"Analysis of Energy Intensity and Its Determinants in 16 OECD Countries," by Musiliu O. Oseni

THE JOURNAL OF ENERGY
AND DEVELOPMENT
Musiliu O. Oseni,
Analysis of Energy Intensity and
Its Determinants in 16 OECD Countries,

Volume 35, Number 1
Copyright 2011
www.iceed.org
ANALYSIS OF ENERGY INTENSITY AND ITS
DETERMINANTS IN 16 OECD COUNTRIES
Musiliu O. Oseni*
Introduction
G
lobal consumption of energy has increased tremendously over the last three
decades with total consumption of the Organization for Economic Co-
operation and Development (OECD) accounting for about 46 percent of the world
total in 2007.
1
The total OECD energy consumption rose at an average rate of 1.62
percent per annum in the 1970s and 1.74 percent per annum in the 1990s. How-
ever, in the 2000s consumption grew by just 0.71 percent annually. Similarly, the
world total energy consumption has been increasing at annual rates of 2.94 per-
cent, 1.19 percent, and 2.90 percent in the 1970s, 1990s, and 2000s, respectively.
Despite this tremendous increase in energy consumption, however, there has
been a downward trend in the ratio of energy consumption to gross domestic
product (GDP)often referred to as energy intensity. As shown in figure 1, for
instance, the world total aggregate energy intensity, which stood at 81 percent of
its 1971 level in 1988, decreased to 64 percent of its 1971 level in 2007. Similarly,
the OECD aggregate intensity accounted for 71 percent of its 1971 level in 1988
and declined to just 53 percent of its 1971 figure in 2007.
*Musiliu O. Oseni has been a lecturer in the Department of Economics of Al-Hikmah University
Ilorin, Nigeria since 2008. He holds an undergraduate degree in economics from the University of
Ibadan, Nigeria, and a masters degree in energy economics and policy from the University of
Surrey, United Kingdom, where his study was sponsored by the Petroleum Technology De-
velopment Fund (PTDF) of Nigeria. This research is based on his M.Sc. thesis and supported by
Lester C. Hunt, Professor of Energy Economics at the Surrey Energy Economics Centre (SEEC).
The views expressed are solely those of the author.
The Journal of Energy and Development, Vol. 35, Nos. 1 and 2
Copyright 2011 by the International Research Center for Energy and Economic Development
(ICEED). All rights reserved.
101
Apart from this reduction in energy intensity, it is not clear yet whether this
reduction is a result of improvements in energy use (efficiency intensity) or
a product of fundamental shifts in the structure of economies (structural effects or
activity intensity). Furthermore, understanding the key economic factors driving
changes in these indexes (especially in the industrialized economies) becomes
important for the implementation of appropriate energy policy.
The existing literature on energy-intensity studies has focused largely on
methodologies for decomposing aggregate energy intensity, to separate the effects
of improvements in the use of energy from the fundamental shifts in the structure
of an economy. The focus over the last three decades has been the comparison of
the methods linked with the Laspeyres index with those based on the Divisia
index.
2
While many studies have addressed the methodology for decomposing energy-
intensity indexes, little is known about the key economic forces that propelled
changes in these indexes over time. Most of the few earlier studies that attempted
to investigate these forces restricted their analysis to aggregate intensity only and
Figure 1
AGGREGATE ENERGY INTENSITY INDEX ORGANIZATION FOR ECONOMIC
COOPERATION AND DEVELOPMENT (OECD) VS. THE WORLD, 1971-2007
(Index 1971=1)
Source: International Energy Agency (IEA), World Energy Balance and Energy Balance of
OECD Countries (Paris: IEA, 2009), available at www.iea.org.
THE JOURNAL OF ENERGY AND DEVELOPMENT 102
could not provide the details of the channel through which these economic vari-
ables affect the aggregate intensity.
3
Furthermore, few studies that have extended
their analysis of the key drivers to efficiency and activity components are country
specific, either in time-series context or in panel context. To the best of our
knowledge, such analysis combining different countries in a panel form (espe-
cially in the selected OECD countries) is lacking in the literature.
Premised on the links between energy intensity and global change, under-
standing the key economic variables that cause changes in aggregate, efficiency,
and structural or activity components of energy intensity in industrialized
economies over a relatively longer time period may contribute to the formu-
lation of policy on this global phenomenon. Therefore, this study adopts the
Fisher ideal index to analyze the relative importance of fundamental im-
provements in energy use and structural shifts in the economy to changes in
aggregate energy intensity in the selected OECD countries (Austria, Belgium,
Canada, Denmark, France, Greece, Ireland, Italy, Japan, the Netherlands,
Norway, Portugal, Spain, Sweden, United Kingdom, and United States). Also,
the most salient economic factors driving changes in the aggregate, efficiency,
and activity components of energy intensity in the selected OECD countries are
investigated.
We shall begin with a literature review on index-decomposition analysis
(IDA), which is followed by the methodology on decomposition and regression
analysis. Then the paper presents and evaluates the results from decomposition
technique and concludes with a summary of the policy issues and findings.
Literature Review
The changes in the composition of economic activities and their effects on
aggregate measures of energy intensity have been the subject of empirical analysis
since the 1978 volume by J. Myers and L. Nakamura.
4
Thereafter, the number of
studies in this area has increased significantly, and there has been development in
terms of both methodology and application. The survey by B. W. Ang and F. Q.
Zhang published in 2000 found a total of 124 studies compared to a total number
of 51 studies reported by B. W. Ang in 1985 and eight studies reported in H. G.
Huntington and J. G. Myers in 1987.
5
The early approach originating from the 1970s computed the impact of
structural change on energy intensity by subtracting the observed base-year ag-
gregate energy intensity from the hypothetical aggregate energy intensity that
would have existed for a target year had sectoral energy intensities remained
unchanged at their respective values in a base year. On the other hand, the dif-
ference between the observed aggregate intensity and the hypothetical aggregate
energy intensity in the target year is attributed to efficiency intensity changes.
ENERGY INTENSITY IN OECD COUNTRIES 103
Examples of studies that have adopted this approach include E. Bossanyi, G. A.
Hankinson and J. M. W. Rhys, J. Myers and L. Nakamura, and G. O
stblom.
6
However, the recent approach is based on an index number system. The ap-
plication of an index number has a long history in economics with Irvin Fisher
being a key contributor to the literature.
7
Specifically, G. A. Boyd et al. were the
first researchers to apply index-number theory to energy-intensity decomposition
by adopting a Divisia index number methodology.
8
However, this decomposition
method, like earlier techniques (which were essentially based on a Laspeyres
index), had residual terms that could account for the variability in the underlying
index of energy-intensity changes.
9
Subsequently, studies using some sort of index-
decomposition techniques have increased tremendously.
10
G. A. Boyd and J. M. Roop adopt a modified Fisher index (known as the Fisher
ideal index) to energy decomposition analysis.
11
This method perfectly de-
composes changes in energy intensity into changes in energy efficiency and
economic mix (activity). The Fisher ideal index satisfies the basic index-number
theory properties, such as factor reversal (perfect decomposition), time reversal,
proportionality, and aggregation test properties. As a consequence, the Fisher ideal
index eliminates the residual problem suffered by many previous energy-intensity
decomposition methods. Additionally, the Fisher ideal index is zero-negative
robust, a desirable property that is not adequately possessed by any of the
remaining perfect index-decomposition methods.
12
Based on this unique property
of this method, several studies have either adopted or recommend it as a desirable
method for energy-intensity decomposition. For instance, B. W. Ang points out
that the Fisher ideal index can be a better methodology under certain conditions
as may be dictated by the data, especially when negative or zero values are
involved.
13
Methodology
Index-Number Methodology and Energy-Intensity Decomposition: Following
W. E. Diewert, the basic price-quantity index number problem arises from finding
the appropriate way to express the change in a value aggregate,
14
V
T
V
0
=
n
l =1
P
i;T
q
i;T
n
i = 1
p
i;0
q
i;0
in the form of two functions P (price) and Q (quantity) such that
V
T
V
0
=P p
0
; p
T
; q
0
; q
T
Q p
0
; p
T
; q
0
; q
T
;
where p
i,T
and q
i,T
are the price and quantity, respectively, of commodity i at time
(0,T ).
To form a value aggregate (in the form of the price-quantity index problem),
energy-intensity decomposition application of the index-number approach relies
on the identity:
E
t
=
n
i =1
E
it
=
Y
it
_ _
Y
it
[
n
i = 1
e
it
Y
it
1:1
where E
t
and E
it
are, respectively, aggregate and sectoral energy consumption, Y
it
is the measure of sectoral activity (e.g., value added of output), and e
it
denotes
sectoral energy intensity. Recall that energy intensity is the ratio of energy used to
unit of output.
Theoretically, the aggregate energy intensity (e
t
) can be computed by dividing
equation (1.1) by some aggregate measure of activity (Y
t
) and defining the ratio of
the i
th
sectors measure of activity to the aggregate activity at time t as s
it
.
e
t
=
E
t
=
Y
t
=
n
i = 1
E
it
=
Y
it
_ _
Y
it
=
Y
t
_ _
=
n
i = 1
e
it
s
it
1:2
where, again, E
t
is the aggregate energy consumption in year t, E
it
is the energy
consumed in sector i in year t, Y
t
represents GDP in year t, Y
it
is the measure of
economic activity in sector i in year t, and n is the number of sectors. Equation (1.2)
disaggregates energy intensity into sector-specific energy efficiency and sectoral
activity. It is important to note that, while it is required that energy consumption in
the sectors must sum to aggregate energy consumption, measures of economic
activity need not sum to GDP; in fact, they need not be in the same units.
Denoting the base-year aggregate energy intensity as e
0
, the aggregate energy-
intensity index can be computed as e
t
/ e
o
. Following the link between the price-
quantity index problem and the energy-intensity decomposition problem, and by
focusing on the energy identity in equation (1.2) and paraphrasing Diewert, we can
express the change in energy aggregate
e
t
e
0
=
n
i =1
e
it
s
it
e
i 0
s
i0
1:3
in the form of two functions Eff and Act, such that
e
t
e
0
=Eff e
i0
; e
it
; s
i0
; s
it
Act e
i0
; e
it
; s
i0
; s
it
1:4
where e
it
and s
it
are the energy and activity intensities, respectively, of sector i at
time t, and where the index functions, Eff and Act, represent aggregate intensity
and activity change, respectively (and they are respectively analogous to price and
quantity in the original index-number problem).
Thus, following Diewert, the energy-intensity index can be decomposed
into efficiency and activity indexes, provided that the sectors that account for
all energy use in the economy (without overlap) and the measures of eco-
nomic activity (Y
it
) with which to construct a measure of energy intensity are
available.
Relying on the links between the price-quantity index problem and energy-
intensity decomposition problem established from equations (1.1) to (1.4), the
Fisher index can be used to compute aggregate energy intensity into efficiency and
activity indexes.
15
This index has been used in the energy-intensity literature based
on its desirable properties and can be constructed by first computing Laspeyres
and Paasche indexes.
16
The Laspeyres indexes for both efficiency (eff) and activity (act) indexes are
L
eff
t
=
n
i
e
it
s
i0
n
i
e
i0
s
i0
1:5
L
act
t
=
n
i = 1
e
i0
s
it
n
i = 1
e
i0
s
i0
1:6
and the Paasche indexes can be computed as
P
eff
t
=
n
i = 1
e
it
s
it
n
i = 1
e
i0
s
it
1:7
P
act
t
=
n
i = 1
e
it
s
it
n
i =1
e
it
s
i0
1:8
where e
i0
and s
i0
represent base-year values of efficiency and activity indexes, and
e
it
and s
it
are their current-year indexes, respectively. The Laspeyres index uses
a base period weight while the Paasche index utilizes current period weight. By
multiplying equations (1.2) and (1.4) and expressing them in geometric mean
yields the Fisher ideal index for efficiency index. A similar process for equations
(1.3) and (1.5) gives activity index.
F
eff
t
= L
eff
t
P
eff
t
_ _1
=
2
1:9
F
act
t
= L
act
t
P
act
t
_ 1
=
2
1:10
Therefore, the aggregate energy-intensity index can be expressed as
e
t
e
0
[I
t
=F
eff
t
F
act
t
: 1:11
Equation (1.11) suggests a way to attribute changes in energy consumption arising
from changes in energy intensity. Denoting energy savings (DE
t
) due to im-
provements in energy intensity as
DE
t
=E
t

