DETERMINANTS OF A WIN-WIN IN ENVIRONMENTAL AND ECONOMIC
GROWTH TRENDS OF NATIONS: 2000-2010
by
Medina Chehbene Jeff-Zanni
B.Sc. Redeemers’ University, 2014

A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF
THE REQUIREMENTS FOR THE DEGREE OF
Master of Science
in
Environmental Science
Thompson Rivers University
Fall Convocation 2018

Thesis Examining Committee:

Peter Tsigaris (PhD), Thesis Supervisor and Professor, Business and Economics, Thompson
Rivers University
Joel Wood (PhD), Committee Member and Assistant Professor, Business and Economics,
Thompson Rivers University
Terry Kading (PhD), Committee Member and Associate Professor, Philosophy, History and
Politics, Thompson Rivers University
Wendy Gardner (PhD), Committee Member and Associate Professor, Natural Resource
Science, Thompson Rivers University
Jesus Crespo Cuaresma (PhD), External Examiner and Professor, Institute for
Macroeconomics, Vienna University of Economics and Business, Austria
© Medina Chehbene Jeff-Zanni 2018

ii
Supervisor; Peter Tsigaris (PhD)
ABSTRACT
The consequences surrounding the rise in economic growth of a nation at the expense of a
deteriorating environmental quality for any given nation in the long-run could reflect
deterioration of human health, quality of life and a further ecosystem destruction which
reflects a win-loss situation. This research identifies those nations that have had progress in
their economic growth and environmental performance simultaneously over the recent decade
in order to identify the nations in a win-win situation. With the use of the environmental
performance trend data available from Yale University and Gross Domestic Product per
capita from the World Bank database, we provide insights into nations which had a win-win
in economic growth and environmental performance trends simultaneously between 20002010 and which did not. Also, we explore nations which had a high win-win trend using
thresholds and some of the underlying factors that can help to explain differences in
performances across nations. This study employs the K-means clustering technique to
identify the different clusters of nations within win-win or other pre-defined clusters for over
200 nations. The environmental performance is divided into environmental health which
focuses on human health and ecosystem vitality which focuses on the health of the
ecosystem. Within this period, low-income, middle-income and high-income nations had an
overall win-win situation in environmental issues that affect human health especially the
child mortality indicator in comparison to its water and air quality counterparts. Nations had
more of a win-loss situation in environmental issues that affect the ecosystem which connotes
win for the economy but a declining ecosystem indicator. The statistically significant
variables found to impact the likelihood of win-win in the environmental health category
using logistic regressions consists of the initial GDP, initial non-income HDI, average
investment spending and improvements in political stability. The explanatory power of the
independent variables is strongest for win-win in child mortality and economic growth but
not much power to explain for water and air quality situation. On the other hand, having a
higher likelihood of a win-win case for the health of the ecosystem and economic growth
included improved governance effectiveness, initial income-level, average investment
spending and the initial environmental performance level of a nation which all varied by
magnitude in its influence on biodiversity, fisheries, forestry, agriculture, air quality, climate
change and water use indicators.

iii
TABLE OF CONTENTS
Abstract ..................................................................................................................................... ii
Table of contents ...................................................................................................................... iii
Acknowledgements .................................................................................................................... v
Dedication .................................................................................................................................vi
CHAPTER 1. INTRODUCTION .............................................................................................. 1
Environmental Kuznets Curve ............................................................................................... 2
Environmental Performance Index ...................................................................................... 10
Independent Approach ......................................................................................................... 16
Thesis Objectives ................................................................................................................. 17
CHAPTER 2. CLUSTER ANALYSIS EVALUATION OF THE GROUPS OF NATIONS
WITH A WIN-WIN TREND IN ENVIRONMENTAL PERFORMANCES AND GDP. ..... 22
Introduction .......................................................................................................................... 22
Methods ............................................................................................................................... 24
k-means clustering………………………………………………………………………24
Data collection…………………………………………………………………………..25
Results .................................................................................................................................. 26
Environmental Health: Child Mortality trend, Growth and HDI 2000………………....26
Environmental Health: Air Pollution, Growth and HDI 2000………………………….30
Environmental Health: Water Quality trend, Growth and HDI 2000…………………..33
Ecosystem Vitality: Biodiversity and Habitat, Growth and HDI 2000…………………36
Ecosystem Vitality: Agriculture trend, Growth and HDI 2000…………………………39
Ecosystem Vitality: Forestry trend, Growth and HDI 2000…………………………….42
Ecosystem Vitality: Fishery trend, Growth and HDI 2000……………………………..45
Ecosystem Vitality: Change in Water Quantity trend, Growth and HDI 2000…………47
Ecosystem Vitality: Climate Change trend, Growth and HDI 2000……………………50
Ecosystem Vitality: Air Pollution trend, Growth and HDI 2000……………………….53
Discussion ............................................................................................................................ 56
Environmental Health…………………………………………………………………...56
Ecosystem vitality………………………………………………………………………58
Change indicators: Water and Forestry…………………………………………………59
Biodiversity and Habitat………………………………………………………………...60
Concluding Remarks…..…………………………………………..................................60

iv

CHAPTER 3: DETERMINANTS OF HIGH PERFORMANCE NATIONS IN
ENVIRONMENTAL AND ECONOMIC PERFORMANCE USING LOGISTIC
ANALYSIS. ............................................................................................................................. 63
Introduction .......................................................................................................................... 63
Methods ............................................................................................................................... 65
Model specification…………………………………………………………………......65
Variables………………………………………………………………………………...67
Thresholds for trends in environmental indicators……………………………………...69
Results .................................................................................................................................. 72
High win-win in economic growth and environmental health………………………….72
High win-win in economic growth and ecosystem vitality……………………………..75
Comparison with a single high win situation for environmental indicators…………….79
Comparison with a single high win situation for high economic performance…………81
Discussion ............................................................................................................................ 83
CHAPTER 4: CONCLUSION AND IMPLICATIONS FOR FUTURE STUDIES ............... 86
References ................................................................................................................................ 89
Appendices ............................................................................................................................... 93
Appendix A: Indicator frameworks and EPI exploration .................................................... 93
Appendix B: Definitions of indicators for EPI policy categories ...................................... 120
Appendix C: Cluster Analysis ........................................................................................... 122
Appendix D: HDI Classification in 2000 and Codes ......................................................... 125
Appendix E: Logistic results.............................................................................................. 133

v
ACKNOWLEDGEMENTS
To my supervisor, Dr. Peter Tsigaris, for accepting to be my supervisor. I have grown and
learnt so much over the last two years and I could not have achieved this without your
support, encouragement, investment and dedication to me and my work, thank you. To
Professor Wendy Gardner, thank you so much for the guidance and help you rendered in
helping me navigate my path in this program to have a clearer understanding of what was
required of me. To Professor Joel Wood and Professor Terry Kading for their valuable
suggestions and comments which lead to an improvement in my manuscript. Finally, I would
like to thank the external examiner Professor Jesus Crespo Cuaresma for his thoughtful
assessment and contributions to my thesis and great suggestions for further research.

vi
DEDICATION
To my father and mother, who approved and let me to apply to this program which was very
distant from home, thank you for this opportunity to follow my passion. I love you so much
and thank you for supporting me financially, emotionally and spiritually, God bless you
richly. I dedicate this thesis to my Lord God Almighty, who is the source of all my wisdom
and the Author and Finisher of my faith.

vii
LIST OF FIGURES
Figure 1.1: The Environmental Kuznets curve (Author’s drawing)…………………………..3
Figure 1.2: The 2012 EPI index. (Emerson et al., 2012)……………………………………..12
Figure 1.3: Performance benchmark for nations (Hsu et al., 2013). …………………………... ..13
Figure 1.4: Classification of trends in economic growth and environmental performance for
the design of this study. Win-win is the top right quadrant with an additional high win-win
cluster. The loss-win is the bottom right quadrant. The win-loss is the top left quadrant and
the loss-loss is the bottom left quadrant…………………………………………………… ..19
Figure 2.1: The distribution for the 6 clusters of nations in Child Mortality trend and the
Average GDP per capita growth rate in the win-win (upper right), win-loss (bottom right),
loss-win (upper left) and loss-loss (bottom left) categories………………………………….29
Figure 2.2: The distribution for the 6 clusters of nations in Air Pollution (human health) trend
and the Average GDP per capita growth rate in the win-win (upper right), win-loss (bottom
right), loss-win (upper left) and loss-loss (bottom left) categories…………………………..32
Figure 2.3: The distribution for the 6 clusters of nations in Water Quality (human health)
trend and the Average GDP per capita growth rate in the win-win (upper right), win-loss
(bottom right), loss-win (upper left) and loss-loss (bottom left) categories………………….35
Figure 2.4: The distribution for the 6 clusters of nations in Biodiversity and Habitat trend and
the Average GDP per capita growth rate in the win-win (upper right), win-loss (bottom right),
loss-win (upper left) and loss-loss (bottom left) categories………………………………….38
Figure 2.5: The distribution for the 6 clusters of nations in Agriculture trend and the Average
GDP per capita growth rate in the win-win (upper right), win-loss (bottom right), loss-win
(upper left) and loss-loss (bottom left) categories……………………………………………44
Figure 2.6: The distribution for the 6 clusters of nations in Forestry trend and the Average
GDP per capita growth rate in the win-win (upper right), win-loss (bottom right), loss-win
(upper left) and loss-loss (bottom left) categories……………………………………………44
Figure 2.7: The distribution for the 6 clusters of nations in Fisheries trend and the Average
GDP per capita growth rate in the win-win (upper right), win-loss (bottom right), loss-win
(upper left) and loss-loss (bottom left) categories……………………………………………46
Figure 2.8: The distribution for the 6 clusters of nations in Water Quantity (ecosystem) trend
and the Average GDP per capita growth rate in the win-win (upper right), win-loss (bottom
right), loss-win (upper left) and loss-loss (bottom left) categories. …………………………49
Figure 2.9: The distribution for the 6 clusters of nations in Climate Change trend and the
Average GDP per capita growth rate in the win-win (upper right), win-loss (bottom right),
loss-win (upper left) and loss-loss (bottom left) categories. ………………………………...52

viii
Figure 2.10: The distribution for the 6 clusters of nations in Air Pollution (ecosystem) trend
and the Average GDP per capita growth rate in the win-win (upper right), win-loss (bottom
right), loss-win (upper left) and loss-loss (bottom left) categories…………………………..52
Figure 3.1: Thresholds for a high-performance win-win for the Environmental Health
categories and Average GDP per capita growth rate. EHEH_trend is the Child Mortality trend
set at 10, EHAIR_trend is Air Pollution trend set at 0, EHWATER_trend is Water Pollution
trend set at 10 and EH_trend is the general Human health trend set at 0. The Average GDP
growth rate is set at 2 for all the indicators………… ………………………………………..70
Figure 3.2: Thresholds of a high-performance win-win for the Ecosystem Vitality categories
and Average GDP per capita growth. The first column consists of a general Ecosystem
Vitality trend (EV_trend) set at 0, Forestry trend (EVFOREST_trend) set at -10, Air Pollution
trend (EVAIR_trend) set at 0. The second column consists of Biodiversity and Habitat trend
(EVBH_trend) set at 0, Fisheries trend (EVFISH_trend) set at 0, Climate Change trend
(EVCC_trend) set at 0. The third column consists of Agriculture trend (EVAG_trend) set at 0
and Water Quantity use trend (EVWATER_trend) set at -25. The threshold for Average GDP
per capita growth is set at 2 for all the indicators…….............................................................71

ix
LIST OF TABLES
Table 1.1: Factors that influence the environmental quality of nations (Author's interpretation
of Stern (2004, 2017)………………………………………………………………. ………....6
Table 2.1: The mean and p-values for Child Mortality trend and the Average GDP per capita
growth rate for the 6 clusters of nations in the different categories. The HDI 2000 shows the
rank of the nations in percentage for each cluster…….………………………………….…..27
Table 2.2: The mean and p-values for Air Pollution (human health) trend and the Average
GDP per capita growth rate for the 6 clusters of nations in the different categories. The HDI
2000 shows the rank of the nations in percentage for each cluster..........................................35
Table 2.3: The mean and p-values for Water Quality (human health) trend and the Average
GDP per capita growth rate for the 6 clusters of nations in the different categories. The HDI
2000 shows the rank of the nations in percentage for each cluster.….........………………...37
Table 2.4: The mean and p-values for Biodiversity and Habitat trend and the Average GDP
per capita growth rate for the 6 clusters of nations in the different categories. The HDI 2000
shows the rank of the nations in percentage for each cluster……………………………...…27
Table 2.5: The mean and p-values for Agriculture trend and the Average GDP per capita
growth rate for the 6 clusters of nations in the different categories. The HDI 2000 shows the
rank of the nations in percentage for each cluster…………………………………………....41
Table 2.6: The mean and p-values for Forerstry trend and the Average GDP per capita growth
rate for the 6 clusters of nations in the different categories. The HDI 2000 shows the rank of
the nations in percentage for each cluster…………………………………………………....43
Table 2.7: The mean and p-values for Fisheries trend and the Average GDP per capita growth
rate for the 6 clusters of nations in the different categories. The HDI 2000 shows the rank of
the nations in percentage for each cluster………………………………………………...….45
Table 2.8: The mean and p-values for Water Quantity trend and the Average GDP per capita
growth rate for the 6 clusters of nations in the different categories. The HDI 2000 shows the
rank of the nations in percentage for each cluster…………………………………………...47
Table 2.9: The mean and p-values for Climate Change trend and the Average GDP per capita
growth rate for the 6 clusters of nations in the different categories. The HDI 2000 shows the
rank of the nations in percentage for each cluster……………………………………..…….49
Table 2.10: The mean and p-values for Air Pollution (ecosystem) trend and the Average GDP
per capita growth rate for the 6 clusters of nations in the different categories. The HDI 2000
shows the rank of the nations in percentage for each cluster……………………………..….51
Table 3.1: The coefficients followed by p-values and odds ratio (in parentheses) for the
determinants of a likelihood of a high win-win for the environmental health indicators and
Average GDP per capita using logistic regression. Coefficients with statistically significant pvalues are bolded. The environmental health indicators are Child Mortality (EH-EH), Water
Quality (EH-Water) and Air Pollution (EH Air)……………….…………………………….59

x

Table 3.2: The coefficients followed by p-values and odds ratio (in parentheses) for the
determinants of a likelihood of a high win-win for the ecosystem vitality indicators and
Average GDP per capita using logistic regression. Coefficients with statistically significant pvalues are bolded. The ecosystem vitality indicators are Biodiversity and Habitat (EV-BH),
Agriculture (EV-AG), Forestry (EV-FOREST), Fishery (EV_FISH), Water Quantity use
(EV-WATER), Climate Change (EV-CC) and Air Pollution (EV-AIR……….....…….....…65
Table 3.3: The coefficients and p-values for the determinants of a likelihood of a single win
for environmental health ecosystem vitality indicators using logistic regression. Coefficients
with statistically significant p-values are bolded.………………….………..…………….....71
Table 3.4: The coefficients and p-values for the determinants of a likelihood of a single win
for environmental health ecosystem vitality indicators simple linear regression. Coefficients
with statistically significant p-values are bolded for Average GDP (Av GDP) and the
probability of obtaining a win for GDP growth (Prob).……………………………….……..73

1
CHAPTER 1. INTRODUCTION
Researchers have sought to explain and estimate the relationship between economic
growth and environmental performance since the seminal paper by Grossman and Krueger
(1995). Some have attributed the economic growth of a nation as the major factor in
improving its environmental performance over time (Dinda, 2005) while other studies have
found the opposite result (Costantini and Monni, 2008; Stern, 2004). Such studies have
become increasingly prevalent because of the impact that the activities of economic growth
have on the environment and its impacts on the quality of life of people in that environment.
A deteriorating environment is characterized by a lower air quality, less safe drinking
water, diseases, more deaths, higher depletion and scarcity of natural resources and more
unstable climate conditions, all of which have adverse effects on human health and the
ecosystem. These conditions are a result of increased economic activities some of which
include increased land use, mining, fishery, electricity generation, agriculture, and
manufacturing, all of which are geared towards increasing the standard of living on nations.
These resulting environmental issues are some of the consequences of a lack of balance
between economic and environmental policies and regulations.
Citizens have become increasingly aware of their right to a high-quality environment,
from the basics like access to clean water to nature and a green environment (Criado et al.,
2011). It has been of great importance to increase research and add to the current knowledge
in pursuit of progress in environmental sustainability (Gallego et al., 2014). A well-known
pattern is that some nations with very high economic growth often have declining
environmental performances in certain environmental issues like climate change, and forestry
(Hsu, 2016) while some developing nations perform well in certain environmental issues like
agriculture (Hsu et al., 2013).
Can nations achieve growth simultaneously in economic and environmental
performance in the long-run, as opposed to the Environmental Kuznets Curve theory which
posits that the relationship between income per capita and environmental deterioration is
positive for low income and negative for high income nations? Can developing nations
achieve high performances in both the economy and environment? What about rich nations?
Can they only aim at improvements of environmental performance given that their growth
has slowed down due to diminishing returns? There is still room for more research to be done
to add to the current knowledge of the relationship between the growth of economic and
environmental performance in the long run.

2
This study seeks to identify those nations that have made progress in their economic
growth and environmental performance simultaneously over a decade in order to identify
behavioral traits, patterns and other factors that have led to their win-win situation. It will
explore not only win-win situations but identify nations which are clustered in high
performances in both the economy and the environment. In addition, this study will attempt
to determine the factors which increases the likelihood of high performances in both areas
simultaneously relative to the likelihood that it does not lead to high economic growth and
environmental performance.
This study will use the Gross Domestic Product (GDP) per capita growth rate for over
200 nations as a measure of increases in the standard of living of nation and the 2012 Pilot
trend Environmental Performance Index of ten different policy categories compiled by
researchers at Yale University to examine the trends in the performances of nations for the
period of 2000 – 2010. This will reveal clusters of nations that have had a win-win situation,
including nations with high performances in both areas, reflecting improvements in
environmental health and a growing economy and factors that lead to this win-win situation.
It will provide relevant information to policy-makers and the public that will help them make
better choices geared towards improving the quality of life from an economy and
environment.
There is a vast body of literature over the years from studies conducted on economic
growth and environmental performance. A prominent theory which was first studied by
Grossman and Krueger (1995) which is called the Environmental Kuznets Curve (EKC)
hypothesis has been used in several studies. Also, the Environmental Performance Index
(EPI) developed by Yale University is also widely applied in several studies. The remainder
of this chapter will review findings from the studies which implemented the EKC theory, the
EPI Index, other independent approaches and the objectives of this thesis.
Environmental Kuznets Curve
The EKC is hypothesized as following an inverted U-shape for the relationship
between income per capita and environmental degradation (Grossman and Krueger, 1995).

3

Figure 1.1: The Environmental Kuznets curve (Author’s drawing).

As seen in Figure 1.1 above, the EKC hypothesis posits that at very low levels of
income per capita that environmental degradation is low, but as income increases that
environmental degradation gets worse until income reaches a ‘turning point’ at which level of
degradation begins to reduce with increases in the standard of living. This theory has been
used to evaluate different environmental issues and the results have mostly differed for
different environmental issues, in different regions and across time (Yandle et al., 2002).
Grossman and Kreuger used panel data from Global Environmental Monitoring
System (GEMS) on ambient pollution levels of urban air and water quality for several
countries to test the relationship between national income and environmental quality. The
authors estimated several reduced-form equations that relate to the level of pollution in a
location (air and water) to a flexible function of the current and lagged income per capita in
the country and to other covariates. They used the reduced-form estimates approach because
it gives the net effect of a nation’s income on pollution as opposed to the structural equations.

4
The results showed increases in GDP per capita may be associated with deteriorating
environmental quality in very poor countries, air and water quality appear to gain from
increasing GDP to which a critical level has been reached. Also, the turning points for the
different pollutants vary but, in most cases, occur at less than $8000 (1985 dollars). Their
findings are consistent with those in other studies like the World Bank Development Report
(2012) which finds an inverted U-shaped relationship between per capita and income and
concentrations of sulfur dioxide in the air.
Shafik (1994) proposed that at a theoretical level, it is not possible to predict how
environmental quality will evolve with changes in per capita incomes, particularly where
public goods are involved. He suggests that, while there is no inevitable pattern of
environmental transformation with respect to economic growth at an aggregate level, there
are clear relationships between specific environmental indicators and per capita incomes. He
focused on the relationship between environment quality and per capita income, taking into
account these other determinants of environmental quality. Income per capita serves to
measure directly the relationship between economic growth and environmental quality and
measures indirectly the endogenous characteristics of growth. Features such as the impact of
rising industrialization and urbanization at middle-income levels and the growing importance
of services in high income economies are typical patterns that are proxied by per capita
income.
Indicators of environmental quality were used as dependent variables for 149
countries for the period 1960-1990. The environmental quality indicators analysed were the
lack of clean water, lack of urban sanitation, ambient levels of suspended particulate matter
(SPM), ambient sulfur oxides (SO2), change in forest area between 1961-1986, the annual
rate of deforestation between 1962-1986, dissolved oxygen in rivers, fecal coliforms in rivers,
municipal waste per capita, and carbon emissions per capita. To analyse this model, three
basic models were tested which include log linear, quadratic, and cubic in order to explore
the shape of the relationship between income and each environmental indicator.
The results show that access to clean water and urban sanitation are indicators that
improve with higher per capita incomes. Data for deforestation, most of which were not
available, were poor at capturing important differences between types of forest. The
disappointing results for both the change in forest area between 1962-1986 and the annual
rate of deforestation between 1961-1986 showed that none of the income terms were
significant in any specification. This led to the conclusion that per capita income appears to
have very little bearing on the rate of deforestation.

5
Shafik's findings also show that the two measures of river quality tend to worsen with
rising per capita income. He pointed out that the initial worsening of fecal content was
probably associated with growing urbanization and consequent pressures on sanitation,
hence, the improvement results when urban sanitation services are introduced. Suspended
particulate matter (SPM), which causes respiratory illness and mortality, is largely the result
of energy use. For local air pollution, there’s an initial deterioration of environmental quality
as industrialization and energy intensity increases, followed by an improvement as cleaner
technology are used and fuel switching occurs. Technology, proxied by the time trend,
appears to have played a favourable role in making improved local air quality possible at an
earlier stage of development. He suggests that it is possible to solve some environmental
problems but that it is not necessarily automatic. The econometric results from this study
indicate that most societies adopt policies and make investments that reduce environmental
damage associated with growth. This author’s work showed how ten different indicators react
to rising income for low, middle and high-income countries. Some indicators improved while
other worsened with rising income.
Research conducted by Stern (2004, 2017), the author presents a critical history of the
Environmental Kuznets Curve (EKC). He also reviewed the development of the EKC
concepts, the theory behind the EKC and the econometric methods used in EKC studies.
Stern highlighted the more important recent developments that have changed the view of the
EKC and alternative approaches that are being used such as decomposition of emissions and
efficient frontiers. The EKC, which is named for Kuznets who hypothesized that income
inequality first rises and then falls as economic development progresses, was a concept that
emerged in the early 1990s with Grossman and Krueger’s (1991) pathbreaking study of the
potential impacts of NAFTA and the concept’s popularization through the 1992 World Bank
Development Report. However, Stern (2004) argues that if the EKC hypothesis were true,
then rather than being a threat to the environment, economic growth would be the means to
eventual environmental improvement.
But contrary to the claims of the EKC concept, Stern (2004) went further to point out
the weaknesses of the EKC. He identified that most of the EKC literature are econometrically
weak which is seen in little or no attention being paid to the statistical properties of the data
used such as serial dependence or stochastic trends in time-series. Also, little consideration
has been paid to issues of model adequacy such as the possibility of omitted variables bias.
He also pointed out that when diagnostic statistics and specification tests are considered, and
appropriate techniques are used, the EKC does not exist and states that instead, we get a more

6
realistic view of the effect of economic growth and technological changes on environmental
quality. The economic factors that drive changes in environmental impacts and may be
responsible for rising or declining environmental degradation over the course of economic
development include proximate variables such as scale of production, composition effect,
technique effect and changes in input mix. Other underlying causes such as environmental
regulation, awareness, and education (Stern, 2017) were also pointed out. Table 1.1 below

Technique effect

Environmental regulation and
innovation policy

Political and institutional framework

presents these proximate factors and how they impact environmental degradation.
EKC effects
Scale effect

Variables
Production

Examples
Rising emissions
with growing
economy
Lower emissions
moving to a service
sector relative to
industrial production

Composition
effect

Output mix

Input mix

Mix of inputs
labour and capital

Robots replacing
humans

State of
technology
Emissions
specific changes
in process

Production
efficiency
Inputs usage

Less polluting inputs
per unit of output
Green energy

Table 1.1: Factors that influence the environmental quality of nations (Author's interpretation
of Stern (2004, 2017).

