REVIEW OF ECONOMIC PERSPECTIVES – NÁRODOHOSPODÁŘSKÝ OBZOR,
VOL. 12, ISSUE 1, 2012, pp. 3–21, DOI: 10.2478/v10135-012-0001-y
Changes in the Energy Consumption in EU-27
Countries
Vladimír Hajko1
Abstract: The complete decomposition method is applied to changes in energy
consumption in the countries of EU-27. This method decomposes the changes in energy
use into three different effects: a change in energy consumption due to an increase of
economic activity (the activity effect), a change in energy consumption due to a relative
increase of significance of a country in the group (the structure effect) and a change in
energy consumption due to a change of energy efficiency measured as energy intensity
(the intensity effect). The results confirm that there is a difference in development of
these effects between the old (EU-15) and the new member countries. The results show
that the activity effect is the most significant effect in old member countries (EU-15),
and is on average 1.13 times higher than in new member countries. The intensity effect
is the main diversifying factor between the two groups and the most significant effect
for the new member countries. The intensity effect is almost universally negative in all
countries, and compensates for the other effects. Because of the importance of the
effect, energy intensity convergence is examined. It is found that even by the ''rough''
distinction between the new and the old member countries, the convergence in energy
intensity in new member countries can be found (in the old member countries there is
no energy intensity convergence).
Key words: Decomposition analysis; Energy use; Energy intensity; EU-27
JEL Classification: O13, Q43
Introduction
After the collapse of the Eastern Bloc and its centrally planned economies, its former
members have gone through significant transitions from their previous rather
underdeveloped states. While the economic levels of these states, many of which are
now members of the European Union are not yet on par with original EU countries, it is
to be expected, that even though being very heterogeneous group, there was a different
development in the transition countries than in the ''old'' EU countries.
Technological underdevelopment was one of the many deficiencies of countries with
centrally planned economies. With the absence of entrepreneurship, these countries
lacked skills, technology and capital to match their capitalist counterparts. Since energy
1
Department of Economics, Faculty of Economics and Administration, Masaryk University,
Lipová 507/41a, Brno, Czech Republic. Tel: +420732737843. E-mail address:
vladimir.hajko@mail.muni.cz
REVIEW OF ECONOMIC PERSPECTIVES
4
use plays an important role in virtually every economic activity, it is, to a certain extent2
,
a common proxy of development both in the economic structure and efficiency.
The 2004 and 2007 enlargement that expanded the EU by 12 countries mostly from the
former Eastern bloc, and (if compared with the original EU-15) also with a rather low
economic level as indicated by GDP per capita, the one obvious differentiation is to
observe ''new''3
and ''old''4
member countries separately.
The analysis of the suspected different patterns of development in energy use between
the original EU-15 members and the new member countries is the object of this article.
The grounds for the analysis of the change in energy use are provided by the results of
the complete decomposition method. This method provides greater insight into how and
why has the energy use changed.
This insight is provided by decomposing the change in energy consumption to three
different effects: activity effect, structure effect and intensity effect. As indicated in the
model specification, the essence is to measure an impact of change in one of the factors
while keeping the other ones fixed on values of base year.
By doing so, the activity effect captures the change in energy consumption due to the
overall increase of economic activity while keeping the energy intensities and economic
structure of the countries fixed. The change in energy consumption caused by the
change in the share of activity relative to the overall activity of the aggregate5
is called
the structure effect 6
. Finally, the intensity effect represents a change in energy
consumption due to a change of energy efficiency (measured as energy intensity).
2
Mainly due to the long-run character of extensive capital investments, the development in
energy sectors exhibits sometimes significant inertia
3
“New member countries'' being Bulgaria, Cyprus, Czech Republic, Estonia, Hungary, Latvia,
Lithuania, Malta, Poland, Romania, Slovakia and Slovenia
4
“Old member countries” being Austria, Belgium, Denmark, Finland, France, Germany, Greece,
Ireland, Italy, Luxembourg, Netherlands, Portugal, Spain, Sweden and United Kingdom
5
In this case, the EU-27 is taken as the aggregate. It is possible to perform the analysis in relation
to the aggregate of the world economy; however, due to the fact of comparing to a much larger
base of the world GDP, the structure effect is then rendered rather unsignificant and might be
even considered superfluous. Though the impact of the different aggreagte selection on the results
is not as pronounced with the two other effects, the reader should be aware the choice of the
aggregate Q affects the computations, as follows from the model specification.
6
Thanks to a suggestion of an anonymous referee, it might prove useful to add a note regarding
understanding of structure effect: If one would consider undertaking the decomposition analysis
for a single country, the economic structure of the country might be represented by what is
commonly referred to as economic sectors: primary (production of raw materials), secondary
(production of goods), tertiary (production of services) sometimes even quaternary (research and
education), or maybe even in finer subdivisions like distinguishing transport from other services
etc. In such case, the structure effect would indeed represent changes in energy consumption
resulting from changes in what is commonly referred to as the structure of an individual economy.
