Professional Documents
Culture Documents
4, DECEMBER 2010
Abstract The object of this article is to examine the relation between social
inequalities and pollution. First of all we provide a survey demonstrating
that, from a theoretical point of view, a decrease in inequality has an uncertain
impact on the environment. Second, on the basis of these conceptual considerations, we propose an econometric analysis based on panel data (xedeects and dynamic panel data models) concerning developing and transition
countries for the 19882003 period. We examine specically the eect of
inequality on the extent of local pollution (sulphur dioxide emissions and
organic water pollution) by integrating the Gini index into the formulation of
the environmental Kuznets curve.
Keywords: pollution, inequality, environmental Kuznets curve, panel data
INTRODUCTION
The scale and intensication of environmental deteriorations caused by
current modes of development present a signicant challenge for the longterm viability of globalization. Neoclassical authors argue that the market
remains the most ecient institution for integrating ecological constraints
(Dales 1968), on the double condition that such externalities are internalized
and that technological progress steadily becomes more widespread. Heterodox economists contest this optimistic version of market failures, and call for
the adoption of a dierent paradigm of economic development (Ostrom
1990).
Analysis of the relations between human activities and the environment is
based on the concept of sustainable development (WCED 1987). The three
Review of Social Economy
ISSN 0034-6764 print/ISSN 1470-1162 online 2010 The Association for Social Economics
http://www.informaworld.com
DOI: 10.1080/00346760903480590
415
pollution the equality hypothesis, reecting the idea that greater social
equality favours the preservation of environmental quality.
Boyce provides several additional arguments that imply a negative
inuence of inequality on environmental quality. First, ecological irreversibility implies that a modication of power favouring those who benet from
deteriorations increases the long-term under-optimality of pollution levels,
but the opposite cannot work if degradations have already occurred and
have become irreparable. Second, environmental valuation produces a
shadow price integrated in the price vector. However, price formation is
inuenced by the ability to pay, which depends on the initial distribution of
wealth. The level of social inequalities is thus an additional channel of
inuence for the well-to-do classes. Likewise, access to information is more
dicult for poor people and, although they tend to be the victims, they also
underestimate the impact of environmental deterioration. Besides, wealthy
people have the ability to shape social compromises through the manipulation of information via marketing and media control. Third, wealthy people
tend to be wary of social protest movements, which may put an end to their
privileges. Their fear causes them to emphasize their preference for the
present (i.e. to overvalue the discount rate). They will increase the production
level of polluting activities in order to maximize their current results, thus
encouraging an increase of polluting emissions or a greater exploitation of
natural resources.
Scruggs (1998) contests these developments by claiming that Boyces
architecture is based entirely on the assumption that wealthy people systematically prefer a degraded environment.2 He contests this idea by recalling
that the literature in the eld suggests on the contrary that the quality of
the environment is a luxury good, the demand for which increases more
rapidly than income levels. A non-egalitarian society should then generate
less environmental deterioration.3 Referring to the new paradigm of social
modernization (Hofrichter and Reif 1990), Scruggs claims that there may be
a threshold of average individual wealth beyond which the direction of the
impact of inequality on pollutions is reversed. Boyces hypothesis is veried
in the case of the poorest countries but, in the most developed countries,
2 Note that this assertion needs to be carefully considered since Boyce explicitly disavows this claim. He
asserts instead that the willingness to support environmental public policies, expressed by wealthy
individuals, could be oset by other arguments in their utility functions. Moreover, this contradiction is
amplied when they are able to move away from polluted sites.
3 The role of an income elasticity of environmental quality exceeding the unit is not as obvious as Scruggs
claims. Indeed, even in this case, a wealthy consumer might be confronted with the necessity of comparing
the usefulness of protecting any ecosystem and the opportunity cost of renouncing the purchase of a
polluting good.
417
who are aected head-on by local pollutions. Therefore, they are the best
informed about ecological deterioration, especially where pollution aects
vital ecosystem services such as drinking water. When the representation of
the most vulnerable within public institutions is poor, their economic
interests are not properly defended and the resulting asymmetry is also
reected in their environmental interests. Since the wealthiest individuals
have a greater ability to protect themselves against local pollution (even
if this is an illusion), society may tend to underestimate environmental
deteriorations. For example, if a territory only has a limited access to safe
drinking water, the poorest will generally be the most severely aected by
water rationing. If the authorities fail to take account of the interests of the
poorest people, the construction of infrastructures designed to improve the
supply of drinking water may turn out to be thoroughly inadequate.
As explained by Bourguignon (2004), Easterly (2007) and Ravallion
(2007), inequality is a major cause of poverty and underdevelopment. This is
why inequality may also have an indirect impact on environmental
deterioration through its impact on poverty. The literature exploring the
connection between poverty and the environment often construes this
relation as a vicious circle (Barbier 2002). The idea is that poor farmers may
have no other choice but to make an excessive withdrawal of the natural
endowments of their ecosystems in order to draw a minimum income
(Mink 1993; Rozelle et al. 1997). In order to minimize the risk of protest
movements, public authorities may, if not positively encourage, at least fail
to constrain deforestation, resulting from the cultivation of new lands, and
the smuggling of protected species. As a result of these survival strategies, the
quality of environmental resources and crop yields may be reduced, thus
further impoverishing farmers.
