UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Is free trade free of environmental cost? Jung, Munhee 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2016_november_jung_munhee.pdf [ 1.9MB ]
Metadata
JSON: 24-1.0314920.json
JSON-LD: 24-1.0314920-ld.json
RDF/XML (Pretty): 24-1.0314920-rdf.xml
RDF/JSON: 24-1.0314920-rdf.json
Turtle: 24-1.0314920-turtle.txt
N-Triples: 24-1.0314920-rdf-ntriples.txt
Original Record: 24-1.0314920-source.json
Full Text
24-1.0314920-fulltext.txt
Citation
24-1.0314920.ris

Full Text

Is free trade free of environmental cost? by  Munhee Jung  B.A., Korea University, 2004  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF SCIENCE in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Resource Management and Environmental Studies)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) September 2016  © Munhee Jung, 2016 ii  Abstract  The impact of international trade on the environment has been the field of focus since the 1970’s. There have been a number of empirical studies exploring the environmental consequence of free trade but the results are mixed and only a few environmental indicators have been used in place of the total environmental impact. In this study, I used combined environmental cost data which converted environmental impact indicators into US$ terms (the data is taken from World Bank database). Also, by taking advantage of panel data (observations from 60 countries over 25 years) and (two-way) fixed effects model, I attempted to reduce the threat of endogeneity problem. Most importantly, environmental impact which is filtered through the trade induced changes of economic activity was analyzed in parallel with unfiltered through effects. And the results revealed that trade openness reduces national level environmental cost rather than increasing it. Meanwhile, income related technique effect was found to be underperforming and when the full sample was split into four income groups, the income-environment relationship appeared to be closer to N-shape as opposed to the inverted U-shaped environmental kutznets curve hypothesis.   iii  Preface  This thesis is an original, unpublished intellectual product of the author, Munhee Jung. The empirical work reported in Chapter 3 was conducted on dataset taken from the World Bank and Penn World Table 9.0. iv  Table of Contents  Abstract .......................................................................................................................................... ii Preface ........................................................................................................................................... iii Table of Contents ......................................................................................................................... iv List of Tables ................................................................................................................................ vi List of Figures .............................................................................................................................. vii Acknowledgements .................................................................................................................... viii Chapter 1: Introduction ............................................................................................................... 1 1.1 Historical Background ................................................................................................ 1 1.2 Problem Statement ...................................................................................................... 3 1.3 Research Question ...................................................................................................... 6 Chapter 2: Decomposing the Trade-Environment Nexus ......................................................... 7 2.1 The Composition Effect .............................................................................................. 7 2.1.1 Inherent Environmental Richness ........................................................................... 7 2.1.2 Combined Cost Advantage: Capital Abundance .................................................... 9 2.1.3 Devaluation of Environmental Cost: Lax Regulation or Eco-Dumping ............... 11 2.2 The Technique Effect ................................................................................................ 13 2.2.1 Upward Convergence: California Effect Hypothesis ........................................... 13 2.2.2 Income Effect and the Environmental Kutznets Curve Hypothesis ..................... 14 2.2.2.1 Income induced Regulatory Stringency and Porter Hypothesis ................... 16 2.2.2.2 Income Induced Structural Change ............................................................... 19 v  2.3 The Scale Effect ........................................................................................................ 23 2.4 Sub-Conclusion: Lessons for the Empirical Analysis .............................................. 26 Chapter 3: Empirical Analysis .................................................................................................. 31 3.1 Hypothesis................................................................................................................. 31 3.2 Empirical Strategy .................................................................................................... 32 3.3 Data ........................................................................................................................... 35 3.3.1 Dependent Variables ............................................................................................. 35 3.3.1.1 Environmental Cost ...................................................................................... 35 3.3.1.2 Proxies for the Scale, Composition and Technique Effects .......................... 38 3.3.2 Explanatory Variables ........................................................................................... 42 3.3.2.1 Trade Intensity .............................................................................................. 42 3.3.2.2 Proxies for the Scale, Composition and Technique Effects .......................... 43 3.3.2.3 Other Independent Variables ........................................................................ 43 3.4 Estimation models ..................................................................................................... 44 3.4.1 Filtered-Through Effects Estimation Models (Intersection of Intersections) ....... 45 3.4.2 Compound Models ................................................................................................ 47 3.5 Results ....................................................................................................................... 47 3.5.1 Hypothesis1) Test Results..................................................................................... 48 3.5.2 Hypothesis2) Test Results..................................................................................... 54 Chapter 4: Conclusion ................................................................................................................ 62 Bibliography .................................................................................................................................66 Appendices ....................................................................................................................................72  vi  List of Tables  Table 3.1 Environmental cost of Trade intensity (Full sample) ................................................... 49 Table 3.2 Environmental cost of Trade intensity (by income group) ........................................... 51 Table 3.3 Income-Environment relationship ................................................................................ 52 Table 3.4 Trade-income relationship ............................................................................................ 54 Table 3.5 Inequality-Environment relationship ............................................................................ 55 Table 3.6 Trade-income equality relationship .............................................................................. 56 Table 3.7 Full sample analysis result ............................................................................................ 57 Table 3.8 Income group analysis result ........................................................................................ 58  vii  List of Figures  Figure 2.1  Income effect on Technological change ..................................................................... 17 Figure 2.2  The ACT model .......................................................................................................... 24 Figure 2.3  Dynamics of the trade and environment relationship ................................................. 30 Figure 3.1  Two tracks of trade's impact on the environment ....................................................... 33 Figure 3.2  Intersection of intersections estimation vs. compound model .................................... 34 Figure 3.3  Environmental cost variables ..................................................................................... 36 Figure 3.4  GDP, Consumption of Fixed Capital and Per Capita GNI ......................................... 40 Figure 3.5  Trade Intensity ............................................................................................................ 42           viii  Acknowledgements  Writing this thesis has been an exciting journey of academic exploration which could not have been done without the support and encouragement from many. Foremost, I thank my supervisor, Prof. Milind Kandlikar for his consistent generosity, understanding and support. Also, my gratitude goes to Prof. Sumeet Gulati for having provided me his expertise as a committee member, and Prof. Hadi Dowlatabadi for his truthful encouragement and constructive comments. And I would like to thank the World Bank’s development data group for their prompt response to whatever inquiries I have regarding data compilation. Last but not least, my special thanks are owed to my parents who flew all the way from Korea to Vancouver to assist me taking care of my little ones, and my beloved husband for having accompanied me throughout the journey even during the hardest time of his life. I am indebted to you all, and wish you all the best. 1  Chapter 1: Introduction  1.1 Historical Background The phenomenal expansion of free trade in the postwar era1 is established upon the economic theory that an open economy can generate more material benefits than autarky. The reason is explained that countries with varying characteristics can specialize in producing goods in which they have comparative advantages and improve social welfare by exchanging the surplus products. This classic theoretical rational for trade liberalization, however, does not adequately account for negative externalities associated with the increased economic activities. Naturally, amid the rising concerns over environmental degradation and the emergence of “low or no growth” philosophy in the 1970s, the environmental consequences of expanding free trade started being questioned.  Prior to the 1972 UN Conference on the Human Environment in Stockholm, the Secretariat of the GATT (General Agreement on Tariffs and Trade) was asked to make a contribution and accordingly prepared a paper 2  focusing on the implications of environmental regulations on international trade. This study was the first of its kind conducted by the GATT Secretariat to probe into the relationship between trade and environment and the standpoint was, in short, ‘concern over green protectionism’; and this shows that trade officials in the 1970’s were oriented toward free trade than ‘environmental sovereignty’. The standpoint is again reflected in the Stockholm                                                  1 WTO (World Trade Organization) International Trade Statistics 2015: from 1950 to 2014, international trade increased 180 times by value and 34 times by volume of export.  2 Industrial Pollution Control and International Trade 2  declaration which found the cause of environmental problems from impoverishment (or under-development) and thus requested countries to “direct their efforts to development…to improve the environment.”     It was not until 1981 when GATT Article XX was referred to by a defendant to seek for the legitimacy of its trade restriction which was taken for the purpose of environmental protection. In the early 1980’s,  the word “sustainability” came to be used and within considerably short time, evolved from a concept to an influential idea in the international debate over the future (Kidd, 1992). Building on the atmosphere, twenty years after the Stockholm conference, GATT secretariat published a special report on “trade and the environment”. The study was conducted as a preparatory work for the 1992 Earth Summit and differently from the previous stance, the Secretariat made it explicit that “GATT rules place essentially no constraints on the ability of countries to use appropriate policies to protect their environment” and assured that trade and environment are ‘mutually supportive’. Yet, the report maintained the former anti-protectionism stance with regard to unilateral trade measures taken for environmental protection purposes and warned that the imposition of special duties against trading partners’ presumably dirtier production would risk “an eventual descent into chaotic trade conditions”.  Along the same line, during the Uruguay Round (1986-1993), the issue of trade related environmental problems were taken up and addressed in the WTO main pillar agreements. But the incorporation of environmental concerns was limited to the extent that the principal rules of trade liberalization are not tarnished by the environmental agenda. For example, SPS (Sanitary and Phyto-Sanitary) agreement explicitly allows for member countries to adopt SPS measures to 3  protect human, animal or plant life or health under the condition that the measures do not arbitrarily discriminate between member countries and attempts to introduce higher level of protection than provided by international standards are to be scientifically justified or scrutinized based on the area specific risk assessment.  Nonetheless, the outcome of Uruguay round never satisfied those who were concerned about adverse environmental impacts of expanding international trade. The core of the criticism was that the international trade regime undermines sovereign rights to resort to whatever policies deemed appropriate to protect their environment; that is, the WTO rule prohibits individual countries to take to trade sanctions against lower environmental standards in other countries (Earnest H., 1995). Interestingly, the rationale of the criticism is essentially based on the same economic theory asserted by trade apologists; that is, the lax regulation lowers the production cost and thus provides comparative advantage over foreign competitors who are burdened with more stringent regulation. What the environmentalists feared is that regulatory competition pursuing the ‘lax regulation advantage’ may provoke a global ‘race to the bottom’.   1.2 Problem Statement Without doubt, the competitiveness concern almost always holds as a strong rationale to relax existing environmental policies or to impede the adoption of new/or more stringent regulations. The problem is, in spite of the theoretical plausibility, neither concrete theory nor coherent empirical evidences have been found to back up the concern. Even if positive correlation is to be found between laxer regulation and price competitiveness, this does not directly support a conclusion that environmental regulation diminishes industrial competitiveness. This is because 4  there is a potential of reverse causality that lax regulation is imposed for the sake of protecting internationally competitive domestic industries; labor concentration in a few comparatively advantageous industries could have increased political pressure to eliminate regulatory burdens on those industries as a way of boosting domestic economies.   Likewise, the other side argument, or trade optimism, is neither established upon concrete evidences; it is not obvious whether trade induced economic prosperity would lead to increase the resource allocation for environmental improvements. The GATT’s argument (1992), that the environmental and economic goals are mutually supportive, is based on the widely accepted assumption that environmental quality is a normal good; as income increases, economic agents become more willing to spend on cleaner environment.  Here as well, the existence of positive correlation between increasing income and expenditure on environment does not necessarily mean that growing economies create better environment. Obviously, what matters is the relative size of environmental investment to the caused degradation; without allocating good enough resources to clean up the mess which is engendered in the process of creating the (material) wealth, trade would generate more environmental problems than it solves.  