For example, we would expect the structural effect to be a significant factor in explanation of rise
of energy consumption, if we observed e.g. a significant increase of industrial production and
transportation. On the contrary, we could observe a low contribution of the structure effect even if
the overall economic development (and energy consumption) would go up if there were no large
shifts in the structure.
Volume 12, Issue 1, 2012
5
The decomposition method itself (the overview of the application of the method can be
found e.g. in Ang and Zhang (2000) covering 124 studies) has been widely used as a
tool to bring more insight into energy use analyses (especially when analyzing the
energy consumption within a country). If applied to the groups of countries, it would
capture similar information as analysis within a country. However, instead of sectors of
an economy, individual countries will be considered.
As indicated above, the usual approach of the energy decomposition handles the
problem by quantifying contributions of three factors – activity, structural, and intensity
effect.
Since the results show that the intensity effect is the most significant difference between
the development in the old and the new member countries, in a short detour on energy
intensity convergence I examine whether these two groups converge to each other in
their energy efficiency (as measured by energy intensity7
), i.e. whether it is reasonable
to assume that differences between energy efficiencies should diminish.
Therefore, the goal of this article is to test the hypothesis that there is an observable
difference in the energy use development between the old (EU-15) and the new member
countries (i.e. the rest of EU-27) by applying the complete decomposition model to the
EU-27 countries in the time period of 1990-2008 and examining the differences in the
effects.
Another hypothesis is that despite the expected gradual and general improvement of
technology, there exists a convergence in energy efficiency between the new and the old
member countries. This means to verify whether the new member countries apply the
new methods and technologies (and replace the old capital stock) in their energy
efficiency investments on greater scale than the old member countries and therefore
converge to the levels of energy efficiency of the ''more developed'' (measured by GDP
per capita) countries of EU-15.
The structure of the article follows this pattern. At first, decomposition model is
presented and applied to the EU-27 countries in the time period 1990-2008. Afterwards,
However, for this analysis, one considers as the economy not just one country but the EU-27
aggregate, and consider the different countries instead of the individual sectors or subsectors of
the economy (e.g. instead of “one way of producing the output”, say mining or agriculture, we
now talk e.g. about Poland, and instead of say the industrial production we now talk e.g. about
Germany). Structural effects shows how energy consumption would change, if EU-27 as a whole
produced the same output as in the base year “differently”. If readers wish so they may consider
this effect under a somewhat less confusing name, e.g. “country effect”.
7
Please note the distinction between the energy intensity indicator and the intensity effect. The
energy intensity indicates how much energy it takes to produce a unit of GDP, here measured in
tones oil equivalent per constant 1,000 USD of the year 2000. The (energy) intensity effect
measures how has the energy consumption changed, therefore is measured in units of oil
equivalent (in this article either in megatons oil equivalent (Mtoe) or in % when this change (in
Mtoe) is divided by the base year energy consumption (in Mtoe). Readers should be aware that
while the approach used here, i.e. energy:GDP ratio, is probably the most common, it is not
without criticism and there are still methodological issues on how to aggregate energy use and
measure the energy efficiency or energy intensity. For further discussion see e.g. Patterson (1996)
and Cleveland, Kaufmann and Stern (2000).
REVIEW OF ECONOMIC PERSPECTIVES
6
I examine energy intensity convergence, namely the beta and sigma convergence of
energy intensity. In the last chapter, conclusions of these parts are then summarized.
The World Bank was the data source for the analysis, the time period covers the years
1990-2009. However, due to insufficient data regarding energy use in 2009 in Bulgaria,
Cyprus, Estonia, Latvia, Lithuania, Malta, Romania and Slovenia, the results for 2009
are incomplete. These countries, and by extension the aggregate of EU-27, are therefore
covered only up to the year 2008. All the tables and figures are the results of my own
calculations.
The country selection is concededly not without controversy, as the EU-27 presents
very heterogeneous group, and given the time frame in the case of new member states, it
was without any doubt also affected by the transformation processes following the
transition to planned economy. Nevertheless, apart from the convergence goal, the
recent trend in European Union of addressing the energy policy on the supranational
level indicates that there is a growing link between energy policies in these countries,
and an analysis within this group may be viable option.
Model
Energy consumption
In essence, an application of the index numbers, the traditional approach (sometimes
called ''traditional Laspeyres index method'' (Ang and Liu (2007)) to the decomposition
analysis was to compute the contribution of each of the aforementioned factors
separately.
The basic relation of change of energy consumption between two years is:
ii
ii
ii
i
SIQ
QQ
QE
QIQE **=*=*= ∑∑
,where
E is overall energy consumption, and Ei is the consumption in sector i
Q is overall economic activity, and Qi is the economic activity in sector i
I is overall energy intensity, and Ii is the energy intensity in sector i
Si is the share of the activity of sector i on the overall economy activity
Energy consumption in year t can be then written as:
t
i
t
i
i
tt
SIQE ∑=
and in the base year 0 as:
Volume 12, Issue 1, 2012
7
0000
= ii
i
SIQE ∑
The activity effect can then be stated as:
)()(= 000
ii
i
t
effect SIQQQ ∑−
The structural effect as:
)(= 000
i
t
ii
i
effect SSIQS −∑
And the intensity effect as:
000
)(= ii
t
i
i
effect SIIQI −∑
The total effect is then:
effecteffecteffect ISQE ++≈∆
However, if we use the ''traditional'' method approach, we are facing the problem of the
unexplained residual – the residual in the country i represents the difference between
actual and explained change, with often significant magnitude (Ang and Liu (2007)).
