March, 2014 Journal of Resources and Ecology Vol.5 No.

J. Resour. Ecol. 2014 5 (1) 011-019

DOI:10.5814/j.issn.1674-764x.2014.01.002
Article
www.jorae.cn

Evaluation of Urban Resource and Environmental Efficiency in

China Based on the DEA Model

ZHANG Xiaoping*, LI Yuanfang and WU Wenjia

College of Resources and Environment, University of Chinese Academy of Sciences, Beijing 100049, China

Abstract: This paper illustrates the spatial variations in urban resource and environmental efficiency (REE)
amongst 285 cities in China using a Data Envelopment Analysis (DEA) model, and examines the factors that
have had the greatest effect on this spatial pattern by regression models. The results gave an average urban
REE of 0.6381, and an average pure technical efficiency (PTE) and scale efficiency (SE) of 0.6964 and 0.9225,
respectively. The results support the existence of a U-shaped relationship between REE and income level, which
means that an increase in urban GDP does not result in an equivalent increase in environmental efficiency.
Economic growth affects REE in three ways: scale effects (population scale and urbanization rate); composition
effects; and spatial effects. Improvements in urban resource use and environmental efficiency depend upon both
technological innovation and effective governance. Policies designed to achieve these improvements should
therefore be implemented at all levels of government and local enterprise.

Key words: resource-environmental efficiency; DEA model; urban economy; China

municipalities, as they attempt to reduce overall resource

1 Introduction consumption and mitigate environment pollution.
China has experienced rapid urbanization since the economic Ecological efficiency in regional economies has been
reforms and opening up process began in 1978. By the end a hot topic in both Chinese and international academic
of 2010, the urbanization rate of China had exceeded 50 circles during the past decade. In the existing literature,
percent and it is expected to increase greatly during the research on resource utilization efficiency is embodied in the
next twenty years. Economic activities in urban areas have evaluation and influencing factors of ecological efficiency
contributed a lot to the Chinese economy. In 2010, cities at (Cui et al. 2014). Ecological efficiency is usually measured
prefectural level and above had produced 24.6 trillion CNY by operational performance indicators that are based on
of GDP (gross domestic production value), a 61.3 percent material and energy consumption balances (Dyckhoff and
share of China’s total GDP. However, the aggressive growth Allen 2001), however a common practice in ecological
of the urban economy has resulted in severe environmental efficiency research is to use a single measure to represent
problems, such as acute resource shortages and aggravated the multiple aspects of such indicators. For example, energy
pollution. Accordingly, China has been under great pressure intensity, which is defined as the ratio between total energy
to construct a resource-conserving and environment-friendly consumption and GDP in a region, is the most widely applied
society. For metropolitan areas, it is thus of great importance index used to measure the efficiency of energy utilization.
to illustrate the significant degree of co-ordination between However, as resource consumption intensity fails to indicate
economic development and ecological protection that is the change in technical efficiency of resource utilization
already occurring. Therefore, measuring and improving in practice, researchers have turned to other indices to
resource use efficiency in parallel with the consideration of better describe the context of “efficiency” in respect of
environmental constraints has become very important for resource consumption (Patterson 1996; Yang and Wang

Received: 2013-11-21 Accepted: 2014-02-25

Foundation: National Natural Science Foundation of China (No.40971075) and the Presidential Foundation of University of Chinese
Academy of Sciences (No.2012).
* Corresponding author: ZHANG Xiaoping. Email: zhangxp@ucas.ac.cn.
12 Journal of Resources and Ecology Vol.5 No.1, 2014

