Ecological Economics 36 (2001) 513– 531

www.elsevier.com/locate/ecolecon

ANALYSIS

Interrelationships between income, health and the

environment: extending the Environmental Kuznets Curve
hypothesis
Lata Gangadharan a,*, Ma. Rebecca Valenzuela b
a
Department of Economics, Uni6ersity of Melbourne, Melbourne, Vic. 3010, Australia
b
Department of Economics, Monash Uni6ersity, Vic. 3145, Australia
Received 4 April 2000; received in revised form 19 September 2000; accepted 20 September 2000

Abstract

This paper examines the link between the health indicators and the environmental variables for a cross-section of
countries widely dispersed on the economic development spectrum. While environment and income are seen to have
an inverted-U shaped relationship (Environmental Kuznets Curve (EKC) hypothesis), it is also well established that
environment and health are positively related. Our study focuses on the implications of this for the relationship
between health and income. In the early phases of income growth, the gains in health and the losses in environmental
quality could cancel each other out and this challenges the idea that as incomes increase health would always
improve. To empirically analyse these issues, we estimate a two-stage least squares model that focuses on the impact
of income and the environment on health status, with environment being an endogenous variable. Our results show
that the environmental stress variable has a significant negative effect on health status. At the same time, gross
national product (GNP) levels are shown to vary positively with health status variables. We find that the health gains
obtained through improved incomes can be negated to a significant extent if the indirect effect of income acting via
the environment is ignored. Research findings in this regard would be a useful policy instrument towards maximising
both the environmental and health gains that come with economic growth and development. © 2001 Elsevier Science
B.V. All rights reserved.

Keywords: Environmental stress; Health indicators; Income levels

JEL classification: O11; Q25; C30

* Corresponding author.
E-mail addresses: gangadha@cupid.ecom.unimelb.edu.au (L. Gangadharan), rebecca.valenzuela@buseco.monash.edu.au (M.R.
Valenzuela).

0921-8009/01/$ - see front matter © 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 9 2 1 - 8 0 0 9 ( 0 0 ) 0 0 2 5 0 - 0
514 L. Gangadharan, M.R. Valenzuela / Ecological Economics 36 (2001) 513–531

1. Introduction mental stress. Hence, in the long run, pollution

levels are expected to improve with incomes. This
Is economic growth part of the solution rather argument has been used to justify the pursuit of
than the cause of environmental problems? This growth strategies that do not give due consider-
question has been raised very often in recent years ation to their effect on the environment.
as empirical evidence in support of the Environ- In this paper, we argue that this strategy is not
mental Kuznets Curve (EKC) hypothesis mount.1 justified and provide some evidence to support
The EKC (Grossman, 1995; Grossman and our case. We test the hypothesis that larger eco-
Kreuger, 1995) describes the relationship between nomic and social gains can be attained by an
declining environmental quality and income as an economy if the growth strategy adopted incorpo-
inverted-U, that is, in the course of economic rates, rather than ignores, environmental con-
growth and development, environmental quality cerns. To do this, we include health as the
initially worsens but ultimately improves with im- intervening variable in the analysis. While envi-
provements in income levels. For instance, Torras ronment and income are seen to have an inverted-
and Boyce (1998) show that the level of air pollu- U shaped relationship, it is also well established
tants (sulphur dioxide and smoke) peak at a per that health and environment are positively re-
capita income in the neighbourhood of US $4000, lated. What does this imply for the relationship
after which they start falling. between health and income? It is possible that in
One of the explanations for the EKC relation- the early phases of income growth, the gains in
ship is that the environment can be thought of as health and the losses in environmental quality
cancel each other out and this challenges the idea
a luxury good. In the early stages of economic
that as incomes increase, health would always
development, a country would be unwilling to
improve.
trade consumption for investment in environmen-
In view of the above, we argue in this paper
tal regulation, hence environmental quality de-
that the recorded health gains brought about by
clines. Once the country reaches a threshold level
the improvement in income levels do not repre-
of income, its citizens start to demand improve-
sent the total realisable health benefits from hav-
ments in environmental quality and this leads to
ing higher per capita income. Without the
the implementation of policies for environmental
appropriate environmental protection policies,
protection and, eventually, to reductions in pollu- damages to a country’s physical environment are
tion. Increasing levels of pollution are thus incurred during the process of income growth and
strongly associated with both poor and develop- economic development. This negatively affects the
ing economies, while declining levels of pollution health and well-being of individuals in the country
are more commonly observed for their developed and the aggregated impact could negate some of
counterparts.2 Another explanation of the EKC the health gains already derived, and hence
hypothesis is that countries pass through techno- dampen achievement levels in the health area. If
logical life cycles, as they move from agriculture- we find that this argument has some empirical
based economies to service-based systems. As the support, it would imply that development policies
service sector is associated with lower environ- addressing environmental issues are effectively
mental impact, this transition from high polluting also addressing the health issues of the economy.
to low polluting technology leads to less environ- In that case, policies that pursue economic devel-
opment cannot afford to ignore environmental
issues, particularly in the early phases of eco-
1
The original Kuznets curve refers to an inverted-U shaped nomic growth.
relationship between per capita income and inequality ob-
served by Kuznets (1955).
We look at recent evidence from a cross-section
2
For example, Grossman and Kreuger (1995) found evi- of countries to determine if this is indeed the case.
dence in support for the EKC hypothesis for 12 of the 14 air While there are some studies (for example, Crop-
and water quality variables for a cross-section of countries. per et al., 1997) that look at the incidences of air
 L. Gangadharan, M.R. Valenzuela / Ecological Economics 36 (2001) 513–531 515