Stern (2004) provided a summary of several studies of sulfur emissions and
concentrations in the order of estimated income turning point. He found that there is a
monotonic relation between sulfur emissions and income just as there is between carbon
dioxide and income for recent studies that used more representative samples. The estimated
turning points from these studies ranged from $3,137 by Panayotou (1993) to $101,166 by
Stern and Common (2001) for sulfur emissions. Stern (2004) concluded from the EKC
literature that concentrations of pollutants may decline in nations from middle income levels,
while emissions tend to be monotonic in income.
The econometric criticisms of the EKC which are heteroskedasticity, simultaneity,
omitted variables bias, and cointegration issues was discussed by Stern (2004). The majority
of studies have found the EKC to be a fragile model suffering from severe econometric

7
misspecification (Millimet et al., 2003; Sobhee, 2004). Stern also proposes that the use of
more appropriate methods tends to indicate higher turning points and possibly a monotonic
curve for emissions of major pollutants. A better model may result from including additional
variables to represent either proximate or underlying causes of change in emission. A detailed
theoretical and empirical review of the EKC concludes that the existence of a simple and
predictable relationship between pollution and per capita income is not robust. The inverted
U-curve becomes monotonic or disappears when the model is adjusted for several tests and
when more variables are added to it.
Dasgupta et al. (2002) shows evidence that developing countries are also performing
better due to informal or decentralized regulation. Also, liberalization of developing
economies has encouraged more efficient use of inputs and less subsidization of
environmentally damaging activities. Other changes include multinational companies raising
standards in the countries in which they invest, better methods of regulating pollution and
better information on pollution being available. This has encouraged government to regulate
and empower local communities, indicating that the regulatory capacity of developing
countries has been strengthened.
Mckitrick and Wood (2017) closely examined the relationship between four common
air pollutants and income across Canadian by using data on local pollution concentrations and
provincial-level and metropolitan-level macroeconomic variables. The purpose was to
identify the scale effect, composition effect, and technique effect while controlling for
unidentified characteristics of individual monitoring stations. These four air pollutants
include the annual average concentrations of sulfur dioxide (SO2), nitrogen dioxide (NO2),
carbon monoxide (CO) and ground-level Ozone (O3) at monitoring stations across the
country from 1984 to 2010. By use of panel methods and pollution concentration data from
individual monitoring stations, their study allowed for a much larger sample size than
previous Canadian studies. The econometric modelling approach used in their study separates
and identifies the relative magnitudes of the scale, composition, and technique effects. Their
results show many similarities in the income–pollution relationship for concentrations of SO2,
NO2, and CO. For all three pollutants, the relationship with income switches from negative to
positive when time fixed effects are accounted for which reflects the improvement in
technique over time. A positive effect of increases in the scale of the economy was
completely offset by improvements in technology and changes in the composition of output.
The results for ground-level ozone were similar to the other pollutants when choosing the
measure used to assess the Canada-Wide Standard (CWS) but different when using annual

8
average concentrations of ozone. The authors point out that this difference may be due to the
focus of government policy to reduce short-term, rather than long-term, exposure to ozone.
The results further revealed no scale effect for CO after controlling for changes in
composition and technique, no composition effect for SO2 but composition effects for CO
and NO2 exist. There was no relationship with income identified when looking at annual
average concentrations of O3 but when the CWS measurement of ozone concentrations was
used instead, the results were similar for the other three pollutants which comprised of a
positive scale effect, a negative composition effect, and a negative technique effect.
In another context, Costantini and Monni (2008) combined the Resource Curse
hypothesis and EKC to test the causal relationship between economic growth, human
development and sustainability. Sample data of 14 nations with resource curse and another 14
nations with resource blessings from 1975-2003 on economic growth and genuine savings
were used. The authors formulated an integrated model which was modified to reflect the role
of human development in the EKC and the quality of institutions on the Resource Curse
hypothesis (RCH). The aim of the study was to provide a link between the RCH modified
with the role of institutions on one side and the relationship between economic growth and
sustainable development on the other. The findings maintain that natural resource endowment
could be a source of low economic growth rates if the institutions in a country do not have the
ability to manage the resources in the right way. They pointed out that investment policies
geared towards human capital formation are effective actions in reaching a higher
development level and consequently in the quality of institutions. The authors also deduced
that an economy based on resource exploitation without appropriate institutions will run into
Dutch disease or rent-seeking effects (which is the negative impact put on an economy by
anything that gives rise to a sharp inflow of foreign currency) with reduced economic growth
and lower Human Development Index (HDIs). They affirmed that Human Development
should be the first objective of international development policies whereas an increase in
human well-being is necessary to provide a sustainability path. Their study expands the
traditional EKC to show how an increase in Human Development and quality of institutions
also plays a significant role in affecting the environmental quality of nations.
A study conducted by Criado et al. (2011), shows that stabilizing pollution levels in
the long run is a pre-requisite for sustainable growth. These authors developed a neoclassical
growth model with endogenous emission reduction in order to analyse the conditions under
which an economy may achieve sustainable growth. This means balanced growth paths
characterized by growing per capita incomes and non-declining environmental quality

9
predicting that, along optimal sustainable paths, pollution growth rates are positively related
to output growth (scale effect) and negatively related to emission levels (defensive effect).
Panel data for 25 Eastern and Western European countries over the period 1980-2005 was
used to test the existence of both the scale and the defensive effect for two air pollutants,
sulfur oxides (SOX) and nitrogen oxides (NOX).
Based on the framework implemented by Criado et al. (2011), sustainability requires
satisfying a general condition of pollution convergence. By considering the question, “How
do pollution dynamics interact with output dynamics along sustainable growth paths?”, the
authors considered a growth model with endogenous pollution abatement, to show that the
optimal path is characterized by a precise dynamic relationship between pollution growth
rates, emission levels, and output growth rates, which induces pollution convergence in the
long run. This dynamic law was tested empirically for the two major air pollutants SOX and
NOX using panel data from European countries.
Criado et al. included an additional element to the model by also predicting a positive
interaction between pollution growth and income growth, added to a negative growth-level
relationship in pollution. This analysis was referred to as a convergence test in which β convergence in pollution is conditional on country-specific output dynamics. All these
specifications allow for structural dissimilarities within groupings of countries through a
group-specific dichotomous variable. Their results were consistent with the predictions of the
theoretical model and confirm the existence of scale effects and defensive effects for SOX and
NOX. Findings showed that the path followed since 1985 by the NOX emissions per capita is
fully compatible with the convergence equation predicted by the theoretical model, but with a
stronger evidence holding within the European Union (EU15) countries. The defensive effect
reflects the effectiveness of abatement expenditures in limiting pollution growth. Regression
estimates support the model predictions, identifying a clear scale effect linked to GDP growth
and a negative effect captured through the impact of the past pollution level component.
Taylor and Brock (2010) set out to provide a cohesive theoretical explanation for
three features of the pollution and income per capita data which includes emissions, emission
intensities and pollution abatement costs. They established that the EKC and the core model
of modern macroeconomics which is the Solow model are intimately related. By introducing
a very simple growth model closely related to the one-sector Solow model, the authors show
how this amended model generates predictions closely in line with U.S. and European
evidence. They showed this by amending the Solow model to incorporate technological
progress in abatement which results in the EKC being a necessary by product of convergence

10
to a sustainable growth path. This was an alternative empirical method tightly tied to theory
to estimate their model on carbon emissions from 173 countries over the period of 1960–
1998.
Taylor and Brock used the Green Solow model to provide a very simple explanation
for all three puzzles. They borrowed techniques used in the macro literature on income
convergence to derive a simple linear estimating equation linking growth in emissions per
capita over a fixed time period to emissions per capita in an initial period and a limited set of
controls. These controls include typical Solow type regressors such as population growth and
the savings rate, but also include a measure of pollution abatement cost. An augmented
Solow model was developed where exogenous technological progress in both goods
production and abatement leads to continual growth with rising environmental quality.
Collection of data on carbon emissions per capita, population growth rates and the investment
share of GDP for a group of 173 countries from 1960 to 1998 was used to conduct their
empirical work.
The results show that relationship between income and pollution is complex even
when using this simple model. The EKC and the Solow model, are intimately related
identifying the forces of diminishing returns and technological progress by Solow as
fundamental to the growth process, may also be fundamental to the EKC finding. Because of
diminishing returns, development starts with rapid economic growth, emissions rise with
output growth but fall with ongoing technological progress in abatement. The findings also
show that as countries mature and approach their balanced growth path, economic growth
slows and the impact of this slower growth on emissions is now overwhelmed by the impact
of technological progress in abatement and emission levels decline. This interplay of
diminishing returns and technological progress generates a time profile of rising and then
falling emission levels as income per capita grows along a path of sustainable growth. The
authors deduced that a tightening of pollution policy raises costs and lowers the level of
pollution, but not its long run rate of growth, showing that, environmental policy has a level
and not growth effect in the model.
Environmental Performance Index
The Environmental Performance Index (EPI) is a composite index composed by Yale
university that ranks nations on their environmental quality for two broad objectives of
Environmental Health and Ecosystem Vitality (Hsu, 2016). However, the 2012 pilot Trend
Index ranks countries on the basis of improvement or decline from 2000 to 2010 (Emerson et

11
al., 2012). The EPI is a composite index that includes multiple tiers of indicators to assess
country-level environmental performance with a score from 0 to 100 and a ranking relative to
other countries. The policy categories represent core areas of environmental policy concern
for which measurable indicators can be assessed. For the purposes of evaluating the trends in
the different country’s performances, the ten policy categories from the 2012 EPI and Trend
EPI are used in this study.
The Environmental Health objective measures the impacts on human health in three
policy categories of Air, Water and Human Health (Hsu, 2016). The Ecosystem Vitality
measures the impacts on the ecosystem and natural resources in seven policy categories of
Air, Water, Fisheries, Forests, Climate Change and Energy, Biodiversity and Habitat, and
Agriculture (Hsu, 2016) as seen in figure 1.2.
Hsu et al. (2013) used the 2012 EPI and trend EPI index proximity-to-target
methodology and simple regression analysis to assess how nations have progressed in the
environmental issues identified in the Millennium Development Goal (MDG7). Using data
from international organizations, research institutions, government agencies and academia,
they created a paired down Rio index and Rio trend index highlighting the policy categories
of the MDG7 which includes; Water (effects on human health), Biodiversity and Habitat,
Forestry, Fisheries, and Climate Change and Energy, demonstrating how these nations have
improved or declined in those areas. Additional explanatory factors used were GDP per
capita, Human Development Index (HDI), non-income HDI, Control of Corruption and Voice
and Accountability. Simple linear regression results show that progress in those
environmental issues identified are uneven and they vary by country, region and issue. The
results also show that income only accounts for a certain percentage of environmental change
in nations and that other factors also play a role.

12

Figure 1.2: The 2012 Environmental Performance Index (EPI) (Emerson et al., 2012).

13
The EPI assesses social and economic driving forces, pressures on the
environment, states of the environmental changes and impacts on human health and
ecosystems (Hsu, 2016). A brief description of its methodology is presented below:
EPI Pilot Trend Methodology
The method employs a multi-step process to produce indicators on a consistent
scale to allow for comparison across sectors (Hsu et al., 2013). The policy indicators are
based on a proximity-to-target methodology as shown in Figure 1.3. Each country’s
performance on any given indicator is measured based on its position and within a range
determined by the lowest performing country (poor performance benchmark, equivalent
to 0 on a scale from 0-100) and the target (top performance benchmark, equivalent to
100). The proximity-to-target score (

, , ) of each nation for each time period is:

,

,,

Where

is the poor performance benchmark,

benchmark or the target and

,

∗ 100

is the top performance

is the indicator of nation i at time t = 2000 – 2010. The

proximity to target shows how far the indicator score is from the poor performance
benchmark as a fraction of the distance between poor and top performances. If it is not far
from the target, then the

,,

will be closer to 100 and if it far from the target it will be

closer to 0.

Figure 1.3: Performance benchmark for nations (Hsu et al., 2013).

14

Then for each indicator, a simple linear regression model of the annual proximity-totarget scores is used to determine a rate of improvement or decline for each indicator. The
slope of the trend line determines the scale. 0 slope reflects “no change”, a positive slope
reflects improvement and a negative slope indicates decline. This is done for every
nation and for every indicator. Then these slopes for each indicator are ranked from “best
improvement” receiving a score of 50 and defined by the 95% percentile of the slopes, 0
slope reflecting “no change” again and -50 is for the “worst trend decline”. Forest Loss,
Forest Growing Stock, Forest Cover, and Change in Water Quantity have trend scores
that range from -50 to 0 as they are change indicators (See Appendix A for EPI indicators
framework and exploration).
The 2012 EPI trend ranks a range of 170 - 230 countries on the change in their
environmental performance over the last decade in two broad policy objectives which are
Environmental Health and Ecosystem Vitality which are defined below:
Environmental Health trend (human health effects): This objective measure which
countries are improving and those declining in policy categories associated with
environmental stresses to human health (See Appendix B for definitions of each indicator
in the policy categories). The policy categories are:


Air Pollution (effects on human health): This policy category consists of two
indicators namely indoor air pollution (INDOOR) and particulate matter
(PM25).



Water (effects on human health): This policy consists of two indicators which
are; access to drinking water (WATSUP) and access to sanitation (ACSAT).



Environmental Burden of Disease: This policy consists of child mortality
(CHMORT).

Ecosystem Vitality trend (ecosystem effects): This objective measure which countries are
improving and those declining in policy categories associated ecosystem health and
natural resource management. The policy categories are:

15


Air Pollution (effects on human ecosystem): This policy consists of sulfur
dioxide emissions per capita (SO2CAP) and sulfur dioxide emissions per GDP
(SO2GDP).



Water (effects on ecosystem): This policy consists of change in water quantity
(WATUSE).



Biodiversity and Habitat: This policy categories consists of biome protection
(PACOV), marine protection (MPAEEZ) and critical habitat protection
(AZE).



Forests: This policy category consists of forest loss (FORLOSS), forest cover
change (FORCOV) and growing stock change (FORGROW).



Fisheries: This policy category consists of coastal shelf fishing pressure
(TCEEZ) and fish stocks overexploited (FSOC).



Agriculture: This consists of agricultural subsidies (AGSUB) and pesticide
regulation (POPs).



Climate Change: This consists of CO2 emissions per capita (CO2CAP), CO2
emissions per GDP (CO2GDP), CO2 emissions per electricity generation
(CO2KWH) and renewable electricity (RENEW).

Gallego et al. (2014) uses the HJ biplot methodology and regression analysis to
examine the impacts of socio-economic factors such as GDP per capita and education and
institutional factors such as government effectiveness, control of corruption and political
ideology jointly on environmental performance in countries worldwide. They used a
sample of 149 nations and the 2008 EPI index as a measure of environmental
performance. The HJ biplot methodology gave a graphical representation of the
countries’ environmental performance in relation to environmental health and ecosystem
vitality to show how the economic and institutional factors affect them. The regression
results show that higher levels of income and education are strongly linked to the
environmental performance of these nations whereas governance effectiveness has little
to no effect on environmental performance. Testing a model where GDP was used in its
quadratic form gave results which conform with the EKC hypothesis which states that in
the early stages of economic development, environmental performance issues increases
along with income level, but then decreases in relation to GDP at higher levels.

16
A similar study yielding a different result was recently conducted by Mavragani et
al. (2016). The authors applied factor analysis methodology to an empirical model to test
the relationship between environmental performance, economic development, governance
and openness of market. They used a sample data of 78 countries including all G20 and
EU members as a representative of countries accounting for over 90% of global trade and
investment. They used the 2014 EPI index as a measure for environmental performance.
The also used the Open Market Index (OMI) proposed by the International Chamber of
Commerce (ICC) as a measure for openness of market, World Governance Indicators
(WGI) indicators which include Rule of Law, Voice and Accountability, Political
Stability and Government Effectiveness as a measure of governance indicators. Results
show a positive correlation between a country’s economic growth, the openness of an
economy, high levels of effective governance and its environmental performance.
Independent Approach
Using data from the Economics of Ecosystems and Biodiversity (TEEB) study,
Costanza et al. (2014) provided an updated estimate of the value of ecosystem services
from the earlier estimate in 1997 by generating global aggregates of the value of
ecosystem services using an accounting approach. These authors estimated the value of
17 ecosystems services for 16 biomes and an aggregate global value given in monetary
terms using a simple benefit transfer method. The authors selected 665 value data points
from over 300 publications which were screened in the Ecosystem Services Value
Database (ESVD). They also provided a comparison of the study conducted by DeGroot
et al. (2012) results with the Costanza et al., (1997) results to estimate the changes in the
flow of ecosystem. They also estimated the global changes in ecosystem services values
from land-use change over the period 1997-2011. The purpose of their study was to raise
awareness about the magnitude of these services relative to other services provided by
human-built capital at that point. Their results showed that global land use changes
between 1997 to 2011 have resulted in a loss of ecosystem services of between $4.3 and
$20.2 trillion/yr. This study brings to light the benefits of ecosystem services by how they
interact with the other forms of capital which are human, social and built capital, to
contribute to the services of a nation. This study connects to the huge policy objectives of
the EPI index whose goal is to promote ecosystem vitality and environmental health.
Another outstanding work that examines nations in a path for sustainable
development was done by Moran et al. (2008). The Human Development Index (HDI)

17
was used as an indicator for human development and the ratio of national footprint per
capita to global biocapacity per capita as an indicator for ecological sustainability. They
established the minimum criteria of sustainable development to be HDI≥0.8 and Footprint
to Biocapacity as ≤1.0. Sample data of 93 countries were used with data span of period of
1975-2003 for these indicators. Comparing the trends in HDI and Ecological Footprint for
the period of 1975-2003 revealed that only one country met the minimum criteria for both
indicators in the most recent year while most nations had exceeded the Ecological
Footprint requirement without reaching the Human Development requirement. They
identified the factors that determined the gap between the footprint and biocapacity as the
need for nations to prioritize ecosystem vitality by protecting ecosystems from climate
change an eliminating the use of toxic chemicals that degrade ecosystems. This also
involves the protection of soil from erosion and degradation, preserving croplands from
agriculture, protection of river basins, wetlands and watersheds to secure freshwater
supplies and maintaining healthy forests and fisheries.
Thesis Objectives
The approach to this thesis is built from the observations and reviews of previous
studies which used the EKC theory, EPI index and other methods to help explain the
relationship between economic growth and environmental performance. These reviews
have helped frame the scope of this study by identifying knowledge gaps which will also
contribute to the current body of knowledge on progress in economic and environmental
performances and sustainability.
Most EKC studies focus on how economic growth impacted various air pollutions
indicators and carbon emission (Stern, 2004) but there is a knowledge gap with the new
EPI index presenting different policy categories for environmental indicators (Hsu, 2016)
to be evaluated. The evaluation of the impacts of the growth rate of output on other
environmental indicators like agriculture, fisheries and biodiversity also has limited
research. There has been a more frequent use of the EPI index in recent years in
identifying how nations have performed in the environmental issues identified in the
MDG7 (Hsu et al., 2013) which impacts policy formation. A further step in research will
be to see how nations respond in all the environmental issues pointed out in the EPI index
and not just the MDG7 when associated with their average growth rate.
Another knowledge gap is identifying the factors common to those countries that
have a high performance in both growth rate of per capita income and environmental

18
quality simultaneously. The use of the EPI aggregate as an indicator for environmental
performance (Mavragani et al., 2016) rather than considering the various policy
categories of environmental health and ecosystem vitality to see if the results differ has
not been vastly studied. There is also limited knowledge in considering the long-run
period as opposed to individual years which can help reveal trends that will be useful to
assess the progress and contributions of national environmental assets over time to devise
more specific and measurable policy goals and targets (Moran et al., 2008).
This research is peripheral to the Environmental Kuznets literature, peripheral to
the macroeconomic growth convergence literature, and peripheral to the ecological
economics literature by taking a different approach to inquiry about economy and the
environment. Looking at the 2000-2010 decade which is a turbulent period for the
economy given events such as the September 11, 2001 terrorist attack on USA, the
response of governments around the world to such an event as well as the influence of the
2008 great recession adds an element of economic and political uncertainty with the
economy area. Nations during that period achieving growth in both the economy and the
environment simultaneously is not as easy given such a decade.
This study aims to investigate the trends in the performances of over 200 nations
for the 10 environmental policy categories associated with their average GDP per capita
growth rate respectively. The aim is to discover clusters of nations which had a win-win
trend relative to those which did not as well as those nations which had very high
performances in both areas simultaneously and the factors which influence the likelihood
of such performances. Examining the impact of economic growth for the environmental
performances of nations for different policy targets will add more information and build
on the existing literature.
According to the classification in Figure 1.4 of this study, nations in the win-win
category are those nations that have progress in their economic growth and environmental
quality simultaneously over a long period of time. They are located in the upper right
quadrant of the figure. According to the EKC theory these would be nations with a high
standard of living mostly operating on the left-hand side of the EKC curve where
increases in the nations income per capita or standard of living leads to improvements in
the environmental quality. Within this quadrant, clusters of nations with high
performance trends in both areas will be identified sometimes using thresholds. The losswin category which is the bottom right quadrant indicates nations that have progress in

19
their economic growth but a decline in their environmental performance over the decade
illustrating trade-offs between economy and the environment. The win-loss category is
the top left quadrant and it indicates nations that have a decline in their economic growth
but still progress in their environmental quality which according to the EKC theory would
be located to the left side with nations that have a low income per capita. The loss-loss
category which is the bottom left quadrant indicates nations that have a decline in both
indicators simultaneously over the decade.

Environmental improvements
(EPI Pilot trend index)

Cluster of high
performance in both
areas

Win‐loss

Win‐win
0,0

Loss‐loss

Growth of income per capita (%)

Loss‐win

Figure 1.4: Classification of trends in economic growth and environmental performance for
the design of this study. Win-win is the top right quadrant with an additional high win-win
cluster. The loss-win is the bottom right quadrant. The win-loss is the top left quadrant and
the loss-loss is the bottom left quadrant.

The threshold for high performance for the growth rate of per capita GDP is set at
2 percent, although somewhat arbitrary, it is based on the long run growth rate of the
industrialized nations (i.e., U.S.A, Canada, England, Japan, Germany, etc.) over the last
century. Hence these high economic growth nations are most likely developing and

20
emerging nations which according to EKC should be in the win-loss of Figure 1.3 and not
in the win-win. The threshold for high performance in the environmental policy
categories will be determined separately for each policy category based on their trend
scores which is set differently by Yale University.
In addition to the classification discussed above, this study will explore also the
factors that influence the likelihood of achieving high performances in both areas
simultaneously. The control variables possibly associated with the likelihood of such
high-performance clusters in both areas includes the initial GDP per capita and initial
non-income HDI (i.e., education and health). Explanatory factors likely impacting the
likelihood of high performance nations in both areas simultaneously relative to other
nations that are not in that cluster of nations include change in government effectiveness
and political stability which are mostly governance indicators over the period. Also, the
investment spending as a fraction of the size of the economy which is considered one of
the engines of economic growth will be considered. The 2012 Environmental
Performance trend index policy categories mentioned previously would be used as the
indicator for environmental performance.
In what follows: Chapter 2 will employ the K-means cluster analysis to identify
various clusters of nations for the average growth rate GDP per capita and each of the 10
environmental trend policy categories as per Figure 1.3 above without having thresholds
for the variables for a sample of 150 - 230 nations. This research uses six clusters to place
the data points of the nations which will appear in different locations on Figure 1.3. The
aim is to determine the type of nations in a win-win situation and those which are not.
These clusters will also be correlated with the human development index which is
composed of GDP per capita, education and health indicators of the nation in an attempt
to determine the kind of nations in a win-win.
Chapter 3 uses logistic regressions to determine which factors are important to
increase the likelihood of achieving simultaneous high performance in both areas of the
economy and the environment. Performance thresholds are set so that the nations being
explored are those that had to achieve a relatively high performance in growth rate of
GDP per capita and environmental performance. For example, high economic
performance was set at an average growth rate of more than 2 percent per year which is in
line with historical average growth rates of developed nations. EKC theory would predict
that these fast-growing nations would be causing environmental degradation and not

21
environmental improvements. The results of achieving high performance in both factors
will be compared with the results of high performance in a single factor which should be
easier to achieve (higher likelihood) by nations than having very good performances in
both factors simultaneously.
Chapter 4 provides insights into policy considerations based on the results,
suggests possible future extensions and discusses limitations of the study.

22
CHAPTER 2. CLUSTER ANALYSIS EVALUATION OF THE GROUPS OF
NATIONS WITH A WIN-WIN TREND IN ENVIRONMENTAL
PERFORMANCES AND GDP.
Introduction
There have been well-known indices developed over the years to assess the
performance of nations in environmental issues in order to drive informed policy-making
towards sustainability. Some of these indices include the Environmental Sustainability
Index (ESI) (Esty et al., 2005), Environmental Performance Index (EPI) (Emerson et al.,
2012), Ecological Footprint (EF) (Wackernagel and Rees, 1998), and Global Green
Economy Index (GGE) (Tamanini et al., 2014) which all evaluate environmental
performances of nations using different measures.
The ESI was created to evaluate the environmental sustainability of countries
relative to the paths of other countries (Esty et al., 2005) while the EPI is a method of
ranking the environmental performance of a state's policies numerically (Emerson et al.,
2012) and has been widely used in several studies. For example, the 2012 pilot trend EPI
which was compiled to allow countries to examine changes in performance on who is
improving and declining from 2000 to 2010 was used by Hsu et al. in 2013 to gauge the
improvement or decline in the environmental policy targets set forth in the United
Nations Millennium Development Goals (MDGs). The Ecological Footprint (EF) which
monitors ecological resource use is also widely used by scientists, businesses,
governments, individuals, and institutions to advance sustainable development
(Wackernagel et al., 1990).
The environmental performance for high-income, middle-income and low-income
nations had changed significantly both positively and negatively over the years evident in
the various indices. For example, the 2012 EPI showed the ranking for the top 5 nations
to be Switzerland, Latvia, Norway, Luxembourg and Costa Rica (Emerson et al., 2012).
At the low end of the 2012 EPI rankings are South Africa, Kazakhstan, Uzbekistan,
Turkmenistan and Iraq. Switzerland’s top ranking was in large part due to its high
performance in air pollution control both on human health and the ecosystem, access to
drinking water and the biodiversity and habitat indicators and was not due to its level of
income. These results were interesting to see because some middle-income countries,
such as Latvia (per capita GDP $12,938) and Costa Rica (per capita GDP $10,238) also
achieved high environmental outcomes. This result, although anecdotal, finds a

23
contradiction for the EKC theory by suggesting that income alone does not determine
good environmental performance. In the 2014 EPI ranking, the top five countries were
Switzerland, Luxembourg, Australia, Singapore, and the Czech Republic. The bottom
five countries in 2014 were Somalia, Mali, Haiti, Lesotho, and Afghanistan (Hsu et al.,
2014). In 2014, developed nations like the United Kingdom was ranked in 12th place,
Japan 26th place, the United States 33rd, while some developing nations like Brazil 77th,
China 118th, and India came in 155th. The different results for some of the nations in the
top ranks and bottom ranks between 2012 and 2014 is interesting to see which can be
further investigated. It reveals that both the high and low performance of nations for
environmental issues can be influenced by certain factors over the years which could
affect their sustainability other than income. The above analysis is static showing
performances in a given year, however the 2012 pilot trend EPI focuses majorly on the
ranking of nations on the performance trends for numerous environmental indicators in
health and ecosystem vitality from 2000 to 2010. Furthermore, there’s a gap in
identifying the trends for these environmental indicators associated with their GDP per
capita growth rates over the same decade.
The GDP per capita growth rate which is also a well-known concept widely used
in the theory of economic development is an important indicator of economic health of
the average person of a nation and reflects progress in the standard of living of a nation
via new businesses, jobs and personal income. However, many economists, and in
particular ecological economists, have cast doubt in GDP as a measure of standard of
living of a nation and prefer to use Genuine Progress index (GPI). The GPI accounts for
the impact of positive and negative environmental and other externalities from the
production of goods and services (Lawn, 2003; Kubiszewski et al., 2013). Arrow et al.
(2004) uses the genuine wealth growth rate per capita which accounts for the depreciation
of natural capital to measure progress in the standard of living of nation. They find that
the GDP per capita growth rate overstates the increase in the standard of living of nations
and a better measure to use is the genuine wealth growth rate per capita.
The EPI policy categories are outcome-oriented indicators and are used as a
benchmark index by policy makers, environmental scientists, advocates and the general
public (Hsu et al., 2013) which makes it a reliable index for assessment. Have the nations
with the high performance in environmental indicators also improved, declined or
remained unchanged in their GDP growth simultaneously?

