The Income Effect of Environmental Wellbeing –
Evidence from the EKC-Hypothesis
Jannik L.T. Zwahlen (s3565734)
Supervised by dr. Gerard Kuper May 2019
EBM877A20
Abstract
This thesis aims to model the causal income effect, measured as real GDP per capita, of the environ-mental wellbeing dimension of the Sustainable Society Index (SSI). The hypothesis is derived from the underlying theory of the environmental Kuznets curve (EKC), which suggests that environmental deg-radation (environmental wellbeing) first rises (falls) when income per capita grows and then starts to fall (rise) after a certain threshold of income is reached.
Introduction
In 1926, the Russian mineralogist and geochemist Vladimir Ivanovich Vernadsky published his most famous book “The Biosphere”. Vernadsky was the first to examine the relationship be-tween life and its natural environment. The leading hypothesis of his book is that the main force that shapes earth is life. In conclusion, he argues that life can be seen as a geological force which can have a strong impact on the climate, earths landform and, moreover, on the content of the atmosphere.
After the publication of his work, the impact on the ecological system through life en-hanced. The dimension life itself grew, mainly because over time more and more humans pop-ulated earth. But also, the extent to which this life influences the environment expanded. This is mainly due to economic growth since the people can afford more products i.e. more natural resources are needed to fulfil the desires. In 1972, almost half a century after the publication of “The Biosphere”, the Club of Rome published a study with the title “The Limits of Growth”. The main objective is to determine the relationship between economic growth and the corre-sponding compatibility with the environment. They take in a forward-looking perspective and were the first to use statistical and mathematical models to estimate the future impact of human activity on nature. The conclusions are rather terrifying. The study, therefore, triggered a con-troversial debate whether constant economic growth is possible. Today, 47 years after “The Limits of Growth”, politicians and economists are still divided when it comes to answering this question. As in 1987 the World Commission on Environment and Development of the United Nations published the Brundtland Report, a worldwide discourse about a paradigm shift of the predominant guiding principle of economic development sparked. Consequently, the term sus-tainable development established with the associated definition that it is a “development that meets the needs of the present without compromising the ability of future generations to meet their own needs” (Brundtland Commission, 1987). Sustainable Development, therefore, consti-tutes a new view on development which includes, next to the economic dimension, also an ecological as well as a social one.
Objectives of the Thesis
The present thesis will investigate the validity of the theory of the EKC in practice. Economic growth, measured as real GDP per capita, will serve as the (main) independent variable in the corresponding regression analysis, while the environmental wellbeing dimension of the Sus-tainable Society Index (SSI) will be the dependent variable. The SSI (see chapter 5 for more details) is available for 154 countries1 (see appendix A). Since the theory of the environmental Kuznets curve argues – among other things – that the richer a society is, the more it cares about environmental issues, or the more efficient and therefore environmentally friendly the technol-ogy standard becomes, those countries will be divided into the four income classes defined by the World Bank2 to better understand the relationship between wealth, measured as GDP per capita, and sustainability.
The existing literature mainly focuses on the effect of economic growth, measured as GDP3, on environmental condition indicators such as the concentration of certain pollutants. The present thesis aims to expand the view of the environmental Kuznets curve to a more gen-eral view of sustainable development and defines sustainable development according to the Brundtland Report. Hence, this thesis is not interested in the development of certain pollutants, since those might be substitutable for instance, or in the development of a sustainability index in a certain country, since the negative influence of economic activity on the environment might be outsourced, but in a global inspection of the relationship between income and sustainability in general. In that sense, the empirical research of this thesis can be distinguished from the existing literature. Due to this and in regard to the importance of the issue, the research of this thesis can be seen as relevant.
The research question then relates, on the one hand, to the environmental Kuznets curve, which constitutes the theoretical foundation and, on the other hand, to the environmental well-being dimension of the Sustainable Society Index. The corresponding research question is: Does the environmental dimension of sustainable development, as defined according to the definition of the United Nations and measured as the environmental wellbeing index of the
1 In this thesis, the term country is interchangeably used with the term economy. By doing so, the author follows the classification of countries according to the World Bank. The World Bank states, that the term country or
economy refers to any territory for which authorities report separate economic or social statistics and does not
imply political independence (World Bank, 2019a).
2 Low-income (GNI per capita of $995 or less in 2017), lower middle-income (GNI per capita between $996 and $3.895 in 2017), upper middle-income (GNI per capita between $3.896 and $12.055 in 2017) and high-income (GNI per capita of $12.056 or more in 2017) (World Bank, 2019b).
Sustainable Society Index, follow the same prediction as the environmental quality, according the environmental Kuznets cure, i.e. is the relationship negative in the beginning and does it become positive after a certain value of economic development?
Structure
The present thesis is structured as follows: Part two gives a brief and intuitive but adequate introduction to the environmental Kuznets curve. The underlying thoughts of this theoretical concept are explained and the present critique on it summarized, from an empirical as well as from a theoretical point of view. Moreover, the modern concept of sustainable development is presented. The third part gives an overview about the existing related literature, i.e. former research that aimed to verify the inverted U-shaped relationship of income and environmental degradation according to the EKC. Again, it is divided between empirical and theoretical re-search. In part four, the methodology is explained, while in part five, the data is presented. The empirical model is described in part six. Part six moreover focuses on the results of the regres-sion, discusses those and mentions the limitations of the applied research, before part seven concludes the findings.
2. Theory
Environmental Kuznets Curve
The environmental Kuznets curve is an amendment of the original Kuznets curve which postu-lates an inverted U-shaped relationship between economic development and income inequality. The hypothesis of this relationship was developed by the economist Simon Kuznets in the 1950s and 1960s. The Kuznets curve implies that economic inequality first rises during the economic development of a certain country and then starts to reduce again after a certain threshold (also referred to as turning point) is reached (Kuznets, 1955).
Figure 2.1 visualizes the environmental Kuznets curve, on the one hand, with the inverted U-shaped relationship between income and environmental degradation which is here represented by the pollution load (left hand side y-axis) and, on the other hand, with a U-shaped relationship between income and environmental quality (right hand side y-axis) which has the same inter-pretation. If, in general, the hypothesis of the environmental Kuznets curve holds, it could in-validate the argument that economic development, measured as GDP growth, poses a threat to the environment in the long run4 (see for example Meadows, 1972).
Figure 2.1: EKC Hypothesis for Pollution Load and for Environmental Quality.
Source: Wang et al. (2015)
Since the environmental wellbeing index of the SSI, the dependent variable in the present re-search, is a measure of the environmental quality, the present thesis focuses on the U-shaped EKC, rather than on the inversely U-shaped one.
The hypothesis of the environmental Kuznets curve has its origin in the field of environ-mental economics and was derived trough data observation (Uchiyama, 2016). Subsequent, three exemplary scatter diagrams from previous research of the EKC will be presented.
In Figure 2.2 below, the EKC is displayed with CO2 emissions per capita on the vertical axis and with GDP per capita (in 2005 USD) on the horizontal axis for China. One can clearly see the upwards sloping trend of the EKC in that case. China is referred to as a developing country. The upwards sloping EKC curve is therefore totally consistent with the underlying
theoretical hypothesis. There is only a small decline to observe which is due to the Asian cur-rency crisis (Uchiyama, 2016).