or water pollution-related diseases in a particular (1998), Torras and Boyce (1998), Suri and Chap-
region or country, this is one of the first papers to man (1998)and Agras and Chapman (1999) argue
analyse the link between the health indicators and that the EKC’s previously estimated could be due
the environmental variables for a cross-section of to important missing variables. Kaufmann et al.
countries widely dispersed on the economic devel- (1998) stress the importance of spatial intensity of
opment spectrum. economic activity, Torras and Boyce (1998) ex-
The major contributions of this paper are the plore the effects of social factors like civil rights,
following. income inequality and education, while Suri and
1. To establish the link between health and envi- Chapman (1998)and Agras and Chapman (1999)
ronment. While this link has been suggested find that trade-related variables and the price of
before, to the best of our knowledge it has not energy have significant explanatory power. Most
been explored for different countries. of the papers mentioned use linear or a log linear
2. To explore the differential impact on health of functional relationship between emissions and in-
several environmental stress variables. The come. An exception is Galeotti and Lanza (1999),
standard practice is to focus on just one vari- which studied relationships based on the gamma
able, normally CO2 emissions. We attempt to distribution. The current state of the research on
make our findings and conclusions more ro- the EKC is unable to conclude if the EKC hy-
bust by including several other pollutants and pothesis is confirmed or rejected.
environmental damage indicators in the With regards to health, there exists a large
analysis. literature that has analysed the relationship be-
3. To analyse the shape of the EKC curve and tween income and health using cross-country data
the health relationship using alternative func- (for example, Gerdtham et al., 1992; Chakrabarti
tional forms to determine which among them and Rao, 1999). A number of previous studies in
best fits the data. this literature have found an economically and
The rest of the paper is organised as follows. statistically significant and negative income elas-
Section 2 surveys the related literature on the ticity of infant mortality rate (see for example,
relationship between income levels and environ- Flegg, 1982; Parpel and Pillai, 1986; Hill and
mental stress and the link between income levels King, 1992; Kakwani, 1993; Subbarao and
and health status. Section 3 describes the analyti- Raney, 1995; Pritchett and Summers, 1996). Simi-
cal framework and the estimation methodology larly, research on life expectancy and income has
used in the paper. Section 4 summarises the data shown that there is a positive relationship between
used in the analysis. Section 5 discusses the results increases in income and life expectancy, with in-
from the estimation, and Section 6 concludes. come elasticity of life expectancy estimated to be
significant and positive (Preston, 1980; Hill and
King, 1992). Most of these studies also control for
2. Related literature survey other factors that affect health status such as the
accessibility of health services and education lev-
Panayotou (1993), Selden and Song (1994) and els of the population.
Grossman and Kreuger (1995) presented initial
evidence that some pollutants follow an inverted-
U shaped curve with respect to income. This was 3. Analytical framework
widely interpreted (for example, World Bank,
1992) to mean that the surest way to improve a In this paper, we are interested in how the
country’s physical environment is to increase in- interplay between income and the environment
come levels. More recent work has focused on affect the health outcomes of a population. Gen-
factors other than income as explanatory vari- erally, it is assumed that health outcomes for a
ables in analysing variations in environmental population improve as the economy grows and
stress in different countries. Kaufmann et al. develops. Such improvements are facilitated by
516 L. Gangadharan, M.R. Valenzuela / Ecological Economics 36 (2001) 513–531

the rise in general standard of living, including the proposition from recent research that environ-
improved access to educational opportunities and mental quality tends to increase once again with
health services. One’s health is also seen as depen- extremely high incomes (De Bruyn et al., 1998).
dent on the quality of his or her physical environ- The upward bend of the Kuznets curve at the very
ment — such as the amount of air pollution or high-income levels will be captured by the i13
the quality of drinking water. At the same time, term, which is expected to have a positive sign.
the quality of a country’s physical environment is The Zi term captures the effects of non-income
a result of certain growth factors in the economy. variables such as population levels, literacy rates
These include, for instance, the more intensive use and income inequality, which are thought to sig-
of land, forest and water resources to increase nificantly influence environmental outcomes. For
overall economic production. Air pollution levels example, countries with higher population densi-
are also bound to increase as production levels ties are expected to suffer from greater environ-
rise. Increase in population numbers is another mental stress as there would be more people
important factor in this context. sharing the existing environmental resources. A
The relationships discussed above are sum- country’s level of urbanisation is also thought to
marised in the following general model: negatively impact on the environment, although it
is possible that urbanisation can bring improved
Hi = f(Xi, Ei (Xi, Zi ), Wi ) (1) waste disposal and sanitation provisions for the
which states that an economy’s health status (Hi ) urban areas, and hence mitigate the detrimental
depends on its level of economic growth (Xi ), the environmental effects of an increase in popula-
quality of its environment (Ei ) and other social tion. Further, there are strong grounds for believ-
factors (Wi ) including the provision and access to ing that the education levels of the population are
health facilities. Zi is used to denote the factors positively associated with better environmental
that determine the quality of the environment. quality, while is not clear in which direction the
Within this framework, we test the relevance of gap between the rich and the poor influences the
the EKC hypothesis, captured in the term Ei (Xi, environment. In previous research, increases in
Zi ), and how it impacts on the health outcomes the inequality of income distribution have been
for a country’s population. associated with higher levels of pollution, as those
To empirically analyse these issues, the follow- who benefit from pollution-generating activities
ing econometric model is formulated for country are better able to prevail against those who bear
i: the costs (Torras and Boyce, 1998). However, this
hypothesis has been challenged by Scruggs (1998),
Ei = i0 +i11Xi + i12X 2i +i13X 3i +i2Zi +ei (2) who suggests that it is impossible to make gener-
Hi = h0 + h1Xi +h2Wi +h3Ei +ui (3) alisations about the effect of income distribution
on environmental degradation without having
where Ei refers to the overall level of environmen- more information about the preferences of differ-
tal stress in the economy; Hi refers to health ent income groups.
status of the population; Xi pertains to the coun- Eq. (3) postulates that the population’s overall
try’s level of economic growth; Zi, are non-income health and general well-being is dependent on
variables that impact on the environment; Wi, are three factors, the country’s level of economic
variables that directly influence health such as growth; the availability and accessibility of medi-
provision and access to medical facilities, etc.; ui, cal facilities; and the quality of the country’s
ei, are the error terms. physical environment. Xi and Ei are the same
Eq. (2) is the EKC, where the dependence of variables defined in Eq. (2) while Wi captures all
environmental quality on economic growth is rep- other variables thought relevant to health such as
resented in a cubic relationship. The inverted-U number of doctors and other medical workers,
shaped EKC requires i11 to be positive and i12 to immunisation levels, as well as literacy rates and
be negative. A cubic income term is added to test other population-related factors.
 L. Gangadharan, M.R. Valenzuela / Ecological Economics 36 (2001) 513–531 517