24
The resulting trends for 2000 to 2010 for environmental performance and GDP
per capita growth rate can reveal vital information about nations from their past
performance which may be useful in making assessments and projections. It will reveal
the behavioral traits and patterns of nations that progressed and the direction in which
these nations have moved over the decade. Was a high-performance trend sustainable for
nations across all indicators? What kind of nations maintained high performances and in
what categories? Is income the only factor impacting a win-win trend? This study will
employ the k-means clustering analytical technique to identify different groups within the
trends of nations that have progressed in the 2012 pilot trend EPI policy categories
associated with their average GDP per capita growth rate to assess nations in a win-win
relative to those which are not.
Methods
The goal of this chapter is to identify and group nations into the win-win, winloss, loss-win and loss-loss categories for the average growth rate of GDP per capita and
each of the 10 EPI trend policy categories as per Figure 1.3 without having thresholds for
the variables from 2000 – 2010 for over 200 nations. It also employs cluster analysis
which is a multivariate analytical technique to find useful groupings that are tightly knit
in a statistical sense and distinct from each other within the different categories as per Fig
1.3 but especially the win-win category. The results from organizing the data into
homogeneous groups can provide either immediate insights or a foundation upon which
to construct other analyses (Kettenring, 2006) (See Appendix C for more background
information on cluster analysis).
k-means clustering
This study specifically employs the K-means clustering technique which is a nonhierarchical clustering method to identify various clusters of nations within the win-win,
win-loss, loss-win and loss-loss categories. The k-means cluster analysis is done by a
mechanical algorithm which will subdivide the two data points of each nation into
clusters based on nearest mean values. The algorithm minimizes the distance between the
data points in each cluster to obtain the optimal division of these points into clusters. The
k stands for the number of clusters in the data and is set by the researcher in such a way to
minimize the sum of square of errors. The aim is to determine the type of nations in the
win-win category and clusters of nations that have high performances in both areas of

25
economy and environment simultaneously. The K-means cluster analysis is a data mining
technique that is used in marketing research, computer science, geography, astronomy
and agriculture studies but is not frequently used with economic research. These clusters
will also be correlated with the human development index (HDI) which is composed of
GDP per capita, education and health indicators in an attempt to determine what kind of
nations achieve a win-win situation relative to those which did not.
Each data point is assigned to its nearest cluster determined by the Euclidean distance;

,

,

and the new centroids for the clusters are computed. By recalculating the
Euclidean distance from each subject to each centroid, the observations are moved to the
clusters they are closest to. This process is repeated until the centroids remain relatively
stable (Rosie, 2007).
Data collection
All data on 2012 Environmental Performance trend index for the 10 policy
categories on environmental health and ecosystem vitality were collected electronically
from Yale web portal (www.epi.yale.edu). All data on GDP per capita for a range of 170
– 230 nations were electronically retrieved from Gap minder Compiled by Mattias
Lindgren for the period of 2000 – 2010 used to compute the growth rate.
The policy categories represent core areas of environmental policy concern for
which measurable indicators can be assessed. For the purposes of evaluating the trends in
the different country’s performances, the ten policy categories from the 2012 EPI and
Trend EPI are used in this study. The Environmental Health objective measures the
impacts on human health in three policy categories of Air, Water and Human Health
(Hsu, 2016). The Ecosystem Vitality measures the impacts on the ecosystem and natural
resources in seven policy categories of Air, Water, Fisheries, Forests, Climate Change
and Energy, Biodiversity and Habitat, and Agriculture. We have a total of 10 policy
categories, 3 from Environmental Health and 7 from Ecosystem Vitality (Figure 1.2).
The Software used to run cluster analysis was the NCSS 11, by NCSS, LLC. It is a
statistical software which provides statistical tools used to analyze and visualize your
data. All the data was imported, filtered and transformed using the software. The

26
optimum number of clusters (K) determined for each policy category is six (6) based on
the elbow method by Thorndike (1953) which defines clusters such that the total intracluster variation or total within-cluster sum of square (WSS) is minimized. The optimum
number of clusters should be chosen such that adding another cluster doesn’t improve the
total WSS. The total WSS measures the compactness of the clustering which should be as
small as possible. The Elbow method looks at the total WSS as a function of the number
of clusters.
These clusters will also be correlated with the human development index which is
composed of GDP per capita, education and health indicators of the nation for the year
2000. The HDI 2000 was also electronically retrieved from the United Nations
Development Programme (UNDP) website.
Results
Win-win for any cluster is characterized by the nations that fall in the area of
positive values for x and y axes for average GDP per capita growth rate and the
environmental trend policy categories. Nations that are above the (0,0) coordinate have
achieved a win-win. According to the standard set up in the pilot trend index by Yale
university, the trend scores for all environmental policy categories ranges from -50 to 50
with -50 being the lowest trend score and 50 being the highest trend score. The
exceptions for these categories with a different trend score includes Forestry (-50 to 0),
and Water (ecosystem effects, -50 to 0) which are change variables (See Appendix D for
list of countries, codes and HDI ranks).
Environmental Health: Child Mortality trend, Growth and HDI 2000
Fig 2.1 shows the trends of nations in their association between Child Mortality
(EH) and the average GDP per capita over the decade. The win-win category which is the
upper right quadrant contains nations that score well on both indicators. Over 80% of
nations are in a win-win and are clusters 1, 2, 3, and 6 (Table 2.1, Figure 2.1). Cluster 3
has 29 nations with a win-win and has the highest improving EH trend of nations and a
very high average growth rate of the economies. It is composed mostly with low (41.4%)
and medium (44.8%) HDIs and only a few high (10.3%) and very high (3.5%) HDI
nations and consists of mostly Middle-eastern and North-African countries. This may
only be explained by the EKC model with a turning point of income per capita that is
very low. Cluster 6 is also a win-win with the highest average GDP trend but also high

27
EH trend consisting of only 8 nations with mainly medium HDIs (62.5%) some of which
include China (CHN), Azerbaijan (AZE), and Equatorial Guinea (GNQ). This also
suggests that the EKC turning point is very low. Clusters 1 and 2 has the most nations in
the win-win trend. Cluster 1 has 43 nations with a high win-win trend which consists of
low, medium, high and only a few very high HDI nations made up of some Central Asian
and Eastern Europe countries. Cluster 2 has 77 nations with a win-win trend but closer to
the origin of not having win-win and not as high as cluster 3 and 6. Cluster 2 consists of
nations with very high HDIs (37.7%) some of which include Great Britain (GBR), Italy
(ITA), and Belgium (BEL) as well as high (15.6%). Cluster 4 is a loss-win situation for
only 9 nations with relatively medium and high HDIs. This is worrisome as this indicates
improvements in economic performance is trade-off with respect to increased child
mortality. It is only 9 nations and most of these nations are from Eastern Europe
including Russia (RUS), Cuba (CUB) and Slovenia (SVN). Cluster 5 shows a win-0
situation for 14 nations of mostly low HDI nations and indicates a declining Child
Mortality trend with no change in average GDP growth rate over the years suggesting a
priority is given to reducing child mortality than to economic growth.

Table 2.1: The mean and p-values for Child Mortality trend and the Average GDP per
capita growth rate for the 6 clusters of nations in the different categories. The HDI 2000
shows the rank of the nations in percentage for each cluster.

Variables
EH-EH mean
p-values
Av growth mean
p-values
HDI 2000 (%)
Very high
High
Medium
Low
Count

1
2
Win-Win Win-Win
15.70
12.32
<0.001
<0.001
4.07
1.05
<0.001
<0.001
7.0
32.6
32.6
27.9
43

37.7
15.6
19.5
27.3
77

Cluster
3
4
5
6
Win-Win Loss-Win Win-0 Win-Win
40.73
-13.91
38.91
26.58
<0.001
0.012 <0.001
<0.001
4.38
4.12
-0.32
9.58
<0.001
<0.001
0.610
<0.001
3.5
10.3
44.8
41.4
29

11.1
55.6
33.3
0.0
9

0.0
14.3
21.4
64.3
14

12.5
0.0
62.5
25.0
8

This policy category shows an overall positive trend for over 80% of nations with
medium and low HDI nations taking the lead in the win-win for the highest Child
Mortality trend and GDP growth over the decade. For this category, EKC holds with a

28
low turning point of income per capita. Reduction in child mortality seems to be a priority
for most developing and underdeveloped nations.

Figure 2.1: The distribution for the 6 clusters of nations in Child Mortality trend and the Average GDP per capita growth rate in the win-win
(upper right), win-loss (bottom right), loss-win (upper left) and loss-loss (bottom left) categories.

Environmental Health: Air Pollution, Growth and HDI 2000
Fig 2.2 shows the trends of nations in their association between Air Pollution and
the average GDP per capita over the decade. The win-win category which is the upper
right quadrant are nations with good performance trends which consists of clusters 3, 4
and 6 (Table 2.2, Figure 2.2). Cluster 4 has a total of 34 nations with a very high win-win
trend especially in the improvements of air pollution (trend) while the GDP per capita
growth rate is spread from no growth to high growth rates. Cluster 4 are mostly nations
with high (32.4%), medium (23.5%) and low (29.4%) and only few very high (14.7%)
HDIs. They consist of some Latin American, Eastern Europe and Sub-Saharan African
countries like Mexico (MEX), Malaysia (MYS), Slovak Republic (SVK), Zambia (ZMB)
and Cameroon (CMR). This cluster supports the EKC theory if the turning point of
income per capita is low as was the case with child mortality for developing nations.
Clusters 3 and 6 has the most nations in the win-win situation as well. Cluster 3 has 32
nations in the win-win trend with medium and low HDIs which consists of some East
European countries. Many nations in Cluster 3 show no change in air pollution trend
achieved with a very high growth rate of GDP per capita. This shows again that EKC is
not holding since very high growth of the economy for emerging nations should be
accompanied with a deteriorating environment assuming the turning point is not very low
for air pollution. Cluster 6 also has a win-win with 81 nations with both very high,
medium and low HDIs which consists of mostly European and some Sub-Saharan
countries. Cluster 6 nations relative to those in cluster 4 have a lower GDP per capita
growth rate indicating that environmental improvements do not always need high
economic growth rates. Cluster 1 has 15 nations in a 0-loss situation indicating no change
in its air pollution trend but a reduction in their standard of living. These are mostly low
HDIs nations. Cluster 2 with only 4 nations have the highest average GDP trend. Two
show improvements in air pollution while the other two show a deterioration. The 4
nations have low and medium HDIs and are China (CHN), Azerbaijan (AZE), Equatorial
Guinea (GNQ) and Armenia (ARM). Cluster 5 is a loss-win situation for 11 nations with
relatively low and high HDIs. whose economic performance is occurring at the expense
of the environment and is consistent with the EKC theory.

31
Table 2.2: The mean and p-values for Air Pollution (human health) trend and the Average
GDP per capita growth rate for the 6 clusters of nations in the different categories. The HDI
2000 shows the rank of the nations in percentage for each cluster.

Variables
EH-Air mean
p-values
Av growth mean
p-values
HDI 2000 (%)
Very high
High
Medium
Low
Count

1
2
0-Loss 0-Win
-1.17
3.09
0.516 0.493
-1.57 11.21
0.001 0.003
20.0
26.7
6.7
46.7
15

0.0
0.0
75.0
25.0
4

Cluster
3
4
5
6
Win-Win Win-Win Loss-Win Win-Win
2.06
25.24
-17.74
1.20
0.009
<0.001
<0.001
0.002
4.80
2.91
3.53
1.59
<0.001
<0.001
<0.001
<0.001
4.8
26.2
38.1
31.0
42

14.7
32.4
23.5
29.4
34

0.0
45.5
18.2
36.4
11

34.6
11.1
29.6
24.7
81

For this category, we see more nations with low, medium and high HDIs in a
higher win-win trend than the nations with very high HDIs. cluster 5 supports the EKC
theory if the turning point is high for those nations but the turning point has to be low for
the other cases to confirm to the EKC model.

Figure 2.2: The distribution for the 6 clusters of nations in Air Pollution trend (effects on human health) and the Average GDP per capita growth
rate in the win-win (upper right), win-loss (bottom right), loss-win (upper left) and loss-loss (bottom left) categories.

Environmental Health: Water Quality trend, Growth and HDI 2000
Fig 2.3 shows the trends of nations in their association between Water Quality and
the Average GDP per capita over the decade. Clusters 1, 2, 5 and 6 have nations in the
win-win category (Table 2.3, Figure 2.3). Cluster 5 has 23 nations with very high winwin and also the highest improved Water Pollution trend which consists of nations with
mostly medium (56.5%) HDIs which consists of some Latin American countries and a
few low (17.4%), high (17.4%) and very high (8.7%) HDI nations as well. High
environmental performance with high economic growth for emerging nations is
consistent with the EKC theory too if the turning point income is low as well. Clusters 1
and 2 has a total of 81 nations with a good win-win trend but not as high as cluster 5 and
consists of nations spread between having low, medium and a few high HDIs and consists
of mostly Middle Eastern and some Sub-Saharan African countries. Cluster 1 has high
environmental performance but lower economic performance relative to that of cluster 2
on average. Cluster 6 has only 4 nations with a very high win-win trend and the highest
average GDP growth trend with nations same as the Child Mortality and Air Pollution
indicators in their highest average GDP trend. Cluster 4 is a win-loss situation for 22
nations with mainly low HDIs and is consistent with EKC theory provided the turning
point income is high. Cluster 3 shows a 0-win situation for 53 nations of mostly very high
HDIs which indicates no change in water quality as if affects human health even while
their economies progressed.
Table 2.3: The mean and p-values for Water Quality trend (human health) and the Average
GDP per capita growth rate for the 6 clusters of nations in the different categories. The HDI
2000 shows the rank of the nations in percentage for each cluster.

Variables
EH-Water mean
p-values
Av growth mean
p-values
HDI 2000 (%)
Very high
High
Medium
Low
Count

1
2
Win-Win Win-Win
19.50
5.36
<0.001
<0.001
1.60
4.97
<0.001
<0.001
8.1
16.2
40.5
35.1
37

2.3
29.5
31.8
36.4
44

Cluster
3
4
5
6
0-Win Win-Loss Win-Win Win-Win
0.65
3.70
34.26
12.39
0.163
0.011
<0.001
0.090
1.74
-0.75
3.92
11.21
<0.001
0.038
<0.001
0.003
49.1
20.8
13.2
17.0
53

4.5
18.2
13.6
63.6
22

8.7
17.4
56.5
17.4
23

0.0
0.0
75.0
25.0
4

34
In this category, we see nations with very high and high HDIs having no change
to minimum increase in their water quality even while their economies continue to
improve while nations with low to medium HDIs generally have a good water quality
trend continuously. These findings are in in contrast with empirical evidence of EKC on
water quality indicators.
In summary, for the three human health categories, a win-win is more likely for
all nations both developing and developed as these categories have a more direct effect on
our health and hence a policy priority. We now turn to ecosystem vitality in which
deterioration of environment does not have a direct impact on our health but trade-offs
impact future generations.

Figure 2.3: The distribution for the 6 clusters of nations in Water Quality trend (effects on human health) and the Average GDP per capita
growth rate in the win-win (upper right), win-loss (bottom right), loss-win (upper left) and loss-loss (bottom left) categories.

Ecosystem Vitality: Biodiversity and Habitat, Growth and HDI 2000
This category shows the trends of nations in their association between
Biodiversity and Habitat (BH) and the Average GDP per capita over the decade. This
category had no nation below 0 which indicates no declining performance in BH trend.
Clusters 1, 3, 4, 5 and 6 have nations in the win-win category (Table 2.4, Figure 2.4).
Cluster 6 has 16 nations with very high win-win and the most improved BH trend
consisting of very high (55.6%) HDI nations like France (FRA), Taiwan (TWA), Belgium
(BEL) and a few high (33.3%) and medium (22.1%) HDI nations. This is consistent with
the EKC theory that environmental improvements will occur with further economic
growth of from developed nations assuming they have past the turning point. Cluster 4
also has 18 nations with a win-win but not as high as cluster 6 consisting of a mix of low
to very high HDI nations like Canada (CAD), Australia (AUS), Uganda (UGA) and
Guatemala (GTA). Here the low HDI nations are not consistent with what the EKC model
predicts except if their turning point is low. These nations should be deteriorating their
ecosystem as their standard of living increases if the turning point is high. Since no nation
had a negative BH trend over the decade, we conclude that Clusters 1, 5 and 3 with a total
of 101 nations have a win-win trend but with very low performances while most of the
nations remained unchanged in their BH trend. They consist of mainly medium and low
HDI nations and only a few high and very high HDI nations and includes mostly SubSaharan an Caribbean countries and some East Asian countries as well. Cluster 2 has 13
nations in a loss-0 situation which showed a declining trend in the Av GDP and no
change in their BH with mainly medium and low HDI nations.

37
Table 2.4: The mean and p-values for Biodiversity and Habitat trend and the Average GDP
per capita growth rate for the 6 clusters of nations in the different categories. The HDI 2000
shows the rank of the nations in percentage for each cluster.

Variables
EV-BH mean
p-values
Av growth mean
p-values
HDI 2000 (%)
Very high
High
Medium
Low
Count

1
Win-Win
1.63
0.003
5.50
<0.001

2
Loss-0
37.07
<0.001
0.34
0.529

3.0
15.2
39.4
42.4
33

61.5
15.4
15.4
7.7
13

Cluster
3
4
5
6
Win-Win Win-Win Win-Win Win-Win
3.85
2.60
2.34
34.65
0.072
<0.001
<0.001
<0.001
10.81
2.72
0.38
4.92
<0.001
<0.001
0.013
<0.001
20.0
0.0
60.0
20.0
7

20.3
25.7
35.1
18.9
18

27.9
16.4
14.8
41.0
61

55.6
33.3
22.1
0.0
16

In this category, we see few nations with very high HDIs in a win-win situation in
Biodiversity and Habitat. Overall this was an unexpected result since nations of the world
improved or had no change in biodiversity and habitat while their economies grew.

50
KIR

SVN

MCO

MDV

BEL
LTU

40

BGR

BHR

GBR

ALB

TWN
NZL

ROU

30

GAB

Clusters

FRA
HRV

Cluster 1
Cluster 2

PER
ARE

Cluster 3
TLS

Cluster 4

ITA

20

Cluster 5
LVA

GTM

Cluster 6

BRA
NAM
EGY

PRT
MEX
GRC

NOR
JPN

UGA
TZA

AUS
CAN

10

COL

PRY
GUY

BHS
PLW

0

DNK

NIC
ISL
SWE
SEN
CHE
IRL
AIA
NLD

KOR
IRN
MMR SVK
HND
URYBFA

ZAF
BLZ
SYR

IDN

SRB
CUB
VNM

LBR
ARM
CHN

AZE

MOZ

MLT FIN
CMR
ETH
VUT
SGP
GMB
BOL
SUR
MLI
BGD
QAT
BIH
SLV
RUS
HUNPHL
CHL
RWA
USA
ESP
DZA
MKD
JOR
AND
WSM
SDN
KHM
CRI
POLZMB
UKR
HTI
SMR
SLBCIV
TGOBRN
BDI
MDG
SOM
COM
CAF
KWT
JAM
NER
TUV
GIN
GNB
BRB
PRI
FJI
B
COD
EN
FSM
TON
PNG
SWZ
SYC
DEU
OMN
KEN
AUT
MWI
YEM
CYP
MRT
GRD
VEN
BMU
LUX
ISR
IRQ
ECU
SAU
STP
MHL
DJI
TCA
PAK
DMA
GRL
LBN
NPL
BWA
TUR
ARG
GHA
LBY
MYS
CZEDOM
KGZ
SLE
LSO
TUN
MNE
NRU
MARPAN
MUS
THA
HKG
LKA
EST
MNG MDA
UZB
LAO
AFG
TTO
CPV
INDGEO
BTN
NGA
TJK
TCD
TKM

CYM

-5ZWE

ERI
ABW

0

5

AGO

KAZBLR

MAC

10

GNQ

15

Avg_growth

Figure 2.4: The distribution for the 6 clusters of nations in Biodiversity and Habitat trend and the Average GDP per capita growth rate in the
win-win (upper right), win-loss (bottom right), loss-win (upper left) and loss-loss (bottom left) categories.

Ecosystem Vitality: Agriculture trend, Growth and HDI 2000
Fig 2.5 shows the trends of nations in their association between Agriculture and
the average GDP per capita over the decade. The win-win category which is the upper
right quadrant contains nations that score well on both indicators. Clusters 1, 2, 5 and 6
have nations in a win-win situation (Table 2.5, Figure 2.5). Cluster 5 has 32 nations in a
win-win situation with the most improved Agriculture trend consisting of nations with
mainly very high (53.1%) and a few high (18.8%), medium (15.6%), low (12.5%) HDI
nations like Finland (FIN), Thailand (THA), Romania (ROU) and Egypt (EGY). This
finding is consistent with the EKC theory. However, Clusters 1 and 2 have a total of 126
nations in a win-win situation too but not as high agriculture improvements as Cluster 5
and are mainly low, medium and high HDI nations some of which include Central Asian,
East Asian, Latin American and Caribbean countries. Cluster 2 dominates Cluster 1 in
that it has higher economic growth while agriculture trend improvements are similar on
average. Cluster 6 has 6 nations with a 0-win situation with the highest average GDP
trend in this category indicating no change in their Agriculture trend are mainly medium
HDI nations. This shows again that economic growth can be achieved without causing a
deterioration of the environment for developing nations. Cluster 3 is a loss-win trend for
15 nations with mainly low HDIs consistent with the EKC theory. Cluster 4 reveals a 0loss trend for 16 nations with decline in their economies and no change in their
Agriculture trend and consists mainly of low HDI nations and a few medium and very
high HDI nations. Seen from a perspective of positive economic growth and clusters with
mainly no change in agriculture sector improvements, cluster 6 is best while the worse
cluster is 4.

40
Table 2.5: The mean and p-values for Agriculture trend and the Average GDP per capita
growth rate for the 6 clusters of nations in the different categories. The HDI 2000 shows the
rank of the nations in percentage for each cluster.

Variables
EV-AG mean
p-values
Av growth mean
p-values
HDI 2000 (%)
Very high
High
Medium
Low
Count

1
2
Win-Win Win-Win
3.56
1.11
0.001
0.026
4.73
1.57
<0.001
<0.001
6.4
25.5
31.9
36.2
47

20.3
21.5
34.2
24.1
79

Cluster
3
4
5
6
Loss-Win 0-Loss Win-Win 0-Win
-28.46
0.95
30.79
-0.63
<0.001 0.333
<0.001 0.363
3.76
-1.61
2.21 10.32
<0.001 <0.001
<0.001 <0.001
0.0
13.3
26.7
60.0
15

25.0
25.0
6.3
43.8
16

53.1
18.8
15.6
12.5
32

In this category, the nations with high and very high HDIs have a very high winwin compared to nations with low to medium HDIs which also had a good win-win.

16.7
0.0
66.7
16.7
6

Figure 2.5: The distribution for the 6 clusters of nations in Agriculture trend and the Average GDP per capita growth rate in the win-win (upper
right), win-loss (bottom right), loss-win (upper left) and loss-loss (bottom left) categories.

Ecosystem Vitality: Forestry trend, Growth and HDI 2000
In this category, due to its set up in the pilot trend index, the lowest performance
benchmark is -50 while the highest target is 0 so nations above -50 approaching 0 are
nations that are not doing as bad since they are closer to the target which is 0. This skew
the computation in this category, hence the negative means (Table 2.6, Figure 2.6). Given
the lowest performance benchmark as -50, all nations were below 0 for the Forestry
category, all clusters are in a loss-win situation (Table 2.6) with some nations in the LossLoss segment. Clusters 5 and 2 contains nations in a loss-win situation with the smallest
loss in Forestry trend. Cluster 5 has 37 nations with mainly medium (43.2%) HDI and a
few high (29.7%), low (21.6%) and very high (5.4%) HDI nations which consists of some
South and Central Asian countries. Cluster 2 has 68 nations with mainly very high
(36.8%), high (25%), medium (22.1%) and a few low (16.7%) HDI nations which
contains mainly European countries. Finally, cluster 5 dominates cluster 2 in that it has
faster growing economies but a similar forest loss on average with cluster 2 whose
economy is slower. Clusters 4 does not have significant economic growth rate (p = 0.67)
as many nations in that cluster fall in the loss-loss area. Cluster 4 is performing worse
than cluster 2 in terms of forest trend loss. Cluster 3 have a total of 56 nations in a lossloss and has higher forest losses than those of clusters 5 and 2, they are mainly medium
and low HDI nations. Cluster 6 has 23 nations with the highest loss-loss with mostly low
and medium HDIs.

43
Table 2.6: The mean and p-values for Forestry trend and the Average GDP per capita
growth rate for the 6 clusters of nations in the different categories. The HDI 2000 shows the
rank of the nations in percentage for each cluster.
Variables
EV-Forests mean
p-values
Av growth mean
p-values
HDI 2000 (%)
Very high
High
Medium
Low
Count

1
Loss-Win
-5.50
0.296
12.10
0.010
0.0
0.0
66.7
33.3
3

Cluster
2
3
4
5
6
Loss-Win Loss-Win Loss-Win Loss-Win Loss-Win
-3.88
-22.71
-15.38
-4.75
-33.75
<0.001
<0.001
<0.001
<0.001 <0.001
1.61
4.53
0.14
5.03
1.58
<0.001
<0.001
0.670
<0.001 <0.001
36.8
25.0
22.1
16.2
68

0.0
15.4
38.5
46.2
26

26.7
13.3
16.7
43.3
30

5.4
29.7
43.2
21.6
37

In this category, based on the highest performance score as 0, high to very high
HDI nations have a lower loss-loss trend while low to medium HDIs show a higher lossloss trend. Nations that have no change in trend forest loss can be considered as a win and
those are in cluster 5 and 2.

4.3
13.0
30.4
52.2
23

0
ARE

CYM
GAB
SMR

KWT
KIR
TUV
BRB
GUY
GMB
TON
SYC
OMN
MLT
YEM
TWNAND
IRQMHL
DJI NPL
WSM
LBY
EGY
KGZ
LSO JOR
IRN
NOR
ISL
HUN
SVNMKD SUR
MNE
NLD

LUX

CRI

JPN
CIV

IRL

HRV

STP

VUT
CYP ISR
DNK AIA
GBR FIN
FRA
GRC
LBN
CODESP
FJI
BMU TCA
GRD
SWE
CAN
CHL
AUT
GNB USACHE
SYR
MLI
ITA
URY
KEN
JAM
PNG
BWA
BFA
CMR BEL
SOM

BHS

HTI

CUB
ROU LVA

ERI

PLW

SWZ
SEN
PRIFSM
GIN
MEX

BRN

MDG

THA

EST

MAR

LBR

KAZ
GNQ

BIH
MOZ

RWA

TJK

Clusters

TCD
TTO

BRA

Cluster 1

ALB

MNG

IDN

DMA
SAU

PAN
SLE TZA
ARG

Cluster 2
Cluster 3

LKA

NAM

-30

AGO

VNM

SGP

PER

DEU

BEN

UZB

KORSVK
BGD
ZMB

BLZ COL
HND

MRT

ZWE

CHN

TUN

NZL

SLB

BLR

GEO
AFG
UKR
RUS

ZAF

-20

AZE

MUS

CZEDOM
SDN
TUR
POL
PHL

CAF

-10

SRB
MDV
MDA
BGR
LTU
CPV BTN TKM
IND

LAO

VEN
BOL

Cluster 4

ETH
ARM

Cluster 5

MWI
BDI

TGO

Cluster 6

SLV

COM

TLS

PRT

NGA

GTM

-40

GHA

AUS

KHM
UGA

DZA

MMR

PAK
NER

ECU

PRY

MYS
NIC

-50
-5

0

5

10

15

Avg_growth

Figure 2.6: The distribution for the 6 clusters of nations in Forestry trend and the Average GDP per capita growth rate in the win-win (upper
right), win-loss (bottom right), loss-win (upper left) and loss-loss (bottom left) categories.

Ecosystem Vitality: Fishery trend, Growth and HDI 2000
Fig 2.7 shows the trends of nations in their association between Fishery and the
Average GDP per capita over the decade. Cluster 2 is a win-win trend with the highest
Fishery trend (Table 2.7) and a high average growth rate (3%). It has a total of 23 nations
spread across high (43.5%), medium (34.8%) and a few very high (13%) and low (8.7%)
HDIs and are mostly Eastern European countries. Clusters 3, 5 and 6 have nations in a
loss-win category with cluster 5 being the worst Fishery trend. Cluster 3 has 51 nations
with mostly very high HDIs which is not what EKC theory would predict. Cluster 5 has
34 nations with mostly low HDIs and consistent with EKC theory. Cluster 6 has 32
nations with mostly medium HDIs and consistent with EKC theory given that most of
these nations are developing with very high economic growth rate. Clusters 3, 5 and 6
have a total of 107 nations in a loss-win and is greater than the total nations in the winwin situation, they consist of some South Asian, Middle Eastern, European and SubSaharan African countries showing mostly consistency with the EKC theory. Cluster 4
has 10 nations is a win-loss situation with mainly low HDIs.
Table 2.7: The mean and p-values for Fishery trend and the Average GDP per capita
growth rate for the 6 clusters of nations in the different categories. The HDI 2000 shows the
rank of the nations in percentage for each cluster.

Variables
EH-Fisheries mean
p-values
Av growth mean
p-values
HDI 2000 (%)
Very high
High
Medium
Low
Count

Cluster
1
2
3
4
5
6
Win-Win Win-Win Loss-Win Win-Loss Loss-Win Loss-Win
8.78
18.68
-1.79
1.74
-23.28
-4.69
0.103
<0.001
0.028
0.573
<0.001
0.002
9.76
3.04
1.16
-1.73
1.92
4.56
0.004
<0.001
<0.001
0.003
<0.001
<0.001
25.0
0.0
25.0
50.0
4

13.0
43.5
34.8
8.7
23

43.1
17.6
21.6
17.6
51

20.0
30.0
0.0
50.0
10

14.7
20.6
26.5
38.2
34

In this category, only few nations are in the win-win are mainly nations with high
and medium HDIs. Nations with very high HDIs did not have a really good trend in the
Fishery trend.