While some studies omit the residual completely, or prefer the appealing simplicity of
the traditional model even at the cost of only partial explanation, in this case some
preliminary calculations indicate that a more complex model of complete decomposition
might be more appropriate (e.g. using the incomplete decomposition for the period
1990/2008 indicates that the average residual accounts to the magnitude of
approximately -620% of the actual changes8
and almost 16% of the cumulative sum of
the effects). Furthermore, Ang and Zhang (2000) suggest either the logarithmic mean
Divisia index or the refined Laspeyres index method as the preferred choices for the
decomposition method, as these possess several desirable properties (e.g. being zerovalue
robust and passing both time-reversal and factor-reversal test (for a more
comprehensive discussion of the properties of different methods, see Ang and Zhang
(2000) and Fisher (1972)).
The approach of this paper follows the methodology proposed by Sun (1998), i.e. the
influence of the three aforementioned factors is attributed equally to the three main
effects using the principle of ''jointly created and equally distributed'', thus "refining"
the Laspeyres index method and satisfying the desirable properties mentioned by Ang
and Zhang (2000).
The general method of complete decomposition using the principle of jointly created
and equally distributed is as follows.
8
Since the effects compensate for each other, the incomplete decomposition model predicts much
greater changes in energy consumption than what actually happened
REVIEW OF ECONOMIC PERSPECTIVES
8
In a two-factor model: assuming that xyA = , we can calculate the change in A ,
A∆ by:
effecteffect
ttt
YXyxyxAAA +−−∆ === 000
which can be decomposed to :
))(()()(= 000000
yyxxxyyyxxA tttt
−−+−+−∆
In three-factor model: assuming that xyzA = , we can calculate the change in A , A∆
by:
effecteffecteffect
tttt
ZYXzyxzyxAAA ++−−∆ === 0000
,
so the A∆ can be decomposed to :
000000000
)()()(= yxzzzxyyzyxxA ttt
−+−+−∆
))())(((
2
1 00000
yzzzyyxx ttt
−+−−+
))())(((
2
1 00000
xzzzxxyy ttt
−+−−+
))())(((
2
1 00000
xyyyxxzz ttt
−+−−+
))()(( 000
zzyyxx ttt
−−−+
In general, N-factor model:
nxxxA K21=
00
2
0
121= n
t
n
tt
xxxxxxA KK −∆
can be decomposed to =A∆ n terms with one order of ),...2,1,( nixi =∆∆
2!
1)( −×
+
nn
terms with two orders of ),( jixx ji ≠∆∆∆
3!
2)(1)( −×−×
+
nnn
terms with three orders of ),( kjixxx kji ≠≠∆∆∆∆ ...+ + one term
!
22)(1)(
n
nnn ×−×−× K
with n orders of )( 12 nni xxxx ∆∆∆∆∆ −K
.
For further discussion of the derivation, please see Sun (1998) and Sun (1996)). Using
the aforementioned principle, the complete decomposition of the energy consumption
Volume 12, Issue 1, 2012
9
model therefore attributes the previously unexplained combined effects in the residual
to the individual effects:
)()(= 000
ii
i
t
effect SIQQQ ∑−
))()(()(
2
1 00000
i
t
iii
t
ii
i
t
IISSSIQQ −+−−+ ∑
))(()(
3
1 000
i
t
ii
t
i
i
t
SSIIQQ −−−+ ∑
)(= 000
i
t
ii
i
effect SSIQS −∑
))()()((
2
1 00000
i
t
i
t
ii
t
i
i
IIQQQISS −+−−+ ∑
))(()(
3
1 000
i
t
ii
t
i
i
t
SSIIQQ −−−+ ∑
000
)(= ii
t
i
i
effect SIIQI −∑
))()()((
2
1 00000
i
t
i
t
ii
t
i
i
SSQQQSII −+−−+ ∑
))(()(
3
1 000
i
t
ii
t
i
i
t
SSIIQQ −−−+ ∑
Substituting of the individual effects leads to the expression:
)()(= 000
ii
i
t
SIQQE ∑−∆
)( 000
i
t
ii
i
SSIQ −+ ∑
000
)( ii
t
i
i
SIIQ −+ ∑
REVIEW OF ECONOMIC PERSPECTIVES
10
))()(()(
2
1 00000
i
t
iii
t
ii
i
t
IISSSIQQ −+−−+ ∑
))()()((
2
1 00000
i
t
i
t
ii
t
i
i
IIQQQISS −+−−+ ∑
))()()((
2
1 00000
i
t
i
t
ii
t
i
i
SSQQQSII −+−−+ ∑
))(()( 000
i
t
ii
t
i
i
t
SSIIQQ −−−+ ∑
The base values in this analysis follow Laspeyres index method, i.e. the initial year
values were used as a base. The computation of the effects was done for the two given
periods by additive decomposition. As the equations above suggest, the resulting effects
are in absolute values (Mtoe9
). For comparison purposes, the relative values are used as
the % amount of the base year energy use.