Figure 2.2: Illustration of the EKC Hypothesis in Practice by the Example of China. Relationship between CO2 Emissions Per Capita and State of Economic Development.
Source: Uchiyama (2016)
Figure 2.3: Illustration of the EKC Hypothesis in Practice by the Example of Sweden. Relationship between CO2 Emissions Per Capita and State of Economic Development.
Figure 2.3 plots the case for Sweden. Again, there are CO2 emissions per capita on the y-axis and GDP per capita on the x-axis. One can clearly observe the course of the EKC. One can also observe that Sweden already reached its turning point. This happened at a GDP per capita of about 20.000 USD (measured in 2005 USD). It seems that Sweden is now in the downward-sloping stage of the EKC. Hence, while the income is increasing in Sweden, the CO2 emissions are decreasing.
Figure 2.4 is plotting the same relationship as in the two examples before for the case of Japan. The figure differs from that of Sweden, although Japan is a developed country as well. First, the CO2 emissions raised as income grew. After a certain level (at a GDP per capita of about 17.000), the relationship became negative. So far, the EKC hypothesis is met. However, the CO2 emissions start to increase again after a GDP per capita level of approximately 23.000.
Figure 2.4: Illustration of the EKC Hypothesis in Practice by the Example of Japan. Relationship between CO2 Emissions Per Capita and State of Economic Development.
Source: Uchiyama (2016)
Criticism
As will be shown in chapter 3, there have been numerous studies on the EKC. Both, from an empirical as well as from a theoretical point of view. However, there is also criticism which mainly argues that the underlying hypothesis of the EKC is rather naive. The criticism of the EKC can be divided into two categories: in empirical aspects and in conceptual and theoretical aspects.
Empirical Aspects
According to Uchiyama (2016) one of the main problems from an empirical point of view lies in the nature of the model formulation. He argues that up to now, there are only a few empirical studies which are in line with the underlying theoretical concept. The model which is mainly used to verify the existence of the EKC is a standard regression model (see chapter 3). It often uses GDP per capita as independent variable. The value of the estimated parameters is, how-ever, difficult to interpret economically. This, in turn, leads to a gap between the theoretical and empirical results (Uchiyama, 2016). But, Uchiyama (2016) defines this kind of estimation model as inevitable since the availability and the quality of environmental indicator data forces researcher to make us of it. Furthermore, the more or less usual problems arise while construct-ing a regression model to investigate the EKC. These are omitted variables, simultaneity and heteroscedasticity (Uchiyama, 2016). Also, a problem with the stationarity of data may arise as Stern & Common (2001) point out.
Conceptual and Theoretical Aspects
so that the question arises whether the developed countries just outsource their pollution-inten-sive industries and that this might be an explanation of the occurrence of the EKC and (iv) that the income distribution within a certain country does not play a role yet in research so that worldwide pollution level is expected to increase even if the existence of the EKC is confirmed since a large share of the world’s population belongs to low-income countries.
Sustainable Development
In the following subparts, the notion of sustainable development will be addressed. The under-standing of the modern concept of sustainable development is important in order to classify the findings of the present thesis.
Definition of terms
This thesis understands the term sustainable development as defined by the Brundtland Report, which is the final document of “Our Common Future” (1978). The corresponding definition of sustainable development is a “development that meets the needs of the present generation with-out compromising the ability of future generations to meet their own needs” (Brundtland Com-mission, 1987). Based on the Brundtland Report, sustainable development is usually presented as a three-dimensional model with the dimensions environment, economy and society (Wachter, 2014). The ecological dimension of sustainability states, that satisfying the needs of the present as well as of the future generations is only possible if nature is preserved as liveli-hood. But sustainable development is more than just environmental protection. Economic well-being is a prerequisite for human need satisfaction, which is the reason why the economic di-mension of sustainability also has its relevance. Economic activity can satisfy needs such as housing or nutrition, but it can also help to reduce unemployment, for example. Ultimately, the social dimension is part of sustainable development. Social values, traditions and knowledge contribute – according to the modern concept of sustainable development – to cohesion and peace in a society (Hilser, 2014).
Figure 2.5: Visualization of the three Dimensions of Sustainable Development. Social Dimension
Bearable Equitable
Environmental Dimension Economic Dimension
Viable
Based on: Kropp (2019)
The interface between the social and the environmental dimension indicates the bearable spot. Equitable is everything in the interface between the social and the economic dimension, while the interface between the environmental and the economic dimension visualizes the via-ble. In order to reach a sustainable development, all those three dimensions as well as criterions (bearable, equitable and viable) have to be taken into account and, moreover, have to be satis-fied (Kropp, 2019).
Measurability of Sustainable Development
Since the concept of sustainable development is not directly measurable, corresponding indica-tors are necessary. Such indicaindica-tors are measurable values that allow one to make a statement about whether a certain region (e.g. a country) is in a sustainable state or not. By bundling the parameters, which are determined by indicators, one obtains an index. If the measurements have been taken over a longer period of time, this can provide information about the condition and development of the observed unit (Meyer, 2004).
There are several such indices available. The Ecological Footprint, for example, which records in which areas and with what intensity the human burdens the environment. It is thus a one-dimensional indicator for measuring sustainable development as it captures only the envi-ronmental dimension of sustainability. Another index are the Genuine Savings, which, in con-trast to the ecological footprint, not only consider the environmental dimension but also the
economic dimension in order to make sustainable development measurable. Genuine Savings – also known as Adjusted Net Savings – show how sustainable an economy is, by taking human and natural capital into account.
Figure 2.6: Calculation of Genuine Savings: A Country Example.
Source: World Bank (2006)
Figure 2.6 shows how the Genuine Savings are calculated for Bolivia in 2003. Are the Genuine Savings negative – as in the example above – the development of the economy is considered to be not sustainable, because more resources are being used than “saved” (World Bank, 2006).
For the present thesis, the environmental wellbeing dimension of the Sustainable Society Index (SSI) serves as dependent variable. The SSI is presented in more detail in chapter 5.
3. Literature
Robalino-López et al. (2014) focused on system dynamics and Li et al. (2007) investigated the relation-ship of economic growth and environmental degradation with the aid of meta-analysis.
The above-mentioned studies as well as the other empirical studies presented below, will all be summarized in table 3.1.
The following subsections will have a closer look at the empirical research, which was pioneered by Grossmann and Krueger, as well as the theoretical research.