9.4
25.0
43.8
21.9
32

Figure 2.7: The distribution for the 6 clusters of nations in Fishery trend and the Average GDP per capita growth rate in the win-win (upper
right), win-loss (bottom right), loss-win (upper left) and loss-loss (bottom left) categories.

Ecosystem Vitality: Change in Water Quantity trend, Growth and HDI 2000
In this category (Figure 2.8), also due to its construction in the pilot trend index,
its lowest performance benchmark is -50 while its target score is 0 so nations on the 0
score are nations that have achieved the target and have a good performance. This also
skews the computation in this category, hence the negative means (Table 2.8). Given the
lowest performance benchmark as -50, all nations were below 0 for the Water Quantity
category, are in a Loss-win situation (Table 2.8). Cluster 2 has 27 nations in a loss-win
situation with the lowest Water Quantity loss trend (Figure 2.8), it consists of mainly low
(40.7%), medium (33.3%) and a few high (11.1%) and very high (14.8%) HDI nations
consisting of most Sub-Saharan African countries. The type of Cluster 2 nations is not
consistent with the EKC theory. Clusters 1 and 3 are also in a loss-win situation but not as
low as cluster 2 in water quantity loss (Table 2.8). Cluster 1 is a loss-win also with high
Water Quantity loss and very little economic growth and has 48 nations with mostly very
high and low HDIs. Cluster 3 with similar loss to cluster 1 but much higher economic
growth has 44 nations with mostly low and high HDIs. These nations include some
Southeastern European, Sub-Saharan African and Caribbean countries. Clusters 4 and 5
have a total of 60 nations with the worst Water Quantity trend with mainly medium and
low HDI nations and a few high and very high HDI nations consisting of some European
and Central Asian countries. Cluster 5 has a faster economic growth than cluster 4.
Cluster 6 has the highest economic growth and in the middle of the pack in terms of water
quantity loss.

48
Table 2.8: The mean and p-values for Change in water quantity trend and the Average GDP
per capita growth rate for the 6 clusters of nations in the different categories. The HDI 2000
shows the rank of the nations in percentage for each cluster.

Variables

EV-Water mean
p-values
Av growth
mean
p-values
HDI 2000 (%)
Very high
High
Medium
Low
Count

1
LossWin
-30.89
<0.001
0.64

2
LossWin
-11.68
<0.001
1.23

Cluster
3
4
LossLossWin
Win
-28.54
-43.80
<0.001
<0.001
4.05
2.22

0.007

<0.001

<0.001

<0.001

<0.001

0.014

39.6
6.3
14.6
39.6
48

14.8
11.1
33.3
40.7
27

4.5
34.1
22.7
38.6
44

27.0
35.1
27.0
10.8
37

4.3
13.0
60.9
21.7
23

0.0
0.0
50.0
50.0
2

5
LossWin
-40.85
<0.001
5.83

6
LossWin
-24.98
0.467
13.32

In this category, few nations with low and medium HDIs have a lower loss-win
than nations with high and very high HDIs which were also in a loss-win.

Figure 2.8: The distribution for the 6 clusters of nations in Change in water quantity trend (effects on ecosystem) and the Average GDP per
capita growth rate in the win-win (upper right), win-loss (bottom right), loss-win (upper left) and loss-loss (bottom left) categories.

Ecosystem Vitality: Climate Change trend, Growth and HDI 2000
Fig 2.9 shows the trends of nations in their association between Climate Change
and the Average GDP per capita over the decade. Cluster 1, 3 and 4 consists of nations
with a win-win trend (Table 2.9). Cluster 4 has 21 nations in the win-win category with
the highest Climate Change trend which are mostly very high (71.4%) and a few high
(14.3%), medium (9.5%) and low (4.8%) HDI nations like Canada (CAD), USA,
Singapore (SGP). Clusters 1 and 3 also have nations in a win-win situation but not as high
as cluster 4 in terms of climate change trend improvement. Cluster 1 has high climate
change trend and high economic growth. It consists of 22 nations with mainly high and
medium HDIs some Southeastern Europe and Central Asian countries. Cluster 3 has 38
nations with mainly medium and very high HDIs and consists of some European and
Latin American countries. Clusters 2 and 5 have a total of 49 nations in the loss-win
category which are mostly medium, low and high HDI nations consisting of some Middle
Eastern and South Asian countries. Cluster 6 has 3 nations in a win-loss situation which
are United Arab Emirates (ARE), Zimbabwe (ZWE) and Eritrea (ERI).
Table 2.9: The mean and p-values for Climate change trend and the Average GDP per
capita growth rate for the 6 clusters of nations in the different categories. The HDI 2000
shows the rank of the nations in percentage for each cluster.

Variables
EV-CC mean
p-values
Av growth mean
p-values
HDI 2000 (%)
Very high
High
Medium
Low
Count

1
Win-Win
13.40
<0.001
6.01
<0.001
0.0
27.3
45.5
27.3
22

Cluster
2
3
4
Loss-Win Win-Win
Win-Win
-22.10
5.77
28.76
<0.001
<0.001
<0.001
1.71
1.58
1.91
<0.001
<0.001
<0.001
12.0
28.0
36.0
24.0
25

31.6
18.4
31.6
18.4
38

71.4
14.3
9.5
4.8
21

5
Loss-Win
-11.76
<0.001
4.72
<0.001

6
Win-Loss
8.95
0.233
-4.26
0.018

8.3
29.2
37.5
25.0
24

0.0
33.3
0.0
66.7
3

In this category, most nations with very high HDIs achieved a higher win-win
which is in accordance with the EKC predictions, but we also find that some very high
HDI nations were in a loss-win situation either performing very well in their climate

51
change or really bad in their climate change trend even while their economies improve.
Some nations with low to high HDIs also fell in a win-win category.

Figure 2.9: The distribution for the 6 clusters of nations in Climate Change trend and the Average GDP per capita growth rate in the win-win
(upper right), win-loss (bottom right), loss-win (upper left) and loss-loss (bottom left) categories.

Ecosystem Vitality: Air Pollution trend, Growth and HDI 2000
Fig 2.10 shows the trends of nations in their association between Air Pollution and
the Average GDP per capita over the decade. Cluster 2, 3 and 4 consists of nations with a
win-win trend (Table 2.10). Cluster 4 has 21 nations in a win-win situation with the
highest Air Pollution trend relatively spread between the very high (28.6%), high
(28.6%), medium (19.1%) and low (23.8%) HDI nations like Hungary (HUN), Nigeria
(NGA), Slovenia (SLV) and Latvia (LVA). Clusters 2 and 3 also have nations in the winwin situation but not as high as cluster 4 (Table 2.10). Cluster 2 had 54 nations in a winwin with mostly very high and a few high and medium HDI nations consisting of mostly
European and some Middle Eastern countries. Cluster 2 has 54 nations in a win-win with
mostly very high and high HDI nations and also very few low HDI nations which consists
of some European, Latin American and Caribbean countries. Cluster 3 has 27 nations in a
win-win with mostly medium and high HDI nations and only few low HDI nations which
consists of some Middle Eastern and Central Asian countries. Clusters 1 and 6 has a total
of 25 nations in a loss-win situation with mostly medium HDI nations and only a few
very high, high and low HDI nations consisting of some Sub-Saharan African and Central
Asian countries. Cluster 5 has 6 nations in a win-loss category with mostly low HDI and
very few high HDI nations.
Table 2.10: The mean and p-values for Air pollution trend and the Average GDP per capita
growth rate for the 6 clusters of nations in the different categories. The HDI 2000 shows the
rank of the nations in percentage for each cluster.

Variables
EV-Air mean
p-values
Av growth mean
p-values
HDI 2000 (%)
Very high
High
Medium
Low
Count

1
2
Loss-Win Win-Win
-9.51
11.53
<0.001
<0.001
3.01
1.77
<0.001
<0.001
18.2
18.2
45.5
18.2
22

40.7
20.4
24.1
14.8
54

Cluster
3
4
5
6
Win-Win Win-Win Win-Loss Loss-Win
5.37
39.92
12.28
-6.57
<0.001
<0.001
0.010
0.734
5.29
2.82
-2.60
10.60
<0.001
<0.001
0.022
0.020
0.0
33.3
40.7
25.9
27

28.6
28.6
19.0
23.8
21

0.0
33.3
0.0
66.7
6

0.0
0.0
100.0
0.0
3

54
This category shows only few nations spread across the very high, high, medium
and low HDIs in a very high win-win.

Figure 2.10: The distribution for the 6 clusters of nations in Air pollution trend (effects on ecosystem) and the Average GDP per capita growth
rate in the win-win (upper right), win-loss (bottom right), loss-win (upper left) and loss-loss (bottom left) categories.

Discussion
How nations responded and regrouped in the different categories of win-win, winloss, loss-win and a loss-loss situation varied across all the indicators in the
Environmental Health and Ecosystem Vitality policy objectives. The win-win results for
nations is consistent with the findings in Hsu, (2013) which also found that an improving
environmental performance trend varied by the issue, country and region. A higher
number of nations fell into a win-win trend for Child Mortality, Water Quality (human
health effects), and Air Pollution (ecosystem effects). Forestry and Water (ecosystem
effects) were found mostly in the loss-win area which is a serious concern and policy can
be directed towards moving out of the environmental loss to a win without potentially
hurting the economy.
Environmental Health
In the environmental health category which evaluates impacts on human health,
the nations which had a high increasing trend in the win-win for the policy category of
Child mortality, Water quality and Air quality consists more of nations with low and
medium HDIs. Although some nations with high and very high HDIs were still in the
win-win category, they do not show as much increasing trend as the low and medium
HDI nations.
The most improved Child Mortality trend were 29 nations in a win-win with
mostly low and medium HDIs consisting of Middle-eastern and North-African countries
like Bangladesh (BGD), Rwanda (RWA), Mongolia (MNG) etc. These nations which are
not high-income nations progressed significantly in this category which suggests that
their progress may not be a function of only income. Some of the nations in the win-win
may have reached diminishing returns and their economy doesn’t grow as fast as
developing nations and their child mortality is very low to start and hence improvements
are only marginal.
The most improved Air Pollution trend had 34 nations which made the most
progress in a win-win and had mostly high, medium and low HDI nations like Mexico
(MEX), Malaysia (MYS), Slovak Republic (SVK), Zambia (ZMB) and Cameroon
(CMR). Dinda (2004) documents evidence that mostly local air pollutants support the
EKC model. Since local urban air quality indicators such as SO2, SPM, CO, NOx directly
affects human health, the inverted U shape EKC model holds. Pollutants that directly
affect our human health have turning points that are very low and thus most nations,

57
developing and developed, can achieve simultaneously win-win with the former growing
faster. According to Dinda (2004), the estimated turning point for most pollutants is in the
range of US$3000–10,000 (at a constant price, 1985 US dollar) and most developing
nations are within that range. However, the evidence on pollutants that have little impact
on our health (e.g., C02) evidence of EKC is not supported potentially because the turning
point has not yet been reached. The highest improved Water Pollution trend had 23
nations with mostly medium HDIs like Ecuador (ECU), Uruguay (URY), Egypt (EGY)
etc.
These results for the Environmental health indicators are similar to some previous
studies like Stern (2004, 2017) that finds little support for the existence of an EKC which
suggests that beyond a certain income threshold, only wealthier countries can impact their
environmental issues positively (Grossman and Krueger, 1995). We see low and middleincome nations taking the lead for the highest improving trends in Environmental Health
category. This is also probably due to a low turning point as well as social and political
factors. Improvements in child mortality starting from a low turning point of income per
capita makes sense since child’s health is a priority for parents and governments. B
O’Hare et al. (2013) conducted a meta-analysis of numerous studies for developing
nations on the association between income and child mortality and found that income is
the most important determinant of child mortality with no turning point. They concluded
that if the GDP per capita (PPP adjusted) increases by 10%, the infant mortality will
decrease to 45 per 1000 live births from 50 per 1000 (i.e., a 10% reduction in infant
mortality).
However, consistent with the findings in Emerson et al. (2012), for developed
nations which have top EPI rankings and very high HDIs over the years, the EPI and
Average GDP trend results may not be particularly meaningful because many of these
longtime leaders have limited room for improvement. Nations like USA, Canada (CAD),
Australia (AUS), Slovenia (SVN) Iceland (ISL), etc. which have very high EPI ranking
each year will have difficulties achieving large gains in trends. This finding is also
peripheral to the macroeconomic growth convergence literature which suggests that
developing economies' per capita incomes will tend to grow at faster rates than developed
economies, which will result in all economies converging in terms of per capita income.
Nevertheless, some very developed nations like South Korea (KOR), France (FRA),
United Kingdom (GBR) still had high win-win trends in the Environmental Health
indicators reflecting improved performance over the decade.

58
Ecosystem Vitality
In the Ecosystem vitality category which examines effects on the ecosystem, the
performances of nations varied significantly for each policy categories. However, nations
with very high HDIs mostly had a high win-win in categories of Air pollution, Climate
change and Agriculture.
Although there were more nations in a win-win, the highest Agriculture trend
consisting of 32 nations were mainly very high HDI nations like Finland (FIN), Germany
(DEU), Slovenia (SVN) taking the lead. A few medium and high HDI nations like Egypt
(EGY), Romania (ROU) and Thailand (THA) also scored high in Agriculture. The gains
in their trends for these nations is largely due to their improvements in agricultural
subsidies (Emerson et al., 2012). Performance in the Fishery category seem to be weakest
in gaining a win-win situation which is similar to the results found in Hsu et al. (2013).
The highest Fishery trend had only 23 nations with a win-win which were nations with
mostly high and medium HDIs and only a few very high and low HDI nations. Nations
like Estonia (EST), Lithuania (LTU), Croatia (HRV), Vanuatu (VUT) taking the lead in
Fishery due to their improvements in coastal fishing shelf pressure (Emerson et al., 2012).
Over 50% of the total nations in Fisheries were in a loss-win category from very high to
low HDI nations like Kuwait (KUW), Monaco (MCO), Russia (RUS), Egypt (EGY) etc.
and mostly had performance declines related to over-fishing. Some low HDI nations also
had difficulties monitoring and controlling the fishing within their exclusive economic
zone (EEZs) while some countries under-report their fish catches (Emerson et al., 2012).
Climate Change category had relatively higher number of nations in a win-win
than in a loss-win, the highest Climate Change trend had 21 nations mostly very high
HDI nations like Canada (CAD), USA, Singapore (SGP) taking the lead. CO2 emissions
correspond strongly to GDP but according to the International Energy Agency (IEA)
(2011), CO2 emissions grew faster than real GDP in 2010 which highlighted the need to
be aware of emissions data to make judgments on current policies and future action plans.
The results for Climate Change is consistent with previous studies like Jessberger (2011).
A few high, medium and low HDI nations like Angola (AGO), Albania (ALB),
Turkmenistan (TKM), Tajikistan (TJK) were also in the same win-win category as the
very high HDI nations. However, many developed nations of very high HDIs are also
found in the lowest Climate Change loss-win category like Taiwan (TWN), Luxembourg
(LUX), South Korea (KOR), Norway (NOR). The findings for Climate Change are quite

59
opposite to the results for Emerson et al. (2012) which found that developing countries in
South Asia and Sub-Saharan Africa perform better in Climate Change than more
developed countries in the Middle East, North Africa and North American regions. The
high win-win trend for Climate Change may also be due to the level of good governance
in the developed countries which have the ability to enforce environmental regulations.
The highest Air Pollution trend had 21 nations in a win-win with more of the very
high and high HDI nations like Hungary (HUN), Slovenia (SVN), South Korea (KOR),
Singapore (SGP), Latvia (LVA) etc. taking the lead. A few medium and low HDI nations
like Nigeria (NGA), Angola (AGO), Cameroon (CMR), Algeria (DZA) also in a high
win-win as the very high HDI nations. This suggests that while income plays a significant
role in improving a nation’s performance, there may be other factors which can also help
to explain some differences in the countries’ performances. The issues of renewable
electricity generation, for which some countries have poor scores have been known to be
tied to challenges with policy processes and choices. It is also important to point out that
though these results present a win-win situation for some nations in the Climate Change
and Air pollution policy categories, but on a global scale, notably climate change, has
declined (Emerson et al., 2012).
Change Indicators: Water and Forestry
The Trend EPI used available historical data to measure performance changes
from 2000 to 2010 but in cases of Water (ecosystem effects) and Forestry, no time series
was available because the indicators themselves are change variables (e.g. Forest Loss,
Forest Growing Stock, Forest Cover and Change in Water Quantity). Their trend scores
range from -50 to 0 (the target is 0% change) and could be used directly to determine the
rate of improvement or decline for each indicator (Emerson et al., 2012). The lowest
Water Quantity loss trend with 27 nations in a loss-win situation had mainly low and
medium HDI nations like Malawi (MWI), Guinea-Bissau (GNB), Guatemala (GTM)
taking the lead. These low HDI nations had a reduced Water Quantity loss as a result of
reduced pressures of water abstraction on aquatic ecosystems which most very high HDI
nations struggle with (Vorosmarty et al., 2010). The Forestry category had 105 nations
with the smallest loss in Forestry trend with mainly medium and very high HDI nations
but also had several high and low HDI nations which was not quite expected for a
declining trend in Forestry given the increasing levels of deforestation. However, as a
result of differences in data collection methodology, there were significant variations in

60
data quality between countries. For example, some countries were allowed to choose
what they consider to be a minimum tree size for inclusion in the growing stock measure
while some countries simply lack the resources to conduct regular forest surveys. The
measure used to represent the change in growing stock for Forest categories covered one
five-year period to the next (2000-2005, 2005- 2010) and considers the target to be zero
change (Emerson et al., 2012).
This indicates that nations in the loss-win trend with the smallest Forestry loss
trend did not necessarily improve but rather did not decline any further and while
countries that had improving trends were not rewarded, those that are losing forest cover
are penalized. The lack of long-term monitoring systems to regularly assess the condition
of forests was also one of the major barriers to establishing sustainable forest practices
and available data (Hsu et al., 2013).
Biodiversity and Habitat
In the case of Biodiversity and Habitat, there was a lack of accurate country-level
data on species abundance and little consistent information on the management of
habitats and the sustainable use of species (Emerson et al., 2012). In the 2012 EPI, the
targets include measures of protected area coverage by terrestrial biome (17% weighted
average of biomes protected), area of coastline (10% of country's terrestrial seas and EEZ
protected) and a measure of the protection of highly endangered species (100% of critical
habitats protected). Countries are not rewarded for protecting beyond these targets so that
higher levels of protection cannot be used to offset lower levels of protection, so, a
positive BH trend for this category only reflects the degree to which a country achieves
these targets within its borders (Emerson et al., 2012).
In recent decades, natural habitats have witnessed considerable declines in
biodiversity and many species are at risk of extinction. Costanza et al. (2014) also found
an estimated loss of world eco-services due to land use change at $4.3–20.2 trillion/yr.
from 1997 to 2011. Rockström et al. (2009) also used planetary boundaries framework to
shows a loss in genetic biodiversity which surpassed the threshold level of a safe
operating space for humanity. Only a few very high HDI nations like France (FRA),
Taiwan (TWA), Belgium (BEL) were in a high win-win situation in this category,
however, most nations remained unchanged in their BH trend.

61
Concluding Remarks
In conclusion, the win-win trends reveal improvements for many countries on
indicators evident in the Environmental Health objectives for decreasing Child Mortality
and increasing Water Quality. In the Ecosystem Vitality objective however, there remains
challenges with respect to categories like Climate Change, Fisheries, Forests, Water
(ecosystem) and Air Pollution categories. Countries in the Middle East, Sub-Saharan
Africa, North Africa performed very well in some indicators like the Environmental
Health indicators while Countries from Europe, North America performed better in
Climate Change and Agriculture similar to findings in Hsu et al. (2013).
The findings using cluster analysis also sheds more light on the behavioral
patterns of nations for a broader range of environmental indicators given their various
GDP growth rates which is not seen in the EKC theory. Nations in the win-win category
for indicators like Child Mortality, Water (human health), Water (ecosystem), Air
Pollution (ecosystem) had developing nations with higher improving trends and are
similar to findings in Dasgupta et al. (2002). The EKC which proposes that income
growth, over a certain level reduces environmental degradation can suggest that economic
growth is good for the environment. However, majority of studies have raised criticisms
of the EKC theory in respect to analytical weaknesses and econometric misspecifications
like heteroskedasticity, simultaneity, omitted variables bias and cointegration issues
(Stern, 2004; Millimet et al., 2003; Sobhee, 2004). Other studies have also achieved
mixed outcomes when considering the EKC for several environmental indicators and
pollutions levels and local pollutants (Shafik, 1994; Mckitrick and Wood, 2017; Taylor
and Brock, 2010). Some studies agree with the EKC theory (Gallego et al., 2014;
Mavragani et al., 2016) that economic growth will lead to an improved environment for
developed nations, however, we find that it also applies to developing nations or could
also be opposite.
Applying cluster analytical technique for this study while bearing in mind, past
theories, hypotheses and research of the variables used to identify win-win situations
gives a graphical representation of the role of income in impacting several environmental
indicators. This segmentation of nations into relative groups called clusters has displayed
the behavioural patterns of nations’ environmental performance in response to their GDP
growth rates. This reveals that a very targeted policy and attitude approach for indicator-

62
by-indicator basis should be adopted to make an improving environment compatible to
the economic growth of that nation.
The results especially for the win-win categories using cluster analysis also
warrants closer investigation into the underlying factors that may have led to a higher
win-win even within the win-win groups. It lays the foundation for the next chapter which
looks at some socio-economic variables like non-income HDI, and governance indicators
like change in government effectiveness and political stability and the investment
spending as a fraction of the size of the economy. They will be considered as factors that
likely impact the likelihood of high performance (win-win) nations for environmental
performance and economic growth simultaneously. The high performance will consider
setting thresholds for the win-win categories in both the average GDP and environmental
indicators to see particular nations that have progressed significantly and determine a
possible cause and effect.

63
CHAPTER 3: DETERMINANTS OF HIGH PERFORMANCE NATIONS IN
ENVIRONMENTAL AND ECONOMIC PERFORMANCE USING LOGISTIC
ANALYSIS.
Introduction
In the past, majority of studies have established a clear connection between the
progress in environmental performances of nations with their respective GDP per capita,
a well-known theory being the Environmental Kuznets Curve. The EKC posits that
countries follow a U-shaped path in which environmental degradation initially worsens
with economic development, beyond a certain income threshold, richer countries can
reduce such degradations some of which could be through cleaner technologies and
changed citizen behaviour (Grossman and Krueger, 1995). While many studies have used
the EKC approach to explain the relationship, Shafik (1994) took into account other
determinants of environmental quality. He considered factors such as the impact of rising
industrialization and urbanization at middle-income levels and the growing importance of
services in high-income economies. He argued that income per capita served to measure
directly the relationship between economic growth and environmental quality and
measures indirectly the endogenous characteristics of growth. However, (Stern, 2017)
proposes that proximate variables such as scale of production, composition effect,
technique effect and changes in input mix are a more realistic view of the effect of
economic growth and technological changes on environmental quality. He pointed out
that when diagnostic statistics and specification tests are considered, and appropriate
techniques are used, the EKC does not exist. These proximate variables were also used in
a study conducted by Mckitrick and Wood (2017) and their results revealed no scale
effect for CO after controlling for changes in composition and technique, no composition
effect for SO2 but composition effects for CO and NO2 existed.
Other researchers have linked progress in environmental performance with
improved Human Development Index (HDI) and human capital accumulation as
necessary factors nations must develop to prevent environmental degradation (Constantini
and Salvatore, 2008; Mukherjee and Chakraborty, 2010). Hsu et al. (2013) also examined
the progress in the Millennium Development Goal 7 (MDG7) by associating the progress
in the MDG7 with the non-income HDI, GDP per capita and governance indicators
likewise. Other approaches like the 2012 EPI employs science-based methodologies to set
targets that is applied to all nations to determine their improvement or decline in the

64
policy targets for environmental performance (Emerson et al., 2012). The EPI uses a
transparent approach and has a consistent framework which has been used by
policymakers and the public to see how close or farther away efforts are made to ensure
improved environmental health and ecosystem health. Moran et al. (2007) uses a different
approach to determine sustainable development in nations. Sustainable development
according to Moran et al. (2007) was defined as advancing human well-being, within the
constraint of ecological limits of the biosphere. Their study employed UN Human
Development Index (HDI) as an indicator of development and the Ecological Footprint as
an indicator of human demand on the biosphere. They also set the threshold to be an HDI
of no less than 0.8 (HDI ≥ 0.8) and a per capita Ecological Footprint (EF) less than the
globally available biocapacity per person (Footprint to biocapacity ratio ≤ 1.0) as
minimum requirements for sustainable development that is
globally replicable. Their findings also showed that despite growing global adoption of
sustainable development as a major policy goal, only one of the 93 countries surveyed
met both of these minimum requirements in 2003. They also found that some lowerincome countries achieved higher levels of development without a corresponding increase
in per capita demand on ecosystem resources. The trend for high-income countries on the
other hand had improvements to their HDI which came with a disproportionately larger
increase in Ecological Footprint, showing a movement away from sustainability.
The approach Moran et al. (2007) employed in determining if a nation is making
progress towards sustainability using specific measurements can also be applied to
determine the progress of nations in environmental and economic performance which has
not been looked into broadly. While a precise definition of sustainability may be elusive
according to Carter (2001), it is still possible and sensible to define measurable bottomline conditions for economic development and environmental performance.
The goal of this chapter is to determine the underlying factors that bring
about a high-performance outcome simultaneously for the association between economic
and environmental performance for the period of 2000 – 2010 by setting thresholds for
both factors to determine a high win-win situation. This study uses the GDP per capita
growth rate as the indicator for economic performance with a threshold set at 2% (i.e.,
average GDP growth rate > 2% represents high growth rate) to identify those nations with
a faster growing rate in the economy relative to those that are not. In economic theory, the
GDP growth rate which is an important indicator of economic wealth gives insight into
the general direction and magnitude of growth of the economy which can be positive or

65
negative. Developed nations which are most highly industrialized countries for example
USA, Canada, Japan, Australia have a long-term GDP growth rate per capita around 2%.
Fast-growing economies experience rates of 6-10% (e.g. China over the last 30-40 years
grew at over 6% per year on a per capita basis) which is not likely to be sustainable over
the long term. The 2012 pilot trend EPI which was compiled to allow countries to
examine the improvement or decline in the environmental policy targets from 2000 to
2010 is used as the indicator for environmental performance. The threshold for high
performance in the environmental policy categories used in this study will be determined
separately for each policy category based on their trend scores which is set differently by
Yale University. A nation with a high win-win is the nation that falls within these
thresholds for a positive economic and environmental performance trend.
This study explores the factors that influence the likelihood of achieving high
performances in both areas simultaneously. The initial GDP per capita and initial nonincome HDI (i.e., education and health) are used as control variables possibly associated
with the likelihood of such high-performance nations that meets the thresholds in both
areas. Explanatory factors include the change in government effectiveness and political
stability which are mostly governance indicators over the decade to likely impact the
likelihood of high performance nations in both areas simultaneously relative to other
nations that are not. Also, the investment spending as a fraction of the size of the
economy which is considered one of the engines of economic growth will be considered.
These indicators are selected to not only reflect changes in standard of living, but also to
show if these changes are compatible with their current environmental conditions.
Methods
This study employs logistic regression which is a statistical technique used to
analyse any dataset in which there are one or more independent variables that determine
an outcome which in this case is a high performance for average GDP per capita growth
rate and the 2012 EPI pilot trend policy categories.
Model Specification
The model specification for this study for the outcome of a high-performance in
average GDP per capita growth and the 2012 EPI pilot trend policy categories is as
follows:

66

∆

1

∆

∆

where p is the probability of presence of the characteristic of interest in this case,
likelihood of a high-performance in both areas of economy and environment for any
given nation. The logit transformation is defined as the logged odds:

1
⋯

The likelihood of a high performance depends on thresholds and are used to generate
larger win-win outcomes for both indicators. Throughout all the environmental indicators
being examined, the threshold for a win in average GDP per capita growth is set to be
greater than 2 percent (2%). The dependent variable is the log of the likelihood of a high
win-win

relative to all other cases. This threshold is imposed because most developed

nations like Canada, many northern European nations, the United States have a long-term
growth rate of this magnitude. As for the threshold for each environmental policy, recall
that the 2012 pilot trend EPI which was conducted by Yale university have different trend
scores for each category based on data availability, expertise judgement and standards.
Low performances and high-performance benchmarks vary for some of the categories, so
the threshold for each category is also defined differently. A high performance (high winwin) covers a win for average GDP per capita growth rate greater than 2% and the win for
the environmental indicator as would be defined per category.
The logistic technique is employed in this study to find the best fitting model to
describe the relationship between the dependent variable and a set of independent
variables. Estimation in logistic regression chooses parameters that maximize the
likelihood of observing the sample values unlike in ordinary regression that chooses
parameters that minimize the sum of squared errors. The statistical software used to
conduct the logistic regression analysis is the Econometric Views 10 (EViews10)
software developed by IHS Markit (1994).

