Coondoo and Dinda (2008).pdf

8/11/2019 Coondoo and Dinda (2008).pdf

1/11

ANALYSIS

Carbon dioxide emission and income: A temporal analysis of

cross-country distributional patterns

Dipankor Coondooa, Soumyananda Dindab,

aEconomic Research Unit, Indian Statistical Institute, Kolkata 700108, IndiabS.R. Fatepuria College, Beldanga, Murshidabad, West Bengal, India

A R T I C L E I N F O A B S T R A C T

Article history:

Received 6 November 2006

Received in revised form

13 June 2007

Accepted 2 July 2007

Available online 3 August 2007

This paper explores the relationship between the inter-country income inequality and CO2emission and temporal shifts in such a relationship. It also examines how the mean per

capita CO2emission and its distributional inequality are related to the corresponding mean

and the distributional inequality of income. The analysis is based on a cross-country panel

data set at the level of country-group. Here environmental damage is treated as a private

goodand the technique of Lorenz and specific concentration curve analysis have been used

as the basic analytical frameworkto argue that distributional inequality of incomeshould be

an explanatory variable in the Environmental Kuznets Curve relationship, along with the mean

income level. In the empirical exercise, Johansen's cointegration analysis technique is used

to explore existence of statistically significant cointegrating vector(s) relating meanemission and Specific Concentration Ratio of emission to mean income level and Lorenz Ratio

of income, using a set of country-group specific time series data set which covers four

country-groups (viz., Africa, America, Asia and Europe) and the World as a whole. The

empirical results confirm that the inter-country income inequality has significant effect on

the mean emission level and inter-country inequality of emission level for most of the

country-groups considered.

2007 Elsevier B.V. All rights reserved.

Keywords:

Cointegration

DistributionEmission

EKC

Inequality

LR and SCR

1. Introduction

Usually in the EKC literature environmental quality is specified

as a function of level of income, ignoring the role that income

distribution may play in the determination of environmentalquality. In some recent studies, however, distributional issues

have been brought explicitly in the discussion of income

environmental quality relationship (see, e.g., Torrasand Boyce

(1998) and alsoScruggs (1998) for a criticism of Torras and

Boyce's conclusion and alsoBoyce (1994)). Whereas Torrasand

Boyce followed the public good choiceapproachto argue that a

society's choice of the environmental degradation level would

be determined by the relative strength of different interest

groups of the society (as reflected by the distribution patterns

of income and social power across interest groups and

inequality therein), income distribution may be thought to

affect a society'senvironmental quality demand through otherroutes as well (Magnani, 2000). For example, a change in in-

come distribution may bring in a new pattern of consumer

demand, fulfillment of which may have important environ-

mental quality implications (Grossman and Krueger, 1995). A

more equitable income distribution may, by contributing to

social harmony, also help create public opinion in favour of

environmental quality improvement. Wider literacy, greater

E C O L O G I C A L E C O N O M I C S 6 5 ( 2 0 0 8 ) 3 7 5 3 8 5

Corresponding author.c/o Dipandor Coondoo, Economic Research Unit, Indian Statistical Institute, 203, B. T. Road, Kolkata 700108, India.Fax: +91 033 2577 8893.

E-mail addresses:[email protected](D. Coondoo),[email protected],[email protected](S. Dinda).

0921-8009/$ - see front matter 2007 Elsevier B.V. All rights reserved.doi:10.1016/j.ecolecon.2007.07.001

a v a i l a b l e a t w w w . s c i e n c e d i r e c t . c o m

w w w . e l s e v i e r . c o m / l o c a t e / e c o l e c o n
mailto:[email protected]:[email protected]:[email protected]://dx.doi.org/10.1016/j.ecolecon.2007.07.001http://dx.doi.org/10.1016/j.ecolecon.2007.07.001mailto:[email protected]:[email protected]:[email protected]


2/11

political liberty and civil rights may facilitate evolving a more

equitable distribution of income and power and hence bring

about improvement of environment.1

Ravallion et al. (2000) discussed the income distribution

environmental quality relationship in a somewhat different

context,viz.,the effectof poverty reduction on globalwarming

due to carbon dioxide emission. Briefly, they examined

whether reducing poverty by raising average income orlowering inequality would exacerbate global warming. The

econometric set up of that study was derived by aggregating

micro-level emission demand functions and thereby relating

country-specific (mean) emission level to per capita income,

population size, intra-country income inequality and time.

Their main empirical results are as follows: (i) given intra-

country income inequality, the income elasticity of per capita

emission is positive and declining in per capita income, (ii)

given per capita income, elasticity of emission with respect to

intra-country income inequality is negative and (iii) the

elasticity of emission with respect to population size is

positive and declining in intra-country income inequality.

Given these, a simulation exercise was done to examine the

effect on global emission of transferring income from the

richest five countriesto the poorest five countries (keeping the

intra-country income inequality of both sets of countries

unchanged). It was found that poverty reduction, whether

achieved through redistribution or growth, would increase

global carbon dioxide emission and hence cause global

warming. However, by lowering intra-country income in-

equality levelsacross board,a reductionof theglobal emission

level could be brought about in the long run. This was made

possible by an improvement of the trade off between reducing

inequality between countries and controlling emission with

growth,roughly when all countries reach the level of present-day

middle income countries.

Heerink et al. (2001) also derived the EKC by explicit

aggregation of the household emission demand function

over households and showed that the aggregate emission

demandfunctionwould be a function of both mean household

income and inter-household income inequality when the

household emission demand function was nonlinear in

income. In their empirical analysis based on a cross-country

cross-sectional data set, they compared the performance of

two alternative specifications of the EKC (viz., one having

intra-country Lorenz ratio of income as an explanatory

variable in addition to per capita mean income and the other

not having the first mentioned explanatory variable) for each

of eight different environmental damage variables. For six out

of these eight environmental damage variables, the effect of

income inequality on the level of environmental damage was

found negative and statistically significant. The income

elasticity of environmental damage was also found to be

significantly declining in income for a number of environ-

mental damage variables. On the whole, studies on EKC that

have explicitly used income inequality as an explanatory

factor by and large suggest that income inequality can be a

determinant of environmental quality. The specific mecha-

nism through which income inequality affects the level of

environmental damage is the differential marginal propensi-

ties to pollute (MPP) of rich and poor. Atthe globallevel, thus, if

MPP is higher for poorer countries, onemay expect a reduction

of inter-country income inequality to lead to a deterioration of

the global environmental quality2.