E
t
1:12
E
t
is actual energy consumption while

E
t
is the hypothetical energy consumption
that would have existed had energy intensity remained at base-year level. Fol-
lowing Metcalf, these energy savings can be attributed to improvements in effi-
ciency and fundamental shifts in economic activities as follows:
DE
t
=DE
t
ln F
eff
t
_ _
ln I
t

_
_
_
_
+DE
t
ln F
act
t
_ _
ln I
t

_ _
=DE
eff
t
+DE
act
t
: 1:13
Econometric Model: The decomposition analysis undertaken in the above sec-
tion is a purely descriptive exercise. It demonstrates how energy consumed per
unit of output has changed over time in different sectors of the economy but cannot
provide the details of the forces driving these changes. To better investigate the
main drivers of changes in energy intensity, efficiency, and structural shifts, this
section looks at the link between these indexes and key economic variables using
a general-to-specific approach in estimating the impacts of various economic
factors on intensity indexes. Begin with a simple static panel model:
I
it
=b
0
+ay
it
+py
2
it
+fp
it
+gu
it
+u
it
2:1
where I
it
represents aggregate intensity, efficiency index, or activity index (all in
natural log) for the ith country in year t, and y
it
and p
it
are natural logs of real
income (GDP) per capita and real energy price for country i in year t, respectively.
y
2
it
is the log-income squared, which accounts for the possible non-linearity in the
response of energy use to income; u
it
represents the growth rates of population
in country i in year t to capture the response of energy use to change in pop-
ulation;
17
u
it
is the error term that is assumed to be u
it
= m
i
+ n
it
, where m
i
and
n
it
are unobservable individual-specific effects and a remainder disturbance,
respectively.
18
However, one difficulty with equation (2.1) is that the response of energy in-
tensity to changes in economic variables is assumed to be immediate. More re-
alistically, since energy demands response to price or income changes depends on
the nature of appliances and capital stocks (which slows down its response to these
economic factors), energy intensity is also likely to respond to these variables with
some lags. Therefore, a more realistic general model (dynamic panel model) with
one lagged-dependent and two lagged-independent, i.e., ARDL (1, 2), is adopted.
The method adopted follows a general-to-specific methodology where the most
insignificant variables are deleted until a preferred model is achieved. Equation
(2.1) can be re-written as
A L I
it
=b
0
+B L y
it
+ C L y
2
it
+D L p
it
+E L u
it
+u
it
2:2
where A(L) is the polynomial lag operator 1 d
1
L, B(L) is the polynomial lag
operator a
0
+a
1
L+a
2
L
2
, C(L) is the polynomial lag operator p
0
+ p
1
L + p
2
L
2
,
D(L) is the polynomial lag operator f
0
+ f
1
L + f
2
L
2
, and E(L) is the lag operator
g
0
+ g
1
L + g
2
L
2
. The short-run price and income elasticities are f
0
and a
0
+
2p
0
y
it
, respectively, while their long-run estimates are D(L)/A(L) and [B(L) +
2C(L)y
it
]/A(L), respectively.
Introducing lagged-dependent variables causes problem in the panel data
model. This is because m
i
is correlated with I
it
and, hence, with (L)I
it
(i.e., I
it1
) so
that the regressors and error terms are correlated, thereby rendering ordinary least
squares (OLS) biased and inconsistent.
19
T. W. Anderson and C. Hsiao suggest
first differencing equation (2.2) to remove m
i
and then using DI
it2
= (I
it2
I
it3
) or
simply I
it2
as an instrument for DI
it1
.
20
These instruments provide consistent but
not necessarily efficient estimates.
21
M. Arellano and S. Bond propose a generalized method of moments (GMM)
procedure, which is considered to be more efficient than the T. W. Anderson and
C. Hsiao estimator.
22
This method, by utilizing the orthogonality conditions
existing between lagged values of I
it
and the disturbances n
it
, generates consistent
estimates when the number of countries N and/or the time T goes to infinity. The
GMM method is applied to equation (2.2) as follows.
Taking the first difference of equation (2.2) yields
A L DI
it
=B L Dy
it
+C L Dy
2
it
+D L Dp
it
+E L Du
it
+Dn
it
2:3
i =1; 2; . . . N; t =3; 4; . . . T:
It is possible that DI
it1
in equation (2.3) is correlated with Dn
it1
through the
correlation between I
it1
and n
it1
. But there exist qualified instrumental variables
that are correlated with DI
it1
and not with Dn
it
; these variables include {I
i1
, I
i2
,
. . . , I
it2
} and serve as instruments for DI
it1
. Using a matrix of these instruments
and performing the generalized least squares (GLS) estimation method on equa-
tion (2.3) yields one-step consistent GMM estimator as
^
d
^
b
_ _
= DI
1
; DX 9Z Z9 d
N
G Z
1
Z9 DI
1
; DX
_ _
1
3 DI
1
; DX 9Z Z9 d
N
G Z
1
Z9DI
_ _
where
G=
2 1 0 . . . . . . 0 0 0
1 2 1 . . . . . . 0 0 0

