UNIVERSITAT
ROVIRA I VIRGILI
DEPARTAMENT D’ECONOMIA

WORKING PAPERS
Col·lecció “DOCUMENTS DE TREBALL DEL
DEPARTAMENT D’ECONOMIA - CREIP”

The International Distribution of Energy Intensities:
some synthetic results

Juan Antonio Duro

Document de treball n.17- 2015

DEPARTAMENT D’ECONOMIA – CREIP
Facultat d’Economia i Empresa

UNIVERSITAT
ROVIRA I VIRGILI
DEPARTAMENT D’ECONOMIA

Edita:
Departament d’Economia
www.fcee.urv.es/departaments/economia/publi
c_html/index.html
Universitat Rovira i Virgili
Facultat d’Economia i Empresa
Avgda. de la Universitat, 1
43204 Reus
Tel.: +34 977 759 811
Fax: +34 977 300 661
Email: sde@urv.cat

CREIP
www.urv.cat/creip
Universitat Rovira i Virgili
Departament d’Economia
Avgda. de la Universitat, 1
43204 Reus
Tel.: +34 977 558 936
Email: creip@urv.cat

Adreçar comentaris al Departament d’Economia / CREIP

Dipòsit Legal: T - 638 - 2015
ISSN edició en paper: 1576 - 3382
ISSN edició electrònica: 1988 - 0820

DEPARTAMENT D’ECONOMIA – CREIP
Facultat d’Economia i Empresa

The International Distribution of Energy Intensities: some
synthetic results
Juan Antonio Duro

Department of Economics and CREIP, Rovira i Virgili University
juanantonio.duro@urv.cat

Abstract

The paper examines the international distribution of energy intensities as a
conventional proxy indicator of energy efficiency and sustainability in the
consumption of resources, by employing some descriptive tools from the
analysis of inequality and polarization. The analysis specifically focuses on the
following points: firstly, inequalities are evaluated synthetically based on diverse
summary

measures

and

Lorenz

curves;

secondly,

different

factorial

decompositions are undertaken that assist in investigating some explanatory
factors (weighting factors, multiplicative factors and decomposition by groups);
and thirdly, an analysis is made of the polarization of intensities when groups of
countries are defined endogenously and exogenously. The results obtained
have significant implications from both academic and political perspectives.

1

1. Introduction
The uncoupling of energy consumption and economic growth plays a crucial in
achieving a carbon-free society. Thus, for example, based on the European
Union Energy Roadmap 2050, the continental objectives of reducing emissions
by 80% by 2050 imply, in all working hypotheses, the need for substantial
improvements in energy efficiency and hence for reduction in the consumption
of resources, not only in relative terms but, most importantly, in absolute terms.
In fact, the necessity for reducing consumption in absolute terms (as a strong
condition) and for making this compatible with economic growth implies
considerable progress in energy intensities. 1 In such circumstances, the
analysis of this indicator, and its international distribution, are a subject of
academic and political interest.

Energy intensity, as a broad environmental indicator can, in fact, be interpreted
based on a variety of factors (Steinberger and Krausmann, 2011). Firstly, in
accordance with the classical environmental impact descriptive models IPAT
and STIRPAT (Ehrlich and Holdren, 1971), intensity can be associated with
technology. Based on this, the environmental impact on a system (I) is the
product of multiplying population factors (P), affluence (A) and technology (T).
In this context, technology would attempt to balance out the requirements and
excesses attributable to the demands on resources by the population and the
economy. Secondly, the intensities can be interpreted in terms of the sectorial

1

Obviously, then, improvements (i.e. reductions) in energy intensity are a necessary but not
sufficient condition for achieving a strong sustainability. Strong sustainability would require a
reduction in global consumption and consumption would need not merely to be inelastic to GDP
growth, but to have negative elasticity (see Böhringer and Jochem, 2007).

2

structure and its changes. Thus, this indicator can change as a result of
changes in the composition of production in favour of sectors that consume
more or less energy (Schäfer, 2005). Thirdly, this indicator, being linked strictly
to the demand for resources (in this case energy), is associated with a concept
of sustainability and reflects, for example, the relative decoupling between the
economy and the consumption of resources. In a world which, in principle, has
finite energy resources, endless economic growth is only possible if energy
intensity is significantly reduced. Therefore, the analysis of this indicator is
interesting insofar as it relates to the sustainability of the endless development
model and the demand for energy.

However, beyond the analysis of intensity levels, it may be interesting to
examine how this indicator is shared between countries. In this sense, the
analysis would be related to the literature on international environmental equity
(Daly, 1992). In fact, in the environmental field, analysis of the international
distribution of indicators has recently been attracting much interest. Technically,
the analysis of international environmental distribution has relied on two major
approaches, complementary in terms of the subject under study but differing in
the tools employed. Firstly, a series of studies have used convergence analysis
based on the suggestive works of Barro and Sala-i-Martin (1996) and Quah
(1996a, 1996b), which focus on the basic use of - and -convergence
techniques. Secondly, there are studies that use tools developed in the analysis
of inequality (Cowell, 1995), which typically examine distributions with a crosssection focus (similar, in fact, to that used in the analysis of -convergence),

3

which have usually concentrated on the properties of the measurements and
the possibility of their decomposition.

Environmental distribution analysis has been applied to indicators such as CO2,
the carbonization index, emissions intensity (CO2/GDP, the environmental
footprint and the consumption of materials (Hedenus and Azar, 2005; Duro and
Padilla, 2006; White, 2007; Jobert et al., 2010; Duro, Alcántara and Padilla,
2010; Steinberger et al., 2010; Cantore, 2011; Duro, 2012; or Camarero et al.,
2013). In particular, the analysis of the international distribution of energy
intensities is of interest, both academically and in terms of policy, for the
following reasons: firstly, because it helps in understanding the sharing of effort
between countries in terms of relative resource consumption, and, therefore, in
analysing the degree of differential responsibility within the overall context.
Specifically, a situation where intensities are reduced globally, but at the cost of
widening the gaps between countries, has different implications to the generally
preferable situation in which the global change includes a narrowing of the intercountry

gaps.

Secondly,

decomposition

analyses

can

address

some

explanatory factors and may also have some useful policy implications. For
example, as we will see, the decomposition of inequality by groups of countries
(regional groupings, or by level of development) can be used as a guide for a
global design of policy and strategies aimed at reducing inequality. Finally,
analysis of polarization (Esteban and Ray, 1994 and Esteban et al., 2007),
which is a distributive concept essentially different from inequality, seems the
best approach for understanding the possibility of the materialization of
international conflicts implicit in the situation – this being an aspect of great

4

importance for current global trading scenarios. Knowledge of the degree of
energy intensity polarization around poles of distant countries, and also the
factors that characterize these, can be useful in guiding reductions in
polarization, in alleviating the inherent distributive conflict and, thus, in
increasing the chance of reaching international agreements related to this
environmental target.