2013). Composite indicators combining economic, social assess the aggregated efficiency of resource utilization and
and environmental dimensions have been widely debated environmental improvement. In the recent literature there
and have already been applied in many empirical studies has been some research which has discussed the energy
to evaluate environmental efficiency and sustainability efficiency or environmental efficiency evaluation problem in
(Krotscheck and Narodoslawsky 1996; Siche et al. relation to the use of the DEA model. Zhou and Ang (2008)
2008). Unfortunately, as most of the empirical analyses presented several DEA-type linear programming models
in this research field have been developed at national and within a joint production framework of both desirable and
provincial levels, research based upon urban statistical undesirable outputs for measuring economy-wide energy
datasets remains relatively weak. efficiency performance, and have applied these models to
In practice, resource and environmental efficiency is the measurement of the energy efficiency performance of 21
usually measured by comparing environmental performance OECD countries. Bian and Yang (2010), Wang et al. (2013),
indicators, and data envelopment analysis (DEA) shows and Jia and Liu (2012) extended the DEA model to establish
a high potential to support such comparisons. Therefore, a comprehensive efficiency measure for resource and
DEA has been widely used in recent literature to evaluate environment efficiency analysis, and used this to evaluate
the relative environmental efficiencies of different regional the relative resource use and environment efficiency of
entities (Wei 2000; Ke and Li 2005). Halkos and Tzeremes provinces in China. Their empirical results showed that the
(2009) measured the environmental efficiency of 17 OECD east area of China has higher energy and environmental
countries and revealed that a Kuznets type relationship efficiency than western China. Li et al. (2005) compared
between environmental efficiency and income did not exist. the efficiency of 202 cities in 1990 and 2000, and found that
Stern (2012) modeled the differences in energy efficiency the spatial pattern of urban efficiency was similar to that
across 85 countries, and concluded that technological of China’s economic distribution pattern. Wu et al. (2011)
change was the most important factor affecting energy- used DEA to assess urban land use efficiency in China, and
use and carbon-emissions under conditions of economic concluded that the input-output efficiency of urban land use
growth. In recent years, there has been a rapid increase in China was relatively high in smaller cities. By analyzing
in the number of studies using DEA in the broad area of 23 high-density urban agglomerations as case studies, Fang
urban development performance evaluation. Yang and Xie and Guan (2011) found that the input and output efficiency
(2002) evaluated the input-output efficiency of China’s 30 of urban agglomerations decreased gradually from the
main cities, and concluded that the development efficiency eastern region to the central and western regions of China.
of cities in western China was much lower than those in However, quantitative studies of the main factors affecting
the eastern region. In an empirical study of 75 resource- urban resource and environmental efficiency have not yet
intensive cities, Gu and Xiao (2009) evaluated the regional been deeply examined.
differentiation of DEA relative efficiency in China, and There are therefore two fundamental research objectives
revealed that tourism-oriented cities had higher potential that this paper seeks to pursue. First, it further explores
for efficiency improvement. Guo et al. (2011) evaluated the the regional variation of environmental efficiency among
comprehensive resource efficiencies of metropolitan areas Chinese cities based on the DEA model. Second, it
in China by using DEA and Malmquist Index models. They examines the main factors affecting the environmental
found that China’s metropolitan population boom and the performance of China’s urban development using regression
rapid spread of built-up areas has resulted in significant models. The results are expected to shed new light on
resource efficiency losses. By employing a DEA model China’s metropolitan development and environmental
and the Malmquist Productivity Index, Sun et al. (2012) management policies.
evaluated the efficiencies of 24 typical resources-based
cities in China, and concluded that small and medium-sized 2 Data and methodology
cities have recorded greater improvements in efficiency 2.1 City sampling
than large ones. A common feature of these studies is that In this paper, we selected 285 Chinese cities for evaluation,
their DEA models only used economic indicators (e.g. including: (i) 1 national capital city, Beijing, and 3
GDP) as output indicators, and did not include indicators of municipalities directly under the control of the central
undesirable environmental impact. government (Tianjin, Shanghai, and Chongqing); (ii) 26
Because any economic production activity is a joint- Provincial Capital Cities (Lhasa City in Tibet Autonomous
production process, it utilizes natural resource inputs Region has not been included in the analysis because of
(e.g. energy, water), and non-natural resource inputs (e.g. a lack of data); and (iii) 255 cities at prefectural level and
capital, labor) to produce desirable output (e.g. products, above. Fig. 1 gives their location and size. These cities are
GDP), along with the emission of by-products which may diverse, show large differences in population size and level
be undesirable for human beings (e.g. waste water, dust, of economic development, and many of them are the key
solid waste, and other pollutants). In this situation it is industrial, commercial and cultural centers of a particular
necessary to develop a multiple-factor model to correctly region. Urban territorial areas also include two spatial parts:
ZHANG Xiaoping, et al.: Evaluation of Urban Resource and Environmental Efficiency in China Based on the DEA Model 13