Empirical Research
Empirical research, so far, mainly focuses on single-equation specification to model an under-lying relationship. As the independent variable, GDP per capita is mainly used and serves as a measure of income. The dependent variable in the empirical literature constitutes generally to single indicators of a particular pollutant, for example CO, NOx, SO2, CO2 or others (Wang et al., 2015). According to Uchiyama (2016), the standard regression model, which has recently been used with panel data, takes the following form:
𝐸
#$= 𝛼
#+ 𝛾
$+ 𝛽
*𝑌
#$+ 𝛽
,𝑌
#$,+ 𝛽
-
𝑌
#$-+ 𝛽
.𝑥
#$+ 𝜀
#$ (3.1)Where Eit donates an index of pollution, 𝛼# indicates the cross-section dimension, while 𝛾$ is
the time effect. Per capita income is expressed in Yit5, x is a set of control variables and 𝜀 is the
error term. In order to observe a turning point where the level of pollution starts to fall within a certain range, it has to hold that b1 > 0, b2 < 0 and b3 = 0 with statistical significance (Uchiyama, 2016). Per capita income level of the turning point in the model of equation 3.1 is given by (-b1/(2b2)) (Uchiyama, 2016). The primary focus of empirical research lies in analys-ing which level the income has to reach in order to observe a turnanalys-ing point (Uchiyama, 2016). If b3 > 0 with statistical significance, the reduction of environmental degradation along with an increase of income per capita, has a temporary character, so that its pollution level can increase again after a certain level of income, and one can observe a EKC which is N-shaped (Uchiyama, 2016).
In addition to this kind of empirical research, single indicators of resource exhaustion such as wilderness, land or deforestation and indices of composite environmental degradation are used as dependent variable (see for example Jha and Murphy, 2003; Babu and Datta, 2003 or
Farhani et al., 2014). By doing so, not a standard model as in equation 3.1 is used. Use is rather made of log-linear models and semi-log-type models. This is due to the fact that the dependent variable is, by its nature, not allowed to fall short of zero (Uchiyama, 2016). Also, relative composite indices as a measure of environmental sustainability such as the ecological footprint, for instance, are used as dependent variable (see Al-Mulali et al., 2015).
Despite these different approaches, focuses and variables, there is hardly a consensus about a robust and common environmental Kuznets curve, as the theory suggests. There were several factors investigated that account for the shape of the EKC, as previous studies have shown (Dinda, 2004). Heil and Selden (2001) showed it, for instance, for trade intensity, Unruh and Moomaw (1998) for market mechanism, Shen and Hashimoto (2004) for the share of the sec-ondary industry and Hettige et al. (1999) for regulation. Also, scale and composition and abate-ment effects have been the focus of the research (see for example Grossman and Krueger, 1991 or Kaufmann et al., 1998).
Theoretical Research
The model of the environmental Kuznets curve has also been generated from some theoretical models (see for example Lopez, 1994; Barbier, 2001 or Bulte and van Soest, 2001). The theo-retical models of the EKC can be categorized into different categories. According to Uchiyama (2016), the most common categorization is to divide the theoretical models into static models on the one hand and dynamic models on the other hand.
In order to derive the model from a theoretical approach, there must be various appropriate assumptions (Wang et al., 2015). To further explore the theoretical basis of the EKC, some economic growth models have recently been generalized. For instance, an endogenous growth model with natural capital as a productive asset (see Dinda, 2005) or the Green-Solow model in order to investigate the natural resource constraint (see Brock and Taylor, 2010; Stefanski, 2010 or Reyes, 2011).
4. Methodology
In order to examine the causality from real GDP per capita on the environmental wellbeing dimension of the SSI (see chapter 5), the following model will be estimated:
𝐸𝑁𝑊
#$= 𝑎
#+ 𝛾
$+ 𝛽
*𝑙𝑛𝑌
#$+ 𝛽
,(𝑙𝑛𝑌
#$)
,+ 𝛽
-
(𝑙𝑛𝑌
#$)
-+ 𝛽
.𝑋
#$+ 𝜀
#$ (4.1)Where 𝐸𝑁𝑊#$ is the environmental wellbeing dimension of the SSI, 𝑎# indicates the cross-section dimension, 𝛾$ is the time index, 𝑙𝑛𝑌#$ is the natural logarithm of real GDP per capita, (𝑙𝑛𝑌#$), is the natural logarithm of real GDP per capita squared, (𝑙𝑛𝑌
#$)- is the natural
loga-rithm of real GDP per capita cubed, 𝑋#$ is a set of control variables6 and 𝜀#$ is the error term. Real GDP per capita is transformed in the logarithmic form, since the logarithm can smooth out outliers. Moreover, does it simplify the interpretation of the estimated coefficients from an economic point of view and make it more intuitive.
The corresponding research hypothesis to the model 4.1 is derived from the underlying, theoretical hypothesis of the EKC. However, since the present thesis aims to model a relation-ship between income and an index that measures the environmental dimension of sustainability, hence a higher value can be interpreted as positive for the environment, while a lower value means the opposite, the predicted signs take in the opposite form than in the “regular” EKC hypothesis. Thus, the sign of 𝛽* is predicted to be negative, while the sign of 𝛽, is predicted to
be positive, in order to get a U-shaped relationship. The sign of 𝛽- is predicted to be zero, since otherwise a differently shaped relationship would appear.
Verbeek (2017) mentions that the choice between a random effects and a fixed effects ap-proach is not easy, especially – as in the present case – T is small, since “the estimates for b appear to be substantial [in such cases]”. According to Verbeek (2017), the appropriate inter-pretation should be that the fixed effects approach, in contrast to the random effects approach, is conditional upon the values for ai.The intuition behind this mainly holds if the individuals
in the sample are “one of a kind” as Verbeek puts it, and, therefore, cannot be seen as a random draw from an underlying population. This assumption is most appropriate when the subscript i denotes countries (Verbeek, 2017), which here is the case. Nevertheless, the present thesis does not depend the decision of the choice between fixed and random effects on this argumentation. In order to choose the right model, the Hausman test will be applied, which tests for the cross-
section. In this test, the coefficients of the fixed effects model and of the random effects model are compared. The null hypothesis, that both estimators are the same, is unlikely to hold, if there is a significant difference between the two estimators (Verbeek, 2017).
Since, there might be the chance of heteroscedasticity and autocorrelation, cross-section robust standard errors will be used in the regression. Moreover, the case of endogeneity may occur. In section 6, several arguments will be mentioned why in the present thesis it is assumed that the explanatory variables are exogenous, that is, Ε 𝜀#$|𝑥#$ = 0, where 𝜀#$ is the error term and 𝑥#$ is the set of explanatory variables. Hence, the expected value of 𝜀#$ given all the explan-atory variables is zero.
The general model will then be divided according to the income classes defined by the World Bank7. Those are low-income (27 countries), lower middle-income (36 countries), upper middle-income (39 countries) and high-income (47 countries)8. By doing so, the same set of control variables will be used as in the general model. Again, the Hausman test will be applied in every sub-model to choose between fixed effects and random effects.
The specific results will then be compared, on the one hand, to the other results of this thesis and, on the other hand, to outcomes of former EKC-studies.
5. Data
The used data is described below in more detail, while descriptive statistics of it can be found in the appendix (appendix C). The data period is from 2006 to 2016. This scope has been se-lected due to the availability of the dependent variable. The SSI was the first time published in 2006, while the data of 2016 constitute the most recent data available. The data of the SSI is published every two years. The independent variables, including all control variables, are avail-able from the World Bank. The dataset consists of 1499 countries.
First, the dependent variable is presented. Second, the (main) explanatory variables are introduced before, third, the set of control variables is presented. Those have been chosen in order to find the exact effect that income has on the dependent variable, the environmental wellbeing index of the SSI. All variables were selected through analysis of empirical and theo-retical literature.
