Globalisation and the environment : the impact of increased openness of trade on environmental quality in OECD countries during 1993 till 2012

(1)

1 Faculty of Economics and Business

First Version

Globalisation and the Environment

The impact of increased openness of trade on environmental quality

in OECD countries during 1993 till 2012

June 15

th

, 2016

Abstract:

The last two decades have seen an unprecedented increase in international trade. Its implications for newly set environmental goals is currently a topic of great discussion. Various studies have investigated worldwide effects of economic growth and international trade on environmental quality, but specific attention to developed countries has been lacking. This study targets the OECD country group as a proxy for developed countries. By performing panel data regressions on three different measurements of environmental degradation (sulphur dioxide, carbon dioxide and nitrogen oxide) it tries to contribute to the debate over globalisation and the environment. In doing so it will make use of an environmental policy stringency indicator. Results indicate that international trade lowers sulphur dioxide and carbon dioxide levels and show an insignificant effect on nitrogen oxide levels. Whilst these are encouraging results, it must be stated that they heavily rely on the non-existence of the problem of pollution offshoring.

Bachelor Thesis Economics and Business Specialization: Economics and Finance

Name: Melle Albada

Student Number: 10555145

(2)

2

Statement of Originality

This document is written by Student Melle Albada who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

At the end of the twentieth century there has been a significant increase in international trade as a result of globalisation. The World Bank (2016a) has measured an increase in openness of trade, which is defined as the sum of imports and exports divided by GDP, of 61% for the Organisation for Economic Co-operation and Development (OECD) countries during the period of 1993 till 2012. At the same time there has also been a growing concern about the environmental sustainability of our planet. With the Paris Agreement of 2015, the first global action plan to deal with climate change in a legally binding manner, as the pinnacle.

With the expectation that both developed and developing countries will continue to economically integrate, it is important to understand the relation between economic globalisation and environmental quality. The various ways in which globalisation can impact the environment have led to ambiguity about its true effect (Huwart & Verdier, 2013). This ambiguity exists because international trade often induces opposing effects. As increased openness of trade, for example, might lower a country’s environmental standards in order to remain internationally competitive, but it could also stimulate ongoing innovation within a country (Frankel, 2009). A just understanding of this linkage between globalisation and environmental quality can assist in taking the correct measures to accomplish the goals set in the Paris Agreement.

Contemporary research suggests that developed countries see improvements in environmental quality as economic growth continues. This thesis sets out to investigate what role international trade has played in the advancement of environmental quality during the expansion of developed economies. As 32 of the 34 OECD countries are considered high-income countries, according to World Bank guidelines (The World Bank, 2016b), this is a good sample for investigating the mentioned link between globalisation and environment for developed countries. The central focus of this thesis is what the impact of the increased international trade on the environmental quality in OECD countries has been during the period of 1993 till 2012. Data will be analysed by performing multiple panel data regressions using sulphur dioxide, carbon dioxide and nitrogen dioxide as measures of environmental degradation. New in this type of research is the use of an environmental policy stringency indicator. This index is developed to map the harshness of environmental policy in all the investigated countries. Both its effect and usefulness in environmental quality research will be examined. With the OECD iLibrary as database there is reliable, extensive data available on both economic and environmental indicators coming from a single source.

(5)

5

Especially since most of the done research uses data from the period around 1990 and prior, this research can contribute to the present-day insights of the effect of openness of trade on environmental quality. Its focus on developed countries also adds an unexplored dimension considering other research has mostly used worldwide samples (including developing countries).

Firstly, this thesis takes a deeper look at the literature at hand. Theories of the ways in which openness to international trade affects the environment are summarized, the environmental policy stringency indicator is further introduced and it takes a look at possible pitfalls within this study. The second part elaborates on the methodology and data used to come to results with explanations of the dependent and independent variables. The third section presents the found results of the empirical analysis and how these are to be interpreted. Finally, a conclusion is drawn which combines the literature review and statistical results in order to answer the main question posed.

1. Literature review

The literature review sets out to create an academic basis for the empirical research that is yet to be done. The link between international trade and environmental quality is established by summarizing theories and looking at their empirical validation. It starts off by explaining the definition of environmental quality in developed countries. Next is the Environmental Kuznets Curve theory. This is followed by three economic growth mechanisms induced by international trade that influence the environment. Non-economic effects of openness of trade on environmental quality are also considered in addition to the direct economic effects. Then the environmental policy stringency indicator and its relevance for pollution levels is introduced. Finally, possible pitfalls of this research are taken into account.

1.1 Economic theoretical concepts

It is first of use to define the concept of environmental quality in a clear manner. The OECD terms environmental quality as “… a state of environmental conditions in environmental media, expressed in terms of indicators or indices related to environmental quality standards.” In practice these quality standards are often defined by domestic regulatory agencies. This thesis uses the US Environmental Protection Agency (EPA) as guidance. In multiple acts they have legislated various environmental indicators. For air quality this includes air pollution and greenhouse gases, water quality looks at domestic wastewater and water conditions for leisure activities and the land indicator observes biodiversity and endangered species. The exact

(6)

6

indicators that will represent environmental quality in the empirical analysis will be set out in the methodological section.

A first attempt to theoretically link economic growth to environmental quality was done in 1993 by Grossman and Krueger in the form of the Environmental Kuznets Curve (EKC). The EKC is in succession of Simon Kuznets’ original Kuznets curve, which depicts an inverted U-shaped relation between inequality and economic growth (Kuznets, 1955). In Grossman and Krueger’s assessment of the predicted environmental impacts of the North American Free Trade Agreement, they were the first to find that pollution problems tend to alleviate after a certain threshold per capita income is reached (Grossman & Krueger, 1993). This threshold (indicated by ‘A’ in Figure 1) creates an inverted U-shaped curve between per capita income, representing economic growth, and the measured flow of pollution. The curve would indicate that environmental degradation increases with growing income until this certain threshold is reached beyond which environmental quality improves.