municipal districts and counties. In this paper, the spatial

range of a sample city refers to its municipal districts only,
county areas are excluded. The data set was generated from
the China City Statistical Yearbook and the China Statistical
Yearbook.
2.2 DEA methodology
2.2.1 The DEA model
Data Envelopment Analysis (DEA) has been widely used
since 1978 to evaluate the relative efficiency of multi-input
and multi-output production units (Charnes et al. 1978).
The main advantage of DEA is that it does not require Population (×104 person)
any prior assumptions about the underlying functional 15-50
50-100
relationships between inputs and outputs. The original idea 100-200
behind DEA was to provide a methodology whereby, within 200-500 0 250 500
a set of comparable decision making units (DUMs), those 500-1500 km
exhibiting best practice could be identified, and would form
an efficiency frontier. Furthermore, the approach allows Fig. 1 Location of sample cities and comparison of their
measurement of the level of efficiency of non-frontier units, population.
and can be used to identify benchmarks against which
inefficient units can be compared (Cook and Seiford 2009).
DEA is applied in this paper to measure the resource- extended this model by providing for variable returns to
environmental efficiency of cities in China. Supposing that scale (VRS). The VRS model differs from that of the CRS
there are K cities to evaluate and each city has N inputs and by way of an additional constraint to Equation (1):
M outputs. Let Xkn, the matrix of input variables, represent
the nth input (n=1, 2, ..., N) in the kth city (k=1, 2, ..., K),
Ykm, the matrix of output variables, represent the mth output
in the kth city, and the Constant Returns to Scale (CRS) By using a VRS model, the REE can be decomposed as
DEA model for the kth cities can be defined as (Equation 1): the pure technical efficiency (PTE) and the scale efficiency
(SE), that is
REE = PTE × SE
According to the DEA model, REE is the overall
measurement of the resource allocation, utilization level
and scale effect of a city; while PTE is the indicator of
(1) efficiency achieved by pure technical improvement in urban
resources utilization, and SE measures the efficiency gained
through metropolitan scale increase. Similarly, if PTE or SE
equals 1, this indicates that the kth city is relatively efficient
in terms of pure technical effect or in scale effect.
2.2.2 Input and output indicators in DEA
where, θ(0<θ≤1) is the comprehensive resource-
environmental efficiency (REE) of the kth city; λ k is a Environmental sustainability is a complex concept, which
k dimensional weight vector; s−(s −≥0) is an input slack includes environmental quality, economic efficiency,
variable vector; s +(s +≥0) is an output residual variable and social equality, and numerous indicators have been
vector; and ε is a non-Archimedean infinitesimal. considered in its evaluation (Wu and He 2006; Yu and Wen
For the kth city, if REE equals 1, that city operates on 2010). On the basis of existing theories, for this study cities
the production frontier and is DEA efficient, because REE were considered to be integrated systems of resources,
reaches 1 only if both slack vectors are equal to zero and economy, and environments. To apply the DEA model, each
none of the input variables of DMU0 are larger than any city is represented as production decision making unit using
linear combination of other DMUs (Cook and Seiford inputs to obtain outputs. To simplify, inputs of DMU are
2009). The closer the value of REE is to 1, the higher the defined as capital, labor, land, energy and water; desirable
relative efficiency is, and vice versa. outputs (also called good outputs) are defined as GDP and
Equation (1) is referred to as the CCR model, and other pleasant urban environmental indicators; undesirable
provides for constant returns to scale (CRS). Banker (1984) outputs (also known as bad outputs) are the various
14 Journal of Resources and Ecology Vol.5 No.1, 2014

Table 1 Explanation of input and output indicators in the DEA model.

Category Indicators Quantitative method
Non natural resource inputs Labor (x1) Total number of employment
Capital (x2) Total investment in fixed assets
Natural resource inputs Land (x3) Urban construction land area
Energy (x4) Total electricity consumption
Water (x5) Total water supply
Desirable outputs Economic indicator (y1) Gross Domestic Production value
Ecological indicator (y2) Forest coverage rate
Undesirable outputs Environmental pollutant (y3) Intensity of industrial wastewater (y31)
Emissions intensity of Industrial sulfur dioxide (y32)
Intensity of industrial solid waste (y33)