7 See footnote 2 for a definition of the income classes according to the World Bank. 8 See appendix B for a sorting of the countries according to their income class.
The environmental wellbeing dimension of the Sustainable Society Index10 serves as dependent variable. In order to understand how the SSI is structured, figure 5.1 visualises its composition.
Figure 5.1: Structure of the Sustainable Society Index.
Source: van de Kerk et al. (2014)
The SSI is composed by the three dimensions of sustainable development. Those are then fur-ther diversified into seven categories which then again are diversified into, altogefur-ther, 21 indi-cators.
The dimension environmental wellbeing consists of two categories, Natural Resources and Climate & Energy respectively. Natural Recourses is measured by looking at the indicators Biodiversity, Renewable Water Resources and Consumption. Climate and Energy is composed of the indicators Energy Use, which is measured as oil consumption per capita, Energy Savings, Greenhouse Gas emissions and the share of Renewable Energy (van de Kerk et al., 2014).
The Sustainable Society Index is expressed in absolute values, which means that every indicator, category and dimension can be expressed with a value between one and ten, whereby one indicates the worst possible value a country can achieve in relation to the stage of sustain-ability of the corresponding indicator and ten is referred to as the best one, i.e. the optimum sustainable situation. The index is calculated by using the geometric mean. The corresponding formula is given by:
𝑥
<=>?=
B @#A*𝑥
#= 𝑥
B *∙ 𝑥
,∙ ∙ ∙ 𝑥
@(5.1)
Whereby x represents the values of the individual categories and n the number of categories of the Sustainable Society Index respectively. By choosing this calculation method, it is not pos-sible that bad values of one category are fully compensated by good values of another cate-gory11. As of today, the Sustainable Society Index represents 99 percent of the total world pop-ulation (van de Kerk et al., 2014).
Real GDP per capita serves as main independent variable12. GDP is an indicator of the comprising output of services and goods produces within a certain economy during a certain time period (mostly yearly). GDP per capita then, is the value of GDP divided by the (midyear) population of the target economy. The present thesis adopted real GDP per capita calculated in consistent 2011 US Dollar and based on purchasing power parity (PPP) to measure real income. According to the research hypothesis, this variable is expected to be negatively related with the environmental wellbeing index.
The variable real GDP per capita squared, on the other hand, is expected to be positively related with the environmental wellbeing index of the SSI in order to obtain the U-shaped rela-tionship. This assumption is based on former EKC studies and the EKC-hypothesis itself. More-over, the quadratic form of the independent variable is widely utilized in previous studies.
In order to obtain a U-shaped relationship between GDP per capita and the environmental wellbeing index of the SSI, the coefficient of the variable real GDP per capita cubed is expected to be zero. If it is not zero but negative, for instance, one would obtain an inverted N-shaped relationship. A lot of EKC studies, however, do not include a cubed GDP term. In order to control for other shapes of relationships, this present thesis includes it.
The variable “Agricultural Land” refers to the percentage of total land of a certain country, which is arable, under permanent pastures and under permanent crops (World Bank, 2018). In the present research, this variable serves as control variable.
“Industrialisation”, the second control variable, expresses the share of the second sector of total GDP in percent. It includes value added13 in manufacturing, construction, mining, water, gas and electricity (World Bank, 2019c).
In order to control for the influence of trade, the “Net Barter Terms of Trade Index” (NBTTI), is included in the regression. The percentage ratio of the value of export unit to the value of import unit, measured relative to the base year 2000 (2000 = 100), is the corresponding calculation (World Bank, 2019d).
Many former EKC-studies include demographic control variables. The present thesis in-cludes “Population Density”, “Population Growth” and “Urbanization” as such. By dividing midyear population14 by the corresponding land area in square kilometres, one obtains the var-iable “Population Density” (World Bank, 2013a).
“Population Growth” is expressed in percentage. It is calculated as the exponential rate of growth of midyear population from year t-1 to t (World Bank, 2013b). Population is defined as in the previous variable.
The control variable “Urbanization” is a percentage of the population of a certain country living in urban areas as defined by national statistical offices. The United Nations Population Division collects and smoothens the data (World Bank, 2019e).
6. Empirical Model
The general model which is estimated looks as follows:
𝐸𝑁𝑊#$ = 𝛼#+ 𝛾$+ 𝛽*𝑙𝑛𝑌#$+ 𝛽,(𝑙𝑛𝑌#$),+ 𝛽
-(𝑙𝑛𝑌#$)-+ 𝐴𝐺𝑅#$+ 𝐼𝑁𝐷#$+ 𝑁𝐵𝑇𝑇𝐼#$+
𝑃𝐷𝐸#$ + 𝑃𝑂𝑃𝐺#$ + 𝑈𝑅𝐵#$+ 𝜀#$ (6.1)
13 By subtracting intermediate inputs from net output of a certain sector, one obtains the value added (World Bank, 2019).
Where ENWit is the environmental wellbeing index, 𝛼# indicates the cross-section dimension,
𝛾$ is the time index, 𝑙𝑛𝑌#$ is the logarithm of real GDP per capita, (𝑙𝑛𝑌#$), is the logarithm of
real GDP squared, (𝑙𝑛𝑌#$)- is the logarithm of real GDP in cubic form, 𝐴𝐺𝑅
#$ is the agricultural
land, 𝐼𝑁𝐷#$ is industrialisation, 𝑁𝐵𝑇𝑇𝐼#$ is the net barter terms of trade index, 𝑃𝐷𝐸#$ is popu-lation density, 𝑃𝑂𝑃𝐺#$ is population growth, 𝑈𝑅𝐵#$ is urbanisation and 𝜀#$ is the error term. The general model was estimated with fixed effects and random effects. The corresponding Hausman test was applied. The outcome, the Prop>chi2, equals 0.0001. The null hypothesis could therefore be rejected, which means that a fixed effects approach is used for the regression. The general model was estimated using cross-section clustered robust standard errors.
This general model (6.1) is then divided in the four already mentioned income models. Hence, they have exactly the same structure and make us of the same control variables. They only differ in the underlying data. Thus, there is the low-income model, which aggregates all the low-income countries (LI), the lower middle-income model with all the lower middle-in-come countries (LMI), the upper middle-inmiddle-in-come model with all the upper middle-inmiddle-in-come coun-tries (UMI) and the high-income model with all the high-income councoun-tries (HI)15.
It is assumed that all explanatory variables are uncorrelated with the residuals, that is, Ε 𝜀#$|𝑥#$ = 0, where 𝜀#$ is the error term and 𝑥#$ is the set of explanatory variables. There are several arguments why this assumption of exogeneity is plausible. On the one hand, the envi-ronmental wellbeing index is a very slow-moving variable, while GDP, the main independent variable, is a comparable fast-moving variable. On the other hand, none of the indicators that are used to calculate the environmental wellbeing index, nor the index itself, are part of the calculation of growth in any of the known economic growth mode. Hence, the possibility that a certain score of the environmental wellbeing has a causal effect of GDP is rather inexistent.
Results
In table 6.1 the regression results of the general model as well as of the four different income models are summarized. For every model, the coefficients of all the explanatory variables as well as the corresponding significance level is indicated and it is stated if the model was esti-mated with fixed or random effects. Moreover, the outcome of the F-test and the corresponding p-value is listed to indicate the overall fit of the respective model.