Energy intensity convergence
To briefly analyze the evolution of energy intensity between the two groups, beta and
sigma convergence in energy intensity is examined (on the subject of convergence see
e.g. Barro and Sala-i-Martin (1992)).
There is an evidence for beta convergence if negative (and significant) coefficient is
obtained when regressing the growth rates on the initial levels. In other words, the initial
level of energy intensity plays its role in how fast is the lowering of the energy intensity.
If beta convergence is present, with higher levels of initial energy intensity (as we
expect, and also can find in most of the new member countries) we should observe a
faster rate by which the energy intensity is reduced.
The following model was estimated for each group (using panel data model where the
individual effects αi are treated as random and with GLS estimator exploiting the
structure of the error covariance matrix):
t
i
t
ii
t
i uII +−+ −1
log)(1=log βα
t
i
t
iit
i
t
i
ulogI
I
I
+− −
−
1
1
=log βα
The evidence for the sigma convergence is found if the characteristics of the distribution
lead to a lower dispersion. Sala-i-Martin describes when examining the evolution of
income: “Sigma convergence studies how the distribution of income evolves over time
and beta convergence studies the mobility of income within the same
distribution“ (Sala-i-Martin (1996)). In other words, if sigma convergence is present, we
9
Millions of tonnes of oil equivalent, in IEA definition (the amount of energy of burning one
tonne of crude oil, approximately 41.868 GJ or 11.63 MWh)
Volume 12, Issue 1, 2012
11
should be able to observe energy intensities becoming “more similar” in these groups.
Also, the beta convergence is a necessary but not a sufficient condition for the sigma
convergence.
Sigma convergence is examined by evolution of the standard deviation of the logarithms
of energy intensity and the coefficient of variation of energy intensity10
. Decrease of
these two measures over time is interpreted as an evidence for sigma convergence.
Energy consumption
In order to describe changes in energy consumption, a model of a complete
decomposition with three factors described above is utilized. In this paper, the activity
effect captures the additional energy demand of increased economic activity, structure
effect the additional demand due to the shifts between the economic “significance“ of
countries and intensity effect the additional (usually negative) use due to the changes of
energy intensity.
Figure 1: Decomposition effects, 1990-2008
The results of decomposition analysis are summarized in
Table 1. With the exception of Germany11
, all old member countries showed a positive
increase in energy use (in other words, the combined effects summed to positive
change). On the contrary, new member countries with the exception of Cyprus, Malta
and Slovenia, showed decrease in energy use.12
10
Decrease of the measures over time is interpreted as an evidence for convergence, for the new
member countries, the continuous decrease can be found from 1996 onwards
11
Note, however, that this probably illustrates ambiguity of Germany after reunification - while
the Western part is clearly in the pattern with the original “old” member countries, the Eastern
part burdened by the socialist heritage is more akin to the “new” member countries. Nonetheless,
due to lack of data, it is impossible to distinguish this more clearly in the analysis.
12
Again, it is worth noting the intra-group heterogeneity of new member countries (e.g. by IMF
classification, the status of “advanced economies“ only applies to Cyprus, Czech Republic, Malta,
Slovenia and Slovakia)
REVIEW OF ECONOMIC PERSPECTIVES
12
Table 1: Activity, structural and intensity effect, absolute and relative values,
1990-2008
Activity effect Structural effect Intensity effect Actual change
Mtoe % Mtoe % Mtoe % Mtoe %
European Union 644.7 39.4% 0.0 0.0% -529.0 -32.4% 115.7 7.1%
Austria 10.8 43.7% 1.4 5.6% -3.7 -15.1% 8.5 34.2%
Belgium 20.1 41.6% -0.6 -1.3% -9.2 -19.1% 10.3 21.3%
Bulgaria 9.5 33.4% -1.1 -3.8% -17.2 -60.4% -8.8 -30.8%
Cyprus 0.7 52.5% 0.6 43.8% -0.1 -6.3% 1.2 90.0%
Czech Republic 18.0 37.0% -0.8 -1.6% -21.4 -43.9% -4.1 -8.5%
Denmark 6.9 39.8% -0.3 -1.9% -4.9 -28.3% 1.7 9.6%
Estonia 3.1 31.9% 0.6 6.5% -7.8 -82.0% -4.2 -43.6%
Finland 12.0 42.2% 2.0 7.1% -7.1 -25.1% 6.9 24.2%
France 92.3 41.2% -13.4 -6.0% -36.4 -16.2% 42.6 19.0%
Germany 131.6 37.5% -23.7 -6.7% -124.1 -35.3% -16.1 -4.6%
Greece 9.7 45.1% 4.1 19.1% -4.8 -22.3% 9.0 41.9%
Hungary 10.6 37.0% -1.6 -5.6% -11.2 -39.1% -2.2 -7.7%
Ireland 4.9 49.3% 8.5 85.1% -8.4 -84.4% 5.0 50.0%
Italy 60.7 41.4% -24.7 -16.9% -6.5 -4.4% 29.5 20.1%
Latvia 2.5 31.2% -0.6 -8.1% -5.2 -66.0% -3.4 -42.9%
Lithuania 5.0 31.1% -1.7 -10.5% -10.3 -63.7% -6.9 -43.0%
Luxembourg 1.5 43.4% 1.7 48.6% -2.4 -71.3% 0.7 20.7%
Malta 0.3 42.0% 0.2 31.3% -0.4 -55.5% 0.1 17.8%
Netherlands 27.5 41.8% 6.9 10.5% -20.3 -31.0% 14.0 21.3%
