Total-factor energy productivity growth of regions in Japan

(1)

Total-factor energy productivity growth 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

National Chiao Tung University, Taiwan

a r t i c l e

i n f o

Article history:

Received 5 February 2009 Accepted 22 April 2009 Available online 20 May 2009 Keywords:

Total-factor energy productivity change index (TFEPI)

Data envelopment analysis (DEA) Malmquist productivity index (MPI)

a b s t r a c t

This article computes the energy productivity changes of regions in Japan using total-factor frameworks based on data envelopment analysis (DEA). Since the traditional DEA-Malmquist index cannot analyze changes in single-factor productivity changes under the total-factor framework, we apply a new index proposed by Hu and Chang [2009. Total-factor energy productivity growth of regions in China. Energy Policy, submitted for publication]: a total-factor energy productivity change index (TFEPI) that integrates the concept of the total-factor energy efficiency index into the Malmquist productivity index (MPI). Moreover, we separate TFEPI into change in relative energy efficiency, or the ‘catching up effect,’ and shift in the technology of energy use, or the ‘innovation effect.’ The data from 47 prefectures during the period of 1993–2003 are used to compute the TFEPI and its components for 4 kinds of energy. The TFEPI of electric power for commercial and industrial use changes 0.6% annually, which can be separated into a total-factor energy efficiency change of 0.2% and a technical change of 0.8%. The TFEPI for coal deteriorates by 1.0%/year, which is mostly caused by a decrease in relative energy efficiency change. We define and identify ‘innovators’ who cause the frontier to shift. Most regions identified as frontier shifters are located outside of Japan’s four major industrial areas.

1. Introduction

The ﬁrst oil crisis hit the Japanese economy in 1973 and led to a

drive for efﬁcient energy use in Japan. As a result, Japan has

achieved one of the highest levels of energy efﬁciency in the

world. Energy conservation policy has been a crucial concern for

Japan as a resource-poor country without a stable supply of

energy. Moreover, Japan ratiﬁed the Kyoto protocol and must, by

2012, decrease its greenhouse gas emissions by 6% from its 1990

level. The Ministry of Environment (MOE) of Japan has proposed a

carbon tax to mitigate carbon dioxide emissions since 2003. The

proposed tax rate in 2003 was 3400 yen (approximately 29 US$ at

the day’s exchange rate) per ton of carbon contained in fossil fuel

emissions, and, since 2004, it has been reduced to 2400 yen

(approximately 22 US$ at the day’s exchange rate). However,

because of opposition from business interests, the MOE has failed

to institute the carbon tax. Japan’s carbon dioxide emissions from

energy use have remained above the 1990 baseline and, in 2007,

increased 15.0% above it. As

Kasahara et al. (2007)

suggested, a

climate change tax combined with international emission trading

might be a rational choice for Japan; however, in reality a

climate change tax has been and will continue to be politically

unacceptable. The Japanese government’s plan depends on

voluntary action to reduce energy use in industrial, commercial,

and residential sectors, which seems to be unrealistic. In addition

to Japan’s obligation to implement the Kyoto mechanism

includ-ing the international emission tradinclud-ing, improvinclud-ing energy

efﬁ-ciency or energy productivity per se has been the key issue for

Japan’s energy-environmental policy. However, the energy

efﬁ-ciency-enhancing policy may have two unintended consequences:

First, improvements in energy efﬁciency may result in lower

energy prices and in turn increased energy consumption. This is

called the rebound effect which was ﬁrst suggested by W.S. Jevons

in 1865; however, this effect still remains debatable (recently, e.g.,

Hanley et al., 2009

). Second, energy efﬁciency measures may not

necessarily lead to reducing carbon emissions when Japan

participates in international emissions trading schemes. In that

case, the social cost of reducing carbon dioxide as well as the cap

amount of carbon emissions will be different if Japan does not

participate in these schemes.

1

Two well-known indicators are commonly used to study

whether energy inputs are efﬁciently used. The ﬁrst is energy

intensity, which measures the amount of energy consumption for

economic output produced in the economy. According to this kind

Contents lists available at

ScienceDirect

journal homepage:

www.elsevier.com/locate/enpol

Energy Policy

Corresponding author. Tel.: +81926735280; fax: +81926735290. E-mail addresses:honmasatoshi999@gmail.com,honma@ip.kyusan-u.ac.jp (S. Honma),jinlihu@mail.nctu.edu.tw (J.-L. Hu).

1

So¨derholm and Pettersson (2008) show that the social cost of power generation depends upon whether or not the country participates in international emissions trading in the Swedish case.

(2)

of indicator, Japan is one of the world’s leading countries in energy

use. For example, if Japan’s primary energy consumption (on a

crude oil equivalent basis) per real GDP is taken as 1 in 2005, then

that of the United States is 2.00, that of the United Kingdom is

1.35, that of France is 1.82, and that of Germany is 1.65.

2

_For

example, at the industry level, if energy consumption per unit of

production in the Japanese iron and steel industry is taken as 1,

that of the United States is 1.25, that of the United Kingdom is

1.22, and that of Germany is 1.17.

3

_{Moreover, if energy}

consump-tion per cement clinker in Japan is taken as 1, that of the United

States is 1.77 and that of Western Europe is 1.30. The second

indicator is energy efﬁciency (or energy productivity), deﬁned as

the economic output divided by the energy input (e.g.,

Berndt,

1990

;

Patterson, 1996

;

Han et al., 2007

). Notice that although each

indicator represents identical measures from different

perspec-tives, we focus only on the application of energy efﬁciency and

productivity in this paper.