General Model
All three coefficients of the real GDP per capita, 𝑙𝑛𝑌, (𝑙𝑛𝑌), and (𝑙𝑛𝑌)-, are statistically
sig-nificant at the 1% level. With the exception of PDE, population density (sigsig-nificant at the 10% level), none of the control variables are statistically significant. In chapter 4, the research hy-pothesis was established. It predicts a negative sign for the coefficient of 𝑙𝑛𝑌#$, a positive for
(𝑙𝑛𝑌), and zero for (𝑙𝑛𝑌)- in order to get the U-shaped relationship between income and
en-vironmental wellbeing. But in table 6.1 we see a completely different outcome: the coefficient of 𝑙𝑛𝑌 is positive, while the one of (𝑙𝑛𝑌), is negative. The coefficient of (𝑙𝑛𝑌)- is 0.198 and
quite close to zero. This means, that the environmental wellbeing would improve in the begin-ning of economic development and would then start to decline after the corresponding turbegin-ning point. This outcome is not in accordance with the theory and is therefore rather unexpected and surprising. The p-value of the corresponding F-test is 0.0000. The corresponding null hypothe-sis, that all coefficients except the constant are equal to zero, can therefore be rejected.
Low-Income Model
Also for the low-income model, the fixed effects estimation is the right approach. The Prob>chi2 of the corresponding Hausman test is 0.0000. The model was estimated using cross-section clustered robust standard errors. The three coefficients of the different forms of loga-rithmic real GDP per capita are all statistically significant at the 5%-level. Moreover, the control variables industrialisation and urbanisation are significant at the 5%-level, while population density is again significant at the 10%-level. In contrast to the general model, the coefficients have now almost the predicted signs: The sign of 𝛽* is negative and the one of 𝛽, is positive, as the research hypothesis predicted. 𝛽- is no longer close to zero, however. Thus, this model says that, for low-income countries, the environmental wellbeing worsens in the beginning of economic development but will improve after a certain threshold is reached. Since 𝛽- is not close to zero and, moreover, statistically significant, it is not possible to find a single valid turning point. The relationship would, over time, become inversely N-shaped.
Lower Middle-Income Model
or at the 10%-level. A calculation and interpretation of the marginal effect of income on the environmental wellbeing in this model is therefore rather meaningless. Moreover, none of the control variables have a statistical significance. Hence, this model is not capable at all to explain the changes in the environmental wellbeing of the observed countries. Also, when the test for 𝛽* and 𝛽, is considered to be a one-tailed test rather than a two-tailed test, the coefficients are not statistically significant16. The estimated coefficients would have the expected signs, how-ever. Hence, the sign of 𝛽* is negative, the one of 𝛽, is positive and 𝛽- is -0.261 and close to zero.
Upper Middle-Income Model
Prob>Chi2 = 0.0024. Hence, the fixed effects estimation leads to efficient outcomes. But, as in the model before, there is no casual effect of real GDP per capita on the environmental wellbe-ing in this model. Only the control variables population density (5%-level) and industrialisation (10%-level) obtain a statistical significance. The corresponding coefficients are rather small, however. Hence, also this sub model is not capable of explaining changes in the environmental wellbeing. Nevertheless, as in the sub-model above, the signs of the coefficients follow the prediction.
High-Income Model
According to the Hausman test, the high-income model can be estimated using random effects (Prob>Chi2 = 0.1462). Again, the model is estimated using cross-section clustered standard errors. While 𝑙𝑛𝑌 is statistical significant at the 10%-level, (𝑙𝑛𝑌), and (𝑙𝑛𝑌)- are not
signifi-cant. (𝑙𝑛𝑌), is significant at the 10% level if a one-tailed test is applied. The sign of the
coef-ficient of 𝑙𝑛𝑌 is negative and the sign of the coefcoef-ficient of (𝑙𝑛𝑌), is positive. Hence, this model
predicts, that the environmental wellbeing first decreases while income increases and then starts to increase after a certain threshold of income is reached, which stands in line with the EKC-hypothesis.
Discussion
The previous chapter presented the results in a technical manner. This subchapter is now aiming to provide a less technical, more intuitive interpretation of the results in order to discuss them. To fully understand the results, scatter plots for the five different models were made. They
resemble the scatter plots shown in chapter 2. Moreover, they allow for a visual interpretation. From the results above we already know that only for the general model and for the low-income model all three coefficients of real GDP per capita are statistical significant, while for the high-income model 𝑙𝑛𝑌 is statistical significant at the 10% level in a two-tailed test and (𝑙𝑛𝑌), is
statistically significant at the 10% level in a one-tailed test. The corresponding signs, however, are not the same in those three models.
The scatter plots are all build up the same way: On the y-axis is the mean of the environ-mental wellbeing score for each of the observed countries for the period from 2006 to 2016, while on the x-axis the mean of the corresponding real GDP per capita of the country is plotted for the period from 2006 to 2016.
The graph on the left-hand side plots the general model. On can clearly observe that, as the theory suggest it, the ENW falls while income in-creases. This graphs also shows, that this devel-opment is decreasing. However, there is not yet reached a turning point. It seems, that as richer a country gets, the worse its influence on the en-vironment is.
This graph captures the low-income countries17. The image is comparable to the one above: as a country becomes richer, the environment has to suffer more. The results of the previous chapter for low income countries enhance this view. They indicate, however, that there will be a turn-ing point at some level of income.
17 The “outlier”, the point in the bottom-right corner, belongs to Yemen. According to the World Bank, Yemen is considered as a low-income country. Its real GDP per capita, however, differs a lot of the one of other low-income countries. In the graph it has therefore the power to influence the curvature of the line of the predicted ENW.
2 4 6 8 En vi ro n me n ta l W e llb e in g 0 50000 100000 150000 GDP Per Capita
ENW predicted ENW
3 4 5 6 7 8 En vi ro n me n ta l W e llb e in g 1000 2000 3000 4000 GDP Per Capita
The negative trend, i.e. richer countries have a more negative influence on the environmental wellbeing, can also be observed in lower middle-income countries. The trend, however, is not as strong and clear to observe as in the cases shown above. The dots are more widely spread com-pared to the examples above.
The trend starts to get a bit clearer again when only focusing on countries of the upper middle-income spectre. It remains, however, quite vague.
The high-income countries, make it possible to draw a clearer line again. The trend remains the same, however. There are a few outliers. But in general, it holds that the richer a country is the more negative influence it has on the environ-mental wellbeing.
According to the data and the corresponding visualisation above, one can clearly see, that the environment has to suffer more as countries get richer. So far, there is no turning point observ-able, that is, the richest countries of the sample score the worst in the environmental wellbeing. A quite striking result. This negative trend is observable, by focusing on the scatter plots, in the general model itself as well as in the four income models.
