Where is the Poverty–Environment Nexus?

Where is the Poverty–Environment Nexus?

Evidence from Cambodia, Lao PDR, and Vietnam
SUSMITA DASGUPTA, UWE DEICHMANN,
CRAIG MEISNER and DAVID WHEELER *
The World Bank, Washington, DC, USA

Summary. — This paper investigates the poverty–environment nexus at the provincial and district
levels in Cambodia, Lao PDR, and Vietnam. The analysis focuses on spatial associations between
poverty populations and five environmental problems: deforestation, fragile soils, indoor air pollution,
contaminated water, and outdoor air pollution. The results suggest that the nexus is quite different
in each country. We conclude that the nexus concept can provide a useful catalyst for
country-specific work, but not a general formula for program design. Joint implementation of poverty
and environment strategies may be cost effective for some environmental problems, but independent
implementation may be preferable in many cases as well.
2005 Elsevier Ltd. All rights reserved.

Key words — poverty, pollution, basic needs, deforestation, sanitation, natural resources





1. INTRODUCTION

During the past few years, the publications of
international development agencies have highlighted
the importance of the ‘‘poverty–environment
nexus,’’ a set of mutually reinforcing links
between poverty and environmental damage
(Bojo¨ et al., 2001; Bosch, Hommann, Rubio,
Sadoff, & Travers, 2001; Ekbom & Bojo¨ ,
1999). In this nexus, poverty reduction and environmental
protection are complementary goals.
For environmentalists, the nexus concept has
provided a welcome defense against arguments,
based on the ‘‘Environmental Kuznets Curve,’’
that the early stages of development are
unavoidably marked by conflicts between poverty
reduction and environmental protection. 1
Common profession of belief in the nexus has
also smoothed uneasy relations between environmental
specialists and traditional project
officers in development aid organizations.
Numerous studies have suggested that environmental
damage can have particular significance
for the poor. Recent participatory
poverty assessments, conducted in 14 developing
countries of Asia, Africa, and Latin America,
reveal a common perception by the poor
that environmental quality is an important
determinant of their health, earning capacity,
security, energy supplies, and housing quality
(Brocklesby & Hinshelwood, 2001). Rural studies
often observe that poor people’s economic
dependence on natural resources makes them
particularly vulnerable to environmental degradation
(Ambler, 1999; Cavendish, 1999, 2000;
Kepe, 1999; Reddy & Chakravarty, 1999).
Other studies have assessed the health damage
suffered by poor households that are directly
exposed to pollution of the air, water, and land
(Akbar & Lvovsky, 2000; Bosch et al., 2001;
Brooks & Sethi, 1997; Mink, 1993; Songsore
& McGranahan, 1993; Surjadi, 1993). In addition,
environmental disasters and environment-
related conflicts may have regressive
impacts because the poor are least capable of
coping with their effects (Albla-Betrand, 1993;
Myers & Kent, 1995).
In some cases, poor households themselves
may increase environmental degradation.
* The authors would like to thank Jostein Nygard,
Giovanna Dore, Piet Buys, Kiran Pandey, and Hua
Wang, for their valuable comments and contributions.
This research was supported by the World Bank’s
Environment Department and its East Asia and Pacific
Unit for Environment and Social Development. Final
revision accepted: October 11, 2004.
World Development Vol. 33, No. 4, pp. 617–638, 2005
2005 Elsevier Ltd. All rights reserved
Printed in Great Britain
0305-750X/$ - see front matter
doi:10.1016/j.worlddev.2004.10.003
www.elsevier.com/locate/worlddev






617
Poverty-constrained options may induce the
poor to deplete resources at rates that are incompatible
with long-term sustainability (Holden,
1996). In such cases, degraded resources precipitate
a ‘‘downward spiral,’’ by further reducing
the income of the poor (Cleaver & Schreiber,
1994; Dasgupta & Ma¨ler, 1994; Durning, 1989;
Ekbom & Bojo¨ , 1999; Mink, 1993; Pearce &
Warford, 1993; Prakash, 1997; World Bank,
1992; World Commission on Environment &
Development, 1987). Rapid population growth,
coupled with insufficient means or incentives to
intensify production, may induce overexploitation
of fragile lands on steep hillsides, or invasion
of areas that governments are attempting
to protect for environmental reasons. Again, a
downward spiral can ensue (World Bank, 1992).
The existing literature also suggests that the
strength of poverty–environment linkages may
be affected by factors as diverse as economic
policies, resource prices, local institutions,
property rights, entitlements to natural resources,
and gender relations (Ambler, 1999;
Arnold & Bird, 1999; Barbier, 2000; Dasgupta
& Ma¨ ler, 1994; Dutt & Rao, 1996; Ekbom &
Bojo¨ , 1999; Eskeland & Kong, 1998; Heath &
Binswanger, 1996; Leach & Mearns, 1991;
Roe, 1998). This research suggests that the relative
strength of links between poverty and
environment may be very context specific
(Bucknall, Kraus, & Pillai, 2000; Chomitz,
1999; Ekbom & Bojo¨ , 1999). In a recent theoretical
work, Ezzati, Singer, and Kammen
(2001) have demonstrated the implications for
the Environmental Kuznets Curve (EKC). In
a long-run model that allows for many interactions
between socioeconomic and environmental
variables, they show that the conventional,
U-shaped EKC describes only one of many potential
development paths. 2
What does the empirical evidence suggest
about the actual prevalence and importance
of the poverty–environment nexus and complementary
problems? Here the actual record is
sparse, because the requisite data are often difficult
to obtain in developing countries. In principle,
household-level studies can adequately
test whether environmental problems have a
disproportionate impact on the poor. In practice,
such tests are rare. For example, some
studies have established a link between poverty
and consumption of wood fuel, and at least,
one credible study has established the relationship
between indoor combustion and health
(Ezzati & Kammen, 2001). However, the research
also suggests the importance of intervening
variables such as cooking practices (indoor
vs. outdoor) and fuel choice (e.g., charcoal
emits far fewer fine particles than wood). Children
die of waterborne disease at higher rates in
poor households, but again, research points to
the significance of intervening variables such
as water source quality and mother’s education
(Filmer & Pritchett, 1997; Merick, 1985). Rigorous
empirical studies that combine local-area
environmental variables (deforestation, outdoor
air quality, water quality, soil erosion,
etc.) with standard household surveys are almost
nonexistent. Similarly, very few local-area
studies relate environmental quality to the
number and characteristics of poor households.
In poverty–environment analysis that is relevant
for policy, the spatial dimension is critical
for two reasons: First, most environmental
problems are inherently geographical. In principle,
different environmental problems should be
analyzed at different regional scales. In the case
of pollution, for example, the theoretically
appropriate scale is affected by the dispersal
characteristics of the pollutant and medium:
Particulate pollution from cement mills may
only be dangerous in one urban region; acid
rain from sulfur emissions may damage forests
hundreds of miles from the source; and eutrophication
from fertilizer runoff may affect
ocean fisheries a thousand miles downstream
from the farms that are the source of the problem.
In practice, data constraints often dictate
the choice of the scale. In Cambodia, Vietnam,
and Lao PDR, for example, appropriate data
are relatively plentiful at the provincial level,
scarce at the district level (except in Cambodia),
and practically nonexistent for subdistricts.
Accordingly, this paper focuses on provincial
data, with extensions to the district level for
Cambodia. While we readily acknowledge that
more disaggregated evidence would be desirable,
we believe that even province-level analysis
provides a useful first approximation for
poverty–environment work.