0 0 0 . . . . . . 1 2 1
0 0 0 . . . . . . 0 1 2
_
_
_
_
is a T 23T 2
matrix and d is an identity matrix.
23
Fortunately, the application of this method is
readily available in the STATA package.
24
Data, Estimation, and Interpretation of Results
Sources of Data: This study utilizes a panel data set covering the period of 1975-
2007 for 16 OECD countries: Austria, Belgium, Canada, Denmark, France,
Greece, Ireland, Italy, Japan, Netherlands, Norway, Portugal, Spain, Sweden, the
United Kingdom, and the United States. These countries and the time periods are
selected based on the availability of data on sectoral value added of output and the
significant trends in the countries energy consumption over these periods. For its
empirical analysis, this study requires data on aggregate and sectoral energy de-
mand, aggregate energy prices, and macroeconomic indicators such as GDP,
sectoral value added of output, aggregate personal consumption expenditures,
population, and population growth rates from 1975 to 2007. Two different sets of
GDP data are used: GDP at constant 2000 U.S. dollars (sourced from the World
Bank database) is used for the decomposition analysis, while GDP at 2000 U.S.
dollars purchasing power parity (PPPs), sourced from the International Energy
Agency (IEA), is used for the regression analysis.
25
Data on energy demand (in
thousand tons of oil equivalent or ktoe), the index of real energy price (2000=100),
and GDP at 2000 U.S.$ PPPs are obtained from the IEA database World Energy
Balance and Balances of OECD Countries.
26
However, the IEA database only
provides data on price from 1978-2007; the remaining data prior to 1978 are
computed using the method in O. I. Adeyemi et al.
27
This method involves
weighting gas in households and industry, coal in households and industry, electricity
in households and industry, gasoline, diesel fuel, and kerosine by their fuel con-
sumption shares. This produces a real aggregate energy price index for each country
in 1972 prices (1972=100) over the period 1970-1980. The two series (1970-1980
with 1972=100 and 1978-2007 with 2000=100) subsequently are spliced using the
ratio from the overlap year 1978 to obtain the real energy price index for the whole
period 1975 to 2007 at 2000 prices (2000=100). The macroeconomic indicators
(except GDP at 2000 U.S.$ PPPs) are obtained from the World Bank database.
Summary Statistics of the Data for Decomposition Analysis: Table 1 presents
summary statistics on the data used for the construction of indexes. The measures
of economic activity related to the underlying demand for energy within each
sector are used. A key driver for residential energy demand is personal con-
sumption expenditures.
28
This is preferable to disposable income since a portion of
disposable income goes to savings, which have no appreciable effect on residential
energy consumption. For the remaining sectors, value added in each sector is used.
Value added is the best measure of the contribution of a given sector to final
production. For the measure of energy consumption, the sectoral energy demand
(in ktoe) is used. On average, the transportation sector represents the highest
energy-consuming sector followed by that of industry. The reason for this may not
be unconnected with the recent increase in demand for automobiles in developed
countries. The standard deviation also suggests a considerable variation in the
transport and industrial energy demand across the countries. This perhaps may be
due to the variation in the transportation services, the demand for automobiles, and
the level of industrialization among the countries. On the other hand, the com-
mercial sector is the least energy-consuming activity.
Energy Intensity, Efficiency, and Activity Indexes: Figure 2 shows the results of
the decomposition analysis for the selected 16 OECD countries taking 1975 as the
base year for the analysis. There is a wide variation in the structure of the intensity
indexes across the nations. In other words, there are considerable variations in the
relative importance of fundamental improvements in energy use and changes in
economic structures among the countries. For instance, the U.S. aggregate energy
intensity in 2007 was 46 percent of its level in 1975; the activity index was 83
percent of its level in 1975 while the efficiency index was 56 percent of its 1975
level. In other words, had the composition of economic activity remained un-
changed between 1975 and 2007, energy intensity would have been 56 percent that
of 1975. Similarly, had energy efficiency remained constant at its 1975 levels for
all sectors, changes in economic activity would have led to about a 17-percent
decrease in energy intensity. By implication, 39 percent of improvements in the
U.S. energy intensity as at 2007 were due to improvements in energy efficiency
while the remaining 15 percent was accountable to changes in economic structure.
T
a
b
l
e
1
S
E
C
T
O
R
S
F
O
R
D
E
C
O
M
P
O
S
I
T
I
O
N
A
N
A
L
Y
S
I
S
a
E
c
o
n
o
m
i
c
A
c
t
i
v
i
t
y
S
e
c
t
o
r
a
l
E
n
e
r
g
y
D
e
m
a
n
d
S
e
c
t
o
r
M
e
a
s
u
r
e
(
i
n
b
i
l
l
i
o
n
2
0
0
0
$
)
M
e
a
n
S
t
a
n
d
a
r
d
D
e
v
i
a
t
i
o
n
M
e
a
s
u
r
e
(
i
n
k
t
o
e
)
a
M
e
a
n
S
t
a
n
d
a
r
d
D
e
v
i
a
t
i
o
n
R
e
s
i
d
e
n
t
i
a
l
A
g
g
r
e
g
a
t
e
P
C
E
a
5
6
7
.
7
1
,
2
0
6
.
7
E
n
e
r
g
y
c
o
n
s
u
m
p
t
i
o
n
2
8
,
7
1
4
.
5
5
5
,
4
2
8
.
7
C
o
m
m
e
r
c
i
a
l
a
V
a
l
u
e
a
d
d
e
d
8
2
8
.
4
1
,
5
8
2
.
8
E
n
e
r
g
y
c
o
n
s
u
m
p
t
i
o
n
2
2
,
5
8
8
.
9
4
6
,
9
3
3
.
6
I
n
d
u
s
t
r
i
a
l
a
V
a
l
u
e
a
d
d
e
d
2
6
7
.
5
4
7
4
.
2
E
n
e
r
g
y
c
o
n
s
u
m
p
t
i
o
n
4
1
,
1
3
1
.
1
7
7
,
9
0
0
.
4
T
r
a
n
s
p
o
r
t
a
t
i
o
n
V
a
l
u
e
a
d
d
e
d
1
3
.
5
1
4
.
2
E
n
e
r
g
y
c
o
n
s
u
m
p
t
i
o
n
4
8
,
1
1
1
.
1
1
1
8
,
1
8
7
.
3
T
o
t
a
l
G
D
P
a
9
9
7
.
8
1
8
9
1
.
8
E
n
e
r
g
y
c
o
n
s
u
m
p
t
i
o
n
1
4
0
,
5
4
5
.
6
2
9
5
,
2
7
4
.
5
a
D
a
t
a
f
r
o
m
1
9
7
5
t
o
2
0
0
7
f
o
r
1
6
c
o
u
n
t
r
i
e
s
o
f
t
h
e
O
r
g
a
n
i
z
a
t
i
o
n
f
o
r
E
c
o
n
o
m
i
c
C
o
o
p
e
r
a
t
i
o
n
a
n
d
D
e
v
e
l
o
p
m
e
n
t
(
O
E
C
D
)
:
A
u
s
t
r
i
a
,
B
e
l
g
i
u
m
,
C
a
n
a
d
a
,
D
e
n
m
a
r
k
,
F
r
a
n
c
e
,
G
r
e
e
c
e
,
I
r
e
l
a
n
d
,
I
t
a
l
y
,
J
a
p
a
n
,
N
e
t
h
e
r
l
a
n
d
s
,
N
o
r
w
a
y
,
P
o
r
t
u
g
a
l
,
S
p
a
i
n
,
S
w
e
d
e
n
,
U
n
i
t
e
d
K
i
n
g
d
o
m
,
a
n
d
U
n
i
t
e
d
S
t
a
t
e
s
.
T
h
e
i
n
d
u
s
t
r
i
a
l
s
e
c
t
o
r
c
o
r
r
e
s
p
o
n
d
s
t
o
I
S
I
C
(
I
n
t
e
r
n
a
t
i
o
n
a
l
S
t
a
n
d
a
r
d
I
n
d
u
s
t
r
i
a
l
C
l
a
s
s
i
f
i
c
a
t
i
o
n
)
d
i
v
i
s
i
o
n
s
1
0
-
4
5
a
n
d
i
n
c
l
u
d
e
s
m
a
n
u
f
a
c
t
u
r
i
n
g
,
m
i
n
i
n
g
,
c
o
n
s
t
r
u
c
t
i
o
n
,
e
l
e
c
t
r
i
c
i
t
y
,
w
a
t
e
r
,
a
n
d
g
a
s
.
T
h
e
c
o
m
m
e
r
c
i
a
l
s
e
c
t
o
r
i
n
c
l
u
d
e
s
c
o
m
m
u
n
i
c
a
t
i
o
n
,
w
h
o
l
e
s
a
l
e
a
n
d
r
e
t
a
i
l
t
r
a
d
e
,
f
i
n
a
n
c
e
,
s
e
r
v
i
c
e
s
,
a
n
d
g
o
v
e
r
n
m
e
n
t
.
k
t
o
e
=
t
h
o
u
s
a
n
d
t
o
n
s
o
f
o
i
l
e
q
u
i
v
a
l
e
n
t
.
P
C
E
=
p
e
r
s
o
n
a
l
c
o
n
s
u
m
p
t
i
o
n
e
x
p
e
n
d
i
t
u
r
e
s
.
G
D
P
=
g
r
o
s
s
d
o
m
e
s
t
i
c
p
r
o
d
u
c
t
.
S
o
u
r
c
e
s
:
E
n
e
r
g
y
c
o
n
s
u
m
p
t
i
o
n
d
a
t
a
f
r
o
m
t
h
e
O
r
g
a
n
i
z
a
t
i
o
n
f
o
r
E
c
o
n
o
m
i
c
C
o
o
p
e
r
a
t
i
o
n
a
n
d
D
e
v
e
l
o
p
m
e
n
t
(
O
E
C
D
)
,
I
n
t
e
r
n
a
t
i
o
n
a
l
E
n
e
r
g
y
A
g
e
n
c
y
(
I
E
A
)
,
E
n
e
r
g
y
B
a
l
a
n
c
e
o
f
O
E
C
D
C
o
u
n
t
r
i
e
s
,
a
v
a
i
l
a
b
l
e
a
t
w
w
w
.
i
e
a
.
o
r
g
.
E
c
o
n
o
m
i
c
d
a
t
a
a
r
e
f
r
o
m
t
h
e
W
o
r
l
d
B
a
n
k
,
W
o
r
l
d
D
e
v
e
l
o
p
m
e
n
t
I
n
d
i
c
a
t
o
r
s
,
a
v
a
i
l
a
b
l
e
a
t
w
w
w
.
w
o
r
l
d
b
a
n
k
.
o
r
g
.
Among other countries whose improvements in energy intensity were largely due
to fundamental improvements in energy use are Austria, Belgium, Denmark, and
Sweden. On the other hand, changes in economic activity are the important drivers
of aggregate energy intensity among the remaining OECD countries: Canada,
France, Ireland, Italy, Japan, Netherlands, Norway, and the United Kingdom (all
with evidence of improvements in energy intensity), and Greece, Portugal, and
Spain (with a deterioration in energy intensity). For instance, the result shows that
the 50-percent improvement in energy intensity in the United Kingdom between
1975 and 2007 was due solely to structural changes in the economy while the
sectoral energy efficiency remained approximately at its 1975 level. In Canada,
France, and Netherlands, activity indexes stood at 60 percent, 68 percent, and 58
percent of their 1975 levels while the efficiency indexes were 96 percent, 91
percent, and 96 percent of their 1975 levels, respectively. That is, aggregate in-
tensity would have improved by 40 percent, 32 percent, and 42 percent, re-
spectively, had efficiency intensity remained constant. Meanwhile, energy
intensity would have improved by just 4 percent, 9 percent, and 4 percent, re-
spectively, had the economic structure remained unchanged.
Finally, figure 3 presents the graph of the overall aggregate intensity, effi-
ciency, and activity for the 16 countries.
29
The aggregate intensity for the 16
OECD members in 2007 was 55 percent of its level in 1975. This shows that, on
average, the aggregate energy intensity in the countries improved by 45 percent
between 1975 and 2007. The activity index in 2007 was 66 percent of its level in
1975, while the efficiency index stood at 83 percent of its 1975 level. In other
words, had the composition of the economy remained at its 1975 level, the energy
intensity would have been 83 percent of its 1975 level. On the other hand, had
energy efficiency remained constant between 1975 and 2007, shifts in economic
activity would have led to a 34-percent decline in energy intensity. By implication,
30 percent (about two-thirds) of improvements in aggregate intensity in the OECD
countries was due to structural shifts in their economies while the remaining 15
percent (one-third) was contributed by fundamental improvements in energy use.
Therefore, structural shifts in the economic activities in the countries led to greater
