate undesirable outputs with the ‘the less the better’ property is therefore an important topic in DEA and has been investigated by a number of previous studies, e.g. [6–8,2,9]. Among the alternative approaches to modeling undesirable outputs, the concept of environmental DEA technology proposed by Färe et al. [10] has been widely adopted in modeling environmental performance.1 See, for example, [12–17,2,18–25]. Besides, environmental DEA technology has also been used in other application areas, e.g. resource allocation with environmental constraints [26]. Most previous studies using DEA to measure environmental performance often adopt radial efficiency measures for use as they can be directly linked to the Shephard distance functions holding some desirable mathematical properties. However, radial efficiency measures have some limitations such as weak discriminating power in efficiency evaluation [27]. In addition, radial measures may overestimate technical efficiency due to the existence of nonzero slacks [28]. As such, [17] develop a non-radial DEA approach to modeling environmental performance, which can generate several non-radial EPIs by solving undesirable outputs orientation non-radial DEA models. Despite their usefulness, a limitation of the non-radial EPIs proposed by Zhou et al. [17] is that they cannot measure the degree to which an entity succeeds in reducing undesirable outputs while accounting for increasing desirable outputs.2 The study by Färe et al. [13] has demonstrated the advantage of simultaneously contracting undesirable outputs and expanding desirable outputs in modeling environmental performance. Motivated by it, we extend [17] and follow the ideas used in [13] to develop a formal non-radial EPI, which can be calculated by solving several non-radial DEA models. The proposed index has been used to model the environmental performance of industrial sectors in different provinces over time in China. In order to measure the dynamic changes in environmental performance and quantify their contributing factors, we also employ the non-radial Malmquist EPI in our empirical study. The remainder of this paper is organized as follows. Section 2 introduces our methodology for constructing static and dynamic non-radial EPIs. In Section 3, we apply the proposed approach to modeling the environmental performance of industrial sectors in China from 1998 to 2009. Section 4 concludes this study.
2. Methodology
2.1. Environmental DEA technology Consider a production process that converts input vector x to desirable output vector y and undesirable output vector q. Then the production technology can be conceptually described as S = {(x, y , q) : x can produce (y , q)}. In order to appropriately characterize the production process with both desirable and undesirable outputs being produced, [10] suggest that the following two assumptions be imposed on the production technology: (a) Outputs are weakly disposable, i.e., if (x, y , q) ∈ S and 0 < θ ≤ 1, then (x, θ y , θ q) ∈ S. The weak disposability assumption implies that the proportional reduction in desirable and undesirable outputs is possible, whereas it may not be feasible to reduce undesirable outputs solely. (b) Desirable and undesirable outputs are null-joint, i.e. if (x, y , q) ∈ S and q = 0, then y = 0. This assumption says that undesirable outputs must be produced in order to produce desirable outputs. The only way to remove all the undesirable outputs is to cease the production process. The production technology S has so far been well defined conceptually, which is usually termed as environmental production technology. In order to put it into practice, we need to use the data on the inputs and outputs of all the DMUs. Suppose there are i = 1, 2, . . . , I decision making units (DMUs) that convert M inputs into N desirable outputs and J undesirable outputs. For DMUi , its input, desirable output and undesirable output vectors are respectively xi = (xi1 , . . . , xiM ), yi = (yi1 , . . . , yiN ) and qi = (qi1 , . . . , qiJ ). In the DEA framework, the environmental production technology can be formulated as
1 In addition to environmental DEA technology, undesirable outputs can also be incorporated by using a data translation approach [7]. The study by Hua et al. [11] provides an example along this line of research. 2 It is worth pointing out that the non-radial directional distance function described in [29] is capable of expanding desirable outputs and contracting undesirable outputs simultaneously, which might also be considered as a latest development in non-radial DEA.-.
F.Y. Meng et al. / Mathematical and Computer Modelling 58 (2013) 1047–1056
(x, y , q) :
S =
I
λi xim ≤ xm ,
1049
m = 1, . . . , M
i=1 I
λi yin ≥ yn ,
n = 1, . . . , N
i =1 I
(1)
λi qij = qj ,
j = 1, . . . , J
i =1
λi ≥ 0,
i = 1, . . . , I
.
In Eq. (1), (λ1 , . . . , λI ) denotes the intensity levels at which the production activities are conducted by the I DMUs, and it provides the weights for constructing the piecewise linear production technology. As Eq. (1) is formulated in a DEA framework, S is also termed as an environmental DEA technology. Clearly, the environmental DEA technology S exhibits constant returns to scale. Other cases, e.g. the environmental DEA technologies exhibiting variant returns to scale, have been investigated by Zhou et al. [2] in the context of environmental performance measurement. 2.2. Static non-radial EPI Within a joint production framework of desirable and undesirable outputs, an EPI may be defined as the degree to which the undesirable outputs can be reduced while not violating the restriction of environmental DEA technology. Mathematically, the DEA model for measuring the environmental performance of DMU0 [12] can be formulated as REI (x0 , y0 , q0 ) = min θ I
s.t.