67
Variables
The parameters used covers the period of 2000 to 2010. GDP Per capita at
Purchasing Power Parity (PPP) in constant 2005 international dollars from the World
Bank database was used as my indicator for economic performance. It is measured as the
average growth rate of GDP per capita over the decade. The average investment share in
GDP over the same period obtained from the UN Statistics Division is also employed in
this study as an indicator that measures the share of investment in total production for any
nation. The rate of investment reflects the infusion of requisite capital to support the
development process in any given nation. The HDI is a summary measure of human
development that assesses the average achievements in a country in three dimensions of
human development: health, education, and income (Sustainable & We, 2012). Because
income is a component of the HDI and we also considered the initial GDP per capita in
year 2000 as a control variable, the non-income HDI in 2000 (UNDP, 2012) is used
separately to control for the initial state of social development of the nations.
Government effectiveness was selected from the Worldwide Governance
Indicators database because of prior literature that has established the link between
governance and environmental regulation (Hsu et al., 2013). It reflects the perceptions of
the quality of public services, the quality of the civil service and how independent they
are from political pressures. It also reflects the quality of policy formulation and
implementation, and how credible the government's commitment to such policies are
(WGI, 2012). This variable reflects the change in public perceptions from 2000 to 2010.
A positive change indicates perception of improvement in government effectiveness.
Political stability was also selected from the Worldwide Governance Indicators database.
There is a very close link between economic growth and political stability (WGI, 2012).
It measures perceptions of the tendencies of political instability and violence which may
be political in nature. Another indicator used is the Public expenditure on health by the
Organization for Economic Co-operation and Development (OECD) which refers to
expenditure on health care incurred by public funds which constitutes of state, regional
and local government bodies and social security schemes.
The environmental policy categories represent core areas of environmental policy
concern for which measurable indicators can be assessed. The method employs a multistep process to produce indicators on a consistent scale to allow for comparison across
sectors (Hsu et al., 2013). The policy indicators are based on a proximity-to-target

68
methodology. For the purposes of evaluating the trends in the different country’s
performances, the ten policy categories from the 2012 EPI and Trend EPI are used in this
study. The Environmental Health objective measures the impacts on human health in
three policy categories of Air, Water and Human Health (Hsu, 2016). The Ecosystem
Vitality measures the impacts on the ecosystem and natural resources in seven policy
categories of Air, Water, Fisheries, Forests, Climate Change and Energy, Biodiversity
and Habitat, and Agriculture. We have a total of 10 policy categories, 3 from
Environmental Health and 7 from Ecosystem Vitality (Figure 1.2). All data on 2012
Environmental Performance trend index for the 10 policy categories on environmental
health and ecosystem vitality were collected electronically from Yale web portal
(www.epi.yale.edu). All data on GDP per capita for a range of 170 – 230 nations were
electronically retrieved from Gap minder Compiled by Mattias Lindgren for the period of
2000 – 2010 which was used to compute the growth rate.
Two types of independent variables where used to estimate the likelihood of a
high performance for nations in average GDP per capita growth and the 2012 EPI pilot
trend policy categories. First, the 2000 values of GDP in natural log (
non-income HDI (

) are to control for the initial state of the nation. If

and the
< 0 then it

shows that the win-win outcome is more likely to occur if the nation is of low income per
capita initially than rich nations which may be due to diminishing returns for rich nations.
Similar, if

< 0 then win-win is more likely to occur with nations that have a low initial

social development relative to nations that have a high initial score in social development
as measured by health and education outcomes. However, in this case it could possibly be
the case that

> 0 which would indicate a high state of initial social development

increases the likelihood of a win-win outcome.
Once the initial state of the economy is controlled for, the economic and policy
variables will determine the likelihood of a high-performance event relative to the
absence of a high-performance. These remaining variables are the average investment as
a percent of GDP over the 2000-2010 period; the change from 2000 to 2010 in
perceptions of government effectiveness; the change in perceptions of political stability
and in some occasions health change also is considered. A positive change in government
effectiveness or political stability shows a perception of improvement and an increased
likelihood of a high-performance. A higher investment as a percent of GDP should
increase the likelihood of a high-performance also.

69
Thresholds for Trends in Environmental Indicators
The environmental health indicators focus on impacts on human health (Figure
3.1). With the trend scores in child mortality for low and high performances ranging from
-50 to 50, nations above 0 (no change) were the nations with improvements in their child
mortality rates by reducing the number of death between the ages 1-5 over the decade.
The threshold is set to be greater than a trend score of 10 (Child Mortality > 10) because
the EPI had a record of over 100 nations with a positive trend score above 0 (Figure 3.1).
This threshold will capture those nations with a higher improving trend that is well above
0 or minimum improvements. The trend scores in water quality category also ranges from
-50 to 50, the threshold is set to be greater than a trend score of 10 (Water Quality > 10)
as it will reveal those nations with a good improving trend that is above 0 (no change)
(Figure 3.1). The trend scores in air quality is also from -50 to 50, the threshold was set to
be greater than a trend score of 0 (Air Quality > 0) as air quality trends has been shown
from numerous studies to perform badly over the decade with only few nations with an
improving trend (Emerson et al., 2012), hence the low threshold (Figure 3.1).
The ecosystem vitality indicators focus on impacts on the environment (Figure
3.2). The trend scores in Biodiversity and Habitat (BH) ranges from -50 to 50, the
threshold was set to be greater than a trend score of 0 (BH > 0) because this category had
no nation with a negative trend score which was due to the nature of the criteria of
improvement and how it was set (Figure 3.2). Trend scores for Agriculture category spans
-50 to 50 also with 0 representing no change, the threshold was set to be greater than a
trend score of 0 (Agriculture > 0) as there were only few nations with an improving trend
that was above zero (Figure 3.2). Trend scores for the forestry category was set at a range
of -50 to 0 because it is a change variable and 0 was the highest trend score to be
achieved (0% change as target) (Figure 3.2). The threshold for forestry was set to be
greater than -10 (Forestry> -10) as these nations showed a less loss trend towards the
target. The trend scores for fisheries category ranges from -50 to 50. The threshold was
set to be greater than 0 (Fisheries > 0) as only few nations had an improving trend that
was above zero (Figure 3.2). Change in Water Quantity trend scores ranges from -50 to 0
and is also a change variable with 0 as the highest target for reducing Water Quantity use.
The threshold was set to be greater than -25 (Water Quantity use > -25) because many
nations had a significant deteriorating trend, however this threshold will reveal those
nations with a lower loss trend (Figure 3.2). The trend scores in climate change category

70
ranges from -50 to 50, the threshold as set to be greater than 0 (Climate Change > 0) as
those are the nations with an improving trend that is above 0 which is no change (Figure
3.2). The trend scores for Air Pollution category also ranges from -50 to 50, the threshold
was set to be greater than 0 (Air Pollution > 0) as those are the nations with an improving
trend that is above 0 which is no change (Figure 3.2).

-5

0

EH_trend

0

50

40

5

10

15

EHEH_trend

25

20

10
0

0
0

0
-25

-20

-50

EHAIR_trend

60

EHWATER_trend

60

40

40

20

20

0

0

-20

10
0

0
-20

-5

0

2

5

10

15

0

2

Figure 3.1: Thresholds for a high-performance win-win for the Environmental Health
categories and Average GDP per capita growth rate. EHEH_trend is the Child Mortality
trend set at 10, EHAIR_trend is Air Pollution trend set at 0, EHWATER_trend is Water
Pollution trend set at 10 and EH_trend is the general Human health trend set at 0. The
Average GDP growth rate is set at 2 for all the indicators.

0

0

2

EV_trend

20

8

16

0

EVBH_trend
40

40
0

0

-20
0

0

20

0

EVFOREST_trend

0

2

EVAG_trend

EVFISH_trend

50

0

-40
0

0

EVWATER_trend

0

-10
-20

0

-40

-50

0

-20

-25

-40

EVAIR_trend

50

0

0

0

-50

-50
0

8

EVCC_trend

50

16

0

8

16

0

0

2

Figure 3.2: Thresholds of a high-performance win-win for the Ecosystem Vitality categories and Average GDP per capita growth. The
first column consists of a general Ecosystem Vitality trend (EV_trend) set at 0, Forestry trend (EVFOREST_trend) set at -10, Air
Pollution trend (EVAIR_trend) set at 0. The second column consists of Biodiversity and Habitat trend (EVBH_trend) set at 0, Fisheries
trend (EVFISH_trend) set at 0, Climate Change trend (EVCC_trend) set at 0. The third column consists of Agriculture trend
(EVAG_trend) set at 0 and Water Quantity use trend (EVWATER_trend) set at -25. The threshold for Average GDP per capita growth
is set at 2 for all the indicators.

Results
High win-win in economic growth and environmental health
Results of factors influencing the likelihood of a high win-win relative to no win-win is
presented below:
Table 3.1: The coefficients followed by odds ratio and p-values (in parentheses) for the
determinants of a likelihood of a high win-win for the environmental health indicators and
Average GDP per capita using logistic regression. Coefficients with statistically significant pvalues are bolded. The environmental health indicators are Child Mortality (EH-EH), Water
Quality (EH-Water) and Air Pollution (EH-Air).

Constant term
lnGDP2000
HDI2000
Inv/GDP
GE change
PS change
Health change
Observations without win‐
win
Observations with win‐win
McFadden R‐Square
S.E. of regression
LR statistics
Thresholds

EH‐EH
2.341
(0.194)
‐0.753, 0.471
(0.021)
0.004, 1.004
(0.869)
0.160, 1.174
(0.000)
1.058, 2.881
(0.149)
0.822, 2.275
(0.029)
0.061, 1.063
(0.284)

EH‐Water
3.922
(0.036)
‐1.173, 0.309
(0.002)
0.066, 1.068
(0.016)
0.051, 1.052
(0.081)
‐0.328, 0.720
(0.653)
‐0.124, 0.883
(0.743)
0.056, 1.058
(0.321)

EH‐Air
0.788
(0.614)
‐0.303, 0.739
(0.283)
0.008, 1.008
(0.696)
0.027, 1.027
(0.291)
0.130, 1.139
(0.844)
0.670, 1.954
(0.054)
0.018, 1.018
(0.737)

84
65
0.250
0.422
56.82
GDP>2,
EH>10

106
41
0.097
0.431
16.86
GDP>2,
WATER>10

100
50
0.051
0.466
9.82
GDP>2,
AIR>0

Child Mortality
The win-win category in child mortality had a total of 65 out of 149 that met the
threshold for a high win-win in GDP per capita growth rate and health improvement (Table
3.1). There were only 8 high-income nations out of the 69 that met the threshold of a highwin-win which includes Panama, Slovak Republic, Slovenia, Singapore, Saudi Arabia, South

73
Korea, Hong Kong China, Latvia and Argentina. The majority of nations were the low and
middle-income nations consisting mainly of Central Asian, Sub-Saharan African, Latin
American, Caribbean and Middle-Eastern nations like Afghanistan, Angola, Nigeria,
Columbia. This is a statistically significant amount of improvement in both areas and
especially for low and middle-income countries which is supported by the theory of
economic growth that a faster growth rate is seen mostly for developing economies as
opposed to developed economies. The initial GDP (p < 0.05), Average investment as a % of
GDP (p < 0.001) and Political Stability change (p < 0.05) were the statistically significant
variables impacting the high win-win trend. A statistically significant initial GDP but with a
negative coefficient is most likely due to convergence hypothesis. The odds ratio for this
variable implies that initial GDP is not a strong factor for win-win because it is 0.471 times
less likely going to increase a win-win relative to no win-win. This may be due to developed
nations reached high levels of both standard of living and low child mortality. Furthermore,
the initial social development indicator is not statistically significant. Hence, a win-win in
this case is more dependent on the initial standard of living of a nation as measured by GDP
per capita in 2000 than the initial level of education and health score in the HDI. However,
after controlling for the initial state of the socioeconomic situation of the nations, we find that
the average investment as a percent of GDP shows that it increases the odds of being in a
win-win 1.174 times more than the odds of not obtaining a win-win for child mortality and
economic growth. Similarly, political stability change is a strong indicator for improvement
in child mortality and economic performance by increasing the chances of a win-win 2.275
times more that the odds of having no win-win. The remaining variables such as initial nonincome HDI and government effectiveness were not statistically significant in this category.

Water (human health effects)
The win-win trend for water quality had 41 out of 147 nations meeting the threshold
for a high win-win for GDP and water quality (Table 3.1). There were only 5 high-income
nations out of the 41 in a high win-win which includes Uruguay, Panama, Trinidad and
Tobago, South Korea and Hungary. The majority were low and middle-income nations
similar to the nations in the Child Mortality but mostly Latin American, Caribbean and
Middle-Eastern nations like Afghanistan, Cuba, Egypt and Brazil. This also shows a
statistically significant improving trend in both areas especially for low and middle-income
countries which is also consistent with the theory of economic growth as it relates to a faster

74
growing rate for developing economies. Initial GDP (p < 0.005), Average investment (p <
0.05) and initial HDI (p < 0.1) were the statistically significant variables impacting a high
win-win trend. The odds ratio for this category shows that the level of impact by the Initial
GDP is 0.309 times less likely to result in a high win-win which indicates it is not a strong
indicator for this category which supports the EKC theory for an initial low level of income.
Average investment has a positive impact of 1.052 times towards achieving a high win-win
for water quality. The initial HDI also promotes a high win-win in water quality by 1.068
times. The change in perceptions of government effectiveness and political stability were not
statistically significant in this case to affect the likelihood of a high win-win.

Air pollution (human health effects)
In this category, we have 50 out of 150 in the win-win that met the threshold for GDP
and air quality (Table 3.1). A total of 8 high-income nations out of the 50 were in a high winwin which were Trinidad and Tobago, Poland, Panama, Hungary, Croatia, Slovak Republic,
South Korea and Argentina. Majority of the nations in this category were low and middleincome nations with same regions as Child Mortality and Water like Syria, Sudan, Chad,
Egypt, Pakistan, Costa Rica and Cuba. Political stability change (p < 0.05) was the only
statistically significant variable impacting the win-win for air quality. Political stability
change had an odds ratio of 1.954 which implies that it positively impacts a high win-win for
air quality. All the other variables such as initial GDP, initial non-income HDI, average
investment % of GDP, change in perceptions of government effectiveness and health
expenditure change were not statistically significant in this category.
In summary, in the environmental category, child mortality had the highest number of
nations (65) in a high win-win while water had the least number of nations (41). The
explanatory power of the independent variables to determine a high win-win is strongest for
child mortality improvements and economic growth, followed by water but only a single
factor, change in political stability, was statistically significant to explain a high win-win in
for air quality improvements and economic growth (See Appendix E for the logistic results
from Eviews).

High win-win in economic growth and ecosystem vitality
Results of factors influencing the likelihood of a high win-win relative to no win-win is presented below:
Table 3.2: The coefficients followed by odds ratio and p-values (in parentheses) for the determinants of a likelihood of a high win-win
for the ecosystem vitality indicators and Average GDP per capita using logistic regression. Coefficients with statistically significant pvalues are bolded. The ecosystem vitality indicators are Biodiversity and Habitat (EV-BH), Agriculture (EV-AG), Forestry (EVFOREST), Fishery (EV_FISH), Water Quantity use (EV-WATER), Climate Change (EV-CC) and Air Pollution (EV-AIR).

Constant term
lnGDP2000
EV_2000
HDI2000
GE change
PS change
Inv/GDP
Cases of no win-win
Cases with win-win
Adjusted R-Square
S.E. of regression
LR statistics
Thresholds

EV-BH
4.532
(0.014)
‐1.453, 0.234
(0.000)
‐0.002, 0.998
(0.791)
0.098, 1.103
(0.000)
0.869, 2.385
(0.229)
0.708, 2.030
(0.069)
0.062, 1.064
(0.029)

LAND
EV-AG
‐0.913
0.642
‐0.575, 0.563
(0.097)
0.009, 1.009
(0.282)
0.051, 1.052
(0.064)
1.642, 5.166
(0.033)
‐0.421, 0.656
(0.314)
0.031, 1.031
(0.340)

99
52
0.182
0.430
35.471
GDP>2
BH=>0

122
29
0.074
0.390
10.925
GDP>2
AG>0

WATER
EV-FOREST
EV-FISH EV-WATER
‐6.468
‐0.399
‐1.171
0.848
0.704
0.032
‐1.189, 0.305 ‐0.422, 0.656 ‐0.397, 0.672
(0.003)
(0.363)
(0.406)
0.064, 1.066 ‐0.052, 0.949 0.061, 1.063
(0.000)
(0.036)
(0.000)
0.044, 1.045 0.025, 1.025
0.023, 1.023
(0.116)
(0.499)
(0.553)
1.121, 3.068 0.698, 2.010 ‐0.826, 0.438
(0.195)
(0.515)
(0.526)
0.202, 1.224 1.319, 3.740 1.772, 5.883
(0.681)
(0.034)
(0.009)
0.087, 1.091 0.150, 1.162 0.116, 1.123
(0.009)
(0.004)
(0.061)
110
92
139
38
26
11
0.298
0.237
0.353
0.364
0.367
0.227
50.222
29.535
27.744
GDP>2
GDP>2
GDP>2
FOR>‐10
FISH>0
WAT>‐25

AIR
EV-CC
EV-AIR
5.847
4.417
0.133
0.046
‐1.196, 0.302 ‐0.730, 0.482
(0.013)
(0.024)
‐0.016, 0.984 ‐0.021, 0.979
(0.310)
(0.082)
0.059, 1.061 0.032, 1.033
(0.097)
(0.218)
‐0.644, 0.525 1.555, 4.735
(0.532)
(0.045)
1.667, 5.296 0.194, 1.214
(0.004)
(0.629)
0.035, 1.036 0.016, 1.016
(0.433)
(0.642)
88
74
34
47
0.187
0.107
0.405
0.468
27.023
17.329
GDP>2
GDP>2
CC>0
AIR>0

Biodiversity and Habitat
A total of 52 out of 151 nations met the threshold of a high win-win for GDP
growth and BH trend (Table 3.2). There was a total of 13 high-income nations out of the
52 in a high win-win like Singapore, South Korea, Slovenia, Croatia, Uruguay, Panama,
Greenland etc while the majority were low and middle-income nations like Malaysia,
Brazil, Cuba and Costa Rica. The outcome for this indicator varied across nations but a
high win-win was dominant for low and middle-income nations. The initial GDP (p <
0.001), initial HDI (p < 0.001) Political Stability change (p < 0.1) and average investment
(p < 0.005) were the statistically significant variables impacting the high win-win trend
for biodiversity and habitat and the economy. The odds ratio for these variables suggest
that Initial GDP was 0.234 times less likely contributing to the odds of being in a high
win-win for both areas. Here again it shows that nations with a low initial income are
more likely to achieve a high win-win relative to high-income nations. Contrary to the
environmental health where non-income HDI was found to be statistically significant, the
initial non-income HDI in this case increases the likelihood of achieving a win-win 1.103
times relative to having no win-win. Political stability change is also seen to have a strong
positive impact being 2.030 times indicates a large positive increase in the odds for a winwin. Similarly, the average investment as a percent of GDP has a positive impact on the
odds ratio as well. Initial BH score and change in government effectiveness was not
found to be statistically significant in influencing the likelihood of a high win-win
outcome.

Agriculture
A total of 29 out of 151 nations met the threshold for being in a high win-win for
GDP and agriculture (Table 3.2). There were only 6 high-income nations out of the 29 in
a high win-win which includes Slovak Republic, Slovenia, South Korea, Singapore,
Latvia and Hungary while the rest were mostly middle-income and low-income nations
from North African and Latin American regions. The initial GDP (p < 0.1), initial HDI (p
< 0.1) and government effectiveness change (p < 0.05) were the statistically significant
variables impacting the high win-win trend. The odds of Initial GDP imply that an initial
high GDP for agriculture is 0.563 times less likely contributing to the odds of being in a
win-win relative to no win-win. Here again low-income nations are more likely to
achieve a win-win than developed nations. The initial HDI impacts a win-win for

77
agriculture 1.052 times more than a no win-win outcome. Nations with a high HDI are
more likely to have a win-win relative to nations that have a low HDI score. Similarly,
government effectiveness change is a strong positive indicator for a win-win as it fosters
a win-win 5.166 times more that the odds of having no win-win. Initial agriculture score,
change in political stability and investment as a % of GDP were not statistically
significant in explaining a high win-win outcome.

Forestry
A total of 38 out of 148 nations met the threshold for a high win-win for GDP and
forestry (Table 3.2). There were only 7 high-income nations in this category consisting of
Latvia, Lithuania, Slovak Republic, Slovenia, Poland and South Korea which have some
nations in common with those in agriculture. The rest of the nations with a high win-win
were middle and low-income nations including Azerbaijan, Egypt, Bhutan, Bulgaria and
Cape Verde. The initial GDP (p < 0.005), initial forestry level (p < 0.001) and average
investment (p < 0.05) were the statistically significant variables impacting the high winwin trend. Initial GDP is 0.305 times less likely contributing to the odds of being in a
high win-win. Low-income nations are more likely to have a high win-win outcome. The
initial forestry level of any given nation also is a positive indicator as it increases the
chances of a high win-win by 1.052 times. This indicates that nations that have a good
forest level initially will want to further protect their forests while having economic
growth relative to nations that their forest levels initially do not score high. Similarly,
average investment also positively contributes to a win-win 1.091 times more that the
odds of not. Changes in government effectiveness and political stability were not found to
be statistically significant. Initial non-income HDI has a positive impact but marginal (p =
0.11).

Fisheries
A total of 26 out of 118 nations met the threshold for win-win for GDP and
fisheries (Table 3.2). There was a total of only 7 high-income nations in this category
which includes Uruguay, Turks and Cacao Islands, South Korea, Estonia, Panama,
Croatia and Lithuania while the majority were low and middle-income nations like India,
Indonesia, Tunisia and Namibia. The initial fishery level (p < 0.1), political stability
change (p < 0.1) and average investment (p < 0.01) were the statistically significant

78
variables impacting the high win-win trend. The result showed that a high initial fishery
score leads to a lower likelihood of a win-win relative to a low initial fishery score. This
factor contributes to the odds of being in a high win-win by 0.949 times. Political stability
change also positively affects the fisheries category by an increase in the chances of a
win-win by 3.740 times. Similarly, average investment contributes to a win-win 1.162
times in this category. The remaining variables such as initial GDP per capita, initial HDI
and government effectiveness were not statistically significant in this category.

Water (ecosystem effects)
A total of only 11 out of 150 nations met the threshold for a high win-win for
GDP and water quantity (Table 3.2). There were only 3 high-income nations in this high
win-win which were Greenland, Turks and Cacao Island and Latvia while the majority
were low and middle-income nations like Bhutan, Belize, Sierra Leone, Myanmar and
Angola. The initial water quantity level (p < 0.001), political stability change (p < 0.01)
and average investment (p < 0.1) were the statistically significant variables impacting the
high win-win trend. A high initial water score contributes positively to the odds of being
in a high win-win 1.063 times in this category relative to a low initial water score.
Political stability change is also a strong indicator for a high win-win by increasing the
chances by 5.883 times. Similarly, average investment also contributes to a high win-win
1.123 times. The remaining variables such as initial GDP per capita, initial HDI and
government effectiveness were not statistically significant in this category.

Climate Change
A total of 34 out of 122 nations met the threshold for a high win-win for GDP and
climate change (Table 3.2). There was a total of 7 high-income nations which include
Estonia, Latvia, Slovak Republic, Slovenia, Singapore, Poland and Hungary while the
majority were low and middle-income nations like Ethiopia, Ghana, Libya, Morocco,
Nepal and Nigeria. The initial GDP (p < 0.1), initial HDI (p < 0.1) and political stability
change (p < 0.01) were the statistically significant variables impacting the high win-win
trend for climate change. The odds ratio of these variables shows that an initial high GDP
is 0.302 times less likely contributing to the odds of being in a win-win which indicates a
weak indicator. The initial HDI being 1.061 indicates a strong indicator for a high winwin by increasing the odds 1.061 times. Political stability shows some strong win-win

79
odds as it contributes to a high win-win 5.296 times for climate change. The remaining
variables such as government effectiveness and average investment were not statistically
significant in this category.

Air Pollution (ecosystem effects)
A total of 47 out of 121 nations met the threshold for a high win-win for GDP and
air quality (Table 3.2) which is a higher number of nations compared to climate change.
There was a total of 14 high-income nations similar to climate change and also includes
Saudi Arabia, Uruguay, Argentina and South Korea, the rest which are majority were the
low and middle-income nations like Cuba, Ghana, Peru, Russia and Ukraine in a high
win-win. The initial GDP (p < 0.1), initial air quality level (p < 0.1) and government
effectiveness change (p < 0.1) were the statistically significant variables impacting the
high win-win trend for air quality. The odds ratio for these variables shows that Initial
GDP is a weak high win-win indicator because it is 0.482 times less likely contributing to
the odds of being in a win-win. The initial air quality improves the chances of win-win to
occur by 0.979 times. Government effectiveness change shows a strong win-win indicator
by increasing the odds of a win-win 4.735 times more that the odds of having no win-win
in this category. The remaining variables such as initial HDI, political stability and
average investment were not statistically significant in this category.
In conclusion, the high win-win for ecosystem vitality had only a few nations
consistent throughout the indicators that met the thresholds. However, low and middleincome nations were dominant in the high-performance win-win trends. But, we see an
overall lesser number of nations within the thresholds with the highest being 52 high inwin nations in biodiversity and habitat and the lowest as 11 high win-win for Water out of
over 150 nations surveyed. The explanatory power of the independent variables to
determine a high win-win is varied across all the indicators for ecosystem vitality (See
Appendix E for the logistic results from Eviews).
Comparison with a single high win situation for environmental indicators
In order to further distinguish the importance of a win-win situation for these two
factors, a comparison with a win in one factor, in this case, the environmental categories
alone, is presented below to highlight the differences.

Table 3.3: The coefficients and p-values for the determinants of a likelihood of a single win for environmental
health ecosystem vitality indicators using logistic regression. Coefficients with statistically significant p-values are
bolded.