Poland 40.3 39.1% 35.0 34.0% -80.6 -78.1% -5.2 -5.1%
Portugal 7.6 45.2% -0.1 -0.7% 0.0 -0.3% 7.4 44.3%
Romania 20.2 32.4% -2.4 -3.8% -40.7 -65.4% -22.9 -36.7%
Slovak Republic 7.8 36.7% 3.1 14.5% -14.0 -65.4% -3.0 -14.2%
Slovenia 2.5 44.1% 1.0 16.9% -1.5 -25.5% 2.0 35.5%
Spain 42.2 46.9% 16.5 18.3% -10.0 -11.1% 48.7 54.1%
Sweden 18.5 39.2% 1.7 3.5% -17.8 -37.7% 2.4 5.1%
United Kingdom 79.6 38.7% 11.2 5.4% -88.3 -42.9% 2.5 1.2%
All countries show positive activity effect, mostly in a rather narrow distribution when
compared to other effects, however, the average activity effect of old member countries
is about 1.13 times higher. This means that the increase of overall economic activity is
more pronounced in the old member states.
The structure effect is highly diversified within each group, with differences as large as
43.8% (percentage of the initial year energy use) for Cyprus and -10.5% for Lithuania
Volume 12, Issue 1, 2012
13
among new member countries, and 85.1% of Ireland or 48.6% of Luxembourg, and -
16.9% and -6.7% of Italy and Germany, respectively, among old member countries. The
change in energy use explained by this effect shows a relative increase of significance in
the structure, if the energy intensities and overall economic activity were fixed.
As for the intensity effect, with the exception of a single country (Cyprus, -6.1%), the
rest of the group of new member countries achieved significant reductions, on average -
58.6% (Cyprus not included), with values as high as -82% for Estonia and -78% for
Poland.13
Among the old member countries, the average intensity effect was -29.6%, the
poorest result being achieved by Portugal (-0.3%) and the best results by Ireland (-
84.4%) and Luxembourg (-71.3%). Given the “focus” of the Luxembourg and Ireland
on the services with high added value like banking and ICT14
, the last results are not as
surprising.
Table 2 presents results of equality of means tests and equality of variances for
individual effects. We can conclude that there is a difference in magnitudes of the
activity effect and the intensity effect, which are big enough to affect the overall change.
The only observable difference in the variability is for the activity effect, which is
higher for the old member countries. However, given the distribution of the activity
effect, the economic activity is projected more into the old member countries than into
the new member countries.
Table 2: Difference of means and variance ratio tests between new and old
member countries, by effects, 1990/2008)
Difference of means = 0 Variance ratio = 1
Test statistics p-value Test statistics p-value
Activity effect -2.70166 0.01221 3.95146 0.01814
Structural effect -0.205992 0.8385 1.88266 0.2967
Intensity effect -2.84077 0.0045 1.09806 0.8906
Actual change -2.19032 0.0285 5.01598 0.006038
The results show that in the period of 1990-2008, despite the difference in the
magnitudes, the development of activity effect was (in relative terms) rather similar in
all member countries amounting to the total of 120.5 and 525.9 megatons of oil
13
Then again, if we take a look on the actual energy consumption numbers in these countries
(103.1 Mtoe in the base year and 97.9 Mtoe in the year 2008 for Poland, and 9.6 Mtoe / 5.4 Mtoe
for Estonia) we can rather observe how inefficient the former ways of production (energy
intensities are still 2.3 times (Poland) and 3.2 times (Estonia) higher than the EU-27 average in
2008)
14
It is worth mentioning that due to the (small) absolute size of the country energy consumption
in these cases (3.4 Mtoe in Luxembourg and 10 Mtoe in Ireland in the base year), all individual
changes leading to higher energy efficiency (lower energy intensity) will be necessarily more
likely to affect the country's position than in case of the large countries (the otherwise identical,
say "moderate size firm", e.g. newly operating bank, even when achieving exactly the same
results, will represent greater share on the indicators of a given country in "small countries")
REVIEW OF ECONOMIC PERSPECTIVES
14
equivalent in new and old member countries, respectively. Total structural effect was
highly diversified in both groups, but its cumulative size did not change the relative
importance of each group very much. Per group it amounted to 32.4 megatons in the
new member countries, and -8.9 megatons in the old member countries. In other words,
the significance of the new member countries as a whole increased, but it was not a
result of any universal pattern. The intensity effect was the most significant effect for
the new member countries, amounting to the total of -210.3 in new and - 344 megatons
in old member countries. The overall change in the new member countries (especially if
not accounting for “atypical” new member countries like Cyprus, Malta and Slovenia)
was much more pronounced in new member countries, and the variability of the
intensity effect in both groups was similar.