Figure 1: Environmental Kuznets Curve (Source: socialist.wordpress.com)

Grossman and Krueger (1993) distinguish three separate economic growth mechanisms which are arguably affected by increased openness of trade: the scale, composition and technique effect.

Contemporary international trade theories imply an increase in output and overall welfare if a country engages in international trade (Krugman, Obstfeld, & Melitz, 2012). The scale effect is based on the thought that as economic activity increases, and the proportions in which goods are produced remain constant, pollution simply must also increase. This effect is seen as the driver behind the initial phase during which deterioration of environmental quality increases.

(7)

7

On the other hand, the composition effect can have a positive impact on environmental quality as countries move along the EKC. As an economy transitions from an agricultural to an urban system, it tends to first see increase environmental degradation. But once this energy intensive industry is replaced by a knowledge or technology based industry, pollution starts to drop (Dinda, 2004). Grossman and Krueger (1993) do argue that the composition effect can also have a negative impact on environmental quality if an expansion of production, due to increased trade, is based on a comparative advantage in environmental standards. Countries will then specialize in less strictly regulated activities leading to more pollution.

Thirdly, there is the technique effect. With liberalization of trade comes the transfer of more modern technologies. The firms that establish themselves in other countries might make use of more advanced technologies, which are often cleaner as advanced technologies usually incorporate growing awareness of environmental urgency (Grossman & Krueger, 1993). This argument however mostly holds for developing countries, because developed countries, including the OECD group, generally already have access to these more advanced technologies. Folmer, Gerking and Komen (1997) add that wealthier nations do have the tendency to spend more on environmental research and development. Research helps in identifying environmental problems and coming up with solutions for them. It also often is imperative for the introduction of environmental policy, resulting in a positive effect on environmental quality.

The improved environmental spending pattern, mentioned as part of the technique effect and observed in more prosperous countries, is a result of EKC theory. Namely that as income grows, people achieve a higher standard of living. A higher standard of living then increases demand for higher environmental quality. These preferences induce structural changes in economies that tend to reduce environmental degradation (Dinda, 2004). Barrett and Graddy (2000) argue that the people’s will to improve non-material standards must also have room to be implemented (i.e. democracy is needed to accomplish environmental quality improvements). This is where the government as an actor becomes important. As they do the actual implementation of economic and environmental policy, an incentive for them to satisfy the preferences of changing policy is imperative. Within developed countries, (re-)election is seen as the best incentive to realise this. Barrett and Graddy (2000) studied the effect of different degrees of political and civil freedoms on various environmental quality indicators. For air pollutants they find, after combining the political and civil freedoms into a single indicator, a monotonically decreasing effect of democratic freedoms on air pollution levels. A similar finding does not seem to hold for the case of water pollutants. In the same study by

(8)

8

Barrett and Graddy they find either insignificant effects of democratic freedoms on multiple water quality measures or even encounter a flipped sign on the freedoms coefficient.

Although these findings suggests that including a measure for political and civil freedoms would be useful, this is not necessarily the case for the OECD country group. Unlike earlier studies, which also included developing countries, most countries of this country group are regarded as developed and enjoy relatively good political systems. For measures of political and civil freedoms across countries, the dataset from the IV Project is often used. The IV Project has built a set of statistics with the purpose of providing input for comparative, quantitative analysis on authority characteristics. Democratic and autocratic patterns of authority and regime changes are coded for all independent countries. Countries receive a democracy score from 0 to 10 from which the autocracy score, also ranged from 0 to 10, is subtracted. This leads to a polity score range of -10 (fully institutionalized autocracy) to +10 (fully institutionalized democracy). With a minimal value of 6 and mean of 9,7 for the selected OECD country group within the chosen time frame, it is reasonable to assume that these small differences will not contribute to explaining variance of environmental degradation indicators.

1.2 Empirical findings and implications

Empirical validity of the scale, composition and technique effects was given by a study of Antweiler, Copeland and Taylor (2001). Data on sulphur dioxide (SO2) levels from the Global Environmental Monitoring System (GEMS) was used to estimate the magnitude of the different effects in cities around the world. They also included a country-specific openness of trade measure to assess how increased openness affects pollution concentrations. Overall their results show that increased openness of trade creates small but measurable changes in pollution concentrations by altering the pollution intensity of national output. Whilst the composition effect is negligible, increases in output and income created by trade lead to scale and technique effects that do influence pollution intensity. They estimate that a 1 percent increase in both output and income, inducing scale and technique effects, results in a 1 percent decrease in pollution concentrations (Antweiler, Copeland, & Taylor, 2001). These results, combined with the negligibility of the composition effect, indicate that freer trade is good for the environment. Frankel and Rose (2002) have also studied whether trade is a good or a bad thing for the environment. In their cross-country OLS and IV estimations, using worldwide 1990 data, they find beneficial effects of international trade on environmental quality. Estimations show negative coefficients for all selected air pollutants (sulphur dioxide, nitrogen oxide and particle matter), although not all were significant at the commonly used significance levels (Frankel &

(9)

9

Rose, 2002). For other types of environmental degradation results were mixed. Deforestation, rural clean water access and carbon dioxide were all negatively affected by increased openness of trade (mostly insignificant). They conclude that there is little evidence that trade has a detrimental overall effect.