We also recognize the possibility that crosscountry
externalities may introduce elements
of the nexus at the regional level, as we have
noted in the previous paragraph. If much of pollution’s
impact is felt far downwind or downstream,
failure to find a nexus in the local data
may simply mask trans-boundary effects. For
some pollutants (e.g., acid rain and deforestation
from industrial sulfur emissions), this may
indeed be a problem. However, a large body
of evidence indicates that local impacts predominate
for pollutants that most heavily affect the
618 WORLD DEVELOPMENT
poor. The foremost examples are fine particulate
air pollution, which causes the bulk of cardiorespiratory
problems, and fecal coliform
pollution of water, which is the major cause of
intestinal disease. Most particulate air pollution
settles locally; locally generated ‘‘plumes’’ of fecal
coliform pollution generally persist for several
miles downstream, but would present
trans-boundary problems only in border areas.
At the provincial level, a minimum criterion
for potential significance of the nexus is disproportionate
environmental damage in high-poverty
areas. Correlation does not necessarily
imply causality, of course, but the lack of time
series data prevents formal testing of structural
causation models. 3 Nevertheless, some reasonable
inferences from positive correlations are
possible. We can also gain an insight from contrary
cases, in which some of a country’s environmental
problems exhibit no spatial
correlation with poverty across provinces. At
least three inferences are possible: The country’s
poverty–environment nexus may not include
these problems; the government may
already have addressed them effectively; or
their part of the nexus may only be operative
at the district or subdistrict level. Data permitting,
further tests could be run at those levels.
Administrative economics provide the second
rationale for spatially disaggregated analysis.
An appropriate geographic scale must strike
an appropriate balance between the benefits of
decentralization and the associated costs. On
the environment side, for example, effective regulation
requires local inspection of damage
sources (pollution, deforestation, etc.), as well
as more centralized facilities for information
collection, storage, and analysis. Environmental
management is undoubtedly improved by a
knowledge of local conditions, but the marginal
cost of administration rises with distance from
administrative centers, because of deteriorating
transport and communications quality. Generally,
province- or district-level administration
strikes the right balance between headquarters
scale economies and the cost of dispersed monitoring
and enforcement operations.
Similar factors govern the choice of administrative
scale for poverty-alleviation programs.
Headquarters staffing remains important, but
such programs also require local monitoring
information and frequent interaction with clients.
By the same logic, province- or district-level
administration may be preferable to national
or subdistrict administration in many cases.
Why are these factors important for analysis
of the poverty–environment nexus? From a policy
perspective, the nexus is important only if it
has consequences for the allocation and administration
of public resources for alleviation of
poverty and environment problems. If there is
no nexus, then optimal policy should treat the
two sectors as separate, divide the overall budget
between them by some criteria, and use separate
calculations to distribute resources
among provincial or district agencies. If poverty–
environment links are strong, on the other
hand, then optimal policy should treat them
jointly in allocating the overall budget. Some
resources for environmental improvement
should be allocated to poverty alleviation when
poverty significantly increases environmental
degradation, and the converse should hold
when environmental factors significantly increase
poverty.
In this paper, we use newly available data to
test for the existence of the nexus in Cambodia,
Lao PDR, and Vietnam. Our main analytical
tools are georeferenced indicator mapping, correlation,
and regression analysis. 4 We recognize
that systematic empirical work is only
beginning in this area; future work may reveal
that the regional scale of our analysis is too
broad, and that the ‘‘true’’ poverty–environment
nexus is more localized in nature. It is also
entirely possible that different dimensions of
the nexus are best analyzed at different geographical
scales. For example, watersheds
may provide the best spatial units for the study
of relationships linking poverty, soil erosion,
and deforestation. 5 At present, the available
data limit us to the broader exercise that is presented
here. In addition, our focus on the comparative
incidence of poverty–environment
problems leads us to focus the exercise at a
common geographical scale.
The remainder of the paper is organized as
follows: Section 2 provides an overview of the
methods that we employ and relates them to
underlying hypotheses about links between
poverty and the environment. Section 3 introduces
the new regional dataset for Cambodia,
Vietnam, and Lao PDR, with particular attention
to the data-collection process, coverage,
and accuracy. In Section 4, we provide a detailed
illustration of our approach for Cambodia.
Section 5 summarizes comparable
evidence for Lao PDR and Vietnam, 6 Section
6 discusses some causal implications of the results,
and Section 7 concludes the paper.
WHERE IS THE POVERTY–ENVIRONMENT NEXUS? 619
2. MAPPING THE PROBLEMS
(a) Absolute poverty
For each country, our provincial or districtlevel
absolute poverty indices are determined
by the number of inhabitants whose daily consumption
expenditure cannot support a food
intake of more than 2,000 calories, plus minimal
nonfood expenditures. By this measure,
the national incidence of absolute poverty is
36% in Lao PDR, 40% in Cambodia, and
37% in Vietnam. 7 To illustrate, Figure 1 displays
the regional distribution of absolute poverty
in Cambodia.
We vary our use of the absolute poverty measure
according to the nature of each environmental
problem. For example, we employ
both the size of the poverty population and
the poverty head-count ratio (or incidence) in
our regression analyses of the relationships
linking poverty to illness from indoor air pollution
and polluted water. 8, 9 For our analysis of
deforestation, we use the density of the poverty
population per unit area. 10
Even though absolute material poverty provides
a useful indicator, we recognize that no
single measure can capture all the dimensions
associated with a broader concept of poverty.
For an analysis of the poverty–environment
nexus, other relevant factors may include lack
of access to common property resources, poor
health, and low levels of education. For example,
research on China and Indonesia has suggested
that education reduces pollution
damage because better-educated communities
are more willing and able to organize to control
polluters. 11 As better data become available,
more fully specified models should incorporate
such factors. In this study, the consequences of
our reliance on a single poverty index are
ambiguous. Exclusion of poverty dimensions
that are correlated with material poverty may
result in overestimation of its impact in our correlation
and regression analysis. On the other
hand, exclusion of these other dimensions also
suppresses their potentially significant twoway
interactions with environmental variables.
(b) Environmental problems
We consider five critical environmental problems,
two related to natural resource degradation
and three to pollution. The ‘‘Green’’
problems are deforestation and soil degradation,
whereas the ‘‘Brown’’ problems are indoor
air pollution, contaminated water, and
outdoor air pollution.
(i) Deforestation
Deforestation serves as a proxy for the loss of
critical ecosystems and biodiversity, as well as
increased risk of soil erosion in steeply sloped
areas. To test for a poverty–environment nexus
in this context, we map forested areas and rates
of deforestation by province and district. In
Figure 1. Cambodia: total poverty population by district, 2000. Source: World Food Program, 2001.
620 WORLD DEVELOPMENT
areas where significant forests remain, we assess
the spatial correlation of poverty and deforestation
using maps, graphical scatter plots, and
regressions.
For the regression analysis, our two principal
variables are the settlement density of the poor
population and overall population density. 12
By incorporating both, we can simultaneously
test the impact of poor and nonpoor households
on forest clearing. 13 Insignificance of settlement
density for the poverty population in
this regression would certainly weaken the
nexus argument that poor households clear forests
more rapidly than others. We cannot test
the converse proposition (exogenously generated
deforestation increases poverty) until we
have better information about the dependence
of the poor on forest products. Future research
should use local data for a more detailed analysis
of this potential link.
We also test for the impact of commercial
logging by controlling for differences in tree
species. In our three study countries, some area
experts have suggested that deforestation is significantly
faster in areas dominated by evergreens,
which are the preferred species for
commercial loggers.
(ii) Fragile soils
Steep hillsides under intensive cultivation are
particularly vulnerable to erosion and soil degradation
without terracing, and the economic
return to farming steeply sloped areas is generally
lower than the return to cultivating alluvial
soils in river valleys. While these observations
are straightforward, their implications for the
poverty–environment nexus depend on local
possibilities for migration. In regions where
people are relatively free to migrate to areas
with higher expected returns, we would expect
steeply sloped areas to be more sparsely populated
than alluvial plains. If population growth
raises the labor intensity of alluvial farming, we
would expect diminishing returns in the lowlands
to induce uphill movement by farmers.
