Producer and Consumer Responsibility in the Kyoto
Protocol: an Input-Output analysis of CO
2emissions
In this paper, we argue that carbon leakage and pollution haven effects might undermine the effectiveness of production-based emission targets, employed in the Kyoto Protocol and many environmental policies. We show how, when analyzing world CO2 emissions between 1995 and
2009, the producer and consumer responsibility approaches yield statistically different results. Finally, we argue that the production approach is limited insofar it does not weight in trade in emissions: the policy maker interested in reducing greenhouse gas emissions should then also consider employing a consumption approach to incentivize emission reduction in the most effective fashion.
Matteo Tesei (S3197077) m.tesei@student.rug.nl
MSc International Economics & Business Faculty of Economics and Business University of Groningen
Groningen, The Netherlands
Contents
Introduction ... 3
Literature Review... 7
Methodology, Data and Analysis ... 12
Methodology ... 12
Data ... 16
Analysis... 17
Discussion and conclusions ... 26
Graphs ... 30
References ... 34
Introduction
“The concept of global warming was created by and for the Chinese in order to make U.S. manufacturing non-competitive.”1
Donald Trump, 45th President of the United States of America
While developed countries have been struggling to contain and reduce their emission production of greenhouse gases, recently industrialized emerging countries such as China may seem to freely pollute to achieve economic growth. As we will show in this paper, when looking at domestic emissions production figures, developing countries seem to have been increasing their contribution to global warming incredibly quickly in the past decades – to a point when they started polluting more than the developed world. But in our increasingly interconnected reality, this begs the question whether developing countries are the main responsible for growth in greenhouse gas emissions – or rather, whether developed countries are fueling emissions in developing countries via international trade. Consequently, we also pose the question whether environmental policies limiting the production of emissions, such as the Kyoto Protocol, really help in containing pollution emissions.
In particular, this thesis will aim to investigate the adequacy of the Kyoto Protocol in limiting greenhouse gas emissions. As the Protocol’s cap system on signatories’ domestic gas emissions was calculated on the production rather than the consumption of emissions from the part of national economies, it will be argued that the Kyoto Protocol has failed in building an international consensus on shared tools that are indeed effective in tackling greenhouse emissions, and thus climate change. At a time when international solidarity towards global
1 While this quote dates back to 2012, before the beginning of President Trump’s mandate, it surely shows the
skepticism that some politicians present towards global warming. Moreover, the quote highlights how developed countries have been struggling to meet their environmental policy targets while maintaining the competitiveness of the national economy – when, at the same time, developing countries have been able to boost economic growth without the need to worry (under the Kyoto Protocol) about additional emission production. References: Trump, D.J. (realDonaldTrump), “The concept of global warming was created by and for the Chinese in order to make U.S. manufacturing non-competitive.”, 6th November 2012, 20:15, Tweet. Available at:
warming mitigation appears to falter, to scrutinize the methodological tools with which to limit anthropogenic air pollution seems to be of the essence.
More generally, this thesis will argue that the consumer responsibility approach, by taking into account international trade in emissions and the possibility for strategic interaction among countries laid out by the Kyoto Protocol’s cap system, better highlights the national responsibilities for greenhouse gas emissions. More specifically, although developed countries seem to have been reducing their emissions in the past decades, this paper will underscore how this reduction is partly to be attributed to the export of emissions from the part of developed countries to developing countries. Thus, the methodology employed in the Kyoto Protocol has not only laid out the very conditions for its inefficacy, but it has also encouraged developed nations to meet their targets by way of proliferating world’s gas emission, as opposed to encouraging greener production.
The Kyoto Protocol, an international treaty signed in 1997 and entered into force in 2005, was built on the consensus that global warming is a reality, and that CO2 emissions are
likely to be the main cause behind it. The goal of the Kyoto Protocol is to combat global warming by reducing the CO2 levels in the atmosphere. Such a goal is achieved by limiting the
CO2 production levels that signatories are allowed to reach. However, since it is recognized
that, due to historical reasons, developed countries are disproportionately more responsible for the current levels of CO2 in the atmosphere, different limits and responsibilities are allocated
to different countries. In particular, developed countries are identified as the group – called Annex I in the text of the treaty2 – being under the obligation to reduce CO
2. The treaty laid
out the details concerning the first commitment period, setting emission targets for the years between 2008 and 2012, after which the treaty has been further amended. On the contrary, no binding target was laid out for developing countries3, which therefore benefitted from a
favorable treatment under the Protocol.
The unique layout of the Kyoto Protocol, which imposes limits on the production of CO2 for developed countries, but not for developing countries, creates the opportunity for a
“carbon leakage” effect. Carbon leakage is defined as the increase of CO2 emissions in a given
country which results from the reduction of CO2 emissions in a different country, namely,
2 We will employ the term “Annex I”, used in the Kyoto Protocol, to identify the group of developed countries. 3 For the sake of simplicity, we will employ the term “non-Annex” to identify the group of countries that are not
part of Annex I, that is, those countries that are bound to limit their CO2 emissions. Thus, non-Annex countries
bound by more stringent environmental policies. Indeed, as developed countries must comply with the Kyoto Protocol production targets, developing countries are not limited by any target whatsoever. This means that developing countries will develop a comparative advantage in producing all those goods whose production process is highly polluting: as it becomes harder for developed countries to carry out such “dirty” production at home (due to the presence of the aforementioned targets), developing countries will be able to take on such processes relatively more easily. As a consequence, developing countries will become “pollution havens”, that is jurisdictions with fewer environmental restrictions compared to other countries. Thus, according to the pollution haven hypothesis, one would expect developed countries to make the cheapest choice, namely, to shift emission-intensive production processes to developing countries, and exploit strategically the absence of targets laid out by the Protocol for non-Annex states. In short, developing and developed countries may be accused of eluding the Kyoto protocol restrictions, to the greater detriment of the environment.