The present paper3 seeks to examine the effect of inter-country income inequality on the corresponding all-country

mean level of environmental damage, separately for country

groups of different continents. Here carbon dioxide emission

(henceforth denoted as CO2emission or simply emission) has

been taken as the environmental damage variable. The choice

of CO2 emission as the environmental damage variable is

primarily motivated by the fact that it is perhaps the most

important of the green house gases leading to such con-

sequences as global warming etc. Like Ravallion et al. (2000) and

Heerink et al. (2001), it is assumed here that demand for

environmental quality/damage is a derived demand, deter-

mined by the level and composition of goods and services

consumed. The basic theoretical setup of this paper is built on

aggregation of the micro level environmental damage de-

mand functions over the population of persons/households/

countries belonging to a given country/country-group. Two

relationships have been examined here, viz., whether for indi-

vidual country-groups the mean emission and inter-country

inequality of emission are significantly related to the corres-

ponding mean income and inter-country income inequality.

While a justification for this analysis can be readily given in

terms of the aggregation of the micro level emission demand

function,it is,in fact, a followup of an earlier study (viz.,Dinda

and Coondoo, 2006) based onthe same basic data set,in which

existence of a cointegrating relationship between income and

CO2emission was examined separately for different country-

groups. In that analysis, such a cointegrating relationship was

found for most of the country-groups. Now, existence of a

cointegrating incomeemission relationship naturally sug-

gests existence of a corresponding relationship between

inter-country inequality of income and CO2 emission for

individual country-groups.

Here we have examined if (1) mean CO2 emission, mean

income and inter-country income inequality and (2) inter-

country CO2 emission inequality, mean income and inter-

country income inequality are significantly interrelated, sep-

arately for country-groups to seeif a changein the inter-country

income distribution pattern would result in a change in mean

emission and the corresponding inter-country inequality of

1 In fact, inTorras and Boyce (1998), literacy, political liberty andcivil rights turned out to be better proxies for power inequalityand the effect of inequality on environmental quality worked outto be stronger in poorer countries.

2 Empirical evidences based on cross-country data suggest thateconomic growth in a poor country often leads to worsening ofenvironment. For a few environmental indicators, however, theevidences suggest that the direction of the relationship eventuallygets reversed and environment starts improving with incomegrowth. The existence of such non-linearity in the relationship ofincome with environmental indicators should have implicationsfor the relationship between income inequality and environmen-tal indicators. Here we focus on those implications.3 This is the third of a set of three papers reporting results of

empirical analyses based on the same data set. The other twopapers are Coondoo and Dinda (2002) and Dinda and Coondoo(2006).

376 E C O L O G I C A L E C O N O M I C S 6 5 ( 2 0 0 8 ) 3 7 5 3 8 5


3/11

CO2 emission. Theempirical analysis hasbeen doneat thelevel

of country-group, separately for the country-groups of Africa,

America (North and South combined), Asia, Europe and the

World taken as a whole. The analysis, based on time series

data on mean income, mean CO2emission and inter-country

inequality of income and CO2emission for different country-

groups, has been done by a combined use of two different

techniques, viz., the concentration curve analysis and coin-tegration analysis. Note in this context that concentration

curve analysis has been used here for explaining not only

between-country variation of mean emission levels but alsofor

explaining the corresponding between-country inequalities of

mean emission levels. Compared to earlier studies, this is a

major innovation. Briefly, three issues have been examined

here viz., whether the inter-country income inequality

significantly affects the mean level of CO2 emission for

individual country-groups and how the inter-country inequal-

ities of income and CO2 emission are related. Needless to

mention, here we make no attempt to hypothesize nor to

identify anyspecific economicmechanism,otherthan demand

aggregation as already mentioned, linking up mean emission,

mean income andinter-country inequalities of income andCO2emission.4

In what follows, we explain first the methodological

framework and then present the empirical results of our

analysis. The paper is organized as follows: Section 2 presents

the methodological framework. The empirical results are

presented in Section 3 and in Section 4, some concluding

observations have been drawn. Finally, the compositions of

the four country-groups considered in the analysis are given

in an Appendix.

2. EKCand the distributional issue

As such, the statement of the EKC hypothesis makes no

explicit reference to the possible relationship between level of

environmental degradation and income distribution. In the

discussion of income-environmental quality relationship,

income distribution generally enters through either or both

of two routes. First, treating environmental quality as a public

good, one may argue that the observed level of environmental

quality is determined by the relative powersof various interest

groups of the society, where the power distribution may be

closely related to income and other relevant socio-economic

inequalities. Alternatively, demand for environmental

damage5 may be regarded as a derived demand, being deter-

mined by the income level, the associated pattern of

consumption of goods and services and the technology used

to produce these goods and services.6 From this point of view,

the environmental damageincome relationship may be

viewed as the engel curve for environmental damage. As the

demand for environmental damage, ceteris paribus, changes

with income following a change in the level and composition

of goods and services consumed, the income elasticity of

demand for environmental damage is likely to vary system-

atically along the engel curve. Thus, environmental damage

will change its status from a luxury/necessary good to aninferior good with the rise of income. As Beckerman (1992)

puts it, if someone wants a better environment, (s)he has to be-

come rich. This should be true for an individual, a household, a

country or a nation and even for the human society at large. In

what follows, we take this latter route and examine the

distributional issues involved in the incomeenvironmental

damage relationship.

