How sustainable is the urban transition? A cross-country analysis of the effect
of urbanization on carbon dioxide emissions and life expectancy.
Carlijn Freutel 10437878 University of Amsterdam Faculty of Economics and Business
January 30, 2016
A thesis submitted for the degree of Bachelor of Science
in
Economics & Business with specialization Economics
This document is written by Carlijn Freutel who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
ABSTRACT
Recent research in the scientific field of sustainable development has given much attention to tradeoffs between the consumption of resources that harm the environment and the development of human well-being. This study focuses on the ratio of consumption-based carbon dioxide levels to average life expectancy – a relation known as the carbon intensity of well-being (CIWB). A potential pathway to sustainability is decreasing the CIWB. We use panel data of 65 nations from 1990 to 2008 to analyze how the effect of urbanization on the consumption-based CIWB changed through time for different regional samples. We find that in Africa and Latin America urbanization increases consumption-based CIWB, with increasing magnitude through time. In Asia urbanization decreases CIWB, although this effect became weaker throughout the time period and is likely to turn unsustainable soon. For the combined regions of Europe, North America, and Oceania the relation between urbanization and CIWB is weak. The results highlight the urgency for regionally-defined urban sustainability policies, especially in the developing world.
1. Introduction
Climate change is a hot topic, both in literal and figurative speech. Human interference with the climate system enhances climate change, which poses risks for human societies and natural systems (IPCC, 2014). Among the risks of climate change are reductions in ice and snow cover leading to increased sea levels, extreme weather events leading to disruptions of food production and damage to settlements, and projected increases in global temperature that can lead up to 6°C by the year 2100 (IPCC, 2014). While some scientists and politicians deny that global warming is partly man-made - for example Jeb Bush calling it ‘intellectual arrogance’ to fully conclude the man-man-made argument from the existing science (Killough, 2015) - most people find that enough scientific evidence is available to start altering our behaviour. The worldwide political response to climate change began at the Rio Earth Summit in 1992 where the UN Framework on Climate Change (UNFCCC) was adopted. The UNFCCC now has 195 members, and a legally binding and universal agreement on climate was adopted at the recent Climate Conference in Paris in December 2015 (COP21). This agreement includes the long-term goal of keeping global warming below a 1.5°C increase compared to pre-industrialized levels. Next to government officials, intergovernmental organizations, UN agencies, NGOs, and civil organizations attended the COP21, showing the broad concern of global warming across civil society worldwide (UNFCCC, 2015).
The term ‘sustainable development’ has been a popular goal in environmental negotiations since the first use of the term in the 1980 World Conservation Strategy. A definition widely used is the one stated in the report ‘Our Common Future’ (WCED, 1987): ‘(….) development that meets the needs of the present without compromising the ability of future generations to meet their own needs’. Sustainable development comprehends the present-day global concerns of environmental problems and socio-economic issues and links them together by striving for a healthy future for mankind (Hopwood et al., 2005). Hence sustainable development requires the integration of economic growth with other goals such as eradication of poverty and inequality, environmental protection, job creation, security, and justice. This paper will focus on the integration of environmental protection and poverty eradication.
Next to a decarbonization of industrialized societies, there is almost no room for energy growth to combat climate change. This presents concerns for sustainable development efforts, which will have to raise the living standards of the poor within these limits. Economic development is recognized as a pathway to improved human well-being, but economic development increases carbon emissions. In addition, poor populations and developing countries are the most vulnerable to climate change as they have the least adaptive capacity and often occupy areas that are most exposed to hazards. Herewith, climate change threatens to undermine the progress in poverty reduction that has been achieved today (Hoornweg, 2011).
These controversies about sustainable development efforts gave rise to the concept of decoupling human development from environmental pressure (Rao et al., 2014). This concept of decoupling implies that human well-being should increase, while stress placed on the environment from the usage of fossil fuels, the carbon intensity, should decrease at the same time. Therefore, effective strategies that aim for a sustainable future require reducing the carbon intensity of human well-being (CIWB) (Jorgenson, 2014). So far research has been done on the effects of economic development and inequality on CIWB, using CO2 emissions as a measure of carbon intensity and life expectancy as a measure of human well-being.
To understand what drives low CO2 emissions and high life expectancy, more socioeconomic conditions that affect CIWB need to be identified. Urbanization is a socioeconomic phenomenon that
often accompanies both economic growth and social modernization. This paper will focus on the effect of urbanization on CIWB for the following reasons: (i) the urban transition presents a challenge for global climate change mitigation efforts as it is accompanied by increasing emissions; (ii) there is a lack of research on the relation between urbanization and life expectancy; (iii) insights in the relation between urbanization and CIWB can promote better sustainable development policies. First, the urban population of the world has more than doubled since 1975, and is projected to double again by 2050 (Poumanyvong & Kaneko, 2010). In 2008, already more than 50% of the world’s population lived in urban areas. This urban transition involves transformations in the relationship between nature and society (Jorgenson et al., 2014). Urban activities account for approximately 36.8% to 48.6% of total greenhouse gas emissions (Marcotulli et al., 2013). Hence, the fast urban transition could increase carbon emissions rapidly and thus presents a challenge for global climate change mitigation efforts. Second, while the relation between urbanization and human well-being has been widely researched in the public health field, no quantitative research has been done on the effect of urbanization on life expectancy as a measure of well-being. This paper could contribute to existing research on life expectancy. Third, estimating the effect that urbanization has on CIWB could enhance effectiveness of global sustainable development efforts by creating more holistic policies that initiate and foster societal change (Jorgenson, 2015).
As Anna Kajumulo Tibaijuka, former Executive Director of UN-Habitat1 stated in 2007: ‘there is an inextricable link connecting urbanization, urban poverty, and climate change’ (Kajumulo Tibaikuka, 2007). According to her, urban issues should be at the forefront of the sustainable development agenda (United Nations, 2007). This underscores the importance of examining the effect of urbanization on CIWB.
While most research on carbon emissions uses production-based measures of CO2 emissions, in this study consumption-based measures will be used. This ensures a more accurate assessment of the carbon emissions implications of people’s livelihoods and well-being in urban areas, and takes into account the complexities of global trade and potential environmental load displacement (Jorgenson & Givens, 2015). This will be further explained in section 2.