This movement would be tempered by erosion
and soil depletion in the highlands, with a consequent
drop in the overall marginal productivity
of agricultural labor. The remaining
highland farmers should farm larger plots, on
average, to compensate for poorer soils and
to maintain parity in expected income with lowland
farmers. Damage to highland soils would
be a resource conservation problem for society
as a whole, but would not have a disproportionate
impact on the poorest farmers if they
remained free to migrate.
A very different picture would emerge, however,
if marginalized ethnic groups were isolated
in highland areas by historical patterns
of separation and discrimination. In this case,
population growth and soil degradation in the
highlands might well create a ‘‘poverty trap’’
there. By implication, a potential poverty–environment
nexus exists in regions where poor
households are highly concentrated in steeply
sloped areas.
(iii) Indoor air pollution
Recent research has suggested that indoor air
pollution from wood fuels is a major cause of
respiratory disease in developing countries.
Many households use wood or charcoal in Cambodia,
Lao PDR, and Vietnam, so indoor air
pollution may be a significant health problem.
Although indoor air monitoring data are not
yet available in the region, household surveys
have recorded the use of wood and charcoal.
We use regression analysis to test whether poverty
and wood fuel use are significantly associated,
after controlling for area population. We
recognize that our results can only be suggestive,
since the impact of wood fuel use depends on
whether burning is indoors or outdoors. Gauging
the true magnitude of the problem will require
household-level pollution monitoring and
health assessment. This should be an important
topic for future research in our focal countries.
(iv) Access to clean water and sanitation
Safe water and adequate sanitation are critical
determinants of health status, particularly
for children. Ingestion of coliform bacteria
from contaminated drinking water or food is
a prime cause of diarrheal disease, which is in
turn a major cause of infant mortality in developing
countries. Although data for Southeast
Asia remain limited, we use the available information
to assess the spatial relationships linking
poverty, sanitation, and diarrheal disease.
At present, many households in the three countries
do not have access to safe water or sanitation.
A poverty–environment nexus exists if the
affected households are disproportionately
poor. We use maps, scatter plots, and regressions
to test for this possibility.
(v) Outdoor air pollution
Outdoor air pollution is primarily an urban
phenomenon, whose severity depends on the
WHERE IS THE POVERTY–ENVIRONMENT NEXUS? 621
scale of polluting activities, their pollution
intensity (or pollution per unit of output), and
the characteristics of the urban air shed. Recent
research has established that exposure to fine
particulates (with diameters of 10 microns
(PM10) or less) is the main cause of pollutionrelated
respiratory disease. Until recently, little
was known about fine-particulate pollution in
Southeast Asian cities. During the past year,
however, the World Health Organization and
the World Bank have used a large international
database to develop a prediction model for
PM10 pollution, based on urban population, income,
fuel use, and local atmospheric characteristics
(wind, rainfall, temperature, altitude,
etc.). 14
We measure the health impact of PM10 pollution
using a standard ‘‘dose–response’’ function,
based on population exposure studies,
that relates the incidence of mortality to the
airborne concentration of PM10. 15 For each
urban area in Cambodia, Lao PDR, and Vietnam,
we use the World Bank/WHO model to
predict the airborne concentration of PM10,
transform the concentration to an estimated
mortality incidence using the dose–response
function, and multiply the predicted incidence
by the urban population to obtain estimated
deaths from air pollution. We cannot adjust
specifically for the impact of air pollution on
the poor, because the existing dose–response
functions do not distinguish between income
groups. It seems likely that our approach
underestimates health damage to the poor,
who are generally unable to afford health care
that can (at least partially) compensate for the
effect of air pollution.
We would, of course, prefer to base our estimates
on actual monitoring data. However, to
our knowledge, previous environmental studies
have not even attempted to estimate air pollution
for cities in the region. We therefore offer
these estimates as a suggestive benchmark for
further research. Aggregation of the urban-area
results to the provincial level enables us to test
for a poverty–environment nexus by assessing
the spatial correlation between poverty and
health damage from outdoor air pollution.
3. REGIONAL DATASETS
FOR CAMBODIA, VIETNAM,
AND LAO PDR
Work on the poverty–environment nexus requires
spatially integrated information, but the
underlying problems have traditionally received
separate treatment. The spatial dimension of
poverty has probably received the most attention.
In our three focal countries, for example,
poverty mapping has been supported by the
World Food Program and the World Bank in
Cambodia, the International Food Policy Research
Institute (IFPRI) in Vietnam, and the
World Bank in Lao PDR.
Support for spatially oriented work on the
environment has been scattered among agencies
that have interests in different topics. The
Mekong River Commission has supported
work on deforestation, terrain slopes, erosion,
and water quality in our three focal countries.
As previously mentioned, the World Health
Organization and the World Bank have estimated
air pollution for cities with over
100,000 in population as part of an international
program.
Various household survey exercises have provided
critical information for the analysis of
health damage from water pollution and indoor
air pollution. For Cambodia and Vietnam,
USAID has sponsored regionally coded Demographic
and Health Surveys (DHS) that include
data on intestinal illness, respiratory disease related
to air pollution, and the use of biomass
fuels that are major determinants of indoor
air pollution. Population census data for Cambodia
provide regionally coded information on
households’ access to safe drinking water and
sanitation.
When this study began, these critical information
sources were scattered among international
agencies, local development research institutes,
and government ministries that had sponsored
or participated in the work. Accessing the data
involved several trips to the three countries by
a team of World Bank staff members and consultants,
including the authors. The World
Bank’s Spatial Analysis Team incorporated
the information into the spatially integrated
dataset that has been used for this paper.
Most topical components of the data (e.g.,
poverty incidence, intestinal illness from water
pollution) reflect sophisticated survey and mapping
work by experienced international teams.
National population census surveys have also
employed standard sampling methods. For this
analysis, the major problem has not been the
quality of the underlying data, but rather the
appropriate level of spatial integration. Some
geophysical data (e.g., forest coverage, terrain
maps) have very high levels of spatial disaggregation,
while the coverage of some household
622 WORLD DEVELOPMENT
survey data (e.g., the DHS survey for Vietnam)
is only sufficient to support reliable estimation
at the provincial level. To maintain an integrated
view, we have analyzed the data at the
greatest common level of disaggregation available
in each country. Accordingly, we have
analyzed the Cambodian data at the district
(subprovincial) level, while limiting the exercise
to the provincial level in Lao PDR and Vietnam.
4. EVIDENCE FOR CAMBODIA
To provide the most disaggregated view of
the evidence, we begin with the district-level
analysis for Cambodia. Figure 1 provides the
best available map of Cambodia’s poverty population
at the district level: Poor households are
concentrated along an axis that runs northwest
to the border with Thailand. Figure 2, which
displays variations in the density of the poverty
Figure 2. Cambodia: settlement density of the poverty population. Source: World Food Program, 2001.
Figure 3. Cambodia: forested area, 1997. Source: Mekong River Commission (MRC), 2001.
WHERE IS THE POVERTY–ENVIRONMENT NEXUS? 623
population, suggests that provision of services
to the poor would have the lowest unit cost in
the southeastern part of the axis.
Figures 3 and 4 provide maps of Cambodia’s
forest cover and rate of deforestation for the
period 1993–97. Figure 4 suggests that deforestation
is a major problem at the margin of the
central population axis; many contiguous districts
have very high deforestation, and many
areas, one district removed, also have high
rates. The other region with rapid deforestation
is the sparsely populated northeast. For the
country as a whole, a comparison of Figures
1 and 4 suggests that priority areas for poverty
alleviation and forest protection are weakly related
because many of the core poverty areas
are already deforested. The scatter diagram in
Figure 5 confirms this, showing a nearly random
relationship with a rank correlation of
0.15.