Last, the ability of democratic congurations to take better account of the
interests of the most vulnerable is also often considered in the economic
literature. The general idea is that the interests of each social group are better
upheld in nations with free elections, by granting a vote to every agent. The
state ought therefore to take better care of the needs of the poorest, and
ecological problems need to be managed more eciently. Using a theoretical
model based on utilities, Eriksson and Persson (2003) show that the impact
of a more equal distribution of income on the environment depends on
the degree of democratic governance. In a full democracy, a decrease of
inequality reduces pollution, ceteris paribus, whereas the opposite occurs
under authoritarian regimes. However, since political systems vary widely,
the reactions of regulators to ecological conicts may dier signicantly
(Bernauer and Koubi 2009). The inuence capacities of interest groups are
419
Empirical Findings
There is a fairly extensive empirical literature studying the relation between
the environment and inequalities. The present survey is restricted to the main
studies that specically address the Environmental Kuznets Curve and the
impact of inequality.
Torras and Boyce (1998) were the rst to provide an analysis of these
issues. Their aim was to demonstrate that inequalities interact with per capita
GDP levels in explaining pollution. More specically, their object was to test
the hypothesis according to which the eect of inequality on pollution diers
in low- and high-income countries. They estimated the equation that links
the pollution level to the level of per capita GDP and negotiating power. The
latter is approximated by the Gini index, the literacy rate and an indicator
measuring civil rights and political freedoms. The authors resort for the most
part to data drawn from the GEMS database for pollutant concentrations
throughout the 19771991 period in 1852 cities in 1942 countries
(depending on the type of pollution). In the rst instance, they estimated
the regressions without inequality variables. Their results are similar to those
outlined in Grossman and Krueger (1995) and conrm the EKC for most of
the pollutants. Second, they included inequality variables and found that the
statistical signicance of per capita GDP eects is reduced. Globally, an
increase of literacy rates in low-income countries tends to reduce pollution
levels. The authors observe the same result for the political rights index.
The Gini index has mixed eects according to the pollutant under analysis.
For SO2 and smoke, greater income inequality is associated with more
pollution in low-income countries. An increase of inequality also reduces the
percentage of the population with access to safe water in the poorest
countries. In other words, the weaker the per capita GDP level is, the more
likely inequality is to have a harmful impact on the quality of the
environment. Yet in the case of heavy particles and dissolved oxygen, they
420
country. The relationship between pollution and GDP in a specic country may
be unrelated to the relation observed with panel data estimations on account of
country-specic characteristics (De Bruyn et al. 1998). This is why Gergel et al.
(2004) recommend using long-term time-series rather than panel or crosssection data. However, one advantage of panel data estimations as opposed to
cross-country estimations is that they help to model heterogeneity across
countries and time. More precisely, a xed-eects approach avoids the
assumption of homogeneity across countries by estimating a separate intercept
for each country. It takes account therefore of a certain degree of heterogeneity,
which helps to reduce the problem of country-specic factors. On the other
hand, one drawback of panel data or time-series estimations is that their focus is
centred on year-to-year changes. We may assume therefore that year-to-year
variations in independent variables (per capita GDP, inequality or others)
induce an adjustment in pollution levels within the same year in which the
process occurs. But the relationship between wealth, inequality and environmental quality is a long-term issue, and adjustment processes relating to
pollution levels are clearly not instantaneous. According to Torras and Boyce
(1998), the use of cross-country estimations helps to focus strictly on longerterm determinants of inter-country variation concerning pollution levels by
eliminating short-term variations. But cross-country estimations assume
homogeneity among countries concerning the relationship between pollution
and inequality. In this paper, we propose to resolve this bias by estimating
dynamic panel data models that allow for a dynamic adjustment by integrating
the lag dependent variable as an explanatory variable (Halkos 2003).
Our sample includes 83 countries67 developing countries and 16
transition countries(Central and Eastern European Countries (CEEC)
and former Soviet republics) for which observations based on the 19882003
period are available. The per capita GDP observations are drawn from the
2007 World Development Indicators, and are expressed in 2000 constant
dollars and are Purchasing Power Parity (PPP)-adjusted.
Because there are no composite indicators of environmental quality,
two pollution variables are taken into account, with the requirement that
they refer exclusively to local pollutants. First, the data concerning sulphur
dioxide (SO2) emissions are drawn from the ASL database and its update
proposed by David Stern for the 1990s and 2000s (ASL and Associates 1997;
Stern 2005). These are expressed in kg per capita. Second, the observations
concerning organic water pollution are based on the 2007 World Development Indicators. These emissions are measured by the amount of oxygen
consumed by bacteria in water in breaking down industrial and domestic
waste (World Bank 2007: 149), and are expressed in kg per industrial worker.