The quality of the physical environment is in- Twenty-two of the 51 countries (43%) come from
cluded here as an endogenous variable. We expect the high-income OECD set, but the population
that health indicators would show an improve- composition is dominated by those in the low-in-
ment with better availability of health care ser- come countries, with the inclusion of China and
vices and with higher rates of child immunisation. India in this group.
Urbanisation being an indicator of modernisation There are a number of environmental stress
could lead to improved health, as there might be variables that are available for analysis. The air
easier access to sophisticated health facilities. pollutants for which data are available are carbon
However, it is also possible that the same could dioxide (CO2), sulphur dioxide (SO2), nitrogen
impact negatively on health quality as urbanisa- oxides (NOx) and total suspended particulates
tion very often leads to overcrowding, poor sani- (TSP).4 Commercial energy use (ENPC) is taken
tation levels and hence more health ailments. We to be another environmental indicator. For water
expect higher educational levels to improve health pollution, we use data on emission levels of or-
quality as education lowers the cost of informa- ganic pollutants (EMW) while data on deforesta-
tion, and people with higher levels of education tion rates (DEFRTE) have also been obtained
might have a better understanding of the value of from the World Bank. CO2 emissions and ENPC
public health infrastructure and are better able to are very high for richer countries compared with
locate and utilise these services. In particular, low-income countries. Other air pollutants like
education has often been cited for its strong effect SO2, NOx and TSP are significantly higher for
on reducing child mortality (for example, Melling- low-income countries than for the high-income
ton and Cameron, 1999). countries in the sample. This could be due to the
In the above model, the structural equations are fact that richer countries have already in place
clearly identified given that Ei and Hi are the only environmental regulations targeting these pollu-
endogenous variables in the system while the rest tants, while this has yet to be implemented for
(Xi, Zi and Wi ) are taken to be exogenous. The poorer countries. EMW and DEFRTE are much
equations are then estimated using general two- higher for low-income countries for similar
stage least squares estimation methods. The re- reasons.
sults obtained are subjected to a robustness test The link of the environment with the country’s
with regards to functional forms and to different level of economic growth is analysed using the
assumptions made regarding the type and nature country’s gross national product (GNP) per cap-
of the variables used. The coefficients from the ita, purchasing power parity, as the proxy vari-
estimated equations will indicate if environmental able for the latter.5 Population factors are
variables play an important role in improving controlled for the estimation through the coun-
health outcomes in a country.

4
The data for SO2, NOx and TSP were provided at the city
4. Data level not the country level. We obtain the country level data
using the city’s population as a proportion of the country’s
The data used in this paper are obtained from population.
5
the World Development Indicators (1998), com- It is acknowledged here that gross national product is just
one aspect of economic growth and development, and that
piled by the World Bank. Summary statistics of
other more broad based measures (for example, the Human
these data and the units of measurement used are Development Index or HDI) are available. We nonetheless use
presented in Table 1. There are a total of 51 GNP in this analysis to facilitate comparison with previous
countries included in the analysis, covering more EKC studies. It is noted that GNP is highly correlated with
than 70% of the world’s population (Appendix A).3 HDI values for the countries covered in this study (z= 0.82),
therefore, results should be robust between these two vari-
ables. For more details on the importance of alternative
3
The countries in the sample are listed in the Table of measures of income and growth on the EKC relationship, see
Appendix A. Munasinghe (1999).
 518
Table 1
Variable means and summary statisticsa

Variable Code All Income groupb Region

countries

High Medium Low Africa Asia Eastern Latin Middle OECD

Europe America East

Totals

L. Gangadharan, M.R. Valenzuela / Ecological Economics 36 (2001) 513–531

Number of Total 51 23 16 12 3 9 8 7 2 22
countries number
Share (%) 100 45 31 24 6 18 16 14 4 43
Size of Total 4326 875 706 2745 83 2685 260 374 122 802
population number
(in
millions)
Share (%) 100 20 16 63 2 62 6 9 3 19
Means
Population POPDEN 206 354 65 112 53.33 751.11 89.88 30.00 50.00 117
density
(number
per km2)
Urban UPOP 41 76 65 48 11.03 100.74 23.28 41.81 31.90 74
population
(%)
GNP GNP 12 204 20151 7514 3228 3457 10 252 5579 7830 4110 18 733
purchasing
power
parity p.c.
Gini GINI 36 32 43 37 52 39 27 53 38 32
coefficient
Life LEXP 72 77 71 66 61 70 70 71 68 76
expectancy
(years)
Healthy life HLE 65 71 62 58 42 63 65 64 60 71
expectancy
(years,
1997–1999)
Physicians/ DOC 202 257 157 19 26 72 309 147 126 263
100 000
(1993)
Table 1 (Continued)

Variable Code All Income groupb Region

countries

High Medium Low Africa Asia Eastern Latin Middle OECD

Europe America East

Child IMM 87 88 87 84 57 91 93 84 95 87
immunisatio

L. Gangadharan, M.R. Valenzuela / Ecological Economics 36 (2001) 513–531

n rate (%
of all
under 12
months)
Gross EDU 86 109 71 61 48 69 84 56 72 109
enrollment
ratio (% of
school-age
childrenc)
Infant IMR 19 6 22 40 59 27 13 26 44 8
mortality
rate (per
1000 live
births)
Child CMR 24 7 26 53 89 34 17 31 52 9
mortality
rate (per
1000
children
under age
5)
Carbon CO2 6.99 10.07 5.61 2.92 2.93 5.88 7.48 3.49 2.95 9.30
dioxide
emission
p.c.c
Total TSP 1.997 0.394 0.992 6.411 0.834 7.931 1.061 1.118 1.143 0.426
suspended
particles
emissionc
Sulfur SO2 0.477 0.160 0.411 1.174 0.159 1.377 0.371 0.279 1.173 0.190
dioxide
emissionc

519
 520
Table 1 (Continued)

L. Gangadharan, M.R. Valenzuela / Ecological Economics 36 (2001) 513–531

Variable Code All Income groupb Region
countries

High Medium Low Africa Asia Eastern Latin Middle OECD

Europe America East

Nitrogen NOx 0.693 0.534 0.558 1.176 0.289 1.417 0.456 0.687 0.00 0.603
dioxide
emissionc
Emission of EMW 430 382 217 800 148 1092 217 216 150 336
organic
water
pollutantsc
Commercial ENPC 2850 4437 1940 1020 869 2067 2852 1169 985 4144
energy use
p.c.c
Deforestation DEFRTE 0.19 −0.36 0.53 0.80 0.60 1.13 −0.09 0.76 0.90 −0.39
rate
(average%
change,
1990–1995)

a
Data were derived from World Development Indicators, WDI (1998), except for HLE and DOC which were taken from the World Health Organisation. All
variables are for year 1996 unless otherwise indicated. Gini coefficients for a few countries (missing in the WDI) were obtained from Deininger and Squire (1996).
b
Denote x as the per capita income. This grouping thus classifies countries in the high income category if (x]$12 000), medium if ($45005xB$12 000) and low
if (xB$4500).
c
EDU refers to secondary level (1995), CO2 expressed in metric tons (1995), TSP, SO2 and NOx in kg/m3 (1995), EMW in kg/day (1993), and ENPC in kg of oil
equivalent (1995).
 L. Gangadharan, M.R. Valenzuela / Ecological Economics 36 (2001) 513–531 521