This research considers the layout of the Kyoto Protocol, and evaluates the “producer responsibility” method employed to measure carbon emissions against the alternative “consumer responsibility” approach. The producer responsibility method takes into account the production of CO2 emissions of each country: every country is thus held responsible for the
CO2 emissions that were produced in their national territory in a given year. Such a method is
employed in the Kyoto Protocol to define targets as well as to measure and account for countries’ emissions. On the contrary, the consumer responsibility approach considers the consumption of each country: every country is then held responsible for the CO2 emissions
necessary for the production of the final products consumed in the national territory, even if such emissions were produced in a third country. As will be shown, while the latter provides greater insight as to who is ultimately responsible for CO2 emissions, the producer
responsibility approach is a simpler tool at the policy maker’s disposal, which might explain why it was employed in the agreement.
improvements that could be adopted on the matter. This is achieved by calculating and comparing the CO2 emission levels of 41 countries over a 15-year period, employing both the
Literature Review
Serrano and Dietzenbacher (2010) have shown that, while it may appear that a country is directly producing fewer emissions, international trade allows said country to substitute local emissions with imports from abroad – thus avoiding local emission production, but nonetheless contributing to worldwide emissions from a demand perspective. As mentioned earlier, this could then undermine the effectiveness of emission restrictions based on national production.
In the past decades, a great share of emissions – as well as their growth – is to be attributed to developing countries when we take a territorial perspective (i.e. we measure producer responsibility), as proved by Peters et al (2011). On the contrary, the emissions produced in developed countries appear to have remained relatively stable or slightly decreased. Moreover, Raupach et al (2007) find that 73% of the growth in global emissions in 2004 is to attribute to production in developing and least-developed economies – although these same countries produced only 41% of global emissions in that same year. Indeed, this could lead us to believe that developing countries are the ones to be held responsible for growth in emissions. In the context of the Kyoto Protocol, which bound developed countries to bear the burden of reducing greenhouse gas emissions, one may argue that, having the West faced up to its responsibilities towards global environment, the Rest (Huntington) is now the one to be blamed for global warming.
However, when the international trade links between developing and developed countries are considered, one may argue that (at least part of) these results are due to a transfer of emissions from the second group to the first. If one were to consider consumer responsibility, they would find that developed countries are responsible for a greater amount of emissions compared to what the producer responsibility would indicate. As Wiedmannet al (2007) show, there has been extensive discussion in the literature regarding the best way to allocate emission responsibility. As an alternative to the production responsibility accounting method employed by the Kyoto Protocol, a different approach – where consumption responsibility is considered instead – has been proposed more than once4. The benefits of such an approach become
apparent when we consider the results of Davis and Caldeira (2010): they find that in 2004 23% of global CO2 emissions were traded internationally. Thus, consumption-based
accounting allows to include such traded emissions within the national responsibility of the countries directly consuming them. This can greatly improve the effectiveness of the Kyoto
4 For example, by Davis and Caldeira (2010), Fernández-Amador et al (2017), Peters and Hertwich (2008),
Protocol if we consider its design – since binding targets are set by the protocol only for developed countries (i.e. Annex I countries), and taking into account the international trade in emissions, a potential issue with the Kyoto Protocol is that global emissions may increase even when emissions produced by developed countries are decreasing, as suggested by Peters and Hertwich (2008).
In fact, Mongelli et al (2006) find how the absence of binding targets for non-Annex I countries may create a comparative advantage in production of emission-intensive products, resulting in a “carbon leakage” effect: that is production of emission-intensive products shifts from developed to developing countries as they become “havens” for dirty production processes. Similarly, Peters et al (2011) find that Europe and other developed countries – albeit being close to meeting their emission targets laid out by the Kyoto Protocol – in reality only transferred emissions to developing countries. Hence, as suggested before, developed countries did not truly reduce their emission consumption; on the contrary, they increased their carbon footprint. On an opposite trend, however, Levinson (2010) does not find a carbon leakage effect in the United States: instead, he observes a shift towards cleaner imported goods.
Therefore, one could argue, as anticipated in the introduction, that environmental policies focusing on national production levels – such as the Kyoto Protocol – may fail to account for the importing and exporting of emissions. Wilting and Vringer (2009) show that such policies will subsequently increase the environmental pressures towards weakly regulated countries, thus further limiting the effect of the environmental policies themselves overall. More specifically, Fernández-Amador et al (2017) find that the regulation stemming from the Kyoto Protocol was not effective in limiting not only consumption, but also production of carbon emissions. Finally, Peters and Hertwich (2008) argue that consumption-based inventories can solve the carbon leakage problem as well as help to efficiently reduce emissions from the consumption side, where the costs of such reductions are lowest and the impacts minimal.
This is also proven by Rijpkema (2014), who shows how “dirty” imports from developing countries have been increasing – thereby confirming the pollution haven hypothesis5.
As mentioned in the introduction, this thesis argues that the Kyoto Protocol has failed in fostering international norms that are able to cap greenhouse gas emissions effectively. In particular, the presence of a clear distinction between two categories of countries, among which only one is bound to limit their production of pollutants, as well as the adoption of a producer- rather than a consumer-based approach when it comes to the computation of such volumes of production, has laid out the ground for strategic interaction among developed and developing countries, to the detriment of the environment.
As shown earlier, a number of existing researches already support this argument. However, the existing literature is rather limited with respect to the number of countries and length of the time period analyzed. For instance, Serrano and Dietzenbacher’s (2010) methodology considers a two-region economy, which allows them to research the trade in emissions of Spain in years 1995 and 2000. While this is definitely an interesting analysis, it would be even more insightful to widen the considered database. With our research, we further extend such framework to consider a remarkably higher number of regions, as well as inspect data over a greater number of years. The goal is to obtain a bigger picture regarding production, consumption, and trade of carbon emissions globally.