In this paper we seek to enhance understanding of the environmental and social consequences of national-level urbanization throughout the world by examining its relation to the consumption-based CIWB. The main research question this paper tries to answer is: What is the effect of urbanization on the consumption-based carbon intensity of human well-being and how does this effect change through time? Three dependent variables will be examined. The first dependent variable will be the consumption-based CIWB, which will be created by using consumption-based emissions data and data on life expectancy for the years 1990 to 2008. Thereafter the separate effects of urbanization on the numerator and denominator of the CIWB will be examined in the remaining models. The second model will have consumption-based CO2 emissions per capita as the dependent variable, while for the third model this is life expectancy. A statistical modeling technique will be employed to assess how the effect of urbanization on the three dependent variables changed throughout the time period from 1990 to 2008. Multiple samples of nations will be examined, including an overall sample of 65 nations as well as samples restricted to geographic regions. Control variables for GDP per capita and inequality will be included
1 UN-HABITAT is the human settlements programme of the United Nations. UN-Habitat is working towards a
better urban future. Its mission is to promote socially and environmentally sustainable human settlements development and the achievement of adequate shelter for all (UN-Habitat, 2015).
in the model, since prior research showed their significant effect on CIWB (Jorgenson, 2015; Jorgenson & Givens, 2015). The regressions results lead to the following conclusions: (1) In African and Latin American nations urbanization is unsustainable as it increases CIWB, and this effect became larger throughout the time period; (2) In Asian countries urbanization decreases CIWB, and hence is sustainable. However, this effect is becoming smaller over time, making urbanization less sustainable; and (3) In Europe, North America, and Oceania urbanization and CIWB have a positive, but weak, relation.
The paper is organized as follows: the second section will review the literature on CO2 emissions, human well-being, CIWB and urbanization. The third section presents the hypotheses. The fourth sections describes the dataset and the model estimation technique is presented in the fifth section. The sixth section will present and discuss the results of the regression models. Suggestions for further research will be given in section 7. The paper will conclude with the main findings and some policy recommendations.
2. Literature Review
2.1 Carbon Intensity of Well-Being
The carbon intensity of human well-being is a vague concept and therefore this terminology needs to be specified with regards to this paper. We will first focus on the term ‘carbon intensity’, which starts by clarifying the meaning of ‘climate change’. According to the Intergovernmental Panel on Climate Change (1990) there is a natural greenhouse effect which raises global temperature. On top of that, humans are now causing an enhanced greenhouse effect by increasing the concentrations of greenhouse gases, which in turn increases global temperature by a higher degree. The panel distinguishes four greenhouse gases that result from human activity: carbon dioxide, methane, nitrous oxide, and fluorinated gases.2 This paper will focus on the gas carbon dioxide (CO
2), because it has shown to be responsible for more than 75% of the enhanced greenhouse effect (figure 1). In this paper, ‘carbon intensity’ refers to the amount of CO2 the average person in a country emits, i.e. CO2 emissions per capita.
Figure 1. Global Greenhouse Gas Emissions by Gas. Based on global emissions in 2010 as calculated in
Contribution of Working Group III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (source: United States Environmental Protection Agency).
2 Explaining the scientific mechanisms of how these greenhouse gases result from human activities and how
they enhance climate change goes beyond the scope of this economic paper. It is taken as common knowledge that carbon dioxide emissions harm the environment and therefore should be limited.
Although CO2 emissions per capita seems to be a straightforward concept, controversy exists about its method of calculation. While production-based measures are most commonly used in conventional climate change studies, in recent years scholars started to calculate consumption-based measures of CO2 emissions for sustainability analyses. As can be seen in figure 2, worldwide industry accounts for almost one third of total anthropogenic greenhouse gas emissions (direct and indirect emissions). This is a lot. In recent decades the global pattern of industrial activities has shifted from developed countries to developing countries such as China, India, and Brazil. This is partly due to lower wages, higher profitability, and differences in legislation (Dodman, 2009). This shift leads to load displacement issues, meaning a shift in production that is not accompanied by a shift in consumption. Shifts in production can be seen by the large changes in proportion of urban greenhouse gas emissions that can be contributed to the industrial sector for many countries. For example, Weber et al. (2008) calculate that in 1987 12% of China’s emissions were generated in the production of export goods, and this number rose to 33% in 2005. Furthermore, industrial emissions accounted for 80% of Shanghai’s and for 65% of Beijing’s total emissions (Dodman, 2009). These figures show that China’s urban emissions are for a large part attributable to export goods. In addition, Peters et al. (2011) compared national-level production- and consumption-based measures of emissions and find that consumption patterns did not shift accordingly. They show that higher-developed countries tend to have greater consumption-based than production-based emissions figures and thus tend to be net importers of carbon emissions. The reverse holds for less-developed countries. This implies that environmental consequences of developed countries are being displaced to less-developed countries, rather than reduced. Furthermore, the means of load displacement will not be available to the latest-developing countries and is therefore not sustainable (Rothman, 1998). The spatial separation between the point of consumption and the emissions generated in production underscores the importance of using consumption-based measures of CO2 emissions, which account for emissions embodied in trade and production networks. In this way attention is shifted to consumption lifestyles that induce unsustainable levels of CO2 emissions, and therefore the carbon emissions implications of human well-being can be analyzed more accurately.
Figure 2. Total global anthropogenic GHG emissions (Gt CO2 eq/yr) by economic sectors. The inner circle
shows direct GHG emission shares (in % of total anthropogenic GHG emissions) of five economic sectors in 2010. The pull-out shows how indirect CO2 emission shares (in % of total anthropogenic GHG emissions) from electricity and heat production are attributed to sectors of final energy use (source: IPCC, 2014b).
However, as the final calculation of consumption-based measures of CO2 emissions needs to incorporate many more systems than calculations of production-based measures, they have greater degrees of uncertainty. Although this decreases their reliability and makes it difficult to use consumption-based measures as the single indicator in climate policy, they are still considered necessary to be used in conjunction with other indicators. Consumption-based data on carbon emissions could provide considerable insight into climate change policies by helping prioritize domestic mitigation policies, differentiate reduction commitments of countries, and harmonize trade and climate policy (Peters, 2008).
Now we will turn to the second part of CIWB; human well-being. International development efforts to raise human well-being are not only needed to increase the wealth of the poor, but also because people who are socially, economically, politically, or otherwise marginalized are especially vulnerable to climate change and therefore will be the first to experience the consequences of it (IPCC, 2014). Although much research has been done in the international development field, only recently the body of research focusing on the environmental impacts of human well-being started to grow. Specifically, much attention is paid to the amount of fossil fuels used to maintain a certain level of human well-being (Jorgenson, 2015). In this field sustainability is conceptualized as the efficiency with which human well-being is produced from the use of (natural) resources (Dietz et al., 2009). This study will use average life expectancy as a proxy of human well-being. We chose life expectancy because it is well measured in most countries, it is widely accepted as an indicator of well-being, and it is employed in past research on the carbon intensity and ecological intensity of well-being (Knight, 2014; Jorgenson, 2015; Jorgenson & Givens, 2015). Furthermore, other indicators for human well-being tend to be very subjective, which would not be appropriate in an economic paper.