Numerous studies have been performed to confirm the existence of the EKC. Grossman and Krueger (1995) concluded that the hypothesis that further economic growth deteriorates environmental quality can be rejected at the 5% level for countries with an income of $10.000 (in 1985 USD). This is based on both air quality and water quality indicators tracked by the GEMS. Bates et al. (1997) did a similar study using a panel data analysis for the period 1970-1992, also with the 1985 USD as base unit. Their results indicate that the EKC only holds for local air pollutants, such as SO2 and nitrogen oxide (NOx), whilst they discovered a monotonically increase for more global, or indirect environmental indicators like carbon dioxide (CO2). The threshold for SO2 was measured at $6.900 and for NOx at $14.700. A third study, which used the OECD State of the Environment database for national-level emissions data on developed countries, found a definitive decrease of SO2 at an income per capita of $10.000 and the threshold for NOx at $5.500, both in 1985 USD (Panayatou, 1993). Panayatou does however mention that the SO2 estimates are only for 33% of the total variations explained by per capita income.

Although there is substantial difference in the threshold estimates, most studies do find evidence for the existence of an EKC, at least for indicators with a direct local environment impact. During 1993, the start of the period of interest, 10 OECD countries had an income per capita below $10.000 in 1985 US dollars and the average of all members was at $18.503 (OECD, 2016). Some countries had not reached all estimated thresholds, but were in the process of doing so. In 2012, when the lowest per capita income of an OECD member country was measured at $16.959 (Mexico) and the average of the OECD countries was $37.299 (OECD, 2016), only Turkey and Mexico had not passed the estimated thresholds in 1985 US dollars. This would suggest that the further economic expansion within the OECD member countries, partially made possible by increased openness of trade, induced environmental quality to improve.

1.3 Non-economic effects of trade

Besides the named direct economic effects of increased openness of trade, there are also effects that arise at any given level of income. The beneficial effects of globalisation are referred to as the gains of trade hypothesis. Frankel (2009) in his survey on the impact of globalisation on

(10)

10

the environment describes two main positive effects. The first possibility is about both technological and managerial innovation. He claims that further opening up to trade increases the rate in which absorption of frontier technologies and the best forms of management take place. This is also the reason why trading countries often not only see a direct increase in real income, but also encounter a sustained increase of growth. Absorption effects arguably have a larger impact in developing economies, but should nevertheless occur among the different degrees of developed economies. Next to the absorption effect, Frankel also identifies the possibility of an international ratcheting up of environmental standards. An example known within the United States is described in order to further explain this possibility. A single leading or largest state introduces higher environmental product standards. As other states feel the need to adopt these new standards, all states may end up with similar, improved standards as well. This process could also occur on a global scale. If the US as a whole were to up their environmental standards, the EU might have to follow in order to remain appealing to their American customers.

The counterpart of this ratcheting up of international standards is the race to the bottom hypothesis. The race to the bottom outlines the scenario in which international trade and investment will put downward pressure on the environmental standards within a country and thus globally damage the environment (Frankel, 2009). Industry leaders are afraid of lower environmental standards from competitors abroad. These lower standards would allow foreign firms to have an international advantage solely due to domestic regulation. Producers then warn for a loss of sales, employment and investment. Levinson and Taylor (2001) empirically confirm that in the US, industries that experienced the largest rises in environmental control costs also experienced the largest increases in net imports. Results from empirical research on the existence of a race to the bottom scenario are however mixed. Some researchers argue that multinational firms focus more on labour costs and market access than local environmental regulation, but others have found evidence of an effect of environmental regulation on direct investment decisions (Frankel, 2009).

1.4 Environmental policy stringency

Fairly new in the field of environmental policies is the environmental policy stringency (EPS) index. With increasing awareness of the risks that extensive pollution brings, governments have started to adopt regulation that addresses issues such as air pollution and greenhouse gas emissions (Sauvage, 2014). Environmental policies have the ability to assist in long term sustainable growth and wellbeing. These policies try to alter decisions of private actors with

(11)

11

the aim to make markets achieve environmental goals that they do not attain unattended (Botta & Koźluk, 2014). The OECD has collected data on selected environmental policies, primarily related to climate and air pollution, to be able to observe the environmental policy stringency across countries over time and to relate EPS’ effects on economic performance.

The indicator itself is based on the degree of stringency of 14 environmental policy instruments primarily related to climate and air pollution. Environmental policy tools covered by the indicator include environmentally-related taxation, emission limits, renewable energy and energy efficiency support, information on deposit and refund schemes and subsidies to R&D (Botta & Koźluk, 2014). Index values range from 0 (not stringent) to 6 (highest degree of stringency). Botta and Koźluk (2014) define the environmental policy stringency indicator as a higher, explicit or implicit, cost of polluting or environmentally harmful behaviour. There is a distinction between the so called ‘stick’ (directly raises costs) and ‘carrot’ (rewards environmentally behaviour) type of policies. Taxes are part of stick policies and increase the value indicator; larger spending values on subsidising instruments fall under the carrot category and also increase the indicator value.

In order to make the overall structure of the EPS indicator more clear, Figure 2 is added. It shows what the EPS indicator consists of and what the weights of every single component are. Quantitative or qualitative information for each individual instrument is normalised when deemed necessary. Each component scores on a 0 to 6 scale, similar to the overall outcome of the EPS indicator, and is multiplied by its corresponding weight. Country scores are then aggregated for market and non-market based policies and further multiplied to generate an end value between 0 and 6.

(12)

12

Figure 2: Overview of the components of the environmental policy stringency index and their corresponding weights (Source: Botta and Koźluk, 2014)

The overall trend across OECD countries has seen a steady increase in the EPS indicator measures (shown in Figure 3). The expectation is that countries with higher environmental policy stringency will also experience lower levels of environmental degradation. This trend thus predicts that any environmental clean-up occurring in OECD countries can also partially be explained by the increase in environmental policy stringency.