λi xim ≤ x0m ,
m = 1, . . . , M
i =1 I
λi yin ≥ y0n ,
n = 1, . . . , N
λi qij ≤ θ q0j ,
j = 1, . . . , J
(2)
i =1 I i =1
λi ≥ 0,
i = 1, . . . , I .
The optimal objective value of Eq. (2) may be referred to as a radial EPI since in Eq. (2) undesirable outputs are reduced at the same rate. On the basis of Eq. (2), researchers have developed a number of EPIs or different application contexts. A good example is [13] who derive a formal index number for measuring environmental performance. Since a radial EPI has several limitations, e.g. it has weak discrimination power and cannot incorporate the decision makers’ preference information, [17] extend Eq. (2) and propose the following non-radial DEA model for measuring environmental performance: NREIq (x0 , y0 , q0 ) = min
J
wqj θj
j =1 I
s.t.
λi xim ≤ x0m ,
m = 1, . . . , M
i=1 I
λi yin ≥ y0n ,
n = 1, . . . , N
λi qij = θj q0j ,
j = 1, . . . , J
(3)
i =1 I i =1
λi ≥ 0,
i = 1, . . . , I
where wqj is a user-specified weight and represents the desirability degree of decision makers for adjusting undesirable output j. As [17] discussed, the damage costs or marginal abatement costs of undesirable outputs could be used to determine the weights used in Eq. (3).
1050
F.Y. Meng et al. / Mathematical and Computer Modelling 58 (2013) 1047–1056
Despite the usefulness of Eq. (3), it only accounts for the reduction in undesirable outputs. Färe et al. [13] show that considering the increase in desirable outputs and the decrease in undesirable outputs simultaneously in environmental performance measurement has some advantages. Although the study by Färe et al. [13] is based on the Shephard distance function, the resulting EPI is derived by solving a series of radial DEA models. In this paper, we extend [13] and propose a non-radial EPI by solving non-radial DEA models. In addition to Eq. (3), the derivation of the non-radial EPI still requires to solve the following desirable outputs orientation DEA model: NREIy (x0 , y0 , q0 ) = max
N
wyn βn
n =1 I
s.t.
λi xim ≤ x0m ,
m = 1, . . . , M
i=1 I
λi yin ≥ βn y0n ,
n = 1, . . . , N
(4)
i=1 I
λi qij = q0j ,
j = 1, . . . , J
i=1
λi ≥ 0,
i = 1, . . . , I .
Following [13], we define the non-radial EPI as the ratio of the efficiency measure for reducing undesirable outputs and that for increasing desirable outputs, which can be written as SNREI (x0 , y0 , q0 ) =
NREIq (x0 , y0 , q0 ) NREIy (x0 , y0 , q0 )
.
(5)
Since Eq. (5) is mainly used for cross-sectional comparisons between various DMUs, here we term it as a static non-radial EPI (SNREI). Compared to Eqs. (3) and (5) seems to be more general and flexible in modeling environmental performance. For example, if two DMUs can reduce their undesirable outputs by the same rate while DMU1 can increase its desirable outputs by a higher rate than DMU2 , it implies that DMU2 is more successful in producing desirable outputs with the same undesirable outputs. As such, DMU2 has a higher EPI that is the case of Eq. (5). 2.3. Dynamic non-radial EPI As mentioned in Section 2.2, SNREI is a static EPI that provides a basis for comparing the environmental performance of different DMUs during the same period of time. Since this study also aims to measure the dynamic changes in the environmental performance of industrial sectors in China, we here propose to use the non-radial Malmquist EPI developed in [17] for this purpose. In literature, the non-radial Malmquist productivity index has been developed by Chen [30] without considering undesirable outputs. Recently, [31] propose a Malmquist productivity index based on the double frontier DEA models. In this paper, the non-radial Malmquist EPI is referred to as dynamic non-radial EPI (DNREI) to be distinguished from SNREI. Suppose that t1 and t2 refer to two periods of time (t1 < t2 ). Let NREIqs (xr0 , y0r , qr0 ) (s, r = t1 or t2 ) denote the nonradial environmental performance measure of DMU0 based on its inputs/outputs at period r and for the environmental DEA technology at period s derived from Eq. (3). Then the DNREI is defined as follows:
t DNREIt12
(x0 , y0 , q0 ) =
t
t
t
t
t
t
t
t
t
t
t
t
t
t
t
t
NREIq1 (x02 , y02 , q02 ) NREIq2 (x02 , y02 , q02 ) NREIq1 (x01 , y01 , q01 ) NREIq2 (x01 , y01 , q01 )
1/2 .