Thus, the consumption-based carbon intensity of human well-being measures the level of consumption-based carbon emissions per level of human well-being.3 Higher values of CIWB indicate larger levels of consumption-based CO2 emissions per unit of well-being, whereas lower values mean lower levels of consumption-based CO2 emissions per unit of well-being. From an environmental point of view low levels of emissions are preferred, and from a development point of view higher levels of human well-being are desired. Hence, lower values for the consumption-based CIWB are preferred. Jorgenson and Givens (2015) state the central question in this area of research: ‘if reducing CIWB is a potential pathway towards greater sustainability, how can nations throughout the world successfully achieve it?’. So far research on the effect of economic development and inequality on the consumption-based CIWB has been conducted. Jorgenson and Givens (2015) estimated the changing effect of economic development, measured by GDP per capita. They find that the effect was varying across OECD and Non-OECD countries, as well as among Non-OECD nations in different geographic regions. GDP per capita had a relatively large, positive relation to the consumption-based CIWB in OECD nations. Although with a smaller effect on CIWB than for OECD countries, all Non-OECD regions experienced patterns of increasingly unsustainable relationships between CIWB and economic development between 1990 and 2008. They conclude that economic development increases the consumption-based CIWB and is therefore not by itself a pathway to sustainability. Jorgenson (2015) finds that inequality has a positive relation with the
3 In this paper, the abbreviation CIWB will always refer to the consumption-based carbon intensity of
based CIWB for OECD nations, and this effect is stabilizing. Inequality was negatively correlated with CIWB for Non-OECD nations until 1999 and now has an increasingly unsustainable effect on CIWB.
This paper will add to this body of research by estimating the effect urbanization had on the consumption-based CIWB through the time period 1990-2008. Urbanization is a defining phenomenon of the last decades. The World Bank’s 2009 Urban Strategy highlights that the biggest urban transformation is taking place in developing countries, while in industrialized countries already most of the population lives in urban areas. This transformation creates opportunities as well as challenges. Opportunities are created by the concentration of wealth, people and productivity, but cities also concentrate vulnerability to natural disasters and to long-term changes in climate (Hoornweg, 2011). First the environmental impacts of urbanization will be discussed, followed by the relation of urbanization and human well-being.
2.2 Urbanization and its Ecological Consequences
The ecological consequences of urbanization have been studied extensively and empirical studies find both positive and negative effects. We will start with the latter. When countries just start to experience an urban transition, urbanization is associated with industrialization that may lead to higher emissions and energy consumption. As people move from rural to urban areas this increases energy consumption and emissions by: (i) agricultural operations become less labor intensive and must mechanize, (ii) food consumers are spatially separated from food producers, increasing transport and thus emissions, and (iii) industry processes use more energy per worker than traditional agricultural and manufacturing processes (Liddle, 2014). The World Bank reports similar consequences of urbanization, as they state that urbanization is accompanied by migration, changes in land use, and spatial development which are likely to increase CO2 emissions (Hoornweg, 2011). In addition, consumption-based CO2 emissions could increase in cities because economic development often accompanies urban growth (Jorgenson et al., 2014). Gonzalez (2005) supports this argument as he finds that in cities mass consumption, compared to non-urban areas, takes place due to the expansion of the service economy and increases in economic development (Jorgenson et al., 2014). Liddle (2014) explains how urbanization is a proxy for the amount of people that have access to energy and electricity grids and hence urbanization has a positive relation with consumption of such energy. In short, cities are centers of built infrastructure, economic activity, energy consumption, and transportation networks and this built urban environment creates an energy-rich path dependency, which is accompanied by high carbon emission levels (Jorgenson et al., 2014).
On the other hand, urbanization can also have positive ecological consequences. High population density in urban areas could provide benefits of economies of scale and proximity, such as: (i) spatially concentrated consumption can limit carbon emissions embodied in the transportation of goods; (ii) transportation for routine daily activities will be reduced for urban citizens; (iii) it is cheaper to provide infrastructure and services that are needed to minimize environmental hazards; (iv) it is cheaper to enforce environmental legislation because of the concentration of enterprises; and (v) the concentration of people and industries provides opportunities for the fast spread and adaptations of innovations, both technical and behavioural. Furthermore, it is argued that urbanization leads to more efficient household-level resource consumption because of smaller urban residences that use less electricity and heating. In addition, people have greater access to public transport in urban areas and walking and cycling can be more easily encouraged (Jorgenson et al., 2014; and Dodman, 2009).
Most studies find positive correlations between urbanization and various forms of environmental pressure, including energy consumption and CO2 emissions. To start with, Jones (1991) derived a positive correlation between urbanization and energy use per capita, using a national level analysis with cross-sectional data. He finds that urbanization increases transport energy use and energy use per unit of output, although it enables cities to benefit from economies of scale in production. York et al. (2003) used a Stochastic Impacts by Regression on Population, Affluence and Technology (STIRPAT) model to find a positive relation between urbanization and national energy footprints and CO2 emissions. Following a similar model, the results of Cole and Neumayer (2004) suggest that urbanization increases CO2 emissions at the global level. Holtedahl and Joutz (2004) find that urbanization increases residential energy consumption. They argue that a households’ accessibility to electricity increases when moving to an urban area, and households increase their energy consumption compared to living in rural areas. Similarly, Liu (2009) suggests that urbanization has a positive relation with energy use, but that its influence is declining over time. The latter is due to improvements in industrial and technological structure, and more efficient use of resources. Correspondingly, Dodman (2009) finds that for several wealthy cities per capita greenhouse gas emissions were significantly lower than their national average.
