UNIVERSITAT
ROVIRA I VIRGILI
DEPARTAMENT D’ECONOMIA 

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

The causal factors of international inequality in CO2
emissions per capita: A regression-based inequality
decomposition analysis

Juan Antonio Duro
Jordi Teixidó Figueras
Emílio Padilla
Document de treball n.26- 2016

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
Av. de la Universitat, 1
43204 Reus
Tel.: +34 977 759 811
Fax: +34 977 758 907
Email: sde@urv.cat

CREIP
www.urv.cat/creip
Universitat Rovira i Virgili
Departament d’Economia
Av. de la Universitat, 1
43204 Reus
Tel.: +34 977 758 936
Email: creip@urv.cat

Adreçar comentaris al Departament d’Economia / CREIP

ISSN edició en paper: 1576 - 3382
ISSN edició electrònica: 1988 - 0820

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

The causal factors of international inequality in CO2 emissions
per capita: A regression-based inequality decomposition
analysis.
Juan Antonio Duroa, Jordi Teixidó-Figuerasa and Emilio Padillab
a

Departament d’Economia and CREIP, Universitat Rovira i Virgili, Reus. Spain.

b

Department of Applied Economics, Univ. Autónoma de Barcelona, Campus de

Bellaterra, Cerdanyola del Vallès, 08193, Spain. Tel. +34 93 581 1276
(Corresponding author).

Abstract
This paper uses the possibilities provided by the regression-based inequality
decomposition (Fields, 2003) to explore the contribution of different explanatory
factors to international inequality in CO2 emissions per capita. In contrast to
previous emissions inequality decompositions, which were based on identity
relationships (Duro and Padilla, 2006), this methodology does not impose any a
priori specific relationship. Thus, it allows an assessment of the contribution to
inequality of different relevant variables. In short, the paper appraises the
relative contributions of affluence, sectoral composition, demographic factors
and climate. The analysis is applied to selected years of the period 1993–2007.
The results show the important (though decreasing) share of the contribution of
demographic factors, as well as a significant contribution of affluence and
sectoral composition.
JEL classification: C19; D39; Q43.
Keywords: CO2 emissions, international emissions inequality, regression-based
decomposition.

1

1. Introduction
There is a huge inequality in the international distribution of CO2 emissions.
Differences in emissions per capita and in their determinants lead to different
perceptions of and interests in the criteria to distribute abatement efforts among
countries—and even of the ambition of mitigation goals—and so hamper
mitigation agreements. A correct design of policies should appropriately take
into account these inequalities, which show different responsibilities for the
problem, as well as the different drivers of them. A wider participation in
international abatement agreements on the part of developing and emerging
economies would be facilitated by the perceived fairness of abatement sharing.
This perceived fairness will increase if countries with greater responsibility in the
problem are charged with the most important part of mitigation efforts.
Agreements, as the past ones, assuming a very uneven distribution of the CO2
abortion capacity in the atmosphere and not involving strong efforts by the main
countries responsible for causing the problem would tend to disincentivise the
participation of developing countries. These countries claim that the global sink
capacity is being disproportionally appropriated by the inhabitants of richer
countries, which are also responsible for past overuse leading to the
intensification of the greenhouse effect resulting in accelerated global warming.
On the other hand, the greater the degree of inequality, the more reluctant the
main emitters may be to participate in agreements asking them to assume most
of the burden of emissions reduction.

Disparities in emissions per capita are due to factors that follow different paths
in different countries. A good knowledge both of the evolution of inequality and
of the drivers of the differences in emissions per capita and their trajectories
over time is essential to inform the debates on policy design and on the criteria
to distribute abatement efforts. The conclusions both for analysing the feasibility
of agreements of a given situation and for informing policy design would be
quite different depending on the different contributions of the relevant factors to
emissions inequality.

2

The increase in papers in recent years dealing with the international distribution
of CO2 emissions is noticeable. These analyses have followed two
complementary paths. First, several works employ the methodologies
developed in the literature on the measurement of income inequality (some of
the reference works include Atkinson, 1970; Sen, 1973; and Cowell, 2011).
These focus on aspects such as the properties of the measurements and their
factorial decomposition. The application and adaptation of this literature to the
analysis of environmental indicators extend the analysis made in the field of
income distribution. Some references in this line include Heil and Wodon (1997,
2000), Alcántara and Duro (2004), Hedenus and Azar (2005), Duro and Padilla
(2006), Padilla and Serrano (2006), Cantore and Padilla (2010), Cantore (2011)
and Duro (2012). Second, there are a series of works also analysing the
international distribution of CO2 (and other environmental indicators), but by
means of the methods developed in the literature on economic growth and
convergence (Barro and Sala i Martín, 1991; Quah, 1995). Some examples of
these works are Strazicich and List (2003), Nguyen Van (2005), Aldy (2006),
Ezcurra (2007), Romero-Ávila (2008), Criado and Grether (2010), Jobert et al.
(2010) and Barassi et al. (2011). Both lines of research analyse similar issues
with different tools and coincide in the relevance of measuring emissions
disparities as a tool for helping policy design.

Such proliferation of distributive analyses applied to the international distribution
of CO2 per capita might be seen as the result of the awareness of ecological
limits as well as of the need to inform discussions on the different
responsibilities and on the mitigation efforts to be assumed by different
countries. Moreover, this research complements the abundant literature focused
on the study of the driving forces of CO2 emissions (Grossman, 1993; Stern et
al., 1996; Suri and Chapman, 1998; Torras and Boyce, 1998; York et al., 2003;
Sharma, 2011). Among them, affluence, population, technology (the factors of
the so-called IPAT identity proposed by Ehrlich and Holdren, 1971),1 economic

1

York et al. (2003) turned that accounting equation into a stochastic regression model, allowing
them to make a test hypothesis and also to introduce further determinants of the environmental
impact.

3

structure, demographic characteristics and climate characteristics are usually
considered among the main drivers of environmental impacts. These analyses
usually consist of econometric models whose emphasis is on the regression
coefficients and their significance. However, as far as we know, the contribution
of these factors to emissions inequality measures has not yet been explicitly
and precisely approached. In this sense, a relevant methodological basis to
approach such analysis is the regression-based inequality decomposition
approach (RBID hereafter), which allows the development of such analysis
without being constrained to an automatic accounting relationship between
explanatory factors and emissions. Actually, previous analyses decomposing
emissions per capita inequality have taken as reference multiplicative identities
and group decompositions (Duro and Padilla, 2006). In contrast, the RBID
technique allows one to widen the list of explanatory factors unrestrictedly.