My initial hypothesis, that there is a different development between the two groups
measured by the aforementioned three effects is therefore supported by the results. The
activity effect was the most significant effect in the old member countries, the primary
reason for the increase being energy use. While it was on average lower in the new
member countries, its development within the period followed a similar pattern in both
groups. Therefore, energy intensity effect, significantly reducing the energy demands of
increased economic activity (in case of new member countries even outweighing the
other effects) was the main diversifying factor between the groups in the period 1990-
2008.
Given these results, it is therefore worth examining the second hypothesis, i.e. whether
there is an observable convergence in energy efficiency between the new and old
member countries.
Energy intensity
In the period of 1990-2009, the GDP of the EU-27 countries (measured in constant USD
of the year 2000) has increased by USD 3094.7 billion, or by 45.4%, while the energy
use15
has increased only by 115.6 megatons of oil equivalent (i.e. by 7.1%). This
indicates that a significant change in the energy efficiency and therefore in the intensity
indicator has taken place – on average the energy intensity has been reduced by more
than 30%, which was mainly due to lowering energy consumption in new member
countries. The overview of the basic data in the selected group of countries is provided
in Table 3.
As the overall numbers suggest, there is a difference in behavior between the two
groups. While the average growth rate of GDP was 2.21% and 1.96% in old and new
member countries respectively, the growth rates of energy use accounted to 0.79% in the
old, and -0.91% in the new member countries.
As the Table 4 shows, the new member countries increased their share on total GDP by
0.7 p.p., while reducing their energy use by 4.6 p.p. in the period 1990-2008. The
variance ratio test also indicates that variability of energy intensity within the groups
15
Energy use refers to use of primary energy before transformation to other end-use fuels, which
is equal to indigenous production plus imports and stock changes, minus exports and fuels
supplied to ships and aircraft engaged in international transport. Primary energy consumption is
calculated as the sum of the gross inland consumption of energy from solid fuels, oil, gas, nuclear
and renewable sources.
Volume 12, Issue 1, 2012
15
between the years 1990 and 2008 changed only in the new member countries. We can
therefore observe that there was a different development in energy intensity
development between the new and the old member countries.
The results displayed in the Table 5 indicate that there is an evidence for beta
convergence in energy intensity in the EU-27. Further estimation for individual groups
shows that this is the result of beta convergence of the new member countries. The
estimated coefficients are 1.58% and 1.78% (i.e. with 1% increase in the energy
intensity level we can expect the additional 1.58% or 1.78% of negative growth in
energy intensity indicator) for EU-27 and the new member countries respectively (even
though the explanatory power of these very simple models is often rather low, especially
if the heterogeneity in energy intensity of the groups is not captured very well by the
simple division).16
Table 3: overview of GDP and energy use in 1990 and 2008
GDP
(billions of
constant 2000
US$)
Energy use
17
(MTOE)
Energy
intensities
(toe/thousand of
constant 2000
US$))
Country
Name
1990 2008
Relative
change
1990 2008
Relative
change
1990 2008
Relative
change
EU-27 6 813.9 9 908.6 45.4% 1 635.1 1 750.7 7.1% 0.24 0.18 -26.4%
Austria 149.0 227.2 52.5% 24.8 33.2 34.2% 0.17 0.15 -12.0%
Belgium 186.5 268.2 43.8% 48.3 58.6 21.3% 0.26 0.22 -15.6%
Bulgaria 14.6 20.3 39.3% 28.6 19.8 -30.8% 1.96 0.97 -50.3%
Cyprus 6.2 12.3 98.5% 1.4 2.6 90.0% 0.22 0.21 -4.3%
Czech
Republic
55.3 79.2 43.1% 48.8 44.6 -8.5% 0.88 0.56 -36.0%
Denmark 123.9 177.0 42.8% 17.3 19.0 9.6% 0.14 0.11 -23.3%
Estonia 6.0 9.4 57.0% 9.6 5.4 -43.6% 1.60 0.57 -64.1%
Finland 99.3 153.8 54.9% 28.4 35.3 24.2% 0.29 0.23 -19.8%
France 1 091.6 1 504.1 37.8% 223.9 266.5 19.0% 0.21 0.18 -13.6%
Germany 1 543.2 2 097.7 35.9% 351.4 335.3 -4.6% 0.23 0.16 -29.8%
Greece 99.6 169.6 70.2% 21.4 30.4 41.9% 0.22 0.18 -16.7%
16
Note the development of energy intensity in Cyprus, Malta and Slovenia that would, however,
fit very well with development of the old member countries
17
Energy use refers to use of primary energy before transformation to other end-use fuels, which
is equal to indigenous production plus imports and stock changes, minus exports and fuels
supplied to ships and aircraft engaged in international transport.
Primary energy consumption is calculated as the sum of the gross inland consumption of energy
from solid fuels, oil, gas, nuclear and renewable sources.