In terms of the literature, international distributive analysis in terms of energy
intensities has relied on contributions such as those of Sun (2002), Alcántara
and Duro (2004), Markandya et al. (2006), Ezcurra (2007), Duro (2012) and
Herrerías (2012). Sun (2002) analysed a reduction in the inequality of energy
intensity between countries in the Organization for Economic Co-operation and
Development (OECD) by considering deviations from the mean. Alcántara and
Duro (2004) used the Theil index, which weights observations according to
GDP and gives greater importance to those countries with a greater share of
global production. Markandya et al. (2006) used convergence-type analysis to
confirm the relationship between energy intensities and income convergence for
the countries of Eastern Europe, finding a positive ratio between the
convergence of the mean European income (increase) and convergence in
intensities (decrease). Ezcurra (2007) analysed inequalities in energy intensities
between 1971 and 2001 using non-parametric techniques. Duro et al. (2010)
analysed the role of intensities in explaining the difference in consumption per
capita based on multiplicative inequality decompositions for 1980–2006. Duro
(2012) analysed inequalities in energy intensities using different summary
indices for the period 1971–2006 (without carrying out any decomposition

5

analysis); Herrerías (2012) analysed the distribution of intensities between 83
countries during the period 1971–2008 based on dynamic distribution
techniques. In light of the previous literature, this paper aims at extensively
exploiting the analytical possibilities related to different decompositions, at
carrying out a polarization analysis, and at updating the calculations to the
period 1990–2011 and for a greater number of countries (137).

In particular, this paper will provide a detailed study of the different distributive
analysis instruments available for the exploration of energy intensities. Thus, we
can highlight the following differential contributions: firstly, a standard inequality
analysis is performed in parallel with a variety of decompositions. Specifically,
the paper distinguishes the importance of weighting factors as opposed to a
vector of intensities in order to explain global inequality. It makes intensive use
of the possibilities associated with the decomposition by country groupings (by
regions, or by levels of development); the work additionally includes different
multiplicative decompositions for analysing the role played by different factors.
Secondly, a polarization analysis is applied to energy intensity based on both
endogenous and exogenous techniques. This concept is very useful in terms of
understanding both potential instability and the probability of reaching certain
agreements on the world stage.

The paper is therefore structured in the following way: the second section
provides a review of a variety of methodological aspects of interest relative to
the measurement of inequality, its decomposition and the analysis of
polarization. The third section presents the main results associated with the

6

implementation (using an extensive territorial coverage and different groupings)
of the aforementioned techniques and instruments in the analysis of
international inequalities in energy intensities in the period 1990–2011. Finally,
there is a section that brings together the main considerations arising from
these analyses.

2. Methodological Aspects

Intensity in the consumption of resources is thus a highly important objective in
guaranteeing the sustainability of the planet and in being able to balance
economic growth with environmental sustainability. This study addresses the
international distribution of this indicator. Although different distributive
dimensions exist, the main one analysed thus far concerns inequality (or
convergence if preferred). The main methodological elements of interest in
respect of the initial focus come from the literature on inequality measurement
(Cowell, 1995).

The first essential aspect concerns synthetic measurement. In this respect, the
literature describes indicators that are consistent with the Lorenz dominance
criterion.2 Duro (2012), for example, has suggested some suitable candidates
for this measurement in the field of analysing environmental indicators. Thus,
the inequalities could be measured by means of three reasonable indicators:
the coefficient of variation (CV), the Gini coefficient and the Theil index:
2

In other words, indices that are consistent with ordering distributions based on Lorenz curves.
In fact, Lorenz curves associated with different distributions occasionally can intersect, which
makes the calculation of the summary indices even more important (Duro, 2012).

7

CV 

G




(1)

1
 pi p j yi  y j
2 i j


T (   0)   p p log
y
i
 i

(2)






(3)

where pi and pj are GDP-shares (adapted to the analysis of energy intensities);
yi the energy intensity in country "i",  the energy intensity world mean and  w
the weighted standard deviation.
These indices, which are weighted (heterogeneous treatment of the
observations) and measure distances between countries (energy intensities in
this case) have differing properties. In particular, those that weight distances
between countries do so in a different manner. The CV, for example, is neutral
and therefore the position of the observations (i.e. countries) in the rankings has
no importance. The curious property of the Gini coefficient is its automatic
higher sensitivity to changes that occur in observations around the mean; the
Theil index, meanwhile, is especially sensitive to lower intensity changes that
occur (i.e. the lower part of the distribution). Given these differences, it makes
sense to manage the different perspectives with a view to having a complete
overview of the situation and its patterns.

Secondly, apart from measurement, another very interesting property in this
context is the capacity of these indices to decompose, in particular the Theil

8

measurement (Bourguignon, 1979). This study will make use of three possible
decompositions:

1. Given that the indices are weighted (in this case, according to GDP-shares),
their evolution may depend not only on how the intensity vector changes but
also on how the weighting vector changes. Indeed, the reader should note that
energy inequalities could change without any variation in the intensities in each
country, simply through changes in their weightings. In this respect, it could be
interesting to look into this question.3

To do so, Duro (2013), for example, used a shift-share decomposition exercise,
which gives a very intuitive way of clarifying the role of these factors.

In particular, and assuming that a weighted-inequality index (I) depends on
vector “p” for weights (in our case GDP-shares) and “e” for energy intensities it
can be decomposed its temporal variation in the following way:

I  p t 1 , e t 1   I  p t , e t   I  p t , e t 1   I  p t , e t   I  p t 1 , e t 1   I  p t , e t 1  (4)
where pt and pt+1 are the GDP-shares at time t and t+1; et and et+1 are the
energy intensities at time t and t+1; and I(.) is a weighted-inequality index.

Thus, the first term in the RHS of the expression (4) captures the change in the
energy intensities inequality that can be due partly to the changes in the energy
3

Obviously the "problem" would disappear when using unweighted indices. Nevertheless, we
understand that it is more reasonable to use a heterogeneous treatment of the different
countries.

9

intensities vector; and the second term is linked to the role attributable to
changes in GDP-shares vectors.

In fact, apart from its simplicity, one of the main attractions of this
decomposition is that it is generally applicable to any weighted-inequality index.

2. The previous indices, especially the Theil index, can be decomposed
multiplicatively thanks to the properties of the logarithm function. In this respect,
a known multiplicative decomposition adopts the Kaya identity (1989) as a
reference. Based on this expression (adapted in per capita terms) per capita
emissions depend on three main factors: first, the carbon intensity
(emissions/energy consumption) which basically is related to the energy mix the
country has and, therefore, the weight of energy generation through low carbon
sources (i.e. nuclear and renewables); secondly, it the energy intensity factor,
related to the sectoral mix and energy efficiency; and finally the affluence
component, typically measured by GDP per capita. Therefore, we should
consider the following expression:
∗

∗

(5)

where ci are carbon per capita emissions in country "i"; ai is carbon intensity; ei
is energy intensity; and yi is per capita GDP.

Synthetically, and as demonstrated by Duro and Padilla (2006), we can
decompose global inequalities of c, using the Theil index, into the sum of the
partial contribution to inequality attributable to each of the multiplicative factors

10

and two correlation factors, one associated with the covariance between
carbonization and the other factors, and the second associated with the
interaction between intensity and affluence. Given this, we have:

	 	

,

,

(6)

In a similar way, energy intensities can be decomposed as consumption per
capita component (cp) and the inverse of GDP per capita (1/y). In this way, the
inequalities in the intensities can be expressed as follows:

	

/

, /

(7)

This type of analysis aims at contextualizing the role of energy intensity in the
context of international equity in carbon emissions (in both static and dynamic
terms). This we believe to be an interesting contribution to the global debate on
international emission liabilities and on mitigation / reduction strategies. It allows
us, not only to clarify the role of the energy intensity factor in the changing
international

emission

inequality

scenario,

but

also

to

assess

the

appropriateness of focusing future strategies on reducing inequality in this
factor.

