Total-factor energy efficiency of regions in Japan

(1)

Energy Policy 36 (2008) 821–833

Total-factor energy efﬁciency of regions in Japan

Satoshi Honma

a,

, Jin-Li Hu

b

a_{Faculty of Economics, Kyushu Sangyo University, 2-3-1 Matsukadai, Higashi-ku, Fukuoka 813-8503, Japan} b_{Institute of Business and Management, National Chiao Tung University, Taiwan}

Received 21 June 2007; accepted 31 October 2007

Abstract

This study computes the regional total-factor energy efficiency (TFEE) in Japan by employing the data envelopment analysis (DEA). A dataset of 47 prefectures in Japan for the period 1993–2003 is constructed. There are 14 inputs, including three production factors (labor employment, private, and public capital stocks) and 11 energy sources (electric power for commercial and industrial use, electric power for residential use, gasoline, kerosene, heavy oil, light oil, city gas, butane gas, propane gas, coal, and coke). GDP is the sole output. Following Fukao and Yue [2000. Regional factor inputs and convergence in Japan—how much can we apply closed economy neoclassical growth models? Economic Review 51, 136–151 (in Japanese)], data on private and public capital stocks are extended. All the nominal variables are transformed into real variables, taking into consideration the 1995 price level. For kerosene, gas oil, heavy oil, butane gas, coal, and coke, there are a few prefectures with TFEEs less than 0.7. The five most inefficient prefectures are Niigata, Wakayama, Hyogo, Chiba, and Yamaguchi. Inland regions and most regions along the Sea of Japan are efficient in energy use. Most of the inefficient prefectures that are developing mainly upon energy-intensive industries are located along the Pacific Belt Zone. A U-shaped relation similar to the environmental Kuznets curve (EKC) is discovered between energy efficiency and per capita income for the regions in Japan.

Keywords: Energy-saving target ratio; Total-factor energy efﬁciency; Data envelopment analysis

1. Introduction

The 1990s are called the ‘lost decade’ in Japan. The average annual growth rate of real GDP was 1% in the 1990s, which was lower than that of other industrialized countries. The causes of the depression continue to be controversial. Although the performance of Japan’s economy in the 1990s was disappointing, an increase in energy consumption has been observed. As such, it is important to reduce energy consumption without harming economic performance due to the following two reasons.

First, saving energy is an important issue in Japan, because Japan’s energy self-sufﬁciency stands at a low 4%. Japan has traditionally promoted energy conservation after the two oil crises in the 1970s. In the past 30 years, the industrial sector in Japan succeeded in saving energy, but energy consumption in the residential, commercial, and

transportation sectors showed upward trends as a result of the pursuit of a comfortable lifestyle.1

Second, in accordance with the Kyoto Protocol—an international treaty designed to mitigate global warming— Japan is required to reduce its greenhouse gas emissions by 6% from 1990 levels, on average, during the period 2008–2012. Energy consumption is the main source of carbon dioxide emissions in Japan, and these emissions have increased approximately by 13.1% between 1990 and 2005. Carbon dioxide emissions resulting from energy consumption accounted for 92.6% of greenhouse gas emissions in 2005.2 To reduce carbon dioxide emissions, www.elsevier.com/locate/enpol

Corresponding author. Tel.: +81 92 673 5280; fax: +81 92 673 5919. E-mail address:honma@ip.kyusan-u.ac.jp (S. Honma).

1_{For a discussion on the energy-efﬁciency policy of Japan and other}

OECD countries, seeGeller et al. (2006).

2_{Nuclear power generation provides approximately one-third of Japan’s}

electric power. The government plans to expand nuclear power generation, because it does not emit carbon dioxide at the time of power generation. However, public opinion with regard to its construction has been negative due to accidents and the falsiﬁcation of records in recent years.

(2)

the implementation of energy efﬁciency has become an important political issue in Japan.

Because the usage of energy differs across areas, regional policies to reduce greenhouse gases play an important role. According to the Law Concerning the Promotion of the Measures to Cope with Global Warming, local govern-ments in Japan shall strive to formulate plans on measures to limit greenhouse gas emissions in accordance with the natural and social conditions of their areas. A government agency subsidizes regional energy-saving plans through local governments.3 Although the local governments estimate energy consumption in an area and set a goal for reducing greenhouse gases, the problem lies in the fact that governments are not aware of regional energy efﬁciency.