Thus, this paper aims to widen the existing literature, namely by learning from the input-output approach employed by Leontief (1970), and utilizing it to investigate the efficacy of producer-based environmental policies. In this sense, this thesis wishes to enrich the literature on the matter by adding new and original evidence of the shortcomings of the Kyoto Protocol. While others have criticized the protocol before, researches and academics have often focused on specific countries or areas, and have failed to accompany their rightful criticism to the agreement with sufficient and satisfactory quantitative support. In particular, by expanding both the analyzed data and the number of years considered, and therefore by looking at the aggregate rather than at country-specific experiences, we are able to go beyond mere suspicions and adequately corroborate our thesis from a quantitative perspective. By the same token, we wish to provide the reader with some indications on where both research and policy-making should be directed.
To do so, it is necessary to inspect the levels of “dirty” emissions of the signatories to the protocol, and to consider whether they have increased or decreased after the adoption of the Kyoto Protocol. Although at a first glance Annex I countries seem to have reduced their emissions, by differentiating among producer- and consumer-based approach we will demonstrate that this has been a consequence of trade in emissions with non-Annex countries. In short, the protocol’s targets have been met only formally, rather than through the development of more environmentally friendly productive processes.
As anticipated, we will calculate the carbon footprints of 41 countries across the 1995-2009 period, using data from WIOD (Timmer et al, 2015) and an input-output framework6.
The research will then proceed with an analysis of the carbon footprint trends over time, and compare the results with the emissions produced locally.
To better answer our research question, we will answer the following secondary research questions:
Research Question 1A: have developed countries (i.e. Annex I) shown a decrease in emission production in the 1995-2009 period?
Research Question 1B: have developed countries (i.e. Annex I) shown an increase in emission consumption in the 1995-2009 period?
Research Question 2: are the coefficients of growth in emission consumption between developed and developing countries statistically different for the 1995-2009 period?
Research Question 3: has the trade in emissions from developing to developed countries been increasing in the 1995-2009 period?
With the results of such an analysis, we aim to understand whether there is a difference in growth of carbon production and consumption between Annex I and non-Annex I countries. Also, we are interested in analyzing the trends in emission production and consumption, and how effective the production approach can be compared to the alternative consumption approach. Based on the current literature, we would expect to find that the most developed countries have indeed been reducing their production of emissions, but that their emissions
6 The input-output framework is an approach widely used and recognized as appropriate for carbon footprint
Methodology, Data and Analysis
Methodology
For our analysis, we employ an input-output approach inspired by Leontief (1970). Such a framework allows us to consider the exact intermediate deliveries between the sectors of an economy, which in turn allows us to calculate not only the (emission) production of each industry, but also their (emission) consumption. Armed with this framework, we can then allocate the emissions to those responsible for the consumption, rather than the production of goods. This is vital for our analysis, as we can then compare the production responsibility and the consumption responsibility approaches to understand whether any differences are observable in our dataset. If that were the case, it would entail a remarkable shift in policy making and require an important change in how we conceive the international fight against global air pollution.
To help the reader in becoming familiar with the employed methodology, a typical input-output table is presented below:
Industries Final Demands Total
Industries Z f x
Primary inputs V
Total x'
Energy Industries
Emissions c'
by each industry7. Given these data, one can compute the input coefficient and the emission
coefficient of each industry through equations 1 and 2 respectively:
𝑨 = 𝒁𝒙̂−𝟏 (1)
𝒘 = 𝒄′𝒙−𝟏 (2)
Where 𝒙̂−𝟏 is the nn inverse diagonal matrix of the column vector x, and the nn matrix A contains the input coefficients, which represent the amounts of input necessary from one industry to another for the production of 1 unit of final output. In equation 2, the n elements of vector w are the emission coefficients, meaning the emissions generated for the production of 1 unit of final output of each industry.
Further, we extend Leontief’s framework to consider a multi-regional world, in a similar fashion to what Serrano and Dietzenbacher (2010) have done. Specifically, given the availability of the data8, we consider a world composed of 41 countries, each counting 35
sectors. These sectors produce goods which may be consumed as intermediate inputs by other (domestic or foreign) sectors or as final products by households or the government (at home or abroad). The above-mentioned framework can be described by the following equation, in matrix notation:
𝒙 = 𝑨𝒙 + 𝒇 (3)
where element xi of vector x is the total output of industry i, element Aij of matrix A is the input
coefficient from industry i to industry j, and element fi of vector f is the final demand for
industry i. We can then derive the solution of equation 3, given by:
𝒙 = (𝑰 − 𝑨)−1𝒇 = 𝑳𝒇 (4)
7 Vector c is defined as a satellite account, as it is not directly embedded in the main input-output table. In fact,
satellite accounts may also include data such as emissions, employment, water or land usage: the advantage of adding the carbon emissions satellite account to the main input-output framework lays in the possibility to calculate the carbon footprint of each industry and country.
8 Further details regarding the dataset, and why the choice is made for this specific dataset, can be found in the
where I is the identity matrix, and L is the Leontief inverse – defined as equal to (𝑰 − 𝑨)−1. With this result, and considering the emission coefficients w, we can calculate the intermediate deliveries of emissions of each world industry to each country, given by:
𝒆𝒎 = 𝒘̂𝑳𝒇 (5)
Where 𝒘̂ is the nn diagonal matrix of the transposed of vector w, and element emij of
matrix em is the amount of carbon emissions produced by industry i for the consumption of country j. It is then possible to calculate the carbon footprint of each country by taking the column sums of the em matrix (indicating all emissions necessary for consumption of country j) and adding the direct emissions by domestic households. Once the consumption-related emissions for each country are obtained, we can also compute the net CO2 emission imports,
namely, by taking the difference between the CO2 consumption and CO2 production.
In order to apply such equations to the WIOD data, we employ a numerical computing environment, allowing for matrix manipulation. While we will provide only a brief description of the performed calculations, the interested reader will find the code in the appendix.
First, matrices Z, f, x and c are imported from the database, as they are already available in the tables. In particular, the WIOD differentiates final demand between final consumption expenditure faced by households, non-profit organizations serving households, and the government; and gross fixed capital formation changes in inventories and valuables. However, our model does not consider different categories of final demand, as we only consider national responsibility vis-à-vis emissions, rather than that of specific stakeholders. Hence, we reshape the f matrix by taking the row sums, in order to obtain a vector containing the total final demand of each country in a given year.
one9. Once matrix A is obtained, we can compute the Leontief inverse L following the
definition mentioned above.