try’s population density levels (POPDEN) and the disability adjusted life expectancy (DALE) and it
overall degree of urbanisation (UPOP). Secondary summarises the expected number of years to be
school enrollment ratios (EDU) proxy for the lived in what might be termed the equivalent of
literacy rate in the country, and the gap between ‘full health’.7
the rich and the poor as measured by the Gini For all the countries in the sample, the average
coefficient (GINI) is used as the indicator of rate of infant mortality is 19 deaths per 1000 live
income inequality in the country. births. However, this rate shoots up to 59 deaths
The main health status indicators used are life per 1000 live births in Africa and 44 deaths per
expectancy (LEXP) and infant mortality rates 1000 births in the Middle East. The African rate
(IMR). Life expectancy is a popular indicator of is extremely high, and far exceeds the average
health although it is not without problems. levels computed for all the other regions — more
Feachem et al. (1992) show that the causes of than double the rate for Asia and Latin America
death in adults are much less likely to decrease and seven times higher than that of the advanced
with increases in per capita income; it may, in countries of Europe and North America. Income
fact, increase. For example, many adult deaths inequality, measured using Gini coefficients, was
could be due to motor vehicle accidents, use of supplemented from Deininger and Squire (1996).
tobacco and alcohol, excessive consumption of The data show that income inequality is highest in
food products related to heart disease, and all Latin America and Africa, and lowest in Europe
these tend to rise with income. Infant mortality is and North America.8
a good alternative indicator as it avoids the po- Table 2 presents the correlation between the
tentially more severe reverse causation problems environmental indicators relating to air pollution.
associated with the relationship between adult CO2 emissions and ENPC are noted to have a
health and income growth. The mortality rate of very high positive correlation, while ENPC and
children under 5 years of age is usually used as an CO2 emissions have a weak negative relationship
indicator of child well-being (UNICEF, 1991 and with the emission levels of SO2, TSP and NOx,
1992). This welfare measure, which we refer to as which are local pollutants. This may be a bit
child mortality rate (CMR), is used as another
Table 2
main health indicator here.6 To account for ill
Correlation between environmental indicators (air pollution)
health in life expectancy, we use a new variable
developed by the World Health Organisation CO2 ENPC TSP SO2 NOx
called healthy life expectancy (HLE). To calculate
HLE, the years of ill health are weighted accord- CO2 1.0
ENPC 0.8909 1.0
ing to severity and subtracted from the expected TSP −0.2871 −0.3193 1.0
overall life expectancy to give the equivalent years SO2 −0.2176 −0.2648 0.8905 1.0
of healthy life. This indicator is also known as the NOx −0.1060 −0.1451 0.8216 0.8677 1.0

6
There exist alternative data sources for the variables used
in this paper. For example, the World Resources Institute
7
(WRI) provides data on the environment, health and the We would like to thank an anonymous referee for bringing
economy (http://www.wri.org/wri/facts/data-tables.html). this to our attention.
8
WRI data on air pollution is very similar to the kind we use as The Gini coefficient does have some limitations as a mea-
it is available for a few years, but not over time. The data on sure of income inequality. Torras and Boyce (1998), Magnani
water pollution is of better quality in our current source (i.e. (2000) discuss some of these limitations, and it is sometimes
World Development Indicators, 1998) compared with that of suggested that the ratio of the income shares of the first and
the WRI. Further, WRI data for water pollution is for differ- fourth quintile of the income distribution might be a better
ent years for different countries, with a time span difference of measure of income inequality. In this paper, however, the use
up to 13 years in some cases. Such disparate data would not of a general measure such as the Gini coefficient is sufficient,
lend itself to cross-country comparisons, as we require in this as the focus of this paper is not on inequality but on impact of
study. environmental stress on health outcomes.
522 L. Gangadharan, M.R. Valenzuela / Ecological Economics 36 (2001) 513–531

coefficients are jointly significant.

Fig. 1 shows how CO2 emissions and ENPC
levels change with increasing per capita income. At
income levels below $5000, environmental stress is
seen to increase with energy use increasing much
more than carbon dioxide emissions. These emis-
sion levels peak at around the $6000 mark, after
which energy use plateaus while CO2 emissions
decrease until income levels reach $18 000. After
this critical income point, rapid rates of increase in
Fig. 1. Estimated relationship between GNP and CO2, ENPC
level. energy use and CO2 emission levels are observed.
These results contrast with the standard EKC curve
surprising, as we would expect that a rise in energy in that we do not find an inverted-U curve.10
consumption would be accompanied by a rise in Rather, we find that the curve is a flattened
pollutant emissions. Suri and Chapman (1998) inverse-S shaped curve where the slope is mostly
explain this seemingly inconsistent result by sug- positive everywhere, except for the inflection point
gesting that it is possible for energy consumption where the slope is zero. The inverse-S shape is
to keep rising but for emission levels of local observed for both CO2 and ENPC variables, with
pollutants to fall, as would be the case when ENPC levels showing larger rates of change over
end-of-pipe technology like scrubbers are used to the income scale, that is, the CO2 curve is flatter and
reduce local pollutants. As the existing policies to less variable.
abate local pollution often concentrate on end-of- The results imply that we can partition the
pipe methods and not on reducing energy consump- environmental stress experience of countries into
tion or emission levels, it should not surprising that distinct phases. During the first phase when per
energy use and CO2 emissions are not being re- capita incomes are low, environmental stress is
duced along with reductions in the levels of local shown to increase but at a diminishing rate. During
pollutants. the second phase when per capita incomes are
higher, environmental stress levels appear con-
trolled and no increases are observed. The third
5. Results phase occurs at extremely high incomes when
emissions increase again and escalate rapidly. This
Table 3 presents results from the estimation of implies that the impact of income on the environ-
the environmental equation (Eq. (2)). It is seen that ment is more significant at the extreme ends of the
per capita income, population density, country’s income scale. In particular, the results show that
level of urbanisation, inequality in the distribution very low and very high-income countries tend to
of income as well as level of education exert experience increasing stress levels in their environ-
significant influences on a country’s level of envi- mental conditions, while there is relatively little
ronmental stress. The results are particularly strong change in the environmental stress levels for the
when CO2 emissions and ENPC are used as the middle-income countries.
dependent variables. In the case of CO2, we find The results further show that population density
that when all other influences are taken to be level and levels of urbanisation are both positively
constant, a $1000 increase in per capita GNP related to environmental stress while the level of
increases the per capita CO2 emission level by 1 income inequality is inversely related to environ-
metric ton. For ENPC, a 255 kg in oil equivalent mental quality. Hence, as a country gets more
increase in energy use results from a $1000 increase
in per capita GNP.9 The F-test shows that the 10
We do not find an inverted-U shaped relationship be-
tween income and any of the pollutants. In the literature, the
9
The coefficients do not come out to be statistically signifi- EKC relationship is supported for pollutants like SO2 and
cant for ENPC, however, the sign of the income coefficients NOx, but not always for CO2, which is usually seen to increase
are very similar to the signs of the income coefficients for CO2 over income.
emissions.
 L. Gangadharan, M.R. Valenzuela / Ecological Economics 36 (2001) 513–531
Table 3
Impact of GNP and other explanatory variables on the environmenta