(6)
t
DNREIt12 (x0 , y0 , q0 ) can be used to measure the environmental performance change of DMU0 from t1 to t2 . If it is greater than unity, it implies that the environmental performance of DMU0 has improved during the period of time. If it is less than unity, it means that the environmental performance of DMU0 has deteriorated. Furthermore, we follow [17,3] to decompose t DNREIt12 (x0 , y0 , q0 ) into two contributing components: t DNREIt12
(x0 , y0 , q0 ) =
t
t
t
t
t
t
t
t
NREIq2 (x02 , y02 , q02 ) NREIq1 (x01 , y01 , q01 )
t
t
t
t
t
t
t
t
t
t
t
t
t
t
t
t
NREIq1 (x02 , y02 , q02 ) NREIq1 (x01 , y01 , q01 ) NREIq2 (x02 , y02 , q02 ) NREIq2 (x01 , y01 , q01 )
1/2 .
(7)
The first term on the right hand side of Eq. (7) is an efficiency change component that measures the change in the relative environmental performance of DMU0 with regards to the frontier of best practice. The second term measures the shift of environmental DEA technology from t1 to t2 .
F.Y. Meng et al. / Mathematical and Computer Modelling 58 (2013) 1047–1056
1051
Table 1 Descriptive statistics of inputs and outputs for thirty provinces 1998–2009. Variable
Unit
Mean
Std. dev.
Industrial energy consumption Labor force Industrial value added Industrial waste gas Industrial waste water Industrial solid waste Carbon dioxide
Million tons of standard coal equivalent Million workers Million RMB 100 million cubic meter Million tons Million tons Million tons
3,648.87 152.32 2,393.83 7,664.30 72,141.24 3,892.87 11,687.56
2,954.45 108.54 2,914.63 7,229.48 59,077.15 3,489.97 9,957.99
3. Empirical study 3.1. Data We apply the non-radial DEA models described in Section 2 to model the static and dynamic environment performance of the industrial sectors in thirty provinces of China from 1998 to 2009. Two inputs used in this study are industrial labor force (ILF) and industrial energy consumption (IEC). The industrial value added is taken as the single desirable output. In the case of undesirable outputs, we choose industrial carbon dioxide (CO2 ) emissions, industrial waste gas (IWG), industrial waste water (IWW) and industrial solid waste (ISW) for use. CO2 is included as an undesirable output since it contributes to global warming and China government has formally declared the target of reducing CO2 emissions per unit of GDP. The other three variables are included since they are main pollutants of industrial sectors and have significant environmental impacts. The data on ILF, IWG, IWW and ISW are collected from various issues of China Statistic Yearbook [32]. The data on IEC are collected and compiled from various issues of China Energy Statistical Yearbook [33]. The data on industrial CO2 emissions are calculated from the industrial energy consumption only for fossil fuel use. Table 1 shows the descriptive statistics of our collected data for the seven variables. 3.2. Static non-radial environmental performance analysis We first use SNREI to evaluate the environmental performance of industrial sectors in different provinces of China from 1998 to 2009. The weights for different undesirable outputs are assumed to be equal to each other. Table 2 provides the results of NREIq and SNREI in 1999 and 2009 calculated by Eqs. (3) and (5) respectively. It is found that for a number of provinces the NREIq and SNREI values as well as the resulting ranking orders are different. For instance, in 2009, Shandong has a larger NREIq than Hubei, which implies that Shandong performs better in generating undesirable outputs. However, the SNREI value of Hubei is greater than that of Shandong since Hubei is more successful in producing desirable outputs given by Eq. (4). Therefore, SNREI is a more encompassing index since it accounts for both the increase in desirable outputs and the decrease in undesirable outputs simultaneously. Table 3 shows the SNREI value for thirty provinces in 2001–2009. It can be easily found that the SNREI ranking changed significantly during the period of time. The ranks for six provinces, namely Beijing, Hebei, Henan, Inner Mongolia, Jilin and Anhui, have improved, which could be explained by the