The method proposed consists of, first, running an auxiliary econometric
estimation to derive an additive decomposition of CO2 emissions per capita.
The model we will employ as reference for the identification of determinant
factors may be seen as an extended version of the econometric models usually
employed to test the environmental Kuznets curve hypothesis or the STIRPAT
models. In short, it includes as explanatory factors affluence, demographic
factors, sectoral composition and a climate variable. Second, the model applies
the methods of additive decomposition of inequality (Shorrocks, 1982, 1983) to
determine factoral contributions. Therefore, from a methodological point of view,
the proposed technique merges two hot research topics: the analysis of
inequalities in the contribution to climate change and the econometric
estimation of impact driving forces. The contribution of such factors to global
emissions inequality depends on two basic parameters: the average direct
relationship between the examined factor and countries’ CO2 per capita (i.e.
coefficient-effect), and the relative magnitude of the international variation of the
factor (i.e. dispersion effect). The methodology is applied to the analysis of
international inequality in CO2 emissions per capita for the period 1993–2007.

4

Our analysis will contribute, first, to informing how the evolution of disparities
leads or does not lead to a situation in which it is more likely that countries
share interests and perceptions of how to distribute abatement efforts; second,
it will contribute to the analysis of the determinants of emissions and how they
change over time; and third, it will show the factors behind the trajectory of
inequality. These factors should be adequately taken into account for a proper
design of policies that facilitate wider and fairer agreements.

The remainder of this paper is organised as follows. Section 2 describes the
RBID methodology. Section 3 measures the international inequality in CO2
emissions per capita and presents the results of the estimation of the driving
forces of emissions and their contribution to the international inequality of CO2
emissions per capita according to the proposed methodology. Section 4
concludes the paper.

2. The regression-based inequality decomposition: methodological
aspects
Inequality decomposition methods allow one to quantify which part of total
inequality is attributable to different components. The traditional additive
decomposition approach (Shorrocks, 1982) allows to one break down the
inequality of any variable according to its additive components. Additionally,
such an approach has been extended to decomposition into multiplicative
factors by taking advantage of logarithmic inequality measures such as the
Theil index. These methods require a consistent identity in order to perform the
decomposition.2 Therefore, the main restriction is that the contributions to
inequality considered are limited to the components of the mathematical
identity.

2

These analytical decomposition methods have been applied to ecological footprint in White
(2007), Teixido-Figueras and Duro (2012) and Duro and Teixidó-Figueras (2012). For the case
of CO2 emissions, Duro and Padilla (2006) made a multiplicative decomposition of the
contribution of Kaya (1989) factors.

5

In contrast, the RBID approach allows one not only to account for the
contribution to inequality of different components, but also to undertake causal
analysis since the contributions to inequality are attributed to the explanatory
variables of an econometric model. Therefore, the model can incorporate any
significant explanatory variable contributing to CO2 emissions inequality.3 The
first step is to construct a linear regression function such as the ones typically
used to estimate the driving forces elasticities of a given environmental
pressure (E):
E   0  1 X 1   2 X 2  ...   K X K  

(1)

Where E is the vector of the environmental pressure in the different countries
considered and Xi (i=0,…,K) the vectors for the driving forces or determinants of
this pressure.
There is a vast empirical literature estimating such functions, such as is the
case of the literature testing the environmental Kuznets Curve hypothesis
(Grossman, 1993; Stern et al., 1996; Suri and Chapman, 1998; Torras and
Boyce, 1998; Sharma 2011) or the stochastic regressions of relationships
derived from the IPAT identity (STIRPAT) (York et al., 2003). Those empirical
models try to disentangle the main determinants of environmental pressures or
impacts analysing the significance of independent variables such as income,
sectoral composition, population and demographic characteristics, among
others, in some cases controlling also for different regional climate
characteristics. Nonetheless, there is still no attempt to answer which is the
contribution of each of these relevant variables to the international inequality in
environmental indicators.
Expression (1) presents environmental pressure, in our case CO2 emissions per
capita, as the sum of K explanatory variables plus the constant and error terms.
So, we can rearrange it and obtain:

3

Most RBID applications analyse income inequality from a micro-approach, so there is an
income-generating function, and income inequality is decomposed in terms of the typical
explanatory variables of those models: race, education level, gender, age, etc. (e.g. Cowell and
Fiorio, 2009; Fields, 2003; Gunatilaka and Chotikapanich, 2009; Morduch and Sicular, 2002;
Wan, 2004).

6

E

K 2


k 0

k

Xk

(2)

The RBID is based on considering the product of the estimated coefficient  k
and its variable X k as the “causal component” of CO2 emissions per capita,
where the coefficient plays the role of weighting the importance of component k.
The explanatory variables jointly with the constant and the residual form a
consistent identity as those required by traditional decomposition methods, so
that the natural decomposition rule can be performed by sources (see
Shorrocks, 1982; Fields, 2003).
Although there are other methods to decompose inequality using regressionbased techniques, we use the Fields (2003) method because of its simplicity
and analogy to natural source decomposition described.4 In this RBID approach
the functional form of the model is restricted to a semi-log linear function:5

ln E 

K 2


k 0

k

Xk

(3)

Once the semi-log model is estimated, the procedure continues by taking
variances on both sides of the equation. Note that the variance of the logarithm
of emissions per capita is a common inequality index (see Cowell, 2011):

 K 2

var(ln E )  var   k X k 
 k 0


(4)

By rearranging the right hand side of expression (4), we obtain the variance of
logarithms as a sum of the covariances between each causal component and
the dependent variable (the logarithm of CO2 emissions per capita):

4

There are several empirical applications to income inequality comparing results obtained
according to the different methods of RBID. Very often they conclude that there are no
significant differences (Cowell and Fiorio, 2009; Fields, 2003; Gunatilaka and Chotikapanich,
2009; Morduch and Sicular, 2002; Wan, 2004).
5
The semi-log model Ln( E )   0   1 F1   2 F2  ...   k Fk   i is equivalent to
k

E  exp(  0  1 F1   2 F2  ...   k Fk   i )  exp(  0 )   exp( k Fk )  exp( i ) . Then, the
k 1

contribution 0 is null since it is a constant to each observation.

7

K 2

 cov(

var(ln E ) 

k 0

k

X k , ln E )

(5)

This result is highly convenient since in the inequality literature those
covariances are the natural decomposition of the variance, which indeed is a
consistent decomposition rule. Hence, in order to obtain the relative contribution
of each causal component, we define:

s k (ln E ) 

cov k X k , ln E 
var(ln E )

(6)

being sk the percentage contribution of factor k to the level of inequality
observed.6
Since the coefficients of the regression play a weighting role, it may be
interesting to know whether the different trajectories of sk are caused by
changes in the dispersion of factor k, or by changes in its importance in the
function measured by :

s kt  s kt 1 

cov(Z tk , ln Et ) cov(Z tk1 , ln Et 1 )


var(ln Et )
var(ln Et 1 )

ˆ
ˆ
 cov(Z tk1 , ln Et ) cov(Z tk1 , ln Et 1 )   cov(Z tk , ln Et ) cov(Z tk1 , ln Et ) 





var(ln Et 1 )   var(ln Et )
var(ln Et ) 
 var(ln Et )

(7)

ˆ
where Z tk   kt X kt and Z tk1   kt 1 X kt . The first term of the right-hand side is the

dispersion effect since the coefficients are not allowed to vary (and so only the
dispersion changes between t - 1 and t). The second term is the coefficient
effect since the dispersion of vector Xk is not allowed to vary (so only the
coefficient changes between both periods).