Further, we calculate w through equation 2, which in turn allows us to use equation 5 to produce matrix em. Finally, we calculate the column sums of em and add the direct emissions by domestic households to obtain the carbon footprint of each country in a given year. By applying the same procedure for each year considered, we obtain the data at the basis of our analysis.
With the results of our calculations (i.e. carbon footprint and net carbon imports) – together with carbon production – we compare the changes over time of CO2 consumption,
CO2 production and net CO2 imports between the developed and developing groups. This
serves the purpose of uncovering changes over time as well as differences between the Annex I group and the remaining countries. To test the research questions presented above, we employ different fixed effects regression models for panel data. By including an intercept for each country, we can control for any constant unobserved country-specific characteristic which may be correlated with the dependent variable, while by adding a dummy variable for Annex I we can compare differences between this group and the remaining countries.
We then consider the following equations:
𝐶𝑖𝑡𝐶𝑂2 = 𝛽 1𝑖+ 𝛽2(𝑎𝑛𝑛𝑒𝑥,𝑡)+ 𝑒𝑖𝑡 (6) 𝑃𝑖𝑡𝐶𝑂2 = 𝛽 1𝑖+ 𝛽2(𝑎𝑛𝑛𝑒𝑥,𝑡)+ 𝑒𝑖𝑡 (7) 𝑁𝐼𝑖𝑡𝐶𝑂2 = 𝛽 1𝑖+ 𝛽2(𝑎𝑛𝑛𝑒𝑥,𝑡)+ 𝑒𝑖𝑡 (8)
where CitCO2 is the national carbon consumption of country i in year t, PitCO2 the national carbon
production of country i in year t, NIitCO2 the net national carbon imports of country i in year t,
annex a dummy variable identifying Annex I countries (i.e. developed countries), and t the time variable indicating the corresponding year.
Equations 6 to 8 only contain dummy variables on the right-hand side. This choice is made as we are mainly interested in comparing the growth in CO2 emissions consumption,
9 While such a manipulation necessarily makes the employed data less accurate, it appears to be an acceptable
production, and net imports between the Annex I and the non-Annex groups. Moreover, one must consider that carbon emissions may vary due to a large number of variables, which may be very different in nature: for example, emissions may vary due to greener production processes or simply more efficient ones, changes in final demand, more stringent environmental policies, availability of resources, trade barriers, sectors of specialization. Since such variables are not always easy to define and obtain, and also considering that the available literature mainly focuses on an input-output approach rather than a regression analysis, we argue that our contribution can bring new insights to the existing research in spite of such a limitation.
Data
The data considered for our analysis derive from the work of Timmer et al (2015). The choice is made for this dataset as it is one of the largest publicly available multi-regional input-output tables collection. Indeed, it presents sectoral input-input-output data for 41 regions over a 15-year period. Moreover, it contains the relative carbon emissions for each sector, country and year considered – which allows for the calculation of carbon footprints, as shown in the previous section.
All in all, the dataset is a good fit for our research purposes, as it contains extensive and easily-accessible data for several countries and years. Although a more recent version of the considered dataset is available, it does not present environmental accounts data. Consequently, we are forced to exclude the most updated version, in that it would not allow us to test the presented research questions.
Annex I Non-Annex
Austria Finland Lithuania Slovenia Brazil
Australia France Luxembourg Spain China
Belgium Germany Malta Sweden India
Bulgaria Greece Netherlands Turkey Indonesia
Canada Hungary Poland United Kingdom Korea
Cyprus Ireland Portugal United States Mexico
Czech Republic Italy Romania Taiwan
Denmark Japan Russia Rest of the World
Estonia Latvia Slovak Republic
Note that the Rest of the World region is considered to include all other countries not otherwise mentioned.
Analysis
Tables 1 to 3 show the direct carbon emission production of the considered countries from year 1995 to year 2009. Table 1 presents national production of CO2 (in kilotons), table
2 shows the percentage contribution of each country to the total world yearly CO2 production,
and table 3 exhibits the percentage change of national CO2 production compared to the previous
year.
Table 1: production of CO2 emissions (kilotons)
Year Total Annex I Group Remaining countries
Table 2: production of CO2 emissions (% of world yearly production)
Year Total Annex I Group Remaining countries
(non-Annex) 1995 100.00% 58.65% 41.35% 1996 100.00% 58.23% 41.77% 1997 100.00% 57.98% 42.02% 1998 100.00% 57.86% 42.14% 1999 100.00% 57.76% 42.24% 2000 100.00% 57.49% 42.51% 2001 100.00% 57.27% 42.73% 2002 100.00% 56.13% 43.87% 2003 100.00% 54.91% 45.09% 2004 100.00% 52.58% 47.42% 2005 100.00% 51.15% 48.85% 2006 100.00% 49.69% 50.31% 2007 100.00% 48.67% 51.33% 2008 100.00% 46.84% 53.16% 2009 100.00% 44.98% 55.02%
Table 3: production of CO2 emissions (% change from previous year)
Year Total Annex I Group Remaining countries
(non-Annex)
1995 N/A* N/A* N/A*
1996 2.56% 1.83% 3.58% 1997 0.92% 0.49% 1.52% 1998 0.39% 0.17% 0.69% 1999 1.06% 0.88% 1.31% 2000 2.73% 2.25% 3.38% 2001 0.49% 0.12% 0.99% 2002 1.68% -0.35% 4.39% 2003 4.16% 1.89% 7.06% 2004 4.97% 0.51% 10.40% 2005 3.22% 0.41% 6.33% 2006 3.04% 0.10% 6.11% 2007 3.45% 1.32% 5.55% 2008 1.40% -2.41% 5.02% 2009 -2.63% -6.50% 0.78%
From Table 1, it is apparent that the world production of CO2 emissions shows an
increasing trend, with a faster growth in the 2003-2008 period compared to previous years. In 2009, however, carbon emissions were greatly reduced, reaching a lower level compared to that for 2007 emissions. To understand which countries are responsible for such a growth in emission production, we look at table 2: Annex I10 countries produced roughly 59% of world
emissions in 1995, thus leaving the remaining non-Annex countries11 responsible for 41% of
the emissions. The shares of emission production slowly but steadily changed over the years. As a way of example, in 2009, Annex I countries only produced as much as 45% of world emissions. With the help of Table 3, we can observe how Annex I countries have been producing fewer additional emissions every year, up until reducing their emissions (compared to the previous year) in 2008 and 2009. On the contrary, non-Annex I countries have been increasing their production emissions significantly in the 2002-2008 period, with a spike of 10% in 2004.