Variable CO2 ENPC TSP SO2 NOx EMW DEFRTE

Constant −0.895 −233.047 13.760 1.034 1.513 844416.1 0.485, 0.651

(3.401) (897.500) (7.941)b (1.011) (1.519) (697726.6)
GNP 0.001 0.255 −0.002 −1.93E-04 −1.85E-04 −129.00 −4.05E-05
(0.001) c (0.208) (0.001) (2.40E-04) (2.61E-04) (201.261) (2.33E-04)
GNP2 −1.12E-07 −2.11E-05 1.21E-07 6.91E-09 1.07E-08 0.002 −9.46E-10,
(5.62E-08)c (1.88E-05) (1.11E-07) (1.70E-08) (2.01E-08) (0.016) 1.52E-08
GNP3 3.02E-12 6.53E-10 −2.41E-12 −5.38E-14 −1.65E-13 1.13E-07 6.71E-14,
(1.47E-12)c (4.81E-10) (2.38E-12) (3.60E-13) (4.65E-13) (3.68E-07) 3.12E-13
POPDEN 0.001 0.199 0.001 9.84E-05 −7.21E-05 −152.317 −5.45E-05,
(0.001)b (0.213) (0.001) (1.43E-04) (1.82E-04) (143.116) 1.20E-04
UPOP 5.034 1989.433 −9.916 −1.421 −1.016 −1010368 0.201, 1.288
(2.869)b (795.405)b (7.185) (1.736) (1.956) (1152394)
GINI −0.090 −36.024 0.020 0.017 0.011 11058.34 0.018, 0.015
(0.059) (14.988)b (0.096) (0.021) (0.024) (14480.17)
EDU 0.034 18.411 0.012 0.010 0.002 6086.675 −0.008, 0.008
(0.034) (7.757)b (0.046) (0.011) (0.016) (7135.055)
F-test 4.44 6.96 0.71 0.83 0.53 1.08 2.45

a
Figures in parenthesis indicate robust S.E.
b
Significant at 10% level.
c
Significant at 5% level.

523
524 L. Gangadharan, M.R. Valenzuela / Ecological Economics 36 (2001) 513–531

crowded (more people on a fixed area of land), the explanatory power of these two models is fairly
higher will be their CO2 emissions and per capita high (adjusted R 2 = 67% for CO2 and 79% for
energy use. This can be due to the fact that as ENPC).
population density increases, there is increasing Eq. (2) was also estimated using data on other
pressure to use the existing land more intensively. specific pollutants such as TSP, SO2, NOx, EMW
The creation of multi-storey residential and com- and DEFRTE. The magnitude and signs of the
mercial buildings in high population density coun- estimated coefficients are very sensitive to the
tries is a good example of this problem. Lifestyle pollutant used, and are very unstable. Further, the
adjustments for residents in these countries imply explanatory power of the models are greatly re-
more energy consumption and this leads to abnor- duced with F-test results simultaneously indicating
mally high levels of CO2 emissions. Singapore is a inappropriate models. We note that many environ-
case in point; its population density in 1996 was mental studies used CO2 and ENPC precisely
4990 persons per km2 and the commercial energy because the data on these variables are well devel-
use was 7162 kg of oil equivalent per capita. In oped. Also, we note here that trend results are
contrast, the corresponding average levels for our similar for CO2 and ENPC because CO2 is a major
sample of 51 countries are 206 and 2850, respec- component of ENPC. As seen in Table 2, these two
tively. Clearly, the high population density in environmental stress variables have a high and
Singapore exerts a major influence on its extremely positive correlation between them.
high level of energy use. Further, the percentage of
population living in urban areas (UPOP) impacts 5.1. Impact on health
positively on the levels of CO2 and ENPC, with
emissions rising as urban population increases. Results of the two-stage least squares (2SLS)
We observe a positive coefficient for the educa- estimation of Eq. (3) are presented in Tables 4 and
tion variable (EDU), which runs counter to expec- 5. In these estimations, we use alternative indicators
tations. The results show that higher levels of of a population’s health status — namely, life
education aggravate, rather than improve, environ- expectancy (LEXP), healthy life expectancy (HLE),
mental conditions. On the other hand, any im- infant mortality rate (IMR) and child mortality
provement in the inequalities between the rich and rate (CMR) — and treat the environmental stress
the poor is found to be detrimental to the environ- variable as endogenous. We find that if we ignore
ment. While counter intuitive in the first instance, the potential endogeniety of the environmental
this makes empirical sense because a move towards variables, the results obtained are inconsistent. The
more equal standards of living implies more people Davidson and MacKinnon (1993) augmented re-
are able to afford the use of electricity, cars and gression test shows that the null hypothesis of an
other luxuries — which leads to increased energy exogenous environmental stress variable is strongly
use.11 For such a cross-section of countries, the rejected for all the alternative types of pollutants.12
Table 6 presents the coefficients of the environmen-
11
This issue has been analysed in greater detail in Torras and tal stress variables for the different health indica-
Boyce (1998), Scruggs (1998) with mixed results. While Torras tors and compares the OLS and the 2SLS estimates.
and Boyce (1998) find that more equitable distributions of
The coefficients obtained from the 2SLS estimation
income and power tend to result in better environmental quality,
Scruggs (1998) shows that equality does not necessarily lead to have signs in the expected direction and the magni-
lower environmental degradation. Magnani (2000) finds that tudes are larger compared with the coefficients
higher levels of income would increase environmental quality from the OLS estimation. This implies that the
provided the negative effect of production of goods and services
on pollution levels is more than counterbalanced by the positive
impact of the environmental stress variable on
effect of growth on the demand for pollution abatement policy.
The demand for environmental quality will be affected by
inequality levels in the country. As the level of per capita income
12
increases above a critical level, income equality becomes posi- The augmented regression is formed by including the
tively correlated with environmental protection; however, be- predicted value of the endogenous right-hand side variable as
yond a certain threshold level of income, the correlation between a function of the all exogenous variables, in a regression of the
income and environmental protection turns negative. original model.
Table 4
Impact of GNP and environment on health: with LEXP and HLE as dependent variablesa

L. Gangadharan, M.R. Valenzuela / Ecological Economics 36 (2001) 513–531

Variable CO2 ENPC TSP SO2 NOx EMW DEFRTE

(A) Dependent 6ariable: life expectancy

Constant 57.228 (3.597)b 53.078 (3.250)b 62.143 (5.048)b 62.728 (6.948)b 65.456 (12.590)b 63.381 (5.336)b 58.002 (3.962)b
Environmental −0.380 (0.156)b −0.002 (0.001)b −0.623 (0.440) −5.144 (5.156) −4.152 (5.439) −4.220E-06 (0.470E-06)c −1.803 (3.430)c
stress variable
GNP 0.001 (9.3E-05)b 0.001 (1.6E-04)c 4.1E-04 (7.4E-05)b 3.3E-04 (1.2E-04)b 4.6E-04 (1.4E-04)b 5.01E-04 (1.1E-04)b 4.0E-04 (8.7E-05)b
IMM 0.064 (0.038)c 0.091 (0.033)b 0.095 (0.062) 0.110 (0.082) 0.056 (0.104) 0.063 (0.069) 0.087 (0.081)
DOC 0.009 (0.004)b 0.013 (0.004)b 0.006 (0.005) 0.005 (0.007) 0.012 (0.009) 0.007 (0.004)c −0.001 (0.017)
EDU 0.012 (0.022) 0.030 (0.023) 0.008 (0.022) 0.016 (0.021) −0.004 (0.040) 0.006 (0.025) 0.002 (0.041)
UPOP 3.846 (2.813) 4.932 (3.301) −5.383 (4.526) −5.541 (5.614) −4.054 (6.700) −3.737 (4.363) 3.664 (4.506)
F-test 23.60 17.27 16.16 9.40 9.43 12.59 21.61

(B) Dependent 6ariable: healthy life expectancy

Constant 40.236 (6.759)b 37.139 (5.767) 44.883 (8.480)b 47.426 (10.820)b 52.610 (20.266)b 47.126 (9.874)b 41.164 (6.873)b
Environmental −0.326 (0.237) − 0.002b (00.001) −0.603 (0.627) −7.256 (7.622) −6.626 (8.987) −5.890E-06c (3.690E-06) 0.320 (4.680)
stress variable
GNP 0.001 (1.357E-04)b 7.393E-04 (1.678E-04)b 4.352E-04 (1.115E-04)b 3.053E-04 (1.807E-04)c 5.029E-04 (2.108E-04)b 5.604E-04 (1.665E-04)b 4.652E-04 (1.564E-04)b
IMM 0.126 (0.070)c 0.146 (0.061)b 0.157 (0.097)c 0.196 (0.138) 0.120 (0.172) 0.146 (0.123) 0.112 (0.099)
DOC 0.022 (0.008)b 0.025 (0.006)b 0.019 (0.009)b 0.017 (0.011) 0.028 (0.016)c 0.020 (0.009)c 0.021 (0.026)
EDU 0.008 (0.039) 0.023 (0.037) 0.004 (0.038) 0.012 (0.042) −0.020 (0.070) −0.004 (0.048) 0.014 (0.039)
UPOP 6.504 (3.639)c 7.176 (3.348)b −2.234 (7.330) −5.738 (8.015) −4.771 (10.689) −3.463 (6.259) 4.662 (4.797)
F-test 18.16 19.50 14.11 7.72 7.14 9.43 17.97

a
Figures in parenthesis indicate robust S.E.
b
Significant at 5% level.
c
Significant at 10% level.

525
 526
Table 5
Impact of GNP and environment on health: with IMR and CMR as dependent variablesa

Variable CO2 ENPC TSP SO2 NOx EMW DEFRTE

(A) Dependent 6ariable: infant mortality rate

Constant 74.065 82.829 60.419 59.308 45.716 56.816 75.416

L. Gangadharan, M.R. Valenzuela / Ecological Economics 36 (2001) 513–531

(11.594)b (13.419) (15.342)b (20.929)b (46.084)b (12.095)b (12.086)b
Environmental stress 0.173 0.004 2.113 16.439 16.079 1.540E-05 12.889
variable (0.619) (0.003) (1.178)c (14.902) (18.806) (8.420E-06)c (10.253)
GNP −0.001 −0.002 −0.001 −0.001 −0.001 −1.270E-03 −0.001
(3.285E-04)b (0.001)b (2.181E-04)b (3.586E-04)c (4.523E-04)b (3.580E-04)b (3.391E-04)b
IMM −0.214 −0.275 −0.348 −0.388 −0.216 −0.283 −0.442
(0.129)c (0.135)b (0.159)b (0.209)c (0.349) (0.141)b (0.256)c
DOC −0.056 −0.066 −0.051 −0.047 −0.073 −0.048 0.003
(0.011)b (0.013)b (0.014)b (0.020)b (0.031)b (0.016)b (0.053)
EDU −0.064 −0.091 −0.038 −0.066 0.011 −0.023 0.034
(0.066) (0.071) (0.064) (0.066) (0.132) (0.074) (0.133)
UPOP −13.266 −17.604 12.711 11.772 11.144 10.838 −24.073
(9.619) (10.654)c (12.768) (16.183) (22.007) (15.122) (16.440)
F-test 22.62 17.98 17.64 10.85 7.64 16.53 13.39
(B) Dependent 6ariable: child mortality rate
Constant 112.428 127.268 89.890 90.029 71.226 82.166 113.820
(21.052)b (23.089)b (27.228)b (36.188)b (71.380) (22.437)c (21.322)b
Environmental stress 0.416 0.007 3.433 24.473 23.110 2.190E-05 17.854
variable (0.851) (0.005)c (1.948)c (22.826) (28.130) (1.250E-05)c (15.862)
GNP −0.001 −0.002 −0.001 −0.001 −0.001 −0.002 −0.001
(4.639E-04)b (0.001)b (3.176E-04)b (0.001) (0.001)b (0.001)b (4.825E-04)c
DPT −0.468 −0.571 0.683 −0.723 −0.467 −0.512 −0.779
(0.223)b (0.227)b (0.299)b (0.363)b (0.547) (0.266)c (0.419)b
DOC −0.071 −0.088 −0.062 −0.057 −0.095 −0.059 0.011
(0.017)b (0.021)b (0.020)b (0.030)c (0.046)b (0.021)b (0.080)
SEC −0.063 −0.111 −0.022 0.068 0.043 −0.010 0.071
(0.095) (0.103) (0.094) (0.095) (0.195) (0.107) (0.208)
UPOP −22.663 −29.605 20.191 15.370 13.224 12.990 −36.789
(13.364)c (15.172)c (19.265) (24.349) (33.370) (21.196) (24.261)
F-test 17.97 13.91 11.99 6.44 5.36 10.61 10.82

a
Figures in parenthesis indicate robust S.E.
b
Significant at 5% level.
c
Significant at 10% level.
 L. Gangadharan, M.R. Valenzuela / Ecological Economics 36 (2001) 513–531
Table 6
Comparative test results from alternative estimation methods useda