Our regression results (Table 1) suggest that
overall population pressure is a major determinant
of deforestation in Cambodia. However,
the results also suggest that forest clearing by
poor people is neither more nor less intensive
than forest clearing by the general population.
In the Cambodian regressions, introduction of
explicit controls for species yields no higher
deforestation rate for evergreens, which are reputed
to be more lucrative for loggers.
Figure 6 uses the incidence of steeply sloped
lands to map the potential for erosion and soil
Figure 4. Cambodia: deforestation rate, 1993–97. Source: Mekong River Commission (MRC), 2001.
0
20
40
60
80
100
120
140
0 20 40 60 80 100 120 140 160 180 200
Number of poor (rank)
Deforestation rate (rank)
ρ = 0.15
Figure 5. Cambodia: rank scatter: deforestation rate versus poverty population.
624 WORLD DEVELOPMENT
depletion in Cambodia. Distinct highland areas
are visible in the northeast, southeast, and particularly
the southwest regions of the county.
The country’s central population axis, on the
other hand, is effectively defined by the lowlands.
Regions with intermediate topography
are intermediate in settlement as well.
Comparison of Figures 1 and 6 suggests a
negative relationship between settlement by
the poor and steeply sloped land: Poor people
are heavily concentrated in lowland areas and
reside at a much lower density in highland
areas. The map shows little evidence of large
poverty populations in steeply sloped areas,
suggesting relatively few cases of inability to
migrate because of ethnic segmentation and
discrimination. The scatter in Figure 7 confirms
the negative relationship between poverty and
steeply sloped land (simple correlation coefficient:
0.29), and is consistent with a model
of relatively free migration in Cambodia.
Figure 8 displays the scatter plot of districtlevel
poverty population versus population
using fuel wood or charcoal. Obviously, the
relationship is very close (the correlation coefficient
is 0.70, with much of the remaining variation
explained by the plot’s obvious separation
into two separate sets of points). However, the
existence of a true poverty–environment nexus
in this context depends on more intensive use
of charcoal and wood fuel by poor households.
The results in Table 2 do, in fact, indicate a
strong association between poverty and wood
fuel use after we control for general population
effects.
Figure 9 suggests a close spatial correlation
between poverty and lack of access to clean
water. Regression analysis (Table 3) also suggests
that poor households have much less access
to safe water than higher-income households in
Cambodia. The implications for child mortality
are suggested by Figure 10, which displays the
regional distribution of childhood deaths in
Cambodia. Again, the spatial correlation with
the poverty population is evident.
Figure 6. Cambodia: percent of land that is steeply sloped. Source: Mekong River Commission (MRC), 2001.
Table 1. Cambodia: population, poverty and
deforestation
Variable Model 1 Model 2 Model 3
Log(Poor/Forest
cover 93)
0.007 0.007
Log(Population/
Forest cover 93)
0.010 0.011 0.018**
Evergreen 0.052* 0.018 0.020
Deciduous 0.036
Mixed 0.062**
Constant 0.014 0.030 0.039
N 369 369 369
R2 0.065 0.056 0.052
Dependent variable: Log(Forest cover 1997/Forest cover
1993). Evergreen, deciduous, and mixed forest dummy
variables.
* Significant at the 10% level.
** Significant at the 5% level.
WHERE IS THE POVERTY–ENVIRONMENT NEXUS? 625
Using the WHO/World Bank model, we project
PM10 pollution levels for urban areas in
Cambodian cities. Figure 11 indicates that estimated
pollution levels are generally higher in
cities located in Cambodia’s population periphery.
Using a standard dose–response model, we
estimate the resulting loss of life and aggregate
the results to the provincial level. Our findings,
displayed in Figure 12, suggest minimal correlation
(0.14) between poverty population and
deaths from air pollution.
Figure 13 summarizes the available evidence
for Cambodia’s poverty population, deforestation,
steeply sloped land, indoor air pollution,
unsafe water, child mortality, and mortality
from outdoor air pollution. The elements of
the matrix are numerically ranked by severity
for ease of comparison. Figure 14 further condenses
the evidence into average rankings for
the first two (‘‘Green’’) indices and the last
three (‘‘Brown’’) indices. The results seem consistent
with a poverty–environment nexus for
Land with slope > 16 % (rank)
= - 0.29
0 20 40 60 80 100 120 140 160 180 200
Number of poor (rank)
0
20
40
60
80
100
120
140
160
ρ
Figure 7. Cambodia: rank scatter: steeply sloped land versus poverty population. Source: Mekong River Commission
(MRC), 2001.
0
20
40
60
80
100
120
140
160
180
200
Pop. using wood & charcoal
for cooking (rank)
= 0.70
0 20 40 60 80 100 120 140 160 180 200
Number of poor (rank)
ρ
Figure 8. Cambodia: rank scatter: fuel wood-using population versus poverty population. Source: Population Census,
1998.
Table 2. Cambodia: population, poverty and use of wood
fuel and charcoal
Variable Model 1 Model 2
Total population 0.843**
Number of poor 0.292**
Log(Total population) 0.971**
Log(Number of poor) 0.013**
Constant 1101.698** 0.141**
N 180 180
R2 0.979 0.994
Dependent variable: Model 1—Population using wood
& charcoal; Model 2—Log(Population using wood &
charcoal).
** Significant at the 5% level.
626 WORLD DEVELOPMENT
indoor air pollution and water contamination.
However, there is no evident relationship between
the spatial distributions of poverty and
deaths from outdoor air pollution. Nor does
there appear to be a significant spatial relationship
between poverty and either of the Green
indices. On the basis of currently available
evidence, we conclude that the regional poverty–
environment nexus in Cambodia is largely
confined to household-level problems associated
with contaminated air and water.
5. EVIDENCE FOR LAO PDR
AND VIETNAM
Similar evidence for Lao PDR in Figure 15
suggests a poverty–environment nexus that is
significantly broader than Cambodia’s. Across
provinces, Figure 15 shows a strong correspondence
between poverty and environmental degradation
in all five categories—deforestation,
erosion potential, indoor air pollution, contaminated
water, and outdoor air pollution. The
association is particularly strong for the lowestand
highest-income provinces. When the environmental
rankings are combined into ‘‘Green’’
and ‘‘Brown’’ indices, the association is clearer
across all provinces. We conclude that the regional
poverty–environment nexus seems very
broad for Lao PDR, so the potential synergy
between poverty alleviation and environmental
policies may be very high. The north- and
northeastern regions of the country appear to
be the main locus for action in this context.
For Vietnam, the evidence in Figures 16–19
suggests a more limited poverty–environment
nexus. The spatial correlation with the poverty
population appears negligible for deforestation,
very weak for sanitation and diarrhea, and negative
for outdoor air population. However, the
large poverty populations in steeply sloped
areas suggest that ethnic separation has opened
potential ‘‘poverty traps.’’ Our evidence also
suggests a relationship between poverty and indoor
air pollution (indexed by cases of acute
respiratory infection).
6. CAUSAL IMPLICATIONS OF
HOUSEHOLD-LEVEL RESULTS
Although the lack of time series data prevents
structural modeling, we believe that
Figure 9. Cambodia: population without access to clean water, 1998. Source: Population Census, 1998.
Table 3. Cambodia: population, poverty and access to
safe water
Variable Model 1 Model 2
Total population 0.241**
Number of poor 1.437**
Log(Total population) 0.587**
Log(Number of poor) 0.186**
Constant 3071.133** 2.398**
N 180 180
R2 0.847 0.664
Dependent variable: Model 1—Population without safe
water; Model 2—Log(Population without safe water).