The conventional energy efﬁciency index introduced in

Patterson (1996)

is partial-factor energy productivity because it

disregards the substitution among energy consumption and other

factors (e.g., labor and capital stock). If energy consumption is

evaluated in terms of partial-factor energy productivity, the end

result is a misleading estimate (

Hu and Wang, 2006

;

Hu and Kao,

2007

;

Han et al., 2007

;

Honma and Hu, 2008

). For this reason,

even though of the above international comparisons, it does not

follow that energy efﬁciency in Japan is higher than in other

developed countries. For example,

Hu and Kao (2007)

show that

Japan is not the best performer in the APEC economy in

1991–2000 using a total-factor framework.

This article evaluates the energy productivity change of regions

in Japan with a total-factor framework. Under the traditional

DEA-Malmquist index, one cannot evaluate the change in single-factor

productivity under the total-factor framework. As a result, we use

a new index, the total-factor energy productivity change index

(TFEPI), which was proposed in

Hu and Chang (2009)

. Following

Hu and Chang (2009)

, we extend the work of

Honma and Hu

(2008)

on total-factor energy efﬁciency (TFEE) to introduce a

total-factor energy productivity index that integrates the concept

of the total-factor energy efﬁciency index into the Malmquist

productivity index (MPI). The MPI was ﬁrst introduced by

Caves et

al. (1982)

to measure total-factor productivity change by the ratio

of the distance functions.

Fa¨re et al. (1994)

broke down the MPI

into efﬁciency change and technical change. They used data

envelopment analysis (DEA), which is a nonparametric, linear

programming method. To evaluate the TFEPI, we also use DEA.

Moreover, we can decompose TFEPI into changes in relative

energy efﬁciency (the catching up effect) and shifts in the

technology of energy use (the innovation effect) under the

total-factor framework. This study extends the panel dataset of

Honma

and Hu (2008)

and analyzes prefecture-level data from 1993 to

2003. There are a single, aggregate output (real GDP) and 14

inputs in our DEA model, including 3 production factors (labor

employment and real private and real public capital stocks), and

11 energy sources. To the best of our knowledge, no studies have

attempted to assess changes in energy productivity for regions in

Japan using a total-factor framework.

4

_{The revised energy}

conservation law evaluates energy efﬁciency with respect to each

apparatus, factory, and building from April 2009. Our results shed

new light on Japan’s energy productivity changes by examining

those changes by region and energy type.

The remainder of this paper is organized as follows: Section 2

introduces the proposed total-factor energy productivity index

using the DEA approach. Section 3 interprets the data sources and

describes the variables involved. Section 4 presents and discusses

the empirical results in the case of Japan. Finally, Section 5

concludes the paper.

2. Total-factor energy productivity index

Hu and Chang (2009)

propose the TFEPI, which combines the

concepts of TFEE and MPI to investigate the energy productivity

changes in regions of China. Because TFEE examines the optimal

energy input level with the input-oriented constant returns to scale

(CRS) DEA model, our TFEPI also follows an input-oriented model.

Additionally, MPI, which is usually computed by an output-oriented

DEA approach, is applied using an input-oriented framework in this

study. In the following subsection, we ﬁrst introduce the

input-oriented MPI and proceed with TFEE. Finally, the TFEPI is presented

with a discussion of how MPI and TFEE are integrated.

2.1. Input-oriented Malmquist productivity index

First, we assume that the production technology S

t

_{models the}

transformation of multiple inputs, x

t

A

R

₊K

, into multiple outputs,

y

t

A

R

₊M

, for each time period t, where

S

t

_{¼ fðx}

t

_;

_y

t

Þ

: x

t

can produce y

t

g

(1)

The computation of input-oriented MPI relies on input-based

distance functions. Following

Fa¨re et al. (1985)

and

Boussemart et

al. (2003)

, the input distance function can be deﬁned at t as

D

ti

ðx

t

_;

y

t

Þ ¼

supf

d

: ðx

t

=

d

;

y

t

Þ 2

S

t

g ¼ ðinff

d

: ð

d

x

t

;

y

t

Þ 2

S

t

gÞ

1

(2)

where distance function (2) is based upon the reciprocal of the

maximum proportional reduction of the input vector by a scalar

d

to catch up to the production frontier. It is notable that D

it

(x

t

,

y

t

_{)X1 and D}

it

(x

t

, y

t

) ¼ 1 if and only if (x

t

, y

t

) is on the production

frontier. Therefore, input-oriented MPI can be measure as follows:

Miðxtþ1;ytþ1;xt;ytÞ ¼ Dt iðxt;ytÞ Dt iðxtþ1;ytþ1Þ ! Dtþ1 i ðxt;ytÞ Dtþ1i ðxtþ1;ytþ1Þ ! " #1=2 ¼ D t iðxt;ytÞ Dtþ1i ðxtþ1;ytþ1Þ Dtþ1 i ðxtþ1;ytþ1Þ Dt iðxtþ1;ytþ1Þ ! Dtþ1 i ðxt;ytÞ Dt iðxt;ytÞ ! " #1=2

(3)

2.2. Total-factor energy efﬁciency

In order to pursue overall technical efﬁciency with energy

inputs, our study adopts the CRS DEA model (

Charnes et al., 1978

).