Environmental Health indicator

stress variable

LEXP HLE IMR CMR

OLS 2SLS t-Valueb OLS 2SLS t-Valueb OLS 2SLS t-Valueb OLS 2SLS t-Valueb

CO2 −0.16c (0.09) −0.38b (0.16) 6.63c −0.35c (0.20) −0.33 (0.24) 3.51c −0.22 (0.20) 0.17 (0.62) 3.78c −0.08 (0.39) 0.42 (0.85) 4.19c
ENPC −6.02E-04c (3.67E-04) −0.002b (0.001) 6.45c −1.33E-03b (5.75E-04) −0.002b (0.001) 3.43c 3.00E-04 (8.12E-04) 0.004 (0.003) 3.76c 1.50E-03 (1.09E-03) 0.007c (0.005) 4.03c
SO2 0.21 (0.24) −5.14 (5.16) 8.35c 0.42 (0.33) −7.26 (7.62) 3.81c 0.53 (1.11) 16.44 (14.90) 3.56c 2.34E-03 (1.40) 24.47 (22.83) 4.06c
NOx 0.28 (0.19) −4.15 (5.44) 8.48c 0.46 (0.31) −6.63 (8.99) 4.01c 0.31 (0.98) 16.08 (18.81) 3.62c −0.19 (1.35) 23.11 (28.13) 4.13c
TSP 0.02 (0.07) −0.62 (0.44) 8.45c 0.10 (0.10) −0.60 (0.63) 3.74c 0.33 (0.37) 2.11c (1.18) 3.45c 0.34 (0.53) 3.43c (1.95) 3.96c
EMW 6.86E-08 (4.58E-07) −4.22E-06c (2.47E-06) 6.98c 3.45E-07 (5.99E-07) −5.89E-06c (3.69E-06) 3.53c 1.17E-06 (1.53E-06) 1.54E-05c (8.42E-06) 3.56c 1.20E-06 (2.11E-06) 2.19E-05c (1.25E-05) 4.21c
DEFRTE −0.25 (0.44) −1.80c (3.43) 7.46c −0.16 (0.77) 0.32 (4.68) 3.43c 0.03 (2.10) 12.89 (10.25) 3.70c −0.27 (3.08) 17.85 (15.86) 4.29c

a
Figures in parenthesis indicate robust S.E. *, Significant at 10%; **, significant at 5%.
b
t-Value from Davidson–Mackinnon augmented regression test for exogeneity.
c
Indicates the null hypothesis of exogeneity is rejected at 5% significance level.

527
528 L. Gangadharan, M.R. Valenzuela / Ecological Economics 36 (2001) 513–531

health is bigger when we take the endogeniety show that income, immunisation rates, access to
into account. The t-statistic from the Davidson – doctors and urbanisation levels all make large
Mackinnon augmented regression test for exo- positive and significant impacts on both infant
geneity shows that the null hypothesis of and child mortality rates. Only the education
exogeneity is rejected at the 5% significance variable fails to make a significant impact on
level. mortality rates.
Panel A in Table 4 presents the results when We also use log linear models to check for
life expectancy is the health indicator used. In robustness of results. Per capita GNP, purchas-
general, the environmental stress variables have ing power parity, is found to be significant in all
the correct sign (negative) and are significant for cases in improving health. Using log TSP, log
certain pollutants such as CO2, EMW, ENPC SO2, log NOx and log EMW as the environmen-
and DEFRTE.13 GNP is always significant and tal stress variable, it is found that the coeffi-
increases in income levels lead to an increase in cients are negative and significant (for log SO2
life expectancy. The level of immunisation and log NOx) in explaining health outcomes.
(IMM) is seen to increase life expectancy when Hence, the environmental variable is significant
the pollutant used is CO2 or ENPC. Education in explaining changes in health levels in a popu-
level (EDU) does not seem to be significant for lation. The estimated coefficients for log CO2
improving health. The availability of doctors and log ENPC, however, have positive coeffi-
(DOC) as a proportion of the population has a cients and are significant. This is contrary to
significant impact on improving life expectancy, what we would expect. We find, therefore, that
particularly for the CO2, ENPC and EMW pol- in some cases the results could be sensitive to
lutants. When the health indicator used is the functional form used.14
healthy life expectancy (panel B of Table 4), the
results are similar. The availability of doctors
increases healthy life expectancy significantly; 6. Conclusion and further research
the absolute impact (measured by size of the
coefficients) of this variable is also greater with In this paper, we examine the links between
HLE, as the dependent variable. The impact of health status, income and environmental indica-
urbanisation (UPOP) on health is large and sig- tors of a country. We first look at the relation-
nificant particularly for HLE as the dependent ship between environment and income — the
variable. This positive impact reflects the soci- EKC hypothesis. We find that low-income coun-
ety’s benefits from improvements in the provi- tries cannot simply postpone attending to envi-
sion of better waste disposal and sanitation ronmental concerns in the hope that the
facilities, which would come with urbanisation. environment will eventually improve with in-
When infant mortality is taken as the health creased incomes. Health is a significant interven-
indicator (panel A of Table 5), we find that in- ing variable and isolating the impact of
creases in TSP emissions and water pollutant environment on health is very important, partic-
emissions levels lead to significantly high infant ularly in the context of developing countries.
mortality. Coefficients derived from the estima- Our results show that the gains in health ob-
tion of the model using the child mortality rates tained through improved incomes can be
are very similar to the infant mortality results negated to a significant extent if the indirect
(panel B of Table 5). In general, the results effect of income, acting via the environ-

13
The economic relationship between deforestation and
health indicators is a bit ambiguous. As deforestation is an
14
indicator of ecological balance in the country, this could have Results for log linear models are not presented in the
long-term effects on health, however, the impact in the short paper, however these are available from the authors on re-
term is not very clear. quest.
 L. Gangadharan, M.R. Valenzuela / Ecological Economics 36 (2001) 513–531 529