** Significant at the 5% level.
WHERE IS THE POVERTY–ENVIRONMENT NEXUS? 627
reasonable inferences about causation are possible
for our household-level results. First, we
consider the evidence related to poverty and
households’ use of polluting wood fuels. In
both Cambodia and Lao PDR, we find a
strong, positive relationship between poverty
and use of these fuels. At the same time, we find
a generally weak relationship between poverty
and deforestation. Since use of wood fuels promotes
deforestation, these asymmetric results
suggest that poverty contributes strongly to
household air pollution, but that fuel wood
use (through deforestation) may not contribute
strongly to poverty.
Poverty, lack of access to clean water, and
intestinal disease are also highly correlated in
Cambodia and Lao PDR. In this case, it seems
reasonable to infer two-way causation. Ceteris
paribus, poverty limits access to clean water
and sanitation. At the same time, sanitation-related
diseases exacerbate poverty by reducing
productivity and imposing significant healthcare
costs on affected households. However,
the weaker relationship between poverty and
Figure 10. Cambodia: child deaths, 1998. Source: Population Census, 1998.
Figure 11. Cambodia: urban PM-10 air pollution. Source: World Bank Estimates, 2004.
628 WORLD DEVELOPMENT
sanitation in Vietnam suggests that public
intervention can break this perverse link. The
critical difference may lie in the Vietnamese
government’s relatively high level of investment
in public health and education (particularly for
women). Extensive literature has documented
the significance of these factors in reducing disease
and mortality, even where access to clean
0
5
10
15
20
25
0 5 10 15 20 25 30
Number of poor (rank)
Number of PM-10 deaths (rank)
ρ = 0.14
Figure 12. Cambodia: rank scatter: PM-10 air pollution deaths versus poverty population.
Province Poor Deforest Slope Wood/
Charcoal
Unsafe
Water
Child
Deaths
PM-10
Deaths
Kampong Chaam 1 1 3 1 1 1 3
Siem Reab 1 2 2 2 1 1 1
Prey Veaeng 1 1 4 1 2 1 3
Kampong Thum 1 3 4 2 1 2 2
Baat Dambang 1 3 2 1 1 1 1
Taakaev 1 1 3 1 1 2 4
Kandaal 2 4 4 1 1 1 2
Kampong Spueu 2 2 1 2 2 2 3
Banteay Mean Chey 2 1 3 2 2 1 1
Kampot 2 2 1 2 2 2 4
Kampong Chhnang 2 3 2 3 2 2 3
Svaay Rieng 2 1 4 2 3 3 4
Pousaat 3 4 1 3 2 2 2
Kracheh 3 4 3 3 3 3 1
Preah Vihear 3 3 2 3 3 3 3
Phnom Penh 3 4 4 1 3 3 1
Kaoh Kong 3 2 1 3 4 4 2
Rotanak Kiri 3 2 1 4 3 3 3
Otdar Mean Chey 4 1 3 4 4 4 _
Stueng Traeng 4 3 2 4 4 3 2
Mondol Kiri 4 4 1 4 4 4 4
Krong Preah Sihanouk 4 2 3 3 3 4 1
Krong Kaeb 4 4 4 4 4 4 2
Pailin 4 3 2 4 4 4 4
Note: “ _” denotes no data for that province.
1  1st quartile; 2  2nd quartile; 3  3 rd quartile; 4  4th quartile
Figure 13. Cambodia: poverty population and environmental problems.
WHERE IS THE POVERTY–ENVIRONMENT NEXUS? 629
water is limited (Merick, 1985; Filmer & Pritchett,
1997).
7. SUMMARY AND CONCLUSIONS
In this paper, we have investigated the regional
poverty–environment nexus in Cambodia,
Lao PDR, and Vietnam. Our analysis has focused
on spatial relations between poverty populations
and environmental problems at the
provincial and district levels (see the summary
in Table 4).
We identify a potential poverty–environment
nexus in cases where the pattern of regional
settlement by poor households is strongly associated
with each of five environmental problems:
deforestation, fragile soils, indoor air
pollution, contaminated water, and outdoor
air pollution. Our results suggest that the nexus
is quite different in each country. In Cambodia,
it seems largely confined to household-level
problems associated with indoor air pollution,
contaminated water, and lack of access to adequate
sanitation. Outdoor air pollution, deforestation,
and fragile lands are not significantly
associated with poverty at the district level.
We conclude that poor households in Cambodia
might benefit most strongly from programs
that jointly address poverty and householdlevel
environmental quality. At the same time,
all of Cambodia’s citizens, including the poor,
would benefit from more effective measures to
reduce the rate of deforestation.
Our results suggest a broader poverty–environment
nexus in Lao PDR, since all five
environmental problems exhibit a spatial correlation
with poverty. The overlap is particularly
strong in the northern and northeastern regions
of the country. We conclude that the welfare of
the poor in Lao PDR might be significantly
enhanced by close integration of poverty-allevi-
Province Poor Green Brown
Kampong Chaam 1 3 1
Siem Reab 1 2 1
Prey Veaeng 1 2 1
Kampong Thum 1 4 2
Baat Dambang 1 2 1
Taakaev 1 1 2
Kandaal 2 4 1
Kampong Spueu 2 1 2
Banteay Mean Chey 2 2 2
Kampot 2 1 3
Kampong Chhnang 2 3 3
Svaay Rieng 2 3 2
Pousaat 3 3 2
Kracheh 3 4 3
Preah Vihear 3 3 3
Phnom Penh 3 4 2
Kaoh Kong 3 1 3
Rotanak Kiri 3 1 4
Otdar Mean Chey 4 2 4
Stueng Traeng 4 2 4
Mondol Kiri 4 3 4
Krong Preah Sihanouk 4 2 3
Krong Kaeb 4 4 4
Pailin 4 3 4
Note: Green indicator index (equal weighting): (a) Deforestation
rate, (b) slope greater than 16%;
Brown indicator index (equal weighting): (a) Number using wood and
charcoal, (b) number of cases of diarrhea, (c) number without access
to water and toilets, and (d) number of PM-10 air pollution deaths.
Figure 14. Cambodia: poverty population versus Green/Brown environmental problems.
630 WORLD DEVELOPMENT
ation and environmental strategies in all Green
and Brown dimensions. A geographic focus on
the north would appear to be most beneficial.
The case of Vietnam is more eclectic than the
other two, suggesting the possibility of a poverty–
environment nexus for fragile soils and indoor
air pollution. We conclude that an
appropriate poverty–environment strategy for
Vietnam might focus on the living conditions
of poor households in steeply sloped areas.
In summary, we find little evidence of a general
poverty–environment nexus in our three
study countries. Indoor air pollution is the only
common issue, and its severity depends on
heating and cooking practices that are little
studied as yet. Our evidence suggests that the
nexus concept can provide a useful catalyst
for country-specific work, but not a general formula
for program design. Joint implementation
of poverty and environment strategies may be
cost effective for some environmental problems,
but independent implementation may be preferable
in many cases as well.
We recognize that our analysis is far from
exhaustive, and that other environmental problems
may warrant close attention. Possible candidates
include depleted and polluted fisheries,
and excessive use of pesticides. Future research
should explore these issues more fully. We also
recognize that sub-district-level analysis might
reveal stronger poverty–environment links, as
well as providing a better guide for spatial targeting
of regional programs. For this reason,
we hope that future research projects will promote
more extensive data collection and analysis
at the local level.