Figure 3: Environmental policy stringency in OECD countries (Source Botta and Koźluk, 2014)

Composite indicator of environemtnal policy stringency Market-based policies Taxes CO2, NOx, SOx, Diesel Trading Scheme CO2, Renewable Energy

Certificates, Energy Effiency Certificates

Feed In Tariffs Solar, Wind

DRS Deposit & Refund Scheme

Non-market based policies

Standards Emission Limit Values on NOx

SOxand PMx

R&D Government Expenditure on

Renewable Energy 0,25 0,25 0,25 0,25 0,5 0,5 0,5 0,5

(13)

13

1.5 The problem of pollution offshoring

One potential pitfall of this research would occur if developed countries offshore pollution intensive production to developing countries, also known as the pollution haven hypothesis. Imagine that the empirical analysis shows a positive result of openness of trade on environmental quality. This result could simply be caused by the OECD countries importing their pollution intensive goods instead of actually proliferating more environmentally friendly demands. Unfortunately, the empirical research on the topic of pollution offshoring is limited. Levinson (2009) started a study for the period 1972-2001 after an increase in contradictory answers from the press and economic journals on the question whether the US enjoys a cleaner environment due to manufacturing processes abroad. He argues that if pollution offshoring were to be the case, the US would see an increase in the imports of polluting goods. His finding is that, besides no significant change in the mix of goods produced domestically, the US imports proportionally more clean goods and proportionally fewer polluting goods than was the case 30 years ago. This green shift suggest that the US has not been offshoring pollution. Levinson claims that the clean-up in US manufacturing instead was largely due to improvements in production techniques.

No pollution offshoring would be good news for the meaningfulness of this thesis. It is however far-stretched to extrapolate the findings in the US for the period 1972-2001 to the entire OECD country group for the period 1993-2012. Luckily, Brunel (2014) did a follow-up study for the EU for the period from 1995 till 2008, using Levinson’s help. In line with Levinson’s US results, she finds that the clean-up in the EU was mostly because of improvements in production techniques. Production in the EU surprisingly even moved in an opposite direction than pollution offshoring would predict. Brunel found that during the previous decade, the EU specialized more in pollution intensive goods and its imports increasingly consisted of less pollution intensive goods. Based on the studies of both Levinson (2009) and Brunel (2014) the conclusion would be that there is little support for the existence of the pollution haven hypothesis within the United States or the European Union. It must be stated that this conclusion comes from a single form of empirical research, from researchers at the same research institute and that the provisional rejection of the pollution haven hypothesis should be approached with caution.

1.6 Conclusion on literature review

This summary of the contemporary academic literature is used to gain insight in the vast theories regarding international trade and environmental quality. EKC theory, combined with

(14)

14

the economic growth mechanisms from Grossman and Krueger (1993), gives a first suggestion for the possible outcome for this study based on the position of the countries on the Environmental Kuznets Curve. The gains of trade and race to the bottom hypothesis can either assist or counter this prediction (Frankel, 2009). Empirical validation of EKC theory and the scale, composition and technique effects done by Antweiler et al. (2001) showed evidence of beneficial effects of international trade on environmental quality. Frankel and Rose (2002) confirm these findings. The trend in environmental policy stringency described by Botta and Koźluk (2014) indicates containment of air pollutants and greenhouse gasses and will be considered in the empirical research. Lastly, the potential pitfall of offshoring pollution was dealt with by studies of Levinson (2009) and Brunel (2014). Their research suggests that pollution offshoring is currently not the case in the US and the EU, but this result should be approached with care.

2. Methodology

The purpose of the methodological section is to set out the framework through which the research question is investigated. First up is a description of the model used in the panel data regression. Effects of the dependent variables, which represent environmental degradation, are described and how these effects are generated. Then all independent variables are discussed in their relation to the environmental damage measurements. Second is the data section. This covers the used databases and comments on their reliability and comparability. It will also elaborate on data selections made within the OECD country group. Last are two short tables with some descriptive statistics, which give insights in the observations of the variables used.

2.1 Model

Of interest for the empirical research is the entire current OECD country group, which represents developed countries. For this type of study, a panel data regression is the best technique to use. It allows for yearly observations within the selected countries and the increased number of observations improves the statistical significance of the estimated results. A similar approach has been used by Frankel (2009) in his report for the Sweden Globalisation Council, where he carried out a panel data regression on global data for the period of 1990 to 2004. Antweiler et al. (2001) also used this method during their estimation of the scale, composition and technique effects, which also included an openness of trade measure.

For every environmental degradation measure (set out in the following paragraphs), two regressions will be performed. Regression equations are interchangeable between the three

(15)

15

measures, because they all share a common set of independent variables. The first model will use the authentic regression equation, which only uses openness of trade and GDP per capita as independent variables. The second model adds the environmental policy stringency indicator. This distinction between the two models is made, considering the EPS indicator has not been used before in combination with openness of trade research. The regression equations that will be used for this panel data regression, go as follows:

𝐸𝑛𝑣𝐷𝑒𝑔_𝑖𝑡 = 𝛽₀+ 𝛽₁𝑂𝑜𝑇_𝑖𝑡 + 𝛽₂𝐺𝐷𝑃𝑝𝑐_𝑖𝑡+ 𝜀_𝑖𝑡 (1)

𝐸𝑛𝑣𝐷𝑒𝑔𝑖𝑡 = 𝛽0+ 𝛽1𝑂𝑜𝑇𝑖𝑡 + 𝛽2𝐺𝐷𝑃𝑝𝑐𝑖𝑡+ 𝛽3𝐸𝑛𝑣𝑃𝑜𝑙𝑆𝑡𝑟𝑖𝑡+ 𝜀𝑖𝑡 (2)

where 𝐸𝑛𝑣𝐷𝑒𝑔 is one of the three chosen measures of environmental degradation, 𝑂𝑜𝑇 stands for the openness of trade of a country (i.e. imports plus exports divided by GDP), 𝐺𝐷𝑃𝑝𝑐 is GDP per capita in US dollars and 𝐸𝑛𝑣𝑃𝑜𝑙𝑆𝑡𝑟 is a measure of the stringency of a country’s environmental policy. The 𝑖 is the country index; the 𝑡 is the year index.