At a first glance, one could argue that developed countries (i.e. Annex I) have been progressively less responsible for increases in world CO2 emissions in the past years – even
contributing to the reduction of emissions in 2002, 2008 and 2009. On the contrary, developing (i.e. non-Annex I) countries appear to be those to blame for large increases in CO2 emissions
in the 2002-2008 period. During this time, the emission production levels of developing countries exceeded those of developed countries. However, we expect that the consumption responsibility approach will not necessarily present the same picture, as in the recent years it became increasingly easier and common for firms and consumers to outsource production and import goods from abroad. As a consequence, the national consumption and national production bundles may differ quite significantly. To further investigate this possibility, we proceed with our analysis by also considering countries’ consumption of emissions.
Tables 4 to 6 illustrate the carbon emission consumption of the considered countries from year 1995 to 2009. Table 4 shows the national consumption of CO2 (in kilotons), table 5
presents the percentage contribution of each country to the total world yearly CO2
10 In the employed dataset, the countries part of the Annex I group are: Austria, Australia, Belgium, Bulgaria,
Canada, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Japan, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Poland, Portugal, Romania, Russia, Slovak Republic, Slovenia, Spain, Sweden, Turkey, United Kingdom, United States.
11 In the employed dataset, the countries that are not part of the Annex I group are: Brazil, China, India, Indonesia,
consumption, while table 6 exhibits the percentage change of national CO2 consumption
compared to the previous year.
Table 4: consumption of CO2 emissions (kilotons)
Year Total Annex I Group Remaining countries
(non-Annex) 1995 2.204107 1.350107 8.540106 1996 2.260107 1.366107 8.945106 1997 2.281107 1.369107 9.117106 1998 2.290107 1.390107 8.999106 1999 2.314107 1.408107 9.059106 2000 2.377107 1.458107 9.190106 2001 2.389107 1.458107 9.307106 2002 2.429107 1.457107 9.715106 2003 2.530107 1.506107 1.024107 2004 2.656107 1.541107 1.115107 2005 2.741107 1.566107 1.175107 2006 2.824107 1.571107 1.254107 2007 2.922107 1.592107 1.330107 2008 2.963107 1.550107 1.412107 2009 2.885107 1.421107 1.464107
Table 5: consumption of CO2 emissions (% of world yearly production)
Year Total Annex I Group Remaining countries
Table 6: consumption of CO2 emissions (% change from previous year)
Year Total Annex I Group Remaining countries
(non-Annex)
1995 N/A* N/A* N/A*
1996 2.56% 1.17% 4.74% 1997 0.92% 0.26% 1.92% 1998 0.39% 1.51% -1.29% 1999 1.06% 1.32% 0.67% 2000 2.73% 3.55% 1.45% 2001 0.49% -0.01% 1.27% 2002 1.68% -0.05% 4.38% 2003 4.16% 3.34% 5.39% 2004 4.96% 2.29% 8.90% 2005 3.22% 1.68% 5.36% 2006 3.04% 0.28% 6.71% 2007 3.45% 1.33% 6.11% 2008 1.40% -2.59% 6.18% 2009 -2.63% -8.38% 3.68%
* not available, as the employed dataset does not include year 1994: it is thus not possible to calculate the change in 1995 compared to 1994.
With a closer inspection of Tables 4 and 6, one can observe how CO2 consumption in
Annex I countries did not vary much between 1995 and 1999, after which it shows a slightly increasing trend until 2007. In the last two years considered, Annex I countries managed to reduce their carbon emission consumption. Turning to non-Annex countries, these show a slightly increasing but rather stable CO2 consumption between 1995 and 2001. From 2002
onwards, however, the increasing trend takes a significantly steeper slope.
By looking at Table 5, we can see how Annex I countries were responsible for roughly 60% of the world CO2 consumption in the 1995-2002 period. After 2002, however, this group
of countries increasingly reduced their share of CO2 consumption: in 2009, they consumed
roughly as much as countries outside of the Annex I group.
To better compare CO2 production and consumption data, we calculate the difference
been producing more CO2 than they actually consume. By looking at changes over the years,
these discrepancies have been increasing in magnitude between 1998 and 2007, with a stronger increasing trend in the second half of this period. Only in 2008 and 2009 has this trend reversed its direction.
Thus, one could argue that, although developing countries have been strongly increasing their carbon emission production over time, this has been caused by the increasing gap between CO2 production and consumption in the Annex I group: since the West was
growingly unable to meet its consumption needs with its own production, it imported increasingly more CO2 emissions from the Rest. Only the last two years considered in our
analysis represent an exception, as a reversing trend is observed in this final period. Such a result may suggest that developed countries are finally introducing successful measures to reduce carbon emissions.
To further investigate this possibility, we then proceed with analyzing equations 6 to 8 to allow us to contrast the different trends over time of Annex I compared to non-Annex countries in matter of carbon consumption, production and net imports.
Table 7: equation (6), fixed effects model, consumption analysis
Growth rate of CO2 consumption in the Annex I Group 4463.411**
(1784.740) Growth rate of CO2 consumption outside of the Annex I Group 54374.150***
(3623.760)
Constant 505323.700***
(14553.740)
Observations 615
R-squared 0.0822
Notes: *** p<0.01, ** p<0.05, * p<0.1. Robust standard errors in parenteses. The table shows the results from estimating equation (6), with the national CO2 consumption levels between 1995 and 2009 as the dependent
variable.