422
Inequality data based on the Gini index are drawn from the World Income
Inequality Database of the World Institute for Development Economics
Research (WIDER), a database that includes approximately 5,000 observations taken from multiple sources and ranked according to their quality.
Whenever possible, we chose to retain high-quality observations that are
homogeneous in terms of unit of measurement and covered population.
However, the inequality data raise two diculties. First, they raise an issue of
comparability since the household surveys used to estimate the distribution
of living standard are based on distinct sampling methods and the indicator
of living standards used here is not homogeneous across countries.4 Second,
the data concerning inequality are signicantly fragmentary, which accounts
for the poorly balanced panel. Indeed, out of 1,328 theoretical observations
(83 countries and 16 years), we only have 412 observations for the Gini
index.
As a proxy of power inequality, we use the political rights index of the
non-governmental organization Freedom House, which measures in particular the extent of political freedom according to four criteria: (i) the extent
of freedom in electoral processes, (ii) pluralism, (iii) citizen participation in
political life, and (iv) the practical methods of governance. The indicator
then grades countries based on the extent of political freedom, with grade 7
corresponding to a highly limited level of political rights and grade 1 to a
fully democratic regime.
ECONOMETRIC ANALYSIS
In the rst instance we discuss the pertinence of the EKC for SO2 emissions
and organic water pollution in developing and transition countries. Second,
we introduce potentially omitted variables, i.e. inequality and political
freedoms, in order to test the equality hypothesis.
423
model (FEM) with both individual and temporal eects.5 The expression of
this model is given by:
POLit b0 b1 GDPhit b2 GDPh2it b3 GDPh3it ai lt eit
where POL represents the level of pollution and GDPh the GDP per capita.
The xed eects are constants to be estimated that are specic to countries
(ai) and years (lt). They indicate the inuence of non-observable
characteristicsrespectively time constants and constants between individuals. The xed eects model estimator is the least squares with dummy
variables estimator (LSDV), equivalent to the ordinary least squares (OLS)
applied to the model expressed as deviations from individual means. The
LSDV estimator can produce biased coecient estimates with small samples.
To address this problem, and following Halkos (2003), we use the dynamic
panel data model elaborated by Arellano and Bond (1991), which involves a
Generalized Method of Moments (GMM) estimation of a dependent
variable as a function of its lagged value and exogenous variables. The
expression of this dynamic model is thus given by:
POLit b0 b1 POLit1 b2 GDPhit b3 GDPh2it b4 GDPh3it ai lt eit
The ArellanoBond estimator consists of rst-dierencing this equation to
remove the individual eects (ai) before estimating the rst-dierence
equation by instrumental variables. More precisely, the rst-dierentiated
lag dependent variable is instrumented by the values of the dependent
variable lagged two or more periods. In this study we use only two lags of the
dependent variable as instruments since the number of observations is
limited, particularly when we introduce the Gini index. The choice of a
limited number of instruments is also informed by Roodman (2009), who
identies the risk entailed by instrument proliferation. The use of too many
instruments may generate results that are invalid and yet appear valid.
Furthermore, it could weaken the power of the Sargan/Hansen instrument
validity test. As a technique for reducing this bias, Roodman recommends
the use of only certain lags instead of all available lags for instruments.
Tables 1a and 1b present respectively the results for SO2 emissions and
organic water pollution. The dynamic models are well-specied since the
Sargan/Hansen tests always conrm the validity of instruments. We also
5 For an account of the technical issues related to these two models, see Baltagi (1995). The random-eects
model is less appropriate than the xed-eects model when the number of individuals is limited.
424
425
7,778
Turning point
0.0726
836
0.0011
(1.62)
6.09E08
(1.29)
2.8239
(1.30)
FEM
24.59
(0.4285)
70.30
(0.7641)
711
0.0016
(1.37)
2.40E08
(0.28)
70.1011
(70.04)
0.1841***
(3.61)
Dynamic
Developing countries
8,544
0.2545
121
0.0229***
(4.73)
71.34E06***
(75.57)
750.6337**
(72.28)
FEM
12.63
(0.987)
0.16
(0.8720)
107
0.0059
(1.58)
72.99E07
(71.57)
723.202
(71.42)
0.8598***
(16.94)
Dynamic
Transition countries
Figures in parentheses are t-ratios for regression coecients and signicance levels for the Sargan/Hansen and autocorrelation tests.
***statistically signicant at 1 percent level; **statistically signicant at 5 percent level; *statistically signicant at 10 percent level.