Consider a population of persons and letzdenote the level

of environmental damage demanded (measured on an appro-

priate continuous scale) by a person having income y ceteris

paribus, let the engel curve for environmental damage be

Ez=y fy;ya0;l; 1

where E(.) denotes the expected value and f(y) measures the

marginal income response of environmental damage

demanded. It is reasonable to expectf(y) to be monotonically

decreasing in income suchthatf(y)b0 at incomelevels greater

than a given threshold income level y when environmental

damage becomes an inferior good. Without loss of generality,

let this engel curve be a polynomial in y, viz.,

Ez=y b0 b1y b2y2 N N 2

The mean environmental damage demanded is

Ez

Z b0 b1y b2y

2 N N gy; hdy

b0 b1Ey b2Ey2 N N 3

whereg(y;) is the density function of income,being the (set

of) parameter(s) of the income distribution. Thus, mean

environmental damage demanded is determined not only by

mean income, but also by the higher order moments of the

income distribution, in general.7 One may proxy the effect of

higher order moments of the income distribution on mean

environmental damage by some measure of relative income

inequality and rewrite Eq. (3) as

Ez wEy; Iy; 3

I(y) and (.) being an income inequality measure and an

appropriate functional form, respectively. Next, the inequality

of environmental damage demanded that is due to income

inequalitycan be described in terms of the specific concentration

curve(SCC), as briefly explained below (Aitchison and Brown,

19578).4 In fact, the idea that CO2emission and income are related isfar wider and less specific than what may be implied by thecointegration of these two variables. We are indebted to ananonymous referee for pointing this out.5 In what follows, the words environmental damage,environmental

degradation, emission,pollution etc. havebeen used interchangeably.6 Ravallion et al. (2000) have also used the postulate that each

individual has an (implicit) derived demand function for environ-mental damage.

7 Note that in the present formulation if the demand function(2) is linear in y, the mean demand function (3) will relateE(z) toE(y) directly (i.e., mean demand will not be affected by incomeinequality).8 See also Kakwani (1980) for a comprehensive discussion on

these measures.

377E C O L O G I C A L E C O N O M I C S 6 5 ( 2 0 0 8 ) 3 7 5 3 8 5


4/11

Consider first

Gy

Z y0

gv; hdv; 4

the proportion of individuals having income up to y. Next,

consider

Gyy

Z y

0vgv; hdv

EY ; 5

the share in aggregate income of those having income up toy.

Finally, consider

Gzy

Z y0

Ez=vgv; hdv

EZ

Z y0

fvgv; hdv

EZ ; 6

the share in aggregate environmental damage demanded of

those having incomeup toy. The Lorenz Curve (LC) ofy relates

the share in aggregate income and the corresponding propor-

tion of persons having income up toyand is defined implicitlyas

/yGy; G

yy 0; 7

y(.) being the functional form. PlottingGyagainstG, one gets

the LC of y. Analogously, the SCC of z relates the share in

aggregate environmental damage demanded and the

corresponding proportion of persons having income up to y

and is defined implicitly as

/zGy; Gzy 0; 8

z(.) being the functional form. PlottingGz

againstG, one getstheSCC ofz. Note that since Gz is obtained by ranking indi-

viduals in ascending order of values ofy, theSCCreflects that

part of inequality ofz which is due to the inequality ofy.

As regards the shapes of these concentration curves, as

y(0,0)=0, y(1,1)=1,dGy

dG

y

EyN0 and

d 2GydG2

ddG

dGydG

1Eygy

N0

for ally, LC is a non-decreasing convex (to theGaxis) function

passing through the points (0,0) and (1,1). In case of equal

distribution,Gy(y)=G(y) for every y and thus LC is the 45 line,

known as theegalitarian lineor theline of equal distribution.

In case ofSCC,z(0,0)=0,z(1,1)=1,dGz

dG

fy

EyN 0for allyand

therefore it is also a non-decreasing function passing through

the points (0,0) and (1,1). However, for a non-inferior good,

f(y)N0 and hence d2

GzdG2 ddG dGz

dG

fVyEy

: 1gyN 0. So the SCC will be

convex (to theGaxis) and lie below the egalitarian line. For an

inferior good, on the other hand,d

2

GzdG2

b0 asf(y)b0 and so the

SCC will be concave (to the G axis) and lie above theegalitarian

line. Thus, when the income elasticity systematically varies

along the engel curve, the curvature ofSCC will change along

the curve as the good turns from non-inferior to inferior with

rising income.

The position of the SCC vis-a-vis the LC and the egalitarian

line, when the good is a luxury, necessary and inferior good,

respectively,can be ascertained as follows. Considerthecurveof

GzagainstGy. The slope of this curve isdGzdGy

fy

y :

Ey

EzN 0for ally.

Since this curve must pass through (0,0) and (1,1), a sufficient

condition forGzto be greater (less) than Gyis that the curve be

convex (concave) from above. Now, d2Gz

dG2yEy2

Ez :

fy

y3gy:gy 1,

whereydenotes the income elasticity ofz at income level y.Thus,

(i) when yN1 (i.e., z is a luxury good),d2GzdG2y

N0, which implies

GzbGy, which, in turn impliesSCC lies belowLC;

(ii) when 0byb1 (i.e.,z is a necessary good),d2Gz

dG2y

b0, and hence

GzNGy, which, in turn implies SCClies aboveLC. Further,

since d2Gz

dG2

fVy

Ey:

1gy

N0, SCC is convex (to the G axis). Hence

in this caseSCClies betweenLCand the egalitarian line;

finally,

(iii) whenyb0 (i.e.,zis an inferior good),d2GzdG2y

b0, and hence

GzNGy. But since d2Gz

dG2

fVy

Ez:

1gy

b0 (because f(y)b0, z

being an inferior good), SCC is concave (to the G axis)now. Hence in this case SCC lies above the egalitarianline;

(iv) sinceGz(y)bG(y) fory such that yN0 andGz(y)NG(y) for

ysuch that yb0 and yis monotonically decreasing in

y, it can be shown that there exists aysuch that Gz(y) =

G(y

) andy=0 at thisy

(i.e.,f(y

) = 0) as follows. SinceG,GzS =[0,1] and Gz= h(G), h:SS is a continuous func-

tion from the non-empty, compact, convex set SR

into itself, by Brouwer's fixed point theorem, there

exists G such that G = h(G) (i.e., Gz(y) = G(y) at some

y =y). Now, Gz(y) = G(y) at y =y implies dGzy

dGy

fy

Ey 1

and hencef(y)=0. Note that, as defined above, yis the

turning point income, when it exists.