Let us ﬁrst deﬁne some mathematical notations. There are K

inputs and M outputs for each of N objects. The ith object is

represented by the column vectors x

i

and y

i

, respectively. The

K N input matrix X and the M N output matrix Y represent the

data for all N objects. The input-oriented CRS DEA model then

solves the following linear programming problem for object I in

each year:

Min

y;l

y

s:t: y

i

þ

Y

l

X

0 y

x

i

X

l

X

0 l

X

0 (4)

2

The above ﬁgures are based onOECD (2007).

3

Japan Business Federation (2008).

4

On the productivity change of Japanese prefectures,Nemoto and Goto (2005)

compute the total-factor productivity change for 1981–2000, andMiyara and Fukushige (2008)compute it for 1976–1997. However, these two models do not include energy as an input; their inputs are only capital stocks and labor.

(3)

where

y

is a scalar that represents the efﬁciency score for the ith

object, with 0p

y

p1.

l

is an N 1 vector of constants, and the

weight vector

l

serves to form a convex combination of observed

inputs and outputs.

After obtaining the efﬁciency score, we apply the approach of

Ali and Seiford (1993)

to compute the total slack, which includes

radial and non-radial slacks. Hence, the TFEE index of region i at

time t can be measured as

TFEE

it

¼

Target energy input

it

Actual energy input

it

¼

Actual energy input

it

Total slack of energy input

it

Actual energy input

it

(5)

2.3. Integrating MPI and TFEE to obtain TFEPI

In this section, we will show how TFEPI brings together MPI

and TFEE. The four input-oriented distance functions in Eq. (3) can

be replaced by the ratio of target energy input and actual energy

input under technologies in different periods. For example, D

it

(x

t

,

y

t

_{) would be presented as}

ðD

ti

ðx

t

;

y

t

ÞÞ

1

¼

Target energy input under technology in t

Actual energy input in t

¼

TFEE

t t

ðD

tþ1 i

ðx

t

_;

_y

t

ÞÞ

1

¼

Target energy input under technology in t þ 1

Actual energy input in t

¼

TFEE

tþ1 t

ðD

tþ1i

ðx

tþ1

;

y

tþ1

ÞÞ

1

¼

Target energy input under technology in t þ 1

Actual energy input in t þ 1

¼

TFEE

tþ1tþ1

ðD

ti

ðx

tþ1

;

y

tþ1

ÞÞ

1

¼

Target energy input under technology in t

Actual energy input in t þ 1

¼

TFEE

t tþ1

(6)

Therefore,

TFEPI ¼

TFEE

tþ1 tþ1

TFEE

t t

TFEE

t tþ1

TFEE

tþ1tþ1

!

TFEE

tt

TFEE

tþ1t

!

"

#

1=2

(7)

where the ﬁrst ratio (outside the brackets) represents the

total-factor energy efﬁciency changes and the second geometric

product of the ratio captures the total-factor energy technical

changes. Note that if the value of TFEPI or any of its components is

less than unity, then a regression or deterioration in performance

is indicated.

3. Description of data and variables

This study augments the panel dataset of

Honma and Hu

(2008)

and analyzes data from 47 prefectures from 1993 to 2003.

Table 1

presents the summary statistics of the inputs and output

used in the DEA models. In our model, 3 production factors (labor

employment, and real private and real public capital stocks) and

11 energy inputs (electric power for residential use, electric power

for commercial and industrial use, gasoline, kerosene, heavy oil,

light oil, city gas, butane gas, propane gas, coal, and coke) combine

to make 14 inputs. These energy inputs are all used for ﬁnal

consumption in each region. The real regional GDP is the sole

output. The data on private and public capital stocks are

unavailable, and hence we extend the stock data estimated in

Fukao and Yue (2000)

.

5

Data on real prefectural GDP and labor (employed persons) are

taken from the Annual Report on Prefectural Accounts (Cabinet

Ofﬁce, Government of Japan). Real GDP and real social and private

capital stocks are adjusted to 1995 yen. We use the same data

sources as

Honma and Hu (2008)

: data on electric power are from

the Handbook of Electric Power Industry (The Federation of Electric

Power Companies of Japan); data on propane and butane gas

consumption are taken from the website of the Japan LP Gas

Association (

http://www.j-lpgas.gr.jp/

); data on city gas

consump-tion are from the Annual Statistics of Gas Industry (Japan Gas

Association); and data on gasoline, kerosene, light oil, and heavy

oil are taken from the Yearbook of Mineral Resources and Petroleum

Products Statistics (Ministry of Economy, Trade and Industry).

Since there are no ofﬁcial statistics on coal and coke consumption

by prefecture, they are taken from the estimated data in

Kainou

(2006)

.

Table 2

is a correlation matrix. As shown in the table, all inputs

have positive correlation coefﬁcients with the output, implying

that all inputs satisfy the isotonicity property with the output for

the DEA model.