A related concept is ‘Environmental Kutznets Curve (EKC)’; the EKC hypothesis describes how the size of economic activity and production efficiency interacts over the phases of economic development to suggest that structural changes (higher proportion of service and information industry) and technological innovation prevail over the scale impact over an income threshold. Empirical studies of the EKC hypothesis flourished in the 1990’s but the results are inconsistent due to different assumptions, estimation methods and consulted dataset (Stern et al., 1996). A point 5  to be noted is that the EKC hypothesis presumes that the structural change or reduced proportion of manufacturing is to take place in high income countries; or conversely, high income is regarded as the source of the less environmentally intensive structure of economy. In a closed system, if a standard EKC is observed, it would be less suspected that the income effect induced policy responses and technological changes are the main drivers of the downward portion of the EKC. However, in an open economy, the seemingly less pollution intensive output system in high income countries could have resulted from outsourcing environmentally intensive sectors to less developed countries where producers bear relatively cheaper cost of polluting activities. If so, even if trade liberalization helps to reduce pollution in high income countries, total global emission could increase due to environmentally unsound production methods employed by those less developed countries; and, the EKC trend in traditionally rich countries will not be repeated in developing countries because they would not have poorer countries to export dirty industries to.   To summarize, the core of the trade and environment problem lies in the ambiguity of environmental policy directions which are to be adjusted as a response to the trade liberalization and the resultant changes of politico-economic conditions. And the ambiguity is spurred by the uncertainty about the degree of income elasticity of environmental demand, and also by the absence of a clear evidence that regulatory stringency hampers domestic industry’s international competitiveness. Both pessimistic and optimistic hypothesis are theoretically plausible, but none is definitive and thus provoked a vast empirical research. Yet, the interplay of various associated factors have made it difficult to conclude that trade is either good or bad for the environment and the question is still up for debate.  6  1.3 Research Question I regard this paper as a stepping stone to answer the ultimate question of how to improve the current trade and environmental regimes at both international and domestic levels to achieve the best balance between developmental and environmental goals. Related questions include i) what is the most agreeable balance between the two fundamental goals, ii) how to maintain individual countries’ sovereign right amid increasing pressure to act harmoniously to tackle global challenges.  For sure, these questions are far from being readily answerable, especially due to the ethical and philosophical aspects being involved. I presume this is why many scholars have been attracted to the question of whether trade is good or bad for the environment because if only trade is proven to be beneficial for the environment, it will be literally ‘problem solved’; put otherwise, environmental concerns can no longer be the reason to oppose growing economic interdependency nor the best trade-off needs to be sought for as long as the two agendas (trade and environment) are in a synergetic relationship.  To get to the point, I strongly doubt that any economic gains of trade could come at no environmental cost. To resolve the doubt, I attempt to explore whether the trade induced income effect, the frequently held rationale for supporting trade liberalization, is in truth bringing more environmental benefits than harms (which is caused in the process of increasing the income). And together with the income channel of trade’s impact, other routes of trade gains and losses will be jointly investigated and compared.  7  Chapter 2: Decomposing the Trade-Environment Nexus  Trade economists developed a structural framework to decompose environmental impacts of trade. The framework, first applied by Kruger and Grossman (1991), separates trade induced environmental impacts into scale, composition and technique effects; the scale effect refers to the increased economic activities as a result of opening up to trade, the composition effect refers to the changes in the mix of production in response to the relative price in the global market, and the technique effect relates to the accelerated innovation of environmentally sound technology and its diffusion through increased interconnectedness between countries. In this chapter, I will further break down each of these three channels of trade’s impact on the environment.  2.1  The Composition Effect Opening trade introduces new relative prices of commodities. The new price signal causes production factors to move from less gainful industries to globally competitive sectors. If a country has a comparative advantage in environmentally intensive sectors, trade liberalization may result in increased pollution for that country. Therefore, the direction of the composition effect hinges on ‘in what sector a country has the ability to produce at lower opportunity cost than its competitors’ and ‘whether the sector is more environmentally intensive than the weighted average environmental intensity of production in autarky’.   2.1.1  Inherent Environmental Richness Intuitively, the most probable reason why it is cheaper to pollute or extract natural resources in some countries than others is the inherent richness of environmental assets. More specifically, if a 8  country is endowed with abundant natural resources, huge per capita land size, or geographically advantageous location, it is highly likely that the country could enjoy relatively lower environmental damages caused by economic activities. Earlier works also paid due attention to the inherent environmental conditions. For example, Krueger and Grossman (1991) controlled for site specific effects by including geographic location and population density (captured by the vicinity to central city) dummy variables and found that “more densely populated cities suffer greater concentrations of sulphur dioxide, all else equal” and “a location on a coast reduces a city’s concentration of pollutants” probably due to the disbursement properties of the local atmosphere.  A decade later, Antweiler et al. (2001) examined how open trade affects the concentration of local sulfur dioxide; data was drawn from the Global Environment Monitoring System (GEMS) and the observations were collected from more than 290 sites in 43 countries over the period from 1971 to 1996. In this research, inherent environmental characteristics were captured by average temperature and precipitation variation; and the authors found that “higher temperatures both dissipate pollution faster and reduce the need for home heating; precipitation highly concentrated in one season reduces the ability of rain to wash out concentrations”.  More recently, Frankel and Rose (2003) conducted a more comprehensive research in the respect that various environmental indicators were explored as independent variables; the indicators include deforestation, energy depletion, water access, and carbon emission. The authors used, instead of panel data, cross-sectional data and addressed the endogeneity problem by using geographical distance between countries as an instrumental variable. In this study, per capita land 9  size was included as a controlled variable and it was found that “population density has an adverse effect on concentration pollutants”.  2.1.2 Combined Cost Advantage: Capital Abundance Inherent environmental richness is certainly a factor which could reduce the unit cost of environmental input (or may justify less stringent regulations). However, it should be noted that what we conveniently call the environmental input is in fact not an input but usually an unwanted output of economic activities. Therefore, even if it takes comparatively low effort to clean up the waste, original input of the production factors could be quite costly. And if the combined cost of production input and charges on the unwanted output is more expensive in one country than others, the country would not be able to produce the ‘environment embedded’ goods at lower cost than its competitors. Put otherwise, because production of tradable goods necessitate inputs other than environment, the cost of environmental input should be considered in combination with other input prices. In the end, countries with the “combined cost advantage” would end up exporting environmentally intensive goods.  In the respect that capital intensive production is deemed to be environmentally intensive, capital abundant countries are likely to enjoy the ‘combined cost advantage’. A plausible inference is that a country which are abundant in both capital and environment is likely to experience the highest degree of trade induced environmental degradation. 3  Meanwhile, even if a country’s                                                  3 If we allow for capital mobility between countries, labor abundant countries will be apt to experience higher degree of environmental degradation because capital will move to where the return on investment can be maximized, put differently where capital per capita (or wage) is the lowest; resultantly, the labor/capital ratio will 10  environmental asset is relatively poor, the country could end up exporting products with fairly large environmental footprints when the capital abundance in the country prevails over the environmental scarcity and thus could create relatively low ‘combined production cost’. Overall, liberalized trade would increase capital abundant country’s opportunity cost of ‘not polluting’; or conversely, free trade lowers the opportunity cost of producing ‘environmentally intensive’ goods in capital abundant countries.   In their empirical research, Krueger and Grossman (1991) examined how the share of capital in the manufacturing input affects the trade pattern of a country; they looked at the share of imports from Mexico to US in 1987. The result was, the bigger the share of capital input, the less Mexico exports to the US. The finding was statistically significant and the authors report that the coefficients denoted 0.52% points and 0.24% points decrease per 10% increase in the share of human and physical capital respectively. Although the research of Krueger and Grossman meaningfully highlighted the role of capital abundance in determining the pattern of trade, the correlation they found between the factor intensity and the composition effect was quite weak because they did not disentangle the effect of ‘environment abundance’ from ‘capital abundance’; an industrialized country like US is very likely to be relatively deficient in environmental assets and therefore, the impact of capital abundance must have been undone by the environmental scarcity.                                                    converge and the highly populated countries will end up having absolutely abundant production factors, increased economic activities and consequently damaged environment.   11  In this regard, Antweiler et al. (2001), quite successfully identified the impact of factor endowment by isolating the capital abundance from environment abundance related factors (e.g., population density and climate). The result shows that the relative capital abundance is a powerful determinant of comparative advantage; they found “a positive composition effect arising from an increase in capital-to-labor ratios”. Also, they confirmed a strong link exists between capital abundance and trade induced pollution; “a 1 percent increase in a nation’s capital-to-labor ratio- holding scale, income, and other determinants constant- leads to perhaps a 1-percent-point increase in pollution”.  2.1.3 Devaluation of Environmental Cost: Lax Regulation or Eco-Dumping One of the most contentious issues related to the environmental impact of trade is whether regulatory stringency determines or affects comparative advantages. Both race-to-the-bottom hypothesis and Pollution-Haven-hypothesis (PHH) presume that the cross-national differences in the regulatory stringency can be critical enough to alter trade patterns. If so, benefits from free trade can be undermined when governments intentionally lower environmental standards to achieve economic policy goals and distort the level playing field; when countries have economic motives to use lax legislation as a tool to improve their terms of trade, eco-dumping could occur and cause world-level inefficiencies of resource use. For example, Chilcilinisky (1994) found that underpriced environmental resources cause overconsumption of the resources not only by the locals but also in the international market (at prices below social costs); in her study, the effectiveness of property rights was used as a proxy for the regulatory differences between countries.   12  However, quantifying regulatory stringency is definitely not an easy task and in whatever ways it is measured, the method is destined to be contested. Very recently, Brunel and Levinson (2016) made an attempt to quantify the regulatory stringency only to find the vast difficulty of the task; the authors classified and examined five categories of earlier quantification approaches and concluded all the five categories fall short of an ideal measure (“ a single overarching measure of multidimensional stringency”). 4  That said, a study by Cole and Elliott (2003)5 is noteworthy in the respect that per capita income is used as a proxy for the regulatory stringency on the ground that the two variables are strongly and significantly correlated; the authors estimated an index of 60 different countries’ environmental regulations and found “a statistically significant correlation between per capita income and the index of regulations of 0.88”.6 Then, the authors found that the environmental effects of environmental regulation works in opposite direction to that of capital abundance and argued that the cancelling out effect of the two different sources of the composition effect partially explains why many preceding studies failed to find a clear evidence of the PHH.                                                    4  The five categories are i) private sector abatement costs, ii) individual regulation targets, iii) a jurisdiction’s composite index of regulatory stringency (e.g., counts of environmental regulations and international treaties signed and etc.), iv) pollution levels as an evidence of stringency, v) measures based on public sector expenditure.  5 Cole and Elliotte(2003) explored the interaction of capital abundance and regulatory stringency in determining the composition effect by exploiting cross-country panel data. This empirical work was built yon Antweiler et al. (2001)’s theoretical model to distinguish the effects of environmental regulation and capital endowment.  6 Although Antweiler et al. (2001) also presumed that changes in per capita income will lead to regulatory responses, they did not empirically confirmed this presumption.  13  2.2 The Technique Effect Technique effect basically refers to the pollution reduction effect per unit production via technological changes. And trade liberalization could induce technological innovation through both direct and indirect channels. The direct channels are through embedded technology in traded goods or technology spillover effect via foreign direct investment. In the meantime, indirect technique effect arises from trade induced income effect as increased income translates into strict environmental regulation and correspondingly, more clean production technologies.   2.2.1 Upward Convergence: California Effect Hypothesis ‘California effect’ is a term invoked by Vogel (1995) to describe how economic integration could create “race-to-the-top” phenomenon; the term is named after the American state that often took the role of a pacesetter for the national environmental regulations in the US. Vogel (1997) confirmed that “there is substantial evidence for a ‘California effect’ and presented several routes how the upwards harmonization could occur; i) if a richer, greener country forces foreign producers to meet its stricter regulatory standards, those producers will make necessary investments to maintain their market access to the richer country and having made the initial investments, the producers would then encourage their home country to strengthen environmental regulations to gain competitive advantage over their home competitors knowing that their products already met stringent standards; or, ii) market access is denied to products from countries with less strict standards to motivate the laxer regulating countries to upgrade their national standards; iii) in some cases, even without extra-territorial enforcement of stringent standards, self-interested developing countries would voluntarily strengthen their regulatory stringency to increase their 14  market access to richer, green countries; iv) as well, increasing number of multinational environmental agreements could be facilitated to disseminate high standards.  Despite the compelling hook of this theoretical optimism, ‘California effect hypothesis’ has gained less attention by researchers than its competitive race-to-the-bottom hypothesis. Although some qualitative studies (Vogel, 1995; Tewari and Pillai 2005) support the ‘upwards-convergence’ hypothesis, only recently attempts have been made to look for the empirical evidence based on quantitative methods. Among them, Perkins and Neumayer (2012) seem to have most delicately dealt with the issue in that the authors accounted for spatial autocorrelation which is highly likely to exist among nearby trading partners and that the income channel of regulatory stringency was explicitly controlled for. I would not say that the study by Perkins and Neumayer provide a generalizable evidence for mutually reinforcing relationship between trade and regulatory stringency because their study is intentionally limited in the scope only to examine the impact on automobile sector which is assumed to be particularly susceptible to product specifications in foreign markets. Still, the finding (positive correlation between the regulatory stringency and the volume of trade) provides an important insight regarding under what conditions ‘California effect’ operates; the existence of substantial market incentives (access to major, richer market), more conducive to product standards than overall national environmental standards.  