6

Independently of the index chosen by the researcher to assess inequality, the natural
decomposition of the variance is the unique unambiguous rule given that it is the only
decomposition rule that allocates indirect effects among components in a non-arbitrary way. In
other words, the contribution of factor k is independent of the inequality index used (see Cowell,
2000; Shorrocks, 1982, 1983).

8

Additionally, we may be interested in knowing the contribution of factor k to the
change in inequality level between two periods. That inequality change
contribution is expressed as:

k 

s kt I (.) t  s kt 1 I (.) t 1
I (.) t  I (.) t 1

(8)

where I(.) is the inequality measure for period t. Notice that expression (8) is not
restricted to the use of any particular inequality index. Our choice for the
empirical analysis will be neutral indexes such as GE(2) or CV (Duro, 2012).

3. International inequality in CO2 emissions per capita and explanatory
factors
As stated above, we use logarithmic variance as a reference index. This choice
is associated to the RBID methodology employed, which, based on the work of
Fields (2003), uses this measure for consistency. In any case, as we will later
show, the factoral decomposition can be applied to any consistent inequality
index. Logarithmic variance is a well-known measure that fulfils the scaleindependent property (that is, is a relative measure) but does not fulfil the
progressivity principle for high observations, which does not have a significant
impact in our case (Cowell, 2011).
Our analysis covers the period from 1993 to 2007 by eight biannual crosssection samples. The data used for each year cover at least 92% of world CO2
emissions, 95% of world population and 96% of world GDP. Although the
samples may be different among years, the results are virtually equivalent with
balanced data. This means the results with balanced data (all years with the
same countries) do not show any significant differences with the results
obtained with the non-balanced sample (where we take the maximum number
of countries in each sample). This is because the countries that are not included
in a specific sample represent a very little proportion of the world CO2 emissions
(and of the world population and GDP). The data used comes from the World
Bank (2013).

9

Figure 1 shows the international dispersion of CO2 emissions per capita for the
period 1993–2007. Initially, it can be noted that in this period the total emissions
increased 40% and the per capita emissions, 16.7%. Coinciding with this
increase was a significant reduction in dispersion, with three different
subperiods. In the first, from 1993–1997, there was an important reduction of
inequality; in the second, 1997–2003, there was a stabilisation of inequality; and
finally, in the third, 2003–2007, there was a new important reduction. Therefore,
we may say that the international responsibility in CO2 emissions, at least in per
capita terms, is becoming more diffused.
Figure 1. International inequality in CO2 emissions per capita measured by
log-variance

3,4
3,2
3
2,8
2,6
2,4
2,2
2
1993

1995

1997

1999

2001

2003

2005

2007

Source: Produced by the authors based on World Bank (2013).

However, underlying such a trend different stories may be occurring. Are the
major polluters reducing their emissions per capita, or, in contrast, are the minor
polluters increasing their emissions? Or may it be both things? Observing the
quantiles distribution in Table 1, the latter appears as more plausible since the
first percentile per capita emissions increased 64%, while percentile 0.9
increased only 5%. Therefore, from the environmental point of view this
reduction in inequality cannot be identified as good news as it was based on a
greater increase in lower emitters and not on a reduction in major emitters.

10

Obviously, in the quantiles analysis there is the anonymity axiom, i.e. we may
talk of different countries for the same quantiles in different years.
Table 1. Distribution of countries (by quantiles) and changes in carbon
emissions per capita, 1993–2007

1993

1999

% Q% Q% Qchange
change
change
2007 1993–1999 1999–2007 1993–2007

0.10

0.10

0.15

0.17

53%

8%

64%

0.20

0.29

0.38

0.43

31%

11%

46%

0.30

0.66

0.81

0.97

23%

20%

47%

0.40

1.14

1.22

1.58

6%

30%

38%

0.50

2.04

2.23

2.77

9%

24%

36%

0.60

3.58

3.92

4.57

10%

16%

28%

0.70

5.85

5.89

6.18

1%

5%

6%

0.80

7.48

7.91

8.06

6%

2%

8%

0.90

11.24

10.60

11.75

-6%

11%

5%

Quantiles
q

Source: Produced by the authors based on World Bank (2013).
Note: quantile refers to countries’ percentage. So Q0.10 in 2007 means 10% of
world countries had CO2 emissions per capita below 0.17 t.

Decomposing the inequality in CO2 emissions per capita according to its
determinants will enlighten the analysis of the causes of this inequality,
complementing previous accounting approaches. The dependent variable used
in the model is the CO2 emissions per capita of the different countries in log
scale, while the independent variables are several of those that have been
typically used in the estimation of the determinants of environmental impacts or
pressures (such as in the above cited STIRPAT models and the extended
models used to test the environmental Kuznets curve). The data used come
from the World Bank (2013). Table 2 details the descriptive statistics of the
variables used in the model.

11

Table 2. Descriptive statistics of explanatory factors

Variable (1993)

Obs

Mean

Std. Dev.

CV

Min

CO2 emissions (ktonnes per capita)

189

4,739.77

7,031.45

ln (CO2 emissions per capita)

189

7.35

1.83

1.48

per capita GDP (constant 2000 US$)

187

6,423.19

10,135.28

1.58

Agriculture GDP share (%)

165

19.43

16.23

Max
1.69

0.83

62,517.04

0.53

11.04

79.58

68,695.23

-

65.12

Industrial GDP share (%)

166

29.02

10.98

0.38

8.70

64.00

Urban population share (%)

210

52.82

24.67

0.47

6.84

100.00

Non dependent population share (aged 15 to 65)

190

59.08

6.71

0.11

45.53

72.13

Average daily min temperature

198

13.67

8.66

0.63

22.60

25.30

Variable (1999)

Obs

Mean

Std. Dev.

CV

Min

195

4,622.44

6,401.30

ln (CO2 emissions per capita)

195

7.46

1.66

per capita GDP (constant 2000 US$)

196

7,511.32

11,999.97

1.60

Agriculture GDP share (%)

171

17.41

15.61

0.90

Industrial GDP share (%)

172

28.78

11.96

0.42

7.20

79.99

Urban population share(%)

210

54.58

24.55

0.45

8.08

100.00

Non dependent population share (aged 15 to 65)

190

60.41

6.52

0.11

48.14

72.74

Average daily min temperature

198

13.67

8.66

0.63

22.60

25.30

Variable (2007)

Obs

Mean

Std. Dev.