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 10843.73 13826.30 2009.20 88566.02 Inputs

x1 Employed persons Person 1372095.79 1458692.35 313693.00 8782396.00

x2 Private capital stock Billion yen in 1995 prices 22324.36 25983.45 3131.81 166007.50

x3 Public capital stock Billion yen in 1995 prices 16435.11 13915.21 4005.28 83458.06

x4 Electric power for residential use Million kWh 5067.27 4964.56 942.00 28428.00

x5 Electric power for residential use Million kWh 11774.85 11079.22 1763.00 52955.00

x6 Gasoline kL 1179741.78 1019686.72 268654.00 7591664.00

x7 Kerosene kL 612685.61 654523.86 60428.00 4092522.00

x8 Gas oil kL 896769.57 752560.76 140763.00 4807624.00

x9 Heavy oil kL 1050366.68 950189.23 57223.00 5793805.00

x10 City gas Million MJ 20480.33 40516.85 515.00 241405.00

x11 Butane gas Tons 103035.14 130104.53 4914.00 770696.00

x12 Propane gas Tons 211618.90 164136.00 39222.00 890332.00

x13 Coal 1000 tons 368.72 573.48 4.87 2664.36

x14 Coke 1000 tons 851.57 1621.14 0.47 7089.15

5

Our extension methods of real public and private capital stocks are the same asHonma and Hu (2008).

(4)

Table 2

Correlation coefﬁcients of input and output variables.

Real GDP 1.00

Employed persons 0.99 1.00 Private capital stock 0.99 0.98 1.00 Public capital stock 0.89 0.90 0.92 1.00 Electric power for residential use 0.96 0.97 0.96 0.92 1.00 Electric power for commercial and industrial use 0.92 0.94 0.93 0.85 0.96 1.00 Gasoline 0.90 0.92 0.92 0.88 0.96 0.95 1.00 Kerosene 0.67 0.70 0.67 0.85 0.69 0.62 0.71 1.00 Gas oil 0.86 0.89 0.86 0.92 0.89 0.87 0.92 0.88 1.00 Heavy oil 0.80 0.81 0.80 0.82 0.77 0.78 0.81 0.77 0.89 1.00 City gas 0.91 0.91 0.92 0.82 0.94 0.91 0.86 0.54 0.75 0.64 1.00 Butane gas 0.80 0.81 0.81 0.65 0.78 0.88 0.83 0.46 0.76 0.73 0.72 1.00 Propane gas 0.79 0.82 0.79 0.78 0.87 0.87 0.92 0.67 0.86 0.75 0.70 0.77 1.00 Coal 0.16 0.19 0.21 0.30 0.27 0.29 0.32 0.20 0.33 0.29 0.18 0.26 0.36 1.00 Coke 0.18 0.20 0.22 0.25 0.27 0.33 0.37 0.17 0.32 0.34 0.20 0.32 0.36 0.66 1.00 Table 3

Total-factor energy productivity index (TFEPI) for electric power for commercial and industrial use by region.