2.2.2 Income Effect and the Environmental Kutznets Curve Hypothesis Whereas the positive spillover effect is less frequently discussed, income effect is a central counter-weight to arguments that trade liberalization incurs downward convergence in environmental standards. The frequently quoted rationale is that as income increases, will rise the 15  demand for improved environment and thus will more resources be allocated for environmental investment. The EKC concept was first introduced in the path-breaking work by Krueger and Grossman (1991) and has been widely pondered ever since.   As a part of wider investigation into the environmental impacts of NAFTA, Kruger and Grossman (1991) identified an income threshold beyond which higher level of income yields decreased level of air pollution (sulfur dioxide and particulates). To cite Krueger and Grossman, the inverted U-shaped relationship between income and pollution results from the changes in the relative magnitude of the scale and the technique effects; and the prevalence of technique effect after a turning point was asserted to be helped by trade liberalization for at least two reasons.  First, foreign producers may transfer modern technologies to the local economy …. Second, and perhaps more importantly, if trade liberalization generates an increase in income levels, then the body politic may demand a cleaner environment as an expression of their increased national wealth. Thus, more stringent pollution standards and stricter enforcement of existing laws may be a natural political response to economic growth. (Kruger and Grossman, 1991)  Note that the same income effect was referred to as a source of the composition effect in the previous section (2.1.3). Very plausibly, income induced stricter regulation would increase firms’ cost of production and the producers may adjust to the shock either by changing production (or pollution management) technologies or by transferring the cost to the final product. If the affected industry is exposed to trade, the adjustment process will surely damage the industry’s global 16  competitiveness unless cost-saving technological breakthrough takes place. That is, the income effect could be either translated into the technological change or the loss of competitiveness; the former case corresponds to the technique effect while the latter relates to the composition effect.  In reality, it is expected that both channels are in operation and if the technological change is not big enough to cancel out the increased cost of environmental input (compliance cost), pollution haven effect might occur to divert domestic production factors from environment intensive sectors to less intensive sectors or to abroad. Therefore, even if negative correlation between income growth and reduced environmental impact is observed, it is hard to tell whether the link comes from the composition effect or the genuine technique effect. And it would be misleading to interpret the result as ‘growing technique effect with income’ or use the result as an evidence for the inverted U-shaped EKC hypothesis.  2.2.2.1 Income induced Regulatory Stringency and Porter Hypothesis For regulatory stringency to translate into the technique effect, at least four causal chains should be found to exist in sequence; i) trade openness enhances  income growth, ii) environment is a normal good (higher income results in increased demand for clean environment), iii) a government effectively responds to the public interest (increased demand for clean environment results in regulatory stringency), iv) producers respond to the increased compliance cost by reducing environmental input per unit product (technological innovation) rather than switching to less environmentally intensive sectors or relocating to less stringent jurisdictions.    17  Figure 2.1. Income effect on Technological change  Components ➀  ➁  ➂  ➃  ➄   Trade liberalization →    Income growth →   Demand increase for clean environment →    Stringent regulation →    Technological change           Links  1  2  3  4   In most empirical works under the rubric of trade-environment relationship, the first two links are usually presumed to exist. A limited number of studies conduct a combined analysis of the second and third links to see how per capita income is correlated with regulatory stringency (Cole and Elliott, 2003; Esty and Porter, 2005). Yet another group of studies jointly examine the third and fourth links to find the correlation between the income and environmental output; for example, Antweiler et al. (2001) distinguished between Communist and non-Communist countries and gained the result that increased income rarely translates into stricter environmental policies in Communist countries. The result is reinforced by Frankel and Rose (2005) to find that “low-democracy countries tend to have higher SO2 pollution”.  From the results, it is naturally inferable that polity matters for income effect to operate; put otherwise, if a government promptly responds to the changing demand for environment, devaluation of environmental assets is less likely to occur and therefore, producers are more likely to adjust to the enhanced standards by reducing environmental input per production.   Whilst these studies provide valuable insights, the importance of the fourth link is somehow disregarded and conveniently presumed that rigorous regulation might well spur technological changes. Related to this issue, as early as 1991, Michael E. Porter suggested a hypothesis that “strict environmental codes may actually foster competitiveness”, and argued that “the conflict 18  between environmental protection and economic competitiveness is a false dichotomy” and “tough standards trigger innovation”. In the follow-up paper, Porter and Linde (1995) specified the reasons why environmental regulation could stimulate innovation; i) by signaling the potential area of technological improvement, ii) by raising corporate awareness, iii) through the reduction of uncertainties associated with environmental investment, iv) by creating outside pressure to overcome inefficient managerial inertia, and v) by removing cost disadvantage inflicted on firms making environmental investments. Porter and Esty (2005) then moved on to empirically test the hypothesis based on Environmental Regulatory Regime Index (ERRI) and World Economic Forum’s Current Competitiveness Index (CCI) to find a strong correlation between environmental regulation and competitiveness of domestic industry. A caveat to this finding is, as was noted by the authors, that the correlation does not reveal causation and thus only supports the “soft version” of the Porter hypothesis.   According to Jaffe and Palmer (1997), Porter hypothesis can be categorized into three versions depending on the specification of explanatory and dependent variables; the first set of studies, which represents the “weak” version of Porter hypothesis, tests for the relationship between regulation and innovation usually by examining pollution abatement costs and R&D expenditure (or the number of successfully applied patents). On the other hand, “strong” version of the hypothesis predicts that regulation does increase entrepreneur profits; empirical assessment mostly involves regulatory stringency indicators and productivity changes (ratio of outputs to inputs) over time; therefore, it will be admirable to have time series data spanning over regulatory stringency changes. Early studies of this strand, however, did not give due attention to time variables and revealed negative relationship between compliance cost and firms’ productivity (Smith and Sims, 19  1985; Gray, 1987; Dufour et al., 1998). It was not until when the time lagged changes of productivity was considered that a positive and significant correlation was found by Lanoie et al. (2001); in this study, the dynamic aspect of Porter hypothesis was explored by using time lagged regulation variables and total factor productivity growth of the manufacturing sector in Quebec; and the result suggested that, contrary to the negative contemporaneous impact of regulation on productivity, “the opposite result is observed”.  Meanwhile, ‘narrow’ version of Porter hypothesis denotes that technological innovation is spurred particularly by market based (or flexible) instruments. On this, due to the paucity of data, it is hard to come across with preceding studies focused on the quality of regulation. Quite recently, Lanoie et al. (2008) explored OECD survey data with a large number of observations (4,200 facilities in seven OECD countries) to examine the “narrow” version as a part of their study on “the whole Porter causality chain”; from regulation to innovation and finally, the consequential environmental and business performances. And the authors found that “performance standards (tax policies) are more likely to induce innovation than more prescriptive technology-based standards” and explained the reason that “perceived stringency of the performance standards has a more important impact that that of the technology based standard”.   2.2.2.2 Income Induced Structural Change Income level is often presumed to cause regulatory changes but it is rarely questioned whether the income effect is valid for different types of environmental impacts. If endogenous policy response is an important factor to bring about the technique effect, it is hard to expect that the technique would be valid for hardly perceptible pollution types within jurisdiction and by contemporaneous 20  people. Put simply, individual countries do not have any politically acceptable reason to address global externalities such as greenhouse gases; marginal damage of which is relatively horizontal at a specific time point. Thus, when an inverted U-shaped relationship between income and carbon dioxide were found (Panayotou et al., 1999), Panayotou, Sachs and Peterson (2000) questioned the validity of “ the behavioral explanation of the EKC: as income rises, the effective demand for environmental quality rises and eventually overwhelms any scale effects of economic growth on pollution”. To answer this question, Panayotou et al. hypothesized that structural change is the underlying mechanism for EKC; if so, the inverted U shaped EKC pattern should be monotonously observed across different kinds of pollutants, even if the adverse impact of emission is not felt within the jurisdiction. Panayotou et al. tested the hypothesis by exploring a unique time series data that spans over 125 years (from 1870 to 1994); the observations include emission of carbon dioxide, non-residential capital stock, population density, and income of six industrialized countries.   By using the unique data set, Panatoyou and his colleagues asserted that the ‘structural change hypothesis’ could be tested without being challenged by the issue of fundamental heterogeneity between countries; they were concerned that cross-sectional data may capture positive income-pollution relationship in developing countries and negative correlation in developed countries and the cross-country differences could be misleadingly interpreted to support the EKC hypothesis. And Panayotou et al. used non-residential capital stock (non-residential structure, equipment and machinery) per unit of GDP as the structural change variable and controlled for trade, population and income variables. The finding was; emission rises in the phase of industrialization and falls in 21  the post-industrial stage. And the authors concluded that pollution haven effect exists but not to the extent to “explain away the EKC for CO2 emissions.”  In the respect that the importance of structural change was shed light on, the research by Panayotou et al. is very worthwhile. Yet, the potential interaction between income growth and the structural change was ignored in this study and it was simply presumed that increasing proportion of capital stock reflects technological change and the simultaneous shifts in the structure of an economy. But accumulation of physical capital stock alone is not a sufficient surrogate for the structural change. Historically, we have observed that industrialization and economic growth is strongly related to the increase of human capital, often represented by literacy rate and average years of schooling. If trade induced income growth causes human capital accumulation, thus created increase of educated populace (or skilled workforce) may change an existing pattern of comparative advantage to boost technology (or information) intensive sectors; which are mostly less environmentally intensive. Then, the EKC pattern obtained by Panayotou et al. (2000) might have to attribute more merits to trade openness as an endogenous growth engine that eventually contributes to the downward trend of environmental impacts.  Of course, the direction of causality between income growth and human capital accumulation is not clear-cut. If reverse causality holds; i.e., human capital accumulation causes higher income; international trade may speed up global income divide because opening to trade theoretically brings more reward to human capital intensive sectors in human capital abundant countries and thus accelerates further accumulation of human capital in those countries whilst the other way around for human capital scarce countries; trade could take the chances of structural change away 22  from the countries which are abundant in unskilled labor or raw materials because these countries will gain more short-term benefits by specializing in the sectors using their abundant factors intensively.  In terms of trade related environmental consequences, the former causality poses a greater optimism; if income leads to environmentally friendly structural change, trade induced income growth can replicate the process in currently less developed countries. Meanwhile, the latter causality inevitably tilts toward the pollution haven hypothesis because trade will induce human capital abundant countries with already high potential of prosperity to be exposed to greater chances of further human capital accumulation (and deeper specialization in sectors with relatively small environmental impacts) while the opposite holds for countries with scarce human capital; then, trade could inflict ‘specialization trap’ and unsustainable path of growth for the poor countries. In this case, structural change propelled by trade and the positive environmental consequences in some countries cannot be extrapolated to other countries.   Theoretically, both causality directions are plausible. It is very likely that human capital generates innovations and stimulates growth and at the same time, economic growth would have positive effect on human capital accumulation as material prosperity would allow for the increase of educational investment. Empirical research outcomes on this subject are mixed. For example, Mehrara and Musai (2013) investigated “the causal relationship between education and GDP in developing countries” for the period of 1970-2010 and found statistically significant evidence that education leads to GDP growth, not vice versa. On the other hand, Ciccone and Papaioannou (2007) found that initial education level of a country is an important determinant of the pattern of 23  international specialization and the “subsequent shifts of the production structure towards schooling-intensive industries”. That said, in the following empirical research, the stock of human capital will not be separately controlled for because whether and how trade will interact with the human capital accumulation remains uncertain.  2.3 The Scale Effect Following from the previous section, trade induced income growth could be accumulated in the form of fixed capital (and human capital) may well lead to the scale increase of economic activities. However, this dynamic path of scale effect is usually unaccounted for in previous research. Turning back to Krueger and Grossman (1991), it seems clear that the term “scale effect” is used to reflect the concurrent trade’s impact on the path of production (and consumption).  There is a scale effect, capturing the simple intuition espoused by the environmental advocates. That is, if trade and investment liberalization causes an expansion of economic activity, and if the nature of that activity remains unchanged, then the total amount of pollution generated must increase.  