34.98
(0.068)*
1.30
(0.1945)
818
0.0019
(1.52)
71.60E07**
(72.24)
71.9382
(70.56)
0.8165***
(21.81)
0.0999
957
0.0091***
(9.27)
75.85E07***
(79.90)
710.6797***
(74.13)
Dynamic
R2 within
Sargan/Hansen
Test
Autocorrelation
order 2
GDPh2
GDPh
SO2 emissions
(t-1)
Constant
FEM
Global
Table 1a: Regression Results for the Determinants of SO2 Emissions, Fixed-eects Model (FEM) and Dynamic Panel Data Model
426
0.84
(0.4024)
Autocorrelation
order 2
604
79.58E14***
(73.84)
73.17E05***
(75.56)
3.13E09***
(4.61)
0.2233***
(7.42)
0.2579**
(2.09)
25.27
(0.5035)
785
0.1663
72.40E13***
(78.67)
76.96E05***
(711.44)
7.38E09***
(9.92)
0.3545***
(27.21)
Dynamic
N
R2 within
Test Sargan/
Hansen
GDPh3
GDPh2
GDPh
Water pollution
(t-1)
Constant
FEM
Global
589
0.0138
75.60E14**
(72.52)
71.41E05***
(72.66)
1.67E09***
(2.67)
0.2364***
(21.50)
FEM
71.08
(0.2789)
32.90
(0.1649)
445
77.57E14**
(72.43)
72.64E05***
(73.31)
2.57E09***
(2.79)
0.2622***
(9.33)
0.0337
(0.34)
Dynamic
Developing countries
196
0.3719
76.21E13***
(77.13)
71.38E04***
(79.72)
1.64E08***
(8.27)
0.4894***
(15.59)
FEM
(continued )
1.18
(0.2369)
6.73
(0.998)
159
71.13E13*
(71.80)
73.32E05***
(73.07)
3.37E09**
(2.29)
0.1607***
(4.25)
0.5951***
(5.48)
Dynamic
Transition countries
Table 1b: Regression Results for the Determinants of Organic Water Pollution, Fixed-eects Model (FEM) and Dynamic Panel
Data Model
427
7,352
13,148
FEM
Global
8,008
13,773
Dynamic
6,082
13,799
FEM
7,879
14,754
Dynamic
Developing countries
6,954
10,652
FEM
8,998
10,884
Dynamic
Transition countries
Figures in parentheses are t-ratios for regression coecients and signicance levels for the Sargan/Hansen and autocorrelation tests.
***statistically signicant at 1 percent level; **statistically signicant at 5 percent level; *statistically signicant at 10 percent level.
Turning point
Min.
Max.
428
against the grain of EKC predictions. For the entire sample, the minimum of
this relationship is observed for a per capita GDP situated between US$
7,000 and 8,000 (according to the model), which is close to the mean
development level of Malaysia, Chile and Russia over the considered period.
The maximum level corresponds to a per capita GDP situated between US$
13,000 and 14,000. The nal decreasing part of the relationship therefore has
a very limited signicance since only two countries within the sample (South
Korea and the Czech Republic) have a mean income above the maximum
level. The concept of a development level beyond which water pollution
decreases is plausible, but the threshold is so high that so far it has not been
reached by the majority of the countries included in our sample.
Finally, the nature of the relationship between environmental deterioration and GDP per capita is related to the specic pollutant in question and
the group of countries under study, but also to the econometric specication
that is adopted. Concerning SO2 emissions, the EKC is validated for
transition countries but not for developing countries. Furthermore, this
relationship is fundamentally dynamic because of adjustment delays of
emissions to per capita GDP variations. In the case of organic water
pollution, the EKC is never conrmed. In line with Gates et al. (2002) and He
et al. (2007), we argue that the relationship between pollution and
development depends on factors related to the social and political situation
of the country in question.
431
7,254
Turning
7,340
0.2958
307
7,142
0.2738
307
(70.68)
70.0885
(70.64)
70.0828
(78.02)
77.77E07***
(6.22)
0.0111***
(71.25)
79.403
(3)
0.0846
228
0.0863
228
(0.54)
0.2015
(0.06)
0.0055
(0.15)
(0.21)
(1.01)
1.88E08
(0.95)
2.63E08
0.0018
(70.04)
0.0016
70.2851
(0.24)
(2)
1.0822
(1)
0.0894
228
(70.04)
70.0041
(0.18)
0.0167
(0.23)
2.83E08
(0.89)
0.0015
(0.17)
1.0412
(3)
Developing countries
7,380
0.3504
82
(73.87)
71.05E06***
(2.71)
0.0155***
(70.41)
711.4439
(1)
6,678
0.5039
79
(2.79)
3.8477***
(0.79)
0.397
(74.08)
78.76E07***
(2.54)
0.0117**
(70.61)
714.4905
(2)
Transition countries
point
0.2406
310
(2.69)
1.2589***
(70.30)
70.0389
(78.36)
(77.53)
(6.60)
77.97E07***
(5.89)
78.34E07***
0.0117***
(72.17)
0.0121***
717.1758**
(72.20)
(2)
715.2826**
R2 within
dictature
Gini_
democracy
Gini_
freedoms
Political
Gini
GDPh2
GDPh
Constant
(1)
Whole sample
Table 2a: Regression Results for the Determinants of SO2 Emissions, Fixed-eects Model
7,142
0.4495
79
(70.52)
70.2535
(70.76)
70.3405
(74.16)
79.38E07***
(2.78)
0.0134***
(0.19)
4.4564
(3)
IS INEQUALITY HARMFUL FOR THE ENVIRONMENT?