ID Region Area 93/94 94/95 95/96 96/97 97/98 98/99 99/00 00/01 01/02 02/03 Average Cumulative 01 Hokkaido A 0.957 1.007 0.976 0.954 1.064 0.984 0.987 0.995 0.980 0.993 0.989 0.898 02 Aomori B 0.999 1.004 0.978 1.000 0.972 0.984 0.965 0.989 0.940 0.976 0.981 0.823 03 Iwate B 0.959 0.982 0.991 0.946 1.016 0.992 1.009 0.976 0.977 0.993 0.984 0.851 04 Miyagi B 0.988 0.971 1.000 0.970 0.990 1.037 1.010 0.982 0.965 0.999 0.991 0.914 05 Akita B 1.017 0.997 1.056 0.928 0.963 0.976 0.958 1.008 0.978 1.010 0.988 0.890 06 Yamagata B 0.938 1.006 1.001 0.955 0.970 0.975 0.980 1.013 0.993 1.000 0.983 0.841 07 Fukushima B 1.001 0.993 1.008 1.135 0.963 1.031 0.990 1.010 0.978 0.985 1.008 1.087 08 Ibaraki C 0.944 0.958 0.967 0.615 1.516 0.900 0.919 0.812 1.065 0.999 0.947 0.582 09 Tochigi C 0.911 1.009 0.982 0.985 1.136 0.972 1.010 1.004 1.003 1.048 1.005 1.048 10 Gunma C 0.940 0.958 0.954 0.913 0.999 0.996 0.941 0.970 0.831 0.967 0.946 0.572 11 Saitama C 0.933 0.957 0.994 0.727 1.275 0.982 1.029 1.013 0.987 1.055 0.987 0.876 12 Chiba C 1.428 0.942 0.841 0.649 1.188 0.958 0.924 0.932 1.132 1.116 0.990 0.908 13 Tokyo C 0.971 1.036 1.006 1.085 0.974 0.980 0.991 1.032 0.983 1.081 1.013 1.139 14 Kanagawa C 0.618 0.953 0.970 1.123 0.863 1.043 1.022 0.990 1.148 1.203 0.979 0.807 15 Niigata D 1.552 0.983 0.931 0.854 1.008 0.963 0.914 1.113 0.991 1.030 1.020 1.223 16 Toyama D 0.937 1.020 1.029 0.945 0.968 0.941 0.927 0.909 0.933 1.011 0.961 0.673 17 Ishikawa D 0.960 1.044 1.003 0.976 0.988 1.003 0.980 1.016 0.982 1.006 0.996 0.956 18 Fukui D 0.956 1.016 1.019 0.979 1.025 0.968 0.952 1.006 0.978 1.004 0.990 0.904 19 Yamanashi D 0.909 1.007 1.003 1.019 0.953 1.024 1.003 0.981 0.987 1.010 0.989 0.896 20 Nagano D 0.968 1.021 1.011 0.937 1.015 0.986 1.005 1.009 0.960 1.006 0.991 0.917 21 Gifu D 1.036 0.979 1.012 1.030 1.024 0.934 0.929 0.942 0.922 1.101 0.989 0.898 22 Shizuoka D 0.915 0.964 0.994 0.627 1.320 0.879 0.867 0.899 0.964 1.011 0.930 0.484 23 Aichi D 1.215 1.033 1.067 1.438 0.641 1.069 1.054 1.051 0.886 1.064 1.033 1.379 24 Mie E 0.925 0.944 0.978 0.861 0.928 0.854 0.909 0.818 1.298 1.001 0.944 0.563 25 Shiga E 0.975 1.011 0.960 0.924 0.734 0.907 0.895 0.855 0.942 0.990 0.916 0.415 26 Kyoto E 0.848 0.986 0.955 0.974 1.204 1.025 1.043 1.035 0.961 1.040 1.003 1.034 27 Osaka E 1.186 1.022 1.008 0.870 1.036 0.933 1.055 0.960 0.923 0.913 0.987 0.877 28 Hyogo E 5.556 1.024 0.773 0.883 1.072 0.934 1.084 0.936 0.919 1.079 1.146 3.908 29 Nara E 0.965 1.055 1.012 0.979 0.968 1.005 0.999 1.018 1.007 1.009 1.001 1.013 30 Wakayama E 1.994 1.014 1.005 0.991 0.865 0.868 0.816 0.946 0.931 0.978 1.006 1.062 31 Tottori F 1.020 1.079 0.965 0.925 0.991 1.013 0.973 1.002 0.944 0.996 0.990 0.903 32 Shimane F 0.943 0.992 0.997 0.967 1.087 0.970 0.961 1.063 0.938 0.984 0.989 0.897 33 Okayama F 0.971 1.123 1.021 1.002 0.989 0.962 0.986 0.996 1.025 0.841 0.989 0.898 34 Hiroshima F 0.981 1.001 1.005 0.998 0.961 0.959 1.026 0.991 0.940 1.037 0.989 0.898 35 Yamaguchi F 2.880 1.032 1.048 1.144 0.968 0.906 1.099 1.006 1.026 0.975 1.132 3.453 36 Tokushima G 0.944 1.012 0.939 0.980 0.959 0.962 0.993 1.005 0.978 1.008 0.978 0.797 37 Kagawa G 0.969 1.020 0.980 0.938 0.996 0.951 0.966 1.000 0.994 0.999 0.981 0.827 38 Ehime G 1.006 0.985 0.952 0.925 0.981 0.963 0.996 0.999 0.971 1.019 0.979 0.812 39 Kochi G 0.989 1.011 0.971 1.006 0.994 1.029 1.008 0.961 0.981 0.996 0.994 0.944 40 Fukuoka H 3.092 0.969 0.995 1.019 0.961 0.953 1.002 0.988 0.976 1.023 1.106 2.749 41 Saga H 0.974 0.997 0.947 1.004 0.927 1.006 0.972 0.958 0.977 1.013 0.977 0.794 42 Nagasaki H 1.086 1.105 0.980 0.970 1.060 0.967 0.990 0.980 0.999 1.031 1.016 1.168 43 Kumamoto H 0.983 0.981 0.976 0.940 0.949 1.012 1.036 0.991 0.967 1.035 0.987 0.873 44 Oita H 0.986 0.990 1.016 1.024 1.000 0.978 1.050 0.982 1.011 1.009 1.004 1.045 45 Miyazaki H 1.026 1.168 0.987 0.914 1.002 1.019 0.980 0.972 1.005 0.965 1.002 1.019 46 Kagoshima H 0.946 0.978 0.942 0.947 0.948 1.003 1.050 0.993 0.955 0.993 0.975 0.776 47 Okinawa H 0.971 0.980 0.988 0.983 0.952 1.024 1.012 0.972 1.001 0.953 0.983 0.846 Summary 1.088 1.006 0.982 0.947 0.999 0.973 0.983 0.979 0.981 1.011 0.994 0.944 Note: A (Hokkaido), B (Tohoku), C (Kanto), D (Chubu), E (Kinki), F (Chugoku), G (Shikoku), and H (Kyushu).

(5)

4. Results

4.1. Total-factor energy productivity change in Japan

We calculated the total-factor energy productivity changes of

the four major types of energy: electric power for commercial and

industrial use, kerosene, heavy oil, and coal.

6

_{Tables 3–6}

_present

the total-factor energy productivity changes of regions in Japan for

1993–2003. From our ﬁndings, the average annual net total-factor

energy productivity changes of electrical power for commercial

and industrial use, kerosene, heavy oil, and coal for the period

from 1993 to 2003 were 0.6%, 0.9%, 0.1%, and 1.0%, respectively.

The TFEPIs of all these four energy sources deteriorate in periods

of 1998–1999 and 2001–2002, respectively. The other years

exhibit both improved and deteriorated TFEPIs. The TFEPIs for

each energy source generally remained largely unchanged in these

11 years.

Now we consider the trends of consumption amount and TFEPI

changes of each energy source during the sample period. If

consumption of each form of energy in 1993 is taken as 1, that of

electrical power for commercial and industrial use in 2003 is

1.166, that of kerosene is 0.998, that of heavy oil is 0.960, and that

of coal is 1.025. The forms of energy for which the TFEPI

deteriorates in the sample period have experienced increases in

consumption, whereas consumption has decreased for energy

sources with improved TFEPI.