3. Global inequalities in energy intensities can be also decomposed into
components of between- and within-group inequality, when countries are
grouped by a criterion of interest. In this respect, for example, it might be of
interest to use the regional criterion provided by the IEA itself, which

11

differentiates new regional groups. Specifically, the between- component
corresponds to the inequality that would exist if the groups were internally
homogenous and the only differences were between group means. The second
consists of capturing the average of the internal inequalities. The Theil index is
best equipped to be decomposed in this way (Shorrocks, 1984). Then, the
decomposition is expressed as follows:

	

	 8

where pg is the GDP-share of group g, T(e)g denotes the internal inequality in
group g, and (e)g represents the average energy intensity in group g.

This decomposition has two major implications, one analytical and the other in
terms of policy. In analytical terms, the weight of the inter-group component
indicates the analytical importance of the established groupings as well as
yielding information about the internal homogeneity of the groups. In policy
terms, this factor might indicate an opportunity to use these aggregations as
units of reference when establishing environmental policy objectives.

Finally, in addition to the previous inequality analysis, polarization is a
distributive concept that is fundamentally different from that of inequality and
has attracted significant attention in recent years. This concept is associated
with the structuring of distribution (in this instance, energy intensities by country)
around homogeneous and opposing groups. Rather than inequality, this
concept is more closely linked to the notion of conflict and potential instability
(Esteban and Ray, 1999). Depending on how the groups are characterized

12

(number of groups, their sizes, internal cohesion and discrepancies), a study of
polarization can offer some guidance on possible strategies.

If we follow the pioneering approach suggested in Esteban and Ray (1993),
polarization depends on the following factor: the number of groups (the fewer
the groups, the greater the polarization); their size (polarization is a matter of
group weight); their internal homogeneity (intra-group cohesion increases
polarization); and the distance between groups. Using these axiomatic
properties, Esteban, Gradín and Ray (2007) suggested a family of multipolarization measures (EGR measures) that allow the concept to be
cardinalized. They are expressed specifically as follows:

n

n

EGR ,     pi p j
i1 j 1

1

ei e j
   G  Gs
e e

(9)

where pi and pj are the GDP-shares of countries i and j; ei and ej are the energy
intensities of countries i and j; e is the world average; is a parameter that
captures the sensitivity of the measure to polarization (its value goes from 1 to
1.6);  is a parameter which captures its sensitivity to the groups’ cohesion
(internal error); G is the Gini coefficient of the original distribution; and Gs is the
Gini coefficient of the grouped distribution (between-group inequality).

At this point, two issues become worthy of comment. Firstly, there is the matter
of the number of groups, which the formula does not establish. Typically, the
studies tackle the analysis of 2, 3 or 4 groups, given that including more groups
usually only produces a marginal gain in the explanatory capacity. The selection

13

of the definitive “ideal” number of groups is determined by the analysis itself (the
higher the EGR value, the better). Secondly, the optimum way of determining
the maximum number of groups should be found, whichever the number
decided on ex-ante might be. For this, Esteban, Gradin and Ray (2007)
recommend using the optimization method proposed by Davies and Shorrocks
(1989).4

Additionally, the literature has produced individual measures for analysing
bipolarization. In this case, the measures calculate the degree to which the
evaluated distribution resembles a symmetrical bimodal distribution, with the
poles located at the ends of the range. One of the best known is the Wolfson
measure (1994), whose additional advantage is its direct derivation from Lorenz
curves. Its usual expression is the following:5
W 

0.5  L(0.5)  G
m

(10)


where L(0.5) corresponds to the median of the Lorenz curve; G is the Gini
coefficient; and m, in this case, is the average per capita intensity.

It may also be appropriate to manage groups by predetermination, as for the
regional IEA groups examined earlier. In this case, we would be implicitly

4

In fact, based on this optimum structuring method, two groups would use the measure as an
inter-group separation value
5

In fact, Esteban, Gradín and Ray (1999) demonstrated that this measure could be rewritten as
an individual case within the EGR family, when 1 and the distributive measure are
replaced by the median.

14

considering the existence of a feeling of identification and proximity between
countries that form a geographical or administrative area. Among the measures
used by exogenous groups, the suggestion put forward in Zhang and Kanbur
(2001), henceforth Z–K, may be used. It basically consists of using group
decomposition components (Shorrocks, 1980 and 1984) of the abovementioned
inequality to reorder them, deriving a measurement consistent with the
axiomatic principles of polarization. Specifically, using the Theil index (Theil,
1967), which is perfectly decomposable in these terms, the Z–K measure is
expressed as the ratio of the between- and within- components:

ZK 

3.

Tb
Tw

(11)

Main Results

The data used in the section come in all cases from the International Energy
Agency (IEA).6 Energy consumption refers to the Total Primary Energy Supply
(million of TPES, Tonnes of Oil Equivalent) and GDP equates to PPP
(Purchasing Power Parity, billion 2005 US dollars). The sample includes 137
countries, representing more than 97% of global energy consumption and GDP.
The reference period of the analysis is from 1990 to 2011. The start point of
1990 coincides with the first year in which data is available for all of the
countries in existence today.

6

http://www.iea.org/statistics/

15

Initially Table 1 shows the main results of energy intensity in a variety of
exogenous country groupings, which represent the regions proposed by the IEA
and the income groups identified by the World Bank. 7 Some initial points of
interest can be observed in these data:

First, energy intensity dropped during the period, going from 0.24 to 0.18,
representing a fall of nearly a quarter. This is a general regional pattern except
in the case of the Middle East.8
Second, this relatively good result, which implies progress towards relative
decoupling, nevertheless coincided with an increase of 48% in total primary
energy consumption and growth of 13% in per capita terms.9
Third, the non-European OECD zones, i.e. Eastern Europe, the Middle East,
China and Africa, appear as those with a higher intensity of energy use, while
the European OECD countries show the least intensity.
Fourth, according to income level and in line with the above, there are the
countries with higher incomes that display, on average, lower intensities (0.16)
and those with low incomes that show higher levels (0.28).
Fifth, the global reduction in intensities has largely depended on the
prominence of Eastern European countries (Russia) and China, with a clear
process of convergence towards the mean.10

7

http://data.worldbank.org/about/country-and-lending-groups

8

A similar pattern has occurred with other resource consumption indicators, as in the case of
the consumption of materials. For example, Krausmann et al. (2009) identified a drop in the
intensity of the consumption of materials during the last hundred years.
9

If fact, this would only occur when the drop in energy intensities is higher than the GDP growth
rate. Unfortunately, Richmond and Kaufmann (2006) found evidence that the turning point
between income and energy use, if it occurs, would happen at very high levels of development,
which exceed the current maximum levels.

16

Sixth, in terms of income, all the groups show falls in intensity, especially the
middle-income bracket.
Seventh, a reduction can be seen in the disparities between regional blocks.
For example, the max-min ratio went from 4.3 in 1990 to 2.8 in 2011.

Insert Table 1 about here

However, the interest lies beyond the partial data mentioned above and is more
concerned with getting an indication of what happened to inequality in energy
intensities by country over the period of the analysis. To achieve this, Figure 1
initially reproduces the Lorenz curves associated with the distribution of energy
intensities by country for the selected years of 1990, 2000 and 2011. It is very
evident that, in Lorenz terms, the distributions for 2000 and 2011 clearly
dominate that of 1990. Therefore, all of the consistent synthetic inequality
indices will be lower in 2000 and 2011 than in 1990. Nevertheless, this
dominance is not reflected when comparing the distribution of 2000 with that of
2011, in view of the intersections observed at the mid-section of the curves.
Given this situation, firstly the summary indices (revised in Section 2) are
necessary and, secondly, these may give contradictory results. In this respect,
in Table 2 three benchmark summary indices of inequality have been
reproduced (weighted) for selected years during the period. Specifically, they
include the coefficient of variation (CV), which is a neutral index (all
10

ln the case of Eastern European countries, Markandya et al. (2006) had already identified
behaviour that was imitative of Western patterns (i.e. a catching-up effect). In the case of China,
Wu (2012), for example, provided a detailed analysis of the explanation of intensities and their
evolution in a national and regional context. In this respect, the importance of the role of energy
efficiency in the evolution of intensities is determined (and by comparison the lesser importance
of the sectorial factor). 