432
112
(70.76)
110
(1.50)
1.7725
43
43
(70.89)
(70.63)
110
70.1303
(70.73)
(70.28)
70.1609
70.0948
(1.35)
2.53E07
(72.15)
70.0079**
(71.75)
70.1982*
(3.35)
43.5433***
(3)
70.0670
43
(1.40)
1.1232
(70.13)
(0.63)
1.26E07
(0.04)
(0.27)
3.04E08
(71.44)
70.0056
(71.25)
70.1514
(1.94)
28.4686*
(2)
70.0175
(70.97)
71.92E07
(70.24)
70.0005
(6.20)
0.6778***
(0.37)
3.4655
(1)
0.0109
(71.33)
72.60E07
(70.14)
70.0005
(5.12)
0.4582***
(1.44)
23.4716
(3)
Developing countries
69
(70.63)
71.54E07
(0.20)
0.0009
(6.43)
0.7743***
(0.48)
11.0690
(1)
67
(0.72)
1.3083
(70.04)
70.0199
(71.59)
73.75E07
(0.62)
0.0030
(3.80)
0.4597***
(0.77)
19.6511
(2)
Transition countries
Gini_dictature
Gini_democracy
freedoms
Political
Gini
71.62E07
(0.26)
(0.30)
GDPh2
0.0010
0.0012
(4.47)
(7.62)
(t-1)
0.4203***
(0.49)
(0.35)
0.7731***
8.2683
6.2962
(2)
GDPh
SO2 emissions
Constant
(1)
Whole sample
Table 2b: Regression Results for the Determinants of SO2 Emissions, Dynamic Panel Data Model
67
(70.15)
70.0645
(0.28)
0.1223
(71.12)
72.69E07
(0.10)
0.0004
(4.44)
0.4923***
(1.26)
31.9257
(3)
433
dictature
Gini_
democracy
Gini_
freedoms
Political
Gini
GDPh3
GDPh2
GDPh
Constant
(8.30)
(8.29)
0.3985***
(3)
(9.23)
0.2270***
(1)
(4.75)
(74.54)
(4.74)
9.90E09***
(71.56)
(2.11)
2.86E09**
(11.11)
0.6540***
(1)
(7.57)
0.7896***
(2)
(7.53)
0.7753***
(3)
(71.65)
(72.15)
2.93E09**
(71.66)
(72.12)
2.89E09**
(78.38)
(7.88)
2.24E08***
(75.92)
(5.98)
2.52E08***
(75.91)
(5.96)
2.50E08***
(77.83)
(70.55)
70.001
(71.69)
70.008*
(71.47)
70.0007
(70.54)
70.0006
70.86
0.0003
70.91
0.0003
(75.99)
(70.58)
70.0023
(73.11)
(77.37)
70.89
(72.23)
(71.71)
(72.26)
70.0030***
(72.25)
0.0003
(74.85)
70.0008*
(74.87)
(continued )
(73.27)
70.0030***
(73.01)
70.0027***
(75.97)
(8.25)
(74.55)
9.93E09***
(78.59)
1.49E08***
(77.49)
0.2143***
(3)
Transition countries
(77.53)
0.2181***
(2)
Developing countries
0.4054***
(13.90)
(2)
0.4836***
(1)
Whole sample
Table 3a: Regression Results for the Determinants of Organic Water Pollution, Fixed-eects Model
434
259
11,147
6,320
0.1289
254
(2)
Whole sample
11,148
6,312
0.1299
254
(3)
0.0874
160
(1)
12,928
4,207
0.0958
160
(2)
Developing countries
12,916
4,134
0.0948
160
(3)
10,929
7,172
0.4813
99
(1)
11,200
6,787
0.3696
94
(2)
Transition countries
6,742
11,586
Max.
0.2881
Min.