Environmental Health
EH
Water
Air
Constant term

BH

Ag

Forest

Ecosystem Vitality
Fish
Air

CC

Water

‐0.130 0.269
(0.929) (0.873)
‐0.054 0.006
(0.001) (0.649)
0.011
0.024
(0.246) (0.622)
‐0.608 0.244
(0.072) (0.478)
0.019 ‐0.628
(0.957) (0.136)

‐1.470
(0.327)
‐0.010
(0.373)
0.035
(0.042)
‐1.741
(0.025)
0.753
(0.067)

‐0.541
(0.496)

2.353
(0.257)
‐0.037
(0.149)
0.015
(0.758)
‐0.122
(0.724)
‐0.086
(0.841)
0.019
(0.781)

‐3.428
(0.006)
‐0.065
(0.000)
0.091
(0.000)
0.070
(0.841)
‐0.490
(0.134)
0.115
(0.033)

0.038
(0.971)
‐0.021
(0.028)
0.023
(0.192)
‐0.380
(0.197)
0.259
(0.403)
‐0.031
(0.536)

‐1.864
(0.097)
‐0.002
(0.720)
0.040
(0.015)
0.110
(0.708)
0.316
(0.292)

‐2.499
(0.088)
0.003
(0.753)
0.024
(0.252)
1.178
(0.001)
‐0.580
(0.149)

‐8.307
(0.000)
0.114
(0.000)
‐0.001
(0.965)
‐0.704
(0.071)
‐0.903
(0.062)

39

93

73

60

100

80

72

28

50

120

114

60

80

97

57

74

52

93

72

34

0.122

0.168

0.070

0.084

0.272

0.510

0.092

0.473

0.834

0.046

S.E. of regression 0.409
LR statistics
21.272

0.443
34.477

0.485
15.010

0.466
17.560

0.400
55.890

0.330
108.92

0.474
15.598

0.421
6.196

0.475
13.8111

0.410
7.552

Initial Env2000
HDI2000
FGE change
PS change
Health change
Observations
without a win
Observations
with a win
Adjusted R‐
Square

‐0.012
(0.336)
‐1.110
(0.136)
0.796
(0.037)

81
There are more nations in a high win situation that met the threshold for all the
environmental indicators without considering a win for economic growth alongside as
opposed to a high win-win which had fewer nations (Table 3.3). When considering progress
for a single environmental indicator, its initial level of environmental performance determines
how farther a nation will improve over the year. A nation with an initial high environmental
performance level may have lesser room for further improvements. In the categories of child
mortality, water pollution (human health), forestry and fisheries, the initial environmental
level is statistically significant (p<0.005) all of which also have a negative coefficient except
for forestry which has a positive coefficient. This implies that a nation with an initial high
level in child mortality, water pollution, forestry and fisheries is likely to have a high win at a
slower rate than a nation with an initial low level which will progress at a faster rate.
In the case of forestry, a nation with an initial high forestry level in terms of forest
loss and forest covers will likely perform better than a nation with an already deteriorating
forest cover or high forest loss. Also, the impact of governance indicators are not strong
indicators of a high win in the environmental cases as opposed to a consideration of a winwin situation which considers economic growth. This shows the interrelationship between the
economic and environmental elements such that the performance of one factor, in this case,
economic growth, cannot be totally isolated from the resulting impact it has on the other,
which are the environmental categories.

Comparison with a single high win situation for high economic performance
Also, in considering a win in just the GDP over the decade, we show results for a win
in GDP to assess the factors that lead to a win in progress of GDP alone to see how it differs
from a win-win situation.

82
Table 3.4: The coefficients and p-values for the determinants of a likelihood of a single win for
environmental health ecosystem vitality indicators simple linear regression. Coefficients with
statistically significant p-values are bolded for Average GDP (Av GDP) and the probability of
obtaining a win for GDP growth (Prob).

Constant term
lnGDP2000
HDI2000
FGE change
PS change
Health change
Observations
without a win
Observations with
a win
Adjusted R‐Square
S.E. of regression
LR statistics

Av GDP Prob(GDP>2)
4.990
3.706
(0.008)
(0.040)
‐1.165
‐1.133
(0.000)
(0.001)
0.060
0.050
(0.012)
(0.053)
0.184
1.524
(0.000)
(0.033)
1.385
1.180
(0.023)
(0.003)
1.100
0.152
(0.005)
(0.000)
91

0.452
1.868

105
0.265
0.420
55.180

There was a general high win trend most nations (105) in their economic performance
over the years that met the threshold (Table 3.4). The initial GDP level, HDI level,
governance indicators were all statistically significant (p<0.05) in impacting a win for GDP.
The initial level of GDP also reveals that a nation with a high level of GDP is likely to
increase at a lesser rate than a nation with initial low level of GDP which also leads to
convergence. Economic growth is very much reflective of the progress of its governance,
economy, human development index (HDI), human health, environmental and social factors,
hence the statistical significance of all the explanatory variables in positively impacting the
economic growth over the decade.
In conclusion, we see the examination of a high win-win situation for nations in their
economic growth and environmental indicators to be very relevant in revealing how the
progress in the association of these two factors varies from one environmental indicator to
another. It also reveals how a high GDP growth rate is favorable low and middle-income
nations and only a few high-income nations.

83
Discussion
The results presented reveal some significant relationships and trends for a high winwin when thresholds are set for GDP per capita growth rate and the 2012 EPI pilot trend
policy categories. The first of the relationships that is clearly seen to favour a high win-win is
the link between the growth rate of GDP and the environmental performance being achieved
by mostly middle and low-income nations. These findings are consistent with the theory of
economic growth for a faster growth rate for emerging economies but finds little support for
the existence of an EKC which should favour high-income nations. However, numerous
studies suggest that nations follow a development path that solely relates to income and
environment inevitably. Consistent with the findings of Hsu et al. (2013) there were few
wealthy countries for example Slovak Republic, Slovenia, South Korea, Singapore and Latvia
who were in a high win-win situation for most of the policy categories for human health and
ecosystem which still highlights the role of income. But, generally, the trends for high winwin suggests that income growth alone is not enough to explain the differences in
environmental performance between countries as can be seen in the indicators of air (human
health), fisheries and water (ecosystem).
The link between the non-income HDI and environmental performance trend for the
high win-win cases were also found (Constantini and Salvatore, 2008; Mukherjee and
Chakraborty, 2010; Hsu et al., 2013; Melnick et al., 2005). As opposed to using HDI as an
explanatory variable which includes income as a component of its indicator, this study uses
the non-income HDI which makes for a stronger factor that impacts a high win-win for
nations between social development and environmental performance trend (Hsu et al., 2013).
However, some indicators like child mortality, air quality (human health), air
pollution(ecosystem), forests, fisheries and water quantity (ecosystem) were shown to not
have been impacted by their initial non-income HDIs as they were statistically not significant.
Another underlying factor considered as a possible explanation for differences in
outcomes for high win-win which has also been widely examined in previous studies is
governance indicators. Sachs and McArthur (2005) in an analysis of the progress toward
MDGs attributes poor performance toward achieving MDG goals to poor governance. This
study however observes mostly no statistical significance of the change in government
effectiveness to increase the likelihood of a high win-win for indicators like child mortality,
air quality (human health), water pollution (human health), biodiversity and habitat, forests,
fisheries, water quantity (ecosystem) and climate change. It seems to be significant for only
two indicators which are agriculture and air pollution (ecosystem). Change in political

84
stability, on the other hand is documented in this study to have increased the chances of a
high win-win in child mortality, air quality (human health), biodiversity and habitat, fisheries,
water quantity (ecosystem) and climate change. This suggests that for mostly low and
middle-income nations, the progress towards a more politically stable economy increases the
government’s ability to enforce environmental regulations for these indicators which may
lead to higher win-win cases (Lopez and Mitra, 2000; Damania et al., 2003). However, while
Kaufmann et al. (1999) affirms that “governance matters’’ and as the author of the World
Governance Indicator (WGI) database, has employed several techniques like validation by
correlation and impact, to authenticate the indicators. There have been criticisms of
governance indicators from the WGI which say that they are based on expert perceptions and
they are inherently subjective. Authors like Razafindrakoto and Roubaud (2010) and Morse
(2006) question whether these indirect validation techniques guarantee a definite link
between subjective governance indicators and real levels of corruption control in a country.
They suggest more appropriate levels of analysis to include the sub-national and local levels,
so interactions can be further identified, and relationships refined.
While income growth rate, social development as non-income HDI, and governance
can help to explain some differences in individual countries’ performance, the average
investment share in GDP over the same period is another factor that measures the share of
investment in total production for any nation. The rate of investment reflects the infusion of
requisite capital to support the development process in any given nation (UNSD, 2010). This
factor increased the chances of a high win-win for many middle and low-income nations for
indicators like child mortality, water quality (human health), biodiversity and habitat, forests,
fisheries and water quantity (ecosystem) and for only a few high-income nations. A positive
average investment impacts the chances of a high win-win by accelerating the pace of
development through infused requisite capital which is reflected in the processes and patterns
of economic activities of low and middle-income countries. This factor also enhances a high
win-win by increasing their partnership in the global economy (UNSD, 2000).
The findings obtained from this study is similar to Gallego et al. (2014), Hsu et al.
(2013) and Mavragani et al. (2016) that point out that socioeconomic factors, such as
economic wealth and education, as well as institutional factors represented by the style of
public administration are determining factors of environmental performance in the countries
analysed but in the case of my study a high win-win situation. This study also found that in
regard to the two groups of variables in the EPI, in environmental health, the disparity in the
ratio of high-income to low and middle-income nations for a high win-win is wider as seen in

85
child mortality (8/69), water pollution (5/41) and air quality (8/50). Whereas variables related
to ecosystem vitality, the gap was not as broad as seen in biodiversity and habitat (13/52),
agriculture (6/29), forest (7/28), Water (3/11) climate change (7/34) and air pollution (14/47).
This shows that some high-income nations like Singapore, Slovenia, Latvia and Uruguay still
performed very well over the decade despite the thresholds, their outcomes can also be
attributed to their level of income growth and the positive impacts it has had on their
environmental degradation which is consistent with the EKC theory.
Society has shown increased interest in environmental issues on the microeconomic
level where stakeholders are increasingly concerned with the environmental performance of
firms and use it to make decisions about their investments. On the macroeconomic level
however, it focuses on the environmental performance of countries and their ability to
produce environmental public goods (Gallego et al., 2014). Citizens have become
increasingly aware of their right to a high-quality environment, which has led to each country
being accountable to its citizens for the environmental policies it puts into practice thereby
fostering the pursuit for a high win-win situation in economy and environment
simultaneously.

86
CHAPTER 4: CONCLUSION AND IMPLICATIONS FOR FUTURE STUDIES
This study set out to investigate the trends in the performances of over 200 nations for
the 2012 trend EPI environmental policy categories associated with their average GDP per
capita growth rate simultaneously. The aim was to discover clusters of nations within a winwin trend relative to those which were not as well as those nations which had very high
performances in both areas and the underlying factors which influenced the likelihood of
such performances. To achieve this goal, the k-means clustering technique was used to
identify homogenous groups of nations within the win-win category for nations in their
economic and environmental performance trends. This approach gave a broader perspective
in form of a graphical representation of several clusters of nations each representing a unique
trend to the association of economic and environmental performances. This approach was in
contrast to previous works that use only one type of variable, either economic or institutional,
or present only a theoretical perspective. This study provided a graphic representation that
differentiates between countries’ environmental and economic performance in relation to
environmental health, on the one hand, and to ecosystem vitality on the other as seen in
chapter 2. The second goal was to identify the underlying factors that may have led to a winwin for some nations and especially a high win-win situation for both variables. To identify
these factors, the countries and variables (average GDP growth rate and environmental
indicators) were contextualized by setting thresholds for both variables as minimum
requirements for a high win-win. A logit model was used to verify which economic and
institutional variables had an impact on a high win-win situation. This study reveals
socioeconomic factors, such as GDP and non-income HDI level, as well as institutional
factors represented by change in political stability and government effectiveness were some
of the determining factors of a high win-win situation which are consistent with findings in
Gallego et al. (2014), Hsu et al. (2013) and Mavragani et al. (2016). This provided an
explanation in form of factors that increase the likelihood of a high win-win situation for
nations in relation to environmental health and ecosystem vitality and their respective average
GDP growth rate.
This paper has demonstrated one way in which the 2012 pilot trend EPI and the
average GDP growth are useful in measuring progress of nations that have achieved a high
performance in global environmental policy goals and a high performance in its economic
growth simultaneously by incorporating thresholds. The goal of the 2012 Trend EPI was to
draw the attention of decision-makers to the environmental issues in their countries on which

87
they both lag and perform well (Hsu et al., 2013), compared to other nations economically,
geographically, as well as globally. An issue-by-issue examination of environmental goals
alongside their GDP growth over a decade provides a more useful insight to detect issues of
concern for policymakers at the country-level. This was achieved by identifying the leading
countries and those lagging behind which in turn helps enable make sense of global trends
toward achieving sustainability. This paper has revealed the environmental progress of low,
middle and high-income nations by using a methodology of clustering and logistic analysis as
the framework to assess which group of nations have improved significantly over the last
decade in the 2012 EPI environmental policy targets as seen in chapter 2 and 3.
Economic development and good governance (mostly with the use of governance
indicators) had been suggested to individually positively affect environmental performance
(Mavragani et al., 2016). However, a positive average investment also played significant role
in increasing the chances of a high win-win especially for low and middle-income countries
by accelerating their pace of development through infused requisite capital which helped
accelerate their processes and patterns of economic activities. The results showed an
interesting trend for countries whereby some environmental issues like child mortality and
water pollution (human health) had win-win situations across high-income, middle-income
and low-income countries. In categories like climate change and agriculture, high-income
countries dominated the win-win category which is consistent with the EKC predictions for
the pathway nations follow on their relationship between economic growth and
environmental quality.
This study contributes to the field of environmental economics and could be of
interest to policy makers as it emphasizes the strong correlation between economic
development in combination with good governance and environmental performance over a
long period of time which consequently leads to sustainable development. By broadening the
sample to include more years for a longitudinal study, this study has made an attempt to
provide research-based approach to ensure environmental sustainability, which is one of the
priorities of environmental authorities around the world (Gallego-Alvarez et al., 2014).
However, like other studies, this study has limitations that might be good starting
points for future research. Firstly, though the variables used in this study are reliable and the
statistical analysis follows a standard procedure, there could be one or more variables which
could be incorporated in addition to economic growth. Other variables such as the control of
corruption, population density, role of science and technology, the importance of market
dynamics, the role of economic agents and the ideology of social movements that affect

88
environmental performance that have not been considered in this study can be looked into. In
addition, the fact that data for the pilot trend environmental policy categories were only
available for the period of 2000 to 2010 restricts the generalization of the findings beyond the
time frame, an updated analysis for recent years to this study, when more data is available, is
necessary in the future.
The results obtained have real-world applications and can be useful for policy makers.
The standout outcome is that income growth is not the only explanatory variable for
understanding environmental performances and sustainability across countries. Institutional
factors such as political stability (regime type: democratic regimes show higher levels of
environmental performance than authoritarian regimes), government effectiveness, and other
institutional factors must also be considered, since they can affect environmental performance
indicators. Governments should also consider that being a wealthy country does not always
lead to better environmental performance, especially considering the long term in the
environmental health aspect apart from natural or ecological resources. In the case of
environmental health, over a long period of time it can be affirmed that there is convergence
for both low-income and wealthy countries such that all nations tend towards a win-win
situation. However, in the case of natural or ecological resources, it varies greatly from one
environmental issue to the other even over a period of time; there is no relation in some cases
at all, it is negative for some and of minimal impact for others. This may be due to the fact
that although wealthy countries may be able to invest money in order to improve their
environment, they also tend to increase environmental issues due to their high level of
consumption (Gallego-Alvarez et al., 2014).
Evident from some of our results however, is that income growth and effective
governance leads to better environmental performance as few high-income nations still
achieved a high-performance trend. Also stated by Gallego-Alvarez et al. (2014), an
effective, innovative and adaptable governance is a necessary condition for nations geared
towards sustainability. Some governance policies may include integrating environmental
policies and enhancing social capital when introducing legislations and regulations, in order
to achieve higher levels of environmental performance.

89
References
Brock, W.A. and Taylor, M.S., 2005. Economic growth and the environment: a review of
theory and empirics. Handbook of economic growth, 1, pp.1749-1821.
Brock, W.A. and Taylor, M.S., 2010. The green Solow model. Journal of Economic Growth,
15(2), pp.127-153.
Chévez, P., Barbero, D., Martini, I. and Discoli, C., 2017. Application of the k-means
clustering method for the detection and analysis of areas of homogeneous residential
electricity consumption at the Great La Plata region, Buenos Aires, Argentina.
Sustainable Cities and Society, 32, pp.115-129.
Cole, M.A., Rayner, A.J. and Bates, J.M., 1997. The environmental Kuznets curve: an
empirical analysis. Environment and development economics, 2(4), pp.401-416.
Cornish, R. (2007). Statistics: 3.1 Cluster Analysis. Mathematics learning support centre,
pp.1-5
Costantini, V. and Monni, S., 2008. Environment, human development and economic growth.
Ecological Economics, 64(4), pp.867-880.
Costanza, R., d'Arge, R., De Groot, R., Farber, S., Grasso, M., Hannon, B., Limburg, K.,
Naeem, S., O'neill, R.V., Paruelo, J. and Raskin, R.G., 1997. The value of the world's
ecosystem services and natural capital. Nature, 387(6630), pp.253-260.
Costanza, R., de Groot, R., Sutton, P., van der Ploeg, S., Anderson, S.J., Kubiszewski, I.,
Farber, S. and Turner, R.K., 2014. Changes in the global value of ecosystem services.
Global environmental change, 26, pp.152-158.
Criado, C.O., Valente, S. and Stengos, T., 2011. Growth and pollution convergence: Theory
and evidence. Journal of Environmental Economics and Management, 62(2), pp.199214.
Dasgupta, S., Laplante, B., Wang, H. and Wheeler, D., 2002. Confronting the environmental
Kuznets curve. The Journal of Economic Perspectives, 16(1), pp.147-168.
De Groot, R., Brander, L., Van Der Ploeg, S., Costanza, R., Bernard, F., Braat, L., Christie,
M., Crossman, N., Ghermandi, A., Hein, L. and Hussain, S., 2012. Global estimates of
the value of ecosystems and their services in monetary units. Ecosystem Services, 1(1),
pp.50-61.
Dinda, S., 2005. A theoretical basis for the environmental Kuznets curve. Ecological
Economics, 53(3), pp.403-413.
Emerson, J.W., Hsu, A., Levy, M.A., de Sherbinin, A., Mara, V., Esty, D.C. and Jaiteh, M.,
2012. Environmental performance index and pilot trend environmental performance
index. New Haven: Yale Center for Environmental Law and Policy.
Esty, D.C., Levy, M., Srebotnjak, T. and De Sherbinin, A., 2005. Environmental
sustainability index: Benchmarking national environmental stewardship. New Haven:
Yale Center for Environmental Law & Policy, pp.47-60.

90
Gallego-Alvarez, I., Vicente-Galindo, M., Galindo-Villardón, M. and Rodríguez-Rosa, M.,
2014. Environmental performance in countries worldwide: Determinant factors and
multivariate analysis. Sustainability, 6(11), pp.7807-7832.
Grossman, G.M. and Krueger, A.B., 1995. Economic growth and the environment. The
Quarterly Journal of Economics, 110(2), pp.353-377.
Grossman, G.M. and Krueger, A.B., 1991. Environmental impacts of a North American free
trade agreement (No. w3914). National Bureau of Economic Research.
Hsu, A., Lloyd, A. and Emerson, J.W., 2013. What progress have we made since Rio?
Results from the 2012 Environmental Performance Index (EPI) and Pilot Trend EPI.
Environmental Science & Policy, 33, pp.171-185.
Hsu, A., J. Emerson, M. Levy, A. de Sherbinin, L. Johnson, O. Malik, J. Schwartz, and M.
Jaiteh. 2014. The 2014 Environmental Performance Index. New Haven, CT: Yale
University, Center for Environmental Law and Policy.
Hsu, A. and Zomer, A., 2016. Environmental performance index. Wiley StatsRef: Statistics
Reference Online.
Index, E.P., 2016. Global Metrics for Environment. New Haven, CT: Yale University, pp.1123.
Jeßberger, Christoph (2011) : Multilateral Environmental Agreements up to 2050: Are They
Sustainable Enough?, Ifo Working Paper, No. 98, Ifo Institute - Leibniz Institute for
Economic Research at the University of Munich, Munich.
Kajikawa, Y., Tacoa, F. and Yamaguchi, K., 2014. Sustainability science: the changing
landscape of sustainability research. Sustainability science, 9(4), pp.431-438.
Kettenring, J.R., 2006. The practice of cluster analysis. Journal of classification, 23(1), pp.330.
Kubiszewski, I., Costanza, R., Franco, C., Lawn, P., Talberth, J., Jackson, T. and Aylmer, C.,
2013. Beyond GDP: Measuring and achieving global genuine progress. Ecological
Economics, 93, pp.57-68.
Lamb, W.F., Steinberger, J.K., Bows-Larkin, A., Peters, G.P., Roberts, J.T. and Wood, F.R.,
2014. Transitions in pathways of human development and carbon emissions.
Environmental Research Letters, 9(1), p.014011.
Lawn, P.A., 2003. A theoretical foundation to support the Index of Sustainable Economic
Welfare (ISEW), Genuine Progress Indicator (GPI), and other related indexes.
Ecological Economics, 44(1), pp.105-118.
List, J.A. and Gallet, C.A., 1999. The environmental Kuznets curve: does one size fit all?.
Ecological Economics, 31(3), pp.409-423.
Mavragani, A., Nikolaou, I.E. and Tsagarakis, K.P., 2016. Open Economy, Institutional
Quality, and Environmental Performance: A Macroeconomic Approach. Sustainability,
8(7), p.601.

91
McKitrick, R. and Wood, J., 2017. An Examination of the Relationship between Air Quality
and Income in Canada. Canadian Journal of Agricultural Economics/Revue canadienne
d'agroeconomie, 65(1), pp.69-92.
Millimet, D.L., List, J.A. and Stengos, T., 2003. The environmental Kuznets curve: Real
progress or misspecified models?. The Review of Economics and Statistics, 85(4),
pp.1038-1047.
Moran, D.D., Wackernagel, M., Kitzes, J.A., Goldfinger, S.H. and Boutaud, A., 2008.
Measuring sustainable development—Nation by nation. Ecological Economics, 64(3),
pp.470-474.
Neri, L., D’Agostino, A., Regoli, A., Pulselli, F.M. and Coscieme, L., 2017. Evaluating
dynamics of national economies through cluster analysis within the input-state-output
sustainability framework. Ecological Indicators, 72, pp.77-90.
Panayotou, T., 1997. Demystifying the environmental Kuznets curve: turning a black box
into a policy tool. Environment and Development Economics, 2(4), pp.465-484.
Panayotou, T., 1993. Empirical tests and policy analysis of environmental degradation at
different stages of economic development (No. 992927783402676). International Labour
Organization.
Results from the 2012 Environmental Performance Index (EPI) and Pilot Trend EPI.
Environmental Science & Policy, 33, pp.171-185.
Řezanková, H. (2014). Cluster Analysis of Economic Data. Statistika, 94(1), pp.73-86.
Shafik, N., 1994. Economic development and environmental quality: an econometric
analysis. Oxford Economic Papers, pp.757-773.
Sobhee, S.K., 2004. The environmental Kuznets curve (EKC): a logistic curve?. Applied
Economics Letters, 11(7), pp.449-452.
Stern, D.I., 2004. The rise and fall of the environmental Kuznets curve. World development,
32(8), pp.1419-1439.
Stern, D.I., 2017. The environmental Kuznets curve after 25 years. Journal of Bioeconomics,
19(1), pp.7-28.
Stern, D.I. and Common, M.S., 2001. Is there an environmental Kuznets curve for sulfur?.
Journal of Environmental Economics and Management, 41(2), pp.162-178.
Tamanini, J., Bassi, A., Hoffman, C. and Valenciano, J., 2014. Global Green Economy Index:
Measuring National Performance in the Green Economy. United Nations: Washington,
DC, USA.
UNDP Annual Report 2011/2012
UNSD Annual Report 2011/2012
Vörösmarty, C.J., McIntyre, P.B., Gessner, M.O., Dudgeon, D., Prusevich, A., Green, P.,

92
Glidden, S., Bunn, S.E., Sullivan, C.A., Liermann, C.R. and Davies, P.M., 2010. Global
threats to human water security and river biodiversity. Nature, 467(7315), p.555.
Wackernagel, M. and Rees, W., 1998. Our ecological footprint: reducing human impact on
the earth (Vol. 9). New Society Publishers.

93
Appendices
Appendix A: Indicator frameworks and EPI exploration
Transforming Raw Indicators
The first step involves conducting some standard normalizations for transforming raw
values, GDP, or another denominator to make data comparable across countries. The second
step involves applying statistical transformations to the data by which to better differentiate
performance amongst countries. Logarithmic transformations are applied to address the
skewness of the underlying datasets. The transformed data are then used to calculate
performance indicators using a proximity-to-target methodology, which reflects how close a
country is to an identified policy target (Hsu et al., 2013). The target, or high-performance
benchmark, is defined by international or national policy goals, established scientific
thresholds, or expert judgment.
Proximity-to-Target Score Methodology
The proximity-to-target score (

, , ) of each nation for each time period the

methodology is shown in
The formula for

,,

used was as follows:
,,

Where

is the poor performance benchmark,

target and

,

,

∗ 100

is the top performance benchmark or the

is the indicator of nation i at time t = 2000 – 2010 all variables transformed as

explained previously. The proximity to target shows how far the indicator score is from the
poor performance benchmark as a fraction of the distance between poor and top
performances. If it is not far from the target, then the

,,

will be closer to zero and if it is

close to the target it will be closer to 100.
Trend Score Methodology
For each indicator, a simple linear regression model of the annual proximity-to-target
scores is used to determine a rate of improvement or decline for each indicator. The slope of
the trend line determines the scale. 0 slope reflects “no change”, a positive slope reflects
improvement and a negative slope indicates decline.

94
This is done for every nation and for every indicator. Then these slopes for each
indicator are ranked from “best improvement” receiving a score of 50 and defined by the
95% percentile of the slopes, 0 slope reflecting “no change” again and -50 is for the “worst
trend decline”.

Forest Loss, Forest Growing Stock, Forest Cover, and Change in Water

Quantity have trend scores that range from -50 to 0 as they are change indicators.
Final Aggregation Methodology
The indicator scores are then aggregated (averaged) according to assigned weightings
to produce scores within each policy category (Hsu et al., 2013). Differential weighting of the
indicators as percentages of a policy category is determined through expert judgments based
on considerations of data quality, relevance of the indicator to measure a particular issue. The
2012 pilot Trend Index scores and ranks countries based on improvement or decline from
2000 to 2010.
Exploratory patterns of some indicators
This chapter aims to explore the already existing trends of some nations for some
indicators in the policy categories of the EPI in order to capture existing patterns that can also
contribute to the win-win or win-loss categories.
Application to Environmental burden of disease policy category
Child mortality is the only indicator in this policy category and it is defined as the
probability of dying between a child’s first and fifths birthdays per 1000 children aged 1year
old. Many environmental and socioeconomic factors influence child mortality for ages 1 – 4
years old. The target and low performance benchmark for this indicator is shown in the table
below:
EPI framework for Child Mortality (CHMORT) Indicator
Indicators

Child
mortality

Low
Statistical Transformed
Target performance
transformation
targets
benchmark
0.0007
probability
of dying
between
age 1 and
5

0.113

natural
logarithm

-6.91

Transformed
low
performance
benchmark

-2.18

95
Child mortality slopes of the trends of the proximity to target of 195 nations is shown in the
figure below. The table below also shows nations with both increasing and a decreasing trend
in CHMORT towards achieving this indicator target or still at the low performance
benchmark.