In addition, we believe that regional coordination
of poverty–environment programs may
be useful, even in some cases where the poverty–
environment nexus does not appear to be
strong in all countries. A good example is provided
by the links between poverty, access to
safe water, and intestinal disease. As we have
noted, the empirical nexus may be weaker in
Vietnam because government intervention has
already been effective. A regional program linking
Vietnamese experts to counterparts in Cambodia
and Lao PDR could be very appropriate
in these circumstances.
Despite these caveats, we believe that our
findings provide some insights for policy makers
who are concerned about the poverty–environment
nexus. Our results suggest that the
nexus is country specific, and institutional factors
may play an important role. Data on more
countries would be required for an in-depth
Province Poor Deforest Sloped Wood/
Charcoal
Unsafe
Water
Child
Diarrhea
PM-10
Deaths
Savannakhet 1 4 4 1 1 1 1
Champasack 1 3 4 1 1 1 2
Huaphanh 1 2 1 2 1 1 1
Luangphrabang 1 1 1 1 1 1 1
Oudomxay 1 1 2 3 2 2 _
Saravane 2 3 3 2 2 2 2
Khammuane 2 4 3 2 2 1 2
Phongsaly 2 1 1 3 3 4 2
Xiengkhuang 2 2 1 3 2 3 3
Vientiane Municipality 3 1 4 1 4 4 1
Vientiane 3 2 2 2 3 3 4
Luangnamtha 3 1 2 4 3 3 3
Xayabouri 3 2 1 1 1 2 3
Bokeo 4 3 3 4 3 3 _
Attapeu 4 4 3 4 4 2 4
Borikhamxay 4 4 2 3 4 4 3
Sekong 4 3 4 4 4 4 4
Xaysomboon 4 4 4 4 4 4 _
Note: a “_” denotes no data for that province.
Figure 15. Lao PDR: poverty population and environmental problems.
WHERE IS THE POVERTY–ENVIRONMENT NEXUS? 631
analysis of the relationship between a country’s
institutions and the dimensions of its poverty–
environment nexus. However, the available evidence
for our three countries does provide
some useful insights. In Vietnam, the poverty–
environment nexus (indoor air pollution, fragile
lands) is concentrated in northern and western
highland areas dominated by ethnic minority
populations. State policy has focused on outdoor
air pollution much more than indoor air
pollution; priority has been given to reducing
infant mortality through provision of clean
water; and administrators seem to have managed
forest clearing to about the same degree
in poverty and nonpoverty areas.
Several poorer regions of Lao PDR are
inhabited by peoples who are ethnic minorities
in Vietnam, and the administrative and development
resources of the Lao government appear
much more limited than those in
Vietnam. The result of more limited intervention
is apparently a broadening of the nexus,
which replicates the crossborder highland problems
of Vietnam but also includes water pollu-
Province Poor Deforest Sloped
No
Toilets
Child
Diarrhea
Acute
Respiratory
Infections
PM-10
Deaths
Thanh Hoa 1 3 1 1 2 1 1
Nghe An 1 2 1 1 1 1 1
Ha Tay 1 4 4 2 1 1 2
An Giang 1 1 4 1 1 1 1
Dak Lak 1 2 2 1 1 1 2
Bac Giang 1 3 4 3 1 4 3
Nam Dinh 1 _ 3 2 1 1
Son La 1 4 1 2 4 2 4
Can Tho 1 1 _ 4 3 3 1
Dong Thap 1 1 _ 4 2 1 2
Thai Binh 1 _ 4 4 1 3
Kien Giang 1 2 4 1 2 2 2
Binh Dinh 1 3 2 1 2 3 1
Ha Tinh 1 4 3 2 3
Quang Ngai 1 2 2 1 3
Quang Nam 2 1 1 1 2 3 2
Phu Tho 2 3 3 3 3 2 2
Gia Lai 2 2 3 1 3
Hai Duong 2 2 4 4 3 1
Soc Trang 2 1 _ 2 1 2 2
Thua Thien - Hue 2 2 2 1 3 4 2
Vinh Phuc 2 3 4 4 3 2 3
Hai Phong city 2 3 _ 4 2 1 1
Lai Chau 2 2 1 2 1 2 4
Binh Thuan 2 2 3 1 4 4 1
Hoa Binh 2 2 2 3 4
Tien Giang 2 4 _ 4 2 3 2
Ha Giang 2 3 1 2 4
Thai Nguyen 2 3 3 3 3 4 2
Lang Son 2 2 4 2 3 2 4
Note: a blank denotes no data for that province; a “_” for Slope means no land greater
than 16%.
Figure 16. Vietnam: poverty population and environmental problems (top 2 quartile provinces).
632 WORLD DEVELOPMENT
tion, deforestation, and outdoor air pollution.
Cambodia’s public resources and administration
also appear weakly developed, but poverty
among highland ethnic minority groups is less
significant in this society. With no concentration
of poverty in ethnic minority regions,
externality-related environmental problems
such as erosion, deforestation, and outdoor
air pollution appear to be more evenly distributed
across poor and nonpoor areas. However,
development administration is weaker than in
Vietnam, and the government has not yet been
successful in addressing any of the householdlevel
environmental problems that are highly
correlated with poverty (indoor air pollution,
contaminated drinking water, lack of access
to sanitation). In Vietnam, by contrast, a relatively
successful focus on water and sanitation
problems seems to have reduced the provincial
poverty–environment nexus to indoor air pollution
in highland areas where biomass fuels predominate.
Even though our results suggest that the poverty–
environment nexus can differ substantially
Province Poor Deforest Sloped
No
Toilets
Child
Diarrhea
Acute
Respiratory
Infections
PM-10
Deaths
Ben Tre 3 1 _ 4 4
Ha Noi city 3 _ 3 1 1 1
Hung Yen 3 _ 4 4 3 4
Tra Vinh 3 1 _ 1 3 4
Long An 3 4 _ 2 2
Lao Cai 3 3 1 2 3 4 4
Tuyen Quang 3 4 2 3 3 3 4
Quang Binh 3 4 3 3 4
Ca Mau 3 _ 3 4 3 2
Yen Bai 3 3 1 2 3
Quang Ninh 3 4 3 4 1
Bac Ninh 3 _ 4 1 2 4
Lam Dong 3 3 2 2 1 3 2
Ninh Binh 3 4 _ 4 3
Vinh Long 3 1 _ 4 2 3 3
Cao Bang 4 4 2 2 4
Phu Yen 4 1 3 1 3
Khanh Hoa 4 1 1 1 1
Quang Tri 4 3 3 3 2 4 3
Ha Nam 4 4 _ 4 4 2 4
Bac Lieu 4 _ 2 4 4 3
Ninh Thuan 4 3 2 1 3
Ho Chi Minh city 4 4 _ 3 2 1
Dong Nai 4 3 4 2 1
Bac Kan 4 4 2 4 4
Kon Tum 4 2 1 3 4
Binh Phuoc 4 4 _ 3 4 3 4
Tay Ninh 4 1 4 4 3
Da Nang city 4 2 3 3 4 1
Ba Ria - Vung Ta 4 1 4 2 1
Binh Duong 4 1 _ 3 4 4 2
Note: a blank denotes no data for that province; a “_” for Slope means no land greater
than 16%.
Figure 17. Vietnam: poverty population and environmental problems (bottom 2 quartile provinces).
WHERE IS THE POVERTY–ENVIRONMENT NEXUS? 633
across countries, we are only beginning to study
the sources of these differences. We have suggested
some reasons why the nexus seems to
vary in Cambodia, Lao PDR and Vietnam,
but the need for more comparative work is
apparent. Since we have no clear basis for identifying
country-specific limitations at this point,
it seems advisable to begin future country analyses
with a broad consideration of potential
poverty–environment links. Starting with a
Figure 18. Vietnam: steeply sloped land.
Figure 19. Vietnam: poverty population, 1998. Source: IFPRI, 2001.