The first air pollutant of interest is sulphur dioxide (SO2), as mentioned earlier. The EPA (1995) declares that SO2 is formed when fuel containing sulphur, such as coal and oil, is burned during smelting and other industrial processes. Exposure to high concentrations of SO2 can cause breathing, respiratory illness and aggravation of existing cardiovascular disease (EPA, 1995). SO2 is also a precursor of sulphates, which acidifies rivers and lakes, corrodes buildings and monuments and reduces visibility (EPA, 1995).

Best known as a cause for environmental degradation is carbon dioxide (CO2). CO2 is a greenhouse gas that is emitted through human activity. The main originators of CO2 emissions are the generation of electricity, transportation such as highway vehicles, air travel, marine transportation and rail (EPA, 1995). The IPCC (2007) states, with very high confidence, that excessive emissions of greenhouse gasses through human activity will lead to climate warming. Warming of the climate system is unequivocal, as is now evident from observations of increases in global average air and ocean temperatures, widespread melting of snow and ice, and rising global average sea level (IPCC, 2007).

A third environmental degradation measure is nitrogen oxide (NOx). NOx is formed when fuel is burned and is emitted by motor vehicles and electric utility facilities (EPA, 1995). According to the EPA (1995) it can lead to lung irritation and lower resistance to respiratory infections. NOx also plays a role in the atmospheric reactions that produce smog and contributes to acid rain and eutrophication in coastal waters. This in its turn produces a destructive environment to fish and other animal life (EPA, 1995).

(16)

16

In Section 1.1 it was argued that environmental quality also comprehends of water quality, wastewater and biodiversity. Unfortunately, the OECD iLibrary database does not contain sufficient data on these matters to run a reliable regression. The remainder of this study focusses on CO2, SO2 and NOx as environmental quality measures, but does acknowledge that there is more to environmental quality besides the analysis of air pollution and greenhouse gasses.

Openness of trade is the first independent variable in the regression equation. Its estimated coefficient will be of most interest for this study. It affects environmental degradation through the economic effects first described by Grossman and Krueger (1993), but also via non-economic effects displayed by Frankel (2009). These effects are thoroughly discussed in the literature review section.

Another independent variable is income, measured as GDP per capita. As follows from EKC theory, income has a strong theoretical and empirical base in determining a country’s pollution level. The inverted U-shaped relation suggests that after reaching a certain income threshold, environmental degradation starts to decrease. Since most included countries enjoy income levels above the empirically calculated thresholds for SO2 and NOx (Grossman & Krueger, 1995; Bates, Cole, & Rayner, 1997; Panayatou, 1993), it is assumed that a higher incomewill lead to lower environmental degradation. For CO2 this relation will not necessarily hold due to its characteristics of a more global environmental indicator (Bates, Cole, & Rayner, 1997).

The final independent variable that will be used is environmental policy stringency. Because EPS is a numerical representation of the imposed costs on environmentally unfriendly behaviour, higher EPS figures should in theory result in lower pollution levels. As can be seen in Figure 2, emphasis of the EPS indicator lies on air pollution and greenhouse gasses. SO2, CO2 and NOx are given a prominent position in taxation and emission limits. This makes it a particularly useful predictor for the chosen measures of environmental degradation.

The decision to use a random or fixed effects model is based on outcomes of the Hausman test. This test is designed to find the model with the highest efficiency. χ2_statistics and probabilities of the different test can be found in the Appendix and the type of model used during each regression is parenthesised in the regression outputs of Section 3 (RE for random effects and FE for fixed effects).

(17)

17

2.2 Data

The main database used to gather data for this study is the OECD iLibrary database. This database is the online library of the OECD and acts as a hub for all available data and papers published by the OECD. The data includes various environmental and economic indicators. Extracted for this panel of data are measures of CO2, SO2, NOx, GDP per capita in US dollars, openness of trade and environmental policy stringency.

Observations of CO2 per capita in the iLibrary database come from National Inventory Submissions to the United Nations Framework Convention on Climate Change (UNFCCC). The UNFCCC has composed a set of guidelines which are meant to guarantee transparency, consistency, comparability, completeness and accuracy for the yearly submissions of the national inventory. Measuring methodologies are based on default methodologies provided by the IPCC. Exemptions are made only when national methodologies reflect their national situation better, provided that they are compatible with the IPCC guidelines and are well documented and scientifically based (UNFCCC, 2013). This framework of reporting makes for reliable country comparisons. SO2 and NOx observations also find their origin in National Inventory Submissions with addition of measurements from the United Nations Economic Commission for Europe-EMEP (UNECE-EMEP) database. UNECE-EMEP reporting follows the same guidelines used by the UNFCCC and also imposes a selected amount of measuring methodologies (UN Economic Commission for Europe, 2014). Data on GDP per capita and the imports, exports and total GDP are all based on Aggregate National Accounts reporting to the OECD 2008 System of National Accounts (SNA 2008). The SNA 2008 aims to set international standards for the recording of economic activity within a country. It’s main goal is to ensure comparability of national accounts across countries. This makes it a useful source for data on per capita income and openness of trade. At last there is the environmental policy stringency indicator. The indicator covers 28 of the 34 OECD countries for the period 1990-2012, making the data appropriate for this study. Botta and Koźluk (2014) state that the index has its flaws in focussing mainly on greenhouse gasses and air pollutant related policies. This is however not a big issue, because these are the only forms of pollution that fall within the scope of this study.

Although the available data is comprehensive, it is not fully complete. Various variables have missing observations, which requires data selection to keep a balanced dataset. Israel has missing observations for all environmental degradation indicators across multiple years. These deficient measures make it unusable for the regression. Both Chile and Mexico have multiple

(18)

18

missing observations for SO2 and NOx and miss the last two observations for CO2. As a result both countries have to be dropped as well.