Fig. 1below gives a diagrammatic presentation of the LC,

SCC and their interrelationship for the varying income

elasticity case. Panel A gives LC and SCC and Panel B gives

Fig. 1 Lorenz curve of income and specific concentration

curve for emission.



5/11

the cumulative income distribution.9 HereSCC intersects the

LC from below at point S. Thus, at incomelevels below YS, z i s a

luxury good. At pointT, theSCCintersects the egalitarian line

from below. So, in the interval (YS,YT), z is a necessary good, YTbeing the turning point income level at which the status of

environmental damage changes from a necessary to an

inferior good.

Now, if the EKC hypothesis holds, initially z will increasewithyand beyond the turning point income levelYT,zwill fall

with y. In terms of engel elasticity, this means that z is a

luxury good below YS, a necessary good for income in the

interval (YS,YT) and an inferior good at income level above YT.

Thus, if EKC hypothesis holds, the SCC for z will intersect the

egalitarian line from below at some turning point income level.

In the above discussion, we have essentially tried to link up

the intra-country income distribution and the corresponding

intra-country demand for environmental damage. The phe-

nomenon of income distribution should also be relevant in a

discussion of inter-country variation in the levels of environ-

mental damage. If income elasticity to emit is higher for poor

countries, a greater inter-country income inequality should

raise the aggregate environmental damage for any given all-

country mean income level. Looked from this angle, an

improvement of the world income distribution may result in

a deterioration of environmental quality (Ravallion et al.,

(2000)).10 As inter-country income inequality is the resultant of

such basic factors as the degree of opennessof economies and

extent of trade liberalization, diffusion of technological know-

ledge across economies etc., one may try to explain temporal

variation of mean environmental damage and inter-country

inequality in environmental damage levels for specific coun-

try-groups in terms of such trade-related macroeconomic

variables. No such analysis has however been attempted here.

Our objective here is rather humble, viz., to examine econo-

metrically whether the temporal variations ofmean per capita

CO2emission, mean per capita income and inter-country in-

equalities of per capita CO2emission and per capita income

are significantly related.

It may be mentioned that as the income distribution

changes from year to year, LC and SCC will both shift over

time. Correspondingly, there will be a temporal drift in the

observed income-emission relationship and so the turning

point income level will also vary from year to year. A related

issue of interest is the extent to which a change of the income

distribution will affect the corresponding distribution of

environmental quality demanded. This issue assumes impor-

tance in a cross-country set up for the following reason. If it is

supposed that the CO2emission level (or for that matter any

kind of environmental damage) is correlated with the income

level, at a specific time point the cross-country distribution of

emission level will be determined by the corresponding cross-

country income distribution. More importantly, a change in

the inter-country income inequality will affect both theaggregate emission level and the inter-country inequality of

emission for the country-group under consideration. To

examine this empirically, one may use the summary mea-

sures of relative inequality of income and emission based on

the LC and the SCC (viz., the Lorenz Ratio (LR) and Specific

Concentration Ratio (SCR)). These measures are defined below:

LR 1 2Z 1

0GyydGy and SCR 1 2

Z 10

GzydGy:

9

Note that whereas LR (0,1) with LR=0 and LR=1, signify-

ing complete equality and complete inequality of incomedistribution, respectively, SCR(1,1). SCRassumes thevalue-

1 in case of an inferior good, all of which is consumed by the

poorest person. It assumes thevalue 1 in case of a luxurygood,

all of which is consumed by the richest person of the society.

In case of both LRandSCR, however, a rise in the value of the

measure signifies a rise in the relevant inequality.

Let us next explain briefly howthe above discussion relates

to the empirical analysis reported here. Consider the time

seriesdata for a given country-group (z t,y t,LRt,SCRt,t =1,2,,T),

where z t and y t denote the observed mean emission and

income, respectively, andLRtandSCRtdenote theLR andSCR

of country-specific means of income and emission, respec-

tively, for yeart.11

Note thatz ,y andLRvary over time and inview of Eq. (3) the short run temporal movements of these

variables should be interdependent if this aggregation rela-

tionship is valid for the given data set. Next, considerFig. 1. A

change in the inter-country income distribution will mean a

shift ofG(y). Now, given the income emission relationship (1),

such a temporal shift of G(y) will cause a corresponding

temporal shift ofy and LRand hence ofSCR. Therefore,a priori

the temporal shifts ofy ,LR and SCR should be cointegrated

with a well defined underlying relationship bindingSCR to y

andLR. In this paper we seek to examine the existence and

nature of relationships binding each ofz t and SCRtto y t and

LRt applying the technique of cointegration analysis to

country-group specific time series data sets on mean income,mean CO2 emission and their inter-country distributional

inequalities.

3. Data description and results

The present exercise is based on a time series data set of

country group-specific mean income and CO2 emission and

corresponding inter-country LR of income and SCR of CO2

9 Note that the cumulative income distribution in Panel B givesthe income level corresponding to a point on the LC or SCC ofPanel A.10 Some economists maintain the optimistic view that individual

preferences of the rich people eventually lead to a virtuous circlerelationship between rising income and environmental degrada-tion. Empirical evidence from cross-country comparisons sug-gests that economic growth in poor countries entails worseningenvironmental outcomes. For a few environmental indicators, theevidence also suggests that the direction of the relationship iseventually reversed so that with enough growth environmentaloutcomes may ultimately begin to improve. The existence of suchnon-linearity in the cross-country relationship between environ-mental indicators and average incomes has implications for therelationship between income/emission inequality and environ-mental outcomes. Here we focus on those implications.

11 Note that here LRt and SCRt, being based on country-specificmeans, measure the between-group component of the relevantinequality.



6/11

emission, compiled from a cross-country panel data set on per

capita income and CO2 emission covering 88 countries andthe

time period 196090. Here the annual per capita GDP of

countries measured at 1985PPPdollars and the corresponding

annual per capita CO2emission measured in metric ton have

been taken as measures of income and emission, respectively.

Country-specific time series data on per capita GDP and CO2emission have been compiled from the Penn World Table andOak Ridge National Laboratory, USA, sources, respectively.