The results of the first regression, using equation 6, are reported in Table 7. The rather low value of the R-squared indicates that the considered model cannot fully explain the data. As it happens, this is not an unexpected result, as the Annex I dummy variable and time variable are unlikely to sufficiently explain the carbon consumption levels12. However, it is worth
12 In fact, carbon emission levels are most likely also affected by variables that are not included in the employed
highlighting how the model is still statistically significant even at the 1% level (since Prob > F = 0). Further, we observe that the coefficient for developing countries (annex = 0) is highly significant at the 1% confidence level, while the coefficient for developed countries (annex = 1) is only significant at the 5% level. On the contrary, when we look at the coefficient values, it is clear that developed and developing countries have – on average – both increased their carbon consumption over the 1995-2009 period (given the positive coefficient values). Nevertheless, this increase has a significantly larger magnitude for developing countries.
All things considered, as developed countries do show an increase in emission consumption, such result successfully answers research question 1B. Further, research question 2 is also confirmed, since the 95% confidence intervals for the coefficients do not overlap. Indeed, this proves that developed and developing countries have been increasing their CO2
consumption with statistically different trends over the analyzed time period.
Table 8: equation (7), fixed effects model, production analysis Growth rate of CO2 production in the Annex I Group 1682.465
(2128.497) Growth rate of CO2 production outside of the Annex I Group 65847.210***
(4322.997)
Constant 505327***
(17362.140)
Observations 615
R-squared 0.1165
Notes: *** p<0.01, ** p<0.05, * p<0.1. Robust standard errors in parenteses. The table shows the results from estimating equation (7), with the national CO2 production levels between 1995 and 2009 as the dependent
variable.
Table 8 shows the results of the next regression of our analysis, based on equation 7. Not surprisingly, we find results that are analogous to those for the first regression, when it comes to the R-squared and the overall significance of the model. The coefficient for developing countries is, again, positive and highly significant at the 1% level, despite the fact
values. To obtain a more powerful model, one could include more relevant variables to explain changes in CO2
emissions. However, this is not necessarily the goal of our analysis, rather we are interested in comparing the trends over time between the Annex I and the non-Annex groups. In fact, CO2 emissions may vary due to a great
that the coefficient for developed countries is now not statistically different from zero, and therefore not statistically significant in the considered equation. These results indicate that, while developing countries show an increasing trend in their carbon production, developed countries do not show a statistically significant trend in CO2 production between 1995 and
2009. Consequently, research question 1A is answered as well.
It is interesting to compare the results of these two regressions, namely, to understand whether the CO2 consumption and production measures are statistically different, and whether
our suggestion that a production responsibility approach to emission reduction is indeed inferior to the alternative consumption approach. In fact, developing counties seem to be increasing their carbon production to a faster pace than they are increasing their consumption – given the larger coefficient in Table 8 compared to Table 7. At the same time, when we look at the production responsibility, developed countries seem not to be contributing to increases in carbon emissions. However, this is most definitely not true when consumption is taken into consideration: developed countries have in fact been increasing CO2 consumption, as shown
in Table 7.
Table 9: equation (8), fixed effects model, net imports analysis
Growth rate of net CO2 imports in the Annex I Group 2780.946***
(791.853) Growth rate of net CO2 imports outside of the Annex I Group -11473.060***
(1608.260)
Constant -3.264603
(6459.139)
Observations 615
R-squared 0.1209
Notes: *** p<0.01, ** p<0.05, * p<0.1. Robust standard errors in parenteses. The table shows the results from estimating equation (8), with the national net CO2 import levels between 1995 and 2009 as the dependent
variable.
Discussion and conclusions
The presented results seem to confirm our suspicion that the Kyoto Protocol policies, which are based on production responsibility and only limit emission production in Annex I countries, are not appropriately specified. In fact, such a policy design allowed developed countries to increase their carbon consumption, albeit the absence of any statistically significant change in their carbon production in the 1995-2009 period.
As seen, although much can be concluded by analyzing the available data for the 15-year period, the employed dataset is still limited insofar as it does not allow us to exhaustively observe the full trend over the years, and, in particular, its changes over time. In fact, in the very final years of the selected time period, we can observe a quite different trend compared to those before. The Annex I group produced less CO2 emissions than the remaining countries for
the first time in 2006, and their difference in production widened until 2009. Moreover, non-Annex countries increased their consumption of CO2 emissions rather significantly in the final
years of the 1995-2009 period. The combination of these trends leads to the reversion of the previously increasing trend in Graph 7: while the Annex I group more than tripled its net CO2
imports (i.e. the difference between CO2 production and CO2 consumption) between 1997 and
2007, there is a substantial decrease in this value in 2009. Hence, accessing to data for periods after 2009 would be of high interest, notably to observe the behavior of these trends all along the remaining commitment periods laid out by the Kyoto Protocol.
Further, it is worth mentioning that the WIOD (i.e. the database employed for this analysis) also includes data for the emissions to air of CH4, N2O, NOX, SOX, CO, NMVOC
and NH3. Yet, our analysis focused on the CO2 measures as this is one of the most aggressive
(in terms of greenhouse effect) and most industrially produced gas among the options13. Further
research could verify if the results obtained through the presented analysis can be replicated with the remaining available data.
Despite the presented limitations, one can observe from the analysis above that the consumption and the production approaches yield rather different results.
13 As CH
4 is a greenhouse gas even more aggressive than CO2, a policy maker willing to tackle the climate change
issue should be interested in containing CH4 production too. However, this gas is particularly difficult to measure
and it is not as much of an industrial pollutant as CO2. These considerations lead us to prefer the CO2 measure for
When we measure production responsibility, it is obvious how most of the increase in global CO2 emissions is to be attributed to developing countries – while they were accountable
for roughly 41% of global emissions in 1995, they produced increasingly more CO2 emissions
than the Annex I group in all years between 2006 and 2009. Developed countries, on the contrary, have been increasing their CO2 emission production to a much slower pace, and they
have actually succeeded in reducing their production in 2008 and 2009.