While most early studies assume that urbanization has homogeneous ecological impacts for all countries, it is now argued by many scholars that urbanization’s effects on emissions are heterogeneous. Using a STIRPAT model, Fan et al. (2006) show that the effect of urbanization on CO2 emissions is restricted by the level of economic development, energy structure, energy consumption per capita, and the gap between the rural and the urban. Using a panel data set of 99 countries over the period of 1975-2005 and a STIRPAT model, Poumanyvong and Kaneko (2010) find that the impact of urbanization on energy use and emissions varies across stages of development, because the latter influences the types of infrastructure and investments. Urbanization has a positive impact on low-, middle-, and high-income groups, but this effect is far greater for the middle-income group. Poumanyvong and Kaneko (2010) argue that their findings support the argument of the ecological modernization theory; the process of modernization may cause substantial environmental issues, but further modernization can diminish such problems. Similarly, Liddle (2013) used city-level datasets from developed and developing countries to find that urbanization increases carbon emissions from transport, but to a greater extent in middle-income countries, than in high- or low-income nations. Liddle (2014) summarizes evidence from cross-country, macro-level studies on the way urbanization, among other factors, influences carbon emissions and energy consumption. He suggests that a different impact of urbanization in less developed countries than in developed countries can be explained by the fact that the urbanization levels of most developed countries did not change substantially in the past decades. However, others argue that not economic development, but regional differences provide an important context for the effect of urbanization on emissions. Thomas Rudel (2005) examined deforestation patterns and finds that historically and regionally specific features generate distinct environmental pressures (Jorgenson et al., 2014). Marcotulli et al. (2013) identified global urban greenhouse gas emissions by region and examined the urban-to-rural gradient. They argue that in the developing world urban emissions levels can be higher than those in non-urban areas, due to better energy and transport infrastructure. However, this pattern may be less apparent in South America, since the region is highly urbanized. In addition, Jorgenson et al. (2014) researched the changing effect of urbanization on CO2 emissions per capita and find that the effects of urbanization on production-based CO2 emissions are not monolithic across regions. While urbanization became less unsustainable (in the context of CO2 emissions per
capita) in Europe, North America and Oceania, it became more unsustainable in Asia and Latin America. Urbanization had little effect on production-based CO2 emissions in Africa.
2.3 Urbanization and Human Well-Being
The relation between urbanization and human well-being will now be discussed. While no quantitative research can be found on the effect of urbanization on life expectancy, most qualitative research on this relation has been done in the public health field.4 The key factors affecting health in cities are: the physical environment, the social environment, and access to health and social services (Galea & Vlahov, 2005). The consequences of urbanization on these key factors differ greatly across geographic regions. While adequate research on the effect of urbanization on well-being in the recent decades is lacking for high-developed countries, plenty of qualitative research can be found for developing countries. In Africa, many urban citizens are lacking basic physical and environmental health-related essentials like safe and adequate water and sanitation, and healthy housing with electricity. This gives many urban Africans low life expectancy numbers. In Latin America, the basic physical health problems are largely understood and acted upon. However, high socio-economic inequality levels in Latin American cities cause rising levels of violence and mental ill health for urban citizens. This decreases life expectancy. In Asia, many countries are rapidly urbanizing and opening up to Western market influences. This creates many behavioural health problems, such as smoking, other drug abuse, and HIV/AIDS (Harpham, 2009). More positively, cities often provide a wider range of social services than non-urban areas across all regions. This contributes to well-being, and hence urbanization may increase nation-wide human well-being. However, the effect that these social services can have on life expectancy differs greatly depending on the quality of the services, the relative strain posed upon the services by high-need populations, and the accessibility to the services (Vlahov & Galea, 2002).
3. Hypotheses
Based on the literature review in section 2, expectations for the regression results are formed. First of all, it was discussed that most developed countries are ‘fully’ urbanized, while urbanization rates in developing countries have been growing fast in the past decades. In addition, the ecological modernization theory suggests that further urbanization can diminish ecological problems for developed countries. Taking this into account, we expect urbanization to have a positive but declining effect on the consumption-based CO2 emissions for the combined sample of highly-developed countries (in Europe, North America, and Oceania) as urbanization increases further. However, expectations are that this effect might not be highly significant, as increases in urbanization rates for this sample were limited in the time period used in this study. The latter can be explained by the notion that these countries are ‘fully urbanized’. For the three regional samples of Latin America, Asia, and Africa, we expect urbanization to increase consumption-based CO2 emissions and this increase is expected to grow throughout the time period. As mentioned before, urbanization is often accompanied by economic development, which could increase consumption-based CO2 emissions. In Africa and Latin America it is expected that the increase in urban slums that
4 It is beyond the scope of this economic paper to fully explain the theoretical background on the relationship between urbanization and human well-being. Therefore, only a small insight in the field of public urban health will be given.
accompany urbanization will have a depressing effect on consumption-based CO2 emissions. However, it is expected that this effect will be offset by the positive influences on consumption-based CO2 emissions.
Secondly, it is expected that urbanization will have a positive effect on life expectancy in the sample for highly-developed nations, because of the high accessibility to social services. This effect is expected to be small, as accessibility to social services in non-urban areas is also adequate. Furthermore, the effect is expected to stay constant throughout the time period, as this accessibility did not change much in recent decades. For the regional samples of Latin America, Asia, and Africa it is expected that urbanization decreases life expectancy because of the reasons mentioned in part 2. This negative effect is expected to be higher in Africa and Latin America than in Asia, because of the presence of many urban slums.
Consequently, we expect the effect of urbanization on the consumption-based CIWB for the combined sample of highly-developed nations to be positive but declining through the time period. Moreover, it is expected that urbanization positively affects the CIWB for the three regional samples of Latin America, Asia, and Africa and that this effect is increasing over time.
4. Dataset
This paper will first examine the effect of urbanization on the consumption-based carbon intensity of well-being, followed by examinations of the separate effects of urbanization on consumption-based CO2 emissions per capita5 and average life expectancy. The effects will be estimated for five samples of nations for the 1990-2008 period. Our time period consists of only 19 years because annual data on consumption-based CO2 emissions are not widely available in large databases. We use the calculations of Peters et al. (2011), which we will further elaborate on below. Hence our analyses excludes countries for which these data were not estimated, which are predominantly small island countries and former Soviet nations (Jorgenson & Givens, 2015). Interaction variables for urbanization rates will be used to examine the extent to which the effect of urbanization on CIWB, CO2 emissions per capita and life expectancy might change through time. Therefore, only countries for which annual data on urbanization rates report no missing values are included. We exclude countries for which annual data on GDP per capita and inequality were missing for more than three years. In addition, Hong Kong and Singapore are excluded because their urbanization rates are one hundred per cent for the entire time period, making both nations large outliers.
The first sample will be an overall sample that includes 65 nations that remain after the above-mentioned exclusions. The second, third, fourth and fifth sample split the overall sample in regional samples, which are largely defined by continent. Table 1 shows the nations that are included in the analyses divided by geographic regions. Regional samples were chosen because of two reasons. First, prior research showed that significant differences exist in the effect of urbanization on national greenhouse gas emissions between geographic regions as explained in section 2 (Marcotulli et al., 2013; Jorgenson et al., 2014). Second, it is widely argued by sustainability scientists, such as Victor et al.(2014) and the IPCC (2014b), that regional analyses should be better integrated into effective climate change mitigation policies (Jorgenson et al., 2014). The four regional samples include 21 nations in the combined region of Europe, North America, and Oceania,
5 In the remaining part of the paper the use of the terms CO
2/carbon emissions will always refer to
15 nations in Latin America, 14 nations in Asia, and 15 nations in Africa. The most highly income nations in Europe, North America, and Oceania are combined into one sample to preserve degrees of freedom.