Finally, a distinction will be made between countries that have measures of environmental policy stringency and countries that do not. Because it has not been used as an explanatory variable for environmental degradation, separate regressions are performed in order to examine whether it is a useful addition. As the environmental policy stringency index is not measured for all OECD countries, this will lessen the number of countries used in this separate regression and thus limit the amount of observations. A summary of the countries included in both regressions is given in Table 1 below.

Standard list of countries

Countries with environmental policy stringency measures

Australia Luxembourg Australia Japan

Austria Netherlands Austria Netherlands

Belgium New Zealand Belgium Norway

Canada Norway Canada Poland

Czech Republic Poland Czech Republic Portugal

Denmark Portugal Denmark Slovak Republic

Estonia Slovak Republic Finland Spain

Finland Slovenia France Sweden

France Spain Germany Switzerland

Germany Sweden Greece Turkey

Greece Switzerland Hungary United Kingdom

Hungary Turkey Ireland United States

Ireland United Kingdom Italy

Italy United States

Japan

Table 1.1 and 1.2: List of all OECD countries included in the dataset and List of all countries with EPS measurements

2.3 Descriptive statistics

Variable Observati

ons Mean Std. Dev. Min Max

Openness of Trade

(in % of GDP) 580 86,2% 52,2% 15,9% 348,4%

Environmental

Policy Stringency 508 1,866 0,842 0,520 4,408

GDP per capita (in $) 580 $28.291,33 $12.959,01 $5.633,489 $90.888,71 Table 2: Descriptive statistics of all independent variables

(19)

19

Variable Observations Mean Std. Dev. Min Max

SO2 per capita (in kg) 580 26,891 27,728 1,337 141,651 CO2 per capita (in tonnes) 580 10,188 4,470 2,736 31,106 NOx per capita (in kg) 580 33,746 21,326 8,708 131,684

Table 3: Descriptive statistics of all dependent variables

3 Results

The result section is dedicated to the different regression outputs with sulphur dioxide, carbon dioxide and nitrogen oxide as dependent variable respectively. For every environmental degradation measure there is one model that abstains from using the EPS variable and one that includes the EPS variable. Each section gives a short analysis on the found coefficients, significance levels and how they compare to the posed theories. The final section is a short conclusion of all of the findings and attempts to give meaning to their joint implications.

3.1 Sulphur dioxide

Sulphur Dioxide (SO₂) per capita

Model 1 (RE) Model 2 (RE)

Openness of Trade -9,100 -22,622

(8,593) (12,287)*

Log GDP per capita -29,398 -13,384

(5,922)*** (9,046)

Environmental Policy Stringency -4,438

(2,243)* Constant 332,912 187,222 (57,453)*** (82,305)** Observations 580 508 R² (overall) 0,1778 0,2259 Prob > Chi² 0,0000 0,0000

Table 4: Determination of environmental degradation in the form of SO2 (standard errors in parentheses; ‘***’, ‘**’, ‘*’ indicate significance levels of 1%, 5% and 10% respectively)

(20)

20

The first examined results have SO2 per capita as dependent variable. Only for the model including an EPS indicator, openness of trade has a significant negative effect on environmental degradation. This is a mild indication that as developed countries increase their engagement in international trade, they will also experience reductions in sulphur dioxide levels. The result is similar to what Antweiler et al. (2001) find in their study on SO2. Along with the effect of openness of trade, increased income per capita also has a positive impact on environmental quality. In line with what EKC theory suggests, developed countries see pollution reductions as economic growth continues. However, this can only be concluded with confidence for the simple model. Last is the coefficient on environmental policy stringency. It is significant at the 10% level and has a negative sign. The result coincides with the intuition that stringent policy for air pollutants leads to a cleaner environment and is an indication that including an EPS measure should be considered in explaining environmental degradation. It also explains part of the negative effect that was first captured by income per capita and is the cause of the reduced significance of the estimated coefficient.

3.2 Carbon dioxide

Carbon Dioxide (CO₂) per capita

Model 1 (RE) Model 2 (FE)

Openness of Trade -1,000 -1,614

(0,282)*** (0,584)**

Log GDP per capita -0,074 1,504

(0,266) (0,526)***

(0,154)*** Constant 11,805 -2,985 (2,846)*** (4,813) Observations 580 508 R² (overall) 0,0938 0,2422 Prob > Chi² 0,0004 0,0000

Table 5: Determination of environmental degradation in the form of CO2 (standard errors in parentheses; ‘***’, ‘**’, ‘*’ indicate significance levels of 1%, 5% and 10% respectively)

Regression results on carbon dioxide per capita show a highly significant negative effect of openness of trade, similar to the sulphur dioxide findings. The gains of trade clearly outweigh

(21)

21

the race to the bottom or any scale and composition effects. This contradicts the results of Frankel (2009) and Frankel and Rose (2002) who found mostly significant positive effects of openness of trade on CO2 levels for their worldwide samples. It appears that the impact of increased openness of trade improves for the more developed countries. The estimated coefficient for the effect of per capita income on carbon dioxide levels in the model excluding EPS is negative, whereas the model including EPS as an explanatory variable estimates a positive effect. As stated in the CO2 variable description in Section 2.1, and earlier found by Bates et al. (1997), there might be a monotonically increasing effect of income on CO2 pollution. With the model excluding EPS finding a non-significant coefficient at the common significance levels and the included model finding a highly significant coefficient, there is another piece of empirical evidence that this monotonically increasing relationship exists. The EPS coefficient is negative and highly significant. This is also in line with the results for SO2. Moreover, it assisted in changing the sign of estimated income per capita coefficient and increased the significance of the income per capita coefficient.