17

observations are treated in the same way regardless of where they are located),
the Gini coefficient (which is especially sensitive to mean observations), and the
Theil index (which is especially sensitive to changes in countries with lower
intensities). The results unanimously indicate a drop in inequalities during the
period, with the Gini and Theil indices showing the process of reduction
stopping between 2005 and 2011. These differences, in fact, are attributed to
the different characteristics of indices dealing with average distances between
observations (Duro, 2012). In particular, the stabilization of the Gini and Theil
indices relative to the CV implies that the Lorenz curves for 2005 and 2011 are
very close until the middle section is reached and then are separated over the
range of observations with higher intensities (i.e. inequality reduction process).
In particular, the approximation to the equity line of the top section of the Lorenz
curve in 2011 is basically explained by the reduction of the relative intensity
experienced by China, which fell from a relative intensity of 1.69 to 1.51
between 2005 and 2011.11 Continuing with comparisons, in particular between
2000 and 2011, it can be seen that the Gini coefficient shows little variation, far
less than the drop experienced in the Theil index and the coefficient of variation.
Whatever the case, and despite the differences in the degree of reduction, all
three measures globally conclude that, synthetically, the international
distribution of intensities has levelled out, leading to a drop in the global
average and its inequality – the best possible scenario for global distribution.

Insert Figure 1 about here

11

 The author can supply further details on request. "Relative" here refers to the world level. 

18

Insert Table 2 about here

At this point, it is worth testing the importance that the weighting factor may
have had in this process. Indeed, given that the data for each country are
weighted according to their GDP-share it could, hypothetically, be possible that
changes in this factor might offer a significant explanation for changes in the
global index. In other words, changes in the previous synthetic indices could
become decomposed into a factor associated with changes in the vector of
GDP-shares and another factor that effectively assimilates the role of changes
in the energy intensities vector. Following Duro (2013), it is possible to clarify
the situation using a decomposition based on the construction of a fictitious
distribution (Equation 4). Table 3 provides the calculations for the evolution in
the overall period, various sub-periods and the three previous summary indices.
Specifically, the data demonstrate that during the overall period the main
changes to weighted inequalities in intensities were indeed attributable to
changes in the intensity vector by country. In any event, in some of the subperiods, such as the 1990s, the role of weighting vectors, in this case GDPshares, played a very important role in reducing summary inequalities (much of
this being attributable to China and its increase in the share of GDP).
Additionally, it can be seen that in the period between 2000 and 2011, changes
in GDP-shares acted to encourage inequalities in intensities rather than reduce
them. In this period, this result was mainly attributable to the US observation
and the change that result from the application of the weighting structure 2000
(GDP-shares) in average energy intensity of 2011. In particular, this application
would reduce the global average energy intensity of 2011 in relation to that

19

actually observed, so that the energy intensity of the US would have come
closer to the global average, reducing inequality compared to that actually
observed in 2011. In fact, in the analysis by sub-period, the role of the weighting
factor is relevant. Therefore, in general terms, it is necessary to proceed with
caution when drawing conclusions on inequalities based only on changes in the
vector of intensities by country.12

Insert Table 3 about here

On the other hand, as reviewed in the foregoing section, it is worth taking
advantage of the decomposition capacities of one of the previous indices to
address some interesting explanatory analyses. In this respect, we propose
three exercises.

First, the intention is to evaluate the explanatory capacity of the intensities
factor as a determining element in CO2 emissions per capita, thereby updating
the work undertaken by Duro and Padilla (2006), whose data ended in 2002. In
this respect, if the Kaya identity is taken as a reference (1989),13 CO2 emissions
(as a major global environmental goal subject to negotiation) can be broken
down into per capita terms, with three main elements: first, the carbonization
index (CO2/energy consumption), which is typically associated with the energy
mix of each country (weight of fossil fuels as part of the total); second, the

12

The aspect of the importance of weighting factors as an element for explaining patterns in the
implementation of distributive analytical tools is also highlighted by authors such as Herrerias
(2012).

13

The IEA typically uses this identity as an example to explain the CO2 levels by country.  

20

intensities which, as mentioned earlier, rely strictly on energy efficiency and the
sectorial mix; and finally affluence, as a factor of economic scale. Following
Duro and Padilla (2006), inequalities in CO2 can, in fact, be broken down into
the sum of the partial contributions of each of the factors plus a series of
factorial correlations. Table 4 shows the main results. These indicate, for
example, that the partial contribution of the energy intensity factor, which is the
factor that concerns this paper, would have been reduced to representing 17%
of international inequalities in CO2 per capita, a weight that is not very far from
that attributable to the carbonization factor, whose weight, in fact, has
increased.14 Whatever the case, the drop in inequalities is reliant, above all, on
the significant reduction in the contribution of the affluence factor, which has
gone from 95% to 80%. In any event, this decomposition illustrates the
significant negative covariance between the energy intensity factor and the
GDP per capita.15

Insert Table 4 about here

14

In this respect, different studies for different samples and countries have established the
importance of the energy efficiency factor and not that of the sectorial component as a powerful
explanation of the reduction in energy intensities and their disparities (Wu, 2012; Mulder et al.,
(2014; Metcalf, 2008; or Duro et al., 2010, among others). The point, in every case, is that the
greater importance of energy efficiency as an explanatory element in the reduction in intensities
would increase the relative weight in the total sectorial component.
15

Regarding this particular decomposition, the results obtained are an update of those of Duro
and Padilla (2006). In that paper, a similar multiplicative decomposition of CO2 was performed
for the period 1971-1999, but with a smaller sample of countries. They observed partial
reduction of the energy intensity inequality contribution up to 1999. It should be noted that there
are differences between the weights obtained in that paper and those in the current paper these differences are explained by differences between the two samples. The present paper
carries out additional exercises that are focussed on energy intensities rather than carbon
emissions.

21

Secondly, these advantages can be used once again to simultaneously
decompose the energy intensity factor into two components: one, the
consumption per capita vector, and the other, the inverse of the affluence
factor, in such a way that the intensities will be lower when consumption per
inhabitant is lower and the GDP per capita is higher. This is a two-factor
decomposition and thus the covariance component is very attractive in that it
reflects the weighted covariance of both factors (Duro and Padilla, 2006). Table
5 illustrates the main results and points to the partial significance of the
affluence factor, and given its reduction, its approximation to the consumption
factor weight.16 Indeed, the partial contribution of the consumption factor drops
quite a lot less. The covariance factor is extremely high. Indeed, if consumption
per capita is a function of income, in the end the bulk of the inter-factorial
covariance would have to be assigned to this factor (Steinberger and
Krausmann, 2011).17 Whatever the case, the partial importance of consumption
at the present day makes it necessary to explore ways of reducing its level and
dispersion across every country.

Insert Table 5 about here

Thirdly, the Theil index, and hence inequalities in intensities, can be
decomposed by groups. In this respect, they need to be broken down into one
16

The partial contribution of each factor is seen as the contribution to inequality that the factor in
question would have if it was the only one that varied throughout the countries (the others
remaining established in the mean). See Duro and Padilla (2006).
17

Indeed, many studies for various samples, using multiple regression analyses, emphasize the
importance of income as an explanatory element of energy intensities (Metcalf, 2008; Wu, 2011,
amongst many others). Thus income convergence would be significantly behind energy
intensity convergence.

22

component that captures the differences between groups of countries and
another which records internal discrepancies. This analysis, in essence,
provides information on the explanatory importance of groupings of countries of
interest in terms of the differences in intensities and their evolution.18 Here, we
are going to use two types of structures: one regional, in line with the groups
identified by the IEA (9) and the three groups of countries grouped by income
identified by the World Bank (as in Table 1). With regard to the regional
groupings, the results show that these provide a good synthetic approximation
of global inequalities. Indeed, almost 70% of international inequalities in energy
intensities can be explained by differences between regional blocks. In fact, the
bulk of the fall in international differences in intensities can be attributed to the
inter-regional-group component (three-quarters of the drop). Meanwhile, the
groups divided according to income do not retain a high explanatory capacity.
Despite their increase, they account for just 15% of global inequalities, which
means they are internally very heterogeneous. Therefore, it seems that energy
intensities and their disparities respond better to regional consumption models
and behavioural patterns.

Insert Table 6 about here

In addition to the previous analysis of inequalities, knowledge of the situation
and the distributive dynamic calls for an analysis of other important concepts
and not just the conventional inequality one. In particular, in recent years there

18

This analysis, drawn up in this context, is similar to the cluster analyses, which are typically
carried out using the convergence approach (Quah, 1996b). 

23

has been a resurgence in the concept of polarization, which is fundamentally
different from that of inequality. In this respect, the aim is not to clarify how
unequal a particular distribution may be – in our case, international intensities –
but rather to explore how this distribution additionally groups around poles that
are both homogenous and distant. As Esteban and Ray (1999) demonstrated,
this is a notion that is closer to that of inherent potential instability and conflict.
To do so, and before going into the different measurements that could
implement this concept, it may be useful to build the density function of the
distribution of energy intensities through non-parametric techniques. This gives
us a visual indication of the shape of the distribution and its change throughout
the period.19 Figure 1 hence reproduces the kernel estimates for three selected
years of the period: 1990, 2000 and 2011. The distribution seems to have
moved from a multimodal situation to being structured around two modes, one
with a mean energy intensity lower than the global average, and a second
smaller one with an intensity 50% higher than the global average. This
movement is due to the combined effect of multiple changes in average energy
intensities and weights across the international energy intensities vector. Three
main patterns underlying the global change in energy intensities, should be
noted: firstly, the relative reduction of the US, which exhibits a fall in from 1 to
0.91 between 2000 and 2011 – this contributing to a clear pole of below
average intensity; secondly, the reduction experienced by China, which causes
a second pole close to the average than before; thirdly, the large reduction in
Russia (from 2.41 in 2000 to 1.92 in 2011), which causes a progressive

19

The estimates are based on Gaussian kernel functions (see Quah, 1996a). These have been
used previously for the analysis of the international distribution of emissions by Padilla and
Serrano (2006) and Ezcurra (2007), among others. The smoothing parameter is determined
endogenously from the method of Silverman (1986).

24

approximation of the fourth peak, to the position in China (see Figure 2).
Overall, the distribution seems to move towards two major poles, one with the
great majority of countries below average, and the other with higher intensity
countries, among which China plays a major role.20

Insert Figure 2 about here

Whatever the case, it is necessary to synthesize the status of the polarization
with cardinal measures. For this, we use the EGR indices for two, three and four
endogenous groups, which are those typically used in the literature, and include
the Wolfson measure, which strictly approximates the degree of bipolarization,
and the Z–K index for the regional groups classified by the IEA. The main data
are provided in Table 7 for the mean EGR parameters. The results demonstrate
the following points of interest: firstly, all the polarization indices point to a drop
in polarization in the aggregate period. In recent years, however, there are
discrepancies. Secondly, focusing on the EGR endogenous indices, the results
indicate that the most interesting structure would be the one that synthesizes
distribution in just two groups, given that the value of the index is higher than
the rest (see Esteban et al., 2007), and in which the relative error was only 27%
in 2011. In 2011, typically, the countries in the group with lower energy
intensities are, broadly speaking, the OECD countries and Latin America.
Meanwhile, the group of countries with higher intensities consist of most of the
countries from Eastern Europe and the former USSR, Asia, Africa and the
Middle East. This being the case, dividing the countries into two groups
20

The author can supply further details on request. 

25

according to their mean energy intensities provides a very reasonable
approximation of the underlying polarization. It can be seen, in this case, that
polarization would have remained fairly stable from 2000. Consequently, the
change in polarization may be due to changes in three factors (see Section 2):
the within-group error (i.e. the internal cohesion component), the sizes of the
groups and, finally, the between-group intensity gap. If we look at the
bipolarization case (Table 7 and Figure 3) we can see a distinct evolution in
these factors which may explain the stabilization of EGR (2) since 2000. Thus,
we note a clear reduction in error (this tends to increase bipolarization), a
reduction in the energy intensity gap between groups (mainly caused by the
effect of China and Russia on the reduction in the average relative intensity in
the group of higher intensities) which tends to diminish bipolarization and,
finally, an equalization process of group sizes (which again increases
bipolarization). In this sense, the Wolfson index, which measures bipolarization
exactly (remembering that it takes as a reference for dividing the groups the
median and not the mean, as EGR(2) does), would have increased even since
2000, emphasizing the relevance of the increasing polarization factors.
Thirdly, the Z–K exogenous polarization index also indicates this worsening
since 2000.

Insert Table 7 about here
Insert Figure 3 about here

26

4. Conclusions and Policy Implications
This paper has analysed the international distribution of energy intensities, as a
sustainability and energy efficiency indicator, based on the instruments
associated with the literature for measuring inequality and polarization. In
particular, with respect to the inequality analysis, it has used different
decompositions of interest, by weights as opposed to energy intensities vector,
by multiplicative factors, and finally by regional country groupings. In terms of
the basic results, we obtain that the fairly generalized reduction in energy
intensities coincided with an improvement in their international equality (good
news), although these advances stopped to a large extent from 2000 and there
are intersections in the Lorenz curves, for example in 2000 and 2011. The main
protagonists of this reduction were the countries of Eastern Europe and the
former USSR (especially Russia) and China. In fact the drop in the disparities of
CO2 emissions per capita in the same period was partially attributable to the
equalizing role played by own energy intensity factor. Although most of the
improvement in international inequalities in energy intensities was due indeed to
changes in the intensities vector the weightings-vector has been relevant in
some sub-periods. Another interesting result found is that exogenous groups of
countries have a high explanatory capacity for global inequalities in intensities
when these follow the regional structure used by the IEA. Indeed, the reduction
in the gap between world regions is the essential reason behind the positive
process of international equalization of energy intensities during the period.
Also the decomposition of intensity inequality into consumption and affluence
factors illustrates the weight of the latter factor in the level of inequalities and

27

how they change. And, finally, the analysis of the polarization of intensities,
which is a fundamental different dimension from that of inequality, demonstrates
their reduction and good informative synthesis through bimodal distribution.
However, the advance towards bimodality and particularly the stability of this
polarization since 2000 give rise to certain doubts about the distributive process
in these terms.

The above results yield, first of all, a pair of academic and analytical
implications that need to be taken into account. Care must be taken when
extracting unequivocally precise results based on the observation of specific
environmental inequality indices. The intersection of the Lorenz curves, and
discrepancies in the size of the changes according to the summary measures
used, demonstrate the need to manage a full complement of measures and
examine the status of the curves. Moreover, the use of weighted-inequality
measures to analyse the international distribution of intensities demonstrates
the importance of weighting factors in different sub-periods and the need to take
these into account when interpreting the results in specific years.

On the other hand, some useful policy implications may be derived from the
main results found. For instance, and in spite of the reduction in world mean
intensities and their inequality across countries, energy consumption in per
capita terms has increased rapidly since 1990. True environmental success will
come about when there is a reduction in intensities on a scale that exceeds the
actual rate of economic growth and, therefore, implies a decrease in per capita
energy consumption. This is the big challenge in environmental and energy

28

policy: the absolute decoupling of energy and economic growth as a key
element in a sustainable development strategy. It would be necessary to
intensify energy-saving technologies, improve the use of energy resources and
encourage activities that do not use energy intensively worldwide. This result
has not been achieved thus far. On the other hand, and despite the overall
reduction in energy intensity inequalities, which is positive, it is also true that
there seems to be a process of stabilization in recent years. It is therefore of
interest that the downward trend in inequality recommenced. This would
depend crucially on outlier countries like China and Russia (in 2011 their energy
intensity relative to the global average was still 1.5 and 1.9 respectively) making
progress on converging to the global average energy intensity. In this regard,
further reductions will have to come from the international convergence on the
main

factors

decomposition

that

explain

analysis

has

the

energy

stressed

intensities.
the

Among

importance

of

them,
the

our

income

convergence process as a global factor and hence we would need to address
the factors which explain it, technology, human resources and sectorial
structure. But this income convergence must be accompanied by a policy on
intensity of resource use which has a sustainability objective. Moreover, we
found that inequality in intensities is a very regional phenomenon. In this work,
we have taken as reference the nine regions proposed by the World Bank,
which basically arise from geographical criteria. Therefore, as a political tool it
would make sense to use the regions as an essential foundation for designing a
global sustainability policy for energy and the consumption of resources, if
necessary. And, finally, the polarization analysis (which is generally related to
the notion of conflict) reveals a clear consolidation (on endogenous criteria,

29

rather than exogenous ones such as regions) of two poles of countries
according to their energy intensities, one above the world average and one
below. In particular, there is evidence that when the distribution of emissions is
simplified into two poles (i.e. analysis of bipolarization) there would have been a
stabilization (or even an increase according to some measures such as
Wolfson), in polarization since 2000. This pattern could hamper country-based
negotiations. This scenario is not unrealistic since countries that consume more
resources argue (to relax their responsibility and avoid having to make
sacrifices), that their resource-use intensity is lower. It is interesting, therefore,
to encourage a process of intensity equalization (consistent with a reduction in
the global levels) between groups of countries, which can reduce the
bipolarization. This might be achieved, either through the known mechanisms of
technological diffusion and energy savings, or directly through a process of
economic convergence (or indeed by a combination of both). Thus, because the
inherent lower potential instability, a reduction in the international polarization of
energy intensities may favour international environmental agreements (Esteban
and Ray, 1999).

ACKNOWLEDGEMENTS
The author is grateful to Jordi Teixidó-Figueras and to the financial support
made by the project ECO2013–45380–P

30

References
Alcántara, V. and Duro, J.A., (2004, “Inequality of Energy Intensities Across
OECD Countries: a Note”, Energy Policy 32, 1257–1260.
Barro, R.J. and Sala-i-Martin, X., (1992), “Convergence”, Journal of Political
Economy, 100, 223–251.
Böhringer, C., Jochem, P., (2007), “Measuring the immeasurable – A Survey of
sustainability indices”, Ecological Economics 63, 1–8
Bourguignon, F., (1979),
Econometrica 47, 901–920.

“Decomposable

income

inequality

measure”.

Camarero, M, Picazo-Tadeo, A.J., Tamarit C.R., (2013),”Are the determinants
of CO2 emissions converging among OECD countries?, Economics Letters
118, 159–162.
Cowell, F., (1995), Measuring Inequality, Blackwell.
Davies J, Shorrocks A., (1989), “Optimal Grouping of Income and Wealth Data”,
Journal of Econometrics 42, 97–108.
Daly, H. E., (1992), “Allocation, distribution, and scale: Towards an economics
that is efficient, just, and sustainable”, Ecological Economics 6, 185–193.
Duro, J.A., (2012), “On the automatic application of inequality indices in the
international environmental distributive analysis”, Ecological Economics, 76, 1–
7.
Duro, J.A. (2013), “Weighting vectors and international inequality changes in
environmental indicators: an analysis of CO2 per capita emissions and Kaya
factors”, Energy Economics, 39, 122–127.
Duro, J.A., Padilla, E., (2006), “International inequalities in per capita CO2
emissions: a decomposition methodology by Kaya factors”, Energy Economics
28, 170–187.
Duro, J.A. Alcántara, V and Padilla, E., (2010), “International inequality in
energy intensity levels and the role of production composition and energy
efficiency: an analysis of OECD countries”. Ecological Economics 69, 2468–
2474.
European Commission (2011), Energy Roadmap 2050, Brussels.
Ehrlich, P.R. and Holdren, J.P., (1971), “Impact of population growth”, Science,
171, 1212–1217.

31

Esteban J, Gradín C, Ray D., (2007), “An Extension of a Measure of
Polarization, With an application to the income distribution of five OECD
Countries”, Journal of Economic Inequality 5, 1–19.
Esteban J, Ray D., (1994), “On the Measurement of Polarization”, Econometrica
62, 819–852.
Esteban J, Ray D., (1999), “Conflict and Distribution”, Journal of Economic
Theory 87, 379–415.
Ezcurra, F., 2007. “Distribution dynamics of energy intensities: a cross-country
analysis,” Energy Policy 35, 5254–5259.
Hedenus, F. and Azar C., (2005), “Estimates of trends in global income and
resource inequalities”, Ecological Economics 55, 351–364.
Herrerías, M.J., (2012), “World energy intensity convergence revisited: a
weighted distribution dynamics approach”, Energy Policy, 49, 393–399.
Jobert, T., Karanfil, F., Tykhonenko, A., (2010), Convergence of per capita
carbon dioxide emissions in the EU. Legend or reality?, Energy Economics 32,
1364–1373.
Kaya, Y., (1989), Impact of Carbon Dioxide Emission Control on GNP Growth:
Interpretation of Proposed Scenarios. Paper Presented to the Energy and
Industry Subgroup, Response Strategies Working Group, Intergovernmental
Panel on Climate Change, Paris, France.
Markandya, A., Pedroso-Galinato, S. and Streimikiene, D., (2006), “Energy
intensity in transition economies: is there convergence towards the EU
average?, Energy Economics, 28, 121–145.
Metcalf, G.E., (2008), “An empirical analysis of energy intensity and its
determinants at the State level”, Energy Journal 29 (3), 1–26.
Mulder, P, de Groot, H. And Pfeiffer, B., (2014), “Dynamics and determinants of
energy intensity in the service sector: a cross-country analysis, 1980–2005”,
Ecological Economics, 100, 1–15.
Quah, D., (1996a), “Convergence empirics across economics with (some)
capital mobility”, Journal of Economic Growth 1, 95–124.
Quah, D., (1996b), “Regional convergence clusters in Europe”, European
Economic Review, 40,
Schäfer, A., (2005), “Structural change in energy use”, Energy Policy, 35, 429–
437.

32

Silverman, B.W.,(1986), “Density estimation for statistics and data analysis”. In:
Monographs on Statistics and Applied Probability, vol 26. Chapman and Hall,
London.
Shorrocks, A., (1980), “The class of additively decomposable inequality
measures”, Econometrica 48, 613–625.
Shorrocks, A., (1984), “Inequality decomposition by population subgroups”,
Econometrica 52, 1369–1386.
Steinberger, J. K. and Krausmann, F., (2011), “Material and Energy
Productivity”, Environmental Science and Technology, 45, 1169–1176.
Steinberger, J. K., Krausmann, F., and Eisenmenger, N. (2010). “Global
patterns of materials use: A socioeconomic and geophysical analysis”,
Ecological Economics, 69(5), 1148–1158.
Sun, J.W., (2002), “The Decrease in the Difference of Energy Intensities
between OECD Countries from 1971 to 1998”, Energy Policy 30, 631–635.
Theil, H., (1967), Economics and Information Theory. Amsterdam, North
Holland.
White, T. J. (2007), “Sharing resources: The global distribution of the ecological
footprint”, Ecological Economics, 64(2), 402–410.
Wolfson M., (1994), “When Inequalities Diverge”, American Economic Review
P&P 94, 353–358.
Wu, Y. (2012), “Energy intensity and its determinants in China’s regional
economies”, Energy Policy, 41, 703–711.
Zhang X, Kanbur, R., (2001), “What Difference do Polarization Measures
Make? An application to China”, Journal of Development Studies 37, 85–98.

33

Tables
Table 1: Energy Intensities by Regional Groups, selected years 1990-2011

OECD Americas
OECD Asia Oceania
OECD Europe
Non-OECD Europe
Africa
Asia
China
Non-OECD Americas
Middle East
Low-income
Middle-income
High-income
World
Ratio max/min reg groups

1990
0,2345
0,1490
0,1597
0,4811
0,2941
0,2461
0,6342
0,1525
0,1892
0,3466
0,3013
0,2111
0,2372
4,26

1995
0,2250
0,1563
0,1514
0,5408
0,3101
0,2309
0,4387
0,1445
0,2491
0,3556
0,2744
0,2010
0,2240
3,74

2000
0,1996
0,1581
0,1369
0,4677
0,2936
0,2258
0,3291
0,1486
0,2329
0,3518
0,2416
0,1847
0,2037
3,42

2005
0,1836
0,1474
0,1304
0,3686
0,2806
0,2063
0,3186
0,1464
0,2553
0,3262
0,2365
0,1721
0,1956
2,83

2011
0,1645
0,1370
0,1151
0,3214
0,2495
0,1821
0,2666
0,1338
0,2599
0,2773
0,2128
0,1572
0,1814
2,79

Source: Own elaboration based on International Energy Agency data (2013)

Table 2: Cross-country inequality in energy intensities through summary
indices, selected years 1990-2011
CV

Gini

Theil

1990

0,0874

0,2847

0,1303

1995

0,0708

0,2605

0,1110

2000

0,0507

0,2309

0,0873

2005

0,0411

0,2243

0,0813

2011

0,0339

0,2223

0,0794

Source: Own elaboration based on International Energy Agency data (2013)

34

Table 3: Decomposing international inequality changes in energy
intensities between GDP-share changes and energy intensity changes
over selected periods and for different inequality indices

C.V.

Gini

Total

E.I.
-0,0156

GDP

Total

share

Change
-0,0367

Change

-0,0211

-0,0169

(43%)

(57%)

-0,0173

1990-2000

1990-2011

-0,0536

(103%)

(-3%)
(16%)

GDP

Total

share

Change

-0,0285

(47%)

(53%)

-0,0180
-0,0085

-0,0088

(84%)

E.I.
-0,0254

-0,0538

0,0004

-0,0447

2000-2011

Theil

-0,0530
-0,0623

(-112%)
(15%)

share
(46%)

(54%)

-0,0138
-0,0079

-0,0094

(85%)

GDP

-0,0199 -0,0231
-0,0430

0,0095

(212%)

E.I.

0,0059

(174%)

(-74%)

-0,0432 -0,0077
-0,0509

(85%)

Source: Own elaboration based on International Energy Agency data (2013)

Table 4: Decomposition of CO2 emissions per capita into Kaya factors,
1990-2001
Carbon
0,1365

2011

(15%)

(21%)

(95%)

(19%)

(-51%)

0,1457

0,7565

0,1144

-0,3770

(18%)

(19%)

(97%)

(15%)

(-48%)

0,1275

0,7253

0,1058

-0,3400

(18%)

(17%)

(96%)

(14%)

(-45%)

0,1212

0,6434

0,1027

-0,2942

(19%)

(17%)

(91%)

(14%)

(-41%)

0,1352

2005

Cov 2
-0,4522

0,1376

2000

Cov 1
0,1689

0,1358

1995

Affluence
0,8433

0,1403

1990

Intensity
0,1889

0,1100

0,5229

0,1068

-0,2228

(21%)

(17%)

(80%)

(16%)

(-34%)

Note cov 1 is related to the component, which includes the covariance among carbon intensity
and the rest of factors; cov 2 is related to the component which includes the covariance among
energy intensity and affluence.
Source: Own elaboration based on International Energy Agency data (2013)

35

(15%)

Table 5: Decomposition of energy intensity into consumption and
affluence, 1990-2001

1990

Consumption
0,3111

Affluence
0,7305

Cov
-0,9113

1995

0,3245

0,7047

-0,9182

2000

0,3333

0,6921

-0,9381

2005

0,3126

0,6177

-0,8490

2011

0,2723

0,4875

-0,6805

Source: Own elaboration based on International Energy Agency data (2013)

Table 6: Decomposition of energy intensities inequality by subgroups
components, 1990-2011

2000

0,0873

2005

0,0813

2011

0,0794

Within
0,1162

(71%)

(29%)

(11%)

(89%)

0,0343

0,0116

0,0994

(69%)

(31%)

(10%)

(90%)

0,0285

0,0096

0,0777

(67%)

(33%)

(11%)

(89%)

0,0262

0,0132

0,0682

(68%)

(32%)

(16%)

(84%)

0,0540

0,1110

Income
Between
0,0141

0,0551

1995

0,1303

Within
0,0383

0,0588

1990

Regional
Between
0,0920
0,0767

Total

0,0254

0,0123

0,0671

(68%)

(32%)

(15%)

(85%)

Source: Own elaboration based on International Energy Agency data (2013)

36

Table 7: Polarization of energy intensities according to the EGR family of
indices, Wolfson’s measure and Z-K measure
EGR

ε/Gini

(2)

EGR

ε/Gini

(3)

EGR

ε/Gini

W

Z-K

(4)

reg

1990

0.0811

29.7%

0.0631

9.9%

0.0429

6.0%

0.1223

2,4049

1995

0.0739

29.8%

0.0604

8.9%

0.0465

11.4%

0.1146

2,2333

2000

0.0665

31.3%

0.0522

10%

0.0389

10.1%

0.0854

2,0589

2005

0.0673

28.0%

0.0481

9.7%

0.0441

12.0%

0.0966

2,1062

2011

0.0664

26.9%

0.0469

10%

0.0358

6.6%

0.1017

2,1285

Note:  is the absolute error associated with the groupings and  Gini is the relative error,
which can be considered as a proxy of the explanatory capacity of the groupings
Source: Drawn up by the authors using the World Bank data set