To review this paragraph in the context of the Heckscher-Ohlin (H-O) trade theorem, concurrent expansion of economic activity via opening to trade is only possible when production factors are less than fully employed under autarky and the inefficiency is corrected for by introducing free trade for whatever reasons. In the existing literature, the reason why trade openness corrects for the autarky inefficiency (or misallocation of resources) has not been adequately questioned. Rather, theoretical research on the trade-environment relationship has uncritically taken the static 24  H-O trade model and its assumption of full employment; for example, Anteweiler et al. (2001) illustrated the environmental impact of trade based on the static H-O trade model and described 𝐵𝐶 in the graph below as the source of the scale effect (hereafter I will call this model the ACT7 model taking after the first letters of authors’ sir names).8    Figure 2.2. The ACT model (Source: Antweiler et al., 2001)                                                  7 Antweiler, Copeland and Taylor  8 The ACT model is mostly built on the H-O model; it presumes a small open economy (which cannot influence the world relative price) which produces two goods (X: dirty good, Y: clean good). Also, other key assumptions of H-O theorem are adopted; i) complete production factor mobility between industries, ii) zero transaction cost, and iii) full employment of production factors. One different feature is that the ACT model limited to production side whereas the H-O model depicts welfare increase effect from expanded purchasing power (consumption side); this is so because the main purpose of ACT model is to describe how trade induced production change brings about any impacts on the environment.  25  To be looked at more closely, 𝐵𝐶 does not correspond to the definition of the scale effect. By definition, the scale effect refers to a linear and outward expansion of economic activities without “the nature of the activity” being changed. And for sure, 𝑂𝐴 and 𝑂𝐵 are not on the linear path of expansion; the composition of output alters as production moves from A to B. Point B, in fact, is a hypothetical point of production which is retrospectively marked; once exposed to trade, under the assumption of perfect factor mobility and complete employment, the equilibrium is supposed to move directly from A to C, not through B. Also, the scale effect is supposed to be counted by the quantity of production, not by its value because the environmental consequence of production is much more closely related to the amount of resource input per unit production than its market price. But in the ACT model, A and B are placed on the same iso-value line, not an iso-quant line.  That said, how would trade contribute to the increase of economic activity? On this, Krueger and Grossman (2001) gave an example of increased demand for cross-border transportation and the consequential increase of air pollution unless trucking practices change. For sure, unlike the standard H-O model assumption, transaction cost exists and the associated economic activity will grow proportionately to the increase of trucking and bartering.    Other possible routes of the scale effects are also related to the violation of the H-O assumptions. Standard H-O model assumes perfectly competitive market but if, for whatever reasons, the condition of perfect competition is removed, resource misallocation may occur to position point A (autarky equilibrium) beneath the Production Possibility Curve (PPC). Then, competition spurred by being exposed to world market could enhance production efficiency of domestic industries to result in outward expansion of production possibility (or repositioning the production point back 26  on the PPC). To exemplify, restricted factor mobility may increase the cost of production and thus reduce the size of production; suppose that production of Y requires relatively highly skilled labor and thus the labor union of Y industry has advantageous negotiating power to significantly increase wages, then Y will be produced less than its optimal level under competitive labor market.   Whereas the relaxation of the ‘perfect competition’ assumption relates to the realization of the status-quo potential of growth, allowing for the inter-temporal dynamics of factor endowment changes could present yet another source of the scale effect, As aforementioned briefly, increased income via trade could be accumulated in the form of capital stock. Then, the capital stock increase would shift the economy’s PPC outward. Of course, if the income is thoroughly consumed or a country’s economy depends on the dissipation of exhaustible resources (particularly non-renewable resources), growing trade could result in the shrink of the economy’s future PPC.  2.4 Sub-Conclusion: Lessons for the Empirical Analysis In summary, except for the transportation increase, trade’s impact on the environment is mostly indirect; trade changes the pattern of production and therefrom environmental consequence arises. First, the direction of composition effect hinges on what sector a country has a combined cost advantage; the combined cost is the sum of ordinary input price for a unit production plus the environmental charge on the waste caused by that unit production. Second, the technique effect is good for the environment but the existence of which is rather ambiguous; the upward-convergence hypothesis is not generally applicable and the income effect induced technology change depends on firms’ response to the environmental standard increment (if technological change does not fully offset the increased compliance cost, there exists the threat of pollution haven effects). Third, the 27  scale effect is, by definition, bad for the environment but except for the transportation increase, it is unclear how trade would scale up the economic activities; to account for the outward PPC expansion, time-lags should be presumed. Therefore, the best guess I can make from the preceding discussion is that environmental impact of trade is theoretically uncertain and the magnitude and directions of the impact would change over time as the environmental impacts affects the combined cost advantage and the PPC of a country.  Given the dynamics of trade and environment relationship, a methodological inference is to be made that a proper surrogate for the environmental impact should be a stock variable (not a flow variable) which measures the accumulated sum of environmental externalities caused by the shock of trade liberalization. The use of a stock variable (e.g., pollution concentration level), however, poses at least three problems. The first is, due to the very nature of dynamics, observed consequences would vary depending on the time point the index is measured. Second, a stock variable such as pollution concentration level is hugely affected by natural conditions; for example, high precipitation rate may substantially reduce ambient air pollution level and thus makes it hard to identify anthropogenic environmental consequences. Third, even if the interference of natural system could be successfully controlled for, the initial shock of trade liberalization cannot be isolated from ensuing influence. Simply put, it is almost impossible to sort out the genuine proportion of trade induced environmental impacts.   That said, an alternative approach is to utilize a flow index (e.g., yearly emission) as a dependent variable to observe a statistically significant correlation between trade openness and the environment. Here again, caution needs to be taken in selecting environmental indicators, of which 28  changes infer social welfare implications; say for example, a great deal of particulate emission, which is instantaneously washed away by natural purification process, would hardly cause welfare losses. Unless a paper is written in preservationists’ point of view or addresses pollution types of which quantitative threshold exists9, environmental impacts, which do not interpret into tangible welfare losses, would not offer any meaningful policy implications. Therefore, in the following chapter, flow variable which counts environmental impacts in terms of monetary value (as a means of converting various environmental impact indicators to a single comparable unit) will be used.  And the agony of calculating lost environmental value is avoided by resorting to (presumably) reliable secondary data; the World Bank database.10  Next, I could identify a gap in the literature that, in most empirical research, the three channels of environmental impact have been considered not as channels but rather as independent factors to be controlled for; and thus, indicators taken as proxies for the scale, composition, and technique effects have been put on the right hand side of estimation models together with the main explanatory variable (trade intensity). Considering that the theoretical framework explains that the environmental impacts of trade occur via changed patterns and sizes of production and                                                  9 In 2009, Rockström et al. quantified planetary boundaries of nine subcategories of earth system and asserted that human has already transgressed at least three planetary thresholds. The planetary boundary hypothesis has, since then, been widely cited and served as a leading framework to think about environmental problems. With regard to the endangered ecological system, however, Nordhaus et al. (2012) dissented even the existence of such thresholds except for four categories of ecological systems; “Real, global biophysical threshold elements exist in the global climate system, and partly also for ocean acidification, ozone depletion, and phosphorous levels. But for all the remaining boundaries, there are no global tipping points.”  10 During the examining process of this thesis, the sincerity of the World Bank’s data and the appropriateness of its estimating methodology were questioned. Although the issue (examining the reliability of the data) is beyond the scope of this paper, I deem it my duty to note the raised questions and the responses from the Word Bank’s data development group (See Footnotes 16 and 17 in p.43, and 19 in p.44). 29  consumption (which results from trade liberalization), it would be more appropriate to incorporate the filtered-through impacts than throwing out the correlated portion of variance in trade intensity with those proxy indicators. Thus, in this paper, along with the standard multiple regression model, the filtered-through effects estimation model will be constructed and tested.  Lastly, it was found that a caution needs to be made when utilizing cross-sectional or panel data when estimating the income-environment relationship. Even though inverted U-shaped relationship is obtained, it does not necessarily mean that higher income would eventually reduce environmental damages because there remains a possibility that the lower pollution in high income countries has not been caused by endogenous technological change (or structural change) but is the result of exporting dirty sectors to less developed countries. To avoid this spurious EKC, in addition to exploiting full sample panel data (consisted of observations from rich and poor countries), I will split countries into four income groups  based on three criteria; i) rich or poor, ii) traditionally rich or not, iii) relatively poor but above or below the income threshold found by Kruger and Grossman ( US$ 4,000-5,000) 11.                                                 11 See Krueger and Grossman (1991) pp5: “we find that ambient levels of both sulphur dioxide and dark matter suspended in the air increase with per capita GDP at low levels of national income, but decrease with per capita GDP at higher levels of income. The turning point comes somewhere between $4,000 and $5,000.” 30    Figure 2.3. Dynamics of trade and environment relationship   31  Chapter 3: Empirical Analysis  In this chapter, I attempt to quantify how strongly trade interacts with the environmental changes by using a global scale time-series data and panel dataset of selectively chosen sixty countries, the observations spanning twenty five years (1990-2014). Analysed dataset is taken from the World Bank database (http://beta.data.worldbank.org) and Stata is used as the statistical software package.  3.1 Hypothesis The theoretical framework of my empirical analysis is largely based on the (capital flow constraint) Heckscher-Ohlin (H-O) model and therefore, resembles in many ways to the ACT model12. Two fundamental differences are that i) assumptions of H-O model are modified and ii) time lags are explicitly considered for the time consumed for trade impacts to translate into the increase of economic activities and technological changes. Modified H-O assumptions are; a) restricted factor mobility between industries and b) different variety of goods produced in different countries. By relaxing the assumptions of perfect mobility and identical production mix, adding income inequality to explanatory variables became logically plausible. 13  In addition, iii) income effect is                                                  12 Assumptions shared are; i) different environmental intensity between products, ii) imperfect specialization even after opening to trade, iii) chances of technological changes, iv) homothetic consumer preferences for different countries, v) no transaction cost, vi) restricted production factor mobility between countries  13 Samuelson (1948) presented “factor price equalization theorem”; in a world where production factors are perfectly mobile (within a country), and where countries produce the same goods (X and Y), production factor prices converge over time across countries. 32  assumed to exist for environmental goods (i.e. environmental asset is a normal good) and iv) environmentally intensive goods are presumed to be capital intensive.  Under these assumptions, hypothesis to be tested are;  1) If income effect is a primary source of trade induced technique effect, trade intensity might lower local externalities (such as ambient air pollution) but not global externalities (such as greenhouse gases) nor the depletion of natural resources. And the aggregate impact (counted in terms of monetary value) will be greater than zero..  2) Under the assumption of restricted factor mobility, trade openness will result in deeper inequality and therefore, will adversely affect the operation of the income effect (translation of income into the technique effect).  3.2 Empirical Strategy To estimate the trade’s effect on the environment, I will construct largely two sets of regression models; i) the first set aims to estimate environmental impacts from changed patterns and scale of economic activities which have resulted from freer trade (filtered-through effect estimation); and ii) the second set of regression models attempt to identify the remaining portion of the trade-environment relationship which is not captured by the first set of models (unfiltered-through effects). Then, the directions and magnitudes of the estimated coefficients will be compared. 33   Figure 3.1. Two tracks of trade’s impact on the environment  The filtered-through effect estimation is theoretically based on the H-O model and refers to the environmental impacts arising from trade induced changes of economic activities and income levels. More specifically, trade’s impact on the scale, composition and the income of the sample economies will be estimated (intersection 1) and in parallel, the link between those three economic indicators and environmental cost variables will be assessed (intersection 2). Then, the estimated results will be multiplied to obtain the intersectional properties (intersection of intersections).  To capture the unfiltered-through effects (especially, technique effect unassociated with the income effect), the three filtering channels (scale, composition and income effects) will be put on the right hand side of multiple regression models (as controlled variables) 14  together with the                                                  14 Multiple regression models use only the uncorrelated portion of variability of the main explanatory variable to other controlled variables; i.e. regress trade intensity on the controlled variables and use the residuals to estimate the coefficient.  34  main explanatory variable, trade openness.  Hereafter, to distinguish from the filtered-through effects estimation model, I will call the second model ‘compound model’.     Figure 3.2. Intersection of intersections estimation vs. Compound model  And the two sets of regression models will be repeatedly run on four different income groups (each income group contains roughly 350 observations of about fifteen countries)15 in addition to the full sample of sixty countries’ data. This income group analysis is done to take into account of income group specific characteristics and thus avoid making hasty generalizations from the full sample results.                                                   15 i) developed countries with more than ten thousand dollar (current US$) per capita income since before the 1980’s, ii) mid-income countries that achieved ten thousand dollar per capita income in the 1980’s and after, iii) low-income countries with less than ten thousand per capita income by now, and iv) the lowest income countries with less than three thousand per capita income by now. 35  3.3 Data 3.3.1 Dependent Variables 3.3.1.1 Environmental Cost Environmental cost indicators to be examined are i) particulate emission damage, ii) natural resource depletion (energy, mineral and net forest depletion), and iii) carbon dioxide emission damage. To refer briefly to the calculation methodology; the particulate emission damage is the weighted average of economic losses (premature death and illness) due to the exposure to ambient particulates (PM2.5) and indoor air pollution.16  Carbon dioxide damage is calculated as the emission unit multiplied by the marginal social cost of CO2 emission which is estimated to be 20$ (in 1995 US$) per ton of carbon.17 Natural resource depletion is the sum of energy (coal, crude oil, and natural gas), mineral (tin, gold, lead, zinc, iron, copper, nickel, silver, bauxite, and phosphate) and net forest depletion; energy and mineral depletion is measured as the ratio of present value of the depleted stock of resources, discounted at 4%, to the remaining reserve lifetime while the net forest depletion is counted as the unit rent times the excess of harvest over natural growth.18                                                  16 (Refer to Footnote 10 in page36) It was questioned whether it is appropriate to equate ‘health damages’ with the ‘lost income’ from forgone labor output. The methodology inevitably leads to underestimation of the health impacts in low income countries compared to high income countries. Preferably, the damages might have to include not only the economic losses but also less tangible welfare losses such as pain and discomfort associated with premature death and morbidity.  