3.3 Nitrogen oxide

Nitrogen Oxide (NOx) per capita

Model 1 (FE) Model 2 (RE)

Openness of Trade 4,720 3,378

(5,403) (6,383)

Log GDP per capita -14,592 -7,917

(3,921)*** (4,415)*

(0,965)*** Constant 177,675 116,899 (36,230)*** (40,206)*** Observations 580 508 R² (overall) 0,0558 0,0010 Prob > Chi² 0,0001 0,0000

Table 6: Determination of environmental degradation in the form of NOx (standard errors in parentheses; ‘***’, ‘**’, ‘*’ indicate significance levels of 1%, 5% and 10% respectively)

(22)

22

The final regression uses nitrogen oxide as dependent variable. Unlike what was found for both sulphur dioxide and carbon dioxide, nitrogen oxide shows a positive effect of openness of trade on environmental degradation. The effect is however not significant for both the simple model and the model that includes EPS. Combining the results of the EPS coefficients from the SO2 and CO2 regressions with the sign and significance of the NOx regression gives the impression that EPS is a good predictor of pollution levels. Income per capita has a significant negative effect on environmental degradation, similar to the findings of the SO2 regression. This result is, along with the negative coefficient of the SO2 regression, a confirmation of existence of the downward sloped part of the EKC for these air pollutants.

3.4 Conclusion

The outcome for the effect of openness of trade on the environmental quality measures is predominantly positive. Regressions on SO2 and CO2 show a clear negative impact of increased international trade on environmental degradation. NOx is the only measure that saw an increase in air pollution levels, but the estimated coefficient of the extensive and simple models is not significant.

Adding an environmental policy stringency measure to all regressions results in similar estimations. All coefficients have a negative impact on environmental degradation and are highly significant. This result confirms the theoretical intuition that more demanding environmental policy also leads to lower pollution levels. It seems that including an EPS measure as explanatory variable for environmental degradation should definitely be contemplated whilst conducting environmental quality research. The appropriate model to look at in the regression outcomes is then considered to be the model that includes EPS (i.e. Model 2). From looking at the outcomes of the elaborate models, in which openness of trade is either negative or non-significantly positive, there is sufficient evidence to say that increased openness of trade has a positive impact on environmental quality.

Findings on the income per capita coefficients coincide with findings in other studies of Panayatou (1993) and Grossman and Krueger (1995). SO2 and NOx pollution levels experience reductions as per capita income increases, which is empirical proof of the downward sloped part of the EKC. The positive estimated coefficient on CO2 can be explained by its reputation as a more global environmental indicator.

(23)

23

4. Conclusion

The main goal of this study was to find out what the impact of the increased openness of trade was on environmental quality in OECD countries during the period of 1993 till 2012. Because trade can influence environmental conditions in various ways, it was first appropriate to distinguish between the different effects and display what the academic literature has found out thus far.

Direct economic effects were considered first. Within the framework of the Environmental Kuznets Curve (EKC), established by Grossman and Krueger (1993), there are three separate economic growth mechanisms: the scale, composition and technique effect. As the OECD member countries are largely considered to be developed countries, empirical validation of the EKC suggests that their position on the EKC will be on the downward sloped part (Grossman & Krueger, 1995; Bates, Cole, & Rayner, 1997; Panayatou, 1993). This implies that economic growth experienced by increased international trade will lead to reduced environmental degradation. The net impact of increased openness of trade will then be decided by the addition (or subtraction) of the gains of trade and race to the bottom.

In order to estimate the impact of openness of trade on environmental quality, panel data regressions were used. The regressions focus on the overall effect and thus do not try to explain the individual effects described in the previous paragraph. Environmental quality is represented by per capita measures of sulphur dioxide (SO2), carbon dioxide (CO2) and nitrogen oxides (NOx) (i.e. environmental degradation). Unique is the use of an environmental policy stringency (EPS) indicator as independent variable. This indicator, based on the degree of stringency of 14 environmental policy instruments primarily related to climate and air pollution, in theory is a good predictor of environmental degradation.

Regression results indicate that the increased openness of trade has predominantly had a positive effect. Because inclusion of the EPS indicator was significant for all three pollutants, its sign corresponds with what theory suggests and it generally increased the significance of the other estimated coefficients, Model 2 is considered as leading. Among the Model 2’s, openness of trade had a significant negative effect on SO2 and CO2 levels and an insignificant positive effect on NOx. Combining these empirical findings leads to the conclusion that the net result of the direct economic effects, gains of trade and race to the bottom has had a positive impact on environmental quality in the OECD countries during the period of 1993 till 2012.

Also of interest are the estimated coefficients of income per capita. Coefficients in the SO2 and NOx regression indicate reductions in pollution levels as per capita income increases.

(24)

24

This is a confirmation of the downward sloped part of the EKC in developed countries and suggests that trade also has an indirect effect on the environment. The positive coefficient for carbon dioxide is in line with its reputation as a global externality and shows that there are also dangers in ever increasing economic growth.

Altogether these results indicate that developed countries can continue to spur their international trading patterns when it comes to retaining or improving environmental quality in the form of air pollution and greenhouse gasses. Special attention should be paid to how economic growth and the economic effects of trade evolve. It has potential to cause environmental degradation, but the combination of sound environmental policy that does not seriously interfere with economic growth mechanisms should have the ability to lead this world towards a cleaner environment.

It must be said that the findings of this study heavily rely on the rejection of pollution offshoring (i.e. importing pollution intensive goods from developing countries instead of domestically producing them) in the United States and Europe. The integrity of the research done by Levinson (2009) and Brunel (2014) is by no means questioned, but confirmation from other well regarded experts would be very welcome.

(25)

25

References

Antweiler, W., Copeland, B. R., & Taylor, S. M. (2001). Is Free Trade Good for the Environment? American Economic Review, 877-908.