environmental pollutants. Table 1 explains the details of existing theories then, this research expects to provide
input and output indicators and their quantitative methods some insight into the metropolitan dimension of resource-
in the DEA model used in this study. environmental efficiency.
To eliminate statistical differences, the indicators in Income level. The relationship between economic growth
Table 1 should be standardized. For inputs and desirable and environmental quality has been an object of research
outputs, the standardizing formula is: for many years. The common recognition is that there is
no simple linear relationship between these two. The well-
ωk= ×100 (k=1, 2, …, 285) known Environmental Kuznets Curve (EKC) hypothesis
postulates an inverted-U-shaped relationship between
For the undesirable outputs, the standardizing formula is: different pollutant density and per capita income. This
hypothesis argues that as income grows, people achieve
ωk= ×100 (k=1, 2, …, 285) a higher standard of living and care more for the quality
of environment where they live. As a result the Kuznets
where, k is the number of cities; ω k represents the Curve has become a vehicle for describing the relationship
standardized indicator value of the kth city, and 0≤ωk≤100; between measured levels of environmental quality and
ρk is the initial indicator value of the kth city; ρmax is the per capita income (Dinda 2004). This paper tests whether
maximum value of the index; and ρmin is the minimum value the level of urban environmental efficiency and per capita
of the index. income follows the same relationship. A quadratic equation
The undesirable outputs are byproducts generated between urban efficiency and income level was constructed
in the processes of resource utilization and economic and its statistical significance was tested.
development, which have negative impacts on the natural Population Scale. The impact of urban population
environment and on human health. In this analysis, indicator scale on environmental efficiency is controversial. Some
y3 is calculated as the weighted average of y31, y32, and y33. researchers argue that a metropolitan area has higher
Accordingly, the 285 cities were analyzed using 8 indicators efficiency than a small city as a result of agglomeration
as five inputs and three outputs. economies and population thresholds. In contrast, most
studies have revealed the lower efficiency of large cities
2.3 The regression model (Guo et al. 2011). In this paper, urban population scale
To further understand the spatial variation of urban is expected to have a negative impact on resource and
ecological efficiency, the efficiency scores obtained by this environmental efficiency. On the one hand, increasing urban
study by using a DEA model were related to the factors population requires more output and thus more natural
influencing urban resource-environmental efficiency using resources are used up in production process. However, more
econometric analysis. The DEA-derived relative efficiency output also implies more waste and undesirable by-products,
of each city is thus determined by factors beyond the input which will have negative impacts on urban environmental
and output indicators. Scale effect, structure effect, and efficiency.
technical effect have been often argued as being main Economic structure. Economic growth can impact
determinants in previous studies (Zhang et al. 2008; Jia environmental quality through a composition effect
and Liu 2012). In this study, four sets of indicators were (Grossman and Krueger 1995). Considering that an
regressed as explained variables to analyze their correlation economy dominated by secondary industries may consume
with urban environmental efficiency. On the basis of more natural resources than those that are services and
ZHANG Xiaoping, et al.: Evaluation of Urban Resource and Environmental Efficiency in China Based on the DEA Model 15