By energy conservation law, large companies in Japan should

report their aggregated energy consumption on a crude oil

equivalent basis to the government and implement energy

conservation measures. However, our results indicate differences

Table 4

Total-factor energy productivity index (TFEPI) for kerosene by region.

6

There are 11 energy inputs in our model; however, we analyze four major energy sources within them.

(6)

in energy productivity changes between energy sources, which

tend to be disregarded in energy policy. The policy implications

ﬂowing from the results are that the former type of energy should

be improved or replaced by the latter since each energy source is

substitutable. From the available statistics, we know only that

partial-factor energy efﬁciency, i.e., the change of GDP per unit of

ﬁnal energy consumption (heating value), in Japan only increased

0.1% annually during the sample period.

7

_{Since the existing formal}

energy productivity indices in Japan are based on the aggregated

energy consumption in the partial-factor framework, our results

shed light on the total-factor productivity of individual sources of

energy.

At the regional level, only four (Tochigi, Aichi, Nara, and

Miyazaki) of the 47 regions enhanced their total-factor energy

productivities for all four kinds of energy during the sample

period, whereas two regions (Ibaraki and Shizuoka) showed

deterioration in all categories. The best and the worst performers

in the average TFEPI of the four energy sources are as follows. For

electric power for commercial and industrial use, while the TFEPI

in Hyogo increased 14.6% annually during 1993–2003, in Shiga it

declined 8.4%/year. For kerosene, the TFEPI in Okinawa increased

9.1%, whereas in Kanagawa it declined 5.2%. For heavy oil, the

TFEPI in Nara increased by 10.4%, and in Ibaraki it declined 13.3%.

For coal, the TFEPI in Miyazaki increased 23.1%, while in Kanagawa

it declined 30.1%.

8

It is possible that an improvement in energy

productivity in a region may be attributable to the changing

industrial structure in that region. We regard a prefecture as a

Table 5

Total-factor energy productivity index (TFEPI) for heavy oil by region.

7

This number is calculated by Japan’s Energy White Paper 2007 (Ministry of Economy, Trade and Industry, 2007).

8

Some coal scores have extreme values because of the unstable results of technical changes.

(7)

Table 6

Total-factor energy productivity index (TFEPI) for coal by region.

Table 7

Summary of annual change of total-factor energy productivity index (TFEPI) and its components for each energy source.

Energy source Index 93/94 94/95 95/96 96/97 97/98 98/99 99/00 00/01 01/02 02/03 Average Electric power for commercial and industrial use TFEPI 1.088 1.006 0.982 0.947 0.999 0.973 0.983 0.979 0.981 1.011 0.994

TFEE change 0.997 1.007 1.003 0.986 0.983 0.991 1.007 1.010 1.032 1.004 1.002 Technical change 1.092 0.999 0.978 0.961 1.016 0.982 0.976 0.969 0.951 1.007 0.992 Kerosene TFEPI 1.042 0.957 1.025 1.013 1.041 0.965 1.026 1.011 0.953 1.062 1.009 TFEE change 0.987 1.008 1.011 1.024 0.998 0.973 1.024 1.031 1.015 1.005 1.007 Technical change 1.056 0.950 1.013 0.989 1.042 0.993 1.002 0.981 0.938 1.056 1.001 Heavy oil TFEPI 1.002 0.986 0.993 1.011 1.065 0.976 0.999 0.991 0.992 0.999 1.001 TFEE change 1.003 0.993 1.005 0.983 1.011 0.989 1.006 0.988 1.033 0.976 0.999 Technical change 0.998 0.992 0.989 1.028 1.053 0.987 0.993 1.003 0.960 1.023 1.002 Coal TFEPI 0.981 0.984 0.920 0.970 1.141 0.999 0.999 0.922 0.951 1.051 0.990 TFEE change 0.937 0.974 0.901 1.030 1.026 0.953 1.054 0.975 1.046 1.006 0.989 Technical change 1.047 1.009 1.021 0.941 1.112 1.048 0.948 0.946 0.910 1.044 1.001

(8)

decision-making unit and look at such improvement cases in a

positive light.

4.2. Components of total-factor energy productivity growth

Now we decompose the TFEPI into total-factor energy

efﬁciency change and technical change. The former represents

the change in relative efﬁciency of energy consumption among 47

regions; the latter represents the shift in the technology of energy

use during one period.

Due to space limitations, we present only the summarized

results for decomposition of the four energy sources in

Table 7

.

9

The TFEPI of electric power for commercial and industrial use

changes by 0.6% annually. The TFEE change is 0.2%, while the

technical change is 0.8%. The decomposition makes clear that

the negative TFEPI change is caused by the technical change rather

than the efﬁciency change. The electricity results suggest that the

Japanese government should make more efforts in preventing

productivity degradation of electricity use rather than enhancing

energy productivity in relative inefﬁcient regions. The TFEPI of

kerosene changes by 0.9% annually, while the productivity

progress during the period is largely attributable to the TFEE

change of 0.7% rather than the technical change of 0.1%. The TFEPI

change of heavy oil, 0.1%, can be decomposed into a TFEE change

of 0.1% and a technical change of 0.2%. Finally, the TFEPI change

of coal, 1.0%, can be largely attributed to a TFEE change of 1.1%,

rather than a technical change of 0.1%. The coal results also imply

a wide dispersion in coal efﬁciency changes.