Barrett, S., & Graddy, K. (2000). Freedom, Growth and the Environment. Environment and

Development Economics, 433-456.

Bates, J. C., Cole, M., & Rayner, A. (1997). The Environmental Kuznets Curve: an Empirical Analysis. Environment and Development Economics, 401-416.

Botta, E., & Koźluk, T. (2014). Measuring Environmental Policy Stringency in OECD Countries: A Composite Index Approach. OECD Economics Department Working

Papers, 1177.

Brunel, C. (2014). Pollution Offshoring and Emission Reductions in EU and US Manufacturing. SSRN Working Papers. Retrieved May 26, 2016, from http://dx.doi.org/10.2139/ssrn.2447679

Dinda, S. (2004). Environmental Kuznets curve hypothesis: A survey. Ecological Economics, 431-455.

Folmer, H., Gerking, S., & Komen, R. (1997). Income and environmental R&D: empirical evidence from OECD countries. Environment and Development Economics, 505-515. Frankel, J. (2009). Environmental Effects of International Trade. Stockholm: Sweden's

Globalisation Council.

Frankel, J. A., & Rose, A. K. (2002). Is Trade Good or Bad for the Environment? Sorting Out the Causality. The Review of Economics and Statistics, 85-91.

Grossman, G. M., & Krueger, A. B. (1993). Environmental Impacts of a North American Free Trade Agreement. Garber, P. (Ed.), The U.S. – Mexico Free Trade Agreement, 13-56.

Grossman, G. M., & Krueger, A. B. (1995). Economic Growth and the Environment.

Quarterly Journal of Economics, 353-377.

Huwart, J. Y., & Verdier, L. (2013). OECD Insights: Economic Globalisation Origins and

Consequences . Paris: OECD Publishing.

Krugman, P. R., Obstfeld, M., & Melitz, M. J. (2012). International Economics: Theory &

Policy. Boston: Pearson Education, Inc.

Kuznets, S. (1955). Economic Growth and Income Inequality. American Economic Review, 1-28.

Levinson, A. (2009). Offshoring pollution: Is the U.S. increasingly importing polluting goods? American Economic Review, 2177-2192.

Levinson, A., & Taylor, S. M. (2001). Trade and the Environment: Unmasking the Pollution Haven Effect. International Economic Review, 223-254.

OECD. (2016). Gross Domestic Product (GDP) Indicator. Retrieved June 3, 2016, from https://data.oecd.org/gdp/gross-domestic-product-gdp.htm

(26)

26

Panayatou, T. (1993). Empirical Tests and Policy Analysis of Environmental Degradation at Different Stages of Economic Development. Working paper: World Employment

Programme Research WP. 238.

Sauvage, J. (2014). The Stringency of Environmental Regulations and Trade in

Environmental Goods. OECD Trade and Environmental Working Papers(2014/03). Solomon, S., D, Q., Manning, M., Chen, Z., Marquis, M., Averyt, K. B., . . . Miller, H. L.

(2007). Contribution of Working Group I to the Fourth Assessment Report of the

Intergovernmental Panel on Climate Change. Cambridge: Cambridge University

Press.

The World Bank, & World Bank Indicators. (2016a). Trade (% of GDP). Retrieved from http://data.worldbank.org/indicator/NE.TRD.GNFS.ZS:

The World Bank. (2016b). Country and Lending Groups. Retrieved from http://data.worldbank.org/about/country-and-lending-groups

UN Economic Commission for Europe. (2014). Guidelines for Reporting Emissions and Projections Data. Retrieved from

https://www.unece.org/fileadmin/DAM/env/documents/2015/AIR/EB/English.pdf UNFCCC. (2013). UNFCCC reporting guidelines on annual inventories for Parties included

in Annex I to the Convention. Retrieved from

http://unfccc.int/resource/docs/2013/cop19/eng/10a03.pdf#page=2

US Environmental Protection Agency. (1995). AIR Trends Summary. Retrieved June 7, 2016, from https://www3.epa.gov/airtrends/aqtrnd95/

(27)

27

Appendix

Hausman tests on Random or Fixed Effects

χ2_Statistic _Probability _Indication

SO2 Model 1 1,510 0,469 Random Effects

SO2 Model 2 5,490 0,139 Random Effects

CO2 Model 1 2,200 0,333 Random Effects

CO2 Model 2 13,890 0,003 Fixed Effects

NOx Model 1 7,250 0,026 Fixed Effects

NOx Model 2 9,070 0,028 Fixed Effects

Table 7: Results from the performed Hausman tests Tests on serial correlation occurrence

F Statistic Probability Indication

SO2 Model 1 23,831 0,000 Serial correlation

SO2 Model 2 47,616 0,000 Serial correlation

CO2 Model 1 15,588 0,001 Serial correlation

CO2 Model 2 34,579 0,000 Serial correlation

NOx Model 1 124,275 0,000 Serial correlation

NOx Model 2 84,839 0,000 Serial correlation

Table 8: Results from tests on serial correlation Tests on heteroskedasticity

χ2_Statistic _Probability _Indication

SO2 Model 1 993,510 0,000 Heteroskedasticity SO2 Model 2 872,980 0,000 Heteroskedasticity CO2 Model 1 776,990 0,000 Heteroskedasticity CO2 Model 2 1284,710 0,000 Heteroskedasticity NOx Model 1 24000,000 0,000 Heteroskedasticity NOx Model 2 5173,550 0,000 Heteroskedasticity

Table 5: Results from tests on heteroskedasticity

As can be deduced from the test results on serial correlation and heteroskedasticity, all models encounter these problems. In order to deal with them, cluster standard errors are used in Stata (by using the command vce(cluster ‘panelID’).

Globalisation and the environment : the impact of increased openness of trade on environmental quality in OECD countries during 1993 till 2012