634 WORLD DEVELOPMENT
Table 4. Summary of country-wide correlation with poverty
PEN indicator Cambodia Lao PDR Vietnam
1. Poverty/deforestation rate Weak +ve correlation at the district level Weak +ve correlation at the
provincial level
Weak +ve correlation
(Forestry data strongly questioned)
(prov.: q = 0.46) (district: q = 0.15)** (prov.: q = 0.37)** (prov.: q = 0.07)*
2. Poverty/fragile land (slope > 16%) Weak ve correlation Weak +ve correlation Weak ve correlation
(prov.: q = 0.29) (district: q = 0.29)** (prov.: q = 0.30)** (prov.: q = 0.10)**
3. Poverty/indoor air pollution
(wood/charcoal use)
Strong +ve correlation, supported by
regression model
Strong +ve correlation, supported
by linear and log regression models
No data on charcoal/wood use,
however moderately correlated
with ARI
(prov.: q = 0.88) (district: q = 0.70)*** (prov.: q = 0.74)*** (prov.: q = 0.51)**
4. Poverty/access to clean water
(CW), no sanitation (NS) & Diarrhea
Strong +ve correlation, supported by
regression model
Strong (Water) and medium (Sani.)
+ve correlations, as well as with
Diarrhea cases
Water (prov.: qcw = 0.93) (district:
qcw = 0.85)***
Water (prov.: qcw = 0.85)*** Weak +ve correlation with toilet
access, stronger with reported
cases of diarrhea
Sanitation (prov.: qns = 0.95) (district:
qns = 0.84)***
Sanitation (prov.: qns = 0.43)** Toilets (prov.: q = 0.23)**
Diarrhea (prov.: q = 0.28)** Diarrhea (prov.: q = 0.75)*** Diarrhea (prov.: q = 0.44)**
5. Poverty/outdoor air pollution
(number of deaths from PM10)
Weak +ve correlation Strong +ve correlation Weak +ve correlation
(prov.: q = 0.14)* (prov.: q = 0.75)** (prov.: q = 0.27)**
Overall poverty–environment nexus (PEN) PEN largely confined to household-level
problems due to contaminated air & water**
Spans most environmental
indices considered***
Weaker compared to Cambodia
and particularly Lao PDR. Data
availability an important issue**
* Low correlation.
** Medium correlation.
*** High correlation.
WHERE IS THE POVERTY–ENVIRONMENT NEXUS? 635
broad ‘‘filter’’ also offers the prospect of more
cost-effective approaches because it facilitates
comparison across problems that are frequently
addressed separately. In our three countries, for
example, simultaneous attention to fragile
lands and indoor air pollution facilitates comparison
of intervention costs and potential benefits
in the two dimensions. 16
NOTES
1. Extensive research has explored the relationship
between environmental degradation and economic
growth. See particularly the special issues of the Journal
of Environment and Development Economics, 2(4), 1997
and Ecological Economics, 25(2), 1998.
2. See Dasgupta, Laplante, Wang, and Wheeler (2002)
for a related discussion of policy impacts on the EKC in
developing countries.
3. We recognize that structural models will be difficult
to estimate reliability for quite some time, since a
relatively long time series would be necessary to distinguish
two-way impacts of poverty and environmental
variables from long trends produced by forces such as
demographic change.
4. Henninger and Hammond (2000) make a strong case
for using poverty–environment maps, which afford
unique insights into the importance of spatial relationships.
5. We are indebted to an anonymous reviewer for this
point.
6. We abbreviate the presentation for Lao PDR and
Vietnam to keep the paper’s length tractable, and
because our analytical methods are identical to those
used for Cambodia. For a full presentation and discussion
of the evidence for Lao PDR and Vietnam, see
Dasgupta, Deichmann, Meisner, and Wheeler (2004).
7. The minimum consumption and poverty estimates
have been produced by the World Bank for Lao PDR,
the World Food Program for Cambodia, and the
International Food Policy Research Institute for Vietnam.
8. See the district-level Model 2 results for Cambodia
in Tables 2 and 3. Our model relates an environmental
problem (H) to the poverty count (P), population (N)
and head-count ratio (P/N) as follows:
logHi ¼ a0 þ a1 log
Pi
Ni

þ a2 log Pi þ a3 logNi
¼ a0 þ ða1 þ a2Þ log Pi þ ða3 a1Þ logNi:
We use the latter expression for estimation.
9. For Lao PDR, the available data have also enabled
us to test the effect of the poverty gap (the difference
between actual income and the absolute poverty line for
a representative individual). For each province, we have
estimated the total poverty gap by summing across gaps
for all individuals who are estimated to fall below the
poverty line. However, we find the correlation of this
variable with the poverty count (.96) to be so high that
the poverty count seems sufficient for our analysis.
10. See section (b) for a further discussion of the
deforestation model.
11. For a further discussion, see Dasgupta and Wheeler
(1997) and Pargal and Wheeler (1996).
12. Our regression analysis is based on a model of
deforestation in which the representative individual in a
region’s population clears a hectares of forest annually.
Forest loss in the region between period 0 and period t is
therefore represented by Ft F0 = aN (a < 0 for deforestation),
where F is the forested area and N is the
regional population. Dividing through by forested area
in period 0 and changing to a logarithmic approximation,
we obtain
F t F 0
F 0
¼
aN
F 0
) log
F t
F 0

¼ b0 þ b1 log
N
F 0

:
To allow for differential poverty effects, we generalize
this expression to
log
F t
F 0

¼ b0 þ b1 log
N
F 0

þ b2 log
P
F 0

þ b3 log X;
where P is the region’s poverty population and X represents
other factors. In this model, b1 reflects the average
area cleared by each resident (poor or nonpoor), and b2
measures the difference (if any) attributable to poverty.
13. We recognize that the estimated impact of settlement
density may be biased by the exclusion of information
on transport costs and other factors that affect
settlement location, income, and deforestation. However,
our test remains useful if the degree of bias is similar for
poor households and households in general. For a further
discussion, see Cropper, Griffiths, and Mani (1999).
14. See Pandey et al. (forthcoming).
636 WORLD DEVELOPMENT
15. For a further discussion of the relationship between
outdoor air pollution and health, see Holgate, Samet,
Koren, and Maynard (1999) and WHO (2000).
16. Our thanks to an anonymous reviewer for stressing
this implication of the work.
REFERENCES
Akbar, S. & Lvovsky, K. (2000). Indoor air pollution:
Energy and health for the poor. ESMAP Newsletter,
No. 1. Washington, DC: World Bank.
Albla-Betrand, J. M. (1993). The political economy
of large natural disasters. Oxford: Clarendon Press.
Ambler, J. (1999). Attacking poverty while improving
the environment: Toward win-win policy options.
Background technical paper prepared for the
September 1999 Forum of Ministers meeting, under
the UNDP-EC Poverty and Environment Initiative;
also in Attacking poverty while improving the environment:
Toward win–win policy options (2001),
Poverty and Environment Initiative, New York:
UNDP.
Arnold, J. E. M., & Bird, P. (1999). Forest and the
poverty–environment nexus. Paper presented at the
UNDP-EC Expert Workshop on Poverty and
the Environment, Brussels, Belgium, January 20–21,
1999; also in Attacking poverty while improving the
environment: Toward win–win policy options (2001),
Poverty and Environment Initiative, New York:
UNDP.
Barbier, E. (2000). The economic linkages between rural
poverty and land degradation: Some evidence from
Africa. Agriculture, Ecosystems and Environment, 82,
355–370.
Bojo¨ , J., Bucknall, J., Hamilton, K., Kishor, N., Kraus,
C., & Pillai, P. (2001). Environment chapter, poverty
reduction strategy papers’ source book. Washington,
DC: World Bank.