By looking at consumption responsibility, we can still observe a more rapidly increasing trend for CO2 emission consumption in developing countries compared to developed
nations – although the Annex I group has always consumed more than 50% of world CO2
emissions in all years except 2009. This suggests that developed countries have been indirectly importing CO2 emissions, and increasingly so until 2007. Only in 2008 and 2009 have
developed countries showed a substantial reduction in net emission imports from the non-Annex I group.
Hence, the production approach seems to be biased in favor of Annex I countries when compared to the consumption approach: as trade in emissions from developing to developed countries has actually been increasing, developed countries are not necessarily succeeding in cutting global emissions. On the contrary, they appear to be (at least partially) substituting away domestic emission production by exporting it to the developing world. Such results suggest a carbon leakage and a pollution haven effect: as developed countries, which must uphold stricter emission policies as a consequence of their participation in the Annex I group, struggle to meet their targets, they seem to be – at least partially – exploiting the cheaper costs and less stringent regulations that characterize production in developing countries, to the detriment of global emission production overall. As a result, this leads us to question the efficacy of the Kyoto Protocol framework in reducing global CO2 emissions.
However, it is also worth noting that, while the Annex I group has been importing increasingly higher quantities of CO2 emissions, emission production has been increasing
countries)14, their weaker environmental policies, as well as a stronger focus on the
manufacturing industries (as opposed to the services sector in developed countries)15.
All in all, although it may seem that the Annex I group formally complied with the targets laid out by the Kyoto Protocol, it has been proved that the employed production responsibility approach presents some faults. That being said, we must consider that measuring and obtaining emissions consumption data is no effortless task: indeed, this requires the recording of extensive industry-specific production data of output and emissions, as well as calculations with large matrices. While the latter can be done relatively easily today, collecting extensive and reliable data remains expensive. Moreover, we must consider that when the Kyoto Protocol was drafted and ratified – although the Input-Output approach employed in the above analysis had already been published by Leontief (1970) – the computing power and overall progress of the available technology made such Input-Output approach neither a precise nor a straightforward instrument to use. For these reasons, it may not have been unreasonable to employ a production responsibility approach for the Kyoto Protocol targets, notwithstanding its limitations.
Yet, one may also argue that such technical choice, which presents clear weaknesses from a methodological perspective when it comes to provide accurate measurement of pollution, had not been taken by accident. It may have been that the signatories, developed and developing, had no incentive to adopt more stringent rules, which would have prevented a mutually beneficial, albeit only in the short them, trade in emissions among countries. Be that as it may, we cannot but consider that the adoption of a producer-based approach has weakened the protocol’s ability in tackling climate change, to the profit of signatories and to the detriment of the environment.
In conclusion, we must acknowledge that the Kyoto Protocol has been a success in promoting an international effort to tackle global warming and stimulating policies aimed at reducing greenhouse gas emissions, although its target specifications and measurement mechanisms can be arguably improved. In fact, the distinction of targets between the Annex I
14 This is shown, among others, by Raupach et al (2007).
15 In fact, the manufacturing sector is typically more energy-intensive than the services sector. This is even more
and non-Annex countries, together with the choice of a production responsibility approach, created the opportunity for carbon leakage and pollution haven effects.
While it must be recognized that the Flexibility Mechanisms16 laid out by the Kyoto
Protocol might have had a mitigating impact, the most effective way to reduce carbon emissions is probably by eliminating the carbon leakage opportunity altogether. This is surely the intent of the Paris Agreement, effective as of 2016 and which sets goals starting from 2020: in this agreement, the distinction between developed and developing countries is absent. Rather, each country determines and plans its own efforts to reduce carbon emissions, so that all countries have targets to strive to. Such a different layout will surely assist in eliminating (or at least, reducing) the practice of substituting domestically produced CO2 emissions with
imports from developing countries, carried on by the Annex I group. In fact, developing countries will not be able to increase their emissions production as easily when they are also constrained by their own targets.
However, the Paris Agreement still does not tackle the faults of a production responsibility approach, leaving measurement and reporting to the individual signatories. Further research could then verify whether the results and limitations identified in our analysis can be replicated with more recent data, and whether the consumption responsibility approach remains a better instrument to reduce greenhouse gas emissions.
16 The Flexibility Mechanisms allow Annex I parties to meet their emission targets by contributing to emission
Graphs
0 5000000 10000000 15000000 20000000 25000000 30000000 35000000 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009Graph 1: production of CO2 emissions (kilotons)
Total Annex I Non-Annex
0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Graph 2: production of CO2 emissions (% of world yearly
production)