17 It was also pointed out that the shadow price of carbon is much bigger than $20/ton. To this challenge, the World Bank’s data group kindly shared yet published methodology document which addresses this concern. To summarize the standpoint; it is well acknowledged that the social cost of “carbon pollution” could be much greater but the cost assumed for the ANS (Adjusted Net Savings) estimate should have been brought closer to “values currently used by governments for policymaking purposes. And the assumed cost is derived from “estimates by Fankhauser (1994) of global losses to crops, infrastructure, and human health incurred per ton of CO2”.  18 More specific calculation methods can be found at http://data.worldbank.org/  36  To see the trend over time, when counted in US$, the environmental losses have grown over time for all examined environmental indicators. Especially, since the 2008 financial crisis, natural resource depletion and the aggregated environmental cost shot up significantly19; only the PM damage curved down slightly in 2008 but bounced back again soon after.  Figure 3.3. Environmental cost variables  a. Particulate emission damage                                                     19 Considering that the financial crisis should have caused global economic downturn and thus brought down aggregate demand for production inputs, the upsurge in depletion of natural resources was doubted to be miscalculated. On this, I quote the response from a specialist in the World Bank; “Natural resource depletion is tied to rents which depend on the level of extraction, as well as unit production costs and prices…Unit prices for fossil energy resources have fluctuated wildly in recent years, causing some of the spikes in rents and depletion that you observe.…To sum up, increasing levels of extraction along with some of the price spikes have contributed to higher estimates for natural resource depletion since 2008.” 8.0e+101.0e+111.2e+111.4e+111.6e+11US$1990 1995 2000 2005 2010 2015yearFull sample: sum of PM damage01.0e+092.0e+093.0e+094.0e+09US$1990 1995 2000 2005 2010 2015yearDeveloped Countries Mid-income CountriesLow-income Countries Lowest income CountriesIncome Group Average: Particulate Emission Damage37  b. Natural Resource Depletion   c. CO2 damage  d. Sum of the environmental costs   02.0e+124.0e+126.0e+128.0e+12US$1990 1995 2000 2005 2010 2015yearFull sample: sum of Natural Resource Depletion01.0e+112.0e+113.0e+11US$1990 1995 2000 2005 2010 2015yearDeveloped Countries Mid-income CountriesLow-income Countries Lowest income CountriesEnergy Depletion+Mineral Depletion+Net Forest DepletionIncome Group Average: Natrual Resource Depletion5.0e+101.0e+111.5e+112.0e+112.5e+113.0e+11US$1990 1995 2000 2005 2010 2015yearFull sample: sum of CO2 damage02.0e+094.0e+096.0e+098.0e+09 1.0e+10US$1990 1995 2000 2005 2010 2015yearDeveloped Countries Mid-income CountriesLow-income Countries Lowest income CountriesIncome Group Average: Carbon Dioxide Damage02.0e+124.0e+126.0e+128.0e+121.0e+13US$1990 1995 2000 2005 2010 2015year(PM damage+Natural Resource Depletion+CO2 damage)Full sample: sum of Environmental Cost01.0e+112.0e+113.0e+11US$1990 1995 2000 2005 2010 2015yearDeveloped Countries Mid-income CountriesLow-income Countries Lowest income CountriesParticulate Emission Damage+CO2 Damage+Natural Resource DepletionIncome Group Average: Environmental Cost38  3.3.1.2 Proxies for the Scale, Composition and Technique Effects The second group of dependent variables are the proxies for the scale, composition and technique effects. These variables are put as dependent variables in the intersection 1 estimation model but serve as independent variables in the compound model and in the intersection 2 estimation model. The proxies are; GDP (Gross Domestic Production) for the scale effect, capital intensity of production (consumption of fixed capital used up in the process of production) for the composition effect, and per capita income for the technique effect.  In the respect that the scale effect is related to the quantity of produced units, GDP, the monetary value of a country’s gross production may not be the most appropriate representation of the scale effect. When using panel data, the dynamics of declining and newly emerging industry make it even harder for GDP to be interchangeably used as the scale of an economy; because the composition of produced goods would change over time even absent trade liberalization. Despite this weakness, I take PPP (Purchasing Power Parity) based GDP as the best available surrogate for the scale effect. And the GDP index was modified to be self-relative; for every country, 1990 GDP was converted to 100 and subsequent years were modified accordingly (e.g. 1991 GDP → 100 *GDP1991/ GDP1990).  In the place the composition effect, I used the ratio of consumed capital to the size of GNI. Theoretically, the composition effect arises from the trade induced proportion changes of environmentally intensive (dirty) versus clean goods. Therefore, using the capital intensity data as a proxy for the composition effect is based on the previously posited assumption that capital intensity is closely correlated with environmental intensity of production. That said, the World 39  Bank’s consumption of fixed capital data is the weighted average of “the replacement value of capital (e.g. equipment and machinery) used up in the process of production”.  Using per capita income as a proxy for the technique effect is open for debate, too. As was reviewed in the preceding chapter (2.2.2.1), a sequence of causal chain should be in operation for growing income to translate into environmentally friendly technological change. To look more closely at the causal links, the weakest chain seems to be between the demand increase for the clean environment and heightening of environmental standards. This link has been tested in preceding studies by including political institution dummies; and the results show that more democratic countries are more responsive to the demand increase for stringent environmental regulations. Yet, the link between the income growth and higher demand for stringent regulation has rarely been questioned. Considering that per capita income is a simple average of total income, even if per capita income is high, unless the total income is equitably distributed, it is unlikely that the income effect would successfully lead to politically acceptable demand for clean environment.  Having said that, in this paper, GINI coefficient20 will be utilized to address the weakness of the income induced technique effect. GINI coefficient index is again taken from the World Bank database; to reduce the gaps in the data, missing values between observed data points are replaced by linear interpolation and still remaining gaps are partly filled with the first following data point or the last preceding data point up to three years.                                                   20 GINI coefficient, developed by Corrado GINI (1912), is a measure of the social income inequality; zero expresses perfect equality whereas one (or 100%) expresses maximal inequality.  40   An interesting feature is that the income group discrepancies are quite prominent for the selected proxy indicators. Especially, capital intensity of gross production is notably high in order from the high income to the lower income groups. To compare the GDP growth and per capita income, the high income countries have experienced faster per capita income growth while the total volume of GDP expanded more rapidly in lower income countries (implying that the world population has fast grown in poorer countries). As to the GINI index, while the overall average GINI coefficient exhibits decreasing trend since the late 1990’s, low income countries have suffered from severe fluctuations in the income distribution.  Figure 3.4. GDP, Consumption of Fixed Capital and Per Capita GNI  a. Self-anchored scale increase of GDP     100200300400base year=1990 as 1001990 1995 2000 2005 2010 2015yearFull sample: gross production100200300400500base year=1990 as 1001990 1995 2000 2005 2010 2015yearDeveloped Countries Mid-income CountriesLow-income Countries Lowest income CountriesIncome Group Average: GDP Growth41  b. Consumption of Fixed Capital as a % of GNI  c. Per Capita GNI   d. GINI coefficient   11.51212.5%of GNI1990 1995 2000 2005 2010 2015yearFull sample: Consumption of Fixed Capital (as % of GNI) average5101520% of GNI1990 1995 2000 2005 2010 2015yearDeveloped Countries Mid-income CountriesLow-income Countries Lowest income CountriesIncome Group Average: Consumption of Fixed Capital500010000150002000025000US$1990 1995 2000 2005 2010 2015yearFull sample: Per Capita GNI average01000020000300004000050000US$1990 1995 2000 2005 2010 2015yearDeveloped Countries Mid-income CountriesLow-income Countries Lowest income CountriesIncome Group Average: Per Capita GNI38404244%1990 1995 2000 2005 2010 2015yearFull sample: GINI coefficient average30405060%1990 1995 2000 2005 2010 2015yearDeveloped Countries Mid-income CountriesLow-income Countries Lowest income CountriesIncome Group Average: GINI Coefficient42  3.3.2 Explanatory Variables 3.3.2.1 Trade Intensity Trade openness is the explanatory variable of particular interest. In the respect that trade openness refers to the elimination of trade barriers, a comprehensive index of tariff and non-tariff barriers will serve as the most appropriate proxy for the trade liberalization. However, in the absence of such data, trade intensity index is the best available alternative to measure the openness as the intensity is expected to be the net result of overall trade friendliness of a country. The World Bank’s trade intensity data is “the sum of exports and imports of goods and services measured as a share of gross domestic product”.  A distinctive feature of the trade intensity trend is that the 2008 global financial crisis seems to have had a huge impact on the size of the international trade; mid-income countries experienced the sharpest drop in the trade-to-GDP ratio but the ratio soon bounced back while it appears that low-income countries have yet recovered from the shock.  Figure 3.5. Trade Intensity   556065707580(Export+Import)/GDP1990 1995 2000 2005 2010 2015yearFull sample: Trade Intensity average405060708090% of GDP1990 1995 2000 2005 2010 2015yearDeveloped Countries Mid-income Countries (without Singapore)Low-income Countries Lowest income CountriesIncome Group Average: Trade Intensity43  3.3.2.2 Proxies for the Scale, Composition and Technique Effects GDP, consumption of fixed capital, and per capita income are used as explanatory variables in estimating the intersection 2 of the filtered-through effects. In the compound model, these variables are put as controlled variables. The details are presented in the previous section (3.3.1.2).   3.3.2.3 Other Independent Variables In estimating the effect of trade intensity on the economic growth and income level, aggregate demand variables and aggregate supply variables are jointly put on the right hand side of the regression models; the variables include government spending, household expenditure, investment in education  and gross capital formation, and the size of population. With regard to the trade’s effect on the consumption of fixed capital, the size of capital stock and population (as a proxy for labor) are considered as other determinant factors to be controlled. Also, to estimate the effect of trade on income distribution, I controlled for education expenditure, government spending and financial openness (Foreign Direct Investment inflow). And all these variables are logarithmically transformed to match the scale of the trade intensity variable (proportion of export and import to GDP). Capital stock data is taken from Penn World Table 9.0 (released on June 2016)21 and the other data are taken from the World Bank database.                                                     21 http://www.rug.nl/research/ggdc/data/pwt/pwt-9.0 44  3.4 Estimation models In constructing estimation models, my biggest concern was the endogeneity problem. Given the broad scale and the dynamics of trade and environment relationship, simultaneous causality and omitted variable bias are sure to arise. An established tactic to deal with potential endogeneity is instrumental variable (IV) approach; by throwing out direct path of regressor being correlated with dependent variable, indirect but pure path of link between dependent and explanatory variable is expected to be identified. The problem with the IV approach, however, is that there is no statistical method to test for the validity of instrumental variables and thus it relies on researchers’ intuition and economic reasoning to find out valid instruments. 22  In addition to the burden of finding valid instruments, the IV approach forgoes the uncorrelated portion of variability of explanatory variables with the instrument. That being said, in this paper, IV approach is not taken. Instead, by resorting to fixed effects model and panel data structure, I expect to reduce endogeneity problem arising from omitted variable bias to a great extent.  Next, as shown in above graphs, a shock from the 2008 financial crisis does seem to have caused considerable distortions in the linear trend of both environmental and economic indicators. Thus, to control for the time specific effects, two way fixed effects model is set as the first candidate estimation model and the plausibility was tested against one way fixed effects model, random effects model and pooled OLS model; to examine the better fitness between random effects and fixed effects models, Hausman test was conducted and according to the result, Breusch-Pagan test                                                  22  Hausman test presents whether the result gained from IV approach is systematically different from non-IV approach; but, the test is conducted under the assumption that the used instrumental variable is valid.  45  or F-test was run respectively to choose between the panel (random effects or fixed effects) models and the OLS models. And, joint F-test was done to examine the significance of coefficients on yearly dummies. Finally, to account for possible serial correlation, contemporaneous correlation and heteroskedasticity, standard error robust estimates are to be presented together with the estimates from regular (or non-robust) standard errors.  3.4.1 Filtered-Through Effects Estimation Models (Intersection of Intersections) To capture the filtered through effects, trade intensity variable is regressed on the proxies of scale, composition and income induced technique effects and the proxies are regressed on the environmental cost indicators (Particulate emission damage, CO2 emission damage, Natural resource depletion, and the sum of the three indicators) respectively.   <Estimation of Intersection 1> GDPi,t = α1 + β1Trade Intensityi,t-1 + Α1Government Spending i,t + Β1Consumer Spending i,t  + Γ1Capital Investment i,t  + Δ1Education Expenditure i,t  + Ε1Population i,t  + ui + νt +εit  (1.1)* Note that one-period lagged trade intensity variable is used to account for the time consumed for trade induced income to be accumulated as production factors and thus lead to the expansion of economic activities.  Consumption of Fixed Capital i,t = α2 + β2Trade Intensityi,t +  Ζ Capital stock + Λ Population + ui + νt +εit           (1.2)  Per Capita GNIi,t = α3 + β3Trade Intensityi,t + Α2Government Spending i,t + Β2Consumer Spending i,t  + Γ2Capital Investment i,t  + Δ2Education Expenditure i,t  + Ε2Population i,t  + ui + νt +εit (1.3)  GINI Coefficienti,t = α4 + β4Trade Intensityi,t + Δ3Education Expenditure i,t + ΗFDI inflow i,t  + Θ Government Spending i,t  + ui + νt +εit       (1.4)  46  <Estimation of Intersection 2> PM2.5 Damagei,t = γ 1 + δ 1GDPi,t + δ 2Consumption of Fixed Capitali,t + δ 3Per Capita GNIi,t-3 + δ 4(Per Capita GNIi,t-3)2 + δ 5GINI Coefficienti,t-3 + ui + νt +εit   (1.5)  Natural resource Depletion i,t = γ 2 + θ 1GDPi,t + θ 2Consumption of Fixed Capitali,t + θ 3Per Capita GNIi,t-3 + θ 4(Per Capita GNIi,t-3)2 + θ 5GINI Coefficienti,t-3 + ui + νt +εit (1.6)  CO2 emission Damagei,t = γ 3 + λ1GDPi,t + λ 2Consumption of Fixed Capitali,t + λ 3Per Capita GNIi,t-3 + λ 4(Per Capita GNIi,t-3)2 + λ 5GINI Coefficienti,t-3 + ui + νt +εit  (1.7)  Environmental Costi,t = γ 4 + μ1GDPi,t + μ 2Consumption of Fixed Capitali,t +μ 3Per Capita GNIi,t-3 + μ 4(Per Capita GNIi,t-3)2 + μ 5GINI Coefficienti,t-3 + ui + νt +εit  (1.8) Note that three-periods lagged income and income distribution variables are used to take policy rigidity into consideration.  <Intersection of Intersections> Once coefficients are obtained from above models (1.1)-(1.8), the estimates are then multiplied to find filtered-through effects of trade on each environmental cost variables. For example, β1δ1 is trade’s impact on the particulate emission damage filtered through the scale effect (increase of GDP); β2θ2 is trade’s impact on the natural resource depletion filtered through the composition effect (% of fixed capital to the GNI); β3μ3 is trade’s impact on the CO2 emission damage filtered through the technique effect (demand increase for clean environment represented by per capita income), and β4μ5 is trade’s impact on the aggregated environmental cost (PM damage + resource depletion + CO2 damage) filtered through the income distributional effect (GINI coefficient) of freer trade. And finally, the sum of filtered effects through the scale, composition and income (both size and distribution) will be interpreted as trade’s impact on each environmental cost indicators; e.g. β1δ1+β2δ2+β3δ3+β3δ4+β4δ5 is filtered-through impacts on PM damage of freer trade). 47   3.4.2 Compound Models To examine the unfiltered-through effects, multiple regression models are established to take trade intensity as a main explanatory variable while controlling for filtered through effects associated economic indicators.  PM2.5 Damagei,t = κ1 + ψ1Trade Intensity i,t + ρ1GDPi,t + ρ 2Consumption of Fixed Capitali,t + ρ 3 Per Capita GNIi,t-3 + ρ 4(Per Capita GNIi,t-3)2 + ρ 5GINI Coefficienti,t-3 + ui + νt +εit   (2.1)  Natural resource Depletioni,t = κ 2 + ψ2Trade Intensity i,t + τ 1GDPi,t + τ 2Consumption of Fixed Capitali,t + τ 3Per Capita GNIi,t-3 + τ 4(Per Capita GNIi,t-3)2 + τ 5GINI Coefficienti,t-3 + ui + νt +εit (2.2)  CO2 emission Damagei,t = κ 3 + ψ3Trade Intensity i,t + φ1GDPi,t + φ 2Consumption of Fixed Capitali,t + φ 3Per Capita GNIi,t-3 + φ 4(Per Capita GNIi,t-3)2 + φ 5GINI Coefficienti,t-3 + ui + νt +εit   (2.3)  Environmental Costi,t = κ 4 + ψ4Trade Intensity i,t + ω 1GDPi,t + ω 2Consumption of Fixed Capitali,t + ω 3 Per Capita GNIi,t-3 + ω 4(Per Capita GNIi,t-3)2 + ω 5GINI Coefficienti,t-3 + ui + νt +εit   (2.4)  3.5 Results Table 3.1-3.6 are the excerpt from the summarized estimation results (Table 3.7-3.8 attached at the end of this chapter). Refer to Appendix A to C to see the results from the Hausman test, Breusch-Pagan test, F-test (to see if all ui =o), Joint F-test (to see if all νt=o), and the estimation methods selected based on the test results. And I should add that the trustworthiness of the results below is contingent on the reliability of the data.23                                                  23 See Footnotes 10 (p.36), 16 and 17 (p.43), and 19 (p.44). 48   3.5.1 Hypothesis1) Test Results To look at the trade’s impact on the environment, the full sample panel analysis (Table 3.1) shows that the filtered-through-effects are constantly adverse to the environment, regardless of the environmental externality being local or global. Meanwhile, the unfiltered impacts appear to be consistently beneficial and statistically significant. When the impacts of opposite directions are compared, the beneficial effect of trade exceeds the aggravating effect. This finding seems to lend support to the claim that trade and environment are in symbiotic relationship. And the significant and beneficial unfiltered-through effect could be interpreted that the technique effect (which does not filter through the income effect) is large enough to offset the environmentally harmful effects of trade expansion; i.e., technical spill-over or the upward convergence of environmental regulation effect (namely the “California effect”) are effectively in operation.   To see the specific results, once insignificant channel of impacts are dropped (which is the composition effect in the full sample analysis), 1% point increase of trade intensity, through the channel of increased economic activity and enhanced per capita income, results in 1.16 billion dollar increase of national level environmental cost; comprised of 10.1 million dollar increase of particulate emission damage, 1.12 billion dollar loss from depleted natural resources, and 30.4 million dollar damage from CO2 emission. On the other hand, increased economic interdependency seems to engender higher chances of technological advancement and diffusion to cut down 2 billion dollars of environmental cost for 1% point increase of trade intensity.  49  Therefore, the full sample result does not support the hypothesis 1 in the respect that the relative magnitude of the technique effect does not appear to vary across different environmental indicators. And even more, the result indicates that the main driver of the technique effect is not the income effect (or endogenous policy responses). Rather, the result implies that knowledge diffusion and technology transfer are the key sources of trade related environmental benefits.   Table 3.1 Environmental cost of Trade intensity (Full sample)  Coefficients on Trade Intensity  unit: US$ PM damage Resource depletion CO2 damage Environmental Cost (PM+Resource+CO2)  Filtered Unfiltered Filtered Unfiltered Filtered Unfiltered Filtered Unfiltered Full sample 1.02E+07 -2.78e+07*** 1.13E+09 -1.94e+09*** 3.15E+07 -4.11e+07*** 1.17E+09 -2.00e+09*** 1.01E+07   1.12E+09   3.04E+07   1.16E+09   Notes 1 (Common to Table 3.1 to 3.6): Model specifications are shown in Appendix A. In cases where robust standard errors produce different test results, the differences are shown in parenthesis; for example, **(*) means that the result is significant at 99% confidence level when error disturbance is not controlled and significant at 95% when robust standard errors are used.   group1 (High-income) > 10,000US$ per capita GNI since before the 1980's group2 (Mid-income) > 10,000US$ per capita GNI in the 1980's and after 3,000US$ < group3 (Low-income) < 10,000US$ group4 (Lowest income) < 3,000US$  * Significance at the 90% confidence level, ** Significance at the 95% confidence level. *** Significance at the 99% confidence level.  Notes 2 (Common to Table 3.1 to 3.2):: With regard to the filtered-through effects, the values shown above are the combined products of the intersection 1 and intersection 2. When both intersection 1 and intersection 2 are statistically significant (at 90% or above significance level), product of the coefficients is also regarded as significant and the sum of those significant products are shown in bold numbers.    50  Next, the income group analysis (Table 3.2) results present a different feature of the trade-environment relationship and suggest that the full sample results are not generally applicable to every individual countries. To look at the unfiltered-through effects, the beneficial technology transfer effects no longer holds valid for the highest income group; coefficients are either positive or insignificant. On the other hand, the lowest income group turns out to be mostly benefitted from the unfiltered through effects; 1% point increase of trade intensity is found to reduce the national level environmental cost by 89.7 million dollars.   Meanwhile, positive and significant unfiltered through effects on PM damage (11 million dollar per % point increase of trade intensity) and CO2 damages (30 million dollar per % point increase of trade intensity) in the low income group requires more careful interpretation. Considering that trade associated transportation increase is absorbed by GDP variable, the obtained positive coefficients are expected to arise from uncontrolled time-variant effects specific to the low-income group; and the most probable candidate is burgeoning population in those countries.24 When the population variable is controlled, the results indeed overturned to indicate that trade intensity reduces PM damage (although insignificant), and only minimally increase the CO2 damage.25  Income group filtered-through effects are also contrastingly different from the full sample estimates. First, except for the group 1 (developed country), filtered-through effect is not                                                  24 To avoid multicollinearity problem, population variable is not controlled jointly with the GDP variable; the degree of correlation between population variable and GDP variable is found to be 0.736 in the high-income and the lowest income groups but lower than 0.7 (cut-off threshold) for the low income group data.  25 Population variable controlled estimation produced 7 million US$ decrease of PM damage and 2725.4***US$ increase of CO2 damage per % point increase of trade intensity.  51  adversarial to the environment; in the mid-income and the lowest income group, trade intensity is found to be environmental cost reducing regardless of the indicator types. Second, the directions of the scale, composition and income induced technique effect are not consistent (Table 3.8). i) Higher GDP increases PM damage only in the low-income group (in the rest three groups, δ1 appears to be smaller than zero.)  ii) In contrast to the expectation, capital intensity is negatively correlated with the aggregated environmental cost (μ2 < 0 for all income groups); only in the high income group, 1% point increase of trade intensity increases capital intensity of production slightly but meaningfully (0.018% point increase of fixed capital consumption) and thus results in CO2 damage increase by 11.5 million dollars at national level.  iii) Income increase tends to damage and then improves the environment in the lowest income group (μ3>0, μ4<0) but works to improve first and then damage the environment in higher income groups (μ3< 0, μ4>0).   Table 3.2 Environmental cost of Trade intensity (by Income group)  Coefficients on Trade Intensity  unit: US$ PM damage Resource depletion CO2 damage Environmental Cost (PM+Resource+CO2)  Filtered Unfiltered Filtered Unfiltered Filtered Unfiltered Filtered Unfiltered group1 1.43E+07 3377265 -7.96E+08 1.09e+08 -4.08E+06 6427267 -9.50E+08 -1.46e+07 1.15E+07   0  1.26E+07   0  group2 -1.30E+07 -3441673 --1.26E+09 -5.23e+07 -1.64E+07 -3828132(**) -1.28E+09 -7.12e+07 -1.26E+07   -1.35E+09   -1.72E+07   -1.37E+09   group3 -3.37E+06 1.12e+07(***) -2.63E+08 -8.45e+08 -2.05E+07  3.02e+07**(*) -3.28E+08 -8.04e+08 -2.02E+06   2.24E+03   -2.16E+07   0   group4 -2.00E+06 -1.15e+07**(*) -1.21E+07 -7.75e+07** -1.85E+05  -1373060* -1.95E+07 -8.97e+07** -2.57E+06   -2.03E+07   -4.75E+05   -2.74E+07   Notes: See the Table 3.1 Notes 1 & 2.   52  To look more closely at the effects of income on the environment (Table 3.3), the full sample data show that there exists inverted U-shaped relationship for all three environmental cost indicators. However, when disaggregated into different income groups, the EKC trend no longer holds valid. Except for the lowest income countries (of which ECK persists), the income-environment relationship varies across income groups and environmental indicators. And this inconsistency indicates that the EKC trend obtained from the full sample may reflect a spurious income-environment relationship; all the low-cost and high income combination are from economically well-off countries whilst the high-cost and low income combinations are from economically disadvantaged countries.   Table 3.3 Income-Environment relationship  Coefficients on  Per Capita GNI (US$) & (Per Capita GNI)2 unit :  US$  PM damage  Resource depletion  CO2 damage  Environmental  Cost   Intersection 2 compound Intersection 2 compound Intersection 2 compound Intersection 2 compound Full sample GNIpc 157225.3*** 171189.4*** 1.84e+07*** 1.91e+07*** 476254**(*) 583386.7*** 1.91e+07*** 1.98e+07*** GNIpc2 -0.826*** -1.019*** -87.353** -99.749*** -2.278(***) -3.115*** -90.37**(*) -103.171*** group1 GNIpc 149325.6(**) 149894.8(**) -6537402 -6475541 -91614.56 -124792.5 -7653591 -7261160 GNIpc2 0.196 -0.199 30.747 30.647 0.299 0.884 39.675 37.513 group2 GNIpc -144664(***) 39888.09 -1.51e+07*** 1.49e+07**(*) -201805*(**) -106136*(**) -1.53e+07*** -1.50e+07**(*) GNIpc2 3.436(**) -0.606 261.962*** 258.35**(*) 5.436**(*) -0.080 261.69**(*) 256.77**(*) group3 GNIpc 11065.34 -54770.93 -8317739 -1.00e+07 -990649*** -1015953*** -1.15e+07 -1.32e+07 GNIpc2 7.495 12.330 141.871*** 1791.37 -1.15e+07 117.94*** 2460.55 2063.83 group4 GNIpc 3273100**(*) 2715549**(*) 3.06e+07*** 3.02e+07*** 1137760*** 1043984*** 3.83e+07**(*) 3.39e+07**(*) GNIpc2 -651.13 -447.64(***) -7660.9(***) -7584.35(**) -335.76*** -307.23*** -9858.33*(**) -8250.62(***) Notes: See the Table 3.1 Notes 1.  Note2: Intersection 2 estimates capture the correlation between trade induced income with the envirionmental cost variables. Meanwhile, the compound model estimates forgo of the trade intensity correlated proportion of income change in predicting the income-environment relationship.  53  And interestingly, results from the trade intensity uncontrolled model (intersection 2) show that the income (X axis) - environment (Y axis) relationship is closer to N-shape; when the statistically significant coefficients are gathered from each income group estimation results, as income grows, PM damage rises in the lowest income bracket but drops in group3 (low income countries) to rise again in the mid and high income groups. Likewise, income and resource depletion relationship appears to be inverted U-shaped in the lowest income group, but upward slope in the low-income group, and U-shaped curve in the mid-income group; when the results are combined, it again denotes N-shaped relationship. And the pattern persists for the CO2 damage and the aggregated environmental cost as well.   Results from the trade intensity controlled estimation (the compound model) are not very different. N-shaped relationship becomes even clearer for the PM damage, resource depletion and the aggregated environmental cost. But the coefficients on CO2 damage and the group2 estimation results are distinguishably different from the trade intensity uncontrolled estimation result; the compound model produces a roughly M-shaped income (X axis) - CO2 damage curve (Y axis) and the direction of coefficients are mostly overturned in group 2. And the reason why trade openness correlated portion of income change would alter the pattern of income-environment relationship needs more in-depth investigation (in the future work).   That said, from the obtained results (Table 3.3), a conclusion can be drawn that, in an environmental perspective, trade induced income increase is not the right reason to support trade liberalization. This is because, although trade may increase income, there is no evidence to believe that income increase may eventually resolve environmental problems. Restricted to the lowest 54  income group, it may be right to say that ‘impoverishment is the cause of environmental problem’; inverted U-shaped income-environment relationship constantly appears in group 4. But unfortunately, international trade does not seem to alleviate poverty in those countries (see Table 3.4 below). Meantime, it was found that environmental degradation cannot be the reason to oppose free trade, either because trade, after all, seems to bring more environmental benefits than harms.  Table 3.4 Trade-Income relationship  Dependent variable: Per Capita Income (US$)  Full sample group1 group2 group3 group4 Trade Intensity (%) 27.08*(**) 76.68**(*) 82.21*** 15.8*** 0.289 See the Table 3.1 Notes 1.   3.5.2 Hypothesis2) Test Results Given the results above, a reason for the ineffective income effect is sought from inequitable distribution of income. It was examined whether trade openness affects the income distribution.26  First, to look at the inequality-environment relationship (Table 3.5), the full sample analysis shows that % point increase of GINI coefficient would increase both PM damage and CO2 damage; roughly by 30 million dollars and 100 million dollars respectively at national level. Yet, when disaggregated into different income groups, the signals are mixed; in high income countries (group1), inequality tends to raise the environmental cost but in mid-income countries (group2), inequality appears to reduce the environmental cost in general. Regarding why inequality would                                                  26 It is widely accepted that trade liberalization is detrimental to domestic industries with low international competitiveness while being beneficial to comparatively advantageous sectors, and therefore usually brings about adverse distributional impacts. 55  reduce environmental cost, further investigation needs to be done and the best guess I can make here is that the disproportionate income distribution may negatively interact with propensity to consume and thus curves down economic activities in the mid-income bracket.  Table 3.5 Inequality-Environment relationship  Coefficients on GINI coefficient (%point) US$ PM damage Resource depletion CO2 damage Environmental Cost  Intersection 2 compound Intersection 2 compound Intersection 2 compound Intersection 2 compound Full  sample 2.89e+07** 2.98e+07**(*) 1.32e+09 1.46e+09 1.10e+08**(*) 9.72e+07*** 1.46e+09 1.61e+09 group1 1.03e+08 1.04e+08 1.37e+10 1.44e+10*(**) 6.07e+08 9.49e+08(***) 1.63e+10 1.65e+10(***) group2 3.83e+07(***) -5.33e+07(***) -3.29e+09*(**) -3.24e+09*(*) 7609349 -3.20e+07** -3.41e+09*(**) -3.34e+09*(**) group3 3.10e+07*(**) 1.82e+07(**) -1.82e+09 -1.78e+09 9.16e+07*** 1.15e+08**(*) -1.68e+09 -1.61e+09 group4 -2920059 -476850.4 4.21e+07 1.11e+08* -1004302 -432726.3 9.49e+07 1.13e+08* See the Table 3.1 Notes.  All in all, when statistically significant estimates are singled out, income inequality is likely to result in higher levels of environmental harm (except for the mid-income countries). And this could partially explain for why income increase does not automatically translate into enhanced environmental quality. And if trade liberalization is found to inflict wider inequality, that would lend support to anti-globalization argument in both economic and environmental perspectives.   