Turning Point
R2 within
(1)
11,067
6,873
0.3736
94
(3)
435
dictature
Gini_
democracy
Gini_
freedoms
Political
Gini
GDPh3
GDPh2
GDPh
pollution (t-1)
Water
Constant
(1.02)
(0.99)
0.2133
(4.33)
0.4252***
(3)
(2.85)
(2.05)
0.5628**
(1.00)
1.2313***
0.1003
0.037
(2)
(0.27)
(1)
(3)
(2.04)
0.5847**
(0.64)
0.0676
Developing countries
(72.92)
(70.84)
(0.93)
4.53E09
(70.85)
(1.01)
3.32E09
(70.71)
(70.94)
70.0006
(70.75)
70.0004
(71.04)
(70.34)
70.0006
(1.62)
(70.94)
(70.82)
(72.61)
0.0009
70.0019
(70.55)
(0.74)
2.53E09
(2.85)
1.45E08***
(72.75)
70.0001***
(0.71)
0.1379
(3.34)
0.5006***
(1)
(2.94)
1.99E08***
(72.96)
70.0001***
(0.80)
0.1446
(3.75)
0.6039***
(2)
Transition countries
(2.49)
1.84E08**
(72.65)
70.0001***
(0.75)
0.1905
(3.22)
0.5777***
(3)
(1.98)
0.0011**
(1.84)
0.0010*
(70.81)
(73.00)
(0.19)
0.0007
(70.41)
70.0003
(73.00)
(continued )
(70.59)
70.0005
(70.53)
70.0003
(72.44)
(2.71)
9.53E09***
70.0005
(72.46)
(71.56)
(2.52)
73.05E13**
72.09E13
(1.53)
(72.69)
9.13E09**
5.97E09
(71.62)
(0.50)
0.1957
(4.26)
0.1066
0.4191***
(2.99)
(2)
0.3398***
(1)
Whole sample
Table 3b: Regression Results for the Determinants of Organic Water Pollution, Dynamic Panel Data Model
436
11215
8741
93
(2)
Whole sample
11225
8881
93
(3)
41
(1)
41
(2)
Developing countries
41
(3)
9813
8022
54
(1)
10620
7504
52
(2)
Transition countries
Max.
95
Min.
Turning point
(1)
10467
7679
52
(3)
Organic water pollution. When referring to Tables 3a and 3b, the reduction of
the number of observations partly modies the results on the relationship
between water pollution and per capita GDP. With GMM estimations, the
N-inverted relationship displayed by our previous estimations is no longer
observed for the entire sample (model 1) and for developing countries
(models 1 to 3). When this relationship is still observed, the turning points
are generally weaker. For the entire sample, the maximum of the N-inverted
relationship is reached for a per capita GDP of around US$ 11,000 with
xed-eects estimations. The nal decreasing part thus has more signicance
since four countries (South Korea, Argentina, the Czech Republic and
Hungary) rather than two present higher mean per capita incomes over the
19882003 period.
The impact of inequality variables is clearly context-specic in so far as it
depends on the specic group of countries under consideration. Developing
countries, especially Latin American and sub-Saharan African countries,
experienced the highest levels of water pollution (Table 4). Yet for these
countries the xed-models are not well-specied in so far as the R2 within are
systematically inferior to 0.10. However, for the GMM estimations, we may
note a signicant and positive eect of the two interaction terms (model 3).
Despite the low number of observations, this result has two important
implications. First of all, it conrms that the relation between inequality and
environmental quality is more accurate for localized pollutants than for more
global pollutants such as air pollutants. Second, the signicance of these two
interaction terms indicates that the inuence of inequality on organic water
pollution is constrained by the political context. To some extent, these
estimations highlight an indirect eect by which a high level of wealth
inequality has an adverse eect on water pollution since it contributes to the
deterioration of political rights (a proxy of power inequality). This result
validates Boyces analysis, which relies on the combination of wealth
inequality and power inequality for explaining pollution levels. Therefore, if
it is assumed that poor people are the most severely aected by water
pollution, in a context of strong inequality and constrained political
freedoms, these people are unlikely to be able to assert their interest in the
preservation of water quality. Nevertheless, the impact of inequality is
signicant in both democratic and authoritarian regimes, which indicates
that the impact of a more equal income distribution on water quality does
not depend on the degree of democratic rule, contrary to the view defended
by Eriksson and Persson (2003).
The impact of income inequality on water pollution in transition countries
is markedly dierent since the coecient on the Gini index is negative and
437
SO2
emissions1
Water
pollution2
GDP per