knowledge-based, or have technology-intensive industries, coefficient of each explained variable, respectively; and εk is
it is possible to say that the higher the proportion that the error term.
secondary industry contributes of urban GDP, the lower will
be the environmental efficiency of the city in question. In 3 Results and analysis
consideration of the possible differences in environmental 3.1 Results of DEA efficiency
load within manufacturing sectors, a dummy variable to 3.1.1 Overall features
describe the impact of industrial type on urban resource
use and environmental efficiency was introduced. If a city The comprehensive resource and environmental efficiencies,
is dominated by resource-intensive sectors, it is expected pure technical efficiencies and scale efficiencies of our
to have a higher environmental load, and this will tend to sample cities were calculated by DEAP 2.1. The average
degrade its resource and environmental efficiency. REE of 285 Chinese cities is 0.6381 and the decomposition
Urbanization rate. As the urbanization rate increases, results showed that average PTE and the SE are 0.6964 and
the activities of production and consumption will be more 0.9225, respectively. According to the DEA model, 30 cities
concentrated in urban areas. This spatial agglomeration will in the sample are DEA efficient, because their REE equals
result in the high development densities and mixed land 1. Fig. 2 provides detailed information about these cities. In
uses which are advocated as a sustainable form of urban line with expectation, tourist cities such as Haikou, Sanya
development. To test the influence of spatial concentration in Hainan Province, Heihe in Heilongjiang Province, and
on urban environmental efficiency, the proportion of Maoming in Guandong Province, are in the list of DEA
population living in the built-up area to that of the total efficient cities. Also expected was the finding that because
municipal territory was calculated. It is argued that a city of high pollutant emissions, resource-intensive cities bring
with a higher urbanization rate will improve environmental heavy environmental loads and have relatively lower
efficiency through agglomeration economies and external efficiency scores.
economies of infrastructure spillover. Are large cities more efficient than small cities? From
Based on the above theoretical analyses, the following the perspective of urban population size in relation to
model was constructed to test the relationship between REE, Table 2 presents a detailed description of the results
resource-environmental efficiency and various factors: that relate to efficiency. Cities were divided into five
groups according to their population scale. Cities with
Ek = β0+β1GDPk+β2(GDP)k2+β3POPUk+β4URBANk populations of less than 500 thousand in municipal districts
+β5INDUSk+β6TYPEk+εk have a relatively higher efficiency of 0.7543. A possible
where: k is the number of cities; and explained variable E explanation is that their economic and social activities are
is the efficiency score of each city obtained by the DEA concentrated geographically, making it relatively easier
model. to manage these activities. In contrast, mega-metropolises
Two models were constructed as Model 1 and Model with populations of more than 5 million have an average
2, in which E is defined as REE and PTE, respectively, efficiency of 0.6644. From a comparison of the results, it
so that different factors between these two efficiencies
could be tested. The explanatory variable GDP is the
comprehensive economic development level for a given
city, which is represented by GDP per capita (measured
in ten thousand CNY). The variable POPU represents the
amount of urban population, which is the number of the
residential population (measured in units of ten thousand
persons) in the urban built-up area. The variable URBAN
is the urbanization rate, which is the ratio of the residential
population in the build-up area to that of the municipality
(measured as a percentage). The variable INDU is the
urban industrial structure calculated by the proportion of
industrial output value in GDP (measured in percentage). Efficieney
TYPE is a dummy variable defined to compare the 0.209-0.431
efficiency differences in cities between those are and are not 0.432-0.515
0.516-0.593
natural resource-intensive cities, holding other explanatory 0.594-0.710
variables fixed. If a given city has been listed as a natural 0.711-0.999 0 250 500
resource-intensive city in the policy papers issued by 1.000 km
China’s National Development and Reform Commission
since 2007, TYPE corresponds to 1; otherwise TYPE=0. Fig. 2 Evaluation results of resource and environmental
β 0 is a constant in each model; β 1−β 6 is the regression efficiency of sample cities.
16 Journal of Resources and Ecology Vol.5 No.1, 2014

Table 2 Comparison of urban efficiency variation by subgroups of cities.

Group division Population scale Observations REE PTE SE
City average 285 0.6381 0.6964 0.9225
Group 1 >5million 11 0.6644 0.8717 0.7617
Group 2 2–5 million 31 0.6404 0.7512 0.8619
Group 3 1–2 million 82 0.6059 0.6461 0.9392
Group 4 500–1000 thousand 108 0.6022 0.6443 0.9397
Group 5 <500 thousand 53 0.7543 0.8120 0.9307

Table 3 Contrast of input slacks by subgroups in three regions of China.

Eastern Central Western Total
Observations 114 110 61 285
Labor (104 person) Sum 208.08 111.98 44.22 364.29
Average 1.82 1.02 0.72
Capital (108 CNY) Sum 5348.85 3698.33 2378.29 1142.54
Average 46.92 33.62 38.98
Land (km2) Sum 501.56 463.56 287.12 1252.25
Average 4.39 4.21 4.71
Energy (108 kWh) Sum 409.95 290.96 85.78 786.68
Average 3.59 2.65 1.41
Water (104 t) Sum 60 828.69 47 777.39 66 352.37 174 958.45
Average 533.59 434.34 1087.74