improvements in energy intensity than the fundamental improvements in energy use.
In other words, the improvements in energy intensity in the selected countries were
largely due to movements away from energy-intensive activities (e.g., manufacturing
sector) to less energy-intensive services (e.g., commercial services).
Analysis of Changes in Energy Index Values: Table 2 provides summary statistics
on the three indexes for the 16 OECDmembers for various years between 1980 and
2007. Each row provides statistics across the countries for a given year of the three
indexes. The top panel of the table shows trends over time and variation across
countries for the aggregate intensity index. Several facts are worth noting. First, the
average energy intensity declined at an annual rate of 0.92 percent between 1970
Figure 2
TRENDS OF ENERGY INTENSITY, EFFICIENCY, AND ACTIVITY
INDEXES FOR 16 COUNTRIES OF THE ORGANIZATION FOR
ECONOMIC COOPERATION AND DEVELOPMENT
(OECD), 1975 TO 2007
Figure 2 (continued)
(OECD), 1975 TO 2007
TRENDS OF ENERGY INTENSITY, EFFICIENCY, AND ACTIVITY INDEXES FOR 16
COUNTRIES OF THE ORGANIZATION FOR ECONOMIC COOPERATION AND
DEVELOPMENT (OECD), 1975 TO 2007
a
Greeces axis for index (Y-axis) shows a significant variation. This was due to an excep-
tionally high Greek efficiency index from 1993 to 1997.
(OECD), 1975 TO 2007
(OECD), 1975 TO 2007
(OECD), 1975 TO 2007
(OECD), 1975 TO 2007
TRENDS OF ENERGY INTENSITY, EFFICIENCY, AND ACTIVITY INDEXES FOR 16
COUNTRIES OF THE ORGANIZATION FOR ECONOMIC COOPERATION AND
and 2007. Not surprisingly, the decrease was more rapid in the 1980s and 2000s given
the oil price shocks of 1979 and significant oil price rise between 2002 and 2007.
Second, there is increasing variation in intensity across the countries. The coefficient
of variation, for instance, in 2007 is more than five times the value in 1980. Moreover,
while some countries have reduced their energy intensity drastically (for example,
U.K. and U.S. intensity fell bynearly50 percent and54 percent, respectively, between
1975and2007), other countries have experiencedincreasingintensity(the intensityof
Greece andPortugal rose by 20 percent and 24 percent, respectively, over this period).
The second panel of the table provides summary statistics on the energy effi-
ciency index. Energy efficiency improved at a rate of 0.07 percent in the 1970s but
worsened between 1980 and 1990. Holding economic activity constant, changes in
efficiency result in a 1.31-percent per annum (on average) increase in energy
consumption relative to economic activity in the countries. This trend was re-
versed in the 1990s when energy efficiency fell at a rate of 1.02 percent annually
and continued at a decreasing rate in the 2000s at an average of 0.52 percent.
Similar to aggregate energy intensity, there is wide variation in efficiency im-
provements across countries over time, with a coefficient of variation in 2007
more than five times that of 1980. Finally, while some countries have experienced
significant improvements in efficiency (Belgium saw a 57-percent improvement
Figure 3
TRENDS OF OVERALL ENERGY INTENSITY, EFFICIENCY, AND ACTIVITY INDEXES
FOR 16 COUNTRIES OF THE ORGANIZATION FOR ECONOMIC COOPERATION AND
a
a
Austria, Belgium, Canada, Denmark, France, Greece, Ireland, Italy, Japan, Netherlands,
Norway, Portugal, Spain, Sweden, United Kingdom, and United States.
T
a
b
l
e
2
S
U
M
M
A
R
Y
S
T
A
T
I
S
T
I
C
S
O
F
E
N
E
R
G
Y
I
N
D
E
X
E
S
,
1
9
8
0
,
1
9
9
0
,
2
0
0
0
,
2
0
0
7
a
Y
e
a
r
M
e
a
n
S
t
a
n
d
a
r
d
D
e
v
i
a
t
i
o
n
M
i
n
i
m
u
m
M
a
x
i
m
u
m
A
n
n
u
a
l
C
h
a
n
g
e
(
C
u
m
u
l
a
t
i
v
e
)
A
n
n
u
a
l
C
h
a
n
g
e
(
D
e
c
a
d
e
)
C
o
e
f
f
i
c
i
e
n
t
o
f
V
a
r
i
a
t
i
o
n
I
n
t
e
n
s
i
t
y
1
9
8
0
0
.
9
5
7
0
.
0
6
8
0
.
8
6
2
1
.
0
9
3
-
0
.
4
3
%
-
0
.
4
3
%
0
.
0
7
1
1
9
9
0
0
.
8
2
3
0
.
1
9
2
0
.
6
2
1
1
.
3
8
0
-
1
.
9
2
%
-
1
.
0
7
%
0
.
2
3
3
2
0
0
0
0
.
7
7
0
0
.
2
5
5
0
.
5
3
3
1
.
4
0
8
-
2
.
5
7
%
-
0
.
8
0
%
0
.
3
3
1
2
0
0
7
0
.
6
9
6
0
.
2
5
8
0
.
4
6
5
1
.
2
4
3
-
4
.
0
0
%
-
1
.
3
7
%
0
.
3
7
1
E
f
f
i
c
i
e
n
c
y
1
9
8
0
0
.
9
9
3
0
.
0
7
8
0
.
8
6
3
1
.
1
7
9
-
0
.
0
7
%
-
0
.
0
7
%
0
.
0
7
9
1
9
9
0
1
.
0
8
0
0
.
4
6
0
0
.
6
0
2
2
.
6
1
7
0
.
8
0
%
1
.
3
1
%
0
.
4
2
6
2
0
0
0
1
.
0
2
8
0
.
3
3
3
0
.
6
1
4
1
.
6
6
1
0
.
5
1
%
-
1
.
0
2
%
0
.
3
2
4
2
0
0
7
0
.
9
9
8
0
.
3
9
6
0
.
4
7
0
2
.
0
3
9
0
.
1
5
%
-
0
.
5
2
%
0
.
3
9
6
A
c
t
i
v
i
t
y
1
9
8
0
0
.
9
6
6
0
.
0
4
5
0
.
8
9
3
1
.
0
4
7
-
0
.
3
4
%
-
0
.
3
4
%
0
.
0
4
7
1
9
9
0
0
.
8
0
3
0
.
1
3
1
0
.
5
2
7
1
.
0
3
2
-
2
.
1
6
%
-
1
.
7
0
%
0
.
1
6
3
2
0
0
0
0
.
7
8
1
0
.
2
0
4
0
.
3
6
1
1
.
2
0
0
-
2
.
4
1
%
-
0
.
0
6
%
0
.
2
6
2
2
0
0
7
0
.
7
5
7
0
.
2
6
6
0
.
2
4
4
1
.
2
2
3
-
2
.
8
4
%
-
0
.
3
1
%
0
.
3
5
1
a
I
n
d
e
x
e
s
a
r
e
n
o
r
m
a
l
i
z
e
d
t
o
1
i
n
1
9
7
0
.
S
o
u
r
c
e
:
A
u
t
h
o
r
s
c
a
l
c
u
l
a
t
i
o
n
.
between 1975 and 2007), others efficiency indexes have worsened significantly
(Irelands efficiency worsened by 104 percent over this period).
30
The economic activity index declined at an average rate of 0.6 percent between
1970 and 2007, with a remarkable average of 1.7 percent in the 1980s. This
possibly suggests that the impacts of oil price shocks in the early 1980s on energy
demand operate more through structural shifts in the economy than through
fundamental improvements in energy use (i.e., moving from energy-intensive to
non-energy-intensive activity). There is less time variation in the reduction in
energy intensity due to shifts in economic activity compared to the variation in
efficiency or aggregate intensity. Also, there is less variation across the countries
at any point in time. Even though the coefficient of variation in 2007 is more than
five times its 1980 figure, its annual coefficient is less than the coefficient of
variation in any given year for the efficiency index.
Using equation (1.10), table 3 shows the allocation of changes in energy use
during 2000-2007 (relative to the amount that would have been consumed had
energy intensity remained at its 1975 level) between efficiency and economic
mix.
31
There are considerable variations in energy saved across the nations. While
some countries have achieved sizably higher energy savings over the period,
Greece, Portugal, and Spain have experienced negative energy savings. Similarly,
the relative importance of efficiency and structural shifts to energy saving varies
across countries. While improved energy use has accounted for a larger percentage
of energy saved in some nations (e.g., Belgium77 percent, Sweden 76 percent, and
the United States 73 percent), structural shifts contribute a larger percentage in
some others (e.g., the Netherlands 97 percent and Norway 96 percent). On aver-
age, roughly one-quarter of the energy reduction arising from improvements in
energy intensity in 13 of the countries can be attributed to the fundamental im-
provements in energy use while the remaining three-quarters are accountable to
shifts in the composition of economic activity.
32
Similar to table 3, table 4 presents the overall contributions of efficiency and
activity indexes to energy savings for the 16 countries for the same periods shown
in table 4.
33
The results show that, on average, improved energy efficiency ac-
counts for 23 percent of the total energy saved while the remaining 77 percent was
due to structural changes in the economies. This result suggests that the inclusion
of the three countries, earlier omitted from the computation of average in table 4,
now results in a slight but insignificant reduction in the efficiency and an increase
in activitys contributions to energy savings between 2000 and 2007. Finally, table
5 presents the relative contributions of efficiency and activity to improvements in
energy savings between 1975 and 2007.
Regression Analysis of Key Drivers of Intensity, Efficiency, and Activity Indexes:
As shown in table 2, there is considerable variation in all the three indexes both
across countries and over time. To investigate the causes of this variation,
Table 3
EFFICIENCY AND ACTIVITY CONTRIBUTIONS TO CHANGES IN ENERGY
INTENSITY/ENERGY USE, 2000-2007
a
Energy
Savings
(ktoe)
b
Share Due to:
Energy
Savings
(ktoe)
b
Share Due to:
Year
Efficiency
(in
percent)
Activity
(in percent)
Efficiency
(in percent)
Activity
(in percent)
Austria Belgium
2000 5679.4 142 -42 14743.5 75 25
2001 4559.5 168 -68 14293 69 31
2002 4303.5 184 -84 16955 63 37
2003 3158.4 220 -120 15571 53 47
2004 3765.6 215 -115 17682 62 38
2005 3881.9 223 -123 19404 66 34
2006 4645.9 210 -110 21326 92 8
2007 6474.4 180 -80 24033 136 -36
Average 193 -93 77 23
Canada Denmark
2000 93998.2 37 63 9812.3 80 20
2001 104168.4 35 65 9599.2 80 20
2002 106969.7 32 68 9925.6 76 24
2003 106805.8 19 81 9710.0 77 23
2004 112906.8 20 80 10130.4 80 20
2005 119630.1 16 84 10655.4 75 25
2006 134103.9 16 84 11464.9 68 32
2007 133659.9 7 93 11872.1 58 42
Average 23 77 74 26
France Greece
2000 65996.0 14 86 -5135.5 47 53
2001 63467.7 4 96 -5230.1 48 52
2002 69451.8 9 91 -5066.2 63 37
2003 68034.1 1 99 -5382.6 65 35
2004 72097.2 6 94 -4480.9 42 58
2005 77059.7 11 89 -4425.6 51 49
2006 84196.9 14 86 -4213.8 56 44
2007 93112.9 19 81 -4052.3 51 49
Average 10 90 53 47
(continued)
Table 3 (continued)
a
Energy
Savings
(ktoe)
b
Share Due to:
Energy
Savings
(ktoe)
b
Share Due to:
Year
Efficiency
(in
percent)
Activity
(in percent)
Efficiency
(in percent)
Activity
(in percent)
Ireland Italy
2000 6719.9 -100 200 40594.1 15 85
2001 7419.6 -97 197 41723.8 -2 102
2002 8544.4 -90 190 42959.3 2 98
2003 9133.2 -95 195 37960.7 -23 123
2004 9609.9 -98 198 35102.2 -15 115
2005 10320.6 -99 199 34584.9 -23 123
2006 10925.7 -105 205 38917.7 -10 110
2007 12404.5 -102 202 43144.3 -6 106
Average -98 198 -8 108
Japan The Netherlands
2000 90352.5 -134 234 26638.8 13 87
2001 93608.7 -123 223 27243.8 11 89
2002 91711.3 -124 224 27696.6 10 90
2003 101637.3 -102 202 26872.8 -10 110
2004 109055.2 -82 182 27945.7 -8 108
2005 115897.5 -62 162 29400.6 -3 103
2006 127209.2 -43 143 32935.3 5 95
2007 139287.5 -28 128 36416.8 8 92
Average -87 187 3 97
Norway Portugal
2000 10769.6 7 93 -3727.7 199 -99
2001 10797.2 9 91 -3712.5 197 -97
2002 11647.0 9 91 -4079.5 188 -88
2003 12066.4 6 94 -4089.7 180 -80
2004 12804.1 3 97 -4151.9 177 -77
2005 13643.6 3 97 -4004.4 175 -75
2006 14687.3 -3 103 -3595.9 143 -43
2007 15377.1 -5 105 -3502.0 129 -29
Average 4 96 174 -74
(continued)
Table 3 (continued)
a
Energy
Savings
(ktoe)
b
Share Due to:
Energy
Savings
(ktoe)
b
Share Due to:
Year
Efficiency
(in
percent)
Activity
(in percent)
Efficiency
(in percent)
Activity
(in percent)
Spain Sweden
2000 -4440.6 509 -409 19726.6 67 33
2001 -5657.8 451 -351 20904.0 77 23
2002 -5117.6 536 -436 22216.6 74 26
2003 -8057.7 424 -324 23439.4 74 26
2004 -9509.3 396 -296 25977.0 79 21
2005 -9616.8 407 -307 28214.2 80 20
2006 -4929.9 740 -640 31093.2 79 21
2007 -3941.8 982 -882 32726.5 79 21
Average 556 -456 76 24
United Kingdom United States
2000 82844.2 -11 111 1221044 77 23
2001 86483.4 -13 113 1267356 71 29
2002 95157.2 -11 111 1295459 69 31
2003 100537.1 -8 108 1345017 67 33
2004 107684.9 1 99 1428331 72 28
2005 112510.0 3 97 1515905 74 26
2006 121870.7 -4 104 1606683 75 25
2007 132168.1 -2 102 1645349 76 24
Average -6 106 73 27
Overall average (in percent)
c
Share due to efficiency 26
Share due to activity 74
a
For Greece, Portugal, and Spain, negative percentage contribution means improvements in that
index since it means that the index does not contribute to negative energy savings in that year. But for
others with positive energy savings, negative contribution means that the index worsened in that year.
b
ktoe = kilotons of oil equivalent/thousand tons of oil equivalent.
c
Greece, Portugal, and Spain are excluded from the calculation of average contribution because
no energy was saved over this period. Inclusion of these three countries in calculating the overall
percentage contribution would make the total contributions by efficiency and activity not sum to 100
percent. To get a consistent result by including these countries, aggregation has to be done before
decomposition analysis.
Source: Computed from the decomposed intensity indexes.
regression of these indexes on key economic variables is conducted using equa-
tions (2.1) and (2.4). Regression variables include the natural log of each intensity
index, real energy prices, real income per capita, log-income squared, and pop-
ulation growth rates. Log-income squared is included to account for the possible
non-linear response of energy intensity to income. Population growth rates are
used to capture the response of energy intensity to population changes; for
instance, fast-growing countries may be adding more energy-efficient in-
frastructures than slow-growing nations. On the other hand, fast-growing countries
may be less energy efficient in their energy use if their capital investment does not
keep pace with growth, e.g., traffic congestion.
34
Table 5 presents the summary of
the data for this regression analysis. The table reports three measures of standard
deviation to show the sort of variation in the data. The overall standard deviation
measures variation in the pooled cross-section time-series data. The between
standard deviation indicates the variation in the average country data across the 16
OECD countries; the within standard deviation measures the variation in the data
from country-specific means. Looking at the intensity data, for example, the be-
tween standard deviation is higher than the within standard deviation. This is not
surprising since the variation in energy consumption per output across countries is
likely to be more than variation over time. However, the within standard de-
viations overall suggest more variations in the data across time within countries
than across countries.
Table 6 presents regression results from equations (2.1) and (2.4). The first two
columns report results from a regression of the aggregate intensity index on price,
Table 4
AGGREGATE AVERAGE EFFICIENCY AND ACTIVITIY CONTRIBUTIONS TO CHANGES
IN ENERGY INTENSITY/ENERGY USE FOR THE 16 COUNTRIES OF THE ORGANIZATION
FOR ECONOMIC COOPERATION AND DEVELOPMENT,
2000-2007
Share Due to:
Year
Energy Savings (in 1,000 tons
of oil equivalent)
Efficiency
(in percent)
Activity
(in percent)
2000 98788.8 23 77
2001 102562.0 23 77
2002 105155.2 19 81
2003 107921.8 15 85
2004 114293.0 22 78
2005 120883.9 26 74
2006 130108.6 29 71
2007 135471.7 30 70
Average 23 77
Source: Computed from the decomposed intensity indexes.
T
a
b
l
e
5
S
U
M
M
A
R
Y
S
T
A
T
I
S
T
I
C
S
O
F
D
A
T
A
F
O
R
R
E
G
R
E
S
S
I
O
N
A
N
A
L
Y
S
I
S
,
1
9
7
5
T
O
2
0
0
7
S
t
a
n
d
a
r
d
D
e
v
i
a
t
i
o
n
V
a
r
i
a
b
l
e
M
e
a
n
O
v
e
r
a
l
l
B
e
t
w
e
e
n
W
i
t
h
i
n
M
i
n
m
u
m
M
a
x
i
m
u
m
I
n
t
e
n
s
i
t
y
0
.
8
4
7
0
.
2
0
2
0
.
1
6
1
0
.
1
2
8
0
.
4
6
5
1
.
4
8
8
E
f
f
i
c
i
e
n
c
y
1
.
0
5
2
0
.
3
7
3
0
.
2
7
8
0
.
2
5
7
0
.
4
7
0
4
.
6
5
2
a
A
c
t
i
v
i
t
y
0
.
8
3
9
0
.
1
7
1
0
.
1
0
9
0
.
1
3
4
0
.
2
4
4
1
.
2
2
3
R
e
a
l
e
n
e
r
g
y
p
r
i
c
e
s
(
2
0
0
0
=
1
0
0
)
1
0
2
.
1
4
3
1
7
.
9
3
0
1
0
.
2
0
5
1
4
.
9
5
6
5
3
.
4
0
6
1
7
0
.
3
0
1
R
e
a
l
i
n
c
o
m
e
p
e
r
c
a
p
i
t
a
(
P
P
P
s
2
0
0
0
U
S
$
)
b
2
2
0
2
8
6
1
6
8
.
6
0
4
4
0
5
3
.
1
9
4
4
7
5
6
.
1
3
7
8
1
7
6
.
6
1
9
4
0
5
0
6
.
2
2
L
n
(
i
n
c
o
m
e
p
e
r
c
a
p
i
t
a
)
9
.
9
5
9
0
.
2
9
6
0
.
2
0
5
0
.
2
1
9
9
.
0
0
9
1
0
.
6
0
9
(
L
n
i
n
c
o
m
e
p
e
r
c
a
p
i
t
a
)
2
9
9
.
2
6
0
5
.
8
4
7
4
.
0
3
5
4
.
3
4
8
8
1
.
1
6
3
1
1
2
.
5
5
5
L
n
(
e
n
e
r
g
y
p
r
i
c
e
)
4
.
6
1
1
0
.
1
7
3
0
.
0
9
9
0
.
1
4
4
3
.
9
7
8
5
.
1
3
8
P
o
p
u
l
a
t
i
o
n
g
r
o
w
t
h
r
a
t
e
0
.
5
5
0
0
.
4
7
5
0
.
2
9
9
0
.
3
7
7
-
0
.
6
0
2
3
.
7
9
9
N
u
m
b
e
r
o
f
o
b
s
e
r
v
a
t
i
o
n
s
5
2
8
a
T
h
i
s
u
n
u
s
u
a
l
h
i
g
h
v
a
l
u
e
r
e
s
u
l
t
s
f
r
o
m
G
r
e
e
c
e
s
e
x
c
e
p
t
i
o
n
a
l
h
i
g
h
e
f
f
i
c
i
e
n
c
y
i
n
t
e
n
s
i
t
y
i
n
d
e
x
i
n
1
9
9
7
.
b
P
u
r
c
h
a
s
i
n
g
p
o
w
e
r
p
a
r
i
t
i
e
s
i
n
2
0
0
0
U
.
S
.
d
o
l
l
a
r
s
.
S
o
u
r
c
e
:
C
o
m
p
u
t
e
d
f
r
o
m
t
h
e
d
a
t
a
u
s
e
d
f
o
r
t
h
e
r
e
g
r
e
s
s
i
o
n
a
n
a
l
y
s
i
s
.
T
a
b
l
e
6
R
E
G
R
E
S
S
I
O
N
R
E
S
U
L
T
S
F
O
R
1
6
C
O
U
N
T
R
I
E
S
O
F
T
H
E
O
R
G
A
N
I
Z
A
T
I
O
N
F
O
R
E
C
O
N
O
M
I
C
C
O
O
P
E
R
A
T
I
O
N
A
N
D
D
E
V
E
L
O
P
M
E
N
T
,
1
9
7
5
T
O
2
0
0
7
a
V
a
r
i
a
b
l
e
s
I
n
t
e
n
s
i
t
y
E
f
f
i
c
i
e
n
c
y
A
c
t
i
v
i
t
y
(
1
)
(
2
)
(
3
)
(
4
)
(
5
)
(
6
)
L
n
(
p
r
i
c
e
)
-
0
.
1
7
0
(
0
.
0
2
6
)
d
-
0
.
0
8
2
(
0
.
0
1
2
)
d
-
0
.
1
8
4
(
0
.
0
4
8
)
d
-
0
.
1
0
3
(
0
.
0
5
7
)
b
-
0
.
0
1
1
(
0
.
0
4
5
)
_
_
L
n
(
p
e
r
c
a
p
i
t
a
i
n
c
o
m
e
)
1
2
.
7
2
9
(
0
.
7
6
7
)
d
5
.
4
0
3
(
2
.
5
8
8
)
c
1
4
.
2
8
9
(
1
.
4
3
1
)
d
_
_
-
1
.
5
0
7
(
1
.
3
4
5
)
_
_
(
L
n
p
e
r
c
a
p
i
t
a
i
n
c
o
m
e
)
2
-
0
.
6
6
8
(
0
.
0
3
9
)
d
-
0
.
2
9
7
(
0
.
1
3
1
)
c
-
0
.
7
1
7
(
0
.
0
7
2
)
-
0
.
0
0
2
(
0
.
0
0
)
0
.
0
4
6
(
0
.
0
6
8
)
-
0
.
0
2
9
(
0
.
0
1
1
)
c
P
o
p
u
l
a
t
i
o
n
g
r
o
w
t
h
r
a
t
e
0
.
0
2
0
(
0
.
0
1
0
)
c
0
.
0
0
4
(
0
.
0
0
5
)
0
.
0
5
3
(
0
.
0
1
9
)
d
0
.
0
4
4
(
0
.
0
1
5
)
c
-
0
.
0
3
3
(
0
.
0
1
8
)
b
-
0
.
0
2
3
(
0
.
0
1
3
)
b
L
n
(
p
r
i
c
e
)
-
1
_
_
_
_
_
_
_
_
_
_
-
0
.
0
6
3
(
0
.
0
3
1
)
c
L
n
(
p
r
i
c
e
)
-
2
_
_
_
_
_
_
0
.
1
0
2
(
0
.
0
5
9
)
b
_
_
_
_
L
n
(
p
e
r
c
a
p
i
t
a
i
n
c
o
m
e
)
-
1
_
_
-
4
.
7
1
8
(
2
.
5
0
8
)
b
_
_
_
_
_
_
0
.
5
2
3
(
0
.
2
2
6
)
c
(
L
n
p
e
r
c
a
p
i
t
a
i
n
c
o
m
e
)
2
-
1
_
_
0
.
2
5
9
(
0
.
1
2
7
)
c
_
_
_
_
_
_
_
_
L
n
(
i
n
t
e
n
s
i
t
y
)
-
1
_
_
0
.
8
5
6
(
0
.
0
1
8
)
d
_
_
_
_
_
_
_
_
L
n
(
e
f
f
i
c
i
e
n
c
y
)
-
1
_
_
_
_
_
_
0
.
8
2
8
(
0
.
0
3
1
)
d
_
_
_
_
L
n
(
a
c
t
i
v
i
t
y
)
-
1
_
_
_
_
_
_
_
_
_
_
0
.
8
8
7
(
0
.
0
2
9
)
d
D
i
a
g
n
o
s
t
i
c
T
e
s
t
s
:
H
a
u
s
m
a
n
(
c
2
)
4
.
7
1
(
4
)
N
A
2
7
.
9
7
(
4
)
N
A
1
4
.
0
3
(
4
)
N
A
W
a
l
d
(
c
2
)
1
6
2
8
.
8
0
(
4
)
1
0
8
4
5
.
8
0
(
7
)
N
A
9
6
7
.
0
0
(
5
)
N
A
2
0
6
5
.
3
5
(
5
)
F
-
t
e
s
t
N
A
N
A
3
8
.
4
5
(
4
,
5
0
8
)
N
A
1
1
3
.
9
7
(
4
,
5
0
8
)
N
A
S
a
r
g
a
n
(
c
2
)
N
A
3
5
0
.
4
6
(
3
5
9
)
N
A
3
7
0
.
0
8
(
3
3
8
)
N
A
3
8
4
.
7
6
(
3
5
3
)
a
F
o
r
t
h
e
r
e
s
u
l
t
s
o
u
t
p
u
t
,
s
t
a
n
d
a
r
d
e
r
r
o
r
s
a
r
e
i
n
p
a
r
e
n
t
h
e
s
e
s
,
w
h
i
l
e
t
h
e
n
u
m
b
e
r
s
i
n
p
a
r
e
n
t
h
e
s
e
s
u
n
d
e
r
d
i
a
g
n
o
s
t
i
c
t
e
s
t
s
a
r
e
d
e
g
r
e
e
s
o
f
f
r
e
e
d
o
m
.
N
A
i
n
d
i
c
a
t
e
s
n
o
t
a
p
p
l
i
c
a
b
l
e
.
C
o
l
u
m
n
s
1
,
3
,
a
n
d
5
a
r
e
s
t
a
t
i
c
m
o
d
e
l
s
r
e
s
u
l
t
s
;
c
o
l
u
m
n
s
2
,
4
,
a
n
d
6
a
r
e
d
y
n
a
m
i
c
m
o
d
e
l
s
r
e
s
u
l
t
s
e
s
t
i
m
a
t
e
d
u
s
i
n
g
A
r
e
l
l
a
n
o
a
n
d
B
o
n
d
s
p
e
c
i
f
i
c
a
t
i
o
n
.
T
h
e
1
6
c
o
u
n
t
r
i
e
s
a
r
e
A
u
s
t
r
i
a
,
B
e
l
g
i
u
m
,
C
a
n
a
d
a
,
D
e
n
m
a
r
k
,
F
r
a
n
c
e
,
G
r
e
e
c
e
,
I
r
e
l
a
n
d
,
I
t
a
l
y
,
J
a
p
a
n
,
N
e
t
h
e
r
l
a
n
d
s
,
N
o
r
w
a
y
,
P
o
r
t
u
g
a
l
,
S
p
a
i
n
,
S
w
e
d
e
n
,
U
n
i
t
e
d
K
i
n
g
d
o
m
,
a
n
d
U
n
i
t
e
d
S
t
a
t
e
s
.
T
h
e
H
a
u
s
m
a
n
t
e
s
t
i
s
a
s
p
e
c
i
f
i
c
a
t
i
o
n
t
e
s
t
o
f
s
t
r
i
c
t
e
x
o
g
e
n
e
i
t
y
u
n
d
e
r
a
s
t
a
t
i
c
p
a
n
e
l
m
o
d
e
l
,
w
h
i
l
e
t
h
e
S
a
r
g
a
n
t
e
s
t
i
s
a
t
e
s
t
o
f
e
x
o
g
e
n
e
i
t
y
o
f
i
n
s
t
r
u
m
e
n
t
a
l
v
a
r
i
a
b
l
e
s
a
n
d
o
v
e
r
-
i
d
e
n
t
i
f
y
i
n
g
r
e
s
t
r
i
c
t
i
o
n
s
f
o
r
a
d
y
n
a
m
i
c
p
a
n
e
l
m
o
d
e
l
.
T
h
e
W
a
l
d
a
n
d
F
-
t
e
s
t
s
a
r
e
t
e
s
t
s
o
f
j
o
i
n
t
s
i
g
n
i
f
i
c
a
n
c
e
o
f
t
h
e
v
a
r
i
a
b
l
e
s
.
A
l
l
t
h
e
t
e
s
t
s
,
e
x
c
e
p
t
F
-
t
e
s
t
,
a
r
e
a
s
y
m
p
t
o
t
i
c
a
l
l
y
d
i
s
t
r
i
b
u
t
e
d
a
s
(
c
2
)
.
b
S
i
g
n
i
f
i
c
a
n
t
a
t
t
h
e
1
0
-
p
e
r
c
e
n
t
l
e
v
e
l
.
c
S
i
g
n
i
f
i
c
a
n
t
a
t
t
h
e
5
-
p
e
r
c
e
n
t
l
e
v
e
l
.
d
S
i
g
n
i
f
i
c
a
n
t
a
t
t
h
e
1
-
p
e
r
c
e
n
t
l
e
v
e
l
.
income, and population growth. In column 1, a 1-percent rise in energy prices is
associated with 0.17-percent decrease in energy intensity. The price coefficient is
estimated precisely and significant at the 1-percent level. Energy intensity exhibits
a non-linear response to income, first rising and then falling. However, energy
intensity could be viewed as falling at every level of income since its turning point
is well below the minimum level of per-capita income in the dataset.
35
A 1-percent
rise in income leads to a 0.58-percent drop in energy intensity. Population growth
shows a positive relationship with energy intensity. This coefficient is statistically
significant at a 5-percent level and indicates that fast-growing countries have
higher energy intensity. Whether this occurs because faster-growing countries
suffer from congestion-induced energy cost or because faster-growing economies
attract energy-intensive activities will be investigated later.
Turning to columns 3 and 5, a 1-percent rise in energy prices will result in 0.18-
percent and 0.01-percent decreases in efficiency and activity intensities, re-
spectively. However, all the parameter estimates (except population growth) on
activity index are not statistically significant at even the 10-percent level, in-
dicating that impacts of prices on intensity operate solely through energy efficiency.
As with aggregate intensity, energy efficiency exhibits a quadratic response to
income, first rising and then falling. The positive impact of population growth on
the efficiency index suggests that an increase in energy intensity in response to
population growth occurs because fast-growing countries suffer from congestion-
induced energy cost. Had the economic structure and all other variables remained
constant, a 1-percent point variation increase in population growth would have led
to a 5.3-percent rise in aggregate intensity. However, this was offset by the re-
duction in energy-intensive activities where a 1-percent point variation increase in
population growth (ceteris paribus) would have led to a 3.3-percent decrease in
aggregate intensity.
Given that energy intensity is not likely to respond immediately to changes in
economic variables, columns 2, 4, and 6 present more realistic results using the
dynamic panel model from equation (2.4). In column 2, the results show that a
1-percent rise in price leads to a 0.08-percent reduction in energy intensity, while
income first started rising and then falling as it is under a static model in column 1.
On average, a 1-percent rise in income reduces intensity by 0.51 percent. One
striking feature of these results is that it appears price plays its immediate role in
reducing aggregate intensity through efficient use of energy only, while the im-
pacts of income in reducing energy intensity is through a reduction in energy-
intensive activities (structural changes).
36
Again, the coefficients on population
growth suggest that fast-growing countries suffer from congestion-induced energy
cost (and therefore have increased intensity) but try to adjust by reducing energy-
intensive activities. This is not surprising since an economy with a rapidly growing
population is likely to reduce capital-intensive activities and encourage labor-
intensive services.
The lower part of table 6 presents various diagnostic tests explored to arrive at
the preferred models presented in the table. The Hausman test is explored to de-
termine the consistent and efficient estimator between random and fixed effects.
37
However, there is mixed evidence regarding the superiority of these effects. While
the Hausman test supports the assumption of strict exogeneity for an aggregate
intensity model, it rejects this assumption under efficiency and activity models.
Thus, random effects estimates are reported for aggregate intensity while fixed
effects are estimated for both efficiency and the activity index. The Wald and
F-tests show that the variables are jointly significant in explaining the variation in
each of the three intensity indexes. Similarly, the Sargan test is used to test for the
strict exogeneity of the instrumental variables in the dynamic panel.
38
All the
models perform very well, showing that the instruments and the disturbance terms
are not correlated. G. E. Metcalf used a simple dynamic panel model in a similar
study but did not test for the exogeneity of the instruments.
39
However, this test is
important since correlation of instruments with error terms can result in biased
estimates, and any policy drawn on such results may be misleading.
Table 7 reports the mean price and income elasticity for the estimated results in
columns 2, 4, and 6 of table 6. The short-run and long-run price elasticities of
energy intensity are -0.08 and -0.57, respectively. That is, a 1-percent rise in
energy price has immediate impacts of a 0.08-percent decrease in energy intensity
and a long-run decrease of 0.57 percent. Not surprisingly, an increase in price is
likely to result in decreased energy consumption in an economy either by being
more energy-efficient conscious or by reducing energy-intensive activities. The
long-run response is higher than in the short run, giving the flexibility that char-
acterizes the choice of energy consumption in the long run. Similarly, the short-
run and the long-run responses of energy intensity to income are -0.51 and -0.50,
respectively. These show that a 1-percent increase in income results in 0.51-
percent and 0.50-percent improvements in energy intensity in the short and long
run, respectively.
40
The estimates show that energy intensity responds more to
Table 7
PRICE AND INCOME ELASTICITIES FOR 16 COUNTRIES OF THE ORGANIZATION FOR
ECONOMIC COOPERATION AND DEVELOPMENT, 1975 TO 2007
(in overall averages)
Intensity Efficiency Activity
Price elasticities
Short run -0.08 -0.10 0
Long run -0.57 -0.01 -0.56
Income elasticities
Short run -0.51 -0.04 -0.58
Long run -0.50 -0.23 -0.48
income changes in the short run than in the long run. This is not surprising since
(rapid) increase in income (e.g., economic boom) may lead to greater composi-
tional shifts (moving from energy-intensive to less energy-intensive activities) in
the structure of an economy in the short term. However, this structural change
may be limited in the longer term given the limit to the extent to which energy-
intensive activities can be substituted by less energy-consuming activities.
41
The short- and long-run price elasticities of efficiency are -0.10 and -0.01 while
that of activity are 0 and -0.56, respectively. This further confirms that the short-
run impacts of price changes on energy intensity operate solely through efficiency,
while the long-run impact is largely through changes in the economic mix. This
means that, in order to reduce energy intensity, people respond to an immediate
price rise by being more energy efficient but decide to move away from energy-
intensive activities as price continues to increase in the longer term. Also, this
possibly suggests the limitation to the extent to which improvements in energy use
(in response to price) can take place, since efficient energy use also depends on the
state of technology in an economy.
Both the short-run and the long-run income elasticities of the efficiency and
activity indexes suggest that the decline in aggregate intensity in response to in-
come changes is largely due to compositional shifts in the economies. Had the
sectoral energy intensity remained constant, the aggregate intensitys elasticity
with respect to income would have been -0.58 and -0.48 in the short and long run,
respectively. On the other hand, had the economic mix remained unchanged, the
short-run and long-run income elasticities of aggregate intensity would have been
-0.04 and -0.23, respectively. The long-run income elasticity of the efficiency
index is higher than the short-run estimates. This shows that an increase in income
leads to more fundamental improvements in energy use in the long run than in the
short term, possibly because rising income results in installation of more energy-
efficient machines. However, the long-run response to income of activity is lower
than that in the short run. Again, this possibly suggests the limitation to energy-
intensive/non-energy-intensive activities substitution.
In comparison with previous studies, the short-run price elasticities are close to
those estimated by Metcalf in a related article for 46 U.S. states.
42
In analyzing the
energy intensity of these states, he estimated the short-run price elasticity of ag-
gregate intensity, efficiency, and activity indexes to be -0.105, -0.08, and -0.018,
respectively. However, his estimates of -0.299, -0.154, and -0.058 for intensity,
efficiency, and activity, respectively, show a significant difference between his
long-run price elasticity and those reported in this study. The estimates in this
study can be said to be within the range of long-run price elasticity. As for income,
the short-run elasticities (except for efficiency) in this study are significantly
higher than those of Metcalf, while the log-run estimates are relatively low. The
reasons for this may be as follows. First, while Metcalfs analysis was for the U.S.
states with similar socioeconomic characteristics, this study combines different
T
a
b
l
e
8
S
U
M
M
A
R
Y
O
F
T
H
E
L
O
N
G
-
R
U
N
A
N
D
S
H
O
R
T
-
R
U
N
I
N
C
O
M
E
E
L
A
S
T
I
C
I
T
I
E
S
A
N
D
P
E
R
C
E
N
T
A
G
E
C
H
A
N
G
E
S
I
N
P
E
R
-
C
A
P
I
T
A
I
N
C
O
M
E
A
g
g
r
e
g
a
t
e
I
n
t
e
n
s
i
t
y
E
f
f
i
c
i
e
n
c
y
A
c
t
i
v
i
t
y
A
n
n
u
a
l
P
e
r
c
e
n
t
C
h
a
n
g
e
i
n
I
n
c
o
m
e
C
o
u
n
t
r
y
L
o
n
g
R
u
n
S
h
o
r
t
R
u
n
L
o
n
g
R
u
n
S
h
o
r
t
R
u
n
L
o
n
g
R
u
n
S
h
o
r
t
R
u
n
A
u
s
t
r
i
a
-
0
.
5
6
-
0
.
5
8
-
0
.
2
3
-
0
.
0
4
-
0
.
5
4
-
0
.
5
8
2
.
2
2
B
e
l
g
i
u
m
-
0
.
5
4
-
0
.
5
6
-
0
.
2
3
-
0
.
0
4
-
0
.
5
3
-
0
.
5
8
1
.
9
C
a
n
a
d
a
-
0
.
5
7
-
0
.
5
9
-
0
.
2
3
-
0
.
0
4
-
0
.
5
5
-
0
.
5
9
1
.
8
7
D
e
n
m
a
r
k
-
0
.
5
6
-
0
.
5
9
-
0
.
2
3
-
0
.
0
4
-
0
.
5
5
-
0
.
5
8
2
.
0
4
F
r
a
n
c
e
-
0
.
5
2
-
0
.
5
4
-
0
.
2
3
-
0
.
0
4
-
0
.
5
1
-
0
.
5
8
1
.
8
3
G
r
e
e
c
e
-
0
.
3
8
-
0
.
3
8
-
0
.
2
3
-
0
.
0
4
-
0
.
3
7
-
0
.
5
6
2
.
1
6
I
r
e
l
a
n
d
-
0
.
4
2
-
0
.
4
2
-
0
.
2
3
-
0
.
0
4
-
0
.
4
-
0
.
5
7
4
.
2
I
t
a
l
y
-
0
.
5
-
0
.
5
2
-
0
.
2
3
-
0
.
0
4
-
0
.
4
9
-
0
.
5
8
1
.
8
9
J
a
p
a
n
-
0
.
5
1
-
0
.
5
3
-
0
.
2
3
-
0
.
0
4
-
0
.
5
-
0
.
5
8
2
.
1
7
N
e
t
h
e
r
l
a
n
d
s
-
0
.
5
6
-
0
.
5
8
-
0
.
2
3
-
0
.
0
4
-
0
.
5
4
-
0
.
5
8
2
.
0
5
N
o
r
w
a
y
-
0
.
6
6
-
0
.
6
9
-
0
.
2
4
-
0
.
0
4
-
0
.
6
4
-
0
.
5
9
2
.
5
9
P
o
r
t
u
g
a
l
-
0
.
2
4
-
0
.
2
2
-
0
.
2
2
-
0
.
0
4
-
0
.
2
3
-
0
.
5
5
2
.
4
3
S
p
a
i
n
-
0
.
3
9
-
0
.
3
8
-
0
.
2
3
-
0
.
0
4
-
0
.
3
7
-
0
.
5
7
2
.
0
8
S
w
e
d
e
n
-
0
.
5
6
-
0
.
5
8
-
0
.
2
3
-
0
.
0
4
-
0
.
5
4
-
0
.
5
8
1
.
7
8
U
n
i
t
e
d
K
i
n
g
d
o
m
-
0
.
5
-
0
.
5
2
-
0
.
2
3
-
0
.
0
4
-
0
.
4
9
-
0
.
5
8
2
.
4
6
U
n
i
t
e
d
S
t
a
t
e
s
-
0
.
6
6
-
0
.
6
9
-
0
.
2
4
-
0
.
0
4
-
0
.
6
4
-
0
.
6
2
.
2
4
O
v
e
r
a
l
l
a
v
e
r
a
g
e
-
0
.
5
1
-
0
.
5
2
-
0
.
2
3
-
0
.
0
4
-
0
.
4
9
-
0
.
5
8
S
o
u
r
c
e
:
C
o
m
p
u
t
e
d
f
r
o
m
t
h
e
c
o
e
f
f
i
c
i
e
n
t
e
s
t
i
m
a
t
e
s
i
n
t
a
b
l
e
7
a
n
d
i
n
d
i
v
i
d
u
a
l
c
o
u
n
t
r
y
s
o
b
s
e
r
v
a
t
i
o
n
s
o
n
i
n
c
o
m
e
.
countries with significantly different socioeconomic characteristics. Second,
while this study based its estimates of income elasticity on the overall mean ob-
servations, Metcalf computed his estimates using average observations for 2001
only.
Comparison of Income Elasticity Estimates across Countries: Table 8 presents the
summary of income elasticity and the percentage change in per-capita income
over the period of 1975-2007.
43
Several facts are noteworthy. First, those countries
(Greece, Portugal, and Spain) with deterioration in intensity over the periods of
analysis have lower responses to income changes. While all other countries (ex-
cept Ireland with -0.42) have average long-run income elasticity of aggregate
intensity of between -0.5 and -0.66, Greece, Portugal, and Spain have -0.38, -0.24,
and -0.39, respectively. Second, income elasticity of efficiency is roughly constant
both across countries and over time. This is not surprising since the response of
efficiency to income changes is not statistically significant different from zero as
shown in table 6, column 4. Third, long-run income elasticity continues to increase
over time, suggesting continuous improvements in energy intensity as income
rises. Expectedly, a rising income may lead to the development of more energy-
saving technology or diversion of investments from energy-intensive activities to
non-energy-intensive services (a change in taste). Generally, the response of en-
ergy intensity to income over the periods of analysis largely comes from the
restructuring of economic activities rather than from improvements in energy use.
For all the countries in this analysis, aggregate energy intensity declines with
income as a result of shifts towards a non-energy-intensive activity mix. Table 8
indicates that a 1-percent rise in income results in between approximately a 0.23-
percent to a 0.64-percent decrease in energy intensity as a result of shifts towards
a more non-energy-intensive activity mix. Finally, the short-run income elasticity
(except for efficiency) over these periods is greater than the long-run elasticity
possibly due to the limitation to economic restructuring in the longer term.
In comparison with the mean elasticity reported in table 7, the overall average
elasticity reported in table 8 shows little but insignificant variation. For instance,
short- and long-run income elasticities of intensity are respectively -0.52 and
-0.51, instead of -0.51 and -0.50 as reported in table 7. On average, this suggests
that the mean elasticity reported in table 7 is robust enough to represent the
elasticity estimates for the whole sample period.
Summary and Conclusion
This study investigated the forces driving energy intensity in 16 OECD
countries from 1975 to 2007. It built on the existing literature in energy de-
composition analysis in several ways. First, it analyzed changes in energy intensity
in the OECD countries for the whole economy using a perfect decomposition
methodology. Second, it used econometric techniques in a dynamic panel form to
identify the key drivers of changes in aggregate intensity, efficiency, and activity
indexes. The results obtained from the Fisher ideal index and regression methods
were presented. These results showed considerable variations across the countries
over the period of analysis. While some countries have reduced their energy in-
tensity significantly, others have not experienced any decrease at all. Similarly, the
relative impacts of efficiency and activity show significant variations across the
nations. Even though some countries have improved their intensity considerably
through the efficient use of energy, overall the evidence suggests that significant
improvements come from structural shifts. As for the regression analysis, all three
indexes respond negatively to energy prices and income over the periods of
analysis with low response to income changes coming from countries with high
energy intensity. This could be a result of the low level of commitment to energy-
intensity reduction in those countries during these spans. Further exploration of the
results also showed that the long-run impacts of both energy prices and income on
aggregate intensity operate more through structural shifts than through efficient
energy use. This suggests that the long-run reduction in energy intensity as a result
of changes in energy prices and income is due largely to movements away from
energy-intensive activities and towards non-energy-intensive services. However,
the short-run price impacts on aggregate intensity operate solely through the ef-
ficient use of energy; that of income operates largely through structural shifts.
A number of conclusions that inform thinking about energy policy can be drawn
both from the decomposition and regression analyses conducted in this research.
First, the decomposition analysis showed (overall) that a significant role in reducing
energy intensity operates through structural shifts in the economic activities. This
means that for the countries to continue improving their energy intensity (given the
limit of the extent to which a nation can move away from energy-intensive activ-
ities), there is a need for rapid promotion of efficient use of energy. In other words,
a policy that will encourage or enforce efficient energy use (such as vehicle fuel
standards) needs to be implemented. It could be argued that some of the countries
where such a policy has been implemented (e.g., the United States) have experi-
enced meaningful improvements in energy intensity through efficiency.
Second, rising energy prices and income drive declines in energy intensity in
the OECD countries. That higher energy prices result in lower energy intensity is
no surprise, but economic theory is not clear whether a higher income should result
in lower or higher energy intensity.
44
Based on the findings of this research,
however, it appears that the governments concerned can improve their energy
intensity by raising energy prices either through tax policy or any other means.
However, this also has many welfare implications and the possibility that private
investors may move away from energy-consuming activities (e.g., manufacturing)
to less energy-intensive services.
Third, energy intensity is an important determinant of carbon emissions since
energy combustion is responsible for a greater percentage of carbon emissions (79
percent of the world and 96 percent of the OECD).
45
Fossil fuel-related carbon
emissions can be decreased through fuel switching and by reducing energy de-
mand. This reduction in energy can be effected by becoming more conscious of
energy efficiency, whether in production or consumption, and by introducing more
energy-efficient technology.
However, this exercise has exposed a number of weaknesses that need further
research. First, the aggregate data used in this study made it difficult to identify
in which of the sectors the improvements in efficiency and/or structural shifts are
more pronounced. Moreover, the reduction in energy intensity as income rises is
not simply moving away from industrial activities to commercial services but
rather something more intrinsic to activities within each sector. Conducting the
type of decomposition analysis in this study at a more disaggregated level to
investigate this issue more deeply would be a productive area of future research.
Second, given the diverse nature of the countries used in this study, the as-
sumption of constant slope coefficients across a wide range of countries that
characterizes a panel data methodology seems to be unrealistic. Each country
differs in terms of structures, institutions, and socioeconomic patterns, and the
imposition of the same response to economic variables of intensity indexes tends
to be biased. Carrying out this analysis for each of the countries is likely to
produce more realistic results. Finally, the regression analysis conducted in this
study ignores the impacts of asymmetric price response and technical progress
on energy-intensity changes. Incorporating asymmetric price response will
enable the investigation of variations in energy intensitys response to price
changes. Similarly, incorporating technical progress will help in capturing the
impacts of technical changes on energy intensity. Therefore, this study at least
suggests that such an analysis considering these issues is feasible and likely to be
highly productive.
NOTES
1
Organization for Economic Cooperation and Development (OECD), International Energy
Agency (IEA), World Energy Balances and Energy Balances of OECD Countries (Paris: IEA,
2009), available at www.iea.org, accessed on March 12, 2010.
2
The Laspeyres index isolates the contribution of a variable to the variation of aggregate in-
dicator by allowing changes in that specific variable while holding the other variables at their
respective base-year values. The Divisia index is a weighted sum of growth rates, where the weights
are the components share in total value, given in the form of a line integral. See C. R. Hulten,
Divisia Index Numbers, Econometrica, vol. 41, no. 6 (1973), pp. 1017-25 for more details, and
B. W. Ang, Decomposition Analysis for Policymaking in Energy: Which is the Preferred
Method? Energy Policy, vol. 32, no. 9 (June 2004), pp. 1131-139.
3
For example, see B. W. Ang and N. Liu, A Cross-Country Analysis of Aggregate Energy and
Carbon Intensities, Energy Policy, vol. 34, no. 15 (October 2006), pp. 2398-404, and L. A.
Greening, M. Ting, and T. J. Krackler, Effects of Changes in Residential End-uses and Behavior
on Aggregate Carbon Intensity: Comparison of 10 OECD Countries for the Period 1970 through
1993, Energy Economics, vol. 23, no. 2 (March 2001), pp. 153-78.
4
J. Myers and L. Nakamura, Saving Energy in Manufacturing (Cambridge, Massachusetts:
Ballinger Publishing Co., 1978).
5
B. W. Ang and F .Q. Zhang, A Survey of Index Decomposition Analysis in Energy and
Environmental Studies, Energy, vol. 25, no. 12 (2000), pp. 1149-176; B. W. Ang, Decom-
position Methodology in Industrial Energy Demand Analysis, Energy, vol. 20, no. 11 (1995), pp.
1081-95; and H. G. Huntington and J. G. Myers, Sectoral Shift and Industrial Energy Demand:
What Have We Learned, in The Changing Structure of American Industry and Energy Use
Patterns, eds. A. Faruqui, J. Broehl, and C. W. Gellings (Columbus, Ohio: Battelle Press, 1987), pp.
353-388.
6
E. Bossanyi, UK Primary Energy Consumption and the Changing Structure of Final De-
mand, Energy Policy, vol. 7, no. 3 (September 1979), pp. 253-58; G. A. Hankinson and J. M. W.
Rhys, Electricity Consumption, Electricity Intensity and Industrial Structure, Energy Economics,
vol. 5, no. 3 (July 1983), pp. 146-52; J. Myers and L. Nakamura, op. cit.; and G. O
stblom, Energy
Use and Structural Change: Factors Behind the Fall in Swedens Energy Output Ratio, Energy
Economics, vol. 4, no. 1 (January 1982), pp. 21-28.
7
The Fisher index was originally a decomposition of an expenditure index into price and
quantity indexes. For details, see Irving Fisher, The Best Form of Index Number, Quarterly
Publications of the American Statistical Association, vol. 17, no. 133 (March 1921), pp. 533-37 and
The Making of Index Numbers, 4
th
ed. (Boston: Houghton-Mifflin, 1972).
8
G. A. Boyd, J. F. McDonald, M. Ross, and D. A. Hanson, Separating the Changing Com-
position of US Manufacturing Production from Energy Efficiency Improvements: A Divisia Index
Approach, The Energy Journal, vol. 8, no. 2 (1987), pp. 77-96.
9
G. E. Metcalf, An Empirical Analysis of Energy Intensity and Its Determinants at the State
Level, The Energy Journal, vol. 29, no. 3 (2008), pp. 1-26.
10
See for example, B. W. Ang, Decomposition Methodology in Industrial Energy Demand
Analysis; B. W. Ang and F. Q. Zhang, op. cit.; and H. G. Huntington and J. G. Meyers, op. cit.
11
G. A. Boyd and J. M. Roop, A Note on the Fisher Ideal Index Decomposition for Structural
Change in Energy Intensity, The Energy Journal, vol. 25, no. 1 (2004), pp. 87-101.
12
The Fisher ideal index is robust to zero and negative values. Other perfect methods cannot
handle zero-negative values without normalization.
13
See, for example, B. W. Ang, Decomposition Methodology in Industrial Energy Demand
Analysis.
14
W. E. Diewert, The Consumer Price Index and Index Number Theory: A Survey, Dis-
cussion paper no. 01-02, Vancouver, British Columbia, Department of Economics, University of
British Columbia, 2001, pp. 1-104, available at http://www.econ.ubc.ca/discpapers/dp0102.pdf,
accessed August 7, 2010.
15
The Fisher index is a geometric mean of the Laspeyres and Paasche indexes.
16
The properties include factor-reversal, time-reversal, zero/negative value robust, etc.
17
For instance, fast-growing countries may be adding more energy-saving infrastructures
than slow-growing economies. On the other hand, fast-growing countries may be less efficient
in their use of energy if their capital investment does not keep pace with growth (e.g., traffic
congestion).
18
m
i
captures the characteristics of the individual country i that are not picked up by the ex-
planatory variables but which are assumed to be time invariant.
19
For details, see S. Nickell, Biases in Dynamic Models with Fixed Effects, Econometrica,
vol. 49, no. 6 (1981), pp. 1417-426.
20
T. W. Anderson and C. Hsiao, Estimation of Dynamic Models with Error Components,
Journal of the American Statistical Association, vol. 76, no. 375 (1981), pp. 589-606 and For-
mulation and Estimation of Dynamic Models Using Panel Data, Journal of Econometrics, vol. 18,
no. 1 (January 1982), pp. 47-82. These instruments will not be correlated with Dv
it
as long as v
it
are
not serially correlated.
21
B. H. Baltagi, Dynamic Panel Data Models, in Econometric Analysis of Panel Data, eds.
B. H. Baltagi, Seuck H. Song, and W. Koh (Chichester, United Kingdom: John Wiley & Sons,
1995), pp. 125-48.
22
M. Arellano and S. Bond, Some Tests of Specification for Panel Data: Monte Carlo Evidence
and an Application to Employment Equations, The Review of Economic Studies, vol. 58, no. 2
(1991), pp. 277-97.
23
Detail discussion on one-step GMM can be found in B. H. Baltagi, op. cit.
24
The regression analysis was undertaken using Stata 11.1.
25
GDP at 2000 PPPs would have been used for the decomposition but the sectoral value-added
data at 2000 US$ PPPs are not adequately available for all the countries. So to avoid bias due to
different data computational units, the study utilizes GDP at 2000 US$ for its decomposition. For
the regression analysis, GDP at 2000 US$ PPPs is used because it provides a better comparison of
economic growth (income) among countries. Also, the time series plots of GDP data (OECD total)
from the two sources are closely related, and this therefore eliminates bias that may be due to
differences in data sources.
26
Data are available at www.iea.org using the 2009 data information.
27
O. I. Adeyemi, D. Broadstock, M. Chitnis, L. C. Hunt, and G. Judge, Asymmetric Price
Responses and the Underlying Energy Demand Trends: Are they Substitutes or Compliments?
Evidence from Modelling OECD Aggregate Energy Demand, Energy Economics, vol. 32, no. 5
(September 2010), pp. 1157-164.
28
L. A. Greening et al., op. cit., note that prices and income are a primary driver (p. 154) of
residential energy demand. They equally note a number of other factors, such as square footage of
housing, number of occupants, etc., that are driven in part by income. While there are other de-
terminants of residential energy demand, arguably, expenditures is not inappropriate for the con-
struction of an index of energy intensity.
29
Note that this decomposition was done by using average final energy consumption, sectoral
energy demand, GDP, and sectoral value added of output for the 16 countries. Using the average of
each of the readily decomposed intensity indexes to arrive at overall average would lead to in-
consistent estimates.
30
The efficiency index for Ireland maintained upward trends throughout the period.
31
These periods are selected because they are the most recent years covered by this study, and
policy formulation is expected to be based on a more recent phenomenon.
32
Greece, Portugal, and Spain are excluded from the computation of this average because
they did not experience improvements in intensity over the period and therefore are considered
to be outliers. This relative contribution varies across the countries; while improvement in
intensity in some countries is largely due to efficiency, others improvement is dominated by
activity.
33
Note that this is done by first finding the average of each of the measures of activities and
energy consumptions before carrying out decomposition analysis. The same years are used for
consistency. Similarly, computation of these contributions on aggregated data (excluding Greece,
Portugal, and Spain) also confirms the overall percentage average reported in table 3.
34
G. E. Metcalf, op. cit.
35
Turning point is calculated as @I
it
=@y
it
=a+2py
it
=00y
1
it
=
a
=
2p
.
36
Note that from table 6, the activity index (column 6) is not affected by current price while the
coefficient on log of income in column 4 is not statistically significant different from zero. This
coefficient is not deleted because deleting it does not improve the results.
37
J. A. Hausman, Specification Tests in Econometrics, Econometrica, vol. 46, no. 6 (1978),
pp. 1251-271.
38
This test was proposed by M. Arellano and S. Bond, op. cit.
39
G. E. Metcalf, op. cit.
40
The short- and long-run elasticity of each index with respect to income equal f
0
+ 2p
0
y
it
and B L + 2C L y
it
=A L . See the section in this article on Econometric Model for the defi-
nitions of these parameters.
41
An economy cannot endlessly substitute non-energy-intensive services for energy-intensive
activities.
42
G. E. Metcalf, op. cit., cites median long-run price elasticity estimates of -0.5 from a survey by
J. Atkinson and N. Manning, A Survey of International Energy Elasticities, in Global Warming
and Energy Demand, eds. B. Terry, P. Elkins, and N. Johnstone (London: Routledge, 1995).
43
These estimates are computed using coefficients estimates from equation (2.4) and observa-
tions on the individual countries per-capita income. The full results are available on request from
the author.
44
A rise in income can result in increasing energy demand and therefore worsens energy intensity.
Alternatively, an increase in income also can lead to the installation of more energy-efficient
machines, resulting in improvements in energy intensity through efficient energy use.
45
This was computed based on data from the International Energy Agency 2009 database. The
percentage was based on the 2005 data. More recent data would have been preferred; unfortunately,
these were not accessible to the author at the time of this research.

"Analysis of Energy Intensity and Its Determinants in 16 OECD Countries," by Musiliu O. Oseni

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

"Analysis of Energy Intensity and Its Determinants in 16 OECD Countries," by Musiliu O. Oseni

Загружено:

Авторское право:

Доступные форматы

THE JOURNAL OF ENERGY

Вам также может понравиться