Using this data set, we have derived the LC and SCC and the

correspondingLRandSCRalong with mean income and mean

emission level for each of the years 19601990, separately for

the country-groups of Africa, America (North, Central and

South America pooled together), Asia, Europe and finally for

the World as a whole.12,13

Table 1 gives the summary statistics (viz., sample mean

and sample standard deviation) of mean income, mean CO2emission,LR of income and SCR of CO2emission for each of

the country-groups. These show that the average emission

and income of America and Europe are quite high and much

above the corresponding global averages, whereas for Asia

and Africa these values are far below the corresponding global

average values. This confirms the commonly held view that

developing countries contribute much less to the global CO2emission compared to their developed counterparts. Average

income inequality14 (LR) is lowest for Europe (0.19) and highest

for Asia (0.39). Understandably, the mean income inequality

for the World as a whole is much higher (0.57) than these

figures. The average SCR is lowest for Europe (0.13)and highest

for the World (0.54), the values for Africa, America and Asia

being in between (viz., 0.25, 0.47 and 0.34, respectively). The

standard deviation of the inequality measures, which indicate

the extent of temporal variation of these measures during the

sample period, are, as is to be expected, not large. However, as

the analysis of stationarity of the variables (done in terms of

unit root test reported later) has shown, the time series of the

inequality measures are non-stationary.

Since the empirical results presented later in this paper are

based on cointegration analysis applied to country-group

specific time series data sets on mean income, mean CO2emission, LR of income and SCR of CO2 emission, as a

preliminary analysis stationarity of the individual variables

have been examined by applying theunit root test (henceforth

referred to as the IPS test) ofIm et al. (2003)to the country-

group specific panel data sets for individual variables.15 The

computed t-values of the IPS unit root test, along with the

corresponding country-group specific computed ADF t-values,

are given inTable 2. As these results suggest, all the variables

have significant stochastic trend movement over time.

3.1. EKC and LR

Environmental quality, as already argued, is likely to get

affected by relevant socio-economic inequalities. Consider, for

example, Eq. (3). If E(z) measures the mean environmental

damage (say, CO2 emission) for a country-group, the mean

income and theLR of income of the country-group should be

the basic explanatory variables of the aggregated emission

income relationship (when the engel curve for environmental

damage is a nonlinear function of income).A priori, givenLR, a

rise in mean income should result in a corresponding rise in

mean emission, when the country-group consists of countries

with varying levelsof mean income. If most of thecountries of

a country-group are rich having negative income elasticity to

emit, the partial effect of a rise in mean income will be

negative. Similarly, if most of the countries of a country-group

are poor (having high positive income elasticity to emit), the

partial effect of a rise in mean income will be positive. The

partial effect ofLR on meanemissionwill be positive (negative)

if richer countries of the country-group have larger (smaller)

income elasticity to emit. Finally, if country-specific income

elasticities to emit are uncorrelated with the corresponding

mean incomes, one should expect the partial effect ofLR on

mean emission to be non-significant.

To examine if the effect of LR on the mean emission is

significant, we have done cointegration analysis using the

Johansenprocedure (Johansenand Juselius,1990), separatelyfor

each country-group, based on the country-group-specific yearly

time series data set of mean emission, mean income and LR.

Note in this context that the Johansen procedure assumes the

individual time series of the multivariate time series data set

being used for cointegration analysis (in the present case, the

time seriesof mean income, mean emission andLR of income of

each country group) to be integrated of order one and the data

12 The country-group-specific concentration analyses have beendone by taking into account the population size of the constituentcountries13 Note that the country-group specific data, obtained by

aggregating the corresponding country level, is likely to containconsiderable measurement error and hence use of such aggre-gated data in the analysis would affect the statistical quality ofthe estimates obtained in the present analysis.

Table 1Country-group-specific summary Statistics

Country-group

Variable Mean Standarddeviation

World Mean income 3482.94 619.19

Mean emission 0.96 0.09

LR of income 0.57 0.01

SCR of Emission 0.53 0.02

Africa Mean income 1367.78 259.87



SCR of emission 0.25 0.03

America Mean income 8163.65 1161.04




Asia Mean income 1457.97 426.59




Europe Mean income 7976.71 1747.64




14 It may be noted that here LR and SCR, being based on the dataon country-specific mean values of the variables concerned, givethe between-country group relative inequality measures.15 A brief description of the IPS test procedure has been given in

the Appendix ofDinda and Coondoo (2006).



7/11

generating process to be a finite-order vector autoregression

(VAR) model. As per the theory of cointegration analysis, there

can be at most (K1) cointegrating vectorsi.e., linear relation-

ships that tend to govern and bind the temporal movements of

theK variables together. In case, no significant cointegration

vector is found, the variables are said to be non-cointegrated.

The results of the cointegration analysis are summarised in

Table3. Note firstthat forall thecountry-groups mean income,

mean emission and LR are found to be cointegrated. For all

country-groups, except Europe, one statistically significant

cointegrating vector involving all the three variables is

estimated. For Europe, two significant cointegrating vectors

are obtained. However, both of these suggest a significant

relationshipbetween meanemission and LR only(one involves

all the three variables, but the coefficient of mean income is

non-significant andthe other doesnot involve meanincome at

all). For all theother country-groups, the partial effectof mean

income on mean emission is significant and positive. The

effect ofLR on mean emission, on theother hand, is significant

for all the country-groups except Africa. However, whereas for

America and Europe, the partial effect ofLRon mean emission

is negative (i.e., an equalizing redistribution of income, ceteris

paribus, would increase the mean emission), this effect is

positivefor Asia andthe World asa whole. The richercountries

of America andEurope thus seem to have significantly smaller

income elasticity to emit, so that when LR falls (i.e., there is a

mean-preserving transfer of income from these richer

countries to the poorer ones of the group), the mean emission

level tends to go up. By similar argument, the richer countries

of Asia may be having significantly largerincome elasticity to

emit andtherefore a shiftof theincome distribution away from

them would bring down the mean emission.

