Total-factor energy efficiency of regions in China

(1)

Energy Policy 34 (2006) 3206–3217

Total-factor energy efﬁciency of regions in China

Jin-Li Hu

,1

, Shih-Chuan Wang

Institute of Business and Management, National Chiao Tung University, 4F, 114, Sec. 1, Chung-Hsiao W. Rd., Taipei City 100, Taiwan Available online 1 August 2005

Abstract

This paper analyzes energy efficiencies of 29 administrative regions in China for the period 1995–2002 with a newly introduced index. Most existing studies of regional productivity and efficiency neglect energy inputs. We use the data envelopment analysis (DEA) to find the target energy input of each region in China at each particular year. The index of total-factor energy efficiency (TFEE) then divides the target energy input by the actual energy input. In our DEA model, labor, capital stock, energy consumption, and total sown area of farm crops used as a proxy of biomass energy are the four inputs and real GDP is the single output. The conventional energy productivity ratio regarded as a partial-factor energy efficiency index is computed for comparison in contrast to TFEE; our index is found fitting better to the real case. According to the TFEE index rankings, the central area of China has the worst energy efficiency and its total adjustmentof energy consumption amount is over half of China’s total. Regional TFEE in China generally improved during the research period except for the western area. A U-shape relation between the area’s TFEE and per capita income in the areas of China is found, confirming the scenario that energy efficiency eventually improves with economic growth.

Keywords: Data envelopment analysis; Total-factor energy efﬁciency (TFEE); China economy

1. Introduction

China’s economy has grown aggressively in the past two decades and is still expanding at a 9.5% annual rate in the fourth quarter of 2004 (Business Asia, 2005). At the same time, its energy supply is under a severe supply constraint following such roaring development, as demand for electric power in the country increased by 69% from 2000 to 2004. Much of the ﬁxed asset investments have moved into energy-intensive indus-tries, yet there is nearly a one-third shortage of electricity consumption in China.

China is the second largest energy-consuming

econ-omy in the world. One forecast (Kadoshin and

Nishiyama, 2000) shows that by 2010, China will consume three times its 1992 energy input. That is,

China’s share of the world’s total energy consumption will grow from 8.6% in 1992 to 15.9% in 2010. Along with this fast demand for energy, the efficiency of energy use should be of concern especially under China’s energy policy. Various plans are being carried out to increase investment and to speed up construction in order to establish a sufficient energy supply. However, the efficiency of energy consumption needs to be promoted simultaneously such that redundant energy consump-tion is eliminated.

Energy alone cannot produce just any output. Energy must be put together with other inputs in order to produce outputs. Therefore, a multiple-input model should be applied to correctly assess the energy efﬁciency in a region. In our study the regional targets of energy inputs can be found through the data envelopment analysis (DEA). The out-of-date technol-ogy level and the inefﬁcient production process generate a redundant portion of energy consumption which needs to be further adjusted. The amount of total adjustments, including slack and radial adjustments, is computed by www.elsevier.com/locate/enpol

Corresponding author. Tel.: +886 2 23812386, x 57641; fax: +886 2 23494922.

E-mail address: jinlihu@mail.nctu.edu.tw (J.-L. Hu).

1

(2)

DEA. The target level of energy use, called target energy input, is obtained when the amount of total adjustments is reduced from the actual energy use amount. A new index of energy efﬁciency, named total-factor energy efﬁciency (TFEE), is constructed as the ratio of the target energy input that is suggested from DEA to the actual energy inputs in a region.

Energy efﬁciency improvement relies on total-factor productivity improvement (Boyd and Pang, 2000). The TFEE index incorporates energies, labor, and capital stock as multiple inputs so as to produce economic

output (GDP2). In contrast, a traditional energy

efﬁciency index only takes energy into account as a single input (Patterson, 1996) to produce GDP output while neglecting other key inputs such as capital and labor. DEA can be easily applied to a multiple input–output framework to compute the index of TFEE. An empirical analysis of regional energy efﬁciency in China presents results from a real case application of the new index.

Sustainable development has to be considered with an energy policy (Gibbs, 2000) since energy supply is inevitable for economic growth. The concept should be introduced by covering areas on a local or regional level and then extending to the national level (Gibbs, 1998). Our study concentrates specifically on the regional level of energy efficiency in China where further improve-ments can be planned and executed accordingly. China is now in a transition period, having started from a high resource-consuming and low-efficiency economic

devel-opment pattern (Yang, 2002; World Bank, 2001;

Fleisher and Chen, 1997). It is of particular importance to improve energy efﬁciency in the various regions within China in order to sustain its economic growth.

This paper is organized as follows. Section 2 provides administrative data and descriptive statistics at the regional level in China. Section 3 reviews DEA methodology and describes how the index of TFEE is constructed on the basis of DEA. An empirical study for the energy efficiency of regions in China based on TFEE is given in Section 4. Section 4 also compares the TFEE result with the commonly used partial-factor energy efficiency (PFEE) result—the energy productivity ratio (Patterson, 1996). Section 5 concludes our research findings.

2. Regional data and descriptive statistics

From the perspective of China’s development and political factors, its provinces, autonomous regions, and municipalities are usually divided into three major areas: the east, central, and west as shown inFig. 1.

The east area is constituted by 12 regions that stretch from the province of Liaoning to Guangxi, including the coastal provinces of Shandong, Hebei, Jiangsu, Zhe-jiang, Fujian, Guandong, and Hainan, and the three municipalities of Beijing, Tianjin, and Shanghai. The east area is well known and has experienced the most rapid economic growth in China and its GDP output is around half of China’s total. The east area has also attracted the most foreign investment, technology, and managerial know-how. The central area consists of nine regions that are all inland provinces: Heilongjiang, Jilin, Inner Mongolia, Henan, Shanxi, Anhui, Hubei, Hunan, and Jiangxi. This area has a large population and is a home base for farming. Foreign investment in this area is less and technology lags do exist. The west area covers more than half of the territory in China including the provinces of Gansu, Guizhou, Ningxia, Qinghai, Shaanxi, Tibet, Yunnan, Xinjiang, Sichuan, and the municipality Chongqing. Compared to the other two areas, this area has low population density and is the least developed area in China.

Since Chongqing was promoted to be the fourth municipality in China only in 1997, this municipality is still combined with the Sichuan province and they together are regarded as one region in this research. The energy input data of Tibet were not available for this research. Thus, a total of eight regions in the west area are included in this study. The three areas are abbreviated as E, C, and W for the east, central, and west areas, respectively. A total of 29 regions are this study’s targets to be analyzed for their energy efﬁciency.