Table 1. Nations included in the analyses.
Europe, North America, Oceania
Latin America Asia Africa
Australia Argentina Bangladesh Botswana
Austria Bolivia China Egypt
Belgium Brazil India Ethiopia
Canada Chile Indonesia Madagascar
Denmark Colombia Iran Malawi
Finland Costa Rica Japan Mauritius
France Ecuador Korea Morocco
Germany Guatemala Lao Nigeria
Greece Mexico Malaysia Senegal
Ireland Nicaragua Pakistan South Africa
Italy Panama Philippines Tanzania
The Netherlands Paraguay Sri Lanka Tunisia
New Zealand Peru Thailand Uganda
Norway Uruguay Vietnam Zambia
Portugal Venezuela Zimbabwe
Spain Sweden Switzerland Turkey
United Kingdom
United States of America
A note should be placed on the reliability of the data of African countries. Jerven (2013) investigated African development statistics in his book ‘Poor Numbers: How We Are Misled by African Development Statistics and What To Do About It’. Jerven (2013) cites Riddel: ‘(….) the most fundamental problem with the available Africa data is that these are widely known to be inaccurate but the degree of inaccuracy cannot easily be judged - itself a sign of the underdevelopment of the region’. Jerven (2013) explains how data coverage in the national accounts for African countries shows large gaps, and data has been collected with changing appreciation of the importance of the state, the informal sector, and the rural sector. He therefore argues that it is not possible to compare within countries and across countries at any point in time. Although Jerven’s critique on the quality of African statistics is understandable, following his advice would mean that no research on this geographic area can or should be done. Because African countries are now experiencing large urbanization rates and are growing as emitting nations, we decided to include African countries in this paper to avoid a large research gap.
4.1 Dependent variables
The dependent variable that is used in the first regression for the group of samples is the consumption-based carbon intensity of well-being. The CIWB is a ratio between anthropogenic carbon emissions and a measure of human well-being. This paper uses the method of calculation of the consumption-based CIWB employed by Jorgenson and Givens (2015).
The consumption-based measures of carbon emissions are calculated by Peters et al. (2011). They adjusted estimates of production-based carbon dioxide emissions for trade to create consumption-based measures. These measures attribute the emissions generated in the process of production to the country that consumes the goods, rather than to the country that produces the goods by using input-output analysis techniques. Included in the measures are, among others, emissions from fossil fuel combustion, cement production, and gas flaring (Jorgenson & Givens, 2015). The annual consumption-based carbon emissions data from Peters et al. (2011) are converted into per capita measures by using total population size data from the World Bank’s World Development Indicators database (http://databank.worldbank.org/, accessed December 4, 2015). The per capita measure is used as the numerator for the CIWB ratio.
Following past research (Jorgenson, 2015; Jorgenson & Givens, 2015), we employ life expectancy at birth as the denominator of CIWB. This variable indicates the average number of years a newborn would live if prevailing patterns of mortality at the time of its birth persist throughout its life. These data were obtained from the World Bank’s World Development Indicators database (http://databank.worldbank.org/, accessed December 4, 2015).
As explained by Jorgenson and Givens (2015), using a ratio as the dependent variable creates the following complication: the variability of consumption-based carbon emissions and life expectancy can differ considerably, which could drive variation in the ratio. In the overall sample the variation in consumption-based carbon emissions per capita is larger than the variation in average life expectancy: a coefficient of variation (standard deviation/mean) of 1.028 versus one of 0.142. To resolve this complication, we follow the method of Dietz and York (2012). We add a constant to the consumption-based CO2 emissions per capita data to make the numerator and denominator of the CIWB equal. In this way the mean of CIWB is shifted, but the variance stays equal (Jorgenson & Givens, 2015). For the overall sample, the coefficient of variation is made equal by adding 33.923 to the per capita consumption-based emissions data, leading to a coefficient of variation of 0.142 for both the numerator and denominator of the CIWB. The measure of consumption-based CIWB for the overall sample is therefore:
CIWB1 = [(ConCO
2PC + 33.923)/Life Expectancy)] * 100
Where CIWB1 is consumption-based carbon intensity of well-being for the overall sample, ConCO 2PC is the consumption-based carbon dioxide emissions per capita, and Life Expectancy is average life expectancy. A multiplication by one hundred was done to scale the ratio, consistent with past research (Dietz & Jorgenson, 2014; Jorgenson, 2014; Jorgenson, 2015; Jorgenson & Givens, 2015).
The same method of calculation is used to estimate the values of the constant for each regional sample. This leads to the following measures of consumption-based CIWB. For the combined sample of Europe, North America, and Oceania the measure of consumption-based CIWB (CIWB2) is:
CIWB2 = [(ConCO
2PC + 112.222)/Life Expectancy)] * 100
The measure of consumption-based CIWB for the sample of Latin American nations (CIWB3) is:
CIWB3 = [(ConCO
2PC + 19.275)/Life Expectancy)] * 100
The measure of consumption-based CIWB for the sample of Asian nations (CIWB4) is:
CIWB4 = [(ConCO
2PC + 39.834)/Life Expectancy)] * 100
The measure of consumption-based CIWB for the sample of African nations (CIWB5) is:
CIWB5= [(ConCO
2PC + 8.745)/Life Expectancy)] * 100
To examine to which extent urbanization rates affect the numerator and denominator separately, two additional regressions for all the samples are executed. The dependent variable for the second regression will be ConCO2PC + C; the consumption-based CO2 emissions per capita plus the constant that was also added in the formula for the CIWB. All variables will be transformed with the base 10 logarithm as will be explained below. Therefore, the constant is added to make all base 10 logarithm forms of the variable positive. For each sample, the constant that is added in the formula for the CIWB (a different constant per sample) will also be added to ConCO2PC. At last, the dependent variable for the third regression will be LE: average life expectancy.