Although the energy consumption of Japan as a whole has increased recently, Japan is more energy efﬁcient than other industrialized countries.4 In 2003, the size of the Japanese economy had doubled as compared with 1973, whereas energy consumption of the manufacturing indus-try increased slightly. In 2003, if Japan’s primary energy consumption (on crude oil equivalent basis) per real GDP is taken as 1, then that of the United States is 2.08, that of United England is 1.43, that of France is 1.36, and that of Germany is 2.41.5

Japan’s national energy policy aims to achieve three underlying goals, the 3Es: energy security, environmental protection, and economic growth.6For the third target, the Japanese government has been undertaking the liberal-ization and deregulation of the energy markets in the past 10 years. One of the reasons for this is that petroleum, electricity, and gas prices in Japan were higher than those in other countries. Liberalization and deregulation have reduced costs of electricity and gas and increased the efﬁciency of the energy markets. As a result, these policies have stimulated economic growth. The effects of the above energy policies are the major concerns. For example,

Sueyoshi and Goto (2001) empirically examined the

performance of electric power companies in Japan by employing data envelopment analysis (DEA). The imple-mentation of energy saving requires data on the energy consumption of each sector across areas; however, such statistics do not exist. Kainou (2006) estimated energy consumption and carbon dioxide emissions for 47 regions and 15 sectors in Japan. This estimation provides basic information useful for regional, environmental, and QJ;energy policies and enables a comparison among

regions. However, it does not calculate the absolute efﬁciency score of energy use in each district.

This paper analyzes the energy efficiency of the Japanese economy by employing DEA. The energy efficiency of areas in Japan is obtained from the total-factor framework. Not much attention has been paid to regional energy efficiency in Japan, and to our knowledge, there has been no study applying the DEA method to measure regional energy efficiency in Japan. Our results identify the areas for which the government needs to improve the energy efficiency and provide useful insights with respect to regional energy and environmental policies. The paper is organized as follows: Section 2 gives our methodology of DEA. Section 3 describes the data we use for our analysis. Section 4 presents the empirical results and discussions. Finally, Section 5 concludes this paper.

2. Methodology of the DEA

This paper uses DEA to determine the input targets for each Japanese region by comparing the annual efficiency frontier that is constituted by all the Japanese regions in each year. Since the frontier is an input-reducing focus, this paper uses input-orientated measures following Farrell’s (1957)original ideas. In order to pursue overall technical efficiency with energy inputs, our study adopts the constant returns to scale (CRS) DEA model (Charnes et al., 1978). Our measure of relative efficiency is based on non-parametric techniques (Fa¨re et al., 1994). Let us first define some mathematical notations: there are K inputs and M outputs for each of the N objects. The ith object is represented by the column vectors xi and yi, respectively. The K N input matrix X and the M N output matrix Y represent the data for all the N objects. The input-oriented CRS DEA model then solves the following linear programming problem for object i in each year:

Dðy_i; xiÞ ¼Miny;l y

subject to y_iþY lX0;

yxiX lX0;

lX0;

(1)

where y is a scalar and l is an N 1 vector of constants. The value of y is the overall technical efficiency (OTE) score for the ith object, with 0pyp1. The value of unity indicates a point on the frontier and hence the technically efficient regions, in accordance with Farrell’s (1957) definition. The frontier is a piecewise linear isoquant, determined by the observed data points of the same year, i.e., all the regions in this study of the same year. The region that constructs the frontier is the ‘best practice’ among those observed regions in that year. The weight vector l serves to form a convex combination of observed inputs and outputs.

Fig. 1 illustrates the efﬁciency measurement: each point

on Fig. 1 represents a combination of inputs that all

produce the same level of output. Regions C and D are on 3_{The New Energy and Industrial Technology Development}

Organiza-tion (NEDO), a quasi-public organizaOrganiza-tion under METI, subsidizes local authorities to establish energy conversation visions; from 2000 to 2006, 215 of Japan’s 1891 local authorities received this subsidy.

4

In the 1990s, the total ﬁnal energy consumption increased by 13.4%.

5

Numbers are quoted fromMETI (2006, p. 150). They are based on the original data of IEA (2005). Primary energy equivalent of nuclear electricity is calculated by the conversion method ofIEA (2005).

6

(3)

the frontier and they cannot maintain the given output level by further reducing their inputs. Regions A and B are therefore inefﬁcient regions.

An important issue in efﬁciency studies is the credibility of the assumption that all production processes can actually reach the best practice production frontier (Zofı´o

and Prieto, 2001; Chien and Hu, 2007). In the present

study, when measuring energy efﬁciency, it is assumed that the best practice is accessible to all regions. This assump-tion seems to be adequate since only Japanese regions are considered. Currently, specialized journals, technological fairs, the global marketing strategies of multinationals, etc., guarantee that new innovations are readily available to all regions (Zofı´o and Prieto, 2001;Chien and Hu, 2007).

The set on the frontier is the ‘best practice’ production among the observed regions. The inefﬁcient region can reduce inputs by the amount indicated by the arrow and still remain in the input set (Boyd and Pang, 2000). With respect to the ith region, its distance (amount) in relation to the projected point on the frontier by radial reduction without reducing the output level, (1y)xi, is known as ‘radial adjustment’. We can illustrate this through Fig. 1. Point B is the actual input set and point B0 _{is the ideal or}

best practice input set for region B by reducing the radial adjustment BB0_.