Those results would differ a lot, of course, if another dimension of the SSI (and of sustain-able development in general) would be the dependent varisustain-able of a similar research. If the hu-man wellbeing dimension of the SSI would be the left-hand side variable, for example, one
3 4 5 6 7 8 En vi ro n me n ta l W e llb e in g 2000 4000 6000 8000 10000 GDP Per Capita
ENW predicted ENW
1 2 3 4 5 6 En vi ro n me n ta l W e llb e in g 20000 40000 60000 80000 100000 120000 GDP Per Capita
ENW predicted ENW
2 3 4 5 6 7 En vi ro n me n ta l W e llb e in g 5000 10000 15000 20000 25000 30000 GDP Per Capita
would assume that economic growth leads to a higher outcome since this index captures – among others – things like ‘sufficient food’, ‘sufficient to drink’ or ‘education’. Obviously, the population of a certain economy can easier afford nutrition and the government (or the private sector) is more likely to be capable of building up a proper educational system when the inhab-itants of the country are richer. That the economic wellbeing dimension would profit from a higher income is out of doubt.
Hence, one could argue that a country should be able to become overall more sustainable when it becomes richer. However, it now raises the question, how one should understand and apply the concept of sustainable development. The author is an advocate of the perspective of the strong sustainability, which postulates that the different dimensions of sustainability are not substitutable. There is no point of being able, for instance, to have a proper educational system, income equality (both are indicators of the human wellbeing dimension of the SSI) or other positive outcomes of measures of sustainability, if the nature itself is destroyed in a way that it is no longer possible to cultivate sufficient amounts of food for the present or the future gener-ations, for instance. The environment has to be considered as the foundation of sustainable development. Economic activity has to take place within the boundaries of the feasible. If the existence of the EKC could be proved, however, i.e. economic development would favour the environment in the long run, this view could be adjusted accordingly, so that one would have to support economic progress in favour of the environment. Since, this evidence could not be produced unambiguous yet, it is considered to be very dangerous to rely on the theoretical con-cept itself.
Limitations
This section focuses on limitations of the regression presented above. First, general limitations (or disadvantages) of panel data regressions will be mentioned, before focus is paid to limita-tions concerning the applied estimation.
in-complicate the analysis. Furthermore, missing observations are often an issue in panel data sets. In such cases, the standard analysis has to be adjusted, even if these observations are missing in a random way. In the present thesis, the dataset was adjusted according to this issue: coun-tries, who had no proper data on real GDP per capita were excluded of the set. Moreover, while choosing the control variables, one of the criterions was to have as few missing values as pos-sible.
In his 2004 published article “The Rise and Fall of the Environmental Kuznets Curve”, Stern mentions that a lot of former econometric research about the EKC ignored possible sta-tionarity issues of the underlying data which might lead to spurious regressions. Non-station-arity issues arise mainly in time series analysis, i.e. data with small N and large T. The data used for the present thesis, however, has a large N and a comparably small T. In such models and in models with breaks, stationarity tests only have low power. This, in turn, is the reason, why the data was not tested for non-stationarity here. The same argumentation applies when it comes to cointegration issues. Nevertheless, the author is aware of this issue. If, for example, as time goes on, the T of the dataset gets bigger, non-stationarity issues as well as cointegration issues might become necessary to worry about and have then to be addressed accordingly.
This circumstance, however, exposes a rather obvious limitation of the applied regression: the comparably small T, i.e. the small data period. The SSI was first published in 2006 and is published every second year, while the most recent data is from 2016. When time goes on, comparable studies can be applied, and the income effect of the environmental wellbeing index of the SSI can be investigated more accurate. At the moment, however, this circumstance poses a limitation.
Due to missing values of the real GDP per capita for five countries, not the whole set of countries could be used for which the SSI is available. Since this issue only involves a clear minority, it does not pose a major limitation of the research. Nevertheless, does it have to be mentioned. Especially, when one wants to compare the results of the present research with comparable studies.
outsourcing effects are captured (see chapter 218). But sustainable development, sustainability in general or the environmental wellbeing itself all remain hard to capture and to measure (see chapter 2). Thus, also the environmental wellbeing index of the SSI is just a chosen proxy variable for environmental health in general. The results of the present research have therefore to be interpreted with caution. It is possible to derive conclusions about the income effect of the environmental wellbeing indicator but one has to be cautious to expand this view to the environmental health in general.
Furthermore, the dependent variable is expressed in absolute values and goes from one to ten (see chapter 5). If one is interested in predictions for future development of it, the model applied in this thesis might predict values that are larger (smaller) than ten (one).
Besides that, it has to be mentioned that the set of control variables might not be the most appropriate. In order to select the control variables, the author studied the corresponding former theoretical as well as empirical research intensively. The decision was made with utmost cau-tion. It remains, however, the eventuality, that the selection was inappropriate, which cannot be tested.
Ultimately, the data was not tested for endogeneity. The assumption was made, that all explanatory variables are uncorrelated with the residuals. The presented arguments of this as-sumption seem plausible. Nevertheless, this asas-sumption is able not to hold. This should be kept in mind.
Besides all the mentioned limitations, the applied regression delivers interesting and im-portant insights of the relationship between income development and environmental wellbeing, measured by the SSI.
7. Conclusion
In the introduction, Vladimir Ivanovich Vernadsky is mentioned, who was the first whoever investigated the causality of life on the environment. This is almost one hundred years ago. Today, this can be considered as common knowledge. The reason why this relationship is that obvious for us today, is because the effects which (human) life has on the environment became visible and noticeable in the recent past.
During the last century, voices of criticism of the focus of policy makers on economic growth expanded. Scientists, like the ones from the Club of Rome, warned the world, that we
might reach a point of no return, when the destructibility of nature is not regarded. Still today, one could not yet agree upon the effective relationship between economic progress and envi-ronmental wellbeing. On the one hand, there are people who see the scarcity of natural resources and fear that the ever-growing economy (which leads to a richer society, who will be able to consume more and more) will exhaust those resources irretrievable. On the other hand, there is the assumption, that this development only holds in the beginning of economic development, but that a turning point can be reached, whereupon more income leads to an improvement of the environmental wellbeing (due to a shift of preferences of economic agents or technological progress, for instance). This is the key message of the environmental Kuznets hypothesis. It is desirable, that this hypothesis holds in practice. If economic growth would really lead to less environmental degradation in the long-run and if the damages of the early stages of economic development on the environment are reversible, it would only be favourable to set the course to economic development and economic growth. Hence, one could “kill two birds with one stone”: Policy makers could focus on economic progress, poverty reduction, and so on and simultaneously have a positive influence on the environmental wellbeing. If this theory does not hold in reality, however, the people around the world would be better off to rate the envi-ronment as equally important as economic development, like the modern understanding of sus-tainable development suggests it, and act accordingly.
This thesis aimed to investigate this question. Hence, what is the relationship between in-come, measured as real GDP per capita, and the environmental wellbeing. The regression out-comes are rather controversial. While the general model estimates, that economic growth, meas-ured as real GDP per capita, has a positive effect on environmental wellbeing in the beginning of economic development and starts to have a negative influence on it after a certain threshold is reached, i.e. there is an inversely U-shaped relationship, the estimates of the low-income model predicts a U-shaped relationship. The lower middle-income model as well as the upper middle-income model are both not able to form a relationship with statistical significance (the signs of the coefficients predict, however, a U-shaped relationship). The high-income model predicts a U-shaped relationship as well.