The finding (Table 3.6), however, does not present a clear evidence that open trade triggers less equitable distribution of wealth. To see the results from the full sample regression, distributional impact of trade is trivial and statistically insignificant. Only in the mid-income countries, adverse distributional impact seems to arise but the size of the effect is again pretty small; 0.017% point increase of GINI coefficient per 1% point increase of trade intensity. Even more, as was shown 56  above (Table 3.5), inequality in mid-income countries (group 2) does not result in higher level of environmental cost.   Interestingly, and conversely to the expectation (that trade may exacerbate inequality), increased economic interdependency seems to improve income distribution in most income groups (group 1, 3, and 4). Especially in group 3 (low-income countries), the effect turns out to be statistically significant and when considered in combination with the results from Table 3.5, trade induced distributional impact is found to improve the environment in low-income countries.  Table 3.6 Trade-Income Equality relationship  Dependent variable: GINI coefficient (%)  Full sample group1 group2 group3 group4 Trade Intensity (%p) 0.006 -0.017 0.017*(*) -0.065*** -0.03 See the Table 3.1 Notes 1.  To summarize, contrary to the hypothesis 2, if trade has any distributional impacts, the impact is found to be environmental cost reducing.  And the distributional impact is, for most income groups, equality enhancing rather than inequality aggravating (except for group 3). Therefore, despite the finding that inequality is likely to induce higher environmental cost, the inequality-environment relationship does not validate anti-globalization claims.     57   Table 3.7 Full sample analysis results   a. All Sample Countries Filtered-through effects Unfiltered effects Intersection 1 Intersection 2 product β1 0.307** δ1 1.89e+07*** β1δ1 5.80E+06 ρ1 1.91e+07*** θ1 2.04e+09*** β1θ1 6.26E+08 τ1 2.06e+09*** λ1 5.70e+07*** β1λ1 1.75E+07 φ1 5.78e+07*** μ1 2.11e+09*** β1μ1 6.48E+08 ω1 2.13e+09*** β2 0.005 δ2 -1.20e+07 β2δ2 -6.00E+04 ρ2 1.12e+07 θ2 -1.01e+09 β2θ2 -5.05E+06 τ2 1.03e+09 λ2 8.20e+07 β2λ2 4.10E+05 φ2 1.54e+08** μ2 -9.04e+08 β2μ2 -4.52E+06 ω2 1.20e+09 β3 27.08*(**) δ3 157225.3*** β3δ3 4257661.124 ρ3 171189.4*** θ3 1.84e+07*** β3θ3 4.98E+08 τ3 1.91e+07*** λ3 476254.2**(*) β3λ3 12896963.74 φ3 583386.7*** μ3 1.91e+07*** β3μ3 5.17E+08 ω3 1.98e+07*** δ4 -0.826*** β3δ4 -22.36808 ρ4 -1.019*** θ4 -87.353** β3θ4 -2365.519 τ4 -99.749*** λ4 -2.278(***) β3λ4 -61.68824 φ4 -3.115*** μ4 -90.3722**(*) β3μ4 -2447.279 ω4 -103.171*** β4 0.006 δ5 2.89e+07** β4δ5 1.73E+05 ρ5 2.98e+07**(*) θ5 1.32e+09 β4θ5 7.92E+06 τ5 1.46e+09 λ5 1.10e+08**(*) β4λ5 6.60E+05 φ5 9.72e+07*** μ5 1.46e+09 β4μ5 8.76E+06 ω5 1.61e+09 PM Damage β1δ1+β2δ2+β3δ3+β3δ4+β4δ5 1.02E+07 ψ1 -2.78e+07*** Resource Depletion β1θ1+β2θ2+β3θ3+β3θ4+β4θ5 1.13E+09 ψ2 -1.94e+09*** CO2 Damage β1λ1+β2λ2+β3λ3+β3λ4+β4λ5 3.15E+07 ψ3 -4.11e+07*** Environmental Cost β1μ1+β2μ2+β3μ3+β3μ4+β4μ5 1.17E+09 ψ4 -2.00e+09*** 58   Table 3.8 Income group analysis results   b. Developed Countries Filtered-through effects Unfiltered effects Intersection 1 Intersection 2 product β1 0.013 δ1 -3.25e+07*** β1δ1 -4.23E+05 ρ1 -3.23e+07*** θ1 1.11e+08 β1θ1 1.44E+06 τ1 1.15e+08 λ1 4.96e+07*** β1λ1 6.45E+05 φ1 4.40e+07 μ1 4.85e+07 β1μ1 6.31E+05 ω1 3.69e+07 β2 0.018(**) δ2 2.80e+08 β2δ2 5.04E+06 ρ2 2.85e+08 θ2 -3.51e+09 β2θ2 -6.32E+07 τ2 -3.49e+09 λ2 7.01e+08**(*) β2λ2 1.26E+07 φ2 5.69e+08 μ2 -4.81e+09 β2μ2 -8.66E+07 ω2 -4.89e+09 β3 76.68**(*) δ3 149325.6(**) β3δ3 11450287.01 ρ3 149894.8(**) θ3 -6537402 β3θ3 -5.01E+08 τ3 -6475541 λ3 -91614.56 β3λ3 -7025004.46 φ3 -124792.5 μ3 -7653591 β3μ3 -5.87E+08 ω3 -7261160 δ4 0.196 β3δ4 15.02928 ρ4 -0.199 θ4 30.747 β3θ4 2357.68 τ4 30.647 λ4 0.299 β3λ4 22.927 φ4 0.884 μ4 39.675 β3μ4 3042.279 ω4 37.513 β4 -0.017 δ5 1.03e+08 β4δ5 -1.75E+06 ρ5 1.04e+08 θ5 1.37e+10 β4θ5 -2.33E+08 τ5 1.44e+10*(**) λ5 6.07e+08 β4λ5 -1.03E+07 φ5 9.49e+08(***) μ5 1.63e+10 β4μ5 -2.77E+08 ω5 1.65e+10(***) PM Damage β1δ1+β2δ2+β3δ3+β3δ4+β4δ5 1.43E+07 ψ1 3377265 Resource Depletion β1θ1+β2θ2+β3θ3+β3θ4+β4θ5 -7.96E+08 ψ2 1.09e+08 CO2 Damage β1λ1+β2λ2+β3λ3+β3λ4+β4λ5 -4.08E+06 ψ3 6427267 Environmental Cost β1μ1+β2μ2+β3μ3+β3μ4+β4μ5 -9.50E+08 ψ4 -1.46e+07 59  c. Mid-income Countries Filtered-through effects Unfiltered effects Intersection 1 Intersection 2 product β1 0.292** δ1 -5041638(***) β1δ1 -1.47E+06 ρ1 1285623(**) θ1 -1.48e+08**(*) β1θ1 -4.32E+07 τ1 -1.49e+08**(*) λ1 -2184880(**) β1λ1 -6.38E+05 φ1 -1702698*** μ1 -1.52e+08**(*) β1μ1 -4.44E+07 ω1 -1.54e+08**(*) β2 -0.009 δ2 4.78e+07(**) β2δ2 -4.30E+05 ρ2 -8.48e+07(**) θ2 -9.52e+09*(**) β2θ2 8.57E+07 τ2 9.65e+09**(*) λ2 -7.33e+07(***) β2λ2 6.60E+05 φ2 -6148817 μ2 -9.58e+09*(**) β2μ2 8.62E+07 ω2 -9.75e+09**(*) β3 82.21*** δ3 -144664(***) β3δ3 -11892827.44 ρ3 39888.09 θ3 -1.51e+07*** β3θ3 -1.24E+09 τ3 1.49e+07**(*) λ3 -201805*(**) β3λ3 -16590389.05 φ3 -106135.9*(**) μ3 -1.53e+07*** β3μ3 -1.26E+09 ω3 -1.50e+07**(*) δ4 3.436(**) β3δ4 282.47356 ρ4 -0.606 θ4 261.962*** β3θ4 21535.89602 τ4 258.349**(*) λ4 5.436**(*) β3λ4 446.89356 φ4 -0.080 μ4 261.69**(*) β3μ4 21513.5349 ω4 256.768**(*) β4 0.017*(*) δ5 3.83e+07(***) β4δ5 7.66E+05 ρ5 -5.33e+07(***) θ5 -3.29e+09*(**) β4θ5 -6.58E+07 τ5 -3.24e+09*(*) λ5 7609349 β4λ5 1.52E+05 φ5 -3.20e+07** μ5 -3.41e+09*(**) β4μ5 -6.82E+07 ω5 3.34e+09*(**) PM Damage β1δ1+β2δ2+β3δ3+β3δ4+β4δ5 -1.30E+07 ψ1 -3441673 Resource Depletion β1θ1+β2θ2+β3θ3+β3θ4+β4θ5 -1.26E+09 ψ2 -5.23e+07 CO2 Damage β1λ1+β2λ2+β3λ3+β3λ4+β4λ5 -1.64E+07 ψ3 -3828132(**) Environmental Cost β1μ1+β2μ2+β3μ3+β3μ4+β4μ5 -1.28E+09 ψ4 -7.12e+07 60  d. Low-income Countries Filtered-through effects Unfiltered effects Intersection 1 Intersection 2 product β1 -0.025 δ1 2.71e+07*** β1δ1 -6.78E+05 ρ1 2.75e+07*** θ1 2.98e+09*** β1θ1 -7.45E+07 τ1 2.98e+09*** λ1 8.30e+07*** β1λ1 -2.08E+06 φ1 8.30e+07*** μ1 3.09e+09*** β1μ1 -7.73E+07 ω1 3.09e+09*** β2 0.032*** δ2 -2.66e+07 β2δ2 -8.51E+05 ρ2 2.11e+07 θ2 -5.47e+09 β2θ2 -1.75E+08 τ2 -3.05e+09 λ2 1.62e+07 β2λ2 5.18E+05 φ2 2.74e+07 μ2 -5.56e+09 β2μ2 -1.78E+08 ω2 -3.13e+09 β3 15.8*** δ3 11065.34 β3δ3 174832.372 ρ3 -54770.93 θ3 -8317739 β3θ3 -1.31E+08 τ3 -1.00e+07 λ3 -990648.5*** β3λ3 -15652246.3 φ3 -1015953*** μ3 -1.15e+07 β3μ3 -1.82E+08 ω3 -1.32e+07 δ4 7.495 β3δ4 118.421 ρ4 12.330 θ4 141.871*** β3θ4 2241.5618 τ4 1791.37 λ4 -1.15e+07 β3λ4 -181700000 φ4 117.940*** μ4 2460.55 β3μ4 38876.69 ω4 2063.832 β4 -0.065*** δ5 3.10e+07*(**) β4δ5 -2.02E+06 ρ5 1.82e+07(**) θ5 -1.82e+09 β4θ5 1.18E+08 τ5 -1.78e+09 λ5 9.16e+07*** β4λ5 -5.95E+06 φ5 1.15e+08**(*) μ5 -1.68e+09 β4μ5 1.09E+08 ω5 -1.61e+09 PM Damage β1δ1+β2δ2+β3δ3+β3δ4+β4δ5 -3.37E+06 ψ1 1.12e+07(***) Resource Depletion β1θ1+β2θ2+β3θ3+β3θ4+β4θ5 -2.63E+08 ψ2 -8.45e+08 CO2 Damage β1λ1+β2λ2+β3λ3+β3λ4+β4λ5 -2.05E+08 ψ3 3.02e+07**(*) Environmental Cost β1μ1+β2μ2+β3μ3+β3μ4+β4μ5 -3.28E+08 ψ4 -8.04e+08 61  e. Lowest-income Countries Filtered-through effects Unfiltered effects Intersection 1 Intersection 2 product β1 0.324 δ1 -1433271*(**) β1δ1 -4.64E+05 ρ1 -528350.5(*) θ1 1955660 β1θ1 6.34E+05 τ1 6619579 λ1 -210637*** β1λ1 -6.82E+04 φ1 -162652** μ1 -1057478 β1μ1 -3.43E+05 ω1 6045586 β2 0.024*(**) δ2 -1.07e+08**(*) β2δ2 -2.57E+06 ρ2 -9.79e+07**(*) θ2 -8.45e+08(***) β2θ2 -2.03E+07 τ2 -9.37e+08*** λ2 -1.98e+07** β2λ2 -4.75E+05 φ2 -1.81e+07** μ2 -1.14e+09*** β2μ2 -2.74E+07 ω2 -1.07e+09*** β3 0.289 δ3 3273100**(*) β3δ3 945925.9 ρ3 2715549**(*) θ3 3.06e+07*** β3θ3 8.84E+06 τ3 3.02e+07*** λ3 1137760*** β3λ3 328812.64 φ3 1043984*** μ3 3.83e+07**(*) β3μ3 1.11E+07 ω3 3.39e+07**(*) δ4 -651.125 β3δ4 -188.175125 ρ4 -447.64(***) θ4 -7660.945(***) β3θ4 -2214.013105 τ4 -7584.35(**) λ4 -335.761*** β3λ4 -97.034929 φ4 -307.233*** μ4 -9858.331*(**) β3μ4 -2849.057659 ω4 -8250.623(***) β4 -0.03 δ5 -2920059 β4δ5 8.76E+04 ρ5 -476850.4 θ5 4.21e+07 β4θ5 -1.26E+06 τ5 1.11e+08* λ5  -1004302 β4λ5 3.01E+04 φ5 -432726.3 μ5 9.49e+07 β4μ5 -2.85E+06 ω5 1.13e+08* PM Damage β1δ1+β2δ2+β3δ3+β3δ4+β4δ5 -2.00E+06 ψ1 -1.15e+07**(*) Resource Depletion β1θ1+β2θ2+β3θ3+β3θ4+β4θ5 -1.21E+07 ψ2 -7.75e+07** CO2 Damage β1λ1+β2λ2+β3λ3+β3λ4+β4λ5 -1.85E+05 ψ3 -1373060* Environmental Cost β1μ1+β2μ2+β3μ3+β3μ4+β4μ5 -1.95E+07 ψ4 -8.97e+07** 62  Chapter 4: Conclusion  The underlying motivation for conducting this research came from the doubt that trade associated environmental externalities could significantly undermine the economic benefits of trade liberalization. In spite of well acknowledged economic benefits of free trade, I seriously doubted that the world scale specialization of production and thus increased consumption (and production) could ever be good for the environment. The doubt, however, is found to be groundless and in fact, I found evidences that trade liberalization reduces environmental costs.  A cautionary remarks should be made, however, that my research outcome does not support the notion that trade induced income increase would lead to higher level of resource allocations for the environmental purposes and thus contribute to achieving environmental goals. Although trade intensity is found to increase per capita income in developed and developing countries alike, the increased income does not necessarily improve environmental amenities. Limited to the lowest income group, poverty alleviation seems to be an effective tool to remedy environmental ills but unfortunately, it appeared that trade openness does not bring any significant economic benefits to those poor countries.  In this paper, I have taken two-track approach to estimate trade’s impact on the environment. The first track was designed to estimate ‘filtered-through’ effects of trade on the environment while the second track approach was to estimate the ‘unfiltered-through effects; in terms of estimation method, filtered-through effects reflect the correlated portion of variance in trade intensity with both economic indicators and environmental cost variables and the unfiltered-through effects are 63  obtained by controlling for the trade induced changes in the size and patterns of economic activities (and thus .forgoes the correlated portion of variance in trade intensity with the controlled variables).  And this dual track approach produced an interesting result to show that income induced technique is not functioning very well but trade still reduces environmental costs through the unfiltered- effect channel; which implies that technology spillover and knowledge diffusion would be the key sources of the technique effect. And therefrom, a logical inference can be made that, in an environmental perspective, countries with the most advanced technologies will be least benefitted from free trade; and the income group analysis indeed found that developed countries do not enjoy the beneficial unfiltered through effects whilst the lowest income countries are most benefitted. This finding suggests a counter evidence to the pollution haven hypothesis; contrary to the fear that trade associated environmental impacts would be disproportionately distributed to the disadvantage of poor (or lax regulating) countries, the income group analysis results show that those poor countries (or less technologically advantaged countries) have good enough reason to open to trade.   Another intriguing results from the income group analysis is that the income-environment relationship is inconsistent across different income groups and when the statistically significant coefficients are put together, the shape of the relationship tends to be N-shaped. If this reflects the genuine tendency of income induced environmental impacts, and if the full sample indication of EKC captures only the simple spread of inherently different income-damage combinations of rich 64  and poor countries, we might have to reconsider giving too much credit to economic development as a remedy to solve environmental problems.   With regard to the reason for the malfunctioning of income effect, inequitable income distribution has been considered. And it was found that income inequality partially explains for the increase of environmental cost but not fully. Also, unlike the commonly accepted belief that trade may exacerbate national level income inequality, trade openness was found to be either equality improving or have statistically insignificant effect on the income distribution.   Overall, in the perspective of distributional impacts whether it being environmental costs or income, I could not find any reason to oppose trade liberalization. More importantly, when it comes to the environmental impacts, open trade turned out to do more good than bad, presumably by providing more chances for countries to learn best practices from each other. Therefore, to the title question whether free trade is free of environmental cost, I might have to answer that ‘free trade offers double dividend’; open trade improves national income level (the first dividend) and also saves environmental cost (the second dividend).   Still however, I would not say this conclusion is definite and generally applicable because of following reasons (caveats in my analysis). First, the panel data used in this paper is not a random subset of the world; in an attempt to include equitable number of each income group countries, the sample ended up including considerably larger proportion of rich countries than that of the real world. But expanding sample size is forgone to avoid the efficiency loss from frequent missing values (left out countries have larger gaps in the dataset). Second, the environmental indicators 65  used in this paper is limited in scope (indicators such as solid waste generation and water pollution are not covered) and therefore, the link between trade and the environment is likely to be underestimated. Third, and most importantly, potential simultaneous causality between trade and environment relationship has not been thoroughly removed. In addition to these weaknesses, one might argue that the environmental cost indicator used in this paper is drawn from disputable weak sustainability indicators (World Bank’s genuine savings data) and thus is fundamentally inadequate to address environmental concerns based on a strong sustainability perspective or intergenerational equity based morality questions. Having said that, I will be happy to receive any constructive suggestions to improve these weaknesses and come to a finer conclusion in the future.   66  Bibliography  Ã, M. M. R. (2004). The capital gains from trade are not enough : evidence from the environmental accounts of Venezuela and Mexico, 48, 1175–1191 Abbes, C., & Studies, P. (2011). Analyzing the Twin Effects of Trade and Population Ageing on the Environment Chahreddine Abbes Thesis submitted to the Faculty of Graduate and Postdoctoral Studies in partial fulfillment of the requirements for the PhD degree in Economics. Abdulai, A., & Ramcke, L. (2009). The Impact of Trade and Economic Growth on the Environment : Revisiting the Cross-Country Evidence. Kiel Working Paper, March(1491) Adam, B., Richard, G., & Robert, N. (2002). Environmental Policy and Technological Change Ambec, S., Cohen, M. A., & Elgie, S. (2011). The Porter Hypothesis at 20: Can Environmental Regulation Enhance Innovation and Competitiveness?, (January). Antweiler, B. W., Copeland, B. R., & Taylor, M. S. (2016). Is Free Trade Good for the Environment, The American Economic Review, 877–908. Ashley, A. S. (2011). An econometric analysis of ecological footprint determinants : Implications for sustainability. Atje, R., & Gaduh, A. B. (2006). Integration and Trade Specialization in East Asia. Barbier, E. B. (1996). Economic Growth and Environmental Degradation : The Environmental Kuznets Curve and Sustainable Development, 24(7). Barrett, S. (1994). Strategic environmental international trade, 54, 325–338. Bond, B. Y. E. W., Trask, K., & Wang, P. (2016). Factor Accumulation and Trade : Dynamic Comparative Advantage with Endogenous Physical and Human Capital Boyce, J. K. (2007). Is inequality bad for the environment? Research in Social Problems and Public Policy, 15(135), 267–288. Brown, G. M. (2000). Renewable Natural Resource Management and Use without Markets. Journal of Economic Literature, 38(4), 875–914.  Brueckner, M., & Schwandt, H. (2014). Income and Population Growth. The Economic Journal, (7422). 67  Brunel, C., & Levinson, A. (2016). Measuring the Stringency of Environmental Regulations, 10(1), 47–67. Brunnermeier, S. B., & Levinson, A. (2004). Examining the Evidence on Environmental Regulations and Industry Location. The Journal of Environment & Development, 13(1), 6–41.  Cadoret, I. (2015). The Direct Trade-Induced Composition Effect and Its Environmental Outcomes in Different Continents, (1998), 1–18. Charles V. Kidd, 1992, “The evolution of sustainability”, Journal of Agricultural and Environmental Ethics, Vol.5, Issue 1, pp 1-26 Ciccone, A., & Papaioannou, E. (2007). Human Capital, the Structure of Production, and Growth, (April 2005). Cole, M. A. (2004). Trade, the pollution haven hypothesis and the environmental Kuznets curve : examining the linkages, 48, 71–81.   Cole, M. A., & Elliott, R. J. R. (2003). Influence Trade Patterns ? Testing Old and New Trade Theories. Damania, R., Fredriksson, P. G., & List, J. A. (2003). Trade liberalization, corruption , and environmental policy formation : theory and evidence, 46, 490–512.  Dasgupta, S., Laplante, B., Wang, H., Wheeler, D., Journal, T., & Winter, N. (2007). Confronting the Environmental Kuznets Curve, 16(1), 147–168. Dean, J. M. (1999). Does Trade Liberalization Harm the Environment? A New Test. Dechezleprêtre, A., & Sato, M. (2014). The impacts of environmental regulations on competitiveness. Dabla-norris, E., & Kochhar, K. (2015). Causes and Consequences of Income Inequality : A Global Perspective. IMF, 1–39.   Dinda, S. (2004). Environmental Kuznets Curve Hypothesis : A Survey, 49, 431–455.   Ekins P. et al., “Trade, environment and development: the issues in perspective”, Ecological Economics 9(1994), pp 1-12 Esty, D. C., Porter, M. E., & Porter, M. E. (2005). National Environmental Performance : An Empirical Analysis of Policy Results and Determinants.  