can be seen that mega-metropolises have no advantage in elements will not bring about an improvement of urban
SE, but their advantages in PTE are significant. REE, due to mismatching of input to output. Accordingly,
emphasis should be put on reducing input slacks instead
3.1.2 Spatial variations
of scale expansion. As shown in Fig. 3, there are 169
Fig. 2 reports on the geographical variations of urban REE. cities which are at the stage of decreasing returns to scale,
The overall spatial distribution map shows that cities located approximately 59.3% of the total of evaluated cities. The
in the eastern regions of China have higher efficiency numbers corresponding to cities in the stage of increasing
ratings than those of other regions, which coincides with and constant returns to scale are 71 and 45, accounting for
most previous studies. At the provincial level, cities in
Hebei, Shanxi, Henan, and Anhui provinces displayed
markedly lower resource-environmental efficiency ratings,
but in general cities in western China have relatively higher
efficiency than those in the central regions. This conclusion
is consistent with the result of Wu et al. (2011), but differs
from that of Fang and Guan (2011).
With regard to the DEA efficiency frontier, the summary
of resource input slacks in the input-oriented model describe
those resources that can potentially be saved in the existing
outputs of each city. Table 3 presents the details. A careful
analysis of the resource input patterns of different cities
showed that slacks of urban construction land and fixed
asset investments in the cities of central China has resulted
in a decline in their resource efficiency. Scale returns
By analyzing the input and output efficiency of DMUs
Decreasing
at the DEA frontier, scale returns analysis under DEA can Constant 0 250 500
distinguish between three types: increasing, decreasing lncreasing km
and constant returns to scale. When a city is at the stage of
decreasing returns to scale, this means that increasing input Fig. 3 Spatial variation of scale returns to urban REE.
ZHANG Xiaoping, et al.: Evaluation of Urban Resource and Environmental Efficiency in China Based on the DEA Model 17

24.9% and 15.7% of the total sample, respectively. Table 4 Estimated coefficients of the explanatory variables.

3.2 Regression results Variable Model 1 Model 2

Constant 86.673*** ( 17.788) 91.452*** (17.654)
The designed models were run in SPSS 19.0. Table 4
GDP –3.582*** (–2.610) –2.772* (–1.900)
presents the estimated coefficients of the explanatory
(GDP)2 0.460***(4.251) 0.442*** (3.843)
variables. The results show that the influences of all
POPU –0.024***(–3.368) –0.005 (–0.701)
explanatory variables in Model 1 were statistically
INDUS –0.401*** (–4.058) –0.470*** (–4.475)
significant. While the estimated coefficient of GDP squared
is positive, statistically speaking it exhibited a U-shaped URBAN 0.186***(3.784) 0.166*** (3.177)
relationship between urban resource and environmental TYPE –10.230***(–3.345) –10.052*** (–3.107)
efficiency commensurate with income level. This result R2 0.227 0.249
is similar to the classical empirical results within the Observations 285 285
environmental economics domain. Following the results F 13.851*** 15.371***
obtained, the linear GDP term in the estimated econometric Where *, **,*** represent significance levels of 0.10, 0.05 and 0.01,
equations has a negative coefficient, which means that respectively; figures in brackets are t-values.
increasing environmental efficiency is not associated with
increased GDP.
The estimated results produced by Model 1 show that Model 2 compared with Model 1 indicated another U-shaped
the efficiency level in an urban economy does not only relationship between urban pure technical efficiency and
depend on income level. Apparently, economic composition income level. That is, for both the western cities with
is also very important, because an economy with a larger lower income and the eastern cities with highly developed
industrial production base is likely to produce a higher economies, pure technical efficiencies are relatively high.
environmental burden. This hypothesis is supported by the According to the DEA theory, the REE can be decomposed
negative correlation of the variable INDUS at a significance as PTE and SE, which means that PTE can be understood
level of 0.01. In other words, between two cities with as the contribution made by technical progress to the REE.
same income level, the higher the proportion of secondary Specifically, for the eastern region in China, the high level of
industrial output in GDP, the lower is the urban resource- its economic development is mainly dependent on advanced
environmental efficiency. The higher significance level production technology and developed tertiary industry,
of the variable TYPE in Model 1 shows that a resource- and does not simply rely on resources and consuming the
intensive city has a strong negative effect on urban environment. Thus their higher level of technology leads to
resource-environmental efficiency when other factors are a larger value of PTE. However, for the relatively backward
held constant, which confirms the hypothesis on the role of western region, improvements in production technology are
policy intervention. of great importance and contribute significantly to REE. A
The negative correlation between POPU and urban statistically significant negative impact of INDUS and TYPE
resource-environmental efficiency suggests that larger cities on PTE demonstrates the importance of industrial upgrading
tend to have lower resource-environmental efficiencies on urban efficiency improvement.
compared with cities with smaller populations, when other
factors are held constant. It is no surprise that increasing 4 Conclusion and discussion
urban populations require more output, and more natural This paper has reported on the aggregated efficiency of
resource consumption. Therefore, severe environmental resource and environment in 285 Chinese cities, based on
problems are more often reported as a characteristic of the DEA model. The scores obtained by using the DEA
mega-cities. model were shown through econometric analysis to be
The rate of urbanization has a positive correlation related to a range of factors that influence urban resource
with environmental efficiency, which coincides with our and environmental efficiency. The study also re-examined
expectation. That is, holding other factors constant, a the hypothesis of the Environmental Kuznets Curve in
greater spatial agglomeration of urban activities is expected Chinese urban development using regression analysis. The
to improve relative resource and environmental efficiency results showed a U-shaped relationship between urban
within the municipal area. This result is consistent with the resource and environmental efficiency and income level,
idea of the “compact city” developed in recent years, which which means that an increase in environmental efficiency
recommends intensive urbanization instead of low-density is not associated with increased GDP. However, these two
spreading (Clark 2013). results are not completely contradictory, and can perfectly
The similar patterns revealed by Model 2, which go together if understood properly. Both models show
concerned factors impacting on PTE, show that all how economic growth affects environmental efficiency in
explanatory variables except POPU demonstrate statistically three other channels: scale effects (population scale and
highly significant relationships. The sign of each variable in urbanization rate); composition effects; and spatial effects,
18 Journal of Resources and Ecology Vol.5 No.1, 2014