fact that the economic outputs of these provinces have increased rapidly while their undesirable outputs kept growing slowly. In 2001–2009, the growth rates of industrial value added for Heibei, Inner Mongolia, Anhui and Henan were respectively 67%, 149%, 78% and 77.6%, higher than an average value of 62%. For Beijing and Jilin, though the industrial growth rates are a little lower than the average level, their undesirable outputs are much lower than most provinces. The index of industrial waste water and carbon emission for Beijing even decrease by 59% and 39% during this period. On the contrary, the SNREI ranks for six other provinces, namely Heilongjiang, Zhejiang, Fujian, Yunnan, Shaanxi and Xinjiang, decreased significantly. We find that the growth rates of industrial value added for all of them but Shaanxi are lower than the average value. For Fujian and Yunnan, their carbon emissions grow by 225% and 222%, which are much higher than the average level of 110%. The economy of Yunnan was behind other provinces. In order to boost economic growth, the local government accelerated the development of heavy chemical industries, like the iron and steel industry, chemical industry, coal industry and thermal power during the periods of Tenth and Eleventh Five-Year Plan. Though the industrial development promotes economic growth, it also aggravates undesirable outputs, like industrial waste water and carbon emission which leads to the deterioration of their environmental performance. We further study the static environmental performance of industrial sectors in four economic regions namely East, Central, West and Northeast regions. As shown in Fig. 1, the East region performs the best while the West region is the worst, which could be explained by their discrepancy in the level of economic development. It implies that the Chinese government should strive to improve the environmental performance of the East and Central regions in order to improve the environmental performance of industrial sectors throughout the country. Fig. 2 shows the trend in the average SNREI values of the thirty provinces. Obviously, industrial sector in China as a whole shows a positive shift in environmental performance overtime. It is likely due to the fact that effective measures
1052
F.Y. Meng et al. / Mathematical and Computer Modelling 58 (2013) 1047–1056 Table 2 Comparison between NREIq and SNREI in 1999 and 2009. Provinces
Beijing Tianjin Hebei Shanxi Inner Mongolia Liaoning Jilin Heilongjiang Shanghai Jiang Su Zhejiang Anhui Fujian Jiangxi Shandong Henan Hubei Hunan Guangdong Guangxi Hainan Chongqing Sichuan Guizhou Yunnan Shaanxi Gansu Qinghai Ningxia Xinjiang
1999
2009
NREIq
SNREI
NREIq
SNREI
0.511 (13) 0.890 (3) 0.439 (19) 0.368 (23) 0.306 (25) 0.444 (18) 0.392 (22) 0.625 (8) 0.680 (7) 0.698 (6) 1.000 (1) 0.477 (15) 0.848 (4) 0.561 (10) 0.699 (5) 0.563 (9) 0.479 (14) 0.470 (16) 1.000 (2) 0.287 (28) 0.446 (17) 0.322 (24) 0.512 (12) 0.223 (29) 0.548 (11) 0.395 (21) 0.295 (26) 0.292 (27) 0.184 (30) 0.402 (20)
0.511 (11) 0.890 (3) 0.439 (17) 0.368 (19) 0.306 (23) 0.327 (21) 0.222 (28) 0.576 (8) 0.680 (6) 0.609 (7) 1.000 (1) 0.477 (13) 0.848 (4) 0.561 (9) 0.699 (5) 0.450 (15) 0.479 (12) 0.470 (14) 1.000 (2) 0.287 (25) 0.446 (16) 0.322 (22) 0.367 (20) 0.223 (27) 0.548 (10) 0.395 (18) 0.142 (30) 0.292 (24) 0.184 (29) 0.267 (26)
1.000 (1) 0.826 (2) 1.000 (1) 0.307 (23) 0.607 (6) 0.444 (17) 0.561 (8) 0.611 (5) 1.000 (1) 1.000 (1) 0.757 (3) 0.450 (16) 0.486 (13) 0.461 (15) 0.606 (7)∗ 1.000 (1) 0.524 (11)∗ 0.715 (4) 1.000 (1) 0.410 (19) 0.538 (9) 0.464 (14) 0.532 (10) 0.297 (24) 0.403 (20) 0.499 (12) 0.356 (22) 0.390 (21) 0.190 (25) 0.440 (18)
1.000 (1) 0.826 (2) 1.000 (1) 0.307 (21) 0.607 (6) 0.426 (17) 0.509 (10) 0.611 (5) 1.000 (1) 1.000 (1) 0.757 (3) 0.450 (15) 0.486 (12) 0.461 (14) 0.520 (9)∗ 1.000 (1) 0.524 (8)∗ 0.715 (4) 1.000 (1) 0.410 (18) 0.538 (7) 0.464 (13) 0.486 (12) 0.297 (22) 0.403 (19) 0.499 (11) 0.257 (23) 0.390 (20) 0.190 (24) 0.440 (16)
Note: The values in the brackets are the rank orders of each province.
Fig. 1. Mean values of SNREI for four regions in China, 1998–2009.