Finally, as the results for Asia suggest, a rise in mean

income with a concurrent decrease in the LR may very well

Table 3Estimated cointegrating vectors relating mean emission, mean income and LR of income, by country-group

Country-group

Number ofcointigrating

vectors estimated

Elements of normalized estimated cointegrating vectora

Mean emission Mean income LR of income Intercept

World 1 1 2.4104 9.09 5.26

(6.05) (3.56) (3.2)

Africa 1 1 2.7104 0.28 0.11

(7.71) (0.51) (0.52)

America 1 1 1.5104 17.46 7.58

(2.43) (2.39) (3.2)

Asia 1 1 2.6104 0.78 0.30

(28.06) (6.08) (6.26)

Europe 2 1 5.2105 14.35 4.66

(1.6) (4.6) (6.2)

1 0 15.86 5.19

(9.35) (14.72)

Significance at 5% and 1% level are denoted by and , respectively.a

Figures in parentheses are the t-ratios.

Table 2Results of panel unit root test: Computed values of ADF t statistic by country-group and IPS tstatistic

Computed value of Computed value for the variable

Mean Income Mean Emission LR of Income SCR of Emission

For the null hypothesis of non-stationarity in level a

ADF t-statistics for country-group Africa 1.269 0.473 2.242 1.944

America

2.219

0.119

0.976

0.817Asia 2.235 0.731 1.665 1.103

Europe 1.467 2.274 2.330 2.769

IPStstatisticb 0.745 1.143 0.757 0.452

Computed value of For the null hypothesis of non-stationarity in first difference c

ADF t-statistics for country-group Africa 2.928 2.132 3.157 2.515

America 3.501 2.223 1.801 2.079

Asia 3.002 4.487 4.641 2.488

Europe 3.656 3.209 1.864 2.576

IPStstatisticd 3.845 3.3 2.991 12.554

a ADF equation with time trend is used.b The 5% critical value is 2.79.c ADF equation without time trend is used.

d The 5% critical value is 2.16.



8/11

result in a decrease of mean emission, depending on the

magnitudes of the change of mean income and LR. This is in

line with the view that an equalizing income growth across

countries mayhelp contain the level of emissionglobally. This

may have policy significance so far as the management of

aggregate level of emission in Asia is concerned.

3.2. SCR and LR

As reported above, in the present analysis mean emission,

mean income and/orLR of income are found cointegrated for

all the country-groups. Since the cointegrating relationship

binds the temporal movements of mean income and mean

emission together, it is imperative that inter-country inequal-

ities of emission and income would also be related to each

other. Such a relationship is also warranted by the fact that, by

definition, the SCC (and SCR) of emission is the portion of

inequality of emissiondue to income inequality, ceteris paribus.

A preliminary graphical examination of the observed temporal

movements ofSCR andLRfor the period 196090 for different

country-groups confirmed this. As a formal analysis, we have

performed the cointegration analysis using the three variables

SCR,LRand mean income, separately for each country-group.

As regards the expected partial effect of mean income onSCR,

if income elasticity to emit is inversely (positively) related to

country mean income level, this partial effect will be negative

(positive). If income elasticity to emit is inversely related to

country mean income level, the expected partial effect ofLR

on SCR will be positive,but if this elasticity is positively related

to country mean income level, the expected effect will be

ambiguous. Finally, if income elasticity to emit is uncorrelated

with country mean income level, these partial effects may

turn out to be negligible.

The cointegration results are presented in Table 4. Note

first that at least one statistically significant cointegrating

vector is found for all the country-groups. Whereas just one

significant cointegrating vector is obtained for the country-

groups of America, Europe (and also for the World as a whole),

the number of such vectors found for Africa and Asia are two

and three, respectively. For both these country groups, one of

the estimated cointegrating vectors does not involve the SCR

(implying thereby that this vector defines a relationship

between mean income and LR only). Of the remaining two

cointegrating vectors for Asia, one involves all the threevariables while the other involves SCR and LR but not mean

income.

As regards the nature of dependence ofSCR as shown by

the estimated cointegrating vectors, for America, the partial

effect of mean income andLRonSCR are both significant, the

former beingnegative andthe latter positive.As arguedabove,

this suggeststhat the richer countriesof this group have larger

income elasticity to emit. For Europe the partial effect of mean

income is found negative and significant, but that of LR is

positive but non-significant. This is suggestive of an inverse

relationship between income elasticity and mean income

level for the countries of this group. For Africa and Asia, the

partial effects of both mean income and LR on SCR are both

found positive and significant a result which cannot be

readily explained. On the whole, whereas for every country

group, except Europe and the World as a whole,LRaffects the

corresponding SCR, the nature of the effect is qualitatively

different across country-groups.

3.3. Existence of a turning point on the EKC

As already mentioned, if the SCC intersects the egalitarian line

from below, such an intersection signifies that emission or

environmental degradation becomes an inferior good at

income levels above that corresponding to this intersection

point. In the EKC literature this income level is known as the

turning point of the EKC. In terms of the EKC, the emission

level start declining with income along the EKC once the

Table 4Estimated cointegrating vectors relating SCR of emission, mean income and LR of income by country-group

Country-group Number of cointegratingvectors estimated

Elements of normalized estimated cointegrating vectora

SCR of Emission Mean Income LR of Income Intercept

World 1 1 0.97105 1.84 0.54

(3.23) (1.41) (6.58)

Africa 2 1 5.0105 1.05 0.12

(17.86) (31.76) (10.63)0 1 18724.02 7307.69

(1.92) (2.43)

America 1 1 0.12 105 1.02 7.7104

(8.88) (84.2) (0.19)

Asia 3 1 7.5105 2.15 0.65

(6.88) (15.19) (10.59)

1 0 1.08 0.088

(2.41) (0.48)

0 1 14,228.68 7446.02

(2.34) (2.96)

Europe 1 1 0.25 105 0.32 0.092

(1.78) (1.34) (1.66)

Significance at 5% and 1% level are denoted by and , respectively.a

Figures in parentheses are thet-ratios.



9/11

turning point income level has been crossed. In other words,

evidenceof a turning point on an empirical EKC is indicative of

environmental improvement16 and a support for the EKC

hypothesis. In the present study, we have investigated the

existence of turning point for individual country-groups by

checking numerically if the empirical SCC intersected the

egalitarian line from below. This analysis hasbeen done for all

country-groups for each of the years 196090. A summary ofthe results is given below.