cap.3
Gini
Politics4
Asia
4.6120
0.1620
3,614.9
0.375
4.26
Bangladesh
Cambodia
China
India
0.5318
1.1787
8.0655
2.8263
0.1585
0.1497
0.1381
0.1967
1,373.2
1,576.3
2,937.1
2,017.6
0.306
0.463
0.368
0.309
2.94
6.00
6.94
2.5
Indonesia
Korea. Rep
Lao
Malaysia
1.8261
0.5027
6.1847
0.1800
0.1231
0.1216
2,737.5
13,380.0
1,286.1
7,286.6
0.333
0.335
0.340
0.479
5.31
2.00
6.75
4.63
Mongolia
Pakistan
Philippines
Sri Lanka
19.3226
3.6925
3.7988
1.2155
0.185
0.1791
0.1889
0.1805
1,440.4
1,778.9
3,795.0
2,930.4
0.321
0.377
0.440
0.349
2.81
4.38
2.31
3.44
9.1758
1.1049
7.0720
0.1744
0.1890
5,799.5
1,633.6
4,232.6
0.437
0.362
0.391
2.69
6.94
5.16
Thailand
Viet Nam
Middle East
and North Africa
Algeria
5.1882
0.243
5,262.2
0.377
5.69
Egypt
Iran
Jordan
Morocco
4.6415
9.6123
12.4528
4.5258
0.1918
0.1587
0.1817
0.1703
3,121.1
5,363.9
4,060.1
3,487.4
0.330
0.432
0.395
0.393
5.69
5.94
4.44
4.88
Tunisia
Turkey
Yemen. Rep
Sub-saharan
12.3346
6.3578
1.4648
5.9740
0.1514
0.1772
0.2526
0.2340
5,456.3
5,908.4
768.8
1,962.3
0.406
0.417
0.364
0.456
5.81
3.44
5.38
4.76
Africa
Burkina Faso
Burundi
Cameroon
1.4720
0.5079
2.0825
0.2154
0.2423
0.2321
955.2
763.9
1,905.4
0.457
0.376
0.457
5.00
6.44
6.37
(continued )
438
Table 4: (Continued )
SO2
emissions1
Water
pollution2
GDP per
cap.3
Gini
Politics4
Cote dIvoire
Djibouti
1.6541
3.0317
0.2238
1,600.5
2,173.1
0.398
0.447
5.94
5.19
Ethiopia
Ghana
Guinea
Kenya
1.2045
0.9208
1.8508
1.3010
0.2211
0.1737
0.2364
782.5
1,785.0
1,869.5
1,060.7
0.405
0.354
0.529
0.477
5.31
3.94
6.13
5.63
Lesotho
Madagascar
Malawi
Mali
1.4830
0.7932
2.6503
0.1699
0.2706
0.2803
2,409.2
846.7
542.3
745.0
0.606
0.458
0.557
0.453
4.31
2.88
4.13
3.19
Mauritania
Mauritius
Mozambique
Niger
4.8120
3.5068
2.8735
1.5874
0.1574
0.2762
0.3038
1,882.9
8,197.6
717.2
730.8
0.425
0.379
0.435
0.427
6.25
1.25
4.13
5.06
Nigeria
Senegal
South Africa
Swaziland
Tanzania
3.5067
2.0745
40.4940
1.8430
0.1905
0.3240
0.1767
0.2647
0.2518
844.1
1,375.7
8,842.6
4,109.1
514.4
0.469
0.456
0.579
0.555
0.342
5.44
3.56
2.63
6.00
5.13
Uganda
Zambia
Zimbabwe
Latin America
1.0010
51.8547
7.3978
11.5850
0.2588
0.2296
0.1994
0.2310
1,003.2
850.1
2,576.9
5,563.2
0.473
0.516
0.531
0.514
5.31
4.31
5.50
2.49
Argentina
Bolivia
Brazil
Chile
5.3485
4.9876
6.5807
91.9979
0.2126
0.2377
0.1941
0.2316
11,045.2
2,238.8
6,860.8
7,777.5
0.490
0.601
0.591
0.550
1.94
1.75
2.38
2.31
Colombia
Costa Rica
2.9077
2.8036
0.1971
0.2118
5,888.3
7,267.0
0.559
0.483
3.25
1.00
(continued )
439
Table 4: (Continued )
SO2
emissions1
Water
pollution2
GDP per
cap.3
Gini
Politics4
Dominican
Republic
5.2976
5,425.8
0.501
2.38
Ecuador
El Salvador
Guatemala
Honduras
6.0418
2.6117
2.1353
2.9587
0.2526
0.1952
0.2699
0.2151
3,325.9
4,195.2
3,731.1
2,841.1
0.582
0.510
0.563
0.553
2.31
2.56
3.38
2.56
Jamaica
Mexico
Nicaragua
Panama
18.4621
13.2596
3.7350
6.6014
0.2889
0.1944
0.2912
3,517.2
8,251.8
2,873.0
5,373.7
0.415
0.525
0.463
0.566
2.00
3.25
3.44
2.69
Paraguay
Peru
Trinidad
and Tobago
2.0081
15.5098
13.9399
0.2757
0.2014
0.2553
4,381.8
4,423.0
8,036.0
0.548
0.486
0.407
3.88
3.63
1.44
Uruguay
Venezuela
Transition
countries
Albania
10.3873
13.8790
33.6450
0.2477
0.2075
0.1700
8,031.3
5,879.8
6,611.1
0.436
0.461
0.298
1.25
2.37
3.14
5.9074
0.1955
3,113.9
0.289
4.13
Armenia
Azerbaijan
Bulgaria
Croatia
85.5807
11.4495
0.2064
0.1562
0.1461
0.1617
2,318.2
2,787.7
6,200.7
8,643.9
0.381
0.362
0.329
0.257
4.08
5.85
2.50
3.31
Czech Republic
Hungary
Kyrgyz Republic
Lithuania
52.0943
35.5842
0.1328
0.1597
0.1622
0.1677
14,493.8
11,599.3
1,712.8
8,754.3
0.221
0.238
0.381
0.313
1.00
1.63
4.92
1.15
Macedonia
Moldova
Poland
25.7148
32.7720
0.1714
0.3139
0.1580
5,833.4
2,018.6
8,918.5
0.354
0.336
0.311
3.50
3.31
1.75
(continued )
440
Table 4: (Continued )
SO2
emissions1
Water
pollution2
GDP per
cap.3
Gini
Politics4
Romania
Russian Federation
22.7584
0.1004
0.1683
6,297.9
7,659.4
0.289
0.400
3.56
4.19
Slovak Republic
Ukraine
Whole sample
27.9719
11.0260
0.1409
0.1558
0.1990
10,456.0
5,717.9
4,201.0
0.249
0.282
0.410
1.64
3.44
3.87
Notes: (1) kg per capita; (2) kg per worker; (3) constant US$ 2,000, PPP; (4) the political
freedoms index grades countries according to the extent of their political freedom, grade 7
corresponding to a very restrained level of political rights and grade 1 to a democratic regime.