Fig. 1. The administrative regions and three major areas in China.

2

(3)

2.1. Data sources

A panel dataset of 29 regions from 19953 to 2002 is collected for analysis. Data of labor employment and GDP are both collected from the China Statistical Yearbook. Data of capital stock are not available in any statistical yearbooks of China. In this study every regional capital stock in a speciﬁc year is calculated by the authors according to the formula (Li, 2003): Capital stock in the current year

¼capital stock ðprevious yearÞ þcapital formation ðcurrent yearÞ

capital depreciation ðcurrent yearÞ. ð1Þ

The 1995 regional capital stocks in 1978 prices are

obtained from Li (2003). All monetary inputs and

outputs such as the GDP and capital stock are transformed into 1995 prices with GDP deﬂators.

Regional energy consumption levels are collected from the China Energy Statistical Yearbook. These energy datasets contain only the conventional energy consumption—mainly coal, petroleum, and natural gas. Those energy resources with a low caloriﬁc value are excluded, most of which are renewable energies. Biomass energy is almost 100% of this renewable energy source in China (Chang et al., 2003). There are also other renewable sources such as solar, wind, geothermal, and oceanic energies, but all are consumed at very low

levels (International Energy Agency (IEA), 2004).

Biomass energy is one of the main sources for non-commercial energy use in China’s rural areas, constitut-ing 38.7% of China’s total energy consumption in 1970, but then later dropping to 19.9% of China’s total in 2000 (Chang et al., 2003). Few data sources of biomass energy consumption could be found, and those had been offered were incomplete for this study and relied on estimations (Sinton and Fridley, 2002).

Biomass energy still counts for around 20% of the total energy consumption in China as of 2000. China Rural Energy Statistical Yearbook once provided regional data of biomass energy consumption, but only up to 1999. Since biomass energy is mainly composed of straw, stalk, crop residue, and fuel wood (Chang et al., 2003), the total sown area of farm crops of the regions is selected as a proxy of the biomass energy consumption in this study.4 The data of regional total sown area of farm crops (abbreviated as ‘farm area’ hereafter) are available in the China Statistical Yearbook for each sample year. A very high positive correlation coefﬁcient of 0.841 is found between total biomass energy

consumption level and farm area of regions in China during our 1995–1996 data collection. The factor farm area of regions is hence an appropriate proxy of the biomass portion of energy input in this study. The data are therefore complete for an analysis of energy efﬁciency in regions of China.

2.2. Descriptive statistics of the regions in China The descriptive statistics of China’s regions for their GDP performances in terms of production output, labor employment, capital stock, and energy consumption, including farm area as production inputs in our energy efficiency analysis, are first given inTable 1. As the data in this table show, the east area, frequently called the coastal area, includes regions that show the fastest development progress in China. Mean GDP output in this area is 3.77 trillion RMB,5 which is much higher than 1.75 trillion RMB of the central area and 0.88 trillion RMB of the west area during the sample years. The standard deviation of GDP output shows the same tendency and matches the economic growth pattern among these areas. The east area has a standard deviation of 210.4 billion RMB, which is also much higher than the other two areas, which are 72.7 billion RMB in the central area and 31.3 billion RMB in the west area. The fast-developing east area receives the largest investment—three times that of the central area and five times the west area.

The situation of energy consumption appears to be the same as that of the GDP result. The east area, as its economic growth shows, consumes the largest portion of energy amount in China. As shown inTable 1, the east area consumed 69.5 billion metric tons of standard coal equivalent (Btce) on an average from 1995 to 2002. This is much higher than 28.1 Btce consumed by the west area and 46.2 Btce consumed by the central area. While analyzing deviations of energy consumption among these areas, we ﬁnd that the east area increased its energy consumption the fastest from 1995 to 2002 versus the central and west areas. As aforementioned, the central area is the home base of farming, and it consumes the largest level of biomass energy. The east area consumes the second biggest amount and the west area has the lowest level in accordance with farm area data shown inTable 1.

A correlation matrix is shown in Table 2, whereby a high correlation exists between these inputs and output. The correlation coefﬁcient between the energy input and GDP output is 0.801 (Po0.005). A positive correlation coefﬁcient of 0.539 is found between the farm area input and GDP output as well. These results all show 3

Complete panel data of these variables started from 1995.

4

We have tried to use the forest area as a proxy of fuel wood as one of the biomass energy sources. However, the sign of the correlation coefﬁcient between real GDP and forest area is negative, hence violating the ‘isotonicity’ requirement between an input and an output.

5

The RMB is an abbreviation of Ren-Min-Bi, meaning ‘people’s currency’ in Chinese. The RMB is the ofﬁcial currency of the People’s Republic of China.

(4)

‘isotonicity’ of the four inputs and the one output in our DEA model. The energy input efﬁciency shall be analyzed in this study in order to understand individual energy efﬁciency states among all regions of China.

3. Total-factor energy efﬁciency (TFEE)

We employ an economic production function that is constructed by DEA in order to analyze regional energy efficiencies in China based on the viewpoint of total-factor productivity. Energy, including farm area as a proxy, is considered in conjunction with conventional inputs labor and capital stock, which are normally used in an economic productivity analysis as the total inputs to produce economic output (GDP). For an economy or a region, it is preferable that the local GDP increases and that energy consumption is saved in order to reach production efficiency. Therefore, the goal for GDP growth and efficiency of energy consumption should be put together in order to sustain economic development.

Table 1

Summary statistics of input and output factors by region (1995–2002)

ID Region Input Factors Output factors

Labor employment (10,000 persons) Capital stock (100 million RMB) Energy consumption (Mtce)

Total sown area of farm crops (1000 ha)

Gross domestic product (100 million RMB)

Mean STDev Mean STDev Mean STDev Mean STDev Mean STDev

1 Beijing E 661.25 59.18 19,431.65 3662.31 3992.63 333.17 484.43 80.45 1602.58 153.59 2 Tianjin E 442.24 40.20 11,107.28 2020.18 2657.50 222.80 557.13 20.87 1061.09 82.69 3 Hebei E 3397.53 21.04 13,978.40 4551.14 9670.38 926.31 8944.05 123.98 3295.88 215.68 4 Liaoning E 1907.96 121.90 56,457.27 7,175.74 9924.25 650.52 3693.11 126.81 3008.53 132.15 5 Shanghai E 721.56 48.05 33,191.94 6886.27 5189.75 576.34 529.55 30.49 2873.36 228.68 6 Jiangsu E 3642.55 104.45 40,401.86 8944.46 8441.63 565.02 7923.98 98.49 5655.75 294.58 7 Zhejiang E 2717.79 59.95 23,761.19 6167.38 5633.00 942.97 3689.34 358.03 3965.71 291.14 8 Fujian E 1635.88 44.58 10,940.81 3611.55 2776.88 402.65 2835.19 104.47 2525.74 177.48 9 Shandong E 4682.25 34.71 30,967.33 8605.29 9297.88 856.50 11,079.03 144.98 5606.69 321.22 10 Guangdong E 3806.29 112.26 28,601.31 7684.30 8892.13 1358.10 5292.34 203.30 6254.00 356.32 11 Guangxi E 2483.00 61.21 6172.15 1395.27 2580.25 197.51 6173.84 198.20 1510.36 136.10 12 Hainan E 333.64 7.43 2976.39 520.71 429.50 87.00 898.21 28.73 347.94 7.99 13 Shanxi C 1442.70 28.95 8887.96 1631.26 7426.63 1030.32 3911.38 121.28 1,155.07 64.82 14 Inner Mongolia C 911.20 327.73 6564.68 1320.36 3481.75 651.83 5727.68 357.67 925.19 48.86 15 Jilin C 1152.95 86.39 7810.24 1580.51 3991.50 285.72 4304.51 345.87 1221.38 48.40 16 Heilongjiang C 1637.79 50.48 11,588.03 2242.60 6059.88 175.20 9274.30 457.75 2178.37 82.54 17 Anhui C 3323.05 64.27 11,235.34 2649.00 4,710.75 373.81 8635.93 256.74 2,102.41 80.19 18 Jiangxi C 1996.03 63.20 8456.92 1847.88 2248.63 183.14 5777.39 244.08 1380.92 65.91 19 Hennan C 5172.75 333.13 18,718.26 4528.35 7398.13 783.89 12,690.29 465.98 3377.12 170.38 20 Hubei C 2,594.09 111.68 15,382.57 3992.74 6,103.75 300.12 7580.59 155.35 2790.39 174.26 21 Hunan C 3504.71 50.32 9905.99 2598.71 4803.25 532.53 7933.06 83.34 2437.29 104.82 22 Sichuan W 6180.41 131.67 14,503.10 4351.39 9852.44 450.88 13,118.30 168.69 3814.44 132.03 23 Guizhou W 1976.33 80.42 4169.23 989.50 4049.88 441.30 4517.35 174.14 659.48 21.96 24 Yunnan W 2270.11 49.94 6833.66 1676.02 3265.75 410.90 5441.01 365.69 1322.75 61.22 25 Shaanxi W 1806.74 29.23 10,080.84 1784.45 3145.13 357.55 4535.93 199.66 1096.54 55.13 26 Gansu W 1190.28 26.59 5588.69 1047.75 2832.63 157.32 3743.80 50.83 651.46 40.80 27 Qinghai W 237.13 5.80 1797.95 333.37 824.88 131.36 552.01 27.09 175.98 10.89 28 Ningxia W 265.48 12.47 1963.38 342.72 838.75 50.58 1014.99 58.95 178.71 6.38 29 Xinjiang W 680.58 11.12 6831.47 1406.17 3276.63 232.13 3281.93 159.15 877.31 40.75 Sum 62,774.23 446.37 428,305.91 94,995.22 143,796.06 10,986.60 134,754.26 54,336.87 64,052.43 3,002.45 East 26,431.93 3884.69 277,987.58 60,805.13 69,485.75 6670.12 52,100.18 806.64 37,707.63 2,104.51 Central 21,735.26 1560.11 98,550.00 22,369.30 46,224.25 2861.73 65,835.11 1,417.98 17,568.14 727.12 West 14,607.04 2083.07 51,768.32 11,912.26 28,086.06 1747.12 36,205.31 735.26 8,776.66 312.93

Notes: (1) All monetary values are 1995 prices. (2) Source: China Energy Statistical Yearbook, 1991–1996, 1997–1999, 2000–2002, China Statistical Yearbook, 1995–2002.(3) E: east area, C: central area, and W: west area. (4) Data for the administration region Chongqing is regarded as a part of Sichuan in this paper since it was promoted as one of the municipalities in China only from 1997. (5) Data of energy consumption in Tibet is not included since it was not available in the sample period.

Table 2

Correlation matrix for inputs and output (1995–2002)

GDP Labor Capital Energy Farm area

GDP 1.000

Labor 0.739 1.000

Capital 0.716 0.324 1.000

Energy 0.801 0.718 0.701 1.000

Farm area 0.539 0.887 0.148 0.628 1.000 Notes: (1) Farm area: total sown area of farm crops as explained in text.

(5)

An efficiency frontier is established by DEA com-posed of those regions with the best production efficiency with energy input considered. The energy efficiency is measured in each region for how far apart they are from this efficiency frontier. The methodology of DEA is first reviewed in Section 3.1. How the total adjustments are computed as gaps between efficiency of each region and that on the frontier is explained in detail in Section 3.2. The TFEE index is therefore constructed accordingly in Sections 3.3 and 3.4. We use the software Deap 2.1, kindly provided byCoelli (1996), to solve the linear programming problems.

3.1. Methodology of DEA

DEA is known as a mathematical procedure using a linear programming technique to assess the efficiencies of decision-making units (DMU) that refer to a set of firms (Coelli, 1996) and a set of regions in this study. All DMUs take an identical variety of inputs to produce an identical variety of outputs (Ramanathan, 1999); but through distinct production processes and technologies decided and used in each DMU, the input and output levels and their production efficiency are eventually decided upon. A non-parametric piecewise frontier composed of DMUs, which owns the optimal efficiency over the datasets, is constructed by DEA for compara-tive efficiency measurement. Those DMUs which are located at the efficiency frontier have their maximum outputs generated among all DMUs by taking the minimum level of inputs, are efficient DMUs, and own the best efficiency among all DMUs.