REVIEW OF ECONOMIC PERSPECTIVES
16
Hungary 44.0 60.5 37.4% 28.7 26.5 -7.7% 0.65 0.44 -32.8%
Ireland 48.5 135.4 179.0% 10.0 15.0 50.0% 0.21 0.11 -46.2%
Italy 937.6 1 171.8 25.0% 146.6 176.0 20.1% 0.16 0.15 -3.9%
Latvia 10.4 13.7 31.7% 7.9 4.5 -42.9% 0.75 0.33 -56.7%
Lithuania 15.9 20.3 27.9% 16.1 9.2 -43.0% 1.02 0.45 -55.5%
Luxembourg 12.4 27.5 121.4% 3.4 4.1 20.7% 0.28 0.15 -45.5%
Malta 2.4 4.5 92.1% 0.7 0.8 17.8% 0.29 0.18 -38.6%
Netherlands 282.0 450.1 59.6% 65.7 79.7 21.3% 0.23 0.18 -24.0%
Poland 118.0 237.7 101.4% 103.1 97.9 -5.1% 0.87 0.41 -52.9%
Portugal 87.5 126.5 44.6% 16.7 24.2 44.3% 0.19 0.19 -0.2%
Romania 44.0 61.2 39.1% 62.3 39.4 -36.7% 1.42 0.64 -54.5%
Slovak
Republic
27.5 46.5 68.7% 21.3 18.3 -14.2% 0.77 0.39 -49.1%
Slovenia 16.6 27.9 67.7% 5.7 7.7 35.5% 0.34 0.28 -19.2%
Spain 440.6 740.9 68.2% 90.1 138.8 54.1% 0.20 0.19 -8.4%
Sweden 201.0 302.4 50.4% 47.2 49.6 5.1% 0.23 0.16 -30.1%
United
Kingdom
1 150.3 1 763.1 53.3% 205.9 208.5 1.2% 0.18 0.12 -34.0%
Average 60.3% 10.0% -30.8%
Average for new member
countries only 58.7%
-7.4% -42.8%
Average for old member countries
only 62.6%
24.2% -21.5%
Table 4: GDP and energy use shares in 1990 and 2008
New member countries
Old member
countries
1990 2008 1990 2008
Increase in % share of new
member countries
GDP share of the group 5.3% 6.0% 94.7% 94.0% 0.7%
Energy use share of the group 20.4% 15.8% 79.6% 84.2% -4.6%
Average energy intensity 0.90 0.45 0.21 0.16
Variance ratio test18
(p-
value)
New Old
Standard deviation 0.5334 0.21756 0.042159 0.035782
0.006041 0.5475
18
H0: the ratio of the two standard deviations is unity (against the two-tailed alternative).
Volume 12, Issue 1, 2012
17
Using the similar non-parametric measures of sigma-convergence as Ezcurra (2007), i.e.
evolution of standard deviation of the logarithms and the coefficient of variation of
energy intensity19
and the distribution characteristics), I conclude there is an evidence
for sigma convergence in the group of the new member countries and in total EU27, as
confirmed also by the results of variance ratio tests.
Table 5: beta convergence results, 1990-2008
New member countries Old member countries EU-27 countries
value t-
statistic
p-
value
value t-
statistic
p-
value
value t-
statistic
p-
value
Constant -0.0423 -10.14 0 -0.01261 -0.9568 0.3395 -0.0409 -10.46 0
Beta -0.0178 -3.255 0.0013 0.0009 0.1222 0.9028 -0.0158 -5.574 0
N 220 285 524
Figure 2: Standard deviations of the logarithms and the coefficients of variation
for energy intensity
There is no evidence for sigma convergence among the old member countries.
19
Decrease of the measures over time is interpreted as an evidence for convergence, for the new
member countries, the continuous decrease can be found from 1996 onwards.
REVIEW OF ECONOMIC PERSPECTIVES
18
Figure 3: Frequency distribution of energy intensity in the old and the new
member countries
1990 2008
NewmembercountriesOldmembercountriesEU-27countries
The results of this chapter support my additional hypothesis that the new member
countries show sigma convergence (note the beta convergence is necessary (but not
sufficient) condition for the sigma convergence), i.e. the energy efficiency in the new
member countries is getting more and more similar (see Figure 3) to the levels of “more
developed” EU-15 countries that were not hampered by the burden of centrally planned
Volume 12, Issue 1, 2012
19
economies and resulting economic damages. Nonetheless, this rate is not very fast and
the fact that the energy intensity still remains much higher in the new member states
indicates consequences of the diversifying factors have not diminished yet. 20
While there is more or less common downward trend, the average energy intensity
remains almost three times higher in the new member countries (Cyprus, Malta and
Slovenia, however, exhibits levels and development more closer to those of old member
countries; it should be noted though that due to their size (their combined GDP amounts
to approx. 0.45% of the EU-27), these countries exhibit higher sensitivity to any
individual activities within the countries). It should be said though that a more complex
analysis of the underlying characteristics of the energy intensity is beyond the scope of
this paper, and may be a subject of further work.