Bosch, C., Hommann, K., Rubio, G. M., Sadoff, C., &
Travers, L. (2001). Water, sanitation and poverty
chapter, poverty reduction strategy papers’ source
book. Washington, DC: World Bank.
Brocklesby, M. A., & Hinshelwood, E. (2001). Poverty
and the environment: What the poor say: An
assessment of poverty–environment linkages in participatory
poverty assessments. Swansea, UK: Centre
for Development Studies, University of Wales.
Brooks, N., & Sethi, R. (1997). The distribution of
pollution: Community characteristics and exposure
to air toxins. Journal of Environmental Economics and
Management, 32, 233–250.
Bucknall, J., Kraus, C., & Pillai, P. (2000). Poverty and
the environment, environment strategy background
paper. Washington, DC: World Bank.
Cavendish, W. (1999). poverty, inequality and environmental
resources: Quantitative analysis of rural
households. Working Paper Series, 99–9. Oxford,
UK: Centre for the Studies of African Economies,
University of Oxford.
Cavendish, W. (2000). Empirical regularities in the
poverty–environment relationship of African rural
households. World Development, 28(11), 1979–2003.
Chomitz, K. (1999). Environment–poverty connections
in tropical deforestation. Discussion notes prepared
for the WDR Summer Workshop on Poverty and
Development, Washington, DC, July 6–8, 1999.
Cleaver, K. M., & Schreiber, G. A. (1994). Reversing the
spiral: The population, agriculture, and environment
nexus in sub-Saharan Africa. Washington, DC:
World Bank.
Cropper, M., Griffiths, C., & Mani, M. (1999). Roads,
population pressures, and deforestation in Thailand,
1976–1989. Land Economics, 75(1), 58–73.
Dasgupta, S., Deichmann, U., Meisner, C., & Wheeler,
D. (2004). The poverty–environment nexus in Cambodia,
Lao PDR, and Vietnam.Washington, DC: World
Bank Development Research Group.
Dasgupta, S., Laplante, B., Wang, H., & Wheeler, D.
(2002). Confronting the environmental Kuznets
curve. Journal of Economic Perspectives, 16(1).
Dasgupta, P., & Ma¨ler, K.-G. (1994). Poverty, institutions
and the environment-resource base. World
Bank Environment Paper, No. 9. Washington, DC:
World Bank.
Dasgupta, S.. & Wheeler, D. (1997). Citizen complaints
as environmental indicators: Evidence from China.
World Bank Policy Research Department Working
Paper, No. 1704, January.
Durning, A. B. (1989). Poverty and the environment:
Reversing the downward spiral. Worldwatch Paper,
No. 92. Washington, DC: Worldwatch Institute.
Dutt, A. K., & Rao, J. M. (1996). Growth, distribution
and the environment: Sustainable development in
India. World Development, 24(2), 287–305.
Ecological Economics. (1998). Special issue: The environmental
Kuznets curve. Ecological Economics,
25(2), 143–229.
Ekbom, A., & Bojo¨ , J. (1999). Poverty and environment:
Evidence of links and integration in the country
assistance strategy process. World Bank Africa
Region Discussion Paper, No. 4. Washington, DC:
World Bank.
Eskeland, G. S., & Kong, C. (1998). Protecting the
environment and the poor: A public goods framework
and an application to Indonesia. World Bank
Policy Research Working Paper, No. 1961. Washington,
DC: World Bank.
Ezzati, M., & Kammen, D. (2001). Indoor air pollution
from biomass combustion as a risk factor for acute
respiratory infections in Kenya: An exposure-response
study. Lancet, 358(9281), 619–624.
Ezzati, M., Singer, B., & Kammen, D. (2001). Towards
an integrated framework for development and environment
policy: The dynamics of environmental
Kuznets curves. World Development, 29(8),
1421–1434.
WHERE IS THE POVERTY–ENVIRONMENT NEXUS? 637
Filmer, D., & Pritchett, L. (1997). Child mortality and
public spending on health: How much does money
matter?’’ World Bank Policy Research Department
Working Paper, No. 1864, December.
Heath, J., & Binswanger, H. (1996). Natural resource
degradation effects of poverty and population growth
are largely policy induced: The case of Columbia.
Environment and Development Economics, 1(1), 65–83.
Henninger, N., & Hammond, A. (2000). Environmental
indicators relevant to poverty reduction: A strategy for
the World Bank. Washington, DC: World Resources
Institute.
Holden, S. T. (1996). Adjustment policies, peasant
household resource allocation and deforestation in
Northern Zambia: An overview and some policy
conclusions. In O. Stokke (Ed.), Forum for Development
Studies, No. 1, 1997. Flekkefjord: Norwegian
Institute of International Affairs.
Holgate, S., Samet, J., Koren, H., & Maynard, R. (Eds.).
Air pollution and health. San Diego, CA: Academic
Press.
Journal of Environment and Development Economics.
(1997). Special issue: The environmental Kuznets
curve, Journal of Environment and Development
Economics, 2(4), 357–515.
Kepe, T. (1999). Environmental entitlements in Mkambati:
Livelihoods, social institutions and environmental
change on the wild coast of the Eastern Cape.
Research Report, No. 1, Sussex University, Institute
for Development Studies and PLASS (Program for
Land and Agrarian Studies), Sussex, UK.
Leach, M., & Mearns, R. (1991). Poverty and environment
in developing countries: An overview study.
Institute of Development Studies, Sussex University,
Sussex, UK.
Merick, T. (1985). The effect of piped water on early
childhood mortality in urban Brazil, 1970–1976.
Demography, 22, 1–14.
Mink, S. (1993). Poverty, population and the environment.
World Bank Discussion Paper, No. 189.
Washington, DC: World Bank.
Myers, N., & Kent, J. (1995). Environmental exodus—An
emergent crisis in the global arena. Washington, DC:
Climate Institute.
Pandey, K. D., Bolt, K., Deichmann, U., Hamilton, K.,
Ostro, B., & Wheeler, D. (forthcoming). The human
cost of air pollution: New estimates for developing
countries. World Bank Development Research
Group Working Paper, Washington, DC.
Pargal, S., & Wheeler, D. (1996). Informal regulation of
industrial pollution in developing countries: Evidence
from Indonesia. Journal of Political Economy,
104(6), 1314+.
Pearce, D. W., & Warford, J. J. (1993). World without
end—economics, environment and sustainable
development. New York: Oxford University
Press.
Prakash, S. (1997). Poverty and environment linkages in
mountains and uplands: Reflections on the ‘‘Poverty
Trap’’ thesis. CREED Working Paper Series, No. 12.
London: IIED.
Reddy, S. R. C., & Chakravarty, S. P. (1999). Forest
dependence and income distribution in a subsistence
economy: Evidence from India. World Development,
27(7), 1141–1149.
Roe, E. (1998). Taking complexity seriously: Policy
analysis, triangulation and sustainable development.
Boston: Kluwer Academic Publishers.
Songsore, J., & McGranahan, G. (1993). Environment,
wealth and health: Towards an analysis of intraurban
differentials within the greater Accra Metropolitan
Area, Ghana. Environment and Urbanization,
5(2), 10–34.
Surjadi, C. (1993). Respiratory diseases of mothers and
children and environmental factors among households
in Jakarta. Environment and Urbanization, 5(2),
78–86.
World Bank (1992). World development report 1992—
Development and the environment. New York: Oxford
University Press.
World Commission on Environment and Development
(1987). Our common future, report of the world
commission on environment and development. Oxford:
Oxford University Press.
World Health Organization (WHO) 2000. Guidelines for
Air Quality, WHO, Geneva. Available from <http://
www.who.int/peh/air/Airqualitygd.htm>.
638 WORLD DEVELOPMENT