When the frontier runs parallel to the axes, this could lead to a problem. InFig. 1, point A0_{is the best practice for}

region A by reducing the radial adjustment AA0_{. Point A}0

can reduce some input so as to maintain the same output level. The reduced amount is called ‘input slack’ (by the amount CA0_{). For region A, the best practice is point C,}

instead of point A0_{, by reducing the radial adjustment AA}0

and slack CA0_.

The summation of slack and radial adjustments is the total amount (‘target’) that can be reduced without decreasing the output levels. With respect to energy input, the above summation is called the ‘energy-saving target’ (EST). The formula is as follows:

ESTði;tÞ¼Slack adjustmentði;tÞþRadial adjustmentði;tÞ,

(2)

where EST(i,t)refers to the EST in the ith region and the tth year.

An inefﬁcient region can reduce or save EST in energy use without reducing real economic growth. The CRS model may suggest the slack and radial adjustments of any individual input for all objects to be efﬁcient. Since the actual practice can be improved to the best practice, the actual energy consumption is always larger than or equal to the ideal energy input.

Efficiency is generally defined in terms of the ratio with which the best practice compares with the actual operation. The indicator of energy efficiency should therefore be the ratio of the aggregate energy-saving target from Eq. (2) to the actual energy consumption. The amount of total adjustments in energy input is regarded as the inefficient portion of actual energy consumption. Based on the slack and radial adjustments of energy obtained from DEA, we can calculate the energy-saving target ratio (ESTR), considering other factors simultaneously. The target inputs of an object in a year are determined by comparing its actual inputs to the efficiency frontier in that year. The formula is as below:

ESTRði;tÞ¼

Energysaving targetði;tÞ

Actual energy input_ði;tÞ , (3)

where ESTR(i,t)refers to the ESTR in the ith region and the tth year.

As Eq. (3) shows, the ESTR represents each region’s inefﬁcient level of energy consumption. Since the minimal value of EST is zero, the value of ESTR lies between zero and unity. The total-factor energy efﬁciency (TFEE) index originally proposed by Hu and Kao (2007) and Hu and

Wang (2006)has the following relation with ESTR:

TFEEði;tÞ¼1 ESTRði;tÞ, (4)

where TFEE(i,t)refers to the TFEE in the ith region and the tth year. A zero ESTR value indicates a region on the frontier with the best TFEE up to one among the observed regions. A zero ESTR implies that no redundant or over-consumed energy use exists (the target potential is zero) in this region; otherwise, an inefﬁcient region with the value of ESTR larger than zero implies that energy should and could be saved at the same economic growth level. A higher ESTR implies higher energy inefﬁciency and a higher energy-saving amount.

As mentioned above, many studies criticize the com-monly used indicator of energy inefﬁciency—namely, the energy intensity as a direct ratio of the energy input to GDP for measuring energy efﬁciency (e.g.,Patterson, 1996;

Renshaw, 1981). The ratio is only a partial-factor

energy-efficiency indicator since energy input is the only input-considered factor. Another argument is that this partial-factor ratio is inappropriate to analyze the impact of changing energy use over time (APERC, 2002). We then compute the energy efficiency by a total-factor framework including other inputs such as labor and capital. A total-factor efficiency indicator provides more information and a x2/y x1/y D B’ A’ C A S B S’ 0

(4)

more realistic comparative base to examine the de facto situation across regions.

3. Data sources and variable deﬁnitions

The DEA is applied to a dataset of 47 prefectures in Japan for the period 1993–2003. Table 1 presents the summary statistics of the inputs and output used in the DEA models. This study employs 14 inputs, including three production factors (labor employment, real private, and real public capital stocks) and 11 energy inputs (electric power for commercial and industrial use, electric power for residential use, gasoline, kerosene, heavy oil, light oil, city gas, butane gas, propane gas, coal, and coke). These energy inputs are all used for ﬁnal consumption in the regions. The regional GDP is the sole output. The data on private and public capital stocks are unavailable, and hence the data estimated in Fukao and Yue (2000) are extended.

Data on prefectural real GDP and labor (employed persons) are taken from the Annual Report on Prefectural Accounts published by the Cabinet Ofﬁce (2006a). All prices are adjusted to the market prices for the calendar year of 1995. Electric power data for commercial and industrial use and residential use are obtained from the Handbook of Electric Power Industry published by the Federation of Electric Power Companies of Japan

(1994–2004).7Data on propane and butane gas

consump-tion are obtained from the website of Japan LP Gas Association (2007)(http://www.j-lpgas.gr.jp/). Data on city gas consumption by administrative division in Japan are released only by calendar year, which are obtained from the Annual Statistics of Gas Industry published by the

Agency for Natural Resources and Energy (each year a).

Therefore, only the data on city gas consumption are collated by calendar year, whereas data on all other economic and other energy consumption are collated by ﬁscal year. Data on gasoline, kerosene, gas oil (light oil), and heavy oil are taken from the Yearbook of Mineral Resources and Petroleum Products Statistics (METI). There are no ofﬁcial statistics on coal and coke consumption by prefecture. As for them, we use the estimated data in Kainou (2006).

There are no formal statistics on the private and public capital stock by administrative division in Japan. We obtain the extended data estimated byFukao and Yue (2000), who examined the mechanism of interregional convergence in Japan and estimate the private and public capital stocks by administrative division in Japan from 1955 to 1995.8 We extend the private capital stock data from 1996 to 2003 and

estimate it using their method, which is as follows: Ki_t¼Ki_t1þ ðKtKt1Þ Ii_t It , (5) where Kt i

is the private capital stock of prefecture i at year t; Ktis the nationwide private capital stock; It

i

is the private ﬁrm investment of prefecture i; and It is the nationwide private investment. Data on the gross capital stock of private enterprises (Kt) are obtained from the Cabinet

Ofﬁce (2006b).9The private ﬁrm investments to

adminis-trative regions (Iti) are from the aforementioned Annual Report on Prefectural Accounts. We assume that rate of elimination of capital stock is the same in all regions in each year.

There are also no formal data on nationwide public capital stock. Therefore, we extend the public capital stock data estimated in Fukao and Yue (2000) by the Cabinet Ofﬁce (2002). Referring to data of Cabinet Ofﬁce (2002), we extend the national public stock data to 2003, assuming that public capital is durable for 34 years and that public capital faces a disaster at half the number of the durable years (17 years). Note that two public corporations in Japan were privatized in the 1980s. In 1985, Nippon Telegraph and Telephone Public Corporation was priva-tized to become the Nippon Telegraph and Telephone Corporation (NTT). In addition, in 1987 the Japan National Railway was privatized and split into several railway companies. The computation process of the nationwide public capital stock is as follows:

K_t ¼K_t1þI_t I_t34þINT T_t34þIJR_t34 þB t34þ17BJR t34þ17B t34þBJR t34, ð6Þ

where Kt* is the nationwide public capital stock in year t; It* is the nationwide public investment of new construction and improvement in year t; Bt* is the nationwide public investment of natural disaster relief expenditure in year t; IJRt* is investment of new construction and improvement of Japan National Railway at time t; BJRt* is the investment of Japan National Railway’s natural disaster relief expenditure in year t; and INTTt* is investment of new construction and improvement for Nippon Telegraph and Telephone Public Corporation in year t. The nation-wide public capital stock at year t (Kt*) extended using the above method is then allocated to each prefecture in the same manner as in the case of private capital stock data as follows: Ki_t¼Ki_t1 þ ðK_t K_t1ÞI i t I t . (7)

Table 2presents the correlation coefﬁcients of the input and output variables. The isotonicity property that an output should not decrease with an increase in an input is 7

To avoid the problem of double counting, we subtract each source of energy consumption used for power generation, with reference to Outline of Electric Power Supply and Demand published by theAgency for Natural Resources and Energy (each year b)in Japan.

8

The data ofFukao and Yue (2000)can be obtained from the following database:http://www.ier.hit-u.ac.jp/fukao/japanese/data/index.html (in Japanese).

9

FollwiingFukao and Yue (2000), we refer to data on the gross capital stock of private enterprises obtained from theCabinet Ofﬁce (2006b), the main data source for Japanese capital stock. Note that its depreciation includes the elimination of capital and not capital consumption.

(5)

not violated.Tables 1 and 2show the summary statistics of these inputs and output.

4. Empirical results

Table 3 shows the result of the overall technical

efﬁciency scores for the administrative regions in Japan. We calculate the TFEE in the case of 11 energy sources as inputs. Due to space constraints, we only present the TFEEs for four energy inputs inTables 4–7.

Table 3 presents 29 prefectures that have efﬁciency

scores of one for the entire period. These prefectures operate on the Japanese efﬁciency frontiers and comprise prefectures in rural and metropolitan areas, such as Tokyo and Saitama. These prefectures share a common feature: the proportion of energy-intensive industries to the total number of manufacturers is lower than the nationwide average.10

We consider the geographical features of the prefectures. Fig. 2shows the geographical distribution of the efficient and inefficient regions in Japan. Most of the prefectures that face the Sea of Japan have an efficiency score of one for the entire period (05, 06, 16, 17, 18, 31, 32, 41, 42, and 43). Eight inland prefectures (09, 10, 11, 19, 20, 21, 25, and 29) have an average score of at least 0.97. In particular, six areas with the exception of Gifu (21) and Shiga (25) have

an efficiency score of one for the entire period. The reason why the areas along the Sea of Japan and inland are efficient, for the most part, is that industry and population in modern Japan have concentrated on the Pacific side of the archipelago. These areas are collectively known as the Pacific Belt Zone, which encompasses areas from the Tokyo metropolitan area to the city of Kitakyushu in Fukuoka prefecture. Among other things, energy-intensive industries have concentrated on certain areas on the side. It is important that an energy-saving policy be implemented intensively in the Pacific Coast area.

We next consider the relationship between per capita income and energy efficiency. The regions are divided into four groups based on the average per capita income at 1995 market prices. The groups are as follows: low income, 2.5 million yen or lower; lower middle income, 2.5–2.75 million yen; upper middle income, 2.75–3.0 million yen; and high income, 3.0 million yen or higher.Fig. 3depicts the rates of the efficient areas (more than 0.95) in each group with respect to the technical efficiency score and the TFEE indices of the main energy consumption. A U-shaped relation similar to the environmental Kuznets curve (EKC) is discovered between the TFEE and the per capita income for the regions in Japan. The reason behind this result is as follows. In regions belonging to the high-income group, there are many areas wherein the service industry accounts for a high value added without the use of a considerable amount of energy. Many prefectures in the low-income group do not have industrial zones. Because there are areas

Table 1

Description and summary statistics of variables

Variable Deﬁnition Unit Mean Std Dev Minimum Maximum

Output

y Total income Billion yen in 1995 prices 10,843.73 13,826.30 2009.20 88,566.02 Inputs

x1 Employed persons Person 1,337,847.21 1,423,184.49 313,693.00 8,782,396.00

x2 Private capital stock Billion yen in 1995 prices 22,324.36 25,983.45 3131.81 166,007.50

x3 Public capital stock Billion yen in 1995 prices 16,435.11 13,915.21 4005.28 83,458.06

x4 Electric power for residential use Million kWh 5067.27 4964.56 942.00 28,428.00

x5 Electric power for commercial and industrial use Million kWh 11,774.85 11,079.22 1763.00 52,955.00

x6 Gasoline kL 1,179,741.78 1,019,686.72 268,654.00 7,591,664.00

x7 Kerosene kL 612,685.61 654,523.86 60,428.00 4,092,522.00

x8 Gas oil kL 896,769.57 752,560.76 140,763.00 4,807,624.00

x9 Heavy oil kL 1,050,366.68 950,189.23 57,223.00 5,793,805.00

x10 City gas Million MJ 20,480.33 40,516.85 515.00 241,405.00

x11 Butane gas Tons 103,035.14 130,104.53 4914.00 770,696.00

x12 Propane gas Tons 211,618.90 164,136.00 39,222.00 890,332.00

x13 Coal 1000 ton 368.72 573.48 4.87 2664.36

x14 Coke 1000 ton 851.57 1621.14 0.47 7089.15

Table 2

Correlation coefﬁcients of input and output variables

x1 x2 x3 x4 x5 x6 x7 x8 x9 x10 x11 x12 x13 x14

y 0.930 0.990 0.892 0.961 0.920 0.906 0.679 0.860 0.796 0.913 0.801 0.786 0.163 0.179

10

In this paper, chemical, ceramic, iron and steel, metal products, and pulp and paper industries are regarded as energy-intensive industries.

(6)

in some prefectures belonging to the lower and upper middle-income groups that have energy-intensive indus-tries, they are relatively inefficient. The curves of kerosene and heavy oil overlap in Fig. 3. As the per capita income rises, the electric power for commercial and industrial use and coal is used inefficiently. The former is used efficiently in the lower middle- and low-income regions.

We now consider the relationship between regional energy efﬁciency and the extent of the energy-intensive industry. Prefectures are divided into four groups based on the average ratio of a region’s GDP out of energy-intensive

industries in the sample period. The groups are as follows: the rate of the group A regions is higher than 10%, that of the group B regions lies between 7.5% and 10%, that of the group C regions lies between 5% and 7.5%, and that of the group D regions is less than 5%.Fig. 4depicts the rates of the efficient areas (more than 0.95) in each group with respect to the technical efficiency score and the TFEE indices of the main energy consumption. It is clear that the rates of inefficient areas are high in group A where energy-intensive industries dominate, especially with respect to heavy oil and coal.

Table 3

Overall technical efﬁciency scores of Japanese Administrative Regions (1993–2003)

ID Name 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 01 Hokkaido 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 02 Aomori 1.000 1.000 1.000 1.000 1.000 1.000 0.999 1.000 1.000 1.000 1.000 03 Iwate 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 04 Miyagi 0.971 0.991 0.975 0.959 0.957 0.946 0.966 0.975 0.980 0.978 0.982 05 Akita 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 06 Yamagata 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 07 Fukushima 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 08 Ibaraki 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.976 0.966 0.963 09 Tochigi 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 10 Gunma 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 11 Saitama 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 12 Chiba 0.986 0.979 0.956 0.920 0.884 0.885 0.892 0.890 0.838 0.894 0.918 13 Tokyo 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 14 Kanagawa 1.000 0.974 0.992 0.994 0.958 0.950 0.968 0.984 0.971 1.000 1.000 15 Niigata 0.893 0.894 0.902 0.902 0.846 0.834 0.802 0.784 0.810 0.846 0.859 16 Toyama 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 17 Ishikawa 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 18 Fukui 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 19 Yamanashi 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 20 Nagano 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 21 Gifu 0.990 0.987 0.959 0.977 1.000 1.000 1.000 0.990 0.983 0.990 1.000 22 Shizuoka 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 23 Aichi 0.912 0.919 0.941 0.936 0.874 0.910 0.912 0.964 1.000 1.000 1.000 24 Mie 0.980 0.961 0.980 0.974 0.982 0.977 0.962 0.979 0.952 1.000 1.000 25 Shiga 1.000 1.000 1.000 1.000 1.000 0.991 0.979 0.968 0.911 0.936 0.964 26 Kyoto 0.970 0.970 0.971 0.954 0.919 1.000 1.000 1.000 1.000 1.000 1.000 27 Osaka 1.000 1.000 1.000 1.000 0.980 1.000 1.000 1.000 1.000 1.000 1.000 28 Hyogo 0.924 0.895 0.923 0.926 0.787 0.883 0.852 0.873 0.863 0.865 0.882 29 Nara 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 30 Wakayama 0.884 0.860 0.871 0.863 0.852 0.908 0.819 0.756 0.826 0.921 0.968 31 Tottori 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 32 Shimane 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 33 Okayama 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.986 34 Hiroshima 1.000 1.000 0.995 0.991 0.957 0.956 0.938 0.945 0.943 0.919 0.946 35 Yamaguchi 0.929 0.955 0.946 0.934 0.932 0.963 0.942 0.937 0.924 0.948 0.934 36 Tokushima 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 37 Kagawa 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 38 Ehime 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 39 Kouchi 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 40 Fukuoka 0.931 0.969 0.948 0.945 0.937 0.942 0.925 0.933 0.940 0.959 0.975 41 Saga 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 42 Nagasaki 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 43 Kumamoto 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 44 Oita 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 45 Miyazaki 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 46 Kagoshima 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 47 Okinawa 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000

(7)

To clarify the characteristics of the inefficiency in prefectures, we describe the five most inefficient prefectures based on the average scores of the 11-year period. The prefecture with the lowest efficiency score is Niigata (15), an industrial city that faces the Sea of Japan. Its main industry is metal manufacturing. The percentage of fabricated metal products in the value of manufactured goods shipments in Niigata is nearly double the national average. Another feature of Niigata is the mining industry. Niigata is ranked as the top producer of both crude oil and natural gas in Japan. These natural resources are rare in

Japan and have facilitated the development of iron and chemical factories.

Wakayama (30) is the most inefﬁcient prefecture after Niigata. Many large factories such as Kao, Sumitomo Metal Industries, Mitsubishi Electric, and Tonen General (a company that is part of the Exxon Mobil group) are located in Wakayama. A distinctive feature of Wakayama is that chemical and allied products, petroleum and coal products, and iron and steel totally account for approxi-mately half of the total manufacturing production in Wakayama.

Table 4

Total-factor energy efﬁciency in electric power for commercial and industrial use of Japanese Administrative Regions (1993–2003)

(8)

After Niigata and Wakayama, Hyogo (28) is the most inefﬁcient prefecture. Hyogo prefecture is one of the leading industrial prefectures in Japan. Major Japanese companies such as Kawasaki Heavy Industries, Kobe Steel, Ishikawajima-Harima Heavy Industries, Toshiba, Fujitsu, and Mitsubishi Electric Corporation are located in Hyogo. Its main industries include iron and steel, chemical and allied products, and general machinery. In particular, the proportion of iron and steel production in Hyogo to total manufacturing production is nearly twice as high as the nationwide average.

Chiba (12) is the fourth most inefﬁcient prefecture. Chiba prefecture is one of the leading industrial pre-fectures in Japan and has developed around the petro-chemical complex located in Tokyo Bay. Many leading Japanese companies such as Nippon Steel, JFE Steel, Sumitomo Chemical, Mitsui Chemicals, and Cosmo Oil, are located in Chiba. Chemical and allied products, petroleum and coal products, and iron, steel, and fabricated metal products totally account for approxi-mately half of the total manufacturing production in Chiba.

Table 5

Total-factor energy efﬁciency in kerosene of Japanese Administrative Regions (1993–2003)

(9)

Yamaguchi (35) is the last of the five most inefficient prefectures. Mitsui Chemicals and Taiheiyo Cement Co. originated in Yamaguchi. Furthermore, the heavy and chemical industry has historically developed in Yamaguchi. Its major industries include chemical and allied products, ceramic, stone and clay products, and iron and steel. These three sectors account for about 30% of the total manufacturing production in Yamaguchi. In particular, the proportion of petroleum and coal products to total manufacturing production is nearly five times as high as the nationwide average.

Several observations from the above descriptions reveal that the inefficient regions are concentrated on energy-intensive industries such as iron and steel, chemical, and ceramic. Energy efficiency depends on the industrial structure in areas. To improve energy efficiency, the national and local governments must endeavor to change the industrial structure from energy-intensive industries to others. Service and high value-added industries should be promoted in these energy-inefficient regions. Energy-conserving equipment and technologies should be first applied to the industries in these energy-inefficient regions.

Table 6

Total-factor energy efﬁciency in heavy oil of Japanese Administrative Regions (1993–2003)

(10)

We next consider the main energy consumption. Five of the most inefficient prefectures in terms of TFEE in heavy oil for the sample period are Yamaguchi (35), Chiba (12), Niigata (15), Mie (24), and Wakayama (30). The reason for this is that the energy-intensive industries have developed in these five areas and that heavy oil is a major input in their production. In particular, most of the scores for Yama-guchi, Chiba, and Mie in the 11 years are less than 0.5, implying that these three prefectures have operated at efficiencies of less than 50% of that for the areas in the efficient frontiers. The TFEE scores of heavy oil in these

inefﬁcient regions have improved in the sample period. Five prefectures with the least TFEE in coal in the sample period are Hyogo (28), Wakayama (30), Yamaguchi (35), Niigata (15), and Chiba (12). The reason for this is the same as that in the case of heavy oil. The average TFEE score of these prefectures is an extremely low 0.2. The consumption of coal in these prefectures is extremely high and that in most of the remaining prefectures is very low. These results might be induced by the fact that the consumption of coal differs greatly across regions. Five prefectures with the lowest TFEEs of electric power for industrial use in the

Table 7

Total-factor energy efﬁciency in coal of Japanese Administrative Regions (1993–2003)

(11)

sample period are Hyogo (28), Aichi (23), Shiga (25), Chiba (12), and Kanagawa (14). These prefectures score poorly in a year in the sample period. Most of these areas overlap with the areas mentioned above.

5. Conclusions

This study analyzes the regional TFEE in Japan by using DEA. The efﬁciency frontier is constructed by using DEA based on data on energy sources and other inputs. A dataset of 47 prefectures in Japan for the period 1993–2003 is constructed. There are 14 inputs, including three production factors (labor employment, private, and public capital stocks) and 11 energy sources (electric power for commerical and industrial use, electric power for residen-tial use, gasoline, kerosene, heavy oil, light oil, city gas, butane gas, propane gas, coal, and coke). GDP is the sole output. FollowingFukao and Yue (2000), data on private and social capital stocks are extended. All the nominal variables are transformed into real variables at 1995 price levels.

A U-shaped relation similar to the EKC theory is discovered between the overall technical efﬁciency and the

per capita income for regions in Japan. Many areas with dominant energy-intensive industries are energy inefficient. For kerosene, gas oil, heavy oil, butane gas, and propane gas, there are a few prefectures with TFEEs less than 0.5. Chiba ranks in the bottom three on all nine energy inputs. Most of the regions along the Sea of Japan and in the inland are almost all energy efficient. The Pacific Belt Zone has most of the inefficient prefectures that mainly develop upon energy-intensive industries.

In order to reduce carbon dioxide, the Japanese government has promoted to maintain or even increase the share of electricity generated by nuclear power, approximately up to one-third. However, this energy policy may not work well since the construction of any new nuclear power plant is opposed by the public, because of worry about its safety from the public. The government also promotes the extension of renewable energy resources (such as solar, wind, biofuel, etc.). However, high cost and technical difficulties of biofuel are major hurdles against utilizing renewable energy. Because Japanese energy policies face substantial difficulties, energy-efficiency im-provement is an important issue. To save energy and

(12)

reduce carbon dioxide emissions, the national and local governments should promote the improvement of efﬁ-ciency for energy-intensive industries in the Paciﬁc Belt Zone and Niigata and/or change the industrial structure from energy-intensive industries to others such as service industries.11

Further research is needed to examine the industry-specific energy efficiency in each region. This paper uses the total energy consumption in each area of the industrial, commercial, transportation, and residential sectors. The energy consumption of sectors other than the industrial sector must be carefully considered. First, cooling and heating energy consumption in residential and commercial buildings varies across areas. Because the Japanese archipelago runs north and south, the difference in temperature varies considerably across regions. Second, it is somewhat difficult to specify the purpose of energy consumption in the transportation sector across areas. Gasoline and gas oil sold in an area are used as fuel for automobile transportation, not only within the area but also outside.