-8.00% -6.00% -4.00% -2.00% 0.00% 2.00% 4.00% 6.00% 8.00% 10.00% 12.00% 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Graph 3: production of CO2 emissions (% change from
previous year)
Total Annex I Non-Annex
0 5000000 10000000 15000000 20000000 25000000 30000000 35000000 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Graph 4: consumption of CO2 emissions (kilotons)
0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Graph 5: consumption of CO2 emissions (% of world yearly
production)
Annex I Non-Annex -10.00% -8.00% -6.00% -4.00% -2.00% 0.00% 2.00% 4.00% 6.00% 8.00% 10.00% 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009Graph 6: consumption of CO2 emissions (% change from
previous year)
-2000000 -1500000 -1000000 -500000 0 500000 1000000 1500000 2000000 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Graph 7: Difference between CO2 Production and
Consumption (kilotons)
References
Davis, S. J. and Caldeira, K. (2010) “Consumption-Based Accounting of Co2 Emissions,” Proceedings of the National Academy of Sciences, Vol 107, No 12, 5687-5692. doi: 10.1073/pnas.0906974107
Fernández-Amador, O., Francois, J. F., Oberdabernig, D. A. and Tomberger, P. (2017) “Carbon Dioxide Emissions and Economic Growth: An Assessment Based on Production and Consumption Emission Inventories,” Ecological economics, 135, pp. 269–279
Leontief, W. (1970), “Environmental Repercussions and the Economic Structure: An Input-Output Approach,” The Review of Economics and Statistics, 52(3), 262-271. doi:10.2307/1926294
Levinson, A. (2010) “Offshoring Pollution: Is the United States Increasingly Importing Polluting Goods?,” Review of Environmental Economics and Policy, 4(1), pp. 63–83. doi: 10.1093/reep/rep017
Mongelli, I., Tassielli, G. and Notarnicola, B. (2006) “Global Warming Agreements, International Trade and Energy/carbon Embodiments: An Input-Output Approach to the Italian Case,” Energy Policy, 34(1), pp. 88–100. doi: 10.1016/j.enpol.2004.06.004
Peters, G.P. and Hertwich, E.G. (2008) “Co2 Embodied in International Trade with Implications for Global Climate Policy,” Environmental science & technology, 42(5), pp. 1401–7
Peters, G. P. and Hertwich, E. G. (2008) “Post-Kyoto Greenhouse Gas Inventories: Production Versus Consumption,” Climatic Change: An Interdisciplinary, International Journal Devoted to the Description, Causes and Implications of Climatic Change, 86(1-2), pp. 51– 66. doi: 10.1007/s10584-007-9280-1
Sciences of the United States of America, 108(21), pp. 8903–8. doi: 10.1073/pnas.1006388108
Raupach, M.R., Marland, G., Ciais, P., Le Quéré, C., Canadell, J.G., Klepper, G. and Field, C.B. (2007) “Global and Regional Drivers of Accelerating Co2 Emissions,” Proceedings of the National Academy of Sciences of the United States of America, 104(24), pp. 10288–93
Rijpkema, J. (2014) “Examining the pollution haven hypothesis, ‘dirty’ industries in the world economy from 1995 until 2009,” MSc Thesis International Business & Management, University of Groningen, http://irs.ub.rug.nl/dbi/542557c75b229
Serrano, M. and Dietzenbacher, E. (2010) “Responsibility and Trade Emission Balances: An Evaluation of Approaches,” Ecological Economics, 69(11), pp. 2224–2232
Starrenburg, J. (2016) “Future climate change negotiations: Expanding the Triptych approach taking a demand perspective,” MSc Thesis International Economics & Business, University of Groningen, http://irs.ub.rug.nl/dbi/57837002a09f7
Timmer, M., Dietzenbacher, E., Los, B., Stehrer, R. and de Vries, G.J. (2015) “An Illustrated User Guide to the World Input-Output Database: The Case of Global Automotive Production,” Review of International Economics, 23(3), pp. 575–605
United Nations (1997) Kyoto protocol, New York: United Nations, https://unfccc.int/resource/docs/convkp/kpeng.pdf
Wiedmann, T. (2009) “A Review of Recent Multi-Region Input-Output Models Used for Consumption-Based Emission and Resource Accounting,” Ecological Economics, 69(2), pp. 211–222. doi: 10.1016/j.ecolecon.2009.08.026
Appendix
Matlab code (shortened)
%CO2 Emissions to Air, years 1995-2009
W_CO2_1995 = [xlsread('X:\My Desktop\Thesis\Data\AIR_may12\AUS_AIR_May12.xls',
'1995', 'C42') ...];
%the same procedure is the followed for every other year in our dataset
W_CO2 = [W_CO2_1995; W_CO2_1996; W_CO2_1997; W_CO2_1998; W_CO2_1999; W_CO2_2000; W_CO2_2001; W_CO2_2002; W_CO2_2003; W_CO2_2004; W_CO2_2005; W_CO2_2006;
W_CO2_2007; W_CO2_2008; W_CO2_2009];
save('X:\My Desktop\Thesis\Matrices\W_CO2.mat','W_CO2');
%Carbon Footprints 1995
Z_1995 = xlsread('X:\My Desktop\Thesis\Data\WIOT_row_apr12\wiot95_row_apr12.xlsx',
'WIOT_1995', 'E7:BCI1441');
x_1995 = xlsread('X:\My Desktop\Thesis\Data\WIOT_row_apr12\wiot95_row_apr12.xlsx',
'WIOT_1995', 'BKG7:BKG1441')';
F_1995 = xlsread('X:\My Desktop\Thesis\Data\WIOT_row_apr12\wiot95_row_apr12.xlsx',
'WIOT_1995', 'BCJ7:BKF1441'); if mod( size(F_1995,2), 5) == 0 f_1995 = sum(reshape(reshape(F_1995,size(F_1995,1)*5,[])',[size(F_1995,2)./5,size(F_1995,1) ,5] ),3)'; end x_1995(x_1995<1) = 1; A_1995 = Z_1995*inv(diag(x_1995')); L_1995 = inv(eye(1435)-A_1995);
em_1995 = [xlsread('X:\My Desktop\Thesis\Data\AIR_may12\AUS_AIR_May12.xls',
'1995', 'C2:C36')' ...];
d_1995 = em_1995./x_1995;
EM_1995 = diag(d_1995)*L_1995*f_1995; EM_SUM_1995 = ones(1,1435)*EM_1995;
em_hhir_1995 = [xlsread('X:\My Desktop\Thesis\Data\AIR_may12\AUS_AIR_May12.xls',
'1995', 'C40') ...];
carbon_footprint_1995 = EM_SUM_1995+em_hhir_1995;
%the same procedure is the followed for every other year in our dataset %Total Carbon Footprints
carbon_footprint = [carbon_footprint_1995; carbon_footprint_1996; carbon_footprint_1997; carbon_footprint_1998; carbon_footprint_1999; carbon_footprint_2000; carbon_footprint_2001; carbon_footprint_2002; carbon_footprint_2003; carbon_footprint_2004; carbon_footprint_2005; carbon_footprint_2006; carbon_footprint_2007; carbon_footprint_2008; carbon_footprint_2009;];
Stata code
use "/Users/matteo/RUG/Thesis/data stata v3.dta" xtset id t
xtreg footprint i.annex#c.t, fe estimates store consumption xtreg direct i.annex#c.t, fe estimates store production
Extended Tables
Table I: production of CO2 emissions (kilotons)