96

‐50

‐40

Madagascar
Botswana
Maldives
Turkey
Morocco
Sierra Leone
Timor‐Leste
Guyana
Indonesia
Ethiopia
Cambodia
Brazil
Macao
Taiwan
Myanmar
Sudan
Occupied Palestinian Territory
Grenada
Jamaica
Belize
Paraguay
Norway
Spain
Iran
Libyan Arab Jamahiriya
Pakistan
Czech Republic
Canada
Martinique
French Guiana
Macedonia
Burundi
Denmark
Saudi Arabia
Puerto Rico
Argentina
Somalia
Italy
Congo
Iraq
Belgium
Uzbekistan
Uruguay
Poland
Hungary
Croatia
Kuwait
Bosnia and Herzegovina
Russia
‐30
‐20
‐10
0

10

20

30

Trend scores for a selected number of nations in CHMORT Indicator

40

50

60

97
Most improved
Trend
Trend
trends
score Most decline trends
Score
United Arab
Emirates
50 Russia
-38.07
Azerbaijan
50 Lithuania
-25.85
Bangladesh
50 Cuba
-23.46
Botswana
50 Estonia
-16.83
Laos
50 Bosnia and Herzegovina
-6.56
Liberia
50 Slovenia
-5.52
Latvia
50 Kyrgyzstan
-4.79
Madagascar
50 Mauritius
-4.12
Nepal
50 Kuwait
-3.4
Rwanda
50 El Salvador
-2.39
Trend scores for a selected number of nations in CHMORT Indicator
From the above table, a declining trend were mostly nations that are underdeveloped, while
the nations that show no change or little are the developed nations. Nations that show major a
major increasing trend in child mortality are nations from the former Soviet Union and Cuba.
For most developed nations, CHMORT is usually not caused by environmental factors but by
factors such as accidents or congenital diseases.
Application to Forests policy category
Forests are cover almost 30% of the Earth’s terrestrial surface and are a major source
of biomass, food products, wood, pulp, medicine etc. Forest loss (FORLOSS) indicator is
used as a demonstration from this policy category. Forest Loss measures the loss of forest
area owing to deforestation from either human or natural causes. The targets and low
performance benchmark is shown below:

Indicators

Target

Low
performance
benchmark

Forest loss

0.015 %
loss

7 % loss

Statistical
transformation

Transformed
Transformed low
performance
targets
benchmark

natural
logarithm

-8.81

-2.66

EPI framework for Forest Loss (FORLOSS) Indicator
The slopes of the trends of the proximity to target of 188 nations for FORLOSS indicator is
shown below. The table below also show nations with both an improving and declining trend

98
in FORLOSS towards achieving this indicator target or still at the low performance
benchmark.

99

FORLOSS_trend
British Virgin Islands
Serbia
Senegal
Puerto Rico
Mali
Liberia
Haiti
Guinea
Micronesia
Congo
Botswana
Burkina Faso
Anguilla
Norway
India
United States Virgin Islands
Namibia
Italy
Costa Rica
Mozambique
Jamaica
China
Poland
Malawi
Tanzania

‐60

‐50

‐40

‐30

Cyprus
Russia
Myanmar
Saint Vincent and the Grenadines
Spain
Ecuador
Guatemala
Nicaragua
New Zealand
Zimbabwe
Uruguay
Paraguay
Saint Pierre and Miquelon
‐20
‐10
0

100

Rank Country
Saint Pierre and
1 Miquelon
2 Portugal
3 Swaziland
4 South Africa
5 Malaysia

Forest
Loss

Forest
Loss

Rank Country

-50.0
-49.5
-47.7
-45.9
-43.4

13
14
15
16
17

Tajikistan
Syria
Trinidad and Tobago
Zimbabwe
Indonesia
United States of
6 Paraguay
-40.5
18 America
7 Australia
-39.8
19 Singapore
8 Cambodia
-38.9
20 Chile
9 Argentina
-36.4
21 New Zealand
10 Montserrat
-36.4
22 Bolivia
11 Uruguay
-36.1
23 Saint Lucia
12 Brazil
-35.0
24 Canada
Trend scores for a selected number of nations in FORLOSS Indicator

-32.9
-32.4
-32.4
-31.9
-31.4
-31.4
-30.9
-30.4
-30.4
-29.8
-29.8
-29.2

124 nations show a declining trend away from the target with most of them being advanced
nations, 64 nations show no change in their FORLOSS over the years. Deforestation rates are
mostly higher in Southeast Asia, South America and Africa even though recent evidence
suggests they may be delining (Emerson, 2012).
Application to Fisheries policy category
Fishing and aquaculture are mainly the activities that have a direct impact on the
marine ecosystem. Overfishing of species can be detrimental to marine biodiversity and
ecosystem stability. Fish Stocks Overexploited or Collapsed (FSOC) indicator is used as a
demonstration from this category. FSOC measures the fraction of species that are fished in
each country’s EEZ that are exploited or collapsed. The target and low performance
benchmark for this indicator is shown below:

Indicators

Target

Transformed
Low
low
Statistical
Transformed
performance
performance
transformation
targets
benchmark
benchmark

0 species
natural
Fishing stocks overexploited
1
logarithm
overexploited* or collapsed
within EEZ
EPI framework for Fish Stocks overexploited (FSOC) Indicator

-2.64

0.07

The slopes of the trends of the proximity to target of 180 nations for the fishing stocks
overexploited are shown below in the figure below. The table below also shows nations with

101
an increasing or decreasing trend in Fish stocks overexploitation towards achieving this
indicator target or still at the low performance benchmark.

102

FSOC_trend

‐60

‐40

Malta
Guadeloupe
Saint Kitts and Nevis
Australia
Qatar
Guam
China
Japan
Cook Islands
Oman
Indonesia
Congo
Faeroe Islands
Samoa
Slovenia
Palau
Nicaragua
Guatemala
Eritrea
Bosnia and Herzegovina
Aruba
Estonia
Kenya
Portugal
Sierra Leone
Nigeria
Ecuador
Côte d'Ivoire
Netherlands
South Africa
Fiji
Chile
Costa Rica
United Kingdom
United Arab Emirates
Thailand
Comoros
Taiwan
British Virgin Islands
Sudan
Norfolk Island
Myanmar
Lebanon
Algeria
Netherlands Antilles
‐20
0

20

Trend scores for a selected number of nations in FSOC Indicator

40

60

103
Countries with the most
improvement
Trend scores
Antigua and Barbuda
50
Guadeloupe
50
French Guiana
50
Croatia
50
Haiti
50
Malta
50
Suriname
50
Turks and Caicos
Islands
50
Vanuatu
50
Trend scores for a selected number of nations in FSOC Indicator
96 nations were shown to have an increasing trend in FSOC which was further away from the
target, 29 nations show no change, and 55 nations were found to have a decreasing trend in
overexploitation towards the target. The best improved nations are mostly in the Caribbean
Sea and Northeast South America as seen in the table above. Fisheries are an important
aspect of most developing economies with half of global fish exports by value attributed to
developing countries (Emerson, 2012). The demand for fresh seafood continues to rise with
population growth and increasing affluence in developing countries hence the increasing
trend in FSOC.
Application to Agriculture policy category
Agricultural demands have enormous impacts on global ecosystems due to practices
that are heavily dependent on natural resources such as soil, water, and climate. Pesticide
regulation of persistent organic pollutants (POPs), endocrine disruptors, or carcinogens
indicator is used as a demonstration in this category. This indicator examines the legislative
status of countries according to Stockholm Convention on persistent organic pollutants. It
rates the degree to which these countries have kept those objectives by limiting the use of
certain toxic chemicals. Pesticides are a significant source of pollution in the environment
affecting both human and ecosystem health. The target and low performance benchmark for
this indicator is shown below in the table below:

104

Indicators

Target

Low
Statistical
performance
transformation
benchmark

Transformed
Transformed
low
targets
performance
benchmark

Pesticide
22 points
0 None
regulation
EPI framework for Pesticide Regulation (POPS) Indicator
The slopes of the trends of the proximity to target of 231 nations for POPs indicator is shown
below in the figure below. The table below also shows nations with a trend in improvement
or decline in POPs towards achieving this indicator target or still at the low performance
benchmark.

105

POPs_trend
Zimbabwe
Vanuatu
Holy See
Uganda
Tonga
Togo
Swaziland
Sao Tome and Principe
El Salvador
Senegal
Qatar
Puerto Rico
Panama
Nepal
Nigeria
Mayotte
Mauritania
Myanmar
Maldives
Lithuania
Libyan Arab Jamahiriya
Kiribati
Japan
Iceland
Isle of Man
Hong Kong
Greenland
Gambia
Georgia
Falkland Islands
Algeria
Cyprus
Comoros
Côte d'Ivoire
Botswana
Bolivia
Bahamas
Burundi
Armenia
Albania
Singapore
Viet Nam
Morocco
Bulgaria
Thailand
Moldova
Belgium
0

10

20

30

40

Trend scores for a selected number of nations in POPs Indicator

50

60

106

Rank Most improved
1
Belgium
2
3
4

Trend
Most
score Rank
improved
50.0 11
Thailand
Trinidad and
50.0 12
Tobago
50.0 13
Zambia
50.0 14
Germany

Chile
Finland
Indonesia
Saint Kitts and
5
Nevis
50.0 15
Egypt
6
Moldova
50.0 16
Bulgaria
7
Mexico
50.0 17
Lebanon
8
Mauritius
50.0 18
Monaco
9
Romania
50.0 19
Peru
10
Sudan
50.0 20
Austria
Trend scores for a selected number of nations in POPs Indicator

Trend
score
50.0
50.0
50.0
37.5
37.5
31.3
31.3
31.3
31.3
29.2

31 nations showed an increasing trend in POPs with some meeting the target and other
nations towards the target while 200 countries showed no change in their POPS.
Application to Biodiversity and habitat policy category
Human activities impacts the world’s terrestrial, freshwater and marine ecosystem
significantly throughout history and has intensified over the last 50 years (Emerson, 2012).
Biome protection (PACOV) indicator is used as a demonstration from this category. PACOV
measures the degree to which a country achieves the target of protecting at least 17% of each
terrestrial biome within its borders and represents a weighted average of protection by biome
(Emerson, 2012). The target and low performance benchmark for this indicator is shown in
the table below:
Indicators Target

Transformed
Low
Statistical
Transformed low
performance
performance
transformation targets
benchmark
benchmark

17%
weighted
average
Biome
0
none
protection of
biomes
protected
EPI framework for Biome Protection (PACOV) Indicator
The slopes of the trends of the proximity to target of 230 nations for Biome Protection are
shown in the Figure below. The table below also shows nations with an increasing trend in
achieving the target of biome protection or still at the low performance benchmark.

107

PACOV_trend
Northern Mariana Islands
Gabon
Albania
United Arab Emirates
Uganda
Myanmar
Timor‐Leste
Indonesia
China
Mozambique
Namibia
Syria
Viet Nam
Guadeloupe
Tanzania
Hungary
Zimbabwe
Ukraine
Ecuador
Austria
Wallis and Futuna Islands
Uzbekistan
Tunisia
Thailand
Swaziland
Sierra Leone
Saudi Arabia
Pakistan
Nicaragua
Malawi
Mongolia
Morocco
Liechtenstein
Kuwait
Kenya
India
Guam
Gambia
United Kingdom
Estonia
Djibouti
Comoros
Chile
Barbados
Burundi
Aruba
0

10

20

30

40

Trend scores for a selected number of nations in PACOV Indicator

50

60

108
Countries with most improvement Trend Score
Gabon
50
Greece
50
French Guiana
50
Iceland
50
Italy
50
Northern Mariana Islands
50
New Caledonia
50
Peru
50
Réunion
50
Slovenia
50
Trend scores for a selected number of nations in PACOV Indicator
129 nations show no change in their percentage of biomes under protected status over the
years, while 101 nations show a progressive trend in their percentage of biomes under
protected status with some nations meeting the target as seen above.
Application to Water policy category: (Effects on ecosystem)
Factors such as air pollution, climate change, economic development have grossly
increased the pressures on global freshwater resources. Change in water quantity (WATUSE)
is the only indictor in this category. It represents the area-weighted percent change in river
flow from a pre-industrial natural state owing to water withdrawals and reservoirs. The target
and low performance benchmark for this indicator is shown in the table below:
Transformed
Low
Statistical
Transformed low
Indicators Target performance
performance
transformation targets
benchmark
benchmark
Change
inverse, natural
0%
-7.41
3.79
-44.38
in water
logarithm
reduction
quantity
EPI framework for Change in Water Quantity (WATUSE) Indicator
The slopes of the trends of the proximity to target of 201 nations for the Change in water
quantity (natural river flows) owing to water withdrawals and reservoirs are shown in the
figure below. The table below also shows nations with a decreasing trend in the change in
annual water quantity 1towards achieving this indicator target or still at the low performance
benchmark.

109

WATUSEINV_trend

‐60

‐50

‐40

‐30

‐20

Turks and Caicos Islands
Papua New Guinea
Liberia
Vanuatu
Saint Kitts and Nevis
Fiji
Tonga
New Caledonia
Guyana
Bahamas
Malawi
Grenada
Latvia
Gambia
Laos
British Virgin Islands
Guadeloupe
Slovenia
Bolivia
Austria
Brazil
Netherlands Antilles
Canada
Nicaragua
Cameroon
Viet Nam
Nepal
Ecuador
Sao Tome and Principe
Suriname
Poland
Slovakia
Jamaica
Eritrea
Somalia
Puerto Rico
Zimbabwe
South Korea
Moldova
Tajikistan
Cape Verde
Bulgaria
Italy
Sri Lanka
Saudi Arabia
Turkey
Greece
India
Turkmenistan
Iran
Bahrain
‐10
0

Trend scores for a selected number of nations in WATUSE Indicator

110
Countries with worst trends Trend Score
Bahrain
-50.0
Cyprus
-49.5
Armenia
-49.5
Afghanistan
-48.4
Iran
-48.4
Lebanon
-48.3
Israel
-47.7
Azerbaijan
-47.5
Turkmenistan
-47.3
Spain
-47.1
Trend scores for a selected number of nations in WATUSE Indicator
193 nations show a decreasing trend in the change annual river quantity further away from
the target while 8 nations show no change in their change in annual water quantity.
Application to Air pollution policy category (Effects on ecosystems)
Air pollution negatively impacts plant growth, contributes to acid rain which is
detrimental to ecosystems. S02 emissions per GDP (SO2GDP) indicator is used as a
demonstration from this category. It is the ratio of SO2 emissions to GDP in 2005 constant
international prices PPP. The target and low performance benchmark for this indicator is
shown in the table below:

Indicators

Low
Statistical Transformed
Target performance
transformation
targets
benchmark

0
grammes
natural
-2.59
11.38625
SO2 per
logarithm
US 2005
$s PPP
EPI framework for Sulfur dioxide emissions (SO2GDP) Indicator
Sulfur
dioxide
emissions
per GDP

Transformed
low
performance
benchmark
2.44

The slopes of the trends of the proximity to target of 137 nations for SO2GDP are shown
below in the figure below. The table below also shows nations with a decreasing trend in
Sulfur dioxide emissions and nations with an increasing trend in SO2 emissions towards
achieving this indicator target or still at the low performance benchmark.

111

SO2GDP_trend
Hungary
Cameroon
South Korea
Slovenia
Ireland
Yemen
Macedonia
Taiwan
Ghana
Belarus
Eritrea
Benin
Turkey
Kenya
Lithuania
Senegal
Poland
Estonia
Colombia
El Salvador
Paraguay
Uruguay
Canada
Saudi Arabia
United States of America
Cambodia
Morocco
Syria
India
Haiti
Argentina
Japan
Iceland
Philippines
Sri Lanka
Oman
Tajikistan
Zimbabwe
Serbia
Bulgaria
Sudan
New Zealand
Chile
Viet Nam
Armenia
Namibia
‐40

‐30

‐20

‐10

0

10

20

30

40

Trend scores for a selected number of nations in SO2GDP Indicator

50

60

112
Countries with the
Countries with the
most declining
Trend
most improved
trends
score
trends
Trend score
Algeria
44.6
Namibia
-33.9
Latvia
48.5
Malaysia
-20.3
South Korea
49.5
Brunei Darussalam
-14.9
Angola
50.0
Armenia
-13.7
Côte d'Ivoire
50.0
Mozambique
-10.8
Cameroon
50.0
Iraq
-9.7
Congo
50.0
Viet Nam
-8.2
Gabon
50.0
Indonesia
-7.5
Hungary
50.0
Myanmar
-6.4
Nigeria
50.0
Chile
-6.0
Trend scores for a selected number of nations in SO2GDP Indicator
18 nations showed a decreasing trend in SO2GDP emissions, 5 nations show no change over
this period and 114 nations showed a positive trend in SO2GDP emissions with some nations
meeting the target.
Application to Climate change policy category (CO2 per GDP)
Impacts of climate change such as sea level rise, coastal flooding, droughts,
desertification etc., are being felt globally. CO2 emissions per GDP (CO2GDP) ratio was
obtained using the Sectoral Approach CO2 emissions and the GDP using purchasing power
parities data from the IEA. The target and low performance benchmark for this indicator is
shown in the table below. The target is an estimated value associated with 50% reduction in
global GHG emissions by 2050, against 1990 levels.

Indicators

Target

CO2 per
GDP

0.078 kg CO2
eq.

Low
Statistical
performance
transformation
benchmark

1.533

Transformed
low
Transformed
performance
targets
benchmark

natural
logarithm

-2.04

0.46

EPI framework for C02 emissions per GDP (CO2GDP) Indicator
The slopes of the trends of the proximity to target of 136 nations for CO2GDP emissions are
shown below in the figure below. The table below also shows nations with a decreasing trend

113
in CO2GDP emissions and nations with an increasing trend in Carbon dioxide emissions
towards achieving this indicator target or still at the low performance benchmark.

114

CO2GDP_trend
Ukraine
Romania
Azerbaijan
Zambia
Cuba
Singapore
Sweden
Estonia
Lithuania
Hungary
Ireland
Macedonia
Finland
Libyan Arab Jamahiriya
Venezuela
Portugal
New Zealand
Botswana
Denmark
South Korea
Israel
Australia
Kuwait
Cambodia
Indonesia
Italy
Peru
Kazakhstan
Norway
Côte d'Ivoire
Thailand
Togo
Ecuador
Senegal
Iraq
El Salvador
Malaysia
Algeria
Kyrgyzstan
Bangladesh
Dem. Rep. Congo
Oman
Yemen
Cameroon
Haiti
Benin
‐60

‐40

‐20

0

20

Trend scores for a selected number of nations in CO2GDP Indicator

40

60

115
Countries with
most
Trend
Countries with most
Trend
improvement
score
improvement
Score
Benin
-50.0 Zambia
45.5
Congo
-50.0 Albania
49.0
Brunei Darussalam
-49.5 Nigeria
49.8
Haiti
-35.5 Azerbaijan
50.0
Tanzania
-27.2 Belarus
50.0
Viet Nam
-23.7 Moldova
50.0
Cameroon
-21.8 Romania
50.0
Bolivia
-18.5 Slovakia
50.0
Saudi Arabia
-18.8 Turkmenistan
50.0
Yemen
-17.9 Ukraine
50.0
Trend scores for a selected number of nations in CO2GDP Indicator
33 nations show a declining trend in CO2GDP emissions away from the target, 2 nations
show no change over this period and 101 nations show a much positive trend in CO2GDP
emissions.
Application to Climate change policy category (CO2 per energy use)
This is CO2 emissions per kilowatt hour (CO2KWH) that represents the ratio of CO2
emissions to the electricity generated by thermal power plants, nuclear and hydro production
and geothermal. The target and low performance benchmark for this indicator is shown below
in the table below:

Indicators

Target

Transformed low
Low
Statistical
Transformed
performance
performance
transformation targets
benchmark
benchmark

CO2
0
emissions
natural
grammes
-0.69
6.74
845.3289722
per
logarithm
CO2 per
electricity
KWh
generation
EPI framework for C02 emissions per kilowatt hour (CO2KWH) Indicator
The slopes of the trends of the proximity to target of 136 nations for the CO2KWH are shown
below in the figure below. The table below also shows nations with a decreasing trend in
emissions of CO2KWH and nations with an increasing trend in emissions of CO2KWH
towards achieving this indicator target or still at the low performance benchmark.

116

CO2KWH_trend
Togo
Moldova
Angola
Azerbaijan
Ireland
Mexico
Portugal
Guatemala
Serbia
Sudan
Lebanon
Jordan
Senegal
Greece
Tajikistan
Germany
Switzerland
Czech Republic
Netherlands
Slovenia
Indonesia
Sweden
Bahrain
India
China
Libyan Arab Jamahiriya
Philippines
Pakistan
Malta
Viet Nam
Romania
France
Gibraltar
Jamaica
Hong Kong
Iran
Bolivia
South Korea
Syria
Nigeria
Malaysia
Chile
Tanzania
Namibia
Ethiopia
Cameroon
‐60

‐40

‐20

0

20

Trend score for a selected number of nations in CO2KWH Indicator

40

60

117

Countries with decreasing trend Trend
score
in CO2/KWH emissions

Countries with increasing trend
in CO2/KWH Emissions

Azerbaijan
44.8
Ghana
Qatar
45.9
Luxembourg
Singapore
47.5
Namibia
Angola
50.0
Norway
Albania
50.0
Peru
Armenia
50.0
Tanzania
Moldova
50.0
Uruguay
Myanmar
50.0
Haiti
Mozambique
50.0
Chile
Togo
50.0
Dem. Rep. Congo
Trend scores for a selected number of nations in CO2KWH Indicator

Trend
score
-50.0
-50.0
-50.0
-50.0
-50.0
-50.0
-50.0
-48.6
-44.5
-40.6

58 countries show an increasing trend in CO2KWH emissions with majority of these
countries being in the emerging and developing countries categories. This is as a result of the
expansion of their economies especially in the manufacturing, oil and gas sector etc. 4
countries show no change within this period. 74 nations show a decreasing trend in
CO2KWH emissions with many of these nations coming from both developed and
developing nations. Developed nations with initial high emissions are enforcing policy
regulations and other strategies to curb the level of already high emissions.
Application to Air pollution policy category (Effects on Human Health)
Air pollutions is a high leading environmental factor that causes infections which
leads to high levels of premature death yearly (Emerson, 2012). Particulate matter (PM25)
indicator which is an outdoor air pollution is used as a demonstration from this category.
They are suspended particulates derived from a model by MODIS Aerosol Optical Depth
(AOD) data. The target and low performance benchmark for this indicator is shown in the
table below:

Indicator

Transformed
Low
Statistical
Transformed low
Target performance
performance
transformation targets
benchmark
benchmark

natural
10
Particulate
2.34
48.7916
logarithm
ug/m3
matter
EPI framework for Particulate Matter (PM25) Indicator

3.89

The slopes of the trends of the proximity to target of 155 nations for the Particulate Matter
are shown below in the figure below. The table below also shows nations with a decreasing

118
trend of PM2.5 and nations with an increasing trend in PM2.5 towards achieving this indicator
target or still at the low performance benchmark.

PM25_trend
Iraq
Viet Nam
India
Nepal
Laos
Niger
Venezuela
Tanzania
Sweden
El Salvador
Portugal
Panama
Nicaragua
Mozambique
Madagascar
Lithuania
Kenya
Ireland
French Guiana
Georgia
Spain
Denmark
Costa Rica
Canada
Bolivia
Bulgaria
Eritrea
Mali
Indonesia
Nigeria
Belgium
Armenia
Thailand
Turkey
Sierra Leone
Italy
Cameroon
Malaysia
Angola
‐60

‐40

‐20

0

20

Trend scores for a selected number of nations in PM25 Indicator

40

60

119

Countries with decreasing trend
in PM2.5 concentrations
Angola
Burundi
France
Mexico
Malaysia
Rwanda
Serbia
Uganda
Cameroon

Trend
score
50.0
50.0
50.0
50.0
50.0
50.0
50.0
50.0
48.9

Countries with increasing trend in
PM2.5 concentrations
Bangladesh
Iran
Afghanistan
Viet Nam
Liberia
Saudi Arabia
United Arab Emirates
Iraq
Kuwait

Dem. Rep. Congo
46.2 Oman
Trend scores for a selected number of nations in PM25 Indicator

Trend
score
-26.4
-26.6
-27.0
-27.1
-29.6
-35.4
-50.0
-50.0
-50.0
-50.0

While many nations have no change in the concentration of PM2.5. Out of the 155 nations, 49
nations have improved their air quality as measured by PM2.5 towards the target, 82 nations
have no change in PM2.5, while 24 nations air quality deteriorated which was closer to the
low performance benchmark. From the above table 21, most middle-eastern countries such as
Iran, Iraq, Saudi Arabia are shown to have increased trend in concentrations of PM2.5 due to
the the desert environment in this region which will always contain significant amounts of
windblown dust, mostly during dust storms. Also, expansion of the oil and gas sector and
industrialization are part of the contributors.

120
Appendix B: Definitions of indicators for EPI policy categories
INDOOR (Indoor air pollution); This indicator measures the use of solid fuels which are
used in households.
PM25 (Particulate matter); These are suspended particulates derived from a model by
MODIS Aerosol Optical Depth (AOD) data which causes acute lower respiratory infections
and other diseases detrimental to human health.
WATSUP (Access to drinking water); This indicator measures the percentage of a
country’s population that has access to an improved source of drinking water.
ACSAT (Access to sanitation); It measures the percentage of a country’s population that has
access to an improved source of sanitation such as connection to a public sewer, connection
to septic system, pour-flush latrine, simple pit latrine, ventilated improved pit latrine.
CHMORT (Child mortality); It is described as the probability of dying between a child’s
first and fifths birthdays per 1000 children aged 1. This was chosen because environmental
factors influence child mortality for ages 1 – 4 years old.
SO2CAP (Sulfur dioxide emissions per capita); This is the ratio of SO2 emissions to
population.
SO2GDP (Sulfur dioxide emissions per GDP); This is the ratio of SO2 emissions to GDP in
2005 constant international prices PPP.
WATUSE (Change in water quantity); This measures the area-weighted percent reduction
of mean annual river flow from natural state owing to water withdrawals and reservoirs.
PACOV (Biome Protection); This measures the weighted percentage of biomes under
protected status (weight is determined by relative size of biomes within a country).
MPAEEZ (Marine protection); This measures the percentage of each country’s exclusive
economic zone that is under protection by a marine protected area.
AZE (Critical habitat protection); This is described as the percentage of the total
biodiversity and habitat area that is within protected area.
FORLOSS (Forest loss); This measures the loss of forest area owing to deforestation from
either human or natural causes.
FORCOV (Forest cover change); This measures the change in forest cover area between
time periods of 2005 to 2010.
FORGROW (Growing stock change); This measures the cubic meters of wood over bark
of all living trees more than X cm (vary by country) in diameter at breast height.

121
TCEEZ (Coastal shelf fishing pressure); This measures the catch from trawling and
dredging gears divided the EEZ area by country and year.
FSOC (Fish stock overexploited); This measures the fraction of species that are fished in
each country’s EEZ that are exploited or collapsed.
AGSUB (Agricultural subsidies); This indicator evaluates the magnitude of subsidies to
assess the degree of environmental pressure they exert.
POPs (Pesticide regulation); This indicator examines the legislative status of countries on
one of the landmark agreements on POPS usage, the Stockholm Convention, and rates the
degree to which these countries have kept those objectives by limiting the use of certain toxic
chemicals.
CO2CAP (CO2 emissions per capita); CO2 emissions per capita ratio was obtained using
the Sectoral Approach CO2 emissions and population data from the IEA.
CO2GDP (CO2 emissions per GDP); CO2 emissions per GDP ratio was obtained using the
Sectoral Approach CO2 emissions and the GDP using purchasing power parities data from
the IEA.
CO2KWH (CO2 emissions per electricity generation); This is carbon dioxide emissions
per kilowatt hour that represents the ratio of CO2 emissions to the electricity generated by
thermal power plants, nuclear and hydro production and geothermal.
RENEW (Renewable electricity); This measures the percentage of the total renewable
electricity net generation in total electricity net generation.

122
Appendix C: Cluster Analysis
The primary methods of modern multivariate analysis include; Cluster analysis (CA),
principal components analysis (PCA), and discriminant analysis (DA). These methods focus
on characterizing group differences and assigning unknowns to one of the known groups
(Kettenring, 2006).
PCA is used to reduce dimensionality and visualize data in a reduced number of
dimensions corresponding to the leading PCs. PCA is sometimes used as a method to
find clusters directly, bypassing any of the usual CA algorithms. There is no easy and
rigorous way to quickly extract clusters from complex data, hence, hundreds of different
algorithms have been proposed to achieve results where each has its own benefits and
problems. There is need to be cautious at all phases: the form in which the data are
analyzed, the choice of algorithm and any associated parameters, and the manner in
which outputs are checked for validity (Kettenring, 2006). CA is used to assess the trend of
academic research within a specific discipline (Abson et al, 2014, Kajikawa et al, 2014) on
the assessment of sustainability and economic development of nations (Neri et al, 2017), on
assessment of environmental issues such as impacts of carbon emissions (Lamb et al, 2014),
analysis for cities and planning (Chévez et al, 2017). CA is also widely used commercially.
An example is in market segmentation services which involves k-means clustering and also
has been described in several publications.
CA is as a three-step process that involves preprocessing of the data, invoking algorithms
to assist in identification of clusters, and assessing the results. CA applications involves
certain practices that are observed to stand out as beneficial and some as dangerous. Good
practice involves looking at the data in different forms, considering alternative metrics
and distance functions, comparing the results from different clustering algorithms, and
checking the stability and validity of findings (Kettenring, 2006). Understanding of the
properties of all the methods and processes involved will produce better results. The process
of CA can be further explained below;


Autoscaling; Scales of the variables can have a huge impact on the outcome of a CA
due to the different nature of the variables. It is required to auto scale each of them
separately. This process can obscure clusters in the data and render
them undetectable in the output of a clustering algorithm. Also, simple
transformations of variables, such as taking logs can be very helpful

for

ameliorating scaling problems. Another consideration is differential weighting to

123
intentionally overemphasize those which are more likely to help the CA
(Kettenring, 2006).


PCA an CA; The role of PCA for reducing dimensionality and the number of
variables entering the CA is surrounded with a lot of confusion. Developing viable
alternatives to PCA for reducing dimensionality in CA problems will prove very
useful.



Variable Clustering; This involves clustering variables instead of the objects or
observations. Variable clustering should be done such that it is not thrown off if
the observations themselves are clustered.



DA and CA; DA and CA utilize extra information, such as group labels on some of
the data or constraints designed to keep certain pairs of points in the same or different
clusters (Kettenring, 2006).



Tree Cutting; Another way of obtaining a partition of data into clusters is to perform a
straight line cut of the dendrogram at an appropriate level and then to treat each
separate branch as a cluster. Several software’s are available to perform this.



Very Large Problems; Sampling is often a sensible strategy when observations (n) is
large, especially if it can be done repeatedly, so as not to miss small clusters
completely. The results can be compared across samples and integrated as
appropriate.



Validation and Interpretation; The need for solid validation and careful interpretation
of CA results is clear and has been recognized by many researchers. There are a
variety of approaches that can be applied like simple graphical displays, sensitivity
analyses in ensuring robust results. When using an iterative procedure, such as kmeans CA, experimenting with different starting points is beneficial, this enables a
probability Statement about the chance that the next restart will discover a previously
unobserved local maximum, and the chance that it will be better than any one
previously found.



Circularity; Circularity is used here to refer to the risk of obtaining CA results that are
more due to the vagaries of the process than to the strength of the cluster Structure in
the data (Kettenring, 2006).

Great methods and supporting software are available to support applications. With such
approaches, one can capitalize on the added structure to make a variety of inferences about
the model such as the number of clusters present and their shapes. Some of the approaches to

124
cluster analysis can be grouped into two categories known as hierarchical methods and the
non-hierarchical methods described below;
Hierarchical methods; This consists of agglomerative methods and divisive methods.
 Agglomerative methods are a type of clustering whereby the subjects start in their
own separate cluster and then the most similar clusters are then combined together.
This is done repeatedly until all subjects are in one cluster and an optimum number of
clusters is chosen in the end (Rosie, 2007). There are several methods to determine
which clusters should be joined together at each stage. Some of them include nearest
neighbour method, furthest neighbour method, average linkage method, centroid
method, and wards method among many others. It is usually advisable to try two or
three of the methods so that the results can be confirmed and be much more
believable. The tool for determining the suitable number of clusters is called a
dendrogram (Řezanková, 2014).
Divisive methods involve all subjects starting in the same cluster and the two farthest clusters
are then separated. This is done repeatedly until all subjects are in a separate cluster.

125

Appendix D: HDI Classification in 2000 and Codes
Very high
development
Norway
Australia
Switzerland
United States
Netherlands
Sweden
Belgium

High Development
United Arab Emirates
Bahrain
Montenegro
Kuwait
Poland
Malta
Antigua and Barbuda

New Zealand
Canada
United Kingdom
Liechtenstein
Denmark
Germany

Portugal
Estonia
Bahamas
Argentina
Hungary
Slovakia

Ireland
Finland
Japan
Iceland
Luxembourg
Israel
France
Austria
Italy
Spain
Hong Kong,
China
Slovenia
Czech Republic
Singapore
South Korea
Brunei
Andorra
Qatar
Greece
Cyprus

Chile
Lebanon
Lithuania
Barbados
Croatia
Uruguay
Saudi Arabia
Grenada
Saint Kitts and Nevis
Palau
FYROM
Libya
Latvia
Malaysia
Panama
Russian Federation
Trinidad and Tobago
Seychelles
Bulgaria
Bosnia and
Herzegovina
Serbia

Medium
Development
Dominica
Cuba
Sri Lanka
Brazil
Kazakhstan
Saint Lucia
Fiji
Belarus
Jamaica
Peru
Belize
Tonga
St. Vincent/
Grenadines
Mauritius
Ukraine
Georgia
Venezuela
Ecuador
Palestine, State of
Iran
Turkmenistan
Albania
Dominican Republic
Tunisia

Low Development
Guatemala
Tajikistan
Morocco
Equatorial Guinea
Swaziland
Nigeria
Sao Tome and
Principe
India
Congo
Ghana
Comoros
Timor-Leste
Bangladesh
Lao Republic
Madagascar
Pakistan
Kenya
Nepal
Mauritania
Yemen
Lesotho
Haiti
Solomon Islands
Cameroon

Colombia
Turkey
Thailand
Samoa
Algeria
Armenia
Azerbaijan
Gabon
South Africa

South Sudan
Zimbabwe
Myanmar
Togo
Zambia
Papua New Guinea
Cambodia
Guinea-Bissau
Eritrea

Paraguay

Sudan

126
Very high
development

High
Development
Romania
Costa Rica
Jordan
Oman
Suriname
Mexico

Medium
Development
Philippines
El Salvador
Egypt
Bolivia
Iraq
Guyana
Indonesia
Micronesia
Moldova
Uzbekistan
Kyrgyzstan
China
Vanuatu
Syrian
Mongolia
Maldives
Kiribati
Viet Nam
Bhutan
Nicaragua
Cabo Verde
Botswana
Honduras
Namibia

Low Development
Uganda
Benin
Côte d'Ivoire
Angola
Tanzania
Malawi
Liberia
Gambia
Senegal
Burkina Faso
Djibouti
Afghanistan
Rwanda
Congo
Guinea
Central African
Republic
Sierra Leone
Chad
Mozambique
Mali
Ethiopia
Burundi
Niger

Nations for the study

Acronyms

HDI 2000

Afghanistan

AFG

L

Albania

ALB

M

Algeria

DZA

M

Andorra

AND

VH

Angola

AGO

L

Antigua and Barbuda

AIA

H

Argentina

ARG

H

Armenia

ARM

M

Aruba

ABW

VH

Australia

AUS

VH

Austria

AUT

VH

Azerbaijan

AZE

M

Bahamas

BHS

H

Bahrain

BHR

H

127
Bangladesh

BGD

L

Barbados

BRB

H

Belarus

BLR

M

Belgium

BEL

VH

Belize

BLZ

M

Benin

BEN

L

Bermuda

BMU

VH

Bhutan

BTN

M

Bolivia

BOL

M

Bosnia and Herzegovina

BIH

H

Botswana

BWA

M

Brazil

BRA

M

Brunei

BRN

VH

Bulgaria

BGR

H

Burkina Faso

BFA

L

Burundi

BDI

L

Cambodia

KHM

L

Cameroon

CMR

L

Canada

CAN

VH

Cape Verde

CPV

M

Cayman Islands

CYM

VH

Central African Republic

CAF

L

Chad

TCD

L

Chile

CHL

H

China

CHN

M

Colombia

COL

M

Comoros

COM

L

Congo, Dem. Rep.

COD

L

Costa Rica

CRI

H

Cote d'Ivoire

CIV

L

Croatia

HRV

H

Cuba

CUB

M

Cyprus

CYP

VH

Czech Republic

CZE

VH

128
Denmark

DNK

VH

Djibouti

DJI

L

Dominica

DMA

M

Dominican Republic

DOM

M

Ecuador

ECU

M

Egypt

EGY

M

El Salvador

SLV

M

Equatorial Guinea

GNQ

L

Eritrea

ERI

L

Estonia

EST

H

Ethiopia

ETH

L

Fiji

FJI

M

Finland

FIN

VH

France

FRA

VH

Gabon

GAB

M

Gambia

GMB

L

Georgia

GEO

M

Germany

DEU

VH

Ghana

GHA

L

Greece

GRC

VH

Greenland

GRL

VH

Grenada

GRD

H

Guatemala

GTM

L

Guinea

GIN

L

Guinea-Bissau

GNB

L

Guyana

GUY

M

Haiti

HTI

L

Honduras

HND

M

Hong Kong, China

HKG

VH

Hungary

HUN

H

Iceland

ISL

VH

India

IND

L

Indonesia

IDN

M

Iran

IRN

M

129
Iraq

IRQ

M

Ireland

IRL

VH

Israel

ISR

VH

Italy

ITA

VH

Jamaica

JAM

M

Japan

JPN

VH

Jordan

JOR

H

Kazakhstan

KAZ

M

Kenya

KEN

L

Kiribati

KIR

M

Kuwait

KWT

H

Kyrgyz Republic

KGZ

M

Lao

LAO

L

Latvia

LVA

H

Lebanon

LBN

H

Lesotho

LSO

L

Liberia

LBR

L

Libya

LBY

H

Lithuania

LTU

H

Luxembourg

LUX

VH

Macao, China

MAC

VH

Macedonia, FYR

MKD

H

Madagascar

MDG

L

Malawi

MWI

L

Malaysia

MYS

H

Maldives

MDV

M

Mali

MLI

L

Malta

MLT

H

Marshall Islands

MHL

H

Mauritania

MRT

L

Mauritius

MUS

M

Mexico

MEX

H

Micronesia, Fed. Sts.

FSM

M

Moldova

MDA

M

130
Monaco

MCO

VH

Mongolia

MNG

M

Montenegro

MNE

H

Morocco

MAR

L

Mozambique

MOZ

L

Myanmar

MMR

L

Namibia

NAM

M

Nauru

NRU

H

Nepal

NPL

L

Netherlands

NLD

VH

New Zealand

NZL

VH

Nicaragua

NIC

M

Niger

NER

L

Nigeria

NGA

L

North Korea

PRK

VH

Norway

NOR

VH

Oman

OMN

H

Pakistan

PAK

L

Palau

PLW

H

Panama

PAN

H

Papua New Guinea

PNG

L

Paraguay

PRY

M

Peru

PER

M

Philippines

PHL

M

Poland

POL

H

Portugal

PRT

H

Puerto Rico

PRI

VH

Qatar

QAT

VH

Romania

ROU

H

Russia

RUS

H

Rwanda

RWA

L

Samoa

WSM

M

San Marino

SMR

VH

Sao Tome and Principe

STP

L

131
Saudi Arabia

SAU

H

Senegal

SEN

L

Serbia

SRB

H

Seychelles

SYC

H

Sierra Leone

SLE

L

Singapore

SGP

VH

Slovak Republic

SVK

H

Slovenia

SVN

VH

Solomon Islands

SLB

L

Somalia

SOM

L

South Africa

ZAF

M

South Korea

KOR

VH

Spain

ESP

VH

Sri Lanka

LKA

M

Sudan

SDN

L

Suriname

SUR

H

Swaziland

SWZ

L

Sweden

SWE

VH

Switzerland

CHE

VH

Syria

SYR

M

Taiwan

TWN

VH

Tajikistan

TJK

L

Tanzania

TZA

L

Thailand

THA

M

Timor-Leste

TLS

L

Togo

TGO

L

Tonga

TON

M

Trinidad and Tobago

TTO

H

Tunisia

TUN

M

Turkey

TUR

M

Turkmenistan

TKM

M

Turks and Caicos Islands

TCA

VH

Tuvalu

TUV

M

Uganda

UGA

L

132
Ukraine

UKR

M

United Arab Emirates

ARE

H

United Kingdom

GBR

VH

United States

USA

VH

Uruguay

URY

H

Uzbekistan

UZB

M

Vanuatu

VUT

M

Venezuela

VEN

M

Vietnam

VNM

M

Yemen

YEM

L

Zambia

ZMB

L

Zimbabwe

ZWE

L

133
Appendix E: Logistic results
Eviews
Dependent Variable: PROB_1 childmortality
Method: ML - Binary Logit (Newton-Raphson / Marquardt steps)
Date: 04/17/18 Time: 12:57
Sample: 1 196
Included observations: 149
Convergence achieved after 4 iterations
Coefficient covariance computed using observed Hessian
GLM adjusted covariance (variance factor =0.988453669828)
Variable

Coefficient

Std. Error

z-Statistic

Prob.

C

2.340819

1.802563

1.298606

0.1941

LNGDP

-0.753160

0.325281

-2.315410

0.0206

HDI_2000

0.004087

0.024771

0.164993

0.8689

AV_INV

0.160248

0.038123

4.203474

0.0000

FGE_CHANGE

1.058351

0.732913

1.444033

0.1487

PS_CHANGE

0.822252

0.377028

2.180878

0.0292

HEALTH_CH

0.061322

0.057206

1.071940

0.2837

McFadden R-squared

0.248979

Mean dependent var

0.436242

S.D. dependent var

0.497591

S.E. of regression

0.422438

Akaike info criterion

1.122850

Sum squared resid

25.34049

Schwarz criterion

1.263975

Log likelihood

-76.65234

Hannan-Quinn criter.

1.180187

Deviance

153.3047

Restr. deviance

204.1284

Restr. log likelihood

-102.0642

LR statistic

50.82376

Avg. log likelihood

-0.514445

Prob(LR statistic)

0.000000

Obs with Dep=0

84

Obs with Dep=1

65

Total obs

149

134

Dependent Variable: PROB_1 air(human health)
Method: ML - Binary Logit (Newton-Raphson / Marquardt steps)
Date: 04/17/18 Time: 13:00
Sample: 1 196
Included observations: 150
Convergence achieved after 4 iterations
Coefficient covariance computed using observed Hessian
GLM adjusted covariance (variance factor =1.00572933837)
Variable

Coefficient

Std. Error

z-Statistic

Prob.

C

0.787797

1.559912

0.505027

0.6135

LNGDP

-0.303108

0.282462

-1.073093

0.2832

HDI_2000

0.008498

0.021720

0.391246

0.6956

AV_INV

0.027166

0.025701

1.057034

0.2905

FGE_CHANGE

0.130019

0.659414

0.197173

0.8437

PS_CHANGE

0.670090

0.347106

1.930507

0.0535

HEALTH_CH

0.017611

0.052394

0.336127

0.7368

McFadden R-squared

0.051425

Mean dependent var

0.333333

S.D. dependent var

0.472984

S.E. of regression

0.466497

Akaike info criterion

1.300896

Sum squared resid

31.11952

Schwarz criterion

1.441392

Log likelihood

-90.56721

Hannan-Quinn criter.

1.357975

Deviance

181.1344

Restr. deviance

190.9543

Restr. log likelihood

-95.47713

LR statistic

9.819835

Avg. log likelihood

-0.603781

Prob(LR statistic)

0.132447

Obs with Dep=0

100

Obs with Dep=1

50

Total obs

150

135
Dependent Variable: PROB_1/ water(human health)
Method: ML - Binary Logit (Newton-Raphson / Marquardt steps)
Date: 04/17/18 Time: 13:02
Sample: 1 196
Included observations: 147
Convergence achieved after 4 iterations
Coefficient covariance computed using observed Hessian
GLM adjusted covariance (variance factor =1.04038755889)
Variable

Coefficient

Std. Error

z-Statistic

Prob.

C

3.921657

1.874373

2.092250

0.0364

LNGDP

-1.173327

0.369151

-3.178445

0.0015

HDI_2000

0.065647

0.027183

2.415008

0.0157

AV_INV

0.050768

0.029132

1.742662

0.0814

FGE_CHANGE

-0.328025

0.728477

-0.450288

0.6525

PS_CHANGE

-0.124211

0.379448

-0.327346

0.7434

HEALTH_CH

0.055830

0.056257

0.992416

0.3210

McFadden R-squared

0.096865

Mean dependent var

0.278912

S.D. dependent var

0.449997

S.E. of regression

0.430587

Akaike info criterion

1.164410

Sum squared resid

25.95674

Schwarz criterion

1.306812

Log likelihood

-78.58415

Hannan-Quinn criter.

1.222270

Deviance

157.1683

Restr. deviance

174.0252

Restr. log likelihood

-87.01259

LR statistic

16.85688

Avg. log likelihood

-0.534586

Prob(LR statistic)

0.009824

Obs with Dep=0

106

Obs with Dep=1

41

Total obs

147

136
Dependent Variable:
PROB EV –BH GDP > 2, EV => 0
Method: ML - Binary Logit (Newton-Raphson / Marquardt steps)
Date: 04/18/18 Time: 13:02
Sample: 1 196
Included observations: 151
Convergence achieved after 4 iterations
Coefficient covariance computed using observed Hessian
GLM adjusted covariance (variance factor =0.977306368645)
Variable

Coefficient

Std. Error

z-Statistic

Prob.

C

4.531828

1.840797

2.461883

0.0138

LN_GDP

-1.452538

0.366990

-3.957977

0.0001

EVBH_2000

-0.001954

0.007389

-0.264475

0.7914

HDI_2000

0.098468

0.027433

3.589385

0.0003

FGE_CHANGE

0.869390

0.722834

1.202753

0.2291

PS_CHANGE

0.707928

0.389059

1.819589

0.0688

AV_INV

0.062258

0.028554

2.180380

0.0292

McFadden R-squared

0.182410

Mean dependent var

0.344371

S.D. dependent var

0.476744

S.E. of regression

0.430141

Akaike info criterion

1.145595

Sum squared resid

26.64301

Schwarz criterion

1.285469

Log likelihood

-79.49240

Hannan-Quinn criter.

1.202419

Deviance

158.9848

Restr. deviance

194.4554

Restr. log likelihood

-97.22772

LR statistic

35.47064

Avg. log likelihood

-0.526440

Prob(LR statistic)

0.000003

Obs with Dep=0

99

Obs with Dep=1

52

Total obs

151

137
Dependent Variable: PROB
EV-AG GDP > 2, EV > 0
Method: ML - Binary Logit (Newton-Raphson / Marquardt steps)
Date: 04/18/18 Time: 13:05
Sample: 1 196
Included observations: 151
Convergence achieved after 5 iterations
Coefficient covariance computed using observed Hessian
GLM adjusted covariance (variance factor =0.956023681118)
Variable

Coefficient

Std. Error

z-Statistic

Prob.

C

-0.913061

1.960877

-0.465639

0.6415

LNGDP

-0.574984

0.345914

-1.662217

0.0965

EVAG_2000

0.009060

0.008426

1.075166

0.2823

HDI_2000

0.050926

0.027508

1.851307

0.0641

FGE_CHANGE

1.642096

0.768074

2.137939

0.0325

PS_CHANGE

-0.420510

0.417978

-1.006058

0.3144

AV_INV

0.031287

0.032765

0.954890

0.3396

McFadden R-squared

0.073953

Mean dependent var

0.192053

S.D. dependent var

0.395225

S.E. of regression

0.390366

Akaike info criterion

0.998734

Sum squared resid

21.94357

Schwarz criterion

1.138608

Log likelihood

-68.40444

Hannan-Quinn criter.

1.055558

Deviance

136.8089

Restr. deviance

147.7342

Restr. log likelihood

-73.86711

LR statistic

10.92533

Avg. log likelihood

-0.453010

Prob(LR statistic)

0.090712

Obs with Dep=0

122

Obs with Dep=1

29

Total obs

151

138
Dependent Variable: PROB
EV-FORESTS GDP > 2, EV > -10
Method: ML - Binary Logit (Newton-Raphson / Marquardt steps)
Date: 04/18/18 Time: 13:18
Sample: 1 196
Included observations: 148
Convergence achieved after 6 iterations
Coefficient covariance computed using observed Hessian
GLM adjusted covariance (variance factor =1.02081117144)
Variable

Coefficient

Std. Error

z-Statistic

Prob.

C

-0.399116

2.085267

-0.191398

0.8482

LNGDP

-1.188948

0.397057

-2.994400

0.0027

EVFOREST_2000

0.064288

0.015155

4.242128

0.0000

HDI_2000

0.044396

0.028271

1.570376

0.1163

FGE_CHANGE

1.121019

0.865370

1.295422

0.1952

PS_CHANGE

0.201529

0.489639

0.411587

0.6806

AV_INV

0.087209

0.033593

2.596038

0.0094

McFadden R-squared

0.297855

Mean dependent var

0.256757

S.D. dependent var

0.438327

S.E. of regression

0.364384

Akaike info criterion

0.894531

Sum squared resid

18.72138

Schwarz criterion

1.036291

Log likelihood

-59.19527

Hannan-Quinn criter.

0.952128

Deviance

118.3905

Restr. deviance

168.6126

Restr. log likelihood

-84.30630

LR statistic

50.22206

Avg. log likelihood

-0.399968

Prob(LR statistic)

0.000000

Obs with Dep=0

110

Obs with Dep=1

38

Total obs

148

139
Dependent Variable: PROB
EV_FISH
Method: ML - Binary Logit (Newton-Raphson / Marquardt steps)
Date: 04/18/18 Time: 13:21
Sample (adjusted): 2 194
Included observations: 118 after adjustments
Convergence achieved after 5 iterations
Coefficient covariance computed using observed Hessian
GLM adjusted covariance (variance factor =1.0800807778)
Variable

Coefficient

Std. Error

z-Statistic

Prob.

C

-1.170811

3.080845

-0.380029

0.7039

LN_GDP

-0.422360

0.464483

-0.909311

0.3632

EVFISH_2000

-0.052511

0.025096

-2.092380

0.0364

HDI_2000

0.024691

0.036500

0.676478

0.4987

FGE_CHANGE

0.698047

1.072088

0.651110

0.5150

PS_CHANGE

1.318861

0.621928

2.120601

0.0340

AV_INV

0.150027

0.052745

2.844380

0.0044

McFadden R-squared

0.237322

Mean dependent var

0.220339

S.D. dependent var

0.416243

S.E. of regression

0.367036

Akaike info criterion

0.923020

Sum squared resid

14.95338

Schwarz criterion

1.087383

Log likelihood

-47.45821

Hannan-Quinn criter.

0.989757

Deviance

94.91641

Restr. deviance

124.4515

Restr. log likelihood

-62.22573

LR statistic

29.53504

Avg. log likelihood

-0.402188

Prob(LR statistic)

0.000048

Obs with Dep=0

92

Obs with Dep=1

26

Total obs

118

140
Dependent Variable: PROB
EV-WATER
Method: ML - Binary Logit (Newton-Raphson / Marquardt steps)
Date: 04/18/18 Time: 13:23
Sample: 1 196
Included observations: 150
Convergence achieved after 8 iterations
Coefficient covariance computed using observed Hessian
GLM adjusted covariance (variance factor =0.574778886279)
Variable

Coefficient

Std. Error

z-Statistic

Prob.

C

-6.467667

3.018477

-2.142692

0.0321

LN_GDP

-0.397214

0.478357

-0.830371

0.4063

EVWATER_2000

0.060461

0.015416

3.921972

0.0001

HDI_2000

0.023228

0.039105

0.593992

0.5525

FGE_CHANGE

-0.825834

1.301628

-0.634463

0.5258

PS_CHANGE

1.771936

0.681703

2.599278

0.0093

AV_INV

0.115888

0.061837

1.874089

0.0609

McFadden R-squared

0.352738

Mean dependent var

0.073333

S.D. dependent var

0.261556

S.E. of regression

0.227248

Akaike info criterion

0.432728

Sum squared resid

7.384738

Schwarz criterion

0.573224

Log likelihood

-25.45461

Hannan-Quinn criter.

0.489807

Deviance

50.90922

Restr. deviance

78.65314

Restr. log likelihood

-39.32657

LR statistic

27.74392

Avg. log likelihood

-0.169697

Prob(LR statistic)

0.000105

Obs with Dep=0

139

Obs with Dep=1

11

Total obs

150

141
Dependent Variable: PROB
EV-CC
Method: ML - Binary Logit (Newton-Raphson / Marquardt steps)
Date: 04/18/18 Time: 13:26
Sample (adjusted): 2 196
Included observations: 122 after adjustments
Convergence achieved after 5 iterations
Coefficient covariance computed using observed Hessian
GLM adjusted covariance (variance factor =1.19748332424)
Variable

Coefficient

Std. Error

z-Statistic

Prob.

C

5.847042

3.889508

1.503286

0.1328

LN_GDP

-1.195622

0.483680

-2.471925

0.0134

EVCLIMAT_2000

-0.016323

0.016068

-1.015872

0.3097

HDI_2000

0.058848

0.035448

1.660142

0.0969

FGE_CHANGE

-0.643627

1.029973

-0.624897

0.5320

PS_CHANGE

1.666996

0.581733

2.865569

0.0042

AV_INV

0.034680

0.044262

0.783513

0.4333

McFadden R-squared

0.187168

Mean dependent var

0.278689

S.D. dependent var

0.450203

S.E. of regression

0.404607

Akaike info criterion

1.076676

Sum squared resid

18.82629

Schwarz criterion

1.237562

Log likelihood

-58.67724

Hannan-Quinn criter.

1.142023

Deviance

117.3545

Restr. deviance

144.3773

Restr. log likelihood

-72.18867

LR statistic

27.02287

Avg. log likelihood

-0.480961

Prob(LR statistic)

0.000143

Obs with Dep=0

88

Obs with Dep=1

34

Total obs

122

142
Dependent Variable: PROB
EV-AIR
Method: ML - Binary Logit (Newton-Raphson / Marquardt steps)
Date: 04/18/18 Time: 13:28
Sample (adjusted): 2 196
Included observations: 121 after adjustments
Convergence achieved after 4 iterations
Coefficient covariance computed using observed Hessian
GLM adjusted covariance (variance factor =0.986970164726)
Variable

Coefficient

Std. Error

z-Statistic

Prob.

C

4.416574

2.209127

1.999239

0.0456

LN_GDP

-0.729646

0.323745

-2.253767

0.0242

EVAIR_2000

-0.020913

0.012033

-1.738028

0.0822

HDI_2000

0.031771

0.025785

1.232126

0.2179

FGE_CHANGE

1.554651

0.774462

2.007395

0.0447

PS_CHANGE

0.194003

0.401154

0.483612

0.6287

AV_INV

0.015752

0.033853

0.465291

0.6417

McFadden R-squared

0.107192

Mean dependent var

0.388430

S.D. dependent var

0.489420

S.E. of regression

0.467592

Akaike info criterion

1.308566

Sum squared resid

24.92526

Schwarz criterion

1.470306

Log likelihood

-72.16823

Hannan-Quinn criter.

1.374255

Deviance

144.3365

Restr. deviance

161.6658

Restr. log likelihood

-80.83290

LR statistic

17.32934

Avg. log likelihood

-0.596432

Prob (LR statistic)

0.008146

Obs with Dep=0

74

Obs with Dep=1

47

Total obs

121