Acknowledgments

The authors thank an anonymous referee of this journal, S.Y. Liao, and the seminar participants at National Taiwan University and Soochow University for their helpful comments. The second author gratefully acknowl-edges partial ﬁnancial support from Taiwan’s National Science Council (NSC 95-2415-H-009-001). The usual disclaimer applies.

References

Agency for Natural Resources and Energy Japan, each yeara. Annual Statistics of Gas Industry, Tokyo.

Agency for Natural Resources and Energy Japanb, each yearb. Outline of Electric Power Supply and Demand, Tokyo.

Asia Paciﬁc Energy Research Center (APERC), 2002. Energy Efﬁciency Indicators and Potential Energy Savings in APEC Economies. APERC, Tokyo.

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

Cabinet Ofﬁce, Government of Japan, 2002. Nihon no shakai shihon (Social Capital of Japan), Tokyo.

Cabinet Ofﬁce, Government of Japan, 2006a. Annual Report on Prefectural Accounts, Tokyo.

Cabinet Ofﬁce, Government of Japan, 2006b. Gross Capital Stock of Private Enterprises, Tokyo.

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

Chien, T., Hu, J.L., 2007. Renewable energy and macroeconomic efﬁciency of OECD and non-OECD economies. Energy Policy 35, 3606–3615.

Fa¨re, R., Grosskopf, S., Lovell, C.A.K., 1994. Production Frontiers. Cambridge University Press, Cambridge.

85.7% 64.3% 71.4% 42.9% 88.9% 73.3% 100.0% 77.8% 80.0% 66.7% 90.0% 80.0% 70.0% 60.0% 50.0% 40.0% 30.0% 100.0%

Low Lower Middle Upper Middle High Per Capita Real Income

Rates of numbers of ef

ficient (more than 0.95) regions

OTE

TFEE Electric Power for Commercial and Industrial Use

TFEE Kerosene TFEE Heavy Oil TFEE Coal

Fig. 3. Relationship between per capita income and technical efﬁciency.

66.7% 33.3% 100.0% 75.0% 80.0% 83.3% 70.0% 89.5% 60.0% 50.0% 94.7% 30.0% 40.0% 50.0% 60.0% 70.0% 80.0% 90.0% 100.0% -5% 5%-7.5% 7.5%-10%

10%-Average Rates of Energy Intensive Industries in Regional GDP

Rates of numbers of ef

ficient (more than 0.95) regions

OTE

TFEE Electric Power for Commercial and Industrial Use

TFEE Kerosene TFEE Heavy Oil TFEE Coal

Fig. 4. Relationship between energy-intensive industries and technical efﬁciency.

11

A relocation of an energy-intensive industry to another area or country may lead to carbon leakage.

(13)

Farrell, M.J., 1957. The measurement of productive efﬁciency. Journal of the Royal Statistical Society, Series A 120 (Part 3), 253–290. Federation of Electric Power Companies of Japan, 1994–2004. Handbook

of Electric Power Industry, Tokyo.

Fukao, K., Yue, X., 2000. Regional factor inputs and convergence in Japan—how much can we apply closed economy neoclassical growth models? Economic Review 51, 136–151 (in Japanese).

Geller, H., Harrington, P., Rosenfeld, A.H., Tanishima, S., Unander, F., 2006. Polices for increasing energy efﬁciency: thirty years of experience in OECD countries. Energy Policy 34, 556–573.

Hu, J.L., Kao, C.H., 2007. Efﬁcient energy-saving targets for APEC economies. Energy Policy 35, 373–382.

Hu, J.L., Wang, S.C., 2006. Total-factor energy efﬁciency of regions in China. Energy Policy 34, 3206–3217.

International Energy Agency (IEA), 2003. Energy Policies of Japan. IEA, Paris.

International Energy Agency (IEA), 2005. Energy Balances of OECD Countries. IEA, Paris.

Japan LP Gas Association, 2007. The Annual Sales of LP Gas by Prefecture. Website /http://www.j-lpgas.gr.jpS.

Kainou, K., 2006. Energy consumption statistics according to adminis-trative divisions (in Japanese). Website /http://www.rieti.go.jp/users/ kainou-kazunari/energy/index.htmlS.

Ministry of Economy, Trade and Industry Japan, 2006. White paper on Energy, Tokyo.

Patterson, M.G., 1996. What is energy efﬁciency? Concepts, indicators and methodological issues. Energy Policy 24, 377–390.

Renshaw, E.F., 1981. Energy efﬁciency and the slump in labour productivity in the USA. Energy Economics 3, 36–42.

Sueyoshi, T., Goto, M., 2001. Slack-adjusted DEA for time series analysis: performance measurement of Japanese electric power generation industry in 1984–1993. European Journal of Operational Research 133, 232–259.

Zofı´o, J.L., Prieto, A.M., 2001. Environmental efﬁciency and regulatory standards: the case of CO2emissions from OECD industries. Resource