Of all the country-groups, evidences of turning point are

found only for Europe. The SCC for the country-group of

Europe is seen to cross the egalitarian line from the year 1966

onward.17 Interestingly, the turning point income level is seen

to increase monotonically over time. For Denmark, Luxem-

bourg, the Netherlands, Sweden, Switzerland and West

Germany the mean per capita income level is observed to be

generally higher than the turning point income level in most

years. This result, thus, seems to provide an empirical

evidence in support of the EKC hypothesis for CO2 emission

in case of the countries of Europe. Further, the fact that the

estimated turning point income level (measured at constant

prices) is rising monotonically over time perhaps suggeststhat

even the rich countries of Europe find it hard to bring down

their CO2emission levels. Finally, it should be mentioned that

these estimated turning point income levels are observed

within the sample income levels, which contradicts the

findings ofHoltz-Eakin and Selden (1995)18 for CO2emission.

3.4. Some regression results

While discussing about the data, we pointed out that the

sample standard deviations of LR and SCR are all rather small,

which implies that for individual country groups the temporal

variation of these inter-country inequality measures are low.

This lack of variation of LR and SCR in the given data set may

have implication for the robustness of the results of coin-

tegration analysis presented here.19 As a supporting exercise

and for purely illustrative purpose, we have, therefore,

estimated linear regression equations for mean emission on

mean income and square of mean income using country-

group specific panel data sets comprising country level data

on mean income and mean emission for countries belonging

to specific country groups. For each country-group both the

fixed effect and the random effect models are estimated.

These regression results are presented inTable 5.

Note first that, as the values of Hausman test statistic

suggest, for all the country-groups except America the null

hypothesis of the random effect being the underlying true

model cannotbe rejected.Next, in all the cases the coefficients

of both mean income and mean income squared are

significant implying thereby an empirical support for the

assumption of non-linearity of the incomeemission relation-

ship made in this paper for deriving the relationship of

country-group level mean emission with the corresponding

country-group level mean income and inter-country disparity

in the mean income levels. Finally, for all country-groups

except Africa, the estimated coefficient of mean income

squared is negative, which implies the marginal income

response of mean emission falls as the mean income level

rises.

As a supporting exercise and for purely illustrative purpose,

we have estimated a nonlinear regression equation for mean

18 Holtz-Eakin and Selden (1995) estimated the turning pointincome level for the World as a whole to be $34000 approximately,a value that fell beyond the sample range of income.19 Note, however, that the results of the unit root test have

shown that these inequality parameters have trend for all thecountry groups.

16 In the EKC literature, evidences of two, rather than one,turning points of empirical EKC have been reported (see, e.g.,Sengupta (1997), Grossman and Krueger (1995)). The secondturning point income level (which is greater than the first one)signifies the beginning of a new phase of rising environmentaldegradation required for improving further the already-reachedhigh income level.17 The turning point income level corresponding to the point of

intersection of the SCC and the egalitarian line has beencalculated by interpolation.

Table 5 Estimated regression equations of MeanEmission on Mean Income and Mean Income-squared bycountry-group based on country-group specific panel datasets

Country-group

Explanatoryvariable

Estimated coefficient a

Randomeffect model

Fixedeffect model

Africa Intercept 0.087

(1.31)

Mean income 2.1 104 2.021104

(6.4) (6.03)

Mean income-squared 1.8108 1.88108

(3.6) (3.73)

AdjustedR-square 0.8903 0.8939

Hausman test statistic 1.17

America Intercept 0.4724

(4.68)

Mean income 4.05 104 3.58104

(17.02) (13.67)


(8.58) (7.22)

AdjustedR-square 0.9586 0.9622Hausman test statistic 55.225

Asia Intercept 0.395

(3.39)

Mean income 4.27 104 4.3104

(20.27) (20.07)


(12.14) (12.12)

AdjustedR-square 0.902 0.9052


Europe Intercept 0.0113

(0.03)

Mean income 4.85 104 4.82104

(11.55) (11.41)


(9.49) (9.41)AdjustedR-square 0.8876 0.8913


Significance at 5% and 1% level are denoted by and, respectively.a Figures in parentheses are the t-ratios.



10/11

emission on mean income, square of mean income and income

inequality (LR) using the time series data of country-group-

specific mean emission, mean income andLR.Theseregressionresults are presented inTable 6. In all the cases, the coefficient

of mean income is significantly positive. The estimated

coefficient of mean income squared is significantly negative

for all country-groups except Africa. This implies that the

marginal income response of mean emission falls as the mean

income level rises. This is an empirical support for the

assumption of non-linearity of the incomeemission relation-

ship. Finally, the coefficient of income inequality is statistically

significant only for Africa, Europe and the worldas a whole. For

Africa and the world, mean emission rises with income

inequality while it declines in Europe.

4. Conclusion

In this paper we have examined how at the country-group

level the mean per capita CO2emission and its distributional

inequality is related to the corresponding mean income and

income inequality, based on a cross-country panel data set.

Treating environmental damage demanded as a private good

(and not as a public good as done in most studies) and using

the technique of Lorenz and specific concentration curve

analysis as the basic analytical framework, we have tried to

argue that a measure of distributional inequality of income,

along with the mean income, should be a used as

explanatory variable in the EKC relationship. The concen-tration curve methodology has thus been used not only for

explaining mean income levels but also for explaining

inequality in emission levels for different country-groups.

Compared to earlier studies, this is a major methodological

innovation.

In the empirical exercise, we have used Johansen's coin-

tegration analysis technique to explore existence of statisti-

cally significant cointegrating vector(s) relating mean

emission and SCR of emission to mean income and LR of

income. The empirical results broadly confirm that inter-

country incomeinequalityhas a significant effecton the mean

emission for all the country-groups considered, although the

effect seem to vary qualitatively across country-groups. Thus,

it is found that whereas for the country-groups of Europe and

America, an equalizing redistribution of income would raise

the mean emission level, the opposite is the case for the muchpoorer country-group of Asia and for Africa the effect of LR on

mean emission turned out to be non-significant. A significant

positive effect ofLRon theSCRfor emission is observed for all

the country groups, except Europe and the World as a whole.

Finally, evidences in favour of existence of turning point

income level on the empirical EKCbased on the SCC have been

found for the country-group of Europe alone for the period

1966 onwards.