of environmental protection have been taken by Chinese government. For example, industrial environment performance has been considered as an important criterion in the assessment of regional governments which force them to deal with environmental issues more seriously. Besides, the adjustment of economic and energy structures has also made a contribution to the positive shift of industrial environmental performance. 3.3. Dynamic non-radial environmental performance analysis To measure the dynamic changes in the environmental performance of industrial sectors in China, we further employ Eqs. (3) and (6) to compute DNREI. In the calculation process, there are some infeasible mix-period linear programming problems since not all of the input and output observations in a certain period can be enveloped by the production frontier constructed from the observations in another period. To solve this problem, [14] proposed a three-year window method
F.Y. Meng et al. / Mathematical and Computer Modelling 58 (2013) 1047–1056
1053
Table 3 SNREI values of thirty provinces, 2001–2009. Province
2001
2002
2003
2004
2005
2006
2007
2008
2009
Mean
Beijing Tianjin Hebei Shanxi Inner Mongolia Liaoning Jilin Heilongjiang Shanghai Jiang Su Zhejiang Anhui Fujian Jiangxi Shandong Henan Hubei Hunan Guangdong Guangxi Hainan Chongqing Sichuan Guizhou Yunnan Shaanxi Gansu Qinghai Ningxia Xinjiang
0.598 0.680 0.426 0.306 0.316 0.365 0.286 0.803 0.658 0.662 1.000 0.284 1.000 0.410 0.627 0.455 0.477 0.519 1.000 0.307 0.441 0.394 0.442 0.246 0.492 0.303 0.306 0.285 0.156 0.436
0.738 0.689 0.455 0.346 0.334 0.428 0.340 0.939 0.742 0.579 1.000 0.457 0.809 0.413 0.753 0.481 0.459 0.554 1.000 0.325 0.750 0.416 0.347 0.309 0.502 0.425 0.212 0.361 0.189 0.590
0.912 0.739 0.443 0.337 0.332 0.380 0.326 1.000 0.837 0.675 1.000 0.315 1.000 0.368 0.740 0.574 0.385 0.487 1.000 0.286 0.458 0.460 0.311 0.309 0.466 0.449 0.320 0.373 0.179 0.608
0.901 0.876 0.392 0.327 0.274 0.276 0.305 1.000 1.000 0.741 1.000 0.323 1.000 0.321 0.718 0.472 0.334 0.462 1.000 0.242 0.494 0.431 0.297 0.283 0.486 0.281 0.191 0.340 0.225 0.442
1.000 0.584 0.343 0.338 0.275 0.315 0.303 0.932 1.000 0.678 0.832 0.298 0.537 0.462 0.671 0.469 0.238 0.475 1.000 0.295 0.541 0.415 0.477 0.316 0.423 0.567 0.337 0.277 0.182 0.545
1.000 0.776 0.335 0.261 0.289 0.267 0.357 0.915 1.000 0.672 0.791 0.318 0.540 0.492 0.652 0.486 0.260 0.544 1.000 0.331 0.697 0.413 0.529 0.275 0.434 0.725 0.392 0.278 0.205 0.606
1.000 0.814 0.330 0.317 0.322 0.308 0.461 0.838 1.000 0.696 0.783 0.338 0.533 0.478 0.514 0.483 0.283 0.519 1.000 0.315 0.654 0.439 0.418 0.272 0.430 0.611 0.390 0.289 0.199 0.549
1.000 0.869 1.000 0.339 0.337 0.398 0.471 0.850 1.000 0.704 0.796 0.414 0.534 0.451 0.500 0.637 0.395 0.561 1.000 0.338 0.580 0.290 0.452 0.339 0.437 0.529 0.268 0.294 0.209 0.568
1.000 0.826 1.000 0.307 0.607 0.426 0.509 0.611 1.000 1.000 0.757 0.450 0.486 0.461 0.520 1.000 0.524 0.715 1.000 0.410 0.538 0.464 0.486 0.297 0.403 0.499 0.257 0.390 0.190 0.440
0.807* 0.772* 0.503 0.322 0.333 0.336 0.334 0.812* 0.853* 0.695 0.913* 0.368 0.774* 0.439 0.644 0.539 0.379 0.522 1.000* 0.312 0.524 0.386 0.411 0.277 0.473 0.516 0.262 0.312 0.186 0.486
Note: The mean values are from 1988 to 2009. * The mean value is larger than 0.7.
Fig. 2. Trend in SNREI mean values over time.
by constructing the environmental DEA technology in period t using the observations in period t, t − 1 and t − 2. In our study, we follow the procedure used by Färe et al. [14] to compute the DNREI values for each province from 1998/1999 to 2008/2009. Table 4 shows the results obtained. It can be observed from Table 4 that most provinces experienced a change for each consecutive two-year period. Some provinces, such as Zhejiang and Guangdong, remain DNREI values greater than 1 for each consecutive two-year period. It indicates that the environmental performance of these provinces keeps improving year by year. But for some other provinces, taking Anhui and Henan as examples, their DNREI values become less than 1 since 2007. Since the year 2005, Anhui province has taken out several polices to promote energy conservation and emission reduction, like strengthening the supervision and phasing out low capacity enterprises. In 2006–2009, per unit energy consumption of industrial added value fell 31.9% accumulatively. The reduction of energy consumption in the industrial sector contributed greatly to the improvement of environmental performance. To examine the driving forces behind the change in environmental performance of different provinces, we use Eq. (7) to decompose DNREI into its two contributing components, namely efficiency change (ECH) and technological change (TCH). We further calculate the geometric mean of DNREI and its components for each of the four regions. Fig. 3 shows the result
1054
F.Y. Meng et al. / Mathematical and Computer Modelling 58 (2013) 1047–1056
Table 4 DNREI estimates from 1998/1999 to 2008/2009. Province
98/99
99/00
00/01
01/02
02/03
03/04
04/05
05/06
06/07
07/08
08/09
Beijing Tianjin Hebei Shanxi Inner Mongolia Liaoning Jilin Heilongjiang Shanghai Jiang Su Zhejiang Anhui Fujian Jiangxi Shandong Henan Hubei Hunan Guangdong Guangxi Hainan Chongqing Sichuan Guizhou Yunnan Shaanxi Gansu Qinghai Ningxia Xinjiang
0.985 0.908 1.048 0.947 1.087 0.900 1.007 1.026 1.095 1.125 1.000 1.136 1.348 0.920 1.074 1.158 1.121 1.063 1.126 1.151 0.698 1.479 1.044 1.055 1.151 2.851 1.025 1.080 1.094 0.995
0.920 1.139 0.987 1.210 1.050 1.118 0.969 0.867 1.034 1.067 1.000 1.178 1.010 1.533 1.102 1.081 0.989 0.970 1.000 0.992 1.146 1.057 1.092 0.921 1.157 1.027 1.067 1.074 1.374 0.845
0.987 1.214 1.130 1.039 0.965 1.006 0.950 0.982 1.036 1.111 1.000 1.060 1.131 1.019 1.090 1.069 0.893 1.046 1.028 1.098 0.987 0.888 1.128 1.043 1.060 1.025 0.905 1.008 0.907 1.064
0.895 1.087 1.031 0.958 1.042 0.966 0.959 0.938 0.935 1.067 1.000 1.012 1.306 1.074 0.916 1.001 1.094 1.051 1.056 1.026 0.588 1.039 1.136 0.877 1.078 1.066 1.024 0.871 0.908 0.881
0.877 1.006 1.056 0.955 0.950 1.081 1.095 0.929 0.915 0.963 1.000 1.045 1.050 1.037 1.070 1.016 1.176 1.209 1.027 1.140 1.756 0.973 1.076 0.953 1.055 0.926 1.096 1.016 1.139 1.005
0.956 0.868 1.078 0.922 1.132 1.130 0.988 0.947 0.845 1.020 1.024 1.004 1.297 1.087 1.017 1.017 1.077 1.054 1.029 1.129 0.966 1.087 1.067 1.013 0.896 1.061 0.911 1.006 0.796 1.050
0.883 1.217 1.116 0.925 0.953 1.073 1.093 1.159 0.961 1.100 1.229 1.052 1.953 0.786 1.072 1.024 1.221 1.014 1.038 0.888 0.961 1.093 0.902 0.888 1.157 0.782 0.978 1.226 1.264 1.008
1.035 0.984 1.058 1.128 0.977 1.075 1.030 1.143 1.003 1.061 1.113 1.123 1.081 1.114 1.007 1.042 1.063 0.923 1.034 0.990 0.826 1.095 1.076 1.171 1.031 0.916 0.901 1.038 0.920 0.927
0.973 0.976 1.036 0.953 0.876 0.931 0.961 1.149 1.017 1.009 1.073 1.109 1.065 1.080 1.146 1.082 1.008 1.098 1.063 1.100 1.112 0.983 1.302 0.947 1.036 1.258 0.989 0.971 1.054 1.129
0.960 0.918 0.531 0.901 0.938 0.985 0.991 0.971 1.012 1.013 1.008 0.991 0.995 1.045 1.007 0.918 0.961 0.938 1.036 0.938 1.137 1.050 0.811 0.766 0.977 1.115 0.988 0.971 0.921 0.964
1.038 1.109 1.081 1.162 0.758 0.947 1.036 1.458 1.052 0.872 1.079 0.963 1.136 1.019 1.075 0.861 0.955 0.937 1.020 0.971 1.121 0.927 0.982 1.204 1.145 1.104 1.104 0.929 1.146 1.363
Fig. 3. Mean values of DNREI, ECH and TCH for four regions, 1998–2009.
obtained. It can be found that the average values of DNREI and TCH for all the regions are larger than unity, which indicates that the improvement in environmental performance of industrial sectors in China is mainly attributable to technological change. In addition, we can also observe that the Central region showed the most obvious increase in environmental performance as well as technological change. To examine the overall environmental performance changes from 1998 to 2009, we compute the cumulative DNREI values in 2009 with 1998 as the base year. Both the DNREI values and their contributing components for the thirty provinces are provided in Table 5. As a whole, we find that the environmental performance of industrial sectors in China improves by 58% in 1998–2009. Not surprisingly, technological change contributes to the improvement of environmental performance positively. On the other hand, the efficiency change component is less than unity and plays a negative role in improving the environmental performance. A possible explanation is that the discrepancy between the industrial sectors in different provinces becomes larger so that most of them become further from the frontier of best practice.
F.Y. Meng et al. / Mathematical and Computer Modelling 58 (2013) 1047–1056
1055
Table 5 Cumulative DNREI and its components in 2009. Provinces
Cumulative DNREI
Efficiency change
Technological change
Beijing Tianjin Hebei Shanxi Inner Mongolia Liaoning Jilin Heilongjiang Shanghai Jiang Su Zhejiang Anhui Fujian Jiangxi Shandong Henan Hubei Hunan Guangdong Guangxi Hainan Chongqing Sichuan Guizhou Yunnan Shaanxi Gansu Qinghai Ningxia Xinjiang Mean
0.595 1.420 0.961 1.046 0.715 1.194 1.068 1.553 0.884 1.451 1.634 1.881 6.964 1.752 1.719 1.260 1.646 1.304 1.555 1.458 0.885 1.747 1.689 0.779 1.987 3.485 0.962 1.159 1.452 1.152 1.579
0.451 0.871 0.411 1.009 0.497 0.795 0.624 0.939 0.659 0.687 1.321 1.068 2.058 0.996 1.106 0.575 0.905 0.618 1.000 0.716 0.514 0.891 0.886 0.712 1.381 2.004 0.753 0.727 0.951 0.816 0.898
1.318 1.631 2.339 1.036 1.437 1.503 1.713 1.655 1.342 2.112 1.237 1.762 3.383 1.759 1.554 2.191 1.819 2.112 1.555 2.038 1.723 1.960 1.907 1.095 1.439 1.739 1.277 1.594 1.528 1.413 1.706
4. Conclusion Global awareness on environmental problems has created much interest in analyzing environmental performance. Most of the previous studies only account for the reduction in undesirable outputs in modeling environmental performance. In this paper, we develop a static non-radial environmental performance index called SNREI, which considers both the reduction in undesirable outputs and the expansion of desirable outputs simultaneously. A dynamic non-radial environmental performance index is also introduced to measure the dynamic changes in environmental performance over time. The two environmental performance indexes can be derived by solving a series of non-radial DEA models. The proposed non-radial approach has also been applied to model the environmental performance of industrial sectors in different provinces from 1998 to 2009. It is found that SNREI has a higher discriminating power in modeling environmental performance compared to the undesirable outputs orientation non-radial DEA models. Our empirical results also show that the environmental performance of industrial sectors in China as a whole improves by 58% from 1998 to 2009, which is mainly attributed to technological change. Our empirical results have some policy implications. First, the proposed indexes, i.e. SNREI and DNREI, are useful for the Chinese government to build an evaluation system of environmental performance of the industrial sectors at province or even lower levels. Second, the Chinese government should pay more attention to the western region and promote interregional exchanges in order to reduce the gap of environment performance between different regions. Third, given that there exists an imbalance between technological change and efficiency change in improving the dynamic environmental performance, the local government may strengthen the management and supervision of industrial sectors, improve regulation rules and encourage citizens to adopt green and low energy consumption lifestyles. Meanwhile, the whole country should continue to promote technological progress. Since this study takes equal weights for use in calculating the EPIs, further research may be carried out to use different weight settings to examine how SNREI and DNREI values are affected by the weights. Given data availability, it is also possible to employ the methodology described in this paper to model China’s industrial environmental performance at lower levels, which may provide more insights for China to build an environment-friendly society. Acknowledgments The authors are grateful for the financial support provided by the National Natural Science Foundation of China (no. 70903031), the Jiangsu Qing Lan Project, the Program for New Century Excellent Talents in University (no. NCET-10-0073),
1056
F.Y. Meng et al. / Mathematical and Computer Modelling 58 (2013) 1047–1056
the Fundamental Research Funds for the Central Universities (no. 2012B04314), and the Humanity and Social Science Foundation of the Ministry of Education (nos. 10YJC630020 and 12YJCZH039). References [1] M. Song, S. Wang, H. Yu, L. Yang, J. Wu, To reduce energy consumption and to maintain rapid economic growth: analysis of the condition in China based on expended IPAT model, Renewable and Sustainable Energy Reviews 15 (2011) 5129–5134. [2] P. Zhou, B.W. Ang, K. Poh, Measuring environmental performance under different environmental DEA technologies, Energy Economics 30 (2008) 1–14. [3] P. Zhou, B.W. Ang, J.Y. Han, Total factor carbon emission performance: a Malmquist index analysis, Energy Economics 32 (2010) 194–201. [4] A. Charnes, W.W. Cooper, E. Rhodes, Measuring the efficiency of decision making units, European Journal of Operational Research 2 (1978) 429–444. [5] W.D. Cook, L.M. Seiford, Data envelopment analysis (DEA)—thirty years on, European Journal of Operational Research 192 (2009) 1–17. [6] H. Scheel, Undesirable outputs in efficiency valuations, European Journal of Operational Research 132 (2001) 400–410. [7] L.M. Seiford, J. Zhu, Modeling undesirable factors in efficiency evaluation, European Journal of Operational Research 142 (2002) 16–20. [8] T. Kuosmanen, Weak disposability in nonparametric production analysis with undesirable outputs, American Journal of Agricultural Economics 87 (2005) 1077–1082. [9] V.V. Podinovski, T. Kuosmanen, Modeling weak disposability in data envelopment analysis under relaxed convexity assumptions, European Journal of Operational Research 211 (2011) 577–585. [10] R. Färe, S. Grosskopf, C.A.K. Lovell, C. Pasurka, Multilateral productivity comparisons when some outputs are undesirable: a nonparametric approach, The Review of Economics and Statistics 71 (1989) 90–98. [11] Z. Hua, Y. Bian, L. Liang, Eco-efficiency analysis of paper mills along the Huai River: an extended DEA approach, Omega 35 (2007) 578–587. [12] D. Tyteca, Linear programming models for the measurement of environmental performance of firms—concepts and empirical results, Journal of Productivity Analysis 8 (1997) 183–197. [13] R. Färe, S. Grosskopf, F Hernandez-Sancho, Environmental performance: an index number approach, Resource and Energy Economics 26 (2004) 343–352. [14] R. Färe, S. Grosskopf, C.A. Pasurka Jr., Pollution abatement activities and traditional productivity, Ecological Economics 62 (2007) 673–682. [15] R. Färe, S. Grosskopf, C.A. Pasurka Jr., Toxic releases: an environmental performance index for coal-fired power plants, Energy Economics 32 (2010) 158–165. [16] P. Zhou, B.W. Ang, K.L. Poh, Slacks-based efficiency measures for modeling environmental performance, Ecological Economics 60 (2006) 111–118. [17] P. Zhou, K.L. Poh, B.W. Ang, A non-radial DEA approach to measuring environmental performance, European Journal of Operational Research 178 (2007) 1–9. [18] P. Zhou, B.W. Ang, D.Q. Zhou, Weighting and aggregation in composite indicator construction: a multiplicative optimization approach, Social Indicators Research 96 (2010) 169–181. [19] G.E. Halkos, G. Nickolaos, Tzeremes, Exploring the existence of Kuznets curve in countries’ environmental efficiency using DEA window analysis, Ecological Economics 68 (2009) 2168–2176. [20] Y. Bian, F. Yang, Resource and environmental efficiency analysis of provinces in China: a DEA approach based on Shannon’s entropy, Energy Policy 38 (2010) 1909–1917. [21] S.K. Mandal, S. Madheswaran, Environmental efficiency of the Indian cement industry: an interstate analysis, Energy Policy 38 (2010) 1108–1118. [22] T. Sueyoshi, M. Goto, Methodological comparison between two unified (operational and environmental) efficiency measurements for environmental assessment, European Journal of Operational Research 210 (2011) 684–693. [23] T. Sueyoshi, M. Goto, DEA approach for unified efficiency measurement: assessment of Japanese fossil fuel power generation, Energy Economics 33 (2011) 202–303. [24] K. Wang, Y.M. Wei, X. Zhang, A comparative analysis of China’s regional energy and emission performance: Which is the better way to deal with undesirable outputs? Energy Policy 46 (2012) 574–584. [25] K. Wang, S. Yu, W. Zhang, China’s regional energy and environment efficiency: a DEA window analysis based dynamic evaluation, Mathematical and Computer Modelling 58 (2013) 1117–1127. [26] J. Wu, Q. An, S. Ali, L. Liang, DEA based resource allocation considering environmental factors, Mathematical and Computer Modelling 58 (2013) 1128–1137. [27] J. Zhu, Data envelopment analysis with preference structure, Journal of the Operational Research Society 47 (1996) 136–150. [28] H. Fukuyama, W.L. Weber, A directional slacks-based measure of technical efficiency, Socio–Economic Planning Sciences 43 (2009) 274–287. [29] P. Zhou, B.W. Ang, H. Wang, Energy and CO2 emission performance in electricity generation: a non-radial directional distance function approach, European Journal of Operational Research 221 (2012) 625–635. [30] Y. Chen, A non-radial Malmquist productivity index with an illustrative application to Chinese major industries, International Journal of Production Economics 83 (2003) 27–35. [31] Y.M. Wang, Y.X. Lan, Measuring Malmquist productivity index: a new approach based on double frontiers data envelopment analysis, Mathematical and Computer Modelling 54 (2011) 2760–2771. [32] National Bureau of Statistics of China (NBSC), China Statistical Yearbook, China Statistical Press, 1998–2010. [33] Department of Energy Statistics, National Bureau of Statistics of China (DES-NBSC), China Energy Statistical Yearbook, China Statistical Press, 1998–2010.