4.2 Independent variables
All three models will include the same independent variables. To measure urbanization, data on urban population (% of total) was obtained from the World Bank’s World Development Indicators database (http://databank.worldbank.org/, accessed December 4, 2015). Urban population refers to people living in urban areas as defined by national statistical offices. However, as Jorgenson et al. (2014) note, the data on urbanization has the following limitation. Typically an area is considered urban if it has a population of 2,000 or more, but national definitions of urbanization may differ. As this study uses country-level fixed effects and herewith focuses on estimating within-country effect of urbanization, potential differences in definitions of urbanization are accounted for. To assess to which extent the effect of urbanization on the dependent variable increased or decreased through time, interactions between urbanization rates and dummy variables for each yearly observation (1990 to 2008) are calculated and used, with 1990 as the reference year. The use of interaction variables is commonly employed in past research on the effect of economic development and inequality on CIWB, and the effect of urbanization on production-based carbon emissions (Jorgenson et al., 2014; Jorgenson, 2015; Jorgenson & Givens, 2015).
Two variables are included as control variables: GDP per capita and inequality. These variables have been the primary focus of previous research on CIWB, and consistently found a positive relationship between both variables and CIWB (Jorgenson, 2015; Jorgenson & Givens, 2015). To measure economic development, data on GDP per capita (measured in constant 2005 US dollars) was obtained from the World Bank’s World Development Indicators database (http://databank.worldbank.org/, accessed December 4, 2015). The national-level Gini index of inequality for household disposable income (post-tax, post-transfer) taken from Solt’s Standardized World Income Inequality Database (Version 4, http://myweb.uiowa.edu/fsolt/swiid/swiid.html) was used to measure inequality. The Standardized World Income Inequality Database (SWIID) standardized observations collected from the United Nations University’s World Income Inequality Database (Version 2.0c), the OECD Income Distribution Database, the Socio-Economic Database for Latin America and the Caribbean generated by CEDLAS and the World Bank, Eurostat, the World Bank’s PovcalNet, the UN Economic Commission for Latin America and the Caribbean, the World Top Incomes Database, national statistical offices around the world and many other sources by using a custom missing-data multiple-imputation algorithm. The Luxembourg Income Study data is used as the standard. In the Gini index the values range from 0 (perfect equality) to 100 (perfect inequality) (Jorgenson, 2015).
5. Methodology
A time-series cross-sectional Prais-Winsten regression model with panel-corrected standard errors (PCSE) is used for all three models. This regression model was chosen because it allows for heteroskedasticity and serial correlation. Using the Wooldridge test for autocorrelation in panel data (Wooldridge, 2002), serial correlation in all three OLS models (with successively CIWB, CO2 emissions, and life expectancy as dependent variable and urbanization, GDP per capita and inequality as independent variables) is found for all samples, except for the sample of African nations in model 1 and 2 (Appendix 1). However, a Prais-Winsten regression model is also used for this sample to make results better comparable. Moreover, after applying the modified Wald test for groupwise heteroskedasticity developed by Greene (2000), the presence of heteroskedasticity is detected for all sample groups in all models (Appendix 1). In line with past research on the consumption-based CIWB and CO2 emissions (Jorgenson et al., 2014; Jorgenson, 2015; Jorgenson & Givens, 2015), PCSE is employed. The dataset has less time periods than panels. In this case, the feasible generalized least-squares estimator that is commonly used in panel data analyses produces standard errors which can induce extreme overconfidence (Jorgenson & Givens, 2015). PCSE is used to overcome this issue. Furthermore, the Prais-Winsten regression model produces fitted values by transforming the data. Part of this transformation is taking the first-order difference of all variables included in the model. In this way, the Prais-Winsten regression model takes care of possible non-stationarity of the variables.
It is important to decide on using fixed or random effects. Due to the fact that panel data is used, unmeasured country- and year-specific intercepts could cause heterogeneity bias. It is highly likely that the model is underspecified, which could result in omitted variable bias. In addition, unobserved factors relating to a country’s socioeconomic characteristics and policy choices will be present, resulting in a need for country-specific intercepts. Year-specific factors shared by all countries can be related to macroeconomic events. In order to decide on fixed or random effects, we apply the Hausman test on the main independent variable (urbanization) and the three
dependent variables for the overall sample in Appendix 2. The results show the need to use country-specific intercepts for all three regressions. In addition, time-fixed intercepts are included after the tests showed their significance for all models (Appendix 2). The equivalent is a two-way fixed effects model. This modeling technique controls out between-country variation and estimates within-country effects, an approach which is regularly used in studies that analyze panel data (Jorgenson & Givens, 2015).
All variables are transformed into base 10 logarithm form (log) to make interpretation of the results easier. In this way, the regression models estimate elasticity coefficients; the coefficient of the independent variable is the estimated net percentage change in the dependent variable associated with a one per cent increase in the independent variable (Jorgenson et al., 2014). It should be noted that results from a Prais-Winsten regression model can be interpreted in the usual way as they are the appropriate ones from a generalized least-squares fit, even though they have been transformed.
First, the following two-way fixed effects Prais-Winsten regression model is estimated for each of the five samples:
CIWBit (log) = β1Urbanizationit(log) + β2 Urbanizationit(log) * year1991t + …. + β19 Urbanizationit(log) * year2008t + β20GDP per capitait(log) + β21 Inequalityit(log) + ci + tt + eit (1)
where CIWB is the consumption-based carbon intensity of well-being, with a different value for CIWB depending on the constant that is added to the consumption-based CO2 emissions as explained in section 4. Furthermore, the model includes the urbanization rate (Urbanization), the interactions between urbanization and the dummy variables for each year, with 1990 as the reference category (Urbanizationit(log)* year 1991t + …. + Urbanizationit(log) * year2008t), GDP per capita (GDP per capita), income inequality (Inequality), the country-specific intercepts (ci), the year-specific intercepts (tt), and the disturbance term unique to each country at each point in time (eit). The i denotes the country and the t denotes the year. As employed in past research (Jorgenson et al., 2014; Jorgenson, 2015; Jorgenson & Givens, 2015), the coefficient for Urbanization captures the effect of the base year 1990, whereas the overall effect of urbanization on the dependent variable for other years equals the sum of the coefficient for Urbanization and the appropriate interaction term (if statistically significant).
For the second and third model the same conditions hold, except that the dependent variable changes. The second model is estimated as follows:
ConCO2PCiit(log) = β1Urbanizationit(log) + β2Urbanizationit(log) * year1991t + …. +
β19Urbanizationit(log) * year2008t + β20GDP per capitait(log) + β21 Inequalityit(log) + ci + tt + eit (2)
where ConCO2PC is the consumption-based CO2 emissions per capita plus the constant, which has a different value per sample, as explained in section 4.
Life Expectancyit(log) = β1Urbanizationit(log) + β2Urbanizationit(log) * year1991t + …. + β19Urbanizationit(log) * year2008t + β20GDP per capitait(log) + β21 Inequalityit(log) + ci + tt + eit (3)
where Life Expectancy is the average life expectancy at birth.