One would find that the analytical framework of concen-

tration curve analysis that we have used here is quite novel

and convenient for the purpose of EKC analysis. One would

also find the empirical results and the conclusions based on

them sensible, interesting and useful from the policy point ofview. However, thepaperis notfree of limitations. As hasbeen

mentioned, as the analysis has been designed at the country-

group level, both the mean emission and inter-country

disparity of emission levels are likely to be significantly

affected by factors relating to international trade like the

size of international trade and the degree of openness of the

countries concerned. In fact, the so called Pollution Haven

Hypothesis asserts that many rich developed countries are

more and more outsourcing their requirements of emission-

intensive material goods from the poorer developing countries

and thereby shifting bulk of their emission to the latter

countries (Cole, 2004). When this happens, the inter-country

disparity measured by SCC orSCR of emission may not be asclosely related to the corresponding inter-country income

disparity measured by LC or LR of income. Needless to

mention, one should bring in an appropriate measure of

openness of the individual countries in to the analysis to

identify the pure partial effect of income distribution on the

emission level.

Acknowledgements

We are grateful to the referees for valuable comments and

suggestions on the earlier versions of this paper. Remaining

errors, if any, are our responsibilities.

Table 6Estimated regression equations of mean emission on mean income, mean income-squared and LR of income fordifferent country-groups

Variable Country-groups

Africa America Asia Europe World

Intercept 0.402 4.061 0.17 0.998 1.472

(2.78) (2.75) (3.30) (0.87) (4.21)

Mean income 0.0005 0.002 0.0005 0.0005 0.0007(2.08) (5.75) (4.43) (3.43) (4.71)

Mean income-squared 5.76108 1.1107 8.6108 3.1108 7.4108

(0.66) (5.57) (2.33) (3.40) (3.46)

0.748 1.529 0.082 5.8497 1.887

LR of income (4.64) (0.89) (0.42) (2.04) (2.47)

Adjusted-R2 0.95 0.58 0.96 0.87 0.90

Figures in parentheses are the t-ratios. Significance at 5% and 1% level are denoted by and , respectively.



11/11

Appendix A

Table A1Country-composition of country-groups

Country group Countries covered

Africa Algeria, Cameroon, Cape

Verde Island, CentralAfrican Republic, Comoros,

Congo, Egypt, Gabon,

Gambia, Ghana, Guinea,

Guinea Bissau, Kenya,

Madagascar, Mali,

Mauritania, Mauritius,

Morocco, Mozambique,

Nigeria, Senegal, South

Africa, Togo, Tunisia,

Uganda, Zimbabwe.

America Canada, USA, Costa Rica,

Dominican Republic, El

Salvador, Guatemala,

Honduras, Jamaica, Mexico,

Nicaragua, Panama,Trinidad & Tobago,

Argentina, Bolivia, Brazil,

Chile, Colombia, Ecuador,

Paraguay, Peru, Uruguay,

Venezuela.

Asia Japan, China, Hong Kong,

India, Indonesia, Iran,

Israel, Jordan, Korea

Republic, Philippines,

Singapore, Sri Lank, Syria,

Thailand.

Europe Austria, Czechoslovakia,

Finland, Greece, Turkey,

Yugoslavia, Belgium,

Cyprus, Denmark, France,West Germany, Iceland,

Ireland, Italy, Luxembourg,

Netherlands, Norway,

Portugal, Spain, Sweden,

Switzerland, U.K.

R E F E R E N C E S

Aitchison, J., Brown, J.A.C., 1957. Lognormal Distribution withSpecial Reference to Its Users in Economics. CambridgeUniversity Press, Cambridge.

Beckerman, W., 1992. Economic growth and the environment:whose growth? Whose environment? World Development 20(4), 481496.

Boyce, J.K., 1994. Inequality as a cause ofenvironmental degradation.Ecological Economics 11 (3), 169178.

Cole, M.A., 2004. Trade, the pollution haven hypothesis andenvironmental Kuznets curve:examiningthe linkages. EcologicalEconomics 48 (1), 7181.

Coondoo, D., Dinda, S., 2002. Causality between income andemission: a country group-specific econometric analysis.Ecological Economics 40 (3), 351367.

Dinda, S., Coondoo, D., 2006. Income and emission: a panel databased cointegration analysis. Ecological Economics 57 (2),167181.

Grossman, G.M., Krueger, A.B., 1995. Economic growth and theenvironment. Quarterly Journal of Economics 110 (2), 353377.

Heerink, N., Mulatu, A., Bulte, E., 2001. Income inequality andenvironment: aggregation bias in Environmental KuznetsCurve. Ecological Economics 38 (3), 359367.

Holtz-Eakin, D., Selden, T.M., 1995. Stoking the fire? CO2emissionand economic growth. Journal of Public Economics 57 (1),85101.

Im, K.S., Pesaran, M.H., Shin, Y., 2003. Testing for unit roots in

heterogeneous panels. Journal of Econometrics 115 (1), 53 74.Johansen, S., Juselius, K., 1990. Maximum likelihood estimation

and inference on cointegration, with applications to thedemand formoney. OxfordBulletin of Economics and Statistics52 (2), 169210.

Kakwani, N.C., 1980. Income Inequality and Poverty: Methods ofEstimation and Policy Application. Oxford University Press,Calcutta.

Magnani, E., 2000. The environmental KuznetsCurve, environmentalpolicy and income distribution. Ecological Economics 32 (3),431443.

Ravallion, Martin, Heil, M., Jalan, J., 2000. Carbon emissionsand income inequality. Oxford Economic Papers 52 (4),651669.

Scruggs, L.A., 1998. Political and economic inequality and the

environment. Ecological Economics 26 (3), 259275.Sengupta, R.P., 1997. CO2emissionincome relationship: policy

approach for climate control. Pacific Asia Journal of Energy. 7(2), 207229.

Torras, M., Boyce, J.K., 1998. Income, inequality, and pollution: areassessment of the environmental Kuznets curve. EcologicalEconomics 25 (2), 147160.

Table A1Country-composition of country-groups


Coondoo and Dinda (2008).pdf

Documents

Transcript of Coondoo and Dinda (2008).pdf