and both show that the path that growth takes is more Fang C L, Guan X L. 2011. Comprehensive measurement and spatial
important than growth itself. distinction of input-output efficiency of urban agglomerations in China.
Acta Geographica Sinica, 66(8): 1011-1022. (in Chinese)
As each city has different economic level and develop- Grossman G M, A B Krueger. 1995. Economic growth and the environment.
ment policies, their levels of resource and environmental Quarterly Journal of Economics, 110(2): 353-377.
efficiency differ greatly. As a result, to provide higher Gu L X, Xiao J S. 2009. Economic efficiency of resource-based cities in
standards of living without exceeding regional carrying China. Arid Land Geography, 32(4): 624-630. (in Chinese)
capacity, a transition to a new steady and sustainable Guo T Y, He S J, Dong G P. 2011. Metropolitan resources efficiencies,
change trends and causes in China under the goal to build an international
development model is often needed. Despite the fact metropolis. Journal of Geographical Sciences, 21(4): 746-756.
that sustainable forms of development are possible, they Halkos G E, N G Tzeremes. 2009. Exploring the existence of Kuznets curve
will however not automatically be adopted. Economic in countries’ environmental efficiency using DEA window analysis.
incentive policies by governments can be used to manage Ecological Economics, 68(7): 2168-2176.
the transition from current practices to new and more Jia Y P, Liu R Z. 2012. Study of the energy and environmental efficiency of
the Chinese economy based on a DEA model. Procedia Environmental
efficient forms of development. For less developed cities, it Sciences, 13: 2256-2263.
is necessary to adopt effective clean energy policies, to use Ke J, Li C. 2005. Study on the coordinated development of regional
recycled and low-pollution energy, and to encourage low- resources, environment and economy in China based on DEA cluster
emission industries. Additionally, institutional re-orientation analysis. China Soft Science, (2): 144-148. (in Chinese)
toward technological innovation and incentives are of Krotscheck C, M Narodoslawsky. 1996. The sustainable process index, a
new dimension in ecological evaluation. Ecological Engineering, 6(4):
great importance in the improvement of urban resource use 241-258.
patterns and environmental efficiency. Li X, Xu X X, Chen H H. 2005. Temporal and spatial changes of urban
Although the DEA model has been widely used in efficiency in the 1990s. Acta Geographica Sinica, 60(4): 615-625. (in
efficiency evaluation, there are inherent limitations in this Chinese)
form of analysis when it is extended to ecological variables. Patterson M G. 1996. What is energy efficiency? Concepts, indicators and
methodological issues. Energy policy, 24(5): 377-390.
In the DEA model, efficiency is evaluated by comparing Siche J R, F Agostinho, E Ortega, A Romeiro. 2008. Sustainability
a DUM with recommended or preference units. However, of nations by indices: Comparative study between environmental
a city is a complex system of natural, human, social and sustainability index, ecological footprint and the emergy performance
economic elements, and thus each city may have a different indices. Ecological Economics, 66(4): 628-637.
optimal path for development. The evaluation in this paper Stern D I. 2012. Modeling international trends in energy efficiency. Energy
Economics, 34(6): 2200-2208.
has in this sense only provided a relative performance Sun W, Li Y, Wang D, et al. 2012. The efficiencies and their changes of