All indices except coal’s technical changes deteriorated in the

period from 1998 to 1999. Meanwhile, all indices except heavy

oil’s TFEPI and TFEE improved in the period from 2002 to 2003.

The TFEE changes of electricity and kerosene have been larger

than unity since 1999, implying that the regions that have been

inefﬁcient caught up with the efﬁcient regions with respect to the

above two energy sources. Whereas every TFEE change improved

Table 8

Number of times that each region was classiﬁed as an innovator by energy source.

ID Region Area Electric power for commercial and industrial use Kerosene Heavy oil Coal Total

01 Hokkaido A 2 0 1 1 4 02 Aomori B 1 0 3 0 4 03 Iwate B 2 3 6 0 11 04 Miyagi B 0 0 0 0 0 05 Akita B 3 1 5 3 12 06 Yamagata B 4 3 5 4 16 07 Fukushima B 5 7 3 3 18 08 Ibaraki C 0 1 1 1 3 09 Tochigi C 3 6 5 4 18 10 Gunma C 0 4 2 1 7 11 Saitama C 3 5 8 1 17 12 Chiba C 0 0 0 0 0 13 Tokyo C 5 5 5 5 20 14 Kanagawa C 1 0 2 1 4 15 Niigata D 0 0 0 0 0 16 Toyama D 0 3 2 4 9 17 Ishikawa D 5 4 3 5 17 18 Fukui D 2 7 7 6 22 19 Yamanashi D 1 6 6 6 19 20 Nagano D 4 4 4 6 18 21 Gifu D 0 3 3 0 6 22 Shizuoka D 1 5 0 4 10 23 Aichi D 0 1 0 0 1 24 Mie E 0 0 0 1 1 25 Shiga E 0 1 0 2 3 26 Kyoto E 1 4 5 0 10 27 Osaka E 0 2 4 2 8 28 Hyogo E 0 0 0 0 0 29 Nara E 6 7 9 6 28 30 Wakayama E 0 0 0 0 0 31 Tottori F 4 4 3 5 16 32 Shimane F 2 7 5 4 18 33 Okayama F 1 5 3 2 11 34 Hiroshima F 0 0 0 0 0 35 Yamaguchi F 0 0 0 0 0 36 Tokushima G 0 7 4 7 18 37 Kagawa G 1 4 4 3 12 38 Ehime G 2 5 1 5 13 39 Kochi G 4 6 7 0 17 40 Fukuoka H 0 0 0 0 0 41 Saga H 1 7 7 6 21 42 Nagasaki H 4 4 0 4 12 43 Kumamoto H 3 7 7 6 23 44 Oita H 6 6 2 0 14 45 Miyazaki H 3 8 6 5 22 46 Kagoshima H 1 6 2 6 15 47 Okinawa H 3 7 4 3 17 Average 1.787 3.511 3.064 2.596 10.957

Note: A (Hokkaido), B (Tohoku), C (Kanto), D (Chubu), E (Kinki), F (Chugoku), G (Shikoku), and H (Kyushu).

9

(9)

during period 2001–2002, every energy source’s technical change

deteriorated.

Now we consider whether certain regions shifted the frontier

over the course of the research period.

Fa¨re et al. (1994)

use a

component distance function in the technical change index to

deﬁne ‘innovators’ who cause the frontier to shift. Accordingly, we

can identify the innovators between period t and t+1 if that

region’s:

Total-factor energy technical change 41,

TFEE

t

tþ1

41 and

TFEE

tþ1

¼

1 Following the above deﬁnition,

Table 8

shows the number of

times each region became an innovator for each energy source.

The regions most frequently observed as innovators for each

energy source in the sample period are as follows: Nara and Oita

(six times each) for electricity for commercial and industrial use;

Miyazaki (eight times) for kerosene; Nara (nine times) for heavy

oil; and Tokushima (seven times) for coal. The above four regions

are located outside Japan’s four major industrial areas.

The eight regions containing Japan’s four major industrial

areas, i.e., Chiba, Kanagawa, Gifu, Aichi, Mie, Osaka, Hyogo, and

Fukuoka, tend not to be innovators frequently or at all for each

energy source. Only two regions in the industrial areas, i.e.,

Saitama and Tokyo, were often innovators in the sample period.

Tokyo was an innovator ﬁve times for each energy source.

Geographically, the 11 regions belonging to the Shikoku and

Kyushu areas, except Fukuoka, were frequently observed as

innovators. These 11 regions are located in rural areas of Japan.

To sum up these results, innovators were found only in a small

proportion of the regions dominated by manufacturing industry.

In other words, many of the regions that shift the frontier are not

developing mainly on the basis of manufacturing industry.

5. Concluding remarks

In this study, we used a new approach that combines the

concept of TFEE and the Malmquist productivity index to assess

energy productivity growth in regions in Japan between 1993 and

2003. We computed TFEPI for four representative energy sources

in a multiple-input framework to avoid single-input bias. This

enabled us to compute single-factor productivity under a

total-factor framework.