68  Everett, G., & Wilks, A. (1999). The World Bank’ s Genuine Savings Indicator : a Useful Measure of Sustainability ? Glyn Everett and Alex Wilks, (October), 1–10. Fabiosa, J. F. (2008). What Effect Does Free Trade in Agriculture Have on Developing Country Populations Around the World ? Frankel, B. J. A., & Romer, D. (2016). Does Trade Cause Growth ?, The American Economic Review, Vol . 89, No. 3, pp. 379-399. Frankel, J. a., & Rose, A. K. (2005). Is Trade Good or Bad for the Environment? Sorting Out the Causality. Review of Economics and Statistics, 87(1), 85–91.  GATT (1971), Industrial Pollution Control and International Trade GATT (1992), Trade and the Environment Gowdy M. J., Walton M., “Sustainability concepts in ecological economics”, Economics interactions with other disciplines Gray, W.B. (1987). The cost of regulation: OSHA, EPA and the productivity slowdown, American Economic Review 77 (5), 998-1006. Grossman, G. M., & Krueger, A. B. (1991). Environmental Impacts of a North American Free Trade Agreement. National Bureau of Economic Research Working Paper Series, No. 3914(3914), 1–57. Gulati, S., & Kellenberg, D. (2013). Political Economy and the Interaction between International Trade and Environmental Policies. Encyclopedia of Energy, Natural Resource, and Environmental Economics, 3, 288–295.   Hanley, N., Dupuy, L., Mclaughlin, E., Hanley, N., Dupuy, L., & Mclaughlin, E. (2014). Genuine Savings and Sustainability Genuine Savings and Sustainability, (November). Jaffe, A. B., & Palmer, K. (2016). Environmental Regulation and Innovation : A Panel Data Study, The Review of Economics and Statistics, Vol. 79, No. 4 (Nov., 1997), pp. 610-619 Published by : The MIT Press Jayadevappa R., Chhatre S., 2000, “International trade and environmental quality: a survey”, Ecological Economics 32 (2000), pp 175-194 Kellenberg, D. (2010). CONSUMER WASTE, BACKHAULING, AND POLLUTION HAVENS I. Journal of Applied Economics, 13(2), 283–304.   69  Lanoie, P., Laurent-lucchetti, J., Johnstone, N., Ambec, S., Canada, B., Chabot, R., & Thornton, G. (2007). Environmental Policy, Innovation and Performance : New Insights on the Porter Hypothesis. Lanoie, P., & Patry, M. (2001). Environmental Regulation and Productivity : New Findings on the Porter Analysis. Lehmijoki, U., & Palokangas, T. (2010). Trade, population growth, and the environment in developing countries. Journal of Population Economics, 23(4), 1351–1370.   Levinson, A. (2007). Technology, International Trade, and Pollution from U.S. Manufacturing. Loh, J., Goldfinger, S., Bourne, G., Atkin, M., Mcrae, L., Ewing, B., Backer, W. De. (2008). LIVING PLANET REPORT 2008. Lumpkin, G. T. & Marvel, Matthew R. (2007). Technology Entrepreneurs’ Human Capital and Its Effects on Innovation Radicalness, (270), 807–828. Lydgate E., 2012, “Sustainable development in the WTO: From Mutual Supportiveness to Balancing”, World Trade Review, Vol.11, Issue 4, pp 621-639 Marjit, S., Mandal, B., & Roy, S. (2014). Trade Openness, Corruption and Factor Abundance : Evidence from a Dynamic Panel, 18(1), 45–58.   Mehrara, M., & Musai, M. (2013). The relationship between Economic Growth and Human Capital in Developing Countries, 5, 55–62.  Oleson, K. L. L. (2011). Shaky Foundations and Sustainable Exploiters : Problems with National Weak Sustainability Measures in a Global Economy.  Panayotou, T., Sachs, J., Panayotou, T., Peterson, A., & Sachs, J. (2000). Consulting Assistance on Economic Reform II Discussion Papers, 495–527. Perkins, R., & Neumayer, E. (2012). Does the “California effect” operate across borders ? Trading- and investing-up in automobile emission standards, (March), 217–237. Porter, M. E., Linde, C. Van Der. (2016). Toward a New Conception of the Environment-Competitiveness Relationship Published by : American Economic Association. Pregg, Ernest H. (1995)., Traders in a brave new world: the Uruguay Round and the future of the international trading system. University of Chicago Press Redding, S. (2016). Dynamic Comparative Advantage and the Welfare Effects of Trade welfare effects of trade, 51(1), 15–39. 70  Reppelin-hill, V., & Reppelin-hill, V. (1999). Trade and Environment: An Empirical Analysis of the Technology Effect in the Steel Industry, 301, 283–301. Røpke I., “Trade, development and sustainability – a critical assessment of the “free trade dogma”, Ecological Economics 9 (1994), pp 13-22 Rose, A. K. (2004). Is Trade Good or Bad for the Environment ? Sorting Out the Causality. Rossi, P., & Anderson, U. (2013). Omitted Variables, Instruments and Fixed Effects. Saarenheimo, T. (2016). Trade, Human Capital Accumulation and Growth in an Underdeveloped Economy Published by : Wiley on behalf of The Scandinavian, 95(4), 535–557. Sjo, F. (1999). Technology transfer and spillovers : Does local participation with multinationals matter? 43, 915–923. Smith, J. B., & Sims, W. a. (1985). The impact of pollution charges on productivity growth in Canadian brewing. RAND Journal of Economics, 16(3), 410–423. Spilimbergo, A., Londono, J. L., & Szekely, M. (1999). Income distribution, factor endowments, and trade openness, 59. Stern, D. I. (2004). The Rise and Fall of the Environmental Kuznets Curve, 32(8), 1419–1439.   Strauss, A. L., & Strauss, A. L. (2014). From GATTzilla to the Green Giant : Winning the Environmental Battle for the Soul of the World Trade Organization, 19(3). Talberth, J., & Bohara, A. K. (2006). Economic openness and green GDP, 58, 743–758.   Tewari, M., & Pillai, P. (2016). Global Standards and the Dynamics of Environmental Compliance in India’ s Leather Industry Global Standards and the Dynamics of Environmental Compliance in India ’ s Leather Industry, 0818(April). Torres-reyna, O. (2007). Panel Data Analysis Fixed and Random Effects longitudinal or cross- is a dataset in which the behavior of entities are observed across time. United Nations. (1992). Convention on biological diversity. Diversity, 30. Retrieved from   Urpelainen, J., & Urpelainen, J. (2016). A California Effect for International Environmental Externalities ? A California Effect for International, 0629(April).   Vogel, D., & Vogel, D. (2016). Trading up and governing across : transnational governance and environmental protection Trading up and governing across : transnational governance and environmental protection, 1763(April).   71  Wagner, M. (2003). The Porter Hypothesis Revisited : Empirical Tests. Yoo, I. T., & Kim, I. (2015). Economic integration and environmental cooperation. International Environmental Agreements: Politics, Law and Economics.   Yu, D. W., Sutherland, W. J., Yu, D. W., Sutherland, W. J., & Clark, C. (2016). Trade versus environment Trade versus environment, 5347(February).   Graciela Chichilnisky. (1994). North-South Trade and the Global Environment Source : The American Economic Review, Vol. 84, No. 4 (Sep., 1994), pp. 851-874 Published by : American Economic Association. Backus, David K. & Kehoe, Patrik J. & Kehoe Timothy J. (1992). In search of Scale Effects in Trade and Growth Zhiqi Chen. (2016). Long-Run Equilibria in a Dynamic Heckscher-Ohlin Model, Source : The Canadian Journal of Economics, Wiley on behalf of the Canadian Economics Association, 25(4), 923–943.  72  Appendices Appendix A  Intersection1 Estimation Results Dependent variable: GDP (% point increase compared to 1990)  Full sample group1 group2 group3 group4 estimation model TWFE TWFE TWFE GLS (i.year) GLS (i.year) Trade Intensityt-1 0.307*** 0.013 0.29** -0.03 0.32 Ln_Population 97.18*** 18.12 331.78*** -5.24(*) 61.3(***) Ln_Government spending 7.84*** 43.03*** 3.34 -36.61**(*) 95.65*** Ln_Household consumption 160.58*** 53.91*(*) 139.38*** 18.62(**) -214.43*** Ln_Education expenditure 18.32*** 31.97*** 69.16*** -8.03(***) -62.79**(*) Ln_Capital formation 33.03*** 70.52*** 27.18*** 35.19*(**) 114.12*** Hausman test 46.3*** 56.39*** 104.38*** 0.74 5.77 F-test(all ui=0) 253.66*** 98.55*** 174.87***   Breusch-Pagan test    278.79*** 158.81*** Joint F test (i.year=0) 12.84*** 3.92*** 5.14*** 2083.07*** 599.82***  Dependent variable: Consumption of Fixed Capital (% of GDP)  Full sample group1 group2 group3 group4 Estimation model FE GLS (i.year) GLS TWFE TWFE Trade Intensityt 0.01 0.02(**) -0.01 0.03*** 0.02*(**) Ln_Population -1.99*(**) 0.31 -0.94 15.49**(*) -11.14*** Ln_Capital stock -0.09 -0.4 0.64(*) 1.55*** -0.39(*) Hausman test 123.28*** 4.54 2.92 17.69*** 58.89*** F-test(all ui=0) 88.35***   96.61*** 117.88*** Breusch-Pagan test  2843.31*** 1724.9***   Joint F test (i.year=0) 1.25 87.97*** 13.76 3.17*** 2.19***  73  Dependent variable: Per Capita GNI (US$)  Full sample group1 group2 group3 group4 Estimation mode TWFE TWFE FE TWFE TWFE Trade Intensity 27.08*(**) 76.68**(*) 82.21*** 15.8*** 0.29 Ln_Population -28022.96*** -27404.71*** 23154.83*** -15951.5*** -6254.35*** Ln_Government spending -1500*** 1797.86 -228.73 -537.86 -309.29(**) Ln_Household consumption -120.02 15697.76*** 7696.63*** 2729.8*** 1306.56*** Ln_Education expenditure -1244.56*** 3316.09*** 2768.73*** 23.38 376.7*(**) Ln_Capital formation 1283.08(**) 10210.54*** -814.49 905.28*** 129.49 Hausman test 173.25*** 18.81*** N/A N/A N/A F-test(all ui=0) 32.8*** 7.15***    Breusch-Pagan test      Joint F test (i.year=0) 16.78*** 2.08*** 0.92 5.46*** 3.66***  Dependent variable: GINI Coefficient (%)  Full sample group1 group2 group3 group4 Estimation model TWFE TWFE FE TWFE FE Trade Intensityt-3 0.01 -0.017 0.017*(*) -0.06*** -0.03 Ln_Government spending 0.19 -0.06 0.21 -2.65* 2.45* Ln_Education expenditure -0.99*** -2.26*** -1.54*** 0.52 -1.07 Ln_FDI -0.02 -0.26** 0.11 -0.21 0.11 Hausman test 41.85*** 59.66*** N/A 18.42*** 18.83*** F-test(all ui=0) 157.04*** 43.16***  165.42*** 65.82*** Breusch-Pagan test      Joint F test (i.year=0) 3.32*** 1.95*** 1.00 1.82** 1.15  Notes (Common to Annex A to C): To conserve space, time fixed effects, standard errors, coefficients on time dummies and t-statistics are not shown. TWFE=Two-Way Fixed Effects, GLS (i.year )= Random Effects with yearly time dummy, FE=one way Fixed Effects,   74  Appendix B   Intersection2 Estimation Results a. All sample countries  PM Damage (US$) Resource Depletion (US$) CO2Damage (US$) Environmental Cost (US$)  TWFE TWFER TWFE TWFER GLS(i.year) GLS(i.year)R TWFE TWFER GDP Growth (GDPt/GDP1990)*100 1.89e+07*** 2.04e+09*** 5.70e+07*** 2.11e+09*** Consumption of Fixed Capital (% of GNI) -1.20e+07 -1.01e+09 8.20e+07 -9.04e+08 Per Capita GNIt-3 157225.3*** 1.84e+07*** 476254.2*** 476254.2** 1.91e+07*** Per Capita GNI2t-3 -0.826*** -87.353*** -2.277994*** -2.277994 -90.3722*** -90.3722** GINI Coefficientt-3 2.89e+07** 1.32e+09 1.10e+08*** 1.10e+08** 1.46e+09 Hausman test 14.74*** 20.05*** 4.98 19.46*** F-test (all ui=0) 146.39*** 7.45***   8.53*** Breusch-Pagan test   7797.34***  Joint F test (i.year=0) 16.77*** 17.76*** 532.68*** 18.00***  b. Developed Country  PM Damage Resource Depletion CO2Damage Environmental Cost  TWFE TWFER GLS(i.year) GLS(i.year)R FE FER GLS(i.year) GLS(i.year)R GDP Growth -3.25e+07*** 1.11e+08 4.96e+07*** 4.85e+07 Consumption of Fixed Capital 2.80e+08 -3.51e+09 7.01e+08 7.01e+08** -4.81e+09 Per Capita GNIt-3 149325.6** 149325.6 -6537402 -91614.56 -7653591 Per Capita GNI2t-3 -0.196 30.747 0.299 39.675 GINI Coefficientt-3 1.03e+08 1.37e+10*** 1.37e+10 6.07e+08* 6.07e+08 1.63e+10*** 1.63e+10 Hausman test 10.76** 3.92 7.84** 4.57 F-test (all ui=0) 197.94***  358.56***   Breusch-Pagan test   7.15***   13.54*** Joint F test (i.year=0) 242.59*** 50.71*** 0.3 47.74***  c. Mid-income Countries  PM Damage Resource Depletion CO2Damage Environmental Cost  GLS(i.year) GLS(i.year)R TWFE TWFER GLS(i.year) GLS(i.year)R TWFE TWFER GDP Growth -5041638*** -5041638 -1.48e+08*** -1.48e+08** -2184880** -2184880 -1.52e+08*** -1.52e+08** Consumption of Fixed Capital 4.78e+07** 4.78e+07 -9.52e+09*** -9.52e+09* -7.33e+07*** -7.33e+07 -9.58e+09*** -9.58e+09* 75  Per Capita GNIt-3 -144664*** -144664 -1.51e+07*** -201805*** -201805* -1.53e+07*** Per Capita GNI2t-3 3.436** 3.436 261.962*** 5.436*** 5.436** 261.69*** 261.69** GINI Coefficientt-3 3.83e+07*** 3.83e+07 -3.29e+09*** -3.29e+09* 7609349 -3.41e+09*** -3.41e+09* Hausman test 7.68 11.63** 5 11.61** F-test (all ui=0)  11.95***  13.01*** Breusch-Pagan test 1028.3***  1256.8***  Joint F test (i.year=0) 57.43*** 4.97*** 58.87*** 5.07***  d. Low-income Countries  PM Damage Resource Depletion CO2Damage Environmental Cost  GLS(i.year) GLS(i.year)R TWFE TWFER TWFE TWFER TWFE TWFER GDP Growth 2.71e+07*** 2.98e+09*** 8.30e+07*** 3.09e+09*** Consumption of Fixed Capital -2.66e+07 -5.47e+09 1.62e+07 -5.56e+09 Per Capita GNIt-3 11065.34 -8317739 -990648.5** -990648.5*** -1.15e+07 Per Capita GNI2t-3 7.495 2158.855 141.871*** 2460.55 GINI Coefficientt-3 3.10e+07*** 3.10e+07* -1.82e+09 9.16e+07*** -1.68e+09 Hausman test 4.49 30.44*** 10.76** 29.55*** F-test (all ui=0)  5.25*** 44.08*** 5.50*** Breusch-Pagan test 845.68***    Joint F test (i.year=0) 893.54*** 12.93*** 38.72*** 13.51***  e. Lowest income Countries  PM2.5 Damage Resource Depletion CO2Damage Environmental Cost  GLS GLSR GLS GLSR TWFER GLSR FE FER GDP Growth -1433271*** -1433271* 1955660 -210637*** 3528.614 -1057478 Consumption of Fixed Capital -1.07e+08*** -1.07e+08** -8.45e+08*** -8.45e+08 -1.98e+07** -2.62e+07* -1.14e+09*** Per Capita GNIt-3 3273100*** 3273100** 3.06e+07*** 1137760*** 773246.8** 3.83e+07*** 3.83e+07** Per Capita GNI2t-3 -651.125*** -651.125 -7660.945*** -7660.945 -335.761*** -199.389* -9858.331*** -9858.331* GINI Coefficientt-3 -2920059 4.21e+07 -1004302 1260218 9.49e+07 Hausman test 3.5 5.74 N/A 10.88** F-test (all ui=0)   241.32*** 35.23*** Breusch-Pagan test 497.95*** 1051.25*** 352.93***  Joint F test (i.year=0) 9.68 7.2 2.76*** 12.59 1.34  76   Appendix C  Compound Model Estimation Results a. All sample countries  PM Damage (US$) Resource Depletion (US$) CO2 Damage (US$) Environmental Cost (US$)  TWFE TWFER TWFE TWFER TWFE TWFER TWFE TWFER Trade Intensity (Export+Import)/GDP -2.78e+07*** -1.94e+09*** -4.11e+07*** -2.00e+09*** GDP Growth (GDPt/GDP1990)*100 1.91e+07*** 2.06e+09*** 5.78e+07*** 2.13e+09*** Consumption of Fixed Capital (% of GNI) 1.12e+07 1.03e+09 1.54e+08** 1.20e+09 Per Capita GNIt-3 171189.4*** 1.91e+07*** 583386.7*** 1.98e+07*** Per Capita GNIt-32 -1.019*** -99.749*** -3.115*** -103.171*** GINI coefficientt-3 2.98e+07*** 2.98e+07** 1.46e+09 9.72e+07*** 1.61e+09 Hausman test 13.13*** 34.14*** 18.48*** 30.57*** F-test (all ui=0) 146.4*** 8.11*** 71.37*** 9.06*** Breusch -Pagan test      Joint F test (i.year=0) 14.45*** 15.15*** 19.91*** 15.36***  b. Developed countries  PM Damage Resource Depletion CO2 Damage Environmental Cost  TWFE TWFER GLS(i.year) GLS(i.year)R GLS GLSR GLS(i.year) GLS(i.year)R Trade Intensity 3377265 1.09e+08 6427267 -1.46e+07 GDP Growth -3.23e+07*** 1.15e+08 4.40e+07 3.69e+07 Consumption of Fixed Capital 2.85e+08 -3.49e+09 5.69e+08 -4.89e+09 PerCapitaGNIt-3 149894.8** 149894.8 -6475541 -124792.5 -7261160  PerCapitaGNIt-32 -0.199 30.647 0.884 37.513 GINIcoefficientt-3 1.04e+08 1.44e+10*** 1.44e+10* 9.49e+08*** 9.49e+08 1.65e+10*** 1.65e+10 Hausman test 15.13*** 7.16 8.04 8.23* F-test (all ui=0) 169.79***    Breusch -Pagan test   6.73*** 449.57*** 11.87*** Joint F test (i.year=0) 2.82*** 48.92*** 25.93 45.70***  77  c. Mid-income Countries  PM Damage Resource Depletion CO2 Damage Environmental Cost  GLS GLSR TWFE TWFER TWFE TWFER TWFE TWFER Trade Intensity -3441673 -5.23e+07 -3828132** -3828132 -7.12e+07  GDP Growth 1285623** 1285623 -1.49e+08*** -1.49e+08** -1702698*** -1.54e+08*** -1.54e+08** Consumption of Fixed Capital -8.48e+07** -8.48e+07 -9.65e+09*** -9.65e+09** -6148817 -9.75e+09*** -9.75e+09** PerCapitaGNIt-3 39888.09 -1.49e+07*** -1.49e+07** -106136*** -106136* -1.50e+07*** -1.50e+07** PerCapitaGNIt-32 -0.606 258.349*** 258.349** -0.080 256.768*** 256.768** GINI coefficientt-3 -5.33e+07*** -3.24e+09** -3.24e+09* -3.20e+07** -3.34e+09*** -3.34e+09* Hausman test 0.78 25.77*** 12.68** 25.47*** F-test (all ui=0)  6.81*** 94.24*** 7.62*** Breusch -Pagan test 554.03***    Joint F test (i.year=0) 27.13 4.94*** 8.51** 5.05***  d. Low-income Countries  PM Damage Resource Depletion CO2 Damage Environmental Cost  GLS(i.year) GLS(i.year)R TWFE TWFER TWFE TWFER TWFE TWFER Trade Intensity -1.12e+07*** -1.12e+07 -8.45e+08 3.02e+07*** 3.02e+07** -8.04e+08 GDP Growth 2.75e+07*** 2.98e+09*** 8.30e+07*** 3.09e+09*** Consumption of Fixed Capital 2.11e+07 -3.05e+09 2.74e+07 -3.13e+09 Per Capita GNIt-3 -54770.93 -1.00e+07 -1015953*** -1.32e+07 Per Capita GNIt-32 12.330 1791.37 117.940*** 2063.832 GINI coefficientt-3 1.82e+07** 1.82e+07 -1.78e+09 1.15e+08*** 1.15e+08** -1.61e+09 Hausman test 6.92 52.54*** 72.28*** 48.56*** F-test (all ui=0)   6.54*** 37.33*** 6.61*** Breusch -Pagan test 707.28***    Joint F test (i.year=0) 818.43*** 11.05*** 50.99*** 11,63***  78  e. Lowest income Countries  PM Damage Resource Depletion CO2 Damage Environmental Cost  GLS GLSR FE FER TWFER GLS(i.year)R FE FER Trade Intensity -1.15e+07*** -1.15e+07** -7.75e+07** -1373060* -6165092 -8.97e+07** GDP Growth -528350.5* -528350.5 6619579 -162652.3** 201774.9 6045586 Consumption of Fixed Capital -9.79e+07*** -9.79e+07** -9.37e+08*** -1.81e+07** -5483206 -1.07e+09*** Per Capita GNIt-3 2715549*** 2715549** 3.02e+07*** 1043984*** 560739.8 3.39e+07*** 3.39e+07** Per Capita GNIt-32 -447.64*** -447.64 -7584.35** -7584.35 -307.233*** -77.615 -8250.623*** -8250.623 GINIcoefficientt-3 -476850.4 1.11e+08 1.11e+08* -432726.3 -1.74e+07** 1.13e+08 1.13e+08* Hausman test 6.6 15.07** N/A 99.78*** F-test (all ui=0)  30.42*** 230.42*** 34.88*** Breusch -Pagan test 638.26***  388.98***  Joint F test (i.year=0) 8.63 1.21 2.79*** 11.1*** 1.31   

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0314920/manifest

Comment

Related Items