1.38

Max

CO2 emissions (ktonnes per capita)

15.23

55,114.07

2.72

CV

10.92

95.50

74,111.49

-

Min

CO2 emissions (ktonnes per capita)

198

5,096.36

6,849.85

1.34

ln (CO2 emissions per capita)

198

7.58

1.66

per capita GDP (constant 2000 US$)

194

9,191.87

14,631.59

Agriculture GDP share (%)

169

12.53

12.28

171

31.11

14.10

0.45

Max

0.98

Industrial GDP share (%)

76.19

22.61

57,660.25

3.12

10.96

96.25

1.59

98,397.09

-

54.99

5.86

94.58

Urban population share(%)

210

56.99

24.16

0.42

10.10

100.00

Non dependent population share (aged 15 to 65)

190

62.94

6.87

0.11

48.81

82.22

Average daily min temperature

198

13.67

8.66

0.63

22.60

25.30

-

Source: Produced by the authors based on World Bank (2013).
Note: further descriptive data is available upon request. The most recent
available climate standard normal has been used as climatic reference of the
country.

The results obtained when applying the RBID methodology are twofold. First,
we obtain the regression results of the estimation of the determinants of CO2
emissions per capita. Second, using the regression results we estimate the
contributions of the explanatory variables to the international inequality in CO2
emissions per capita by using expression (6) of the previous section. The OLS
results (i.e. auxiliary regressions) are presented in Table 3. The model
estimated explains close to 85% of the cross-national log-variance in CO2
emissions per capita (except for 1993, in which it explains around 80%),

12

indicating that used variables provide a good fit. The significance and the sign
of the coefficients obtained are coherent with the empirical literature dealing
with the determinants of CO2 emissions. Besides, this high significance points
out that multicollinearity may not be a very important problem. Indeed, we
calculated quadratic partial correlations between the exogenous variables and
the

dependent

variable,

and

the

low

values

obtained

indicate

that

multicollinearlity is not a substantial problem in our estimation.7
Since it is a semi-log model, we must interpret the significant coefficients as
semi-elasticities, i.e. an increase (decrease) of one unit in the explanatory
variable yields a % increase (decrease) in the dependent variable. Hence,
focusing on 2007 coefficients, an increase in one dollar of GDP per capita yields
a 0.01% increase of CO2 emissions per capita, while an increase in the climate
normal minimum temperature would yield a -2.10% decrease in emissions per
capita, and so on.
We are estimating cross-national determinants and so are not making
assumptions about the individual behaviour of countries over time. A wrong
assumption often made in the environmental Kuznets curve literature, when
making panel data estimations, is to assume the same functional form and
parameters for each country in their relationship between income (and other
variables) and environmental pressure (or impact) over time. This assumption
has been clearly rejected when appropriately tested (see Perman and Stern,
1999 and 2003; List and Gallet, 1999; Dijkgraaf and Vollenberg, 2005;
Martínez-Zarzoso and Bengochea-Morancho, 2003 and 2004; Piaggio and
Padilla, 2012). Thus, our results just show cross-national relationships between
independent variables that change across countries and CO2 emissions per
capita in a given moment of time. These relationships may be caused by
different underlying reasons (levels of development, international specialisation,

7

As a robustness check, other models have been estimated with different regressors than those
in Table 3. Results obtained were virtually equivalent. As can be expected the higher
correlations belong to cubic and quadratic terms of GDP per capita; however, it must be taken
into account that the non-collinearity assumption is about linear relationships among regressors,
and despite its high correlation with linear GDP per capita, the cubic and quadratic terms are a
non-linear relationship. Hence, it does not violate the basic assumption (Gujarati and Porter,
2009).

13

different regulations, etc.) and be the result of different patterns followed by
different countries, as the literature seems to support.
Affluence variables indicate the existence of a non-monotonic relationship since
both quadratic and cubic GDP per capita variables are significant. This shows
an N-shape cross-national pattern (Friedl and Getzner, 2003; Sengupta, 1996;
Taskin and Zaim, 2000), though with a basic increasing segment and small
coefficients for quadratic and cubic terms, and so the environmental Kuznets
curve hypothesis is not supported by the data. Therefore, it may be stated that
in most cases greater affluence is accompanied by greater CO2 emissions per
capita.
Attending to sectoral composition determinants, a priori, a greater weight of
industrial sectors is expected to be associated with greater emissions, while
those economies with greater weight of services, and specifically of knowledgebased technology-intensive industries, are expected to have lower emissions
than those based on energy intensive industry (Dinda, 2004). In any case, it
should be taken into account that several services also make an intensive use
of energy (Suh, 2006; Alcántara and Padilla, 2009), and the idea that services
are immaterial sectors should be dismissed. Sectoral composition coefficients—
industrial GDP share and agricultural GDP share—show the expected values. A
positive coefficient of industrial share must be interpreted as the CO2 emissions
percentage augmented when the share of the sector increases 1% and the
base sector (services) decreases in this same percentage. In contrast, the
agricultural share coefficient shows a CO2 emissions per capita comparative
decrease.
Demographic variables have positive significant coefficients, indicating their
influence in spurring emissions per capita. The non-dependent population (aged
15–65), which captures the most consumerist and productive segment of the
population, exhibits an important role in driving emissions: a 1% increase
represents a cross-national increase of 8–10% in CO2 emissions per capita.
This suggests that population age structure may play a significant role in
explaining differences in emissions. This contrasts with previous results in the
literature, such as Dietz et al. (2003) for the case of the ecological footprint and

14

Cole and Neumayer (2004) for the case of total CO2 emissions, who did not find
significant coefficients for age distribution. Some differences with the study of
Cole and Neumayer (2004) that may explain the different result is the use of per
capita instead of total emissions, as well as the set of variables included our
model.
Urban population share exhibits a lower effect but is still significant and positive,
except for years 2001 and 2007 when the estimated coefficient is not
significant. Our estimates are consistent with the previous evidence in the
literature (see Parikh and Shukla, 1995, for the evidence on developing
countries; and Cole and Neumayer, 2004, for an international cross-section
sample including developing and developed countries). We find, however, that
the coefficient decreases over time. The positive impact of urbanisation on
emissions stems from the fact that an urban lifestyle and facilities lead people to
consume more energy and thus generate more CO2 emissions in urban areas
than in rural ones, especially in developing countries, which represent the
biggest part of the world. The migration of rural workers to urban areas in
search of better jobs tends to yield a sprawl growth of cities with large suburbs
and the need to commute every day by private vehicle. There is also more use
of fossil fuels instead of fuel wood and longer distances travelled for the
provision of food and other products (Jones, 1989; Parikh and Shukla, 1995).
Moreover, the use of public and private motor vehicles—cars, buses, and
motorcycles—is likely to be more extended in urban than in rural areas (Cole
and Neumayer, 2003). However, the impacts of urbanisation on emissions are
of a different type, and although most studies indicate that urbanisation tends to
increase energy consumption and emissions due to the abovementioned
reasons, there are other impacts that may go in the opposite way as
urbanisation may be accompanied by greater access to information, technical
innovation and efficient land and energy use, which may contribute to the
reduction of energy consumption and emissions in the long run (Jiang et al.,
2008; Jiang and Hardee, 2011). Actually, there are mixed results on the impact
of urbanisation on energy consumption and emissions (Jiang and Hardee,
2011). The decreasing coefficient of the variable may indicate that some of the

15

gains associated with urbanisation may now be more effectively compensating
the negative effects.
Both demographic variables, jointly with others like household size (Liu et al.,
2003) have only been studied to a limited extent in the literature but are
projected to have quite an important impact on the future evolution of emissions
(Jiang and Hardee, 2011).
Lastly, the climate control variable, proxied by the climate normal8 of minimum
temperature, indicates that an increase in normal temperatures of 1 °C would
decrease CO2 emissions per capita approximately 2%. This result shows the
fact that colder climates require greater amounts of energy for heating and
lower for cooling, the first impact being more important. Previous studies, such
as Neumayer (2004), also found that a cold climate is significantly associated
with greater CO2 emissions.

8

Climatologists define a climatic normal as the arithmetic average of a climate element (such as
temperature) over a prescribed 30-year interval in order to filter out many of the short-term
fluctuations and other anomalies that are not truly representational of the real climate. The last
climatic normal available is for the period 1971–2000

16

Table 3. Results from auxiliary OLS regressions on CO2 per capita and explanatory factors, 1993–2007

Variable

1993

1995

1997

1999

2001

2003

2005

2007

GDP per capita

.00022778**

.00021738***

.00015367**

.00015564***

.00015664***

.00016302***

.00016393***

.00016841***

Squared GDP per capita

-1.431e-08**

-1.322e-08***

-8.516e-09**

-8.346e-09**

-7.730e-09***

-7.707e-09***

-7.461e-09***

-7.006e-09***

Cubic GDP per capita

2.497e-13**

2.243e-13**

1.346e-13*

1.278e-13**

1.116e-13***

1.087e-13***

9.962e-14***

8.448e-14***

Affluence

Sectoral Composition
Agric. GDP share (%)

-.01342054*

-.02222722*** -.02916964*** -.02927415*** -.02897955*** -.02527107*** -.02486403*** -.03068516***

Indust. GDP share (%)

.0347483***

.03459411***

.02564298***

.018115***

.02332022***

.02433404***

.01680331***

.01653982***

Urban population share

.01733784***

.01225186***

.00822251**

.00671887*

0.00546392

.00571538*

.00623419*

0.00337771

Non-dependent pop.

.07808158***

.06992016***

.08566869***

.09595006***

.10306478***

.10293884***

.10317962***

.08912037***

-.02463132**

-.02426302*** -.02362798*** -.02221845***

-.01600683**

-.01564607**

-.01747623**

-.02104922***

Population Structure

Climate
Av. daily min. temp.
Constant

0.9419073

1.8809878**

1.5956777*

1.1516657

0.42887263

0.23336353

0.31073141

1.3898487*

Countries1

154

161

161

160

163

165

161

155

Squared R

0.79798729

0.84017732

0.85189569

0.84314694

0.85502122

0.8497542

0.84449959

0.84741744

Adjusted Squared R

0.78684176

0.8317656

0.84410073

0.83483684

0.84748985

0.84204928

0.83631536

0.83905675

log-likelihood

-183.69907

-165.53358

-154.45728

-157.2094

-155.9453

-159.37318

-154.28651

-138.8818

Note: * p<.1; ** p<.05; *** p<.01
Source: Produced by the authors based on World Bank (2013).

17

The regression results are used to calculate each factor’s weight, which jointly
with variable vector dispersion will yield the contributions to per capita CO2
emissions inequality observed. Table 4 presents the relative factor contributions
to inequality (expression 6).
Table 4. Relative factor contribution to inequality in CO2 per capita
Factors

1993

1995

1997

1999

2001

2003

2005

2007

Affluence

14.87

15.91

13.13

13.95

15.84

17.84

18.71

21.44

Sectoral Composition

19.72

28.71

30.85

27.54

27.70

24.66

21.05

24.36

agriculture GDP share

8.86

16.66

22.38

22.30

20.74

17.47

16.64

19.80

Industrial GDP share

10.86

12.05

8.47

5.24

6.96

7.19

4.42

4.55

Population Structure

39.67

34.05

35.68

37.51

38.53

39.14

40.64

33.88

Urban pop. Share

17.39

12.99

8.76

7.00

5.51

5.65

6.11

3.24

Non-dependent pop.

22.27

21.05

26.92

30.51

33.02

33.49

34.54

30.64

Climate

5.55

5.35

5.54

5.31

3.44

3.33

4.04

5.06

Residual

20.20

15.98

14.81

15.69

14.50

15.02

15.55

15.26

Total

100.00

100.00

100.00

100.00

100.00

100.00

100.00

100.00

Source: Produced by the authors based on World Bank (2013).
The affluence factor—which groups the GDP per capita variables—increased its
contribution significantly to emissions inequality, reaching its largest share in
2007 with 21%.9 Table 5 decomposes the change in this relative contribution of
each factor into the two basic elements explaining it according to equation (7).
Thus, these changes could be explained by a dispersion-effect—and so by
changes in the weight of the international variability of the factor—by changes in
the direct relationship between the factor and CO2 emissions per capita

9

This weight is clearly lower than the obtained by Duro and Padilla (2006) with a different
methodology. Their study decomposed per capita CO2 emissions inequality by a multiplicative
identity (Kaya factors) using the Theil index. As a result, they obtained an affluence net
contribution close to 60%, being the main contributor to CO2 inequality. However, this difference
can be explained by some methodological factors. First, the Kaya identity used in Duro and
Padilla (2006) assumes elasticity proportionality by construction, while in our regression model
the elasticities are allowed to vary among factors (see York et al., 2003). Second, the affluence
contribution is more precisely defined and isolated in our paper, given the more detailed list of
potential factors. Their study can therefore be gathering effects that in our case are separated,
such as the ones associated with demographic and structure factors.

18

according to auxiliary regressions (Table 3), or by both. Taking the whole
period, in which the relative contribution of this factor to the international
inequality in CO2 emissions per capita increases by 6.5%, the result is
explained by the relative increase in the dispersion component of this variable.
Sectoral factors represent 24.4% of total inequalities, ranging between 19.7% in
1993 to 30.9% in 1997. In any case, the two factors change in different
directions: there is a significant decrease in the importance of the industrial
share and a relevant increase in the role of the agricultural share. While in 1993
the agricultural GDP share made a lower contribution than the industrial one,
both contributions being of similar weight (8.9% and 10.9%, respectively), the
relative relevance of both factors reverse, and for the last year considered the
contribution to inequality of the agricultural share is more than four times the
one of the industrial share (19.8% and 4.6%, respectively). It is remarkable that
the increase in the importance of the agricultural share to explain CO2
emissions differences, which is concentrated in the period 1993–1997, is mainly
given by a coefficient effect (Table 5). Thus, its explanatory power in the
regression increased significantly while its coefficient became more negative.
That is, a greater share of agriculture—and so lower of services—is increasingly
associated with lower relative emissions. This different sectoral structure has
increased its relative contribution to emissions inequality, given the small
importance of the dispersion component. This may be seen as support for the
rejection of the notion of service economies as immaterial economies, as the
service sector includes activities which require great use of energy, both
directly—such as transport services—as well as indirectly—such as hotels and
restaurants (Suh, 2006; Nansai et al., 2007; Alcántara and Padilla, 2009;
Fourcroy et al., 2012). Moreover, some of these high-polluting service activities
have experienced an important development in the last decades.
According to our results, demographic characteristics play the most important
role in explaining inequality in CO2 emissions per capita, accounting for 34% of
it. In the first years of the sample the urbanisation variable contributed as much
to inequality as the non-dependent population variable, 17.4% and 22.3%,
respectively. Nonetheless, the relative contribution of urbanisation reduced its

19

level to a much lower value (3.2%). In contrast, non-dependent population
increased its relative contribution over the period. In 2007 it explains 30% of
international inequality in CO2 emissions per capita. The reduction in the
relative contribution of the urbanisation variable, which mainly occurred
between 1993 and 1997, is largely attributable to the coefficient effect. As
shown in Table 3, the positive relationship between urbanisation and greater
CO2 per capita decreases until being non-significantly different from zero in
2007. The decreasing importance in explaining global inequality—jointly with an
important

reduction

in

inequality

levels—means

that

some

of

the

abovementioned gains associated with urbanisation may now be more
effectively compensating the dominant negative effects.
In contrast, the relative contribution of the age structure of population has
increased, and the share of non-dependent population becomes the main
explanatory factor of inequality. In this case, both parameters have contributed
to this relative change, both the dispersion component (relative increase in the
international dispersion in this variable) and the coefficient effect, for the clear
increase in the relationship between non-dependent population share and
emissions per capita, which increases from 0.078 to 0.089. The increase in the
dispersion effect is due to the stability in the international dispersion of this
variable in front of the decrease in the dispersion of the logarithm of emissions
per capita. As regards the intensification of CO2 emissions associated to nondependent population, it seems clear that its greater mobility and energy
intensive consumption holds over time as a driver explaining differences
between countries.
The contribution of climate variable is quite stable over time, and explains only
5% of the differences in CO2 emissions per capita. A direct conclusion is that
international differences in CO2 emissions per capita are mainly caused by
anthropogenic CO2 drivers.
Last, the residual contribution, which plays a significant role, needs a previous
comment. In the typical applications of the STIRPAT models, T of Technology is
estimated in the residual term rather than separately (see York et al., 2003).
Therefore, we may interpret that the residual could show in part a technological

20

effect where the resources are more efficiently used, though it may also be
showing the impact of other omitted variables. Consequently, international
spillovers may be occurring in benefit of more equitable per capita emissions.
Such greater efficiency may be spurred on by either private gains in resource
saving or environmental policy regulations. The contribution of this residual to
total inequality in CO2 emissions per capita was quite stable, around 15% after
its decline in first years.

21

Table 5. Decomposition of the contribution of each factor to inequality
changes into coefficient and dispersion effects
Contrib.
Change
1993–1997
Affluence
Agriculture GDP share (%)
Industrial GDP share (%)
Non-dependent population share (aged 15 to 65)
Urban population share (%)
Average daily min. temperature
Residual
1997–2003
Affluence
Agriculture GDP share (%)
Industrial GDP share (%)
Non dependent population share (aged 15 to 65)
Urban population share (%)
Average daily min. temperature
Residual
2003–2007
Affluence
Agriculture GDP share (%)
Industrial GDP share (%)
Non-dependent population share (aged 15 to 65)
Urban population share (%)
Average daily min. temperature
Residual
1993–2007
Affluence
Agriculture GDP share (%)
Industrial GDP share (%)
Non-dependent population share (aged 15 to 65)
Urban population share (%)
Average daily min. temperature
Residual

Dispersion effect

Coefficient effect

-0.0173
0.1343
-0.0236
0.0462
-0.0858
-0.0001
-0.0538

0.0199
0.0143
0.0063
0.0225
0.0107
0.0023
-0.0538

-115%
11%
-26%
49%
-12%
-3887%
100%

-0.0371
0.1201
-0.0299
0.0237
-0.0965
-0.0023
0.0000

215%
89%
126%
51%
112%
3987%
0%

0.046
-0.049
-0.013
0.065
-0.031
-0.022
0.003

0.011
-0.022
-0.009
0.010
-0.006
-0.024
0.003

24%
45%
70%
15%
20%
111%
100%

0.035
-0.027
-0.004
0.056
-0.025
0.002
0.000

76%
55%
30%
85%
80%
-11%
0%

0.036
0.023
-0.026
-0.028
-0.024
0.017
-0.005

0.031
-0.012
-0.005
0.019
-0.002
0.004
-0.005

86%
-50%
19%
-66%
7%
25%
100%

0.005
0.035
-0.021
-0.047
-0.022
0.013
0.0000

14%
150%
81%
166%
93%
75%
0%

0.0648
0.1087
-0.0626
0.0832
-0.1406
-0.0048
-0.0487

0.1267
-0.0020
-0.0128
0.0455
-0.0075
0.0038
-0.0487

195%
-2%
21%
55%
5%
-79%
100%

-0.0619
0.1107
-0.0498
0.0377
-0.1331
-0.0086
0.0000

-95%
102%
79%
45%
95%
179%
0%

Source: Produced by the authors based on World Bank (2013).

Once we know the different relative contribution of factors to CO2 inequality, it is
interesting to analyse the contribution of those factors to inequality change over
the period analysed (expression 8 above). As we saw in Figure 1, the inequality
between countries in CO2 emissions per capita has decreased in the period
considered. Other studies also point in the same direction (Heil and Wodon,
2000; Duro and Padilla, 2006; Padilla and Serrano, 2006). In our period, the

22

reduction in inequality measured with log-variance was -18%. Table 6 presents
the relative contribution of each factor to such inequality change. Those factors
that are presented with a negative contribution change are the factors that have
contributed to making the distribution of emissions more unequal. In contrast,
those factors with a positive sign have contributed to a lower inequality in CO2
emissions per capita. The main driver of the reduction in emissions inequality
was urbanisation, which accounts for 82.1% of the whole reduction. The
industrial share and the residual—which may partly show a technology effect—
also made a significant contribution to the reduction of CO2 inequality, with
39.7% and 42.8%, respectively. In contrast, we could say that CO2 inequality
could have decreased even more if affluence, agricultural share or nondependent population had contributed in an opposite direction of what they did.

Table 6. Contribution of factors to the change in inequality measured by
log-variance (%)
Factors
Affluence

1993–1999

1999–2007

1993–2007

24.48

-233.27

-15.19

-62.14

132.67

-1.50

-131.79

104.81

-41.20

69.65

27.86

39.70

Population Structure

62.24

157.21

66.12

Urban pop. Share

126.16

131.09

82.14

Non-dependent pop.

-63.92

26.12

-16.02

Climate

7.96

13.61

7.75

Residual

67.46

29.78

42.82

100.00

100.00

100.00

-9

-10

-18

Sectoral Composition
Agriculture GDP share
Industrial GDP share

Total
Total change in logvariance

Note: The last row shows the total change in inequality for the different periods.
The rest of the rows show the percentage of this change that is attributable to
each factor.
Source: Produced by the authors based on World Bank (2013).

23

4. Conclusions
This paper contributes to the literature of the international distribution of
environmental pressures and especially to the literature focused on the
empirical measurement of the international inequality of CO2 emissions. The
analysis of the international distribution of CO2 emissions per capita is of great
relevance to inform the debates on climate change responsibilities, the design
of future agreements and the international distribution of abatement efforts.
We have used causal components instead of the usual analytical (identity)
components examined in the literature of the measurement of environmental
indicators inequality. The estimation of a model of the determinants of CO2
emissions per capita has enabled us to decompose the international inequality
in these emissions in terms of affluence, productive structure, demographic
characteristics and a climate variable showing differences in average daily
minimum temperatures (variables that have often been used in STIRPAT
models, and in the models employed to test the environmental Kuznets curve
hypothesis, among others). We have used the RBID methodology developed by
Fields (2003) which, despite being widely applied in empirical studies of income
inequality, has not yet been applied to environmental issues as far as we know.
The empirical application of such a method opens the door to new possibilities
in the research of distributional issues of the environment–society relationship.
The empirical results contribute significantly to expanding knowledge of the
factors contributing to the international disparities of CO2 emissions per capita.
As may be expected, 95% of such disparities are accounted for by
anthropogenic driving forces (since climate control contributed only around 5%).
The country’s affluence factor was found as a variable contributing significantly
to inequality, which means that remaining differences in GDP per capita are still
avoiding greater reductions in emissions inequality. According to our results, its
relative contribution, despite having increased to 20%, is not the main driving
force explaining emissions per capita differences. As for demographic variables,
population age distribution (measured by non-dependent population share)
appears as the main contributor to the analysed inequality because of its

24

importance in spurring CO2 emissions per capita rather than by its dispersion
among countries. This factor contributed to increasing inequality of emissions
per capita during the period analysed. In contrast, of the factors considered,
urbanisation became the lowest contributor to international disparities in CO2
emissions. The reason must be found in its lower importance in explaining
emissions (coefficient effect) rather than in its dispersion. Finally, the role
played by the residual may (with caution) be seen partly as the consequence of
international technology spillovers, since it has contributed to narrow differences
between countries in terms of emissions per capita.
The unequal use of the global sink capacity of the Earth is closely related to the
difficulty of reaching consensus on how to share the burden of emissions
mitigation and so appears as one of the main barriers to achieving effective
international agreements on emissions control and mitigation. Moreover, the
design of agreements could not be done without appropriately taking into
account this unequal contribution to the problem and the reasons leading to it, if
wide participation is to be achieved. The present paper contributes information
on the main factors responsible for international inequalities in emissions per
capita and so indicates some of the roots of the difficulties of achieving global
mitigation agreements. Besides, it gives some clues to which factors could lead
to a greater convergence or divergence of emissions per capita among
countries over time. According to our results, some implications could be
highlighted. First, it seems of great relevance to analyse the different
consumption patterns associated with demographic factors and how they can
change over time. Analysing in depth the different energy consumption and CO2
emissions patterns associated with urbanisation and to the share of potentially
active population seems of great importance to understanding emissions
drivers, the differences in emissions across countries and how can they change
over time and so condition the possibility of achieving agreements. These
results also indicate the need to focus policies on controlling the emissions
associated

with

these

patterns.

Second,

the

objective

of

economic

convergence, which is a highly desirable objective by itself, would have a clear
impact on reducing emissions inequality and so facilitating agreements between
countries. Third, the change in emissions inequality associated with different

25

sectoral compositions may depend on whether future economies tend toward
convergence to more similar economic structures or whether the trend is to
increase international specialisation. In any case our results show that those
countries more specialised in services tend to increase their differences in
emissions with those specialised in agriculture, in contrast with the oftenpopular idea that the tertiary sector is a cleaner sector. Finally, though the
residuals of our estimation may be the result of different things, they may be
indicating that one of the ways in which more is to be gained is to decrease
emissions differences via more effective technological diffusion.

Acknowledgments
The authors acknowledge support from projects ECO2010-18158 and
ECO2012-34591 (Spanish Ministry of Economy and Competitiveness),
2009SGR-600, XREPP, and XREAP (DGR).

References
Alcántara, V., Duro, J. A. (2004). Inequality of energy intensities across OECD
countries: A note. Energy Policy, 32(11), 1257–1260.
Alcántara, V. and Padilla, E. (2009) "Input–output subsystems and pollution: An
application to the service sector and CO2 emissions in Spain", Ecological
Economics, 68 (3), 905–914.
Aldy, J. (2006). Per capita carbon dioxide emissions: Convergence or
divergence? Environmental and Resource Economics, 33(4), 533–555.
Barassi, M.R., Cole, M.A., Elliott, R.J.R., 2011. The stochastic convergence of
CO2 emissions: a long memory approach. Environmental and Resource
Economics 49, 367–385.
Atkinson, A., 1970. On the measurement of inequality. Journal of Economic
Theory. 3, 244–263.
Cantore, N. (2011). “Distributional aspects of emissions in climate change
integrated assessment models”, Energy Policy, 39 (5), 2919–2924.

26

Cantore, N., Padilla, E. (2010) "Equality and CO2 emissions distribution in
climate change integrated assessment modelling", Energy, 35 (1), 298–313.
Cole, Neumayer, E. (2004) “Examining the Impact of Demographic Factors On
Air Pollution”, Population and Environment, 26 (1), 5–21.
Cowell, F. (2000). Chapter 2 measurement of inequality. Handbook of income
distribution. edited by A. B. Atkinson and F. Bourguignon (pp. 87–166) Elsevier
doi:10.1016/S1574-0056(00)80005-6
Cowell, F. (2011). Measuring inequality Oxford University Press.
Criado, C. O., & Grether, J. (2010). Convergence in per capita CO2 emissions:
A robust distributional approach CEPE Center for Energy Policy and
Economics, ETH Zürich.
Dietz, T., Rosa, E.A. and York, R. (2007) “Driving the human ecological
footprint”, Frontiers in Ecology and the Environment, 5 (1), 13–18.
Dinda, S. (2004). Environmental kuznets curve hypothesis: A survey. Ecological
Economics, 49(4), 431–455.
Duro, J. A. (2012). On the automatic application of
inequality indexes in the analysis of the international distribution of
environmental indicators. Ecological Economics, 76(0), 1–7.
Duro, J. A., Padilla, E. (2006). International inequalities in per capita CO2
emissions: A decomposition methodology by kaya factors. Energy Economics,
28(2), 170–187.
Duro, J. A., Padilla, E. (2008). Analysis of the international distribution of per
capita CO2 emissions using the polarization concept. Energy Policy, 36(1),
456–466.
Ezcurra, R. (2007). Is there cross-country convergence in carbon dioxide
emissions? Energy Policy, 35(2), 1363–1372.
Fields, G. S. (2003). Accounting for income inequality and its change: A new
method, with application to the distribution of earnings in the United States.

27

Research in Labor Economics, 1–38.
Fourcroy, C., Gallouj, F., and Decellas, F. (2012) “Energy consumption in
service industries: Challenging the myth of non-materiality”, Ecological
Economics, 81,155–164.
Friedl, B., Getzner, M. (2003). Determinants of CO2 emissions in a small open
economy. Ecological Economics, 45(1), 133-148.
Grossman, G. (1993). Pollution and growth: What do we know? C.E.P.R.
Discussion Papers.
Gujarati, D. N., Porter, D. C. (2009). Basic econometrics (5th ed.). Boston:
McGraw-Hill.
Gunatilaka, R., and Chotikapanich, D. (2009). Accounting for Sri Lanka's
expenditure

inequality

1980-2002:

Regression-Dased

Decomposition

Approaches. Review of Income and Wealth, 55(4), 882–906.
Hedenus, F., Azar, C. (2005). Estimates of trends in global income and
resource inequalities. Ecological Economics, 55(3), 351–364.
Heil, M. T., Wodon, Q. T. (1997). Inequality in CO2 emissions between poor and
rich countries. The Journal of Environment and Development, 6(4), 426–452.
Heil, M. T., Wodon, Q. T. (2000). Future inequality in CO2 emissions and the
impact of abatement proposals. Environmental and Resource Economics,
17(2), 163–181.
Jiang, L. and Hardee, K. (2011) “How do recent population trends matter to
climate change”, Population Research and Policy Review, 30 (2), 287–312.
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
Jones, D.W. (1989) “Urbanization and energy use in economic development”,
The Energy Journal, 10, 29–44.
List, J., Gallet, C. (1999) “The Environmental Kuznets Curve: does one fits all?”,
Ecological Economics, 31, pp. 409–423.

28

Liu, J., Daily, G. C., Ehrlich, P., Luck, G. W. (2003) “Effects of household
dynamics on resource consumption and biodiversity”, Nature, 421, 530–533.
Martínez-Zarzoso, I., Bengochea-Morancho, A. (2004) “Pooled mean group
estimation of an environmental Kuznets curve for CO2”, Economics Letters,
82, 121–126.
Martinez-Zarzoso, I., Bengochea-Morancho, A. (2003) “Testing for an
Environmental Kuznets Curve in latin-american countries”, Revista de
Análisis Económico, 18, 3–26.
Morduch, J., and Sicular, T. (2002). Rethinking Inequality Decomposition with
evidence from rural China. The Economic Journal, 112 (476), 93–106.
Nansai, K., Kagawa, S., Suh, S., Fujii, M., Inaba, R., and Hashimoto, S. (2009)
“Material and energy dependence of services and its implications for climate
change”, Environmental Science and Technology, 43, pp. 4241–4246.
Neumayer, E. (2004) “National carbon dioxide emissions: geography matters”,
Area, 36 (1), 33–40.
Nguyen Van, P. (2005). Distribution dynamics of CO2 emissions. Environmental
and Resource Economics, 32(4), 495–508.
Padilla, E., Serrano, A. (2006). Inequality in CO2 emissions across countries
and its relationship with income inequality: A distributive approach. Energy
Policy, 34(14), 1762–1772.
Parikh, J. and Shukla, V. (1995) “Urbanization, energy use and greenhouse
effects in economic development - results from a cross-national study of
developing countries”, Global Environmental Change, 5, 87–103.
Perman, R., Stern, D. (1999) “The Environmental Kuznets Curve: implications
of non-stationarity”, The Australian National University, Centre for Resource
and Environmental Studies, Working Paper in Ecological Economics Nº
9901.
Perman, R., Stern, D. (2003) “Evidence from panel unit root and cointegration
test that the Environmental Kuznets Curves does not exist”, The Australian
Journal of Agricultural and Resource Economics, 47, 325–347.

29

Piaggio, M., Padilla, E. (2012) “CO2 Emissions and Economic Activity:
heterogeneity across countries and non stationary series”, Energy Policy, 46,
370–381.
Romero-Ávila, D., 2008. Convergence in carbon dioxide emissions among
industrialized countries revisited. Energy Economics 30, 2265–2282.
Sen, A., 1973. On Economic Inequality. Oxford, Clarendon Press.
Sengupta, R., Unlisted. (1996). Economic development and CO2 emission:
Economy-environment relation and policy approach to choice of emission
standard for climate control. Boston University, Institute for Economic
Development, Boston University - Institute for Economic Development.
Sharma, S. S. (2011). Determinants of carbon dioxide emissions: Empirical
evidence from 69 countries. Applied Energy, 88(1), 376-382.
Shorrocks, A. F. (1982). Inequality decomposition by factor components.
Econometrica, 50(1), 193–211.
Shorrocks, A. F. (1983). The impact of income components on the distribution
of family incomes. The Quarterly Journal of Economics, 98(2), 311–326.
Stern, D. I., Common, M. S., Barbier, E. B. (1996). Economic growth and
environmental degradation: The environmental Kuznets curve and sustainable
development. World Development, 24(7), 1151–1160.
Strazicich, M. C., List, J. A. (2003). Are CO2 emission levels converging among
industrial countries? Environmental and Resource Economics, 24(3), 263–271.
Suh, S. (2006) “Are services better for climate change”. Environmental Science
and Technology, 40 (21), 6555–6560.
Suri, V., & Chapman, D. (1998). Economic growth, trade and energy:
Implications for the environmental kuznets curve. Ecological Economics, 25(2),
195–208.
Taskin, F., Zaim, O. (2000). Searching for a kuznets curve in environmental
efficiency using kernel estimation. Economics Letters, 68(2), 217–223.

30

Torras, M., Boyce, J. K. (1998). Income, inequality, and pollution: A
reassessment of the environmental kuznets curve. Ecological Economics,
25(2), 147–160.
Van, P. N. (2005). Distribution dynamics of CO2 emissionsTHEMA (THéorie
Economique, Modélisation et Applications), Université de Cergy-Pontoise.
Wan, G. (2004). Accounting for income inequality in rural China: A regressionbased approach. Journal of Comparative Economics, 32(2), 348–363.
World Bank (2013). World Development Indicators and Climate Change
Knowledge Portal. http://data.worldbank.org/, accessed on February, 2013.
York, R., Rosa, E. A., Dietz, T. (2003). STIRPAT, IPAT and ImPACT: Analytic
tools for unpacking the driving forces of environmental impacts. Ecological
Economics, 46(3), 351–365.

31