37

Figures
Figure 1: Evolution of density functions of the international distribution of
energy intensities, selected years 1990–2011
1

0,9

0,8

0,7

0,6

0,5

0,4

0,3

0,2

0,1

0
0

0,1

0,2

0,3

0,4
1990

0,5

0,6

0,7

0,8

0,9

1

2000

Source: Own elaboration based on International Energy Agency data (2013)

38

0

.5

density

1

1.5

Figure 2: Evolution of density functions of the international distribution of
energy intensities, 1990–2011

0

.5

1

1.5

2
2.5
3
3.5
relative energy intensity

1990

2000

4

4.5

5

2011

Source: Own elaboration based on International Energy Agency data (2013)

39

ANNEX
EGR2 2011
Low-energy intensities group:
Hong Kong, Colombia, Peru, Ireland, Switzerland, Panama, Botswana,
Dominican Republic, Albania, Malta, United Kingdom, Costa Rica, Gabon,
Other Asia, Uruguay, Denmark, Spain, Portugal, Sri Lanka, Italy, Congo,
Greece, Tunisia, Austria, Germany, Israel, Cyprus, Turkey, El Salvador,
Ecuador, Lebanon, Philippines, Japan, Luxembourg, Croatia, Singapore,
Morocco, Norway, Netherlands, Argentina, Mexico, Chile, France, Angola,
Bangladesh, Brazil, Lithuania, Chinese Taipei, Slovenia, Yemen, Latvia,
Hungary, Australia, Poland, Sweden, Algeria, Paraguay, Slovak Republic,
Romania, Nicaragua, Azerbaijan, Guatemala, Senegal, FYR of Macedonia,
Cameroon, Belgium, Jamaica, Georgia, New Zealand, United States, Tajikistan,
Bolivia, Egypt, Honduras, Armenia, Czech Republic, Cuba, Sudan
High-energy intensities group:
Other Non-OECD Americas, Malaysia, DPR of Korea, Myanmar, India, Korea,
Gibraltar, Pakistan, United Arab Emirates, Finland, Canada, Brunei
Darussalam, Syrian Arab Republic, Indonesia, Venezuela, Jordan, Bulgaria,
Thailand, Serbia, Qatar, Estonia, Vietnam, Belarus, Kuwait, Bosnia and
Herzegovina, Ghana, Islamic Rep. of Iran, Kyrgyzstan, People's Rep. of China,
South Africa, Benin, Other Africa, Haiti, Mongolia, Nepal, Saudi Arabia,
Republic of Moldova, Kenya, Nigeria, United Rep. of Tanzania, Bahrain, Libya,
Russian Federation, Oman, Côte d'Ivoire, Iraq, Kazakhstan, Ethiopia, Ukraine,
Zambia, Togo, Mozambique, Iceland, Uzbekistan, Turkmenistan, Trinidad and
Tobago, Netherlands Antilles, Dem. Rep. of Congo, Zimbabwe

EGR3 2011
Low-energy intensities group:
Hong Kong, Colombia, Peru, Ireland, Switzerland, Panama, Botswana,
Dominican Republic, Albania, Malta, United Kingdom, Costa Rica, Gabon,
Other Asia, Uruguay, Denmark, Spain, Portugal, Sri Lanka, Italy, Congo,
Greece, Tunisia, Austria, Germany, Israel, Cyprus, Turkey, El Salvador,
Ecuador, Lebanon, Philippines, Japan, Luxembourg, Croatia, Singapore,
Morocco, Norway, Netherlands, Argentina, Mexico, Chile, France, Angola,
Bangladesh, Brazil, Lithuania, Chinese Taipei, Slovenia, Yemen, Latvia,
Hungary, Australia

40

Medium-energy intensities group:
Poland, Sweden, Algeria, Paraguay, Slovak Republic, Romania, Nicaragua,
Azerbaijan, Guatemala, Senegal, FYR of Macedonia, Cameroon, Belgium,
Jamaica, Georgia, New Zealand, United States, Tajikistan, Bolivia, Egypt,
Honduras, Armenia, Czech Republic, Cuba, Sudan, Other Non-OECD
Americas, Malaysia, DPR of Korea, Myanmar, India, Korea, Gibraltar, Pakistan,
United Arab Emirates, Finland, Canada, Brunei Darussalam, Syrian Arab
Republic, Indonesia, Venezuela, Jordan, Bulgaria.
High-energy intensities group:
Thailand, Serbia, Qatar, Estonia, Vietnam, Belarus, Kuwait, Bosnia and
Herzegovina, Ghana, Islamic Rep. of Iran, Kyrgyzstan, People's Rep. of China,
South Africa, Benin, Other Africa, Haiti, Mongolia, Nepal, Saudi Arabia,
Republic of Moldova, Kenya, Nigeria, United Rep. of Tanzania, Bahrain, Libya,
Russian Federation, Oman, Côte d'Ivoire, Iraq, Kazakhstan, Ethiopia, Ukraine,
Zambia, Togo, Mozambique, Iceland, Uzbekistan, Turkmenistan, Trinidad and
Tobago, Netherlands Antilles, Dem. Rep. of Congo, Zimbabwe.

EGR4 2011
Low-energy intensities group:
Hong Kong, Colombia, Peru, Ireland, Switzerland, Panama, Botswana,
Dominican Republic, Albania, Malta, United Kingdom, Costa Rica, Gabon,
Other Asia, Uruguay, Denmark, Spain, Portugal, Sri Lanka, Italy, Congo,
Greece, Tunisia, Austria, Germany, Israel, Cyprus, Turkey, El Salvador,
Ecuador, Lebanon, Philippines, Japan, Luxembourg, Croatia, Singapore,
Morocco, Norway, Netherlands, Argentina, Mexico, Chile, France, Angola,
Bangladesh.
Lower-Mid energy intensities group:
Brazil, Lithuania, Chinese Taipei, Slovenia, Yemen, Latvia, Hungary, Australia,
Poland, Sweden, Algeria, Paraguay, Slovak Republic, Romania, Nicaragua,
Azerbaijan, Guatemala, Senegal, FYR of Macedonia, Cameroon, Belgium,
Jamaica, Georgia, New Zealand, United States, Tajikistan, Bolivia, Egypt,
Honduras, Armenia, Czech Republic, Cuba, Sudan

Upper-Mid energy intensities group:
Other Non-OECD Americas, Malaysia, DPR of Korea, Myanmar, India, Korea,
Gibraltar, Pakistan, United Arab Emirates, Finland, Canada, Brunei
Darussalam, Syrian Arab Republic, Indonesia, Venezuela, Jordan, Bulgaria.

41

High-energy intensities group:
Thailand, Serbia, Qatar, Estonia, Vietnam, Belarus, Kuwait, Bosnia and
Herzegovina, Ghana, Islamic Rep. of Iran, Kyrgyzstan, People's Rep. of China,
South Africa, Benin, Other Africa, Haiti, Mongolia, Nepal, Saudi Arabia,
Republic of Moldova, Kenya, Nigeria, United Rep. of Tanzania, Bahrain, Libya,
Russian Federation, Oman, Côte d'Ivoire, Iraq, Kazakhstan, Ethiopia, Ukraine,
Zambia, Togo, Mozambique, Iceland, Uzbekistan, Turkmenistan, Trinidad and
Tobago, Netherlands Antilles, Dem. Rep. of Congo, Zimbabwe.

42