To conclude, this thesis could not prove the existence of the EKC in practice undoubtedly. This result is consistent with the overall outcome of former EKC-studies. They found all sorts of shapes of the relationship between income progress and certain measures of environmental degradation (environmental quality), such as a U-shaped relationship, an inverted U-shaped one, a N-shaped relationship or monotonic relationships.
References
Al-Mulali U., Weng-Wai C., Sheau-Ting L., Mohammed A. H. (2015): “Investigating the en-vironmental Kuznets curve (EKC) hypothesis by utilizing the ecological footprint as an indicator of environmental degradation”, Ecological Indicators, 48, pp. 315-323
Babu S. S., Datta S. K. (2013): “The relevance of environmental Kuznets curve (EKC) in a framework of broad-based environmental degradation and modified measure of growth – a pooled data analysis”, International Journal of Sustainable Development & World Ecol-ogy, 20 (4), pp. 309-316
Barbier E. B. (2001): “The economics of tropical deforestation and land use: an introduction to the special issue”, Land Economics ,77 (2), pp. 155-171
Brock W. A., Taylor M. S. (2010): “The Green Solow model”, Journal of Economic Growth, 15 (2), pp. 127-153
Brundtland Commission (1987): “Our Common Future – The Brundtland Report of the World-commission on Environment and Development”, Oxford University Press, Volker Hauff Brunetti A. (2017): “Volkswirtschaftslehre - Eine Einführung für die Schweiz“,
Bern, Hep Verlag AG
Bulte E. H., van Soest D. P. (2001): “Environmental degradation in developing countries: households and the (reverse) environmental Kuznets curve”, Journal of Development Economics, 65 (1), pp. 225-235
Dinda S. (2004): “Environmental Kuznets Curve Hypothesis: A Survey”, Ecological Econom-ics, 49 (4), pp. 431-455
Dinda S. (2005): “A theoretical basis for the environmental Kuznets curve”, Ecological Eco-nomics, 53 (3), pp. 403-413
Farhani S., Mrizak S., Chaibi A., Rault C. (2014): “The environmental Kuznets curve and sus-tainability: a panel data analysis”, Energy Policy, 71, pp. 189-198
Grossmann G. M., Krueger A. B. (1991): „Environmental impacts of a North American Free Trade Agreement”, National Bureau of Economic Research, Working Paper, 3914, NBER, Cambridge, Massachusetts
Heil M. T., Selden T. M. (2001): „International Trade Intensity and Carbon Emissions: A Cross-Country Econometric Analysis”, Journal of Environment and Development, 10 (1), pp. 35-49
Hilser K. (2014): “Wirtschaftswachstum und nachhaltige Entwicklung - ein Widerspruch? Fall-beispiele Länder des südlichen Afrikas“, Ruprecht-Karls-Universität, Heidelberg Jha R., Murthy K. V. B. (2003): “An inverse global environmental Kuznets curve”, Journal of
Comparative Economics, 31 (2), pp. 352-368
Kaufmann R. K., Davidsdottir B., Garnham S., Pauly P. (1998): „The determinants of atmos-pheric SO2 concentrations: reconsidering the environmental Kuznets curve”, Ecological Economics, 25 (2), pp. 209-220
Kropp, A. (2019): “Grundlagen der Nachhaltigen Entwicklung: Handlungsmöglichkeiten und Strategien zur Umsetzung“, Springer Gabler Verlag
Kuznets S. (1955): “Economic Growth and Income Inequality”, The American Economic Re-view, 45, pp. 1-28
Li H., Grijalva T., Berrens R. P. (2007): „Economic growth and environmental quality: a meta-analysis of environmental Kuznets curve studies”, Econ. Bull. 17 (5), pp. 1-11
Liddle B. (2001): “Free trade and the environment–development system”, Ecological Econom-ics, 39, pp. 21-36
Lopez R. (1994): “The environment as a factor of production: the effects of economic growth and trade liberation”, Journal of Environmental Economics and Management, 27 (2), pp. 163-184
Meadows D. (1972): “The Limits of Growth”, Universe Books
Meyer W. (2004): “Indikatorenentwicklung: Eine praxisorientierte Einführung“, Saarbrücken, Centrum für Evaluation
Mostafa M. M. (2010): “A Bayesian approach to analyzing the ecological footprint of 140 na-tions”, Ecological Indicators, 10 (4), pp. 808-817
Reyes R. C. (2011): “The Green Solow model with natural resources constraint: a theoretical note”, DLSU Business and Economics Review, 21 (1), pp. 111-116
Robalino-López A., García-Ramo J. E., Golpe A. A., Mena-Nieto M. (2014): “System dynam-ics modelling and the environmental Kuznets curve in Ecuador (1980–2025)”, Energy Policy, 67, pp. 923-931
Shen J., Hashimoto Y. (2004): “Environmental Kuznets Curve on Country Level: Evidence from China”, Discussion Papers in Economics and Business, Osaka University
Stern D. I. (2002): “Explaining changes in global sulfur emissions: an econometric decompo-sition approach”, Ecological Economics, 42 (1), pp. 201-220
Stern D. I. (2004): “The Rise and Fall of the Environmental Kuznets Curve”, World Develop-ment, 32 (8), pp. 1419-1439
Stern D. I. (2010): “Between estimates of the emissions-income elasticity”, Ecological Eco-nomics, 69 (11), pp. 2173-2182
Stern D. I., Common M. S. (2001): “Is there an environmental Kuznets curve for sulfur?”, Journal for Environmental Economics and Management, 41 (2), pp. 162-178
Uchiyama K. (2016): “Environmental Kuznets Curve Hypothesis and Carbon Dioxide Emis-sions”, Springer Briefs in Economics, Development Bank of Japan Research Series, Springer
Unruh G. C., Moomaw W. R. (1998): “An alternative analysis of apparent EKC-type transi-tions”, Ecological Economics, 25 (2), pp. 221-229
Van de Kerk G., Manuel A., Kleinjans R. (2014): „Sustainable Society Index SSI-2014”, Sus-tainable Society Foundation, The Hague, Netherlands
Verbeek M. (2017): “A Guide to Modern Econometrics”, 5th Edition, Wiley Costum
Wachter D. (2014): “Nachhaltige Entwicklung - Das Konzept und seine Umsetzung in der Schweiz“, Somedia Buchverlag, Glarus/Chur
Wang Y., Kang L. Y., Wu X., Xiao Y. (2013): „Estimating the environmental Kuznets curve for ecological footprint at the global level: a spatial econometric approach”, Ecological Indicators, 34, pp. 15-21
Wang L., Zhou D., Wang Y., Zha D. (2015): „An empirical study of the environmental Kuznets curve for environmental quality in Gansu province”, Ecological Indicators, 56, pp. 96-105
World Bank (2006): “Where is the wealth of nations? Measuring Capital for the 21st Century” Washington D.C., The World Bank