By separating out parts of TFEPI, we can identify both the

catching up and innovation effects. The former effect indicates

change in relative TFEE, and the latter effect indicates technical

change. We can identify factors that reduce the energy efﬁciency of

electric power for commercial and industrial use and coal. The TFEPI

of electricity changed by 0.6% annually, which can be separated

into a total-factor energy efﬁciency change of 0.2% and a technical

change of 0.8%. The TFEPI for coal deteriorated 1.0%/year, which

can be separated into a total-factor energy efﬁciency change

of 1.1% and a technical change of 0.1%. From our ﬁndings, we

conclude that deterioration in electricity efﬁciency is caused by a

decrease in technical change, whereas that for coal is caused by a

decrease in relative efﬁciency. The average annual net total-factor

energy productivity changes of kerosene and heavy oil were 0.9%

and 0.1%, respectively. Comparing TFEPI and consumption changes,

we ﬁnd that consumption of inefﬁcient energy sources (electric

power for commercial and industrial use and coal) has increased;

and in contrast consumption of efﬁcient sources (kerosene and

heavy oil) has decreased. The best overall performers with respect

to the TFEPIs for electric power for commercial and industrial

use, kerosene, heavy oil, and coal were Hyogo, Okinawa, Nara, and

Miyazaki, respectively; the worst performers were Shiga, Kanagawa,

Ibaraki, and Kanagawa, respectively.

We deﬁned and identiﬁed areas as ‘innovators’ that have

caused the frontier to shift. Many innovators are found in regions

outside Japan’s four major industrial areas. Saitama and Tokyo are

exceptions and are often observed as innovators. We conclude

that the regions that shift the frontier in Japan are those that are

not developing mainly on the basis of manufacturing industry.

Regions in the industrial areas should improve their energy use

technology and adjust their industrial structures.

In order to prevent depletion of natural resources and to meet

the Kyoto target, Japan should improve its energy productivity.

This new approach of total-factor energy productivity index serves

to advance these purposes. The most important issue for future

research is to examine what factors inﬂuence energy productivity.

The relationships among energy price, per capita income, and

energy productivity should be evaluated.

Acknowledgments

The authors thank the seminar participants at the Annual

Meeting of the Society for Environmental Economics and Policy

Studies at Osaka University, the Annual Meeting of the Japanese

Association for Applied Economics at Kanazawa University, the

Conference on Productivity, Efﬁciency and Industry Development,

and Nankai University. The usual disclaimer applies.

References

Ali, A.I., Seiford, L.M., 1993. The mathematical programming approach to efﬁciency measurement. In: Fried, H., Lovell, K., Schmidt, S. (Eds.), The Measurement of Productive Efﬁciency: Techniques and Applications. Oxford University Press, Oxford.

Berndt, E.R., 1990. Energy use, technical progress and productivity growth: a survey of economic issues. Journal of Productivity Analysis 2, 67–83. Boussemart, J.P., Briec, W., Kerstens, K., Poutineau, J.C., 2003. Luenberger and

Malmquist productivity indices: theoretical comparisons and empirical illustration. Bulletin of Economic Research 55, 391–405.

Caves, D.W., Christensen, L.R., Diewert, W.E., 1982. The economic theory of index numbers and the measurement of input, output, and productivity. Econome-trica 92, 73–86.

Charnes, A., Cooper, W.W., Rhodes, E., 1978. Measuring the efﬁciency of decision making units. European Journal of Operational Research 2, 429–444. Fa¨re, R., Grosskopf, S., Lovell, C.A.K., 1985. The Measurement of Efﬁciency of

Production. Kluwer-Nijhoff, Boston.

Fa¨re, R., Grosskopf, S., Norris, M., Zhang, Z., 1994. Productivity growth, technical progress, and efﬁciency change in industrialized countries. American Econom-ic Review 84, 66–83.

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).

Han, Z.Y., Fan, Y., Jiao, J.L., Yan, J.S., Wei, Y.M., 2007. Energy structure, marginal efﬁciency and substitution rate: an empirical study of China. Energy 32, 935–942. Hanley, N., McGregorb, P.G., Swalesb, J.K., Turner, K., 2009. Do increases in energy

efﬁciency improve environmental quality and sustainability? Ecological Economics 38, 692–709.

Honma, S., Hu, J.L., 2008. Total-factor energy efﬁciency of regions in Japan. Energy Policy 36, 821–833.

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.

Japan Business Federation, 2008. Results of the ﬁscal 2008 follow-up to the Keidanren Voluntary Action Plan on the environment.

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

Kasahara, S., Paltsev, S., Reilly, J., Jacoby, H., Ellerman, D., 2007. Climate change taxes and energy efﬁciency in Japan. Environmental and Resource Economics 37, 377–410.

(10)

Ministry of Economy, Trade and Industry, 2007. Energy White Paper 2007. /http:// www.enecho.meti.go.jp/topics/hakusho/2007energyhtml/index2007.htmS. Miyara, I., Fukushige, M., 2008. The types of public capitals and their productivity

in Japanese prefecture. Japanese Economic Review 59, 194–210.

Nemoto, J., Goto, M., 2005. Productivity, efﬁciency, scale economies and technical change: a new decomposition analysis of TFP applied to the Japanese prefectures. Journal of the Japanese and International Economies 19, 617–634.

Organisation for Economic Co-operation and Development, 2007. Energy Balances of OECD Countries, 2004/2005, Paris.

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

So¨derholm, P., Pettersson, F., 2008. Climate policy and the social cost of power generation: impacts of the Swedish national emissions target. Energy Policy 36, 4154–4158.