2.
3.
4.
REFERENCES
Arellano, M. and Bond, S. (1991) Some Tests of Specication for Panel Data: Monte
Carlo Evidence and an Application to Employment Equations, Review of Economic
Studies 58(2): 277297.
Arrow, K., Bolin, B., Costanza, R., Dasgupta, P., Folke, C., Holling, C. S., Jansson, B.
O., Levin, S., Maler, K. G., Perrings, C. and Pimentel, D. (1995) Economic Growth,
Carrying Capacity and the Environment, Science 268: 520521.
ASL and Associates (1997) Sulfur Emissions by Country and Year, Washington, DC:
US Department of Energy, Report no. DE96014790.
Baltagi, B. H. (1995) Econometric Analysis of Panel Data, Chichester: John Wiley & Sons.
Barbier, E. B. (2002) Development, Poverty and Environment, in J. van der Bergh (ed.)
Handbook of Environmental and Resource Economics, Cheltenham: Edward Elgar
Publishing.
Barrett, S. and Graddy, K. (2000) Freedom, Growth and the Environment, Environment
and Development Economics 5(4): 433456.
Benabou, R. (1993) Workings of a City: Location, Education and Production, The
Quarterly Journal of Economics 108(3): 619652.
Bernauer, T. and Koubi, V. (2009) Eects of Political Institutions on Air Quality,
Ecological Economics 68(5): 13551365.
8 The technological progress can help to bring about a slight reduction of the ecological constraint, although
it can in no way help to avoid it.
443
444
Lijphart, A. (1999) Patterns of Democracy: Government Forms and Performance in Thirtysix Countries, New Haven, CT: Yale University Press.
Mani, M. and Wheeler, D. (1998) In Search of Pollution Havens? Dirty Industry in the
World Economy, 19601995, Journal of Environment and Development 7(3): 215247.
Mikkelson, G. M., Gonzalez, A. and Peterson, G. D. (2007) Economic Inequality
Predicts Biodiversity Loss, Plos One 2(5): 15.
Mink, S. D. (1993) Poverty, Population and the Environment, World Bank Discussion
Paper, no. 189, Washington, DC: World Bank.
Ostrom, E. (1990) Governing the Commons, The Evolution of Institutions for Collective
Action, Cambridge: Cambridge University Press.
Ravallion, M. (2007) Inequality is Bad for the Poor, in J. Micklewright and S. Jenkins
(eds) Inequality and Poverty Re-examined, Oxford: Oxford University Press.
Rock, M. (1996) Pollution Intensity of GDP and Trade Policy: can the World Bank be
Wrong?, World Development 24(3): 471479.
Roodman, D. (2009) A Note on the Theme of Too Many Instruments, Oxford Bulletin
of Economic and Statistics 71(1): 135158.
Rozelle, S., Huang, J. and Zhang, L. (1997) Poverty, Population and Environmental
Degradation in China, Food Policy 22(3): 229251.
Scharpf, W. (2000) Institutions in Comparative Policy Research, Comparative Political
Studies 33(6/7): 762790.
Scruggs, L. A. (1998) Political and Economic Inequality and the Environment,
Ecological Economics 26(3): 259275.
Selden, T. and Song, D. (1994) Environmental Quality and Development: is there a
Kuznets Curve for Air Pollution Emissions?, Journal of Environmental Economics and
Management 27(2): 147162.
Stern, D. I. (2005) Global Sulfur Emissions from 1850 to 2000, Chemosphere 58(2): 163
175.
Torras, M. and Boyce, J. K. (1998) Income, Inequality and Pollution: A Reassessment of
the Environmental Kuznets Curve, Ecological Economics 25(2): 147160.
Tsebelis, G. (1995) Decision Making in Political Systems: Veto Players in Presidentialism,
Parliamentarism, Multicameralism and Multipartyism, British Journal of Political
Science 25(3): 289325.
WCED (1987) Our Common Future, Oxford: Oxford University Press.
World Bank (1992) World Development Report 1992. Development and the Environment,
Washington, DC: World Bank.
World Bank (2007) World Development Indicators 2007, Washington, DC: World Bank.
445
Copyright of Review of Social Economy is the property of Routledge and its content may not be copied or
emailed to multiple sites or posted to a listserv without the copyright holder's express written permission.
However, users may print, download, or email articles for individual use.