World Bank (2013a): “Population Density (People Per Sq. Km Of Land Area)”, https://data-catalog.worldbank.org/population-density-people-sq-km-land-area, retrieved: 05/26/19 World Bank (2013b): “Population Growth (Annual %)”,
https://datacata-log.worldbank.org/population-growth-annual, retrieved: 05/26/19
World Bank (2019a): “How does the World Bank classify countries?”, https://data- helpdesk.worldbank.org/knowledgebase/articles/378834-how-does-the-world-bank-classify-countries, retrieved: 05/26/19
World Bank (2019b): “World Bank Country and Lending Groups”, https://data- helpdesk.worldbank.org/knowledgebase/articles/906519-world-bank-country-and-lend-ing-groups, retrieved: 05/26/19
World Bank (2019c): “Industry (Including Construction), Value Added (% of GDP)”, https://datacatalog.worldbank.org/industry-including-construction-value-added-gdp, re-trieved: 05/26/19
World Bank (2019d): “Net Barter Terms Of Trade Index (2000 = 100)”, https://datacata-log.worldbank.org/net-barter-terms-trade-index-2000-100-0, retrieved: 05/26/19
Appendix
Appendix A. List of countries of the Sustainable Society Index
Albania Congo Democratic Rep. Iceland
Algeria Costa Rica India
Angola Cote d'Ivoire Indonesia
Argentina Croatia Iran
Armenia (Cuba) Iraq
Australia Cyprus Ireland
Austria Czech Republic Israel
Azerbaijan Denmark Italy
Bangladesh Dominican Republic Jamaica
Belarus Ecuador Japan
Belgium Egypt Jordan
Benin El Salvador Kazakhstan
Bhutan Estonia Kenya
Bolivia Ethiopia (Korea, North)
Bosnia-Herzegovina Finland Korea, South
Botswana France Kuwait
Brazil Gabon Kyrgyz Republic
Bulgaria Gambia Laos
Burkina Faso Georgia Latvia
Burundi Germany Lebanon
Cambodia Ghana Lesotho
Cameroon Greece Liberia
Canada Guatemala Libya
Central African Republic Guinea Lithuania
Chad Guinea-Bissau Luxembourg
Chile Guyana Macedonia
China Haiti Madagascar
Colombia Honduras Malawi
Mali Rwanda (Venezuela)
Malta Saudi Arabia Vietnam
Mauritania Senegal Yemen
Mauritius Serbia Zambia
Mexico Sierra Leone Zimbabwe
Moldova Singapore
Mongolia Slovak Republic
Montenegro Slovenia
Morocco South Africa
Mozambique Spain
Myanmar Sri Lanka
Namibia Sudan
Nepal Sweden
Netherlands Switzerland
New Zealand (Syria)
Nicaragua (Taiwan)
Niger Tajikistan
Nigeria Tanzania
Norway Thailand
Oman Togo
Pakistan Trinidad and Tobago
Panama Tunisia
Papua New Guinea Turkey
Paraguay Turkmenistan
Peru Uganda
Philippines Ukraine
Poland United Arab Emirates
Portugal United Kingdom
Qatar United States
Romania Uruguay
Appendix B. Countries sorted according their Income Class Low-income countries:
Benin, Burkina Faso, Burundi, Central African Republic, Chad, Congo Dem. Rep., Ethiopia, Gambia, Guinea, Guinea-Bissau, Haiti, Liberia, Madagascar, Malawi, Mali, Mozambique, Ne-pal, Niger, Rwanda, Senegal, Sierra Leone, Tajikistan, Tanzania, Togo, Uganda, Yemen, Zim-babwe
Lower middle-income countries:
Angola, Bangladesh, Bhutan, Bolivia, Cambodia, Cameroon, Congo Rep., Côte d’Ivoire, Egypt, El Salvador, Georgia, Ghana, Honduras, India, Indonesia, Kenya, Kyrgyz Republic, Lao, Lesotho, Mauritania, Moldova, Mongolia, Morocco, Myanmar, Nicaragua, Nigeria, Paki-stan, Papua New Guinea, Philippines, Sri Lanka, Sudan, Tunisia, Ukraine, UzbekiPaki-stan, Vi-etnam, Zambia
Upper middle-income countries:
Albania, Algeria, Armenia, Azerbaijan, Belarus, Bosnia-Herzegovina, Botswana, Brazil, Bul-garia, China, Colombia, Costa Rica, Dominican Republic, Ecuador, Gabon, Guatemala, Guy-ana, Iran, Iraq, Jamaica, Jordan, Kazakhstan, Lebanon, Libya, Macedonia, Malaysia, Mauritius, Mexico, Montenegro, Namibia, Paraguay, Peru, Romania, Russia, Serbia, South Africa, Thai-land, Turkey, Turkmenistan
High-income countries:
Appendix C. Data Description General Model:
Variable Mean SD Min. Max Observations
Low-Income Model:
Variable Mean SD Min. Max Observations
Lower Middle-Income Model:
Variable
Mean
SD
Min.
Max Observations
Upper Middle-Income Model:
Variable Mean SD Min. Max Observations
High-Income Model:
Variable Mean SD Min. Max Observations
Description of the variables:
𝑬𝑵𝑾 = Environmental wellbeing index of the SSI
𝒍𝒏𝒀 = Real GDP per capita in log-form
(𝒍𝒏𝒀)𝟐 = Real GDP per capita in log-form squared
(𝒍𝒏𝒀)𝟑 = Real GDP per capita in log-form cubed
𝑨𝑮𝑹 = Agricultural land
𝑰𝑵𝑫 = Industrialisation (share of the second sector as percentage of total GDP)
𝑵𝑩𝑻𝑻𝑰 = Net barter terms of trade index 𝑷𝑫𝑬 = Population density
𝑷𝑶𝑷𝑮 = Population Growth 𝑼𝑹𝑩 = Urbanisation
T is time, n is the number of countries in the corresponding model and N is the total number of observations, i.e. n times T19.
Appendix D. Close look at individual indicators of the environmental wellbeing index Biodiversity:
The indicator “Biodiversity” itself consists of two sub indicators. On the one hand, the size of protected land areas expressed in percent of the total land area of a certain country, and on the other hand, the ten-years change of forest area (van de Kerk et al., 2014).
Renewable Water Resources:
The indicator “Renewable Water Resources” expresses the water consumption per year as a percentage of total available water resources of a certain economy, in order to monitor the suf-ficiency and the depletion of fresh water resources. This total then includes internal as well as external water resources (van de Kerk et al., 2014).
Consumption:
By subtracting the Carbon Footprint from the Ecological Footprint, one obtains the proxy used by the SSF for consumption. The Carbon Footprint is already monitored in the SSI by the indi-cator of Emissions of Greenhouse Gases (van de Kerk et al., 2014).
Energy Use:
The indicator „Energy Use“ is defined by the TPES (Total Primary Energy Supply). It is cal-culated as production plus imports minus exports plus/minus stock changes (van de Kerk et al., 2014).
Energy Savings:
The indicator “Energy Savings” monitors the results of energy use reduction activities and plans (van de Kerk et al., 2014).
Greenhouse Gases:
This indicator focuses on CO2 emissions. The amount of CO2 emitted is a common measure for the emissions of Greenhouse Gasses (GHG). This in turn means, that other emissions like N2O, HFCs, CH4, SF6 and PFCs are not included (van de Kerk et al., 2014).
Renewable Energy:
