- Email: [email protected]
Total-factor energy efficiency in developing countries
Energy Policy 39 (2011) 644–650
Contents lists available at ScienceDirect
Energy Policy journal homepage: www.elsevier.com/locate/enpol
Total-factor energy efficiency in developing countries Xing-Ping Zhang n, Xiao-Mei Cheng, Jia-Hai Yuan, Xiao-Jun Gao School of Economics and Management, North China Electric Power University, Beijing 102206, China
a r t i c l e i n f o
a b s t r a c t
Article history: Received 5 May 2010 Accepted 27 October 2010 Available online 13 November 2010
This paper uses a total-factor framework to investigate energy efficiency in 23 developing countries during the period of 1980–2005. We explore the total-factor energy efficiency and change trends by applying data envelopment analysis (DEA) window, which is capable of measuring efficiency in crosssectional and time-varying data. The empirical results indicate that Botswana, Mexico and Panama perform the best in terms of energy efficiency, whereas Kenya, Sri Lanka, Syria and the Philippines perform the worst during the entire research period. Seven countries show little change in energy efficiency over time. Eleven countries experienced continuous decreases in energy efficiency. Among five countries witnessing continuous increase in total-factor energy efficiency, China experienced the most rapid rise. Practice in China indicates that effective energy policies play a crucial role in improving energy efficiency. Tobit regression analysis indicates that a U-shaped relationship exists between total-factor energy efficiency and income per capita. & 2010 Elsevier Ltd. All rights reserved.
Keywords: DEA window analysis Total-factor energy efficiency Tobit model
1. Introduction Global warming is one of the world’s most important environmental problems. The problem is largely attributable to the greenhouse gas carbon dioxide, which is released by the burning of fossil fuels. Because of this, a growing body of research has focused on improving energy efficiency, which is a crucial approach to alleviating global warming. During the past several decades, some appropriate methods have been developed to monitor energy-efficiency trends and compare energy efficiency performance across countries. These methods are generally classified as parametric and non-parametric methods (Sadjadi and Omrani, 2008). Parametric methods such as stochastic frontier analysis estimate a cost or production function. Therefore, deviations in the function form affect the results of such models. In contrast, it is not necessary to estimate the cost or production function when using non-parametric methods. Data envelopment analysis (DEA) is a non-parametric method that is capable of handling multiple inputs and multiple outputs. Energy intensity and energy efficiency are the two well-known energy-efficiency indicators that are commonly used in macrolevel policy analysis. Energy intensity is defined as the energy consumption divided by the economic output, and energy efficiency is the reciprocal of energy intensity. These traditional energy-efficiency indicators take energy consumption into account as a single input that produces an economic output; therefore, some other key inputs are ignored, such as capital and labor. Energy consumption must be combined with other inputs to produce an
n
Corresponding author. Tel.: + 86 1051963541; fax: +86 1080796904. E-mail address: [email protected] (X.-P. Zhang).
0301-4215/$ - see front matter & 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.enpol.2010.10.037
economic output, and substitution effects exist between energy and other input factors (e.g., labor and capital stock). If energy consumption is evaluated in terms of partial-factor energy efficiency, the result is a misleading estimate (Hu and Wang, 2006; Honma and Hu, 2009). Boyd and Pang (2000) indicated that energyefficiency improvement relies on total-factor productivity improvement. To overcome the disadvantage of partial-factor energy efficiency, an increasing number of researchers have devoted themselves to analyzing total-factor energy efficiency using DEA. There are essentially two research strands in the literature analyzing total-factor energy efficiency using DEA. The first strand focuses on measuring total-factor productivity change based on the DEA–Malmquist index, which was first introduced by Caves et al. (1982). For example, Forsund and Kittelsen (1998) applied DEA efficiency scores to calculate the Malmquist productivity index in Norwegian electricity distribution companies. Edvardsen and Fbrsund (2003) applied an input-oriented DEA model and the Malmquist productivity index to analyze the performance of 122 electricity distributors in Denmark, Finland, Norway, Sweden and the Netherlands. Wei et al. (2007) applied this approach to investigate changes in the energy efficiency of China’s iron and steel sectors. The index of total-factor productivity takes only radial adjustment into account and disregards the non-radial slack; it is therefore unable to measure single-factor efficiency under a totalfactor framework (Honma and Hu, 2009). The second strand utilizes the index of total-factor energy efficiency (TFEE) first proposed by Hu and Wang (2006). TFEE is defined as the target energy input divided by the actual energy input. Taking non-radial slack into account, the TFEE is capable of measuring single-factor efficiency in a total-factor framework. Following Hu and Wang (2006), Honma and Hu (2008) measured
X.-P. Zhang et al. / Energy Policy 39 (2011) 644–650
the TFEE of 47 regions in Japan for the period 1993–2003. Integrating the concept of TFEE with the Malmquist productivity index, Hu and Chang (2009) proposed the total-factor energy productivity index (TFEPI) to investigate energy productivity changes in regions of China. Honma and Hu (2009) extended their previous work (Honma and Hu, 2008) by applying TFEPI. Following Hu and Wang (2006) and Honma and Hu (2008), we use DEA theory to investigate the total-factor energy efficiency of developing countries. This paper extends the contributions of these earlier studies in three ways. First, this study investigates the totalfactor energy efficiency of 23 developing countries. The Kyoto Protocol is severely criticized for not including emission reduction obligations for developing countries because many developing countries have become major carbon dioxide emitters. For example, China is the second largest energy-related CO2 emitter and India ranks fourth in the world. Therefore, the choice to study developing countries is motivated by the rising importance of their contribution to global warming. Secondly, we use the DEA window analysis introduced by Charnes and Cooper (1985) for the first time to develop total-factor energy-efficiency measures for the 23 developing countries during the period of 1980–2005. This approach can indicate efficiency trends over a specified period of time while simultaneously examining stability and other properties of the efficiency evaluations within the specified windows (Hartman and Storbeck, 1996; Webb, 2003). Third, we use dynamic Tobit model to investigate the relationship between total-factor energy efficiency and income for 23 developing countries over the period 1980–2005. The remainder of this paper is organized as follows: The next section describes the methods used in the study; Section 3 presents the data; Section 4 presents the empirical results; and Section 5 concludes the paper.
2. Method 2.1. Total-factor energy efficiency based on DEA Hu and Wang (2006) proposed the TFEE, which is defined as the target energy input divided by the actual energy input, and also utilized what is known as the constant returns to scale (CRS) DEA model. It is universally known that the assumption of the CRS model is appropriate only when all decision-making units (DMUs) are operating at an optimal scale. However, some factors, such as imperfect competition and constraints on finance, may cause a DMU not to operate at an optimal scale. If it is likely that the size of the DMUs under investigation will influence their ability to create outputs efficiently, then the assumption of CRS is inappropriate (Halkos and Tzeremes, 2009). Banker et al. (1984) suggested an extension of the CRS model to account for variable returns to scale (VRS) situations. The less restrictive VRS frontier allows the best practice level of outputs to inputs to vary with the size of the countries (Halkos and Tzeremes, 2009). Because the 23 developing countries in this study have different sizes, we use the VRS model. The VRS model could be obtained by adding the convexity constraint based on the CRS model (Charnes et al., 1978). The model employs the following mathematical notation: For each of N DMUs, there are K inputs and M outputs. For the ith DMU, the inputs and outputs are represented by the column vectors xi and yi, respectively. The input-oriented VRS model solves the following linear programming problem for DMUto : Min s:t:
y N X j¼1
N X j¼1 N X
lj xj þ s ¼ yxo lj ¼ 1
j¼1
lj Z 0, j ¼ 1, 2,. . .,N
ð1Þ
where y is a scalar with 0r y r1, l is a N 1 vector of constants that form a convex combination of observed inputs and outputs and s represents the non-radial slack. 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). The value of y represents the technically efficient score for the ith DMU. (1 y)xi is called radial adjustments. The ith DMU is the most efficient point on the frontier and is technically efficient if y ¼1 and the slack equals zero; if y ¼1 and the slack is larger than zero, the ith DMU is weakly technically efficient; y o1 indicates that the ith DMU is technically inefficient. The sum of the radial and non-radial adjustments is called the total adjustments, which can be reduced to reach optimal technical efficiency without decreasing the output levels. The total adjustments for the technically efficient DMUs equal zero, but are larger than zero for the other DMUs. The target energy input (TEI) in this study is therefore actual energy input (AEI) minus the total adjustments (TA), which represents a practical minimum level of energy input to be taken as a target to perform at the optimal energy consumption efficiency. Hence, the TFEE index of DMU i at time t can be measured as (Hu and Wang, 2006) TFEEði,tÞ ¼
TEIði,tÞ AEIði,tÞTAði,tÞ ¼ AEIði,tÞ AEIði,tÞ
ð2Þ
The total adjustments are not less than zero according to model (1). Total adjustments of zero indicate that the actual energy input is indeed the target energy input, so TFEE is unity and the energy is used at the optimal efficiency. TFEE is lower than unity if the total adjustments are larger than zero. This implies there is redundant energy input, which should be reduced without decreasing the output. Therefore, the index of TFEE is always between zero and unity. The greater the value of TFEE, the more efficient the energy consumed.
2.2. Total-factor energy efficiency based on DEA window analysis DEA window analysis, which was introduced by Charnes and Cooper (1985), is a variation of the traditional DEA that can handle cross-sectional and time-varying data to allow for dynamic effects. DEA window analysis operates on the principle of moving averages (Charnes et al., 1994a; Yue, 1992) and establishes efficiency measures by treating each DMU in different years as a separate unit. The performance of a DMU in a period can be contrasted with its own performance in other periods as well as to the performance of other DMUs (Asmild et al., 2004). Therefore, DEA window analysis can explore the evolution of performance through a sequence of overlapping windows. Moreover, the number of data points is increased several times over in DEA window analysis, which is very useful when dealing with small sample sizes. A brief DEA window analysis review is presented here. A window with N w observations is denoted starting at time t(1rt rT) with window width w (1rwrT t). The matrix of inputs for this window is 1 2 N 1 2 N Xtw ¼ ðx1t ,x2t ,. . .,xN t ,xt þ 1 ,xt þ 1 ,. . .,xt þ 1 ,. . .,xt þ w ,xt þ w ,. . .,xt þ w Þ
and the matrix of outputs is given by
lj yj s þ ¼ yo
645
1 2 N 1 2 N Ytw ¼ ðy1t ,y2t ,. . .,yN t ,yt þ 1 ,yt þ 1 ,. . .,yt þ 1 ,. . .,yt þ w ,yt þ w ,. . .,yt þ w Þ
646
X.-P. Zhang et al. / Energy Policy 39 (2011) 644–650
The input oriented DEA window problem for DMUto is given by solving the following linear program:
worker. The labor force and It are available from the World Development Indicators database (World Bank, 2005).
min y s:t:
4. Empirical results
yxot Xtw l Z0 N X
lj ¼ 1
Ytw lyot Z 0
ð3Þ
lj Z0, j ¼ 1,2,. . .,N w
j¼1
A ‘‘window’’ is constructed by the number of time periods (years in our study) under analysis, which conceived of in an intertemporal manner, according to Tulkens and Vanden Eeckaut (1995). This study includes 23 countries (N ¼23) for the time period of 1980–2005(T¼26). It should be noted that because all units within a given window are measured against each other, this approach implicitly assumes that there are no technical changes within each window. This is a general problem in DEA window analysis (Asmild et al., 2004). A narrow window length therefore must be used to reduce this problem. Charnes et al. (1994b) found that a window length of three or four tended to yield the best balance of informativeness and stability of the efficiency scores. Following Halkos and Tzeremes (2009), we chose a narrow window length of three (w¼3) to create credible results. Thus, the first three years of 1980, 1981 and 1982 form the first window. The window then moves on a one-year period by dropping the original year and adding a new year; therefore, the next three years of 1981, 1982 and 1983 construct the second window. The process continues until the last window, which contains the last three years (2003, 2004 and 2005 in this case), is constructed. In doing so, 24 windows are performed for each country and the number of DMUs (countries) in each window increases from 23 to 69(N w¼23 3). The TFEEs of 23 developing countries in each window could be obtained using DEA window analysis. For each country, each year has three values on TFEE, with the exception of 1980 and 2005, which has only one value, and 1982 and 2004, which has two values. We then calculate the average result of the TFEEs of each country in the same year, which creates a new TFEE result for the 23 developing countries analyzed in our study.
3. Data and variables We use annual data on the labor force, energy consumption (kt of oil equivalent) and capital stock as the three input variables, while gross domestic product (GDP) is the single output in this study. Due to the unavailability of data on some developing countries, we select 23 developing countries to analyze in this paper. Except capital stock, all data are extracted from the World Development Indicators database (World Bank, 2005) for the period 1980–2005. To standardize the data, capital stock and real GDP are measured in purchasing power parity (constant 2005, international $). To the best of our knowledge, the data on capital stock from 1991 to 2005 could not be obtained from any statistical yearbook or database. Following Hu and Kao (2007) and Chien and Hu (2007), we apply the following perpetual inventory method to calculate the capital stock: K t ¼ It þð1dÞKt1 where Kt is the gross capital stock in current year; Kt 1 is the gross capital stock in the previous year; d represents the depreciation rate of capital stock; and It is the gross fixed capital formation in the current year. d is set to 6% in accordance with the suggestions of many relevant studies such as Iyer et al. (2004) and Wu (2004). The data on capital stock per worker in 1980 is available from Penn World Table 5.6 (1998). Hence, the capital stock in 1980 is obtained by multiplying the labor force total by capital stock per
4.1. Total-factor energy efficiency DEA window analysis is applied in this case, and Table 1 shows energy efficiency for China.1 We can explore the evolution of energy performance through a sequence of overlapping windows. Viewing the column data of Table 1, we can examine the stability of the total-factor energy efficiency of China across the different datasets. The row data enable us to test the trends across the same dataset. According to the last row of Table 1, it is obviously seen that China has significantly improved its TFEE between 1980 and 2005. To explore the trend changes of TFEEs for 23 developing countries, we show the overall TFEE scores of all 23 developing countries in Table 2. The countries with TFEE of unity constitute the efficiency frontier of energy consumption among 23 countries in the same year, which means that energy is used at the optimal level in these countries. Table 2 indicates that the efficiency frontier is different in different year. For example, six countries (Botswana, Honduras, Morocco, Panama, Paraguay and Peru) constitute the efficiency frontier of energy consumption in 1980, while another five countries (Botswana, China, Indian, Mexico and Thailand) constitute the efficiency frontier of energy consumption in 2005. The total-factor energy efficiency of all 23 developing countries displays three evolving trends during the research period. First, the TFEEs of seven countries (Argentina, Bolivia, Botswana, Chile, Kenya, Mexico and Panama) slightly fluctuate over time. It is noteworthy that the average TFEEs of three countries (Botswana, Mexico and Panama) are on the verge of unity, at 0.996, 0.990 and 0.960, respectively. Furthermore, Botswana is on the efficiency frontier of energy consumption in 18 out of 26 years, Mexico is on the efficiency frontier in 10 out of 26 years, and Panama is also on the efficiency frontier in the initial period. Therefore, these three countries are the most energy-efficient economies among 23 developing countries during the whole research period. However, Kenya is particularly noteworthy because its TFEEs are lower than 0.41 during the entire research period, and it has the lowest average TFEE score (0.329) in the 23 developing countries. Secondly, eleven countries display a downward trend in totalfactor energy efficiency: Iran ( 2.02%), Venezuela ( 1.90%), Morocco ( 1.86%), Honduras ( 1.49%), Syria ( 1.37%), Paraguay ( 1.36%), Guatemala ( 1.24%), Philippine ( 1.13%), Ecuador ( 0.88%), Peru ( 0.85%) and Dominican ( 0.28%).2 Therefore, the total-factor energy efficiency of these countries worsened between 1980 and 2005. It is worth noting that Honduras, Morocco, Paraguay and Peru are at the efficient production frontier (TFEE ¼1) in the first few years. However, their TFEEs become smaller and smaller in the later period, that is, they are distancing themselves from the efficient frontier over time. Philippine and Syria are concerned especially for their very low TFEEs in entire research period. Finally, there is a continuous increasing trend in the total-factor energy efficiency of five countries: China (8.64%), Zambia (4.28%), India (3.73%), Thailand (2.77%) and Sri Lanka (1.80%). Special attention should be paid to the first four countries (China, India, Thailand and Zambia): their TFEEs are very low (less than 0.5) in the initial period, but they show a significant improvement from the beginning of the 1990s. It can be noteworthy that China and India 1
The results of the other 22 developing countries are available upon request. The percentage in parenthesis is the annual average growth rate of TFEE between 1980 and 2005. 2
X.-P. Zhang et al. / Energy Policy 39 (2011) 644–650
647
Table 1 A three-year window analysis of total-factor energy efficiency in China. Windows/years
1980
1981
1982
Window Window Window Window Window Window Window Window Window Window Window Window Window Average
0.116
0.123 0.123
0.13 0.13 0.136
1 2 3 4 5 6 7 8 9 10 11 12 13
1983
0.139 0.145 0.147
1984
0.158 0.159 0.159
1985
1 0.621 0.451
1986
1 0.688 0.56
1987
1 0.782 0.703
1988
1 0.919 0.875
1989
1 0.963 0.762
1990
1 0.795 0.612
1991
1 0.772 0.64
1992
0.116
0.123
0.132
0.143
0.159
0.69
0.749
0.828
0.931
0.908
0.802
0.804
1 0.825 0.695 0.84
Windows/years
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Window Window Window Window Window Window Window Window Window Window Window Window Window Average
1 0.845 0.77
1 0.974 0.986 0.987
1 0.964 0.982
1 1
12 13 14 15 16 17 18 19 20 21 22 23 24
0.872
0.977 0.907 0.82
0.901
0.988 0.901 0.795
0.895
1 0.884 0.795
0.893
1 0.9 0.818
0.906
1 0.909 0.822
0.91
were on the efficiency frontier in 2005, Thailand constituted the efficiency frontier in the last few years, and Zambia was on the efficiency frontier in the early 1990s. Although Sri Lanka improved TFEE between 1980 and 2005, its energy efficiency is very low.
4.2. Experiences from China Among 23 developing countries, China experienced the most rapid rise in TFEE from the onset of economic reform in the late 1970s. The past three decades has witnessed the rapid economic growth in China. At the same time, energy consumption has largely increased. China has become the world’s largest consumer for coal since 1986, and the second largest consumer for oil since 2002 (BP, 2009). What’s more, China’s demand for energy has outpaced its energy production since 2001. Environmental problem, especially ‘‘climate change’’, has become the focus of the world since the UN Conference on Environment and Development held in 1992. In order to alleviate energy shortage and climate change, Chinese government has issued and carried out a series of energy policies to improve energy efficiency since late 1970s. Table 3 reports partial key energy policies implemented by China. As early as 1992, China issued ‘‘Ten solutions for environment and development in China’’, in which enhancing energy utilization efficiency is one of the ten solutions. Then in the ninth Five-YearPlan (1996–2000), Government of China proposed an energy conservation target of 5% annually and reduction for principal pollutant from 1996 to 2000, which is the first energy saving and emission reducing goal among the developing countries. Since the mid-1990s, efficiency improvements have been particularly marked in energyintensive industries such as metallurgy, cement, paper, textiles, oil and coal processing and electrical power generation (Steenhof, 2006; Liao et al., 2007). As Weidou and Johansson (2004) pointed out: China has also met its soaring needs for energy services while at
1 0.906 0.818
0.908
1 0.901 0.875
0.925
1 0.971 0.976
0.982
1 0.985 0.981 0.989
the same time has improved the efficiency of its energy use by a factor of nearly three. Lin et al. (2008) stated that China’s experience from 1980 to 2000 was an exception because in the early stage of economic development, industrialization and urbanization tend to lead to extensive infrastructure and housing development, which will consume much energy and material. Government of China officially takes sustainable development as the crucial state policy in the 16th Conference of Communist Party of China (CPC) held in 2002. The Medium and Long-Term Plan for Energy Conservation, approved by Chinese government in 2004, indicated that the GDP energy intensity should decrease by 2.2% annually until 2010. Medium and Long-Term Plan for Energy Conservation is a milestone in China’s energy policy: financial incentive and market mechanism came into operation from then on (China paid much attention to the measures of command-and-control in energy policy before). Although China’s energy efficiency is very low compared with developed countries, China has been devoting to improve energy efficiency and has achieved a great success of annual energy saving by 3.9% from 1980 to 2005. Furthermore, the Eleventh Five-Year Plan for Energy Development, approved by National Development and Reform Commission (NDRC) of China in 2007, set an ambitious goal of reducing energy intensity by 20% between 2006 and 2010. At the end of 2009, 15.6% reduction in energy intensity as of 2005 has been achieved. Practice in China shows that effective energy policies play an important role in forcing all levels of government and energy-using units to take active measures to improving energy efficiency, such as, improving technological level, restructuring industries and products and so on. Numerous studies have investigated the components driving the change of energy intensity by using decomposition analysis (Huang, 1993; Sinton and Levine, 1994; Lin and Polenske, 1995; Garbaccio et al., 1999; Zhang, 2003; FisherVanden et al., 2003; Liao et al., 2007). Most previous studies indicated that the decline of energy intensity can mainly be attributed to technological changes, while there is disagreement
648
X.-P. Zhang et al. / Energy Policy 39 (2011) 644–650
Table 2 Total-factor energy efficiency of 23 developing countries. Country
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
Argentina Bolivia Botswana Chile China Dominican Ecuador Guatemala Honduras India Iran Kenya Mexico Morocco Panama Paraguay Peru Philippines Sri Lanka Syria Thailand Venezuela Zambia
0.919 0.722 1.000 0.823 0.116 0.794 0.928 0.916 1.000 0.386 0.939 0.243 0.978 1.000 1.000 1.000 1.000 0.741 0.409 0.652 0.491 0.954 0.313
0.850 0.698 1.000 0.760 0.123 0.803 0.924 0.856 1.000 0.389 0.775 0.247 1.000 0.884 1.000 0.922 1.000 0.722 0.431 0.636 0.494 0.969 0.323
0.812 0.661 0.993 0.712 0.132 0.790 0.921 0.762 0.973 0.393 0.787 0.257 0.967 0.910 0.997 0.780 0.986 0.715 0.431 0.591 0.507 0.948 0.312
0.839 0.656 1.000 0.676 0.143 0.774 0.933 0.747 0.999 0.411 0.992 0.272 0.989 0.869 0.956 0.711 0.949 0.688 0.454 0.577 0.587 0.920 0.303
0.848 0.672 1.000 0.710 0.159 0.746 0.915 0.733 1.000 0.415 0.887 0.278 1.000 0.867 0.983 0.695 0.977 0.692 0.457 0.510 0.561 0.823 0.294
0.811 0.673 0.996 0.747 0.690 0.778 0.910 0.719 1.000 0.422 0.855 0.290 1.000 0.851 1.000 0.690 0.963 0.600 0.464 0.537 0.520 0.711 0.287
0.849 0.671 0.981 0.754 0.749 0.766 0.915 0.727 1.000 0.715 0.773 0.303 0.991 0.851 1.000 0.660 0.986 0.605 0.455 0.496 0.500 0.690 0.283
0.859 0.699 1.000 0.772 0.828 0.751 0.861 0.713 1.000 0.741 0.760 0.318 1.000 0.775 0.968 0.623 1.000 0.599 0.439 0.477 0.496 0.812 0.279
0.830 0.701 1.000 0.766 0.931 0.763 0.912 0.712 1.000 0.848 0.714 0.333 0.994 0.806 0.909 0.621 0.950 0.602 0.438 0.519 0.523 0.963 0.488
0.776 0.704 1.000 0.795 0.908 0.794 0.957 0.713 1.000 0.891 0.707 0.341 0.988 0.753 0.945 0.617 0.907 0.593 0.438 0.460 0.505 0.897 0.722
0.753 0.744 1.000 0.796 0.802 0.747 0.981 0.729 0.691 0.920 0.773 0.361 0.997 0.747 0.995 0.647 0.863 0.575 0.475 0.450 1.000 0.917 0.996
0.803 0.781 1.000 0.832 0.804 0.756 0.981 0.748 0.728 0.850 0.843 0.373 1.000 0.763 1.000 0.645 0.856 0.566 0.509 0.445 1.000 0.976 1.000
0.853 0.794 0.962 0.853 0.840 0.748 0.952 0.767 0.741 0.841 0.854 0.377 0.994 0.695 0.992 0.658 0.816 0.544 0.540 0.481 0.999 0.989 1.000
0.890 0.790 0.990 0.875 0.872 0.772 0.933 0.783 0.754 0.832 0.814 0.377 0.991 0.667 1.000 0.652 0.794 0.535 0.560 0.493 1.000 0.986 1.000
Country
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Average
Argentina Bolivia Botswana Chile China Dominican Ecuador Guatemala Honduras India Iran Kenya Mexico Morocco Panama Paraguay Peru Philippines Sri Lanka Syria Thailand Venezuela Zambia
0.886 0.786 1.000 0.868 0.901 0.716 0.912 0.783 0.726 0.860 0.700 0.384 1.000 0.706 0.991 0.637 0.843 0.527 0.600 0.514 1.000 0.768 1.000
0.850 0.762 0.991 0.920 0.895 0.725 0.860 0.784 0.711 0.892 0.693 0.395 0.958 0.661 0.990 0.628 0.853 0.514 0.624 0.507 1.000 0.842 0.989
0.870 0.734 1.000 0.924 0.893 0.744 0.800 0.788 0.722 0.909 0.675 0.406 0.972 0.717 0.969 0.615 0.842 0.517 0.609 0.530 1.000 0.677 0.999
0.893 0.727 1.000 0.908 0.906 0.728 0.809 0.780 0.700 0.860 0.690 0.401 0.991 0.687 0.974 0.591 0.889 0.496 0.632 0.514 0.985 0.708 0.978
0.861 0.695 0.991 0.864 0.910 0.712 0.800 0.757 0.677 0.847 0.568 0.399 0.970 0.716 0.930 0.595 0.840 0.495 0.645 0.472 1.000 0.541 0.971
0.805 0.691 1.000 0.740 0.908 0.719 0.755 0.708 0.688 0.831 0.542 0.402 0.961 0.676 0.934 0.610 0.805 0.484 0.630 0.422 0.997 0.522 0.977
0.774 0.676 1.000 0.775 0.925 0.704 0.716 0.697 0.697 0.813 0.522 0.401 1.000 0.660 0.920 0.625 0.807 0.482 0.618 0.421 1.000 0.514 0.964
0.769 0.717 1.000 0.820 0.982 0.697 0.713 0.687 0.688 0.850 0.518 0.304 0.990 0.661 0.848 0.641 0.799 0.489 0.599 0.437 0.998 0.520 0.981
0.703 0.741 0.992 0.812 0.989 0.689 0.709 0.688 0.695 0.863 0.538 0.268 1.000 0.643 0.893 0.663 0.816 0.498 0.601 0.424 1.000 0.480 0.985
0.734 0.739 1.000 0.859 0.987 0.680 0.688 0.694 0.699 0.920 0.552 0.268 0.997 0.664 0.888 0.684 0.833 0.517 0.611 0.437 1.000 0.488 0.972
0.771 0.701 1.000 0.844 0.982 0.694 0.713 0.684 0.683 0.956 0.542 0.269 1.000 0.645 0.927 0.696 0.810 0.527 0.592 0.451 1.000 0.547 0.945
0.834 0.726 1.000 0.883 1.000 0.739 0.737 0.663 0.677 1.000 0.552 0.280 1.000 0.613 0.955 0.701 0.800 0.552 0.650 0.456 1.000 0.580 0.930
0.825 0.718 0.996 0.811 0.745 0.743 0.855 0.744 0.817 0.756 0.714 0.329 0.990 0.750 0.960 0.677 0.884 0.572 0.535 0.497 0.814 0.759 0.742
on the role of structural change (a shift in the mix of industries). Many studies found that structural change has played a minor role in reducing energy intensity. However, a few studies found that structural change actually increased energy intensity in a certain time period (Garbaccio et al., 1999;Fisher-Vanden et al., 2003). Ma and Stern (2008) confirmed that technological change played a dominant role in decreasing energy intensity over the period of 1980–2003, while structural effect plays a minor role. Furthermore, Ma and Stern (2008) indicated that inter-fuel substitution was found to contribute little to the changes in energy intensity. It is noteworthy that China has paid much attention to restructuring industries and inter-fuel substitution recently. For example, Renewable Energy Law and Law on Promoting the Circular Economy were issued in 2005 and 2009, respectively. Medium and Long Term Plan for Renewable Energy Development was released by NDRC in 2007, in which an ambitious target of renewable energy accounting for about 10% of the total energy consumption by 2010 and about 15% by 2020 is set. Furthermore, NDRC took active ways
to accelerate the development of energy management contract (EMC) in 2010, which is a new management model for China. 4.3. Relationship between TFEE and income Based on gross national income per capita, the World Bank classified the countries into four types: high income countries, upper middle income countries, lower middle income countries and low income countries. Because of this, we use the gross national income per capita (GNIPC) as a proxy for income level. The dependent variable TFEE is a fractional variable bounded between 0% and 100%, so Tobit regression with both truncation points is applied in this paper. To address the potential autocorrelation, we apply the following dynamic Tobit model3: TFEEit ¼ ai þ b1 TFEEi,t-1 þ b2 GNIPCit þ b3 GNIPCit2 þ eit 3
See Honore´ (1993) for a discussion of the dynamic Tobit model.
ð4Þ
X.-P. Zhang et al. / Energy Policy 39 (2011) 644–650
649
Table 3 Main energy policies in China from 1980. Year
Policies
Policymakers
1980 1982 1984 1986 1992
On Strengthening Energy Conservation Outline of Energy-Saving Technology Policy Interim Provisions for Some Issues of Comprehensive Utilization of Resources Regulations for Saving Energy Ten solutions for environment and development in China (Article IV addresses energy efficiency and energy structure) Electricity Law The ninth Five-Year-Plan Energy Conservation Law Energy Management Measures for Key Energy-Using Units Management Measures for Energy-Saving Product Authentication Civil Construction of Energy-Saving Regulations Rules for Implementing Energy Conservation Law in Transportation Industry Design Standards for Building Energy Conservation in Hot Summer-Winter Cold Districts The Tenth Five-Year Plan for Building Energy Conservation Medium and Long-Term Plan for Energy Conservation Renewable Energy Law Notification on Near-term Key Work for Building Conservation Society Decision on Strengthening Energy Conservation Eleventh Five-Year Plan for Energy Development Medium and Long Term Plan for Renewable Energy Development Regulations on Energy Conservation of Public Institutions Law on Promoting the Circular Economy On Strengthening Energy Management Contract and Promoting Energy Saving Service Industry
National Economy Commisssion (NEC) National Development and Reform Commission (NDRC) NEC State Council Chinese Foreign Ministry, and State Environmental Protection Administration National People’s Congress State Council National People’s Congress NEC Energy-saving product certification Committee of China National Ministry of Construction Ministry of Communications Ministry of Housing and Urban-Rural Development (MHURD)
1994 1996 1997 1999 2000 2001 2002 2004 2005 2006 2007 2008 2009 2010
Table 4 Tobit estimation results. Variable
Coefficient
Std. error
Prob.
TFEEt 1 GNIPC GNIPC squared
1.0442 1.1578 0.0701
0.0191 0.1677 0.0102
0.0000 0.0000 0.0000
where t ¼ 1,. . .,T time period and i ¼ 1,. . .,N members of the panel; ai is nation-specific effects; the error mit is assumed to be i.i.d. across i and over t. GNIPC is transformed into its logarithm, and the logarithm of TFEE is given by ln(100 TFEE). F-test and Hausman specification test results reject the random model at 5% significance level, therefore, we estimate Eq. (4) with fixed effects by using the maximum likelihood estimate. Table 4 reports the regression results. The positive coefficient of GNIPC squared suggests that a U-shaped relationship exists between TFEE and GNIPC for the sample studied in this paper. That is, energy efficiency worsens with the increase in income per capita due to growth of industries until a certain point. Then, it rises after income per capita reach the point.4
MHURD and NDRC NDRC National People’s Congress State Council State Council NDRC NDRC NDRC National People’s Congress NDRC
eleven countries show a downward trend in TFEE, particularly, Honduras, Morocco and Paraguay experienced a sharp decrease in energy efficiency. It is important to note that the TFEE index measures the energy efficiency of 23 developing countries based on their own frontier. Therefore, the energy-inefficient countries could take feasible improvement measures to reduce energy input without hindering economic output. Taking China as the example because of its great success in improving energy efficiency, we briefly review the energy policies carried out by China and indicate that effective energy policies play a crucial role in improving energy efficiency. The Tobit estimation results suggest there is a U-shaped relationship between TFEE and income per capita, which means the energy efficiency first decreases with the income per capita, and then increases after a certain level. The index of TFEE is capable of measuring energy efficiency in a total-factor framework. However, as in all DEA analyses, the TFEE index for a single country are generally affected by other countries in the sample. The reason is that the production frontier constructed depends on the sample. The results therefore should be interpreted with caution.
Acknowledgments 5. Conclusions This study measures energy efficiency in a total factor production framework using DEA window analysis to detect changes in efficiency over time. For this purpose, a sample of 23 developing countries for the period 1980–2005 is used. The total-factor energy efficiency and change trends of 23 developing countries are investigated. Botswana, Mexico and Panama perform the best in term of TFEE, while Kenya, Philippine, Sri Lanka and Syria perform the worst in energy efficiency during the entire period. Five countries experienced a continual increase in total-factor energy efficiency, and China experienced the most rapid rise in TFEE. While 4
We thank an anonymous reviewer for pointing this issue.
The authors thank the two anonymous referees and the editor of this journal for their valuable suggestions and helpful comments. The authors gratefully acknowledge the financial support the Natural Science Foundation of Beijing (9092013), the Human Resource Development Project of the Beijing City Committee (20081D1600900358), Beijing Planning Project of Philosophy and Social Science (10BaJG371) and the Fundamental Research Funds of the Central Universities. References Asmild, M., Paradi, C.V., Aggarwall, V., Schaffnit, C., 2004. Combining DEA window analysis with the Malmquist index approach in a study of the Canadian banking industry. Journal of Productivity Analysis 21, 67–89.
650
X.-P. Zhang et al. / Energy Policy 39 (2011) 644–650
Banker, R.D., Charnes, A., Cooper, W.W., 1984. Models for estimation of 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. BP, 2009. Statistical Review of World Energy. Caves, D.W., Christensen, L.R., Diewert, W.E., 1982. The economic theory of index numbers and the measurement of input, output, and productivity. Econometrica 50, 1393–1414. Charnes, A., Cooper, W.W., Rhodes, E., 1978. Measuring efficiency of decision making units. European Journal of Operational Research 2, 429–444. Charnes, A., Cooper, W.W., 1985. Preface to topics in data envelopment analysis. Annals of Operation Research 2, 59–94. Charnes, A., Cooper, W.W., Seiford, L.M., 1994a. Extension to DEA models. In: Charnes, A., Cooper, W.W., Lewin, A.Y., Seiford, L.M. (Eds.), Data Envelopment Analysis: Theory, Methodology and Applications. Kluwer Academic Publishers, Norwell. Charnes, A., Cooper, W.W., Lewin, A.Y., Seiford, L.M., 1994b. Data Envelopment Analysis: Theory, Methodology, and Application. Kluwer Academic Publishers, Norwell. Chien, T., Hu, J.L., 2007. Renewable energy and macroeconomic efficiency of OECD and non-OECD economies. Energy Policy 35, 3606–3615. Coelli, T.J., 1996. A Guide to Deap Version 2.1: A Data Envelopment Analysis (Computer) Program. Department of Econometric, University of New England, Armidale, Australia. Edvardsen, D.F., Fbrsund, F.R., 2003. International benchmarking of electricity distribution utilities. Resource and Energy Economics 25, 353–371. Fisher-Vanden, Karen, Jefferson, Gary H., Liu, Hongmei, Tao, Quan, 2003. What is driving China’s decline in energy intensity. Resource and Energy Economics 26, 77–97. Forsund, F.R., Kittelsen, S.A.C., 1998. Productivity development of Norwegian electricity distribution utilities. Resource and Energy Economics 20, 207–224. Garbaccio, R.F., Ho, M.S., Jorgenson, D.W., 1999. Why has the energy–output ratio fallen in China. Energy Journal 20 (3), 63–91. Halkos, G.E., Tzeremes, N.G., 2009. Exploring the existence of Kuznets curve in countries’ environmental efficiency using DEA window analysis. Ecological Economics 68, 2168–2176. Hartman, T.E., Storbeck, J.E., 1996. Input congestion in loan operations. International Journal of Production Economics 46–47, 413–421. Honma, S., Hu, J.L., 2008. Total-factor energy efficiency of regions in Japan. Energy Policy 36, 821–833. Honma, S., Hu, J.L., 2009. Total-factor energy productivity growth of regions in Japan. Energy Policy 37, 3941–3950. Honore´, B.E., 1993. Orthogonality conditions for Tobit models with fixed effects and lagged dependent variables. Journal of Econometrics 59, 35–61. Hu, J.L., Wang, S.C., 2006. Total-factor energy efficiency of regions in china. Energy Policy 34, 3206–3217.
Hu, J.L., Kao, C.H., 2007. Efficient energy-saving targets for APEC economies. Energy Policy 35, 373–382. Hu, J.L., Chang, T.P., 2009. Total-factor energy productivity growth of regions in China. National Chiao Tung University, Taipei, Working paper. Huang, J.P., 1993. Industrial energy use and structural change: a case study of the People’s Republic of China. Energy Economics 15, 131–136. Iyer, K.G., Rambladi, A.N., Tang, K.K., 2004. Measuring spillovers from alternative forms of foreign investment. Working paper series from Centre for Efficiency and Productivity Analysis, University of Queensland. Liao, H., Fan, Y., Wei, Y.M., 2007. What induced China’s energy intensity to fluctuate: 1997–2006? Energy Policy 35 4640–4649. Lin, X., Polenske, K.R., 1995. Input–output anatomy of China’s energy use changes in the 1980s. Economic Systems Research 7 (1), 67–84. Lin, J., Zhou, N., Levine, M., Fridley, D., 2008. Taking out 1 billion tons of CO2: the magic of China’s 11th Five-Year Plan. Energy Policy 36, 954–970. Ma, C., Stern, D.I., 2008. China’s changing energy intensity trend: a decomposition analysis. Energy Economics 30, 1037–1053. Penn World Table 5.6, 1998. Accessible at /http://datacentre.chass.utoronto.ca/ pwt56/S. Sadjadi, S.J., Omrani, H., 2008. Data envelopment analysis with uncertain data: an application for Iranian electricity distribution companies. Energy Policy 36, 4247–4254. Sinton, J.E., Levine, M.D., 1994. Changing energy intensity in Chinese industry: the relative importance of structural shift and intensity change. Energy Policy 22, 239–255. Steenhof, P., 2006. Decomposition of electricity demand in China’s industrial sector. Energy Economics 28, 370–384. Tulkens, H., Vanden Eeckaut, P., 1995. Non-parametric efficiency, progress and regress measures for panel data: methodological aspects. European Journal of Operational Research 80, 474–499. Webb, R.M., 2003. Levels of efficiency in UK retail banks: a DEA window analysis. International Journal of the Economics of Business 10, 305–322. Wei, Y.M., Liao, H., Fan, Y., 2007. An empirical analysis of energy efficiency in China’s iron and steel sector. Energy 32, 2262–2270. Weidou, N., Johansson, T.B., 2004. Energy for sustainable development in China. Energy Policy 32, 1225–1229. World Bank, 2005. World Development Indicators. Accessible at /http://www. worldbank.orgS. Wu, Y., 2004. Openness, productivity and growth in the APEC economies. Empirical Economics 29, 593–604. Yue, P., 1992. Data envelopment analysis and commercial bank performance: a primer with applications to Missouri banks. Federal Reserve Bank of St. Louis Review 74, 31–45. Zhang, Z.X., 2003. Why did the energy intensity fall in China’s industry sector in the 1990s? The relative importance of structural change and intensity change. Energy Economics 25, 625–638.
Contents lists available at ScienceDirect
Energy Policy journal homepage: www.elsevier.com/locate/enpol
Total-factor energy efficiency in developing countries Xing-Ping Zhang n, Xiao-Mei Cheng, Jia-Hai Yuan, Xiao-Jun Gao School of Economics and Management, North China Electric Power University, Beijing 102206, China
a r t i c l e i n f o
a b s t r a c t
Article history: Received 5 May 2010 Accepted 27 October 2010 Available online 13 November 2010
This paper uses a total-factor framework to investigate energy efficiency in 23 developing countries during the period of 1980–2005. We explore the total-factor energy efficiency and change trends by applying data envelopment analysis (DEA) window, which is capable of measuring efficiency in crosssectional and time-varying data. The empirical results indicate that Botswana, Mexico and Panama perform the best in terms of energy efficiency, whereas Kenya, Sri Lanka, Syria and the Philippines perform the worst during the entire research period. Seven countries show little change in energy efficiency over time. Eleven countries experienced continuous decreases in energy efficiency. Among five countries witnessing continuous increase in total-factor energy efficiency, China experienced the most rapid rise. Practice in China indicates that effective energy policies play a crucial role in improving energy efficiency. Tobit regression analysis indicates that a U-shaped relationship exists between total-factor energy efficiency and income per capita. & 2010 Elsevier Ltd. All rights reserved.
Keywords: DEA window analysis Total-factor energy efficiency Tobit model
1. Introduction Global warming is one of the world’s most important environmental problems. The problem is largely attributable to the greenhouse gas carbon dioxide, which is released by the burning of fossil fuels. Because of this, a growing body of research has focused on improving energy efficiency, which is a crucial approach to alleviating global warming. During the past several decades, some appropriate methods have been developed to monitor energy-efficiency trends and compare energy efficiency performance across countries. These methods are generally classified as parametric and non-parametric methods (Sadjadi and Omrani, 2008). Parametric methods such as stochastic frontier analysis estimate a cost or production function. Therefore, deviations in the function form affect the results of such models. In contrast, it is not necessary to estimate the cost or production function when using non-parametric methods. Data envelopment analysis (DEA) is a non-parametric method that is capable of handling multiple inputs and multiple outputs. Energy intensity and energy efficiency are the two well-known energy-efficiency indicators that are commonly used in macrolevel policy analysis. Energy intensity is defined as the energy consumption divided by the economic output, and energy efficiency is the reciprocal of energy intensity. These traditional energy-efficiency indicators take energy consumption into account as a single input that produces an economic output; therefore, some other key inputs are ignored, such as capital and labor. Energy consumption must be combined with other inputs to produce an
n
Corresponding author. Tel.: + 86 1051963541; fax: +86 1080796904. E-mail address: [email protected] (X.-P. Zhang).
0301-4215/$ - see front matter & 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.enpol.2010.10.037
economic output, and substitution effects exist between energy and other input factors (e.g., labor and capital stock). If energy consumption is evaluated in terms of partial-factor energy efficiency, the result is a misleading estimate (Hu and Wang, 2006; Honma and Hu, 2009). Boyd and Pang (2000) indicated that energyefficiency improvement relies on total-factor productivity improvement. To overcome the disadvantage of partial-factor energy efficiency, an increasing number of researchers have devoted themselves to analyzing total-factor energy efficiency using DEA. There are essentially two research strands in the literature analyzing total-factor energy efficiency using DEA. The first strand focuses on measuring total-factor productivity change based on the DEA–Malmquist index, which was first introduced by Caves et al. (1982). For example, Forsund and Kittelsen (1998) applied DEA efficiency scores to calculate the Malmquist productivity index in Norwegian electricity distribution companies. Edvardsen and Fbrsund (2003) applied an input-oriented DEA model and the Malmquist productivity index to analyze the performance of 122 electricity distributors in Denmark, Finland, Norway, Sweden and the Netherlands. Wei et al. (2007) applied this approach to investigate changes in the energy efficiency of China’s iron and steel sectors. The index of total-factor productivity takes only radial adjustment into account and disregards the non-radial slack; it is therefore unable to measure single-factor efficiency under a totalfactor framework (Honma and Hu, 2009). The second strand utilizes the index of total-factor energy efficiency (TFEE) first proposed by Hu and Wang (2006). TFEE is defined as the target energy input divided by the actual energy input. Taking non-radial slack into account, the TFEE is capable of measuring single-factor efficiency in a total-factor framework. Following Hu and Wang (2006), Honma and Hu (2008) measured
X.-P. Zhang et al. / Energy Policy 39 (2011) 644–650
the TFEE of 47 regions in Japan for the period 1993–2003. Integrating the concept of TFEE with the Malmquist productivity index, Hu and Chang (2009) proposed the total-factor energy productivity index (TFEPI) to investigate energy productivity changes in regions of China. Honma and Hu (2009) extended their previous work (Honma and Hu, 2008) by applying TFEPI. Following Hu and Wang (2006) and Honma and Hu (2008), we use DEA theory to investigate the total-factor energy efficiency of developing countries. This paper extends the contributions of these earlier studies in three ways. First, this study investigates the totalfactor energy efficiency of 23 developing countries. The Kyoto Protocol is severely criticized for not including emission reduction obligations for developing countries because many developing countries have become major carbon dioxide emitters. For example, China is the second largest energy-related CO2 emitter and India ranks fourth in the world. Therefore, the choice to study developing countries is motivated by the rising importance of their contribution to global warming. Secondly, we use the DEA window analysis introduced by Charnes and Cooper (1985) for the first time to develop total-factor energy-efficiency measures for the 23 developing countries during the period of 1980–2005. This approach can indicate efficiency trends over a specified period of time while simultaneously examining stability and other properties of the efficiency evaluations within the specified windows (Hartman and Storbeck, 1996; Webb, 2003). Third, we use dynamic Tobit model to investigate the relationship between total-factor energy efficiency and income for 23 developing countries over the period 1980–2005. The remainder of this paper is organized as follows: The next section describes the methods used in the study; Section 3 presents the data; Section 4 presents the empirical results; and Section 5 concludes the paper.
2. Method 2.1. Total-factor energy efficiency based on DEA Hu and Wang (2006) proposed the TFEE, which is defined as the target energy input divided by the actual energy input, and also utilized what is known as the constant returns to scale (CRS) DEA model. It is universally known that the assumption of the CRS model is appropriate only when all decision-making units (DMUs) are operating at an optimal scale. However, some factors, such as imperfect competition and constraints on finance, may cause a DMU not to operate at an optimal scale. If it is likely that the size of the DMUs under investigation will influence their ability to create outputs efficiently, then the assumption of CRS is inappropriate (Halkos and Tzeremes, 2009). Banker et al. (1984) suggested an extension of the CRS model to account for variable returns to scale (VRS) situations. The less restrictive VRS frontier allows the best practice level of outputs to inputs to vary with the size of the countries (Halkos and Tzeremes, 2009). Because the 23 developing countries in this study have different sizes, we use the VRS model. The VRS model could be obtained by adding the convexity constraint based on the CRS model (Charnes et al., 1978). The model employs the following mathematical notation: For each of N DMUs, there are K inputs and M outputs. For the ith DMU, the inputs and outputs are represented by the column vectors xi and yi, respectively. The input-oriented VRS model solves the following linear programming problem for DMUto : Min s:t:
y N X j¼1
N X j¼1 N X
lj xj þ s ¼ yxo lj ¼ 1
j¼1
lj Z 0, j ¼ 1, 2,. . .,N
ð1Þ
where y is a scalar with 0r y r1, l is a N 1 vector of constants that form a convex combination of observed inputs and outputs and s represents the non-radial slack. 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). The value of y represents the technically efficient score for the ith DMU. (1 y)xi is called radial adjustments. The ith DMU is the most efficient point on the frontier and is technically efficient if y ¼1 and the slack equals zero; if y ¼1 and the slack is larger than zero, the ith DMU is weakly technically efficient; y o1 indicates that the ith DMU is technically inefficient. The sum of the radial and non-radial adjustments is called the total adjustments, which can be reduced to reach optimal technical efficiency without decreasing the output levels. The total adjustments for the technically efficient DMUs equal zero, but are larger than zero for the other DMUs. The target energy input (TEI) in this study is therefore actual energy input (AEI) minus the total adjustments (TA), which represents a practical minimum level of energy input to be taken as a target to perform at the optimal energy consumption efficiency. Hence, the TFEE index of DMU i at time t can be measured as (Hu and Wang, 2006) TFEEði,tÞ ¼
TEIði,tÞ AEIði,tÞTAði,tÞ ¼ AEIði,tÞ AEIði,tÞ
ð2Þ
The total adjustments are not less than zero according to model (1). Total adjustments of zero indicate that the actual energy input is indeed the target energy input, so TFEE is unity and the energy is used at the optimal efficiency. TFEE is lower than unity if the total adjustments are larger than zero. This implies there is redundant energy input, which should be reduced without decreasing the output. Therefore, the index of TFEE is always between zero and unity. The greater the value of TFEE, the more efficient the energy consumed.
2.2. Total-factor energy efficiency based on DEA window analysis DEA window analysis, which was introduced by Charnes and Cooper (1985), is a variation of the traditional DEA that can handle cross-sectional and time-varying data to allow for dynamic effects. DEA window analysis operates on the principle of moving averages (Charnes et al., 1994a; Yue, 1992) and establishes efficiency measures by treating each DMU in different years as a separate unit. The performance of a DMU in a period can be contrasted with its own performance in other periods as well as to the performance of other DMUs (Asmild et al., 2004). Therefore, DEA window analysis can explore the evolution of performance through a sequence of overlapping windows. Moreover, the number of data points is increased several times over in DEA window analysis, which is very useful when dealing with small sample sizes. A brief DEA window analysis review is presented here. A window with N w observations is denoted starting at time t(1rt rT) with window width w (1rwrT t). The matrix of inputs for this window is 1 2 N 1 2 N Xtw ¼ ðx1t ,x2t ,. . .,xN t ,xt þ 1 ,xt þ 1 ,. . .,xt þ 1 ,. . .,xt þ w ,xt þ w ,. . .,xt þ w Þ
and the matrix of outputs is given by
lj yj s þ ¼ yo
645
1 2 N 1 2 N Ytw ¼ ðy1t ,y2t ,. . .,yN t ,yt þ 1 ,yt þ 1 ,. . .,yt þ 1 ,. . .,yt þ w ,yt þ w ,. . .,yt þ w Þ
646
X.-P. Zhang et al. / Energy Policy 39 (2011) 644–650
The input oriented DEA window problem for DMUto is given by solving the following linear program:
worker. The labor force and It are available from the World Development Indicators database (World Bank, 2005).
min y s:t:
4. Empirical results
yxot Xtw l Z0 N X
lj ¼ 1
Ytw lyot Z 0
ð3Þ
lj Z0, j ¼ 1,2,. . .,N w
j¼1
A ‘‘window’’ is constructed by the number of time periods (years in our study) under analysis, which conceived of in an intertemporal manner, according to Tulkens and Vanden Eeckaut (1995). This study includes 23 countries (N ¼23) for the time period of 1980–2005(T¼26). It should be noted that because all units within a given window are measured against each other, this approach implicitly assumes that there are no technical changes within each window. This is a general problem in DEA window analysis (Asmild et al., 2004). A narrow window length therefore must be used to reduce this problem. Charnes et al. (1994b) found that a window length of three or four tended to yield the best balance of informativeness and stability of the efficiency scores. Following Halkos and Tzeremes (2009), we chose a narrow window length of three (w¼3) to create credible results. Thus, the first three years of 1980, 1981 and 1982 form the first window. The window then moves on a one-year period by dropping the original year and adding a new year; therefore, the next three years of 1981, 1982 and 1983 construct the second window. The process continues until the last window, which contains the last three years (2003, 2004 and 2005 in this case), is constructed. In doing so, 24 windows are performed for each country and the number of DMUs (countries) in each window increases from 23 to 69(N w¼23 3). The TFEEs of 23 developing countries in each window could be obtained using DEA window analysis. For each country, each year has three values on TFEE, with the exception of 1980 and 2005, which has only one value, and 1982 and 2004, which has two values. We then calculate the average result of the TFEEs of each country in the same year, which creates a new TFEE result for the 23 developing countries analyzed in our study.
3. Data and variables We use annual data on the labor force, energy consumption (kt of oil equivalent) and capital stock as the three input variables, while gross domestic product (GDP) is the single output in this study. Due to the unavailability of data on some developing countries, we select 23 developing countries to analyze in this paper. Except capital stock, all data are extracted from the World Development Indicators database (World Bank, 2005) for the period 1980–2005. To standardize the data, capital stock and real GDP are measured in purchasing power parity (constant 2005, international $). To the best of our knowledge, the data on capital stock from 1991 to 2005 could not be obtained from any statistical yearbook or database. Following Hu and Kao (2007) and Chien and Hu (2007), we apply the following perpetual inventory method to calculate the capital stock: K t ¼ It þð1dÞKt1 where Kt is the gross capital stock in current year; Kt 1 is the gross capital stock in the previous year; d represents the depreciation rate of capital stock; and It is the gross fixed capital formation in the current year. d is set to 6% in accordance with the suggestions of many relevant studies such as Iyer et al. (2004) and Wu (2004). The data on capital stock per worker in 1980 is available from Penn World Table 5.6 (1998). Hence, the capital stock in 1980 is obtained by multiplying the labor force total by capital stock per
4.1. Total-factor energy efficiency DEA window analysis is applied in this case, and Table 1 shows energy efficiency for China.1 We can explore the evolution of energy performance through a sequence of overlapping windows. Viewing the column data of Table 1, we can examine the stability of the total-factor energy efficiency of China across the different datasets. The row data enable us to test the trends across the same dataset. According to the last row of Table 1, it is obviously seen that China has significantly improved its TFEE between 1980 and 2005. To explore the trend changes of TFEEs for 23 developing countries, we show the overall TFEE scores of all 23 developing countries in Table 2. The countries with TFEE of unity constitute the efficiency frontier of energy consumption among 23 countries in the same year, which means that energy is used at the optimal level in these countries. Table 2 indicates that the efficiency frontier is different in different year. For example, six countries (Botswana, Honduras, Morocco, Panama, Paraguay and Peru) constitute the efficiency frontier of energy consumption in 1980, while another five countries (Botswana, China, Indian, Mexico and Thailand) constitute the efficiency frontier of energy consumption in 2005. The total-factor energy efficiency of all 23 developing countries displays three evolving trends during the research period. First, the TFEEs of seven countries (Argentina, Bolivia, Botswana, Chile, Kenya, Mexico and Panama) slightly fluctuate over time. It is noteworthy that the average TFEEs of three countries (Botswana, Mexico and Panama) are on the verge of unity, at 0.996, 0.990 and 0.960, respectively. Furthermore, Botswana is on the efficiency frontier of energy consumption in 18 out of 26 years, Mexico is on the efficiency frontier in 10 out of 26 years, and Panama is also on the efficiency frontier in the initial period. Therefore, these three countries are the most energy-efficient economies among 23 developing countries during the whole research period. However, Kenya is particularly noteworthy because its TFEEs are lower than 0.41 during the entire research period, and it has the lowest average TFEE score (0.329) in the 23 developing countries. Secondly, eleven countries display a downward trend in totalfactor energy efficiency: Iran ( 2.02%), Venezuela ( 1.90%), Morocco ( 1.86%), Honduras ( 1.49%), Syria ( 1.37%), Paraguay ( 1.36%), Guatemala ( 1.24%), Philippine ( 1.13%), Ecuador ( 0.88%), Peru ( 0.85%) and Dominican ( 0.28%).2 Therefore, the total-factor energy efficiency of these countries worsened between 1980 and 2005. It is worth noting that Honduras, Morocco, Paraguay and Peru are at the efficient production frontier (TFEE ¼1) in the first few years. However, their TFEEs become smaller and smaller in the later period, that is, they are distancing themselves from the efficient frontier over time. Philippine and Syria are concerned especially for their very low TFEEs in entire research period. Finally, there is a continuous increasing trend in the total-factor energy efficiency of five countries: China (8.64%), Zambia (4.28%), India (3.73%), Thailand (2.77%) and Sri Lanka (1.80%). Special attention should be paid to the first four countries (China, India, Thailand and Zambia): their TFEEs are very low (less than 0.5) in the initial period, but they show a significant improvement from the beginning of the 1990s. It can be noteworthy that China and India 1
The results of the other 22 developing countries are available upon request. The percentage in parenthesis is the annual average growth rate of TFEE between 1980 and 2005. 2
X.-P. Zhang et al. / Energy Policy 39 (2011) 644–650
647
Table 1 A three-year window analysis of total-factor energy efficiency in China. Windows/years
1980
1981
1982
Window Window Window Window Window Window Window Window Window Window Window Window Window Average
0.116
0.123 0.123
0.13 0.13 0.136
1 2 3 4 5 6 7 8 9 10 11 12 13
1983
0.139 0.145 0.147
1984
0.158 0.159 0.159
1985
1 0.621 0.451
1986
1 0.688 0.56
1987
1 0.782 0.703
1988
1 0.919 0.875
1989
1 0.963 0.762
1990
1 0.795 0.612
1991
1 0.772 0.64
1992
0.116
0.123
0.132
0.143
0.159
0.69
0.749
0.828
0.931
0.908
0.802
0.804
1 0.825 0.695 0.84
Windows/years
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Window Window Window Window Window Window Window Window Window Window Window Window Window Average
1 0.845 0.77
1 0.974 0.986 0.987
1 0.964 0.982
1 1
12 13 14 15 16 17 18 19 20 21 22 23 24
0.872
0.977 0.907 0.82
0.901
0.988 0.901 0.795
0.895
1 0.884 0.795
0.893
1 0.9 0.818
0.906
1 0.909 0.822
0.91
were on the efficiency frontier in 2005, Thailand constituted the efficiency frontier in the last few years, and Zambia was on the efficiency frontier in the early 1990s. Although Sri Lanka improved TFEE between 1980 and 2005, its energy efficiency is very low.
4.2. Experiences from China Among 23 developing countries, China experienced the most rapid rise in TFEE from the onset of economic reform in the late 1970s. The past three decades has witnessed the rapid economic growth in China. At the same time, energy consumption has largely increased. China has become the world’s largest consumer for coal since 1986, and the second largest consumer for oil since 2002 (BP, 2009). What’s more, China’s demand for energy has outpaced its energy production since 2001. Environmental problem, especially ‘‘climate change’’, has become the focus of the world since the UN Conference on Environment and Development held in 1992. In order to alleviate energy shortage and climate change, Chinese government has issued and carried out a series of energy policies to improve energy efficiency since late 1970s. Table 3 reports partial key energy policies implemented by China. As early as 1992, China issued ‘‘Ten solutions for environment and development in China’’, in which enhancing energy utilization efficiency is one of the ten solutions. Then in the ninth Five-YearPlan (1996–2000), Government of China proposed an energy conservation target of 5% annually and reduction for principal pollutant from 1996 to 2000, which is the first energy saving and emission reducing goal among the developing countries. Since the mid-1990s, efficiency improvements have been particularly marked in energyintensive industries such as metallurgy, cement, paper, textiles, oil and coal processing and electrical power generation (Steenhof, 2006; Liao et al., 2007). As Weidou and Johansson (2004) pointed out: China has also met its soaring needs for energy services while at
1 0.906 0.818
0.908
1 0.901 0.875
0.925
1 0.971 0.976
0.982
1 0.985 0.981 0.989
the same time has improved the efficiency of its energy use by a factor of nearly three. Lin et al. (2008) stated that China’s experience from 1980 to 2000 was an exception because in the early stage of economic development, industrialization and urbanization tend to lead to extensive infrastructure and housing development, which will consume much energy and material. Government of China officially takes sustainable development as the crucial state policy in the 16th Conference of Communist Party of China (CPC) held in 2002. The Medium and Long-Term Plan for Energy Conservation, approved by Chinese government in 2004, indicated that the GDP energy intensity should decrease by 2.2% annually until 2010. Medium and Long-Term Plan for Energy Conservation is a milestone in China’s energy policy: financial incentive and market mechanism came into operation from then on (China paid much attention to the measures of command-and-control in energy policy before). Although China’s energy efficiency is very low compared with developed countries, China has been devoting to improve energy efficiency and has achieved a great success of annual energy saving by 3.9% from 1980 to 2005. Furthermore, the Eleventh Five-Year Plan for Energy Development, approved by National Development and Reform Commission (NDRC) of China in 2007, set an ambitious goal of reducing energy intensity by 20% between 2006 and 2010. At the end of 2009, 15.6% reduction in energy intensity as of 2005 has been achieved. Practice in China shows that effective energy policies play an important role in forcing all levels of government and energy-using units to take active measures to improving energy efficiency, such as, improving technological level, restructuring industries and products and so on. Numerous studies have investigated the components driving the change of energy intensity by using decomposition analysis (Huang, 1993; Sinton and Levine, 1994; Lin and Polenske, 1995; Garbaccio et al., 1999; Zhang, 2003; FisherVanden et al., 2003; Liao et al., 2007). Most previous studies indicated that the decline of energy intensity can mainly be attributed to technological changes, while there is disagreement
648
X.-P. Zhang et al. / Energy Policy 39 (2011) 644–650
Table 2 Total-factor energy efficiency of 23 developing countries. Country
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
Argentina Bolivia Botswana Chile China Dominican Ecuador Guatemala Honduras India Iran Kenya Mexico Morocco Panama Paraguay Peru Philippines Sri Lanka Syria Thailand Venezuela Zambia
0.919 0.722 1.000 0.823 0.116 0.794 0.928 0.916 1.000 0.386 0.939 0.243 0.978 1.000 1.000 1.000 1.000 0.741 0.409 0.652 0.491 0.954 0.313
0.850 0.698 1.000 0.760 0.123 0.803 0.924 0.856 1.000 0.389 0.775 0.247 1.000 0.884 1.000 0.922 1.000 0.722 0.431 0.636 0.494 0.969 0.323
0.812 0.661 0.993 0.712 0.132 0.790 0.921 0.762 0.973 0.393 0.787 0.257 0.967 0.910 0.997 0.780 0.986 0.715 0.431 0.591 0.507 0.948 0.312
0.839 0.656 1.000 0.676 0.143 0.774 0.933 0.747 0.999 0.411 0.992 0.272 0.989 0.869 0.956 0.711 0.949 0.688 0.454 0.577 0.587 0.920 0.303
0.848 0.672 1.000 0.710 0.159 0.746 0.915 0.733 1.000 0.415 0.887 0.278 1.000 0.867 0.983 0.695 0.977 0.692 0.457 0.510 0.561 0.823 0.294
0.811 0.673 0.996 0.747 0.690 0.778 0.910 0.719 1.000 0.422 0.855 0.290 1.000 0.851 1.000 0.690 0.963 0.600 0.464 0.537 0.520 0.711 0.287
0.849 0.671 0.981 0.754 0.749 0.766 0.915 0.727 1.000 0.715 0.773 0.303 0.991 0.851 1.000 0.660 0.986 0.605 0.455 0.496 0.500 0.690 0.283
0.859 0.699 1.000 0.772 0.828 0.751 0.861 0.713 1.000 0.741 0.760 0.318 1.000 0.775 0.968 0.623 1.000 0.599 0.439 0.477 0.496 0.812 0.279
0.830 0.701 1.000 0.766 0.931 0.763 0.912 0.712 1.000 0.848 0.714 0.333 0.994 0.806 0.909 0.621 0.950 0.602 0.438 0.519 0.523 0.963 0.488
0.776 0.704 1.000 0.795 0.908 0.794 0.957 0.713 1.000 0.891 0.707 0.341 0.988 0.753 0.945 0.617 0.907 0.593 0.438 0.460 0.505 0.897 0.722
0.753 0.744 1.000 0.796 0.802 0.747 0.981 0.729 0.691 0.920 0.773 0.361 0.997 0.747 0.995 0.647 0.863 0.575 0.475 0.450 1.000 0.917 0.996
0.803 0.781 1.000 0.832 0.804 0.756 0.981 0.748 0.728 0.850 0.843 0.373 1.000 0.763 1.000 0.645 0.856 0.566 0.509 0.445 1.000 0.976 1.000
0.853 0.794 0.962 0.853 0.840 0.748 0.952 0.767 0.741 0.841 0.854 0.377 0.994 0.695 0.992 0.658 0.816 0.544 0.540 0.481 0.999 0.989 1.000
0.890 0.790 0.990 0.875 0.872 0.772 0.933 0.783 0.754 0.832 0.814 0.377 0.991 0.667 1.000 0.652 0.794 0.535 0.560 0.493 1.000 0.986 1.000
Country
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Average
Argentina Bolivia Botswana Chile China Dominican Ecuador Guatemala Honduras India Iran Kenya Mexico Morocco Panama Paraguay Peru Philippines Sri Lanka Syria Thailand Venezuela Zambia
0.886 0.786 1.000 0.868 0.901 0.716 0.912 0.783 0.726 0.860 0.700 0.384 1.000 0.706 0.991 0.637 0.843 0.527 0.600 0.514 1.000 0.768 1.000
0.850 0.762 0.991 0.920 0.895 0.725 0.860 0.784 0.711 0.892 0.693 0.395 0.958 0.661 0.990 0.628 0.853 0.514 0.624 0.507 1.000 0.842 0.989
0.870 0.734 1.000 0.924 0.893 0.744 0.800 0.788 0.722 0.909 0.675 0.406 0.972 0.717 0.969 0.615 0.842 0.517 0.609 0.530 1.000 0.677 0.999
0.893 0.727 1.000 0.908 0.906 0.728 0.809 0.780 0.700 0.860 0.690 0.401 0.991 0.687 0.974 0.591 0.889 0.496 0.632 0.514 0.985 0.708 0.978
0.861 0.695 0.991 0.864 0.910 0.712 0.800 0.757 0.677 0.847 0.568 0.399 0.970 0.716 0.930 0.595 0.840 0.495 0.645 0.472 1.000 0.541 0.971
0.805 0.691 1.000 0.740 0.908 0.719 0.755 0.708 0.688 0.831 0.542 0.402 0.961 0.676 0.934 0.610 0.805 0.484 0.630 0.422 0.997 0.522 0.977
0.774 0.676 1.000 0.775 0.925 0.704 0.716 0.697 0.697 0.813 0.522 0.401 1.000 0.660 0.920 0.625 0.807 0.482 0.618 0.421 1.000 0.514 0.964
0.769 0.717 1.000 0.820 0.982 0.697 0.713 0.687 0.688 0.850 0.518 0.304 0.990 0.661 0.848 0.641 0.799 0.489 0.599 0.437 0.998 0.520 0.981
0.703 0.741 0.992 0.812 0.989 0.689 0.709 0.688 0.695 0.863 0.538 0.268 1.000 0.643 0.893 0.663 0.816 0.498 0.601 0.424 1.000 0.480 0.985
0.734 0.739 1.000 0.859 0.987 0.680 0.688 0.694 0.699 0.920 0.552 0.268 0.997 0.664 0.888 0.684 0.833 0.517 0.611 0.437 1.000 0.488 0.972
0.771 0.701 1.000 0.844 0.982 0.694 0.713 0.684 0.683 0.956 0.542 0.269 1.000 0.645 0.927 0.696 0.810 0.527 0.592 0.451 1.000 0.547 0.945
0.834 0.726 1.000 0.883 1.000 0.739 0.737 0.663 0.677 1.000 0.552 0.280 1.000 0.613 0.955 0.701 0.800 0.552 0.650 0.456 1.000 0.580 0.930
0.825 0.718 0.996 0.811 0.745 0.743 0.855 0.744 0.817 0.756 0.714 0.329 0.990 0.750 0.960 0.677 0.884 0.572 0.535 0.497 0.814 0.759 0.742
on the role of structural change (a shift in the mix of industries). Many studies found that structural change has played a minor role in reducing energy intensity. However, a few studies found that structural change actually increased energy intensity in a certain time period (Garbaccio et al., 1999;Fisher-Vanden et al., 2003). Ma and Stern (2008) confirmed that technological change played a dominant role in decreasing energy intensity over the period of 1980–2003, while structural effect plays a minor role. Furthermore, Ma and Stern (2008) indicated that inter-fuel substitution was found to contribute little to the changes in energy intensity. It is noteworthy that China has paid much attention to restructuring industries and inter-fuel substitution recently. For example, Renewable Energy Law and Law on Promoting the Circular Economy were issued in 2005 and 2009, respectively. Medium and Long Term Plan for Renewable Energy Development was released by NDRC in 2007, in which an ambitious target of renewable energy accounting for about 10% of the total energy consumption by 2010 and about 15% by 2020 is set. Furthermore, NDRC took active ways
to accelerate the development of energy management contract (EMC) in 2010, which is a new management model for China. 4.3. Relationship between TFEE and income Based on gross national income per capita, the World Bank classified the countries into four types: high income countries, upper middle income countries, lower middle income countries and low income countries. Because of this, we use the gross national income per capita (GNIPC) as a proxy for income level. The dependent variable TFEE is a fractional variable bounded between 0% and 100%, so Tobit regression with both truncation points is applied in this paper. To address the potential autocorrelation, we apply the following dynamic Tobit model3: TFEEit ¼ ai þ b1 TFEEi,t-1 þ b2 GNIPCit þ b3 GNIPCit2 þ eit 3
See Honore´ (1993) for a discussion of the dynamic Tobit model.
ð4Þ
X.-P. Zhang et al. / Energy Policy 39 (2011) 644–650
649
Table 3 Main energy policies in China from 1980. Year
Policies
Policymakers
1980 1982 1984 1986 1992
On Strengthening Energy Conservation Outline of Energy-Saving Technology Policy Interim Provisions for Some Issues of Comprehensive Utilization of Resources Regulations for Saving Energy Ten solutions for environment and development in China (Article IV addresses energy efficiency and energy structure) Electricity Law The ninth Five-Year-Plan Energy Conservation Law Energy Management Measures for Key Energy-Using Units Management Measures for Energy-Saving Product Authentication Civil Construction of Energy-Saving Regulations Rules for Implementing Energy Conservation Law in Transportation Industry Design Standards for Building Energy Conservation in Hot Summer-Winter Cold Districts The Tenth Five-Year Plan for Building Energy Conservation Medium and Long-Term Plan for Energy Conservation Renewable Energy Law Notification on Near-term Key Work for Building Conservation Society Decision on Strengthening Energy Conservation Eleventh Five-Year Plan for Energy Development Medium and Long Term Plan for Renewable Energy Development Regulations on Energy Conservation of Public Institutions Law on Promoting the Circular Economy On Strengthening Energy Management Contract and Promoting Energy Saving Service Industry
National Economy Commisssion (NEC) National Development and Reform Commission (NDRC) NEC State Council Chinese Foreign Ministry, and State Environmental Protection Administration National People’s Congress State Council National People’s Congress NEC Energy-saving product certification Committee of China National Ministry of Construction Ministry of Communications Ministry of Housing and Urban-Rural Development (MHURD)
1994 1996 1997 1999 2000 2001 2002 2004 2005 2006 2007 2008 2009 2010
Table 4 Tobit estimation results. Variable
Coefficient
Std. error
Prob.
TFEEt 1 GNIPC GNIPC squared
1.0442 1.1578 0.0701
0.0191 0.1677 0.0102
0.0000 0.0000 0.0000
where t ¼ 1,. . .,T time period and i ¼ 1,. . .,N members of the panel; ai is nation-specific effects; the error mit is assumed to be i.i.d. across i and over t. GNIPC is transformed into its logarithm, and the logarithm of TFEE is given by ln(100 TFEE). F-test and Hausman specification test results reject the random model at 5% significance level, therefore, we estimate Eq. (4) with fixed effects by using the maximum likelihood estimate. Table 4 reports the regression results. The positive coefficient of GNIPC squared suggests that a U-shaped relationship exists between TFEE and GNIPC for the sample studied in this paper. That is, energy efficiency worsens with the increase in income per capita due to growth of industries until a certain point. Then, it rises after income per capita reach the point.4
MHURD and NDRC NDRC National People’s Congress State Council State Council NDRC NDRC NDRC National People’s Congress NDRC
eleven countries show a downward trend in TFEE, particularly, Honduras, Morocco and Paraguay experienced a sharp decrease in energy efficiency. It is important to note that the TFEE index measures the energy efficiency of 23 developing countries based on their own frontier. Therefore, the energy-inefficient countries could take feasible improvement measures to reduce energy input without hindering economic output. Taking China as the example because of its great success in improving energy efficiency, we briefly review the energy policies carried out by China and indicate that effective energy policies play a crucial role in improving energy efficiency. The Tobit estimation results suggest there is a U-shaped relationship between TFEE and income per capita, which means the energy efficiency first decreases with the income per capita, and then increases after a certain level. The index of TFEE is capable of measuring energy efficiency in a total-factor framework. However, as in all DEA analyses, the TFEE index for a single country are generally affected by other countries in the sample. The reason is that the production frontier constructed depends on the sample. The results therefore should be interpreted with caution.
Acknowledgments 5. Conclusions This study measures energy efficiency in a total factor production framework using DEA window analysis to detect changes in efficiency over time. For this purpose, a sample of 23 developing countries for the period 1980–2005 is used. The total-factor energy efficiency and change trends of 23 developing countries are investigated. Botswana, Mexico and Panama perform the best in term of TFEE, while Kenya, Philippine, Sri Lanka and Syria perform the worst in energy efficiency during the entire period. Five countries experienced a continual increase in total-factor energy efficiency, and China experienced the most rapid rise in TFEE. While 4
We thank an anonymous reviewer for pointing this issue.
The authors thank the two anonymous referees and the editor of this journal for their valuable suggestions and helpful comments. The authors gratefully acknowledge the financial support the Natural Science Foundation of Beijing (9092013), the Human Resource Development Project of the Beijing City Committee (20081D1600900358), Beijing Planning Project of Philosophy and Social Science (10BaJG371) and the Fundamental Research Funds of the Central Universities. References Asmild, M., Paradi, C.V., Aggarwall, V., Schaffnit, C., 2004. Combining DEA window analysis with the Malmquist index approach in a study of the Canadian banking industry. Journal of Productivity Analysis 21, 67–89.
650
X.-P. Zhang et al. / Energy Policy 39 (2011) 644–650
Banker, R.D., Charnes, A., Cooper, W.W., 1984. Models for estimation of 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. BP, 2009. Statistical Review of World Energy. Caves, D.W., Christensen, L.R., Diewert, W.E., 1982. The economic theory of index numbers and the measurement of input, output, and productivity. Econometrica 50, 1393–1414. Charnes, A., Cooper, W.W., Rhodes, E., 1978. Measuring efficiency of decision making units. European Journal of Operational Research 2, 429–444. Charnes, A., Cooper, W.W., 1985. Preface to topics in data envelopment analysis. Annals of Operation Research 2, 59–94. Charnes, A., Cooper, W.W., Seiford, L.M., 1994a. Extension to DEA models. In: Charnes, A., Cooper, W.W., Lewin, A.Y., Seiford, L.M. (Eds.), Data Envelopment Analysis: Theory, Methodology and Applications. Kluwer Academic Publishers, Norwell. Charnes, A., Cooper, W.W., Lewin, A.Y., Seiford, L.M., 1994b. Data Envelopment Analysis: Theory, Methodology, and Application. Kluwer Academic Publishers, Norwell. Chien, T., Hu, J.L., 2007. Renewable energy and macroeconomic efficiency of OECD and non-OECD economies. Energy Policy 35, 3606–3615. Coelli, T.J., 1996. A Guide to Deap Version 2.1: A Data Envelopment Analysis (Computer) Program. Department of Econometric, University of New England, Armidale, Australia. Edvardsen, D.F., Fbrsund, F.R., 2003. International benchmarking of electricity distribution utilities. Resource and Energy Economics 25, 353–371. Fisher-Vanden, Karen, Jefferson, Gary H., Liu, Hongmei, Tao, Quan, 2003. What is driving China’s decline in energy intensity. Resource and Energy Economics 26, 77–97. Forsund, F.R., Kittelsen, S.A.C., 1998. Productivity development of Norwegian electricity distribution utilities. Resource and Energy Economics 20, 207–224. Garbaccio, R.F., Ho, M.S., Jorgenson, D.W., 1999. Why has the energy–output ratio fallen in China. Energy Journal 20 (3), 63–91. Halkos, G.E., Tzeremes, N.G., 2009. Exploring the existence of Kuznets curve in countries’ environmental efficiency using DEA window analysis. Ecological Economics 68, 2168–2176. Hartman, T.E., Storbeck, J.E., 1996. Input congestion in loan operations. International Journal of Production Economics 46–47, 413–421. Honma, S., Hu, J.L., 2008. Total-factor energy efficiency of regions in Japan. Energy Policy 36, 821–833. Honma, S., Hu, J.L., 2009. Total-factor energy productivity growth of regions in Japan. Energy Policy 37, 3941–3950. Honore´, B.E., 1993. Orthogonality conditions for Tobit models with fixed effects and lagged dependent variables. Journal of Econometrics 59, 35–61. Hu, J.L., Wang, S.C., 2006. Total-factor energy efficiency of regions in china. Energy Policy 34, 3206–3217.
Hu, J.L., Kao, C.H., 2007. Efficient energy-saving targets for APEC economies. Energy Policy 35, 373–382. Hu, J.L., Chang, T.P., 2009. Total-factor energy productivity growth of regions in China. National Chiao Tung University, Taipei, Working paper. Huang, J.P., 1993. Industrial energy use and structural change: a case study of the People’s Republic of China. Energy Economics 15, 131–136. Iyer, K.G., Rambladi, A.N., Tang, K.K., 2004. Measuring spillovers from alternative forms of foreign investment. Working paper series from Centre for Efficiency and Productivity Analysis, University of Queensland. Liao, H., Fan, Y., Wei, Y.M., 2007. What induced China’s energy intensity to fluctuate: 1997–2006? Energy Policy 35 4640–4649. Lin, X., Polenske, K.R., 1995. Input–output anatomy of China’s energy use changes in the 1980s. Economic Systems Research 7 (1), 67–84. Lin, J., Zhou, N., Levine, M., Fridley, D., 2008. Taking out 1 billion tons of CO2: the magic of China’s 11th Five-Year Plan. Energy Policy 36, 954–970. Ma, C., Stern, D.I., 2008. China’s changing energy intensity trend: a decomposition analysis. Energy Economics 30, 1037–1053. Penn World Table 5.6, 1998. Accessible at /http://datacentre.chass.utoronto.ca/ pwt56/S. Sadjadi, S.J., Omrani, H., 2008. Data envelopment analysis with uncertain data: an application for Iranian electricity distribution companies. Energy Policy 36, 4247–4254. Sinton, J.E., Levine, M.D., 1994. Changing energy intensity in Chinese industry: the relative importance of structural shift and intensity change. Energy Policy 22, 239–255. Steenhof, P., 2006. Decomposition of electricity demand in China’s industrial sector. Energy Economics 28, 370–384. Tulkens, H., Vanden Eeckaut, P., 1995. Non-parametric efficiency, progress and regress measures for panel data: methodological aspects. European Journal of Operational Research 80, 474–499. Webb, R.M., 2003. Levels of efficiency in UK retail banks: a DEA window analysis. International Journal of the Economics of Business 10, 305–322. Wei, Y.M., Liao, H., Fan, Y., 2007. An empirical analysis of energy efficiency in China’s iron and steel sector. Energy 32, 2262–2270. Weidou, N., Johansson, T.B., 2004. Energy for sustainable development in China. Energy Policy 32, 1225–1229. World Bank, 2005. World Development Indicators. Accessible at /http://www. worldbank.orgS. Wu, Y., 2004. Openness, productivity and growth in the APEC economies. Empirical Economics 29, 593–604. Yue, P., 1992. Data envelopment analysis and commercial bank performance: a primer with applications to Missouri banks. Federal Reserve Bank of St. Louis Review 74, 31–45. Zhang, Z.X., 2003. Why did the energy intensity fall in China’s industry sector in the 1990s? The relative importance of structural change and intensity change. Energy Economics 25, 625–638.