DEA does not need to specify either the production functional form or weights on different inputs and outputs. It produces detailed information on the efficiency of the unit, not only relative to the efficiency frontier, but also to specific efficient units which can be

identiﬁed as role models or comparators (Hawdon,

2003). Comprehensive reviews of the development of

efﬁciency measurement can be found in Lovell and

Schmidt (1993). There are K inputs and M outputs for each of these N DMUs. The envelopment of the ith DMU can be derived from the following linear programming problem: Miny;ly such that y_iþY lX0; yxiX lX0; lX0; (2)

where y is a scalar and l is an N 1 vector of constants, and whereby the value of y obtained is the efficiency score for the ith DMU. This satisfies yp1 with a value of 1, indicating a point on the frontier and hence a technically efficient DMU (Coelli et al., 1998). The above procedure constructs a piecewise linear

approx-imation to the frontier by minimizing the quantities of the K inputs required to meet the output levels of the ith DMU. The weight l serves to form a convex combina-tion of observed inputs and outputs. The efﬁciency score y measures the maximal radial expansion of the outputs given the level of inputs. It is an input-orientated measurement of efﬁciency.

Eq. (2) is known as the constant returns to scale (CRS) DEA model (Charnes et al., 1978). This model finds the overall technical efficiency (OTE) of each DMU. The variable returns to scale (VRS) DEA model (Banker et al., 1984) further decomposes the OTE into pure technical efficiency (PTE) and scale efficiency (SE). That is, OTE ¼ PTE SE. In order to pursue OTE with energy inputs, our study adopts the CRS DEA model. Both output-oriented and input-oriented CRS DEA models generate exactly the same efficiency scores, target inputs, and target outputs. However, the results of a VRS DEA model can be drastically changed by shifting the DEA model from output orientation to input orientation.

3.2. Slack and radial adjustments of energy input DEA identifies the most efficient point on the frontier as a target for those inefficient DMUs to achieve through a sequence of linear programming computation (Coelli, 1996). For the ith DMU, the distance from an inefficient point where it is located to the projected point on the frontier by radial adjusting the level of inputs, ð1 yÞxi, is called ‘radial adjustment’. Moreover, the

mostly seen piecewise linear form of the non-parametric frontier causes the second stage to shift from the projected point to a point at the practical minimum level of the inputs on the frontier. The distance of shifting along with the frontier in between is called ‘slack’.

How a point with a practical minimum level for inputs on the frontier can be identiﬁed in DEA is illustrated inFig. 2. The maximum level Y output by the DMUs located on the frontier is normalized to unity and generated from the energy input and other inputs which are also normalized by dividing Y. Point B is the actual input set and point B0 _{is the projected point on}

the frontier for DMU B as the target in order to improve its efﬁciency accordingly by reducing the radial adjust-ment BB0_{. However, as mentioned, the practical frontier}

is a piecewise linear format that requires the second-stage adjustment to determine a practical minimum point for inputs. InFig. 2, point A0_{is the projected point}

on the frontier for another DMU A as the target to reach by reducing the radial adjustment AA0_{. However,}

the input level at point A0 _{could be further reduced to}

input level at point C while maintaining the same output level. The amount CA0_{that shall further be adjusted for}

(6)

the input level at point A0 _{along with the frontier is}

called ‘slack’.

The summation amount of slack (CA0_{) and radial}

adjustment (AA0_{) for inputs is called the amount of total}

adjustments (CA), meaning that it is the total amount for inputs which should be adjusted by a DMU so as to reach its optimal production efﬁciency. The adjustments require both a promotion of technology level and an improvement of production process so that OTE is optimized. The amount of total adjustments therefore decreases and the output level is maximized so that the DMU operates at the frontier position of production efﬁciency. The practice minimum input level is called the target input level for a DMU.

When incorporating energy as an input into an economy’s production in a region, the target level of energy input is named ‘target energy input’, which represents a practical minimum level of energy input to be taken as a target in a region in order to perform at the optimal efficiency of energy consumption. The level of target energy input is identified through DEA in conjunction with other inputs so as to produce economic output. An energy-efficient region, therefore, explains that it has operated at the maximum economic output by taking the minimum level of energy consumption based on the viewpoint of total-factor productivity and with OTE optimized. The amount of total adjustments for energy consumption is obtained from the gap between target and actual energy inputs and shall be reduced from its actual energy input level in the region. 3.3. Regional total-factor energy efficiency

The amount of total adjustments in energy input is regarded as the inefﬁcient portion of actual energy consumption in a region. The more the amount of total

adjustments, the less efficient the energy consumed in the region. Thus, if there does not exist an amount of total adjustments of energy input (equal to zero), then the region is utilizing energy at the ‘target energy input’ level, which is the optimal efficiency of energy con-sumption when its output is maximized. Therefore, energy efficiency in a region is defined in Eq. (3) as below, which is named total-factor energy efficiency (TFEE) for region i at time t since the index is established based on the viewpoint of total factor productivity:

TFEE ði; tÞ ¼Target Energy Input ði; tÞ

Actual Energy Input ði; tÞ, (3)

which implies in the ith region and in the tth year. As Eq. (3) shows, the index TFEE represents the efﬁciency level of energy consumption in a region. As the target energy input is the best practical minimum level of energy input in a region, the actual energy input is therefore always larger than or equal to this target energy input. This makes the index TFEE score to always be between zero and unity. When the actual energy input level of a DMU is equal to the suggested target energy input level, a TFEE score of unity is achieved. Conversely, if the actual energy input level is far away from the suggested target energy input level, then the index approaches zero, which represents low efﬁciency. This index is shown in percentile format in this study for easier reading.

3.4. Total-factor energy efficiency in an area

Index TFEE is also employed to analyze energy efﬁciency in an area. Assume area a covers r regions. Area a’s TFEE at time t is deﬁned in Eq. (4):

TFEE ða; tÞ ¼Sr2a Target Energy Input ðr; tÞ Sr2a Actual Energy Input ðr; tÞ

, (4)

which implies that the rth region belongs to area a. Eq. (4) shows that the TFEE in an area is calculated by dividing the summation of target energy inputs by the total actual energy inputs of the area. This is a summation of actual energy input consumed in regions of the area. We calculate TFEE of an area in China in Section 4.3 in order to ﬁnd its relation with per capita income of regions in China.

4. Empirical study of TFEE in regions of China 4.1. Regional total adjustments of energy input

Regions in the east area—the fast-developing area— have an amount of total adjustments of their energy input at 15.96 Btce in 1995 and then it decreased to 13.49 Btce in 2002. The amount remained at around

Output = 1 B C A D A’ B’ Energy Y Other Inputs Y

Fig. 2. Radial adjustment and slack identiﬁed in an input-oriented CRS DEA.

(7)

one-third of China’s total and fell slowly during the research period. Liaoning (4) and Hebei (3) are the two major regions generating around two-thirds of the area’s total. Shandong (9) reduced its 1995 adjustments’ amount to half in 2002 with continuous improvement. Shanghai (5), Fujian (8), and Guangdong (10) are the three regions in the area having no energy adjustments’ amount in all years of the research period. Guangxi (11) and Hainan (12) have zero adjustments in 1995, but the amount rose in the later years. All detailed data are shown inTable 3.

Regions in the central area own the largest portion of the total adjustments’ amount of energy use in China, which were 22.9 Btce (47.69% of China’s total) in 1995 and 16.7 Btce (41.83% of China’s total) in 2002. The adjustments’ amount actually dropped rapidly during the research period, but was still above 40% of China’s total. All regions in this area had non-zero adjustments among all the years of the research period except for Inner Mongolia (14) and Hunan (21) in the later years. Shanxi (13) created the largest energy adjustments’ amount in this area, which was around one-third of the area’s total in 1995 and then it rose up to half of the area’s total in 2002. Heilongjiang (16), Hubei (20), Jilin (15), and Hennan (19) are the four regions that generated high amounts of adjustments in the area. Their amounts totaled one-third of the area’s adjust-ments’ amount in 1995 and were half of the area’s total in 2002. The detailed data are shown inTable 3.

The west area contains the lowest level of total energy

adjustments. Its adjustments’ amount began at

9.15 Btce, increased up to 10.51 Btce in 1998, and then decreased to 9.72 Btce in 2002, remaining relatively ﬂat as shown inTable 3. Sichuan (22) is the only region to have no adjustments among all the years of the research period. Gansu (26), Shannxi (25), and Xinjiang (29) are three regions that had higher levels of energy adjust-ments’ amount at about two-thirds of the area’s total. Guizhou (23) is the only region having its adjustments’ amount increased during the period.

The total adjustments’ amount of the energy input signiﬁcantly decreased during 1995–2002 for all areas in China, and the results are shown inFig. 3. It shows a total of 48.01 Btce in 1995 and then it dropped to 39.89 Btce in 2002. The major reduction comes from regions in the east area as well as the central area, though regions in the central area still constitute over 40% of China’s total energy adjustments. Regions in the west area contain the least portion of China’s total, but it is worth noting that their adjustments’ amount slightly increased.

4.2. Regional TFEE

Table 4 shows the regional result during 1995–2002 based on the TFEE index. The TFEE score is computed

for a comparison of energy efﬁciency among regions. The ratio of the actual energy input to the target energy input is analyzed instead of using an absolute amount for comparison. A total of four regions in China are found to always have optimal efﬁciency during the research period. Three of these regions are located in the east area: Shanghai (5), Fujian (8), and Guangdong (10). The one region left is Sichuan (22), located in the west area.

These four regions, having the optimal efficiency result in our analysis, actually constitute the efficiency frontier of energy consumption among all regions of China. They have the best technology level and production processes so that their inputs and output are operating at the optimal level. The other regions in China which are not yet at the frontier efficiency position can, therefore, base themselves on these frontier regions as targets to adjust their technology levels and production processes accordingly. A target is hence identified by TFEE analysis based on DEA so that feasible steps for improvement of the OTE do exist. This is one of the benefits from TFEE analysis in a comparative study of energy efficiency.

As shown inTable 4, regions in the east area have the highest TFEE rank over all areas in China. In contrast, regions in the central area have the worst rank of efficiency. Regions in the west area have the second best rank for TFEE. The TFEE scores are in average 77.6%, 55.7%, and 64.4% over the research period for the east, central, and west areas, respectively. The central area was predicted to be the second best energy-efficient area before performing this empirical study. It does have the second largest amount of investments and labor employ-ment, it consumes the second largest level of energy, and it contributes the second largest GDP output in China. However, its efficiency level does not perform the same as with the rank of inputs and output, while it does have the worst TFEE.

4.3. Energy efficiency vs. regional development

A U-shape relation is found to exist between the area’s TFEE and per capita income in China. The east area has the highest level of per capita income and also the highest score of an area’s TFEE among the three areas. The central area has the second highest level of per capita income, but the worst score of an area’s TFEE. The west area has the lowest level of per capita income, but is the second best in TFEE. The per capita income actually represents the economic development level in the region or the area. The U-shape relation is illustrated inFig. 4.

The discovered U-shape relation between TFEE and per capita income in an area, therefore, explains that an improvement of energy efﬁciency is followed by economic growth in an area, though it declines in the

(8)

beginning period. This discovery matches the real condition of regional development in China. As shown inTable 4, most regions and areas in China have their

improved TFEE followed up with economic growth in the period from 1995 to 2002 since China showed dramatic economic growth during this period. The east and central areas both improved during the sample’s period. Especially, the central area had significant improvement in the period, although its average level is the lowest rank among all areas. The west area just started its development in recent years and therefore its TFEE score degraded in the beginning of the research period, and kept flat in the later years. However, there are no significant changes on the TFEE of regions in the west area.

4.4. Comparison of total-factor and partial-factor energy efficiencies

The commonly used indicator of energy efﬁciency, the index of energy productivity ratio (Patterson, 1996), is also computed for comparison with TFEE. In contrast

Table 3

Total adjustments amount of energy use by region (1995–2002)

ID Region 1995 1996 1997 1998 1999 2000 2001 2002 1 E Beijing 1094.53 1104.27 1352.68 1336.31 1263.44 1286.48 1015.54 893.98 2 E Tianjin 1065.11 862.70 849.51 901.04 804.54 930.04 874.57 726.68 3 E Hebei 4163.53 3939.70 3970.14 3918.68 3875.62 5234.86 3744.19 4330.84 4 E Liaoning 4861.31 4835.12 4804.35 4429.55 4353.00 5395.54 5025.00 4674.92 5 E Shanghai 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 6 E Jiangsu 1380.76 1336.77 1407.30 1628.80 1275.31 1213.20 887.79 643.60 7 E Zhejiang 300.77 267.72 407.36 523.66 500.49 453.79 244.08 0.02 8 E Fujian 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 9 E Shandong 3097.46 3192.84 3386.02 3383.65 2873.38 1733.27 1874.18 1963.33 10 E Guangdong 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 11 E Guangxi 0.00 0.00 0.00 144.91 148.98 254.73 165.16 142.77 12 E Hainan 0.00 0.00 43.57 67.09 63.07 90.75 114.03 109.68 13 C Shanxi 7101.27 5530.22 5718.63 5353.53 5221.58 5441.85 6111.53 7198.79 14 C Inner Mongolia 1638.16 1794.48 2396.75 2123.93 2764.61 2453.90 2487.75 0.00 15 C Jilin 2777.67 2779.12 3032.67 2526.71 2278.99 2164.73 1893.43 2115.88 16 C Heilongjiang 3306.50 2990.52 3605.66 3768.90 3794.19 3720.32 2522.89 2130.47 17 C Anhui 1088.29 1229.93 1128.05 1188.91 1137.28 1766.90 1222.61 1208.48 18 C Jiangxi 981.60 729.64 682.28 593.77 588.93 670.12 643.48 638.50 19 C Hennan 2257.33 2037.60 1887.76 2131.44 2018.51 2637.65 1828.89 1792.59 20 C Hubei 2994.68 2995.86 2893.44 2829.05 2632.96 2935.00 1323.62 1599.69 21 C Hunan 752.89 589.36 232.20 113.95 0.00 0.00 0.00 0.00 22 W Sichuan 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 23 W Guizhou 1610.57 2110.22 2384.96 2629.90 2251.12 2560.81 2710.67 2615.53 24 W Yunnan 757.67 667.57 910.47 761.95 700.91 1087.97 815.62 928.05 25 W Shaanxi 1994.61 2400.16 1919.72 1801.29 1427.53 1382.74 1554.33 1721.28 26 W Gansu 2093.02 2018.59 1775.31 1806.03 1914.44 2144.81 1631.17 1583.77 27 W Qinghai 469.33 476.00 485.12 560.22 715.36 645.59 638.22 680.50 28 W Ningxia 533.22 565.65 573.56 643.56 636.07 669.05 592.73 577.14 29 W Xinjiang 1690.38 2072.03 2036.59 2304.41 1965.99 1937.36 2029.38 1610.38 Summary 48,010.66 46,526.06 47,884.10 47,471.22 45,206.28 48,811.47 41,950.82 39,886.84 East 15963.47 15,539.13 16,220.93 16,333.69 15,157.81 16,592.66 13,944.51 13,485.80 Central 22898.39 20,676.71 21,577.44 20,630.18 20,437.05 21,790.47 18,034.20 16,684.41 West 9148.80 10,310.22 10,085.73 10,507.35 96,11.43 10,428.33 9972.11 97,16.64 Notes: (1) The unit is 10,000 tce. (2) E: east area, C: central area, and W: west area. (3) Data for the administration region Chongqing is regarded as a part of Sichuan in this paper since it was promoted as one of the municipalities in China only from 1997. (4) Data of energy consumption in Tibet is not included since it was not available in the sample period.

5000 10000 15000 20000 25000 30000 35000 40000 45000 50000 55000 1995 1996 1997 1998 1999 2000 2001 2002 Year

Total adjustments of energy (Mtce)

.

Total Central East West

Fig. 3. Slack and radial adjustment of energy by area and year (1995–2002).

(9)

to the concept of TFEE, the commonly used index of energy productivity ratio can be regarded as a partial-factor energy efﬁciency (PFEE) index since energy is the

only input factor considered in the index. The index PFEE computes the efficiency ratio by directly dividing GDP output by energy input as an indicator of energy efficiency. In contrast, the index TFEE incorporates the key inputs, labor and capital stock, together with the energy input, including the farm area proxy, to form a multiple-inputs model to assess energy efficiency on the basis of a total-factors effect.

Table 5shows the PFEE score of the three main areas and that of China’s total. The east area drops the most, over 10%, from 62.46% in 1995 to 51.45% in 2002. The west area drops the least at just 3% from 28.68% in 1995 to 25.51% in 2002. The central area’s efﬁciency increased in period from 1995 to 1998, but then dropped in the period from 1999 to 2002: the lowest ratio is 36.44% in 2002. The PFEE score of China’s total dropped from 45.31% in 1995 to 39.64% in 2002. As can be seen from these results, energy efﬁciency among

Table 4

Total-factor energy efﬁciency by region (1995–2002)

ID Region 1995 1996 1997 1998 1999 2000 2001 2002 1 E Beijing 68.9 69.9 64.7 65.8 68.3 69.5 76.5 80.1 2 E Tianjin 58.5 65.5 65.5 63.1 68.5 66.7 70.0 76.0 3 E Hebei 53.7 55.9 56.0 57.2 58.7 47.1 64.0 62.6 4 E Liaoning 49.7 50.3 49.3 51.4 53.6 49.9 52.8 55.9 5 E Shanghai 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 6 E Jiangsu 82.8 83.5 82.4 79.9 84.4 85.9 90.0 93.3 7 E Zhejiang 93.4 94.5 92.0 90.0 90.8 92.4 96.3 100.0 8 E Fujian 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 9 E Shandong 64.7 65.2 63.0 62.5 68.3 78.9 81.2 82.2 10 E Guangdong 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 11 E Guangxi 100.0 100.0 100.0 94.1 94.0 90.5 93.8 95.2 12 E Hainan 100.0 100.0 88.8 83.5 85.4 81.1 78.1 80.4 13 C Shanxi 15.6 19.2 18.1 19.2 19.7 19.2 23.3 22.9 14 C Inner Mongolia 37.8 36.4 29.0 30.4 27.3 30.7 38.9 100.0 15 C Jilin 32.4 33.4 30.0 32.6 38.3 40.8 51.0 51.4 16 C Heilongjiang 44.3 49.0 44.0 36.9 37.4 39.7 58.2 64.5 17 C Anhui 74.1 72.8 74.4 74.0 75.7 63.8 76.1 77.3 18 C Jiangxi 59.0 66.1 68.0 70.7 72.4 69.8 72.4 75.4 19 C Hennan 65.1 69.4 71.9 70.6 72.6 66.5 77.8 79.2 20 C Hubei 47.0 50.1 52.6 53.2 56.0 53.2 78.1 76.2 21 C Hunan 86.1 89.2 95.2 97.7 100.0 100.0 100.0 100.0 22 W Sichuan 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 23 W Guizhou 49.4 42.8 39.8 39.1 44.0 40.8 38.9 41.5 24 W Yunnan 71.3 75.9 73.4 77.3 78.7 66.1 76.6 76.4 25 W Shaanxi 36.4 31.9 38.3 40.4 46.5 49.4 52.3 53.6 26 W Gansu 23.6 28.0 31.2 32.8 34.4 28.8 43.8 47.5 27 W Qinghai 31.8 31.8 31.4 24.2 23.8 26.6 31.4 33.2 28 W Ningxia 29.7 29.4 28.7 22.2 25.0 23.0 33.4 36.6 29 W Xinjiang 40.3 35.7 36.9 29.7 38.8 41.6 42.0 55.5 Total 64.1 65.8 65.2 65.6 67.7 66.4 72.7 76.0 East 74.6 76.0 75.1 75.1 77.6 76.8 81.7 83.6 Central 49.4 53.5 52.4 53.3 53.9 52.0 62.7 68.2 West 64.1 61.7 62.8 62.8 65.6 63.1 65.9 69.1

Notes: (1) The unit is percentage. (2) E: east area, C: central area, and W: west area. (3) Data for the administration region Chongqing is regarded as a part of Sichuan in this paper since it was promoted as one of the municipalities in China only from 1997. (4) Data of energy consumption in Tibet is not included since it was not available in the sample period. (5) Scores with a gray shadow covered are those regions reaching the optimal efﬁciency with a unity (100%) TFEE score.

50 55 60 65 70 75 80

West Central East

Total-factor energy efficiency

of area (Percent)

.

Per capita income

Fig. 4. The U-shape relation existing between total-factor energy efﬁciency of an area and per capita income of an area in China.

(10)

areas in China and that of China’s total are dropping in the sample period.

The result is different from the result found from the TFEE score. First, all areas’ PFEE scores are declining in the period, meaning that energy efficiency drops in all areas of China. The more developed the area is, such as the east area, the more the energy efficiency drops. In TFEE, the energy efficiency drops the most for an area in a developing stage but not in a developed stage. This is because in a developing stage, out-of-date technolo-gies and inappropriate production processes need to be upgraded while more output is still generated. The upgraded technologies and more advanced production processes should have been launched when an area or a region enters a developed stage and a better efficiency level is therefore expected. Second, energy efficiency is proportional to an economy’s developmental progress in an area. TFEE does not mean that a proportional relation exists between the state of economic

develop-ment and energy efﬁciency, but a U-shape relation does, which ﬁts better to the real condition.

The comparative result also shows that the substitu-tion among inputs (labor, capital stock, and energy) to produce the output (GDP) is signiﬁcant. The PFEE scores could be over-estimated if energy is taken as the single input in the production. A certain portion of GDP output is produced not only by the energy itself, but also by other inputs (labor and capital stock). Hence, a multiple-inputs framework is integral to correctly evaluate energy efﬁciency, with which the index TFEE is established.

5. Concluding remarks

The index of total-factor energy efﬁciency (TFEE) is ﬁrst constructed in this study by taking the ratio of actual energy input to target energy input and this is

Table 5

Regional partial factor energy efﬁciency–the energy productivity ratio (1995–2002)

ID Region 1995 1996 1997 1998 1999 2000 2001 2002 1 Beijing E 39.7 40.2 39.6 40.0 39.6 39.4 41.2 41.1 2 Tianjin E 35.8 40.2 42.3 42.5 41.2 39.3 39.4 39.1 3 Hebei E 31.7 35.2 36.7 36.2 35.4 34.5 33.5 30.4 4 Liaoning E 28.9 29.5 30.9 33.2 32.3 29.1 29.5 29.7 5 Shanghai E 55.1 55.3 59.2 58.9 56.3 55.5 53.2 50.9 6 Jiangsu E 64.1 67.4 70.1 69.0 68.5 66.8 66.9 63.7 7 Zhejiang E 77.0 77.8 76.8 74.3 71.4 67.8 64.6 60.8 8 Fujian E 94.8 96.9 99.3 100.5 93.0 89.2 84.0 77.3 9 Shandong E 57.0 59.3 61.0 61.9 61.4 69.8 59.3 55.0 10 Guangdong E 78.1 76.7 77.2 73.6 70.4 68.5 65.4 59.7 11 Guangxi E 67.4 70.3 64.9 60.8 57.4 51.5 52.2 47.4 12 Hainan E 120.2 102.9 88.2 83.9 79.4 72.4 65.6 62.2 13 Shanxi C 13.0 17.4 17.8 18.8 16.8 16.4 14.0 12.4 14 Inner Mongolia C 31.6 31.8 27.2 30.4 24.2 26.5 23.7 21.9 15 Jilin C 27.5 29.2 28.0 32.3 32.8 33.4 32.9 29.7 16 Heilongjiang C 33.9 37.3 35.3 36.9 34.7 35.3 36.9 37.3 17 Anhui C 47.8 47.2 50.9 47.7 45.1 41.7 40.2 38.7 18 Jiangxi C 52.1 64.2 67.5 71.1 66.8 60.5 58.4 54.3 19 Hennan C 46.4 50.4 51.0 46.8 45.0 43.7 42.8 41.3 20 Hubei C 42.3 45.1 47.4 47.7 46.8 45.7 48.1 42.7 21 Hunan C 40.5 44.1 52.2 51.1 59.1 60.8 53.9 49.6 22 Sichuan W 37.1 40.7 42.2 38.9 37.5 37.7 39.3 36.8 23 Guizhou W 19.2 17.8 16.8 15.2 16.5 15.4 15.3 15.3 24 Yunnan W 45.7 49.1 40.2 41.5 41.0 40.8 37.2 32.7 25 Shaanxi W 31.7 30.4 35.8 35.6 40.5 40.7 35.4 31.6 26 Gansu W 20.2 23.2 25.4 25.2 23.2 21.9 23.1 22.2 27 Qinghai W 24.0 24.0 24.0 23.2 18.4 20.1 20.2 19.3 28 Ningxia W 22.4 22.0 22.0 21.4 20.7 20.5 21.0 20.8 29 Xinjiang W 29.2 25.8 27.3 26.5 26.4 27.6 26.6 25.4 Average 45.3 46.6 46.8 46.4 44.9 43.9 42.2 39.6 East 62.5 62.6 62.2 61.2 58.8 57.0 54.6 51.5 Central 37.2 40.7 41.9 42.5 41.3 40.4 39.0 36.4 West 28.7 29.1 29.2 28.4 28.0 28.1 27.2 25.5

Notes: (1) The unit is percentage. (2) E: east area, C: central area, and W: west area. (3) Data for the administration region Chongqing is regarded as a part of Sichuan in this paper since it was promoted as one of the municipalities in China only from 1997. (4) Data of energy consumption in Tibet is not included since it was not available in the sample period.

(11)

conducted through DEA. The formula is constructed in Eq. (3). The multiple-inputs model is adopted for an assessment of energy efﬁciency since energy is not the only input to produce output and it has to be accompanied by other inputs to produce real economic outputs. On the other hand, a certain portion of GDP output is generated not only by energy input, but also by inputs of labor and capital stock. A substitution effect, whereby GDP output is generated from other inputs rather than energy input, is therefore well considered under such a multiple-inputs model. Index TFEE therefore evaluates energy efﬁciency in an appropriate approach.

A unity TFEE score identifies a DMU which operates optimally at the efficiency of energy consumption. The DMU with a unity TFEE score consumes the minimum level of energy input and generates the maximum economic output including other inputs considered. An efficiency frontier of energy consumption is con-stituted by these DMUs. For those DMUs which have a TFEE score of less than unity and are not yet at the frontier efficiency of energy consumption; they are capable of taking one of those DMUs located at the frontier as a target to adjust their technology levels and production processes according to efficiency improve-ment. The slack and radial adjustments provide adequate input targets for a DMU with a TFEE less than unity to achieve the optimal energy efficiency.

This paper reports the result of an empirical study of regional energy efficiency in China based on the index TFEE. The commonly used index of the energy productivity ratio (Patterson, 1996) as PFEE is calcu-lated as well for comparison. The index PFEE observes only the partial effect between energy input and GDP output, but not the total-factor effects observed in TFEE based on the multiple-inputs model. As a result of comparison, index TFEE shows the advantage of assessing energy efficiency as being more practical than PFEE. A U-shape relation is found between the TFEE and per capita income in an area, which explains that an improvement in energy efficiency is followed by economic growth though it declines in the beginning

period. As Table 4 shows, TFEE in most regions of

China, especially for those located in the east and central areas, is improving during the research period. The more economic growth there is, the more is energy efﬁciency a concern and can improve. These results cannot be observed from a conventional PFEE index.

The developed area (east area) in China has the highest TFEE rank and the least developed area (west area) has the second best rank. However, the developing area (the central area) has the worst TFEE rank even though this area creates the second highest level of GDP output in China. Shanghai (5), Fujian (8), Guangdong (10), and Sichuan (22) are the four regions which constitute the efﬁciency frontier of energy consumption

in our analysis. These regions are going to be targets and learning models for other regions which are not yet at the optimal level of energy efﬁciency so that they can improve their technology level and production processes accordingly.

It is noted that the index TFEE assesses the energy efﬁciency of each region in China based on China’s own frontier, and not on the other economies’ or countries’. This means that the feasible improvement steps do exist for those energy-inefﬁcient regions to move onto the frontier. Similarities do exist on a variety of inputs, technology, and production processes among these regions, and the target is reachable and total adjust-ments are possibly reduced from the actual inputs in these regions by improving technology and production processes.

Because sustainable economic development relies upon efficient energy consumption, energy efficiency on a regional basis is the main focus in this study. Further improvement actions can be taken at the regional level and a more detailed analysis can be conducted by taking a reference for each region’s root causes at a deeper level. Industrial structure, energy policies, energy consumption type, and treatments from local governments can be further incorporated together. A world energy efficiency frontier can be also estab-lished with the same method to overview China’s efficiency rank of energy consumption globally, but it would require more data to be collected. As long as a good balance between economic growth and efficiency of energy consumption is reached, sustainable develop-ment with sufficient energy supply can be achieved.

Acknowledgements

We would like to express our appreciation for the valuable comments from an anonymous referee and an editor of this journal. The authors also thank Shih-Nang Chen, Cherng G. Ding, Kuo-Hsiung Lin, and seminar participants at the Shih-Hsin University for their useful suggestions and comments. Partial ﬁnancial support from Taiwan’s National Science Council (NSC 93-2415-H-009-001) is gratefully acknowledged. The usual disclaimer applies.

References

Banker, R.D., Charnes, A., Cooper, W.W., 1984. Some models for estimating technical and scale inefﬁciencies in data envelopment analysis. Management Science 30, 1078–1092.

Boyd, G.A., Pang, J.X., 2000. Estimating the linkage between energy efﬁciency and productivity. Energy Policy 28, 289–296.

Business Asia, 2005. Economic outlook: China. Business Asia 37 (6), 9. Chang, J., Leung, D.Y.C., Wu, C.Z., Yuan, Z.H., 2003. A review on the energy production, consumption, and prospect of renewable

(12)

energy in China. Renewable and Sustainable Energy Reviews 7, 453–468.

Charnes, A., Cooper, W.W., Rhodes, E., 1978. Measuring the efﬁciency of decision making units. European Journal of Opera-tional Research 2, 429–444.

Coelli, T.J., 1996. A Guide to Deap Version 2.1: A Data Envelopment Analysis (Computer) Program. Department of Econometrics, University of New England, Armidale, Australia.

Coelli, T.J., Rao, D.S.P., Battese, G.E., 1998. An Introduction to Efﬁciency and Productivity Analysis. Kluwer Academic Publishers, Boston, MA.

Fleisher, B., Chen, J., 1997. The coast–noncoast income gap, productivity, and regional economic policy in China. Journal of Comparative Economics 25, 220–236.

Gibbs, D., 1998. Regional development agencies and sustainable development. Regional Studies 32, 365–368.

Gibbs, D., 2000. Ecological modernization, regional development and regional development agencies. Geoforum 31, 9–19.

Hawdon, D., 2003. Efﬁciency, performance and regulation of the international gas industry—a bootstrap DEA approach. Energy Policy 31, 1167–1178.

International Energy Agency (IEA), 2004. World Energy Outlook 2004. OECD/IEA, Paris.

Kadoshin, S., Nishiyama, T., 2000. The trend in current and near future energy consumption from a statistical perspective. Applied Energy 67, 289–296.

Li, K.W., 2003. China’s capital and productivity measurement using ﬁnancial resources. Center Discussion Paper No. 851, Economic Growth Center, Yale University.

Lovell, A.K., Schmidt, S.S., 1993. Production Frontiers and Produc-tive Efficiency, the Measurement of ProducProduc-tive Efficiency: Tech-niques and Applications. Oxford University Press, New York. Patterson, M.G., 1996. What is energy efficiency? Concepts, indicators,

and methodological issues. Energy Policy 24, 377–390.

Ramanathan, R., 1999. A holistic approach to compare energy efﬁciencies of different transport modes. Energy Policy 28, 743–747.

Sinton, J.E., Fridley, D.G., 2002. A guide to China’s energy statistics. Journal of Energy Literature 8, 22–35.

World Bank, 2001. China country brief.http://www.worldbank.org. Yang, D.T., 2002. What has caused regional inequality in China?