evaluation of Chinese cities in terms of their resource and China’s resources-based cities employing DEA and Malmquist Index
environmental efficiency. Further research should aim at Models. Journal of Geographical Sciences, 22(3): 509-520.
addressing categorical datasets and constructing models Wang K, Yu S W, Zhang Wei. 2013. China’s regional energy and
with more comprehensive factor bases. environmental efficiency: A DEA window analysis based dynamic
evaluation. Mathematical and Computer Modeling, 58(5-6):1117-1127.
Wei Q L. 2000. Data envelopment analysis (DEA). Chinese Science
References Bulletin, 45(17): 1793-1808. (in Chinese)
Bian Y W, Yang F. 2010. Resource and environment efficiency analysis of Wu D W, Mao H Y, Zhang X L, et al. 2011. Assessment of urban land use
province in China: A DEA approach based on Shannon’s entropy. Energy efficiency in China. Acta Geographica Sinica, 66(8): 1111-1121. (in
Policy, 38(4): 1909-1917. Chinese)
Banker R D. 1984. Estimating most productive scale size using data Wu Y Y, He X J. 2006. The evaluation of Beijing sustainable development
envelopment analysis. European Journal of Operational Research, 17(1): based on DEA model. Systems Engineering – Theory Methodology and
35-44. Applications, 26(3): 117-123. (in Chinese).
Charnes A, W W Cooper, E L Rhodes. 1978. Measuring the efficiency of Yang K Z, Xie X. 2002. DEA efficiency of China city’s input-output.
decision making units. European Journal of Operational Research, 2(6): Geography and Territorial Research, 18(3): 45-47.
429-444. Yang L, Wang K L. 2013. Regional differences of environmental efficiency
Clark T A. 2013. Metropolitan density, energy efficiency and carbon of China’s energy utilization and environmental regulation cost based
emission: multi-attribute trade-offs and their policy implications. Energy on provincial panel data and DEA method. Mathematical and Computer
Policy, 53: 413-428. Modeling, 58(5-6): 1074-1083
Cook W D, L M Seiford. 2009. Data envelopment analysis (DEA)-Thirty Yu Y, Wen Z G. 2010. Evaluating China’s urban environmental sustainability
years on. European Journal of Operational Research, 192(1): 1-17. with Data Envelopment Analysis. Ecological Economics, 69(9): 1748-
Cui Q, Kuang H B, Wu C Y, Li Y. 2014. The changing trend and influencing 1755.
factors of energy efficiency: The case of nine countries. Energy, 64: 1026- Zhang B, Bi J, Fan Z Y, et al. 2008. Eco-efficiency analysis of industrial
1034 system in China: A data envelopment analysis approach. Ecological
Dinda S. 2004. Environmental Kuznets Curve hypothesis: A survey. Economics, 68(1): 306-316.
Ecological Economics, 49(4): 431-455. Zhou P, Ang B W. 2008. Linear programming models for measuring
Dyckhoff H, K Allen. 2001. Measuring ecological efficiency with data economy-wide energy efficiency performance. Energy Policy, 36 (8):
envelopment analysis (DEA). European Journal of Operational 2911-2916.
Research, 132(2): 312-325.
ZHANG Xiaoping, et al.: Evaluation of Urban Resource and Environmental Efficiency in China Based on the DEA Model 19

基于DEA模型的中国城市资源环境效率评价
张晓平，李媛芳，吴文佳

中国科学院大学 资源与环境学院，北京 100049

摘 要：本研究以全国地级以上城市为实证数据，试图刻画我国城市资源－环境效率的空间差异并分析影响这一空间格局
的主要因素。研究中采用数据包络分析法（DEA），对全国285个城市的资源－环境效率进行了评价。结果表明，城市平均资
源－环境效率指数为0.6381，该指数的分解结果表明城市的纯技术效率指数为0.6964，低于规模效率指数0.9225。对城市资源－
环境效率影响因素的回归分析结果表明，城市资源－环境效率与城市收入水平呈U形关系。经济增长可以通过三个渠道影响城
市的资源－环境效率：规模效应（包括人口规模和城市化水平）、经济结构效应以及空间效应。城市资源－环境效率的提高并
不是自然发生的，而需依赖于技术创新和有效管治等手段。因此各级政府、企业和组织必须实施一系列应对措施才能确保城市
的高效发展。

关键词：资源-环境效率；DEA模型；城市经济；中国