ment, is ignored. This study thus shows that Acknowledgements

policy makers who have chosen to pursue rapid
growth strategies at the expense of the environ- We would like to thank Pete Summers, three
ment are not delivering the full realisable health anonymous referees, seminar participants at the
gains that can be derived from higher incomes. Research School of Pacific and Asian Studies,
Also environmental damage is bound to result Australian National University, the participants
in health problems for the domestic population. at the Conference of Economists, 1999, Latrobe
A less healthy labour force will not be able to University and at the Sixth Biennial Meeting of
increase productivity levels, and hence result in International Society of Ecological Economics,
lesser income for the economy. Addressing Canberra, 2000, for their comments. We are,
however, responsible for all remaining errors.
chronic health problems for the population is
Funding for this research was provided by the
also costly and will divert valuable resources
Faculty Research Grant Scheme, Faculty of
from income generating investment projects.
Economics and Commerce, University of Mel-
Clearly, policies for growth must incorporate
bourne.
appropriate programs for protection of the
country’s natural environment and this does not
have to be at odds with growth and develop- Appendix A. Countries included in the study.
ment targets.
One of the ways this research can be extended 1 Argentina
is to obtain time series data on environmental 2 Australia
indicators and health status for varied countries 3 Austria
along the development spectrum. As we are in- 4 Belgium
terested in different kinds of environmental and 5 Brazil
health indicators, obtaining data for all these 6 Bulgaria
indicators for many years is quite challenging. 7 Canada
Most developing countries do not keep records 8 Chile
of environmental variables, and this hampers 9 China
our objective here. However, with the continued 10 Colombia
improvements in data availability, over time this 11 Croatia
problem will be reduced and research on this 12 Czech Republic
can be encouraged. A second extension is to 13 Denmark
create a single indicator or index that could cap- 14 Ecuador
ture the overall quality of a country’s physical 15 Egypt, Arab Republic
environment. Such an index would be useful for 16 Finland
17 France
analysing environmental issues within a country
18 Germany
and can also provide important insights for
19 Greece
cross-country trends. A third extension would be
20 Hungary
to study the impact of environmental policies on
21 India
income or growth in a country and whether it 22 Indonesia
has an impact on the health outcomes. This 23 Iran, Islamic Republic
could help us in understanding whether the pur- 24 Ireland
suit of pro-environmental policies would have 25 Italy
beneficial or adverse effects on growth itself. 26 Japan
This could be examined using a three-equation 27 Kenya
system with income, environmental stress and 28 Korea, Republic
health being endogenous variables. 29 Malaysia
530 L. Gangadharan, M.R. Valenzuela / Ecological Economics 36 (2001) 513–531

30 Mexico Gerdtham, U., Sogaard, J., Andersson, F., Jonsson, B., 1992.
31 Netherlands An econometric analysis of health care expenditure: a
cross-section study of the OECD countries. J. Health Econ.
32 New Zealand 11, 63 – 84.
33 Norway Grossman, G., 1995. Pollution and growth. In: Goldin, I.,
34 Philippines Winters, L.A. (Eds.), The Economics of Sustainable Devel-
35 Poland opment. OECD, Paris, pp. 19 – 46.
36 Portugal Grossman, G., Kreuger, A.B., 1995. Economic growth and the
37 Romania environment. Q. J. Econ. 112, 353– 377.
Hill, K., King, E., 1992. Women’s education in the third world:
38 Russian Federation an overview. In: King, E., Hill, A.M. (Eds.), Women’s
39 Singapore Education in Developing Countries: Barriers, Benefits and
40 South Africa Policy. John Hopkins University Press for the World Bank,
41 Spain Baltimore, pp. 1 – 50.
42 Sweden Kakwani, N., 1993. Performance in living standards: an inter-
43 Switzerland national comparison. J. Dev. Econ. 41 (2), 307– 336.
Kaufmann, R.K., Davidsdottir, B., Garnham, S., Pauly, P.,
44 Thailand 1998. The determinants of atmospheric SO2 concentrations:
45 Turkey reconsidering the Environmental Kuznets Curve. Ecol.
46 Ukraine Econ. 25, 209– 220.
47 United Kingdom Kuznets, S., 1955. Economic growth and income inequality.
48 United States Am. Econ. Rev. 45 (1), 1 – 28.
49 Venezuela Magnani, E., 2000. The Environmental Kuznets Curve, environ-
mental protection policy and income distribution. Ecol.
50 Ghana Econ. 32 (3), 431– 443.
51 Slovak Republic Mellington, N., Cameron, L., 1999. Female education and child
mortality in Indonesia. Bull. Indonesian Stud. 35 (3), 115–
144.
References Munasinghe, M., 1999. Is environmental degradation an in-
evitable consequence of economic growth: tunneling
Agras, J., Chapman, D., 1999. A dynamic approach to the through the Environmental Kuznets Curve. Ecol. Econ. 29,
Environmental Kuznets Curve hypothesis. Ecol. Econ. 28, 89 – 109.
267– 277. Panayotou, T., 1993. Empirical tests and policy analysis of
Chakrabarti, A., Rao, D.N., 1999. Measuring the impact of environmental degradation at different stages of economic
policy variables on the burden of disease. Mimeo, Jawaharlal development. Working Paper, Technology and Environ-
Nehru University. ment Programme, International Labour Office, Geneva.
Cropper, M.L, Simon, N.B., Alberini, A., Sharma, P.K., 1997. Parpel, F., Pillai, V., 1986. Patterns and determinants of infant
The health effects of air pollution in Delhi, India. Policy mortality in developing nations, 1950– 1975. Demography
Research Working Paper 1860, The World Bank. 23 (4), 525– 542.
Davidson, R., MacKinnon, J.G., 1993. Estimation and Infer- Preston, S., 1980. Mortality declines in less developed countries.
ence in Econometrics. Oxford University Press, New York. In: Easterlin, R. (Ed.), Population and Economic Change in
De Bruyn, S.M., van den Bergh, J.C.J.M., Opschoor, J.B., 1998. Developing Countries. The University of Chicago, Chicago,
Economic growth and emissions: reconsidering the empirical pp. 289– 360.
basis of Environmental Kuznets Curve. Ecol. Econ. 25, Pritchett, L., Summers, L.H., 1996. Wealthier is healthier. J.
161– 175. Hum. Resour. XXXI (4), 841– 868.
Deininger, K., Squire, L., 1996. Measuring Income Inequality: Scruggs, L., 1998. Political and economic inequality and the
A New Data-base. World Bank, Washington, DC., environment. Ecol. Econ. 26 (3), 259– 275.
mimeographed. Selden, T.M., Song, D., 1994. Environmental quality and
Feachem, R., Kjellstrom, F., Murray, C., Over, M., Phillips, M., development: is there a Kuznets Curve for air pollution
1992. The Health of Adults in the Developing World. Oxford emissions. J. Environ. Econ. Manage. 27 (2), 147– 162.
University, Oxford. Subbarao, K., Raney, L., 1995. Social gains from female
Flegg, A.T., 1982. Inequality of income, illiteracy and medical education: a cross-national study. Econ. Dev. Cult. Change
care as determinants of infant mortality in underdeveloped 44 (1), 105– 128.
countries. Popul. Stud. 36 (3), 441–458. Suri, V., Chapman, D., 1998. Economic growth, trade and
Galeotti, M., Lanza, A., 1999. Desperately seeking (Environ- energy: implications for the Environmental Kuznets Curve.
mental) Kuznets. Mimeo, International Energy Agency.
Ecol. Econ. 25, 195– 208.
 L. Gangadharan, M.R. Valenzuela / Ecological Economics 36 (2001) 513–531 531

Torras, M., Boyce, J.K., 1998. Income, inequality and pollu- Oxford University Press, Oxford.
tion: a reassessment of the Environmental Kuznets Curve. World Bank, 1992. World Development Report 1992: Devel-
Ecol. Econ. 25, 147–160. opment and the Environment. Oxford University Press,
UNICEF, 1991 and 1992. The State of the World’s Children. New York.