6. Results
6.1 Descriptive statistics
Appendix 3 shows the plots over time for the dependent variables and the main independent variable in log10 form and in their first order difference form for one country as an example. The plots show that all variables in their log10 form seem to be trend stationary (non-stationary). The Prais-Winsten regression model transforms all variables into their first-order difference values. It can be seen in Appendix 3 that this transformation seems to eliminate the time trend for all variables and is therefore sufficient to make the dependent variables and urbanization stationary. Similar plots are observed for all countries.
The scatter plots and correlation matrices for the dependent variables and urbanization can be found in Appendix 4 and 5. Since the correlation values of the main variables in normal form and log(10) form do not significantly differ in the overall samples, the normal form is used to make interpretation easier. As expected, the relation between urbanization and CIWB is positive for all geographic samples. However, the strength of this positive relation between urbanization and the CIWB differs for the geographic samples. The samples for Latin American and Asian countries show a moderate upward sloping line. However, for the combined sample of Europe, North America, and Oceania and for the sample of African nations this line is only slightly increasing. This implies that the relationship between urbanization and CIWB is weak in these regions. Surprisingly, the relation between urbanization and the CIWB for the overall sample is negative. However, this result is not very reliable as the correlation between urbanization and CIWB is quite low in the overall sample (r=-0.17), which can be explained by; (i) the differences in relations between urbanization and the CIWB for the geographic samples; and (ii) by the fact that urbanization has separate effects on both the numerator and the denominator of the CIWB. This first reason is shown in the varying results of the correlation values between urbanization and CIWB for the different samples: for the combined sample of Europe, North America, and Oceania r=0.03, for the sample of Africa r=0.11, for the sample of Latin America r=0.22, and for the sample of Asia r=0.34. The second reason is shown by looking at the correlation of CO2 and urbanization (r=0.66) and the correlation of life expectancy and urbanization (r=0.73) of the overall sample, which are both good correlation values and therefore do imply significant effects of urbanization on both variables. In addition, the low correlation values between urbanization and CIWB for all samples could be the result of the use of a ratio in the CIWB. Opposite effects of urbanization taking place on the numerator and denominator could cancel out the effects in the CIWB as a dependent variable.
Urbanization is both positively related to the consumption-based CO2 emissions per capita and life expectancy for all samples as can be seen in the scatter plots. For CO2 emissions, this positive relation is strong for all samples. For the combined sample of Europe, North America, and Oceania it was not expected that this relation is strong. The positive relation between life expectancy and urbanization is also surprising, as it was expected that urbanization would decrease life expectancy
in Latin America, Asia, and Africa. The positive correlation is strongest for Africa, which is especially unexpected with regards to the existence of many urban slums in Africa.
In the combined sample of Europe, North America, and Oceania, the correlation between CO2 and urbanization (r=0.39) and the correlation between life expectancy and urbanization (r=0.34) are both low. The low correlations combined with the similar slopes in the scatter plots gives the expectation that urbanization will not have a significant effect on CIWB in Europe, North America, and Oceania. For Latin America, correlations between CO2 and urbanization (r=0.66) and life expectancy and urbanization (r=0.43) are respectively good and modest. In the sample for Asia, the correlation between CO2 and urbanization (r=0.91) is very high and the correlation between life expectancy and urbanization is good (r=0.65). In Africa, the correlation between CO2 and urbanization (r=0.67) and life expectancy and urbanization (r=0.61) are both good. Taking this into account, it is expected that urbanization will have a significant effect on the CIWB in Latin America, Africa, and Asia.
Furthermore, it is important to add economic development as a control variable in all three models, as urbanization and GDP per capita are highly correlated in the overall sample (r=0.59). This correlation is especially high in the samples for Latin America, Asia, and Africa. Not including GDP per capita would lead to omitted variable bias. The next section examines the results of the regression models.
6.2 Empirical results
Appendix 6 shows the final results of the three regression models for all samples. All regression models show close to perfect R-square statistics, which is unrealistically high. This is largely due to unreported country-specific and year-specific intercepts, which help account for omitted variable bias. Furthermore, Wooldridge (2012) states in his book ‘Introductory Econometrics: A Modern Approach’ that in a Prais-Winsten regression model the R-squared is calculated from the final regression of the transformed dependent variables on the transformed independent variables. He argues that it is therefore not clear what the R-square statistic indicates in a Prais-Winsten regression model. Hence, we argue that not much attention should be given to the R-square statistics in this research.
We will now first look at the general results for the three regression models separately, after which we will assess the effect of urbanization on consumption-based CIWB, consumption-based CO2 emissions per capita and life expectancy over time by analyzing the elasticities of urbanization for the three dependent variables per sample.
Table 1 in Appendix 6 shows the regressions results for the first model, which has CIWB as the dependent variable. The effect of urbanization on the CIWB is positive and statistically significant in the base year, 1990, for the overall sample, Latin America and Africa. For Asia this effect is statistically significant but negative. For the combined sample of Europe, North America and Oceania this effect is not statistically significant, although most interaction variables do have significant elasticity coefficients. The values of these coefficients do not change considerably over time. For the other samples most interaction variables are positive, increasing over time and statistically significant. For all samples GDP per capita has a positive and statistically significant effect on the CIWB, except for Africa for which this effect is not statistically significant. The latter was not expected, as GDP per capita had a strong positive significant effect for Africa in earlier research on the CIWB (Jorgenson & Givens, 2015). Inequality does not have a significant effect on CIWB in the overall sample and the combined sample of Europe, North America, and Oceania. For the last
mentioned sample this is a surprising finding, as earlier research on the CIWB found that the effect of inequality was positive and highly statistically significant for high-income countries (Jorgenson, 2015). For Latin America the effect of inequality on CIWB is positive and statistically significant, whereas for Asia and Africa this effect is negative.
Table 2 (Appendix 6) reports the regressions results for the model with consumption-based CO2 emissions per capita as the dependent variable. The effect of urbanization on CO2 emissions per capita is positive and statistically significant for the overall sample in 1990. This effect is mainly driven by countries in Latin America, where a 1% increase in urbanization increases CO2 emissions per capita by 0.17% in the base year. In Asia urbanization decreases CO2 emissions slightly in 1990. The effect is not statistically significant for the combined sample of Europe, North America, and Oceania and for Africa. GDP per capita increases CO2 emissions per capita for all samples, consistent with prior research as mentioned in section 2. Inequality only has a significant effect on CO2 emissions per capita in Asia, although this effect is almost negligible.