Conclusion
The results verify my initial hypothesis that there is an observable difference between
the decomposed effects of the new and the old member states.
Decomposition of energy consumption shows that energy intensity effect, significantly
reducing the energy demands of an increased economic activity (in case of new member
countries even outweighing the other effects) was the main diversifying factor between
the groups (in the period 1990-2008). The activity effect, despite being the most
substantial effect in the old member countries and the main agent behind their increased
energy use, follows relatively similar development in both groups, but with the new
member countries slightly lagging behind the EU-15 countries.
While the new member countries did show almost twice as big reduction in their energy
intensity (compared to the old member countries), it is necessary to keep in mind that
this reduction was affected by their high starting values; replacing of the outdated
communist era technology thus played a dominant role in this reduction. Even at the end
of the examined period, energy intensity remains on average almost three times higher
in the new member countries. As the activity effect is on average 1.13 times higher in
the old member countries, it shows that the economic growth in the examined period
was more pronounced in the old member countries. This indicates that should strict(er)
energy policies be implemented to achieve the EU Convergence goal21
, in order to
compensate for their lower activity, the new member countries would need even more
energy efficient technologies than the old member countries. This is not only hard to
imagine, but to some extent raises doubts about emphasis on stricter energy efficiency
and energy savings measures (often increasing costs) as an instrument of convergence,
especially if such policies are financed by increased taxes in direct or indirect form.
The most distinct results among the old member countries were the effects for Ireland
and Luxembourg (both of these countries were focused on high added value services
(like ICT and banking) and exhibited large economic growth rates), showing above-
20
In this case, being new member state is almost identical to being post-communist, formerly
centrally planned economy
21
E.g. statements like: ”For both Convergence and the Regional Competitiveness and
Employment objectives an important ERDF priority is to stimulate energy efficiency and
renewable energy production and the development of efficient energy management systems.“
(http://ec.europa.eu/regional_policy/archive/themes/energy/index_en.htm)
REVIEW OF ECONOMIC PERSPECTIVES
20
average results (within the group) for each of the decomposition effects. Among the new
member countries, these positions are held by Malta and surprisingly Poland, indicating
the extension of their economic activities, increasing economic significance and energy
intensity improvements.
The analysis of the main diversifying effect (the intensity effect) indicates that the
division by the groups of the new and the old member countries is not perfect
(especially concerning the non-Eastern-Bloc countries like Cyprus and Malta, showing
development more close to that of old member countries). However, even this ”rough”
grouping shows different convergence behavior in the energy intensity. There is an
evidence for sigma convergence among the new member countries. This means the
distribution of energy intensities of the new member countries is getting closer to that of
the old member countries. The results for the beta convergence (as a necessary
condition of sigma convergence) correspond to this pattern, even though the explanatory
power of the simple model used in this paper may not be very high (in further work, it
might be appropriate to consider the conditional convergence). This convergence is
more evident from the year 1996 onwards (the period 1990-1996 may be affected by the
transition process after the collapse of the centrally planned economies). In short, the
new member countries become more similar in their energy efficiency to the old
member countries and the process has not stopped yet.
This supports my second hypothesis that even though there is more or less a common
downward trend in all EU-27 countries, the new member countries are getting closer in
their energy efficiencies to the EU-15 countries (but they are still not yet on par). While
the convergence is not very fast given the fact there is no observable convergence
among the EU-15 countries, it is possible that the rates of the energy intensity indicator
will even out eventually. There is no evidence for either beta or sigma convergence in
energy intensity among the EU-15 countries (if anything, there may be a slight
indication for energy intensity divergence for the old member countries).
References
ANG, B.W., NA LIU. (2007). Energy decomposition analysis: IEA model versus other
methods. Energy Policy. 35: pp. 1426–1432.
ANG, B.W., ZHANG, F. Q. (2000). A survey of index decomposition analysis in
energy and environmental studies. Energy. 25: pp. 1149–1176.
BARRO, R.J., SALA-I-MARTIN, X.. (1992). Convergence. Journal of Political
Economy.100, pp. 407–443.
CLEVELAND, C.J., KAUFMANN, R.K., STERN, D.I. (2000). Aggregation and the
role of energy in the economy. Ecological Economics. 32, pp. 301–317
EZCURRA R. (2007). Distribution dynamics of energy intensities: a cross-country
analysis. Energy Policy. 35: 5254–5259
FISHER I. (1972). The making of index numbers. 3rd ed. Boston, MA: Houghton
Mifflin.
Volume 12, Issue 1, 2012
21
PATTERSON, M. G. (1996). What is energy efficiency?: Concepts, indicators and
methodological issues. Energy Policy 24: 377–390.
SALA-I-MARTIN, X. (1996). Regional cohesion: Evidence and theories of regional
growth and convergence. European Economic Review. 40: pp. 1325-1352.
SUN, J.W. (1996). Quantitative analysis of energy consumption, efficiency and savings
in the world, 1973-1990. Turku School of Economics Press, series A-4: 1996.
SUN, J.W. (1998). Changes in energy consumption and energy intensity: A complete
decomposition model. Energy Economics. 20: pp. 85-100.