In Table 3 (Appendix 6) the regression results for the last model, with life expectancy as dependent variable, are reported. In the base year, urbanization has a positive effect on life expectancy for the combined sample of Europe, North America, and Oceania and for Asia, whereas this effect is negative for Latin America and Africa. Although the effects for the four regional samples are all highly significant, the effect of urbanization on life expectancy in the base year is not statistically significant for the overall sample. This can be explained by contrary effects in the regional samples. The negative effect in Latin America and Africa is intuitive because many people in cities in Latin America and Africa live in slums with poor living conditions. The effect of GDP on life expectancy is positive and statistically significant in the overall sample. In the regional samples this effect is only statistically significant for Africa and Asia; in Africa the effect of GDP is positive while in Asia it is negative. No clear reason can be found for the latter. Inequality only has a significant effect on life expectancy in the combined sample of Europe, North America and Oceania and in Latin America, both effects are negative.
We will now examine the changes over time of the effect of urbanization on the three dependent variables, consumption-based CIWB, consumption-based CO2 emissions per capita, and life expectancy, per sample. This will be done by looking at the elasticity coefficients for the estimated effects of urbanization for each year, which are based on the test of statistical significance for the interactions between urbanization and time as reported in Appendix 6.
The elasticity coefficients for the overall sample are given in table 2. The coefficients for the effect of urbanization on CIWB are positive for the entire period. In 1990, a 1% increase in urbanization rates led to a 0.06% increase in CIWB. This effect of urbanization decreased until 1995, after which it started to increase until in 2008 a 1% increase in urbanization rates led to a 0.11% increase in CIWB. Compared to the lowest value in 1995, the CIWB more than doubled in 13 years. The trend of the CIWB can be explained by looking at the elasticity coefficients of urbanization for CO2 emissions per capita and life expectancy. In 1990, urbanization had a positive effect on CO2 emissions, while the elasticity for life expectancy was zero. The elasticities for CO2 emissions decreased until 1995, while the elasticities for life expectancy increased until approximately 1997. This explains the declining trend of the CIWB elasticities. Around 1999, elasticities for CO2 emissions began to increase, while elasticities for life expectancy decreased again and became even negative from 2002 onwards. Thus urbanization led to higher CO2 emissions per capita worldwide and lowered average life expectancy, hereby increasing the CIWB from around the turn of the century.
The effects are also shown in figure 3, which shows that the increase in CIWB is mainly driven by lower life expectancy.
Year CIWB1 CO21 LE
1990 .062 .038 .000 1991 .062 .038 .002 1992 .052 .033 .005 1993 .046 .029 .007 1994 .044 .028 .009 1995 .043 .028 .008 1996 .047 .033 .010 1997 .047 .033 .010 1998 .050 .034 .008 1999 .049 .030 .006 2000 .054 .033 .000 2001 .056 .031 .000 2002 .062 .033 -.006 2003 .070 .034 -.013 2004 .081 .038 -.018 2005 .088 .038 -.025 2006 .098 .042 -.032 2007 .108 .044 -.040 2008 .114 .043 -.048
Table 2 and Figure 3 – overall sample. Elasticity coefficients for the estimated effect of urbanization on the
consumption-based CIWB, consumption-consumption-based CO2 emissions per capita, and life expectancy for the overall sample. Elasticity coefficients
are derived from the estimated models in Appendix 6. The elasticity coefficients equal the sum of the coefficient for independent variable Urbanization (in the base year 1990) and the coefficient of the appropriate interaction term, provided that these coefficients are statistically significant.
Table 3 shows the elasticity coefficients for the combined sample of Europe, North America, and Oceania, and a graph of these estimates can be found in figure 4. In most years, urbanization did not have a significant effect on consumption-based CO2 emissions per capita in these highly-developed countries. From 1990 to 1994 the effect of urbanization on carbon emissions was slightly positive, while after 2003 this effect became negative. Urbanization did have a significant effect on life expectancy in the combined sample; a 1% increase in urbanization led to a 0.06% increase in life expectancy in 1990. This effect decreased over time, becoming negative in 2008. The effect of urbanization on consumption-based CIWB was insignificant in 1990 and 1991, after which the effect became positive. The elasticity coefficient increased in magnitude over time until 2002, after which the effect became smaller (although ambiguous between 2002 and 2006). It is interesting to note that urbanization decreased both carbon emissions and life expectancy in 2008, the former is desired as opposed to the latter. The effect of urbanization on CIWB was positive in 2008.
Year CIWB2 CO22 LE 1990 .000 .000 .061 1991 .000 .003 .065 1992 .007 .005 .059 1993 .010 .000 .056 1994 .011 .006 .056 1995 .012 .000 .046 1996 .013 .000 .049 1997 .008 .000 .049 1998 .010 .000 .048 1999 .015 .000 .045 2000 .012 .000 .043 2001 .023 .000 .033 2002 .041 .000 .026 2003 .000 -.031 .026 2004 .022 -.014 .024 2005 .000 -.036 .021 2006 .026 -.015 .019 2007 .028 .000 .020 2008 .028 -.016 -.015
Table 3 and Figure 4 – Europe, North America, and Oceania. Elasticity coefficients for the estimated effect of urbanization
on the consumption-based CIWB, consumption-based CO2 emissions per capita, and life expectancy for the combined
sample of Europe, North America and Oceania. Coefficients are derived from the estimated models in Appendix 6. The elasticity coefficients equal the sum of the coefficient for independent variable Urbanization (in the base year 1990) and the coefficient of the appropriate interaction term, provided that these coefficients are statistically significant.
Table 4 and Figure 5 – Latin America. Elasticity coefficients for the estimated effect of urbanization on the
consumption-based CIWB, consumption-consumption-based CO2 emissions per capita, and life expectancy for sample of Latin America. Coefficients
are derived from the estimated models in Appendix 6. The elasticity coefficients equal the sum of the coefficient for independent variable Urbanization (in the base year 1990) and the coefficient of the appropriate interaction term, provided that these coefficients are statistically significant.
Year CIWB3 CO23 LE
1990 .303 .170 -.127 1991 .319 .178 -.134 1992 .323 .170 -.141 1993 .330 .170 -.149 1994 .338 .170 -.156 1995 .342 .170 -.162 1996 .359 .182 -.168 1997 .382 .200 -.174 1998 .401 .212 -.181 1999 .397 .201 -.187 2000 .374 .170 -.194 2001 .383 .170 -.200 2002 .359 .141 -.206 2003 .394 .170 -.211 2004 .385 .170 -.215 2005 .405 .170 -.219 2006 .431 .199 -.222 2007 .454 .219 -.226 2008 .461 .223 -.230
