Future climate change negotiations: Expanding the Triptych approach taking a demand perspective

(1)

Future climate change negotiations:

Expanding the Triptych approach

taking a demand perspective

In this paper we argue that due to globalization, climate change negotiations are in need of a new benchmark. We find that even from a consumer / demand perspective, the original Triptych variables are still relevant. Additionally, we explore two new variables that may be relevant for such negotiations, of which only the urbanization share was found relevant. Equipped with this information, we establish a benchmark and group 39 countries according to their performance, and find that in the aggregate, countries saw their performance deteriorate relative to the benchmark.

Jasper Starrenburg (s2299755) j.starrenburg@student.rug.nl University of Groningen

Faculty of Economics and Business MSc International Economics & Business Groningen, The Netherlands

Supervisor: Prof. Dr. Mr. C.J. Jepma

(2)

(3)

3

Michael Horn, former head of Volkswagen America This quote by the former boss of VW America, heard during the Volkswagen Passat revealing in Brooklyn, New York on September 21st 2015, came after the scandal. The scandal involved a piece of software that improved performance by reducing the amount of carbon dioxide emitted when the car was being tested for emissions. Volkswagen is the first in a series of revelations that revealed similar cheating practices at other automotive companies, such as Mitsubishi and Opel. It painfully brings to light that even after decades of convening on the matter of climate change, still not everybody agrees on its importance, or have other matters higher up the agenda.

This same quote by the former VW boss can perhaps be extended to the last two centuries of human history as well. Before the advent of what is currently is known as homo

sapiens, 200,000 years ago, the earth was able to sustain all forms of life. Their 'arrival' at first

did not seem to threaten earth's capabilities as well. However, in an ecosystem that consists of organisms producing, consuming, and decomposing, humans took on the part of the consumer too much as it became more intelligent, which earth could not compensate with its producing and decomposing roles. The Industrial Revolution two centuries ago, a negligible time span if we take into account that the earth is 4.54 billion years old, is seen at as the catalyst that accelerated the negative anthropogenic impact on earth.

(5)

5

Figure 1. GDP per capita in 1990 International Geary-Khamis dollars from 1810-2010.

the academic world, in which the validity of the paper is being doubted (Tol, 2014). Together with the observation that earth experiences large natural carbon dioxide fluctuations, these findings and observations led to uncertainties about the severity and validity of the human part in climate change in the past. Notwithstanding these skepticisms, the anthropogenic causes of the recent increase has been firmly established: In the fifth assessment report of the Intergovernmental Panel on Climate Change (IPCC), it is repeatedly claimed that it is

extremely likely that human influence has been the dominant cause of the observed warming

since the mid-20th century, with many climatologic phenomena attributed to anthropogenic causes (IPCC 2012, 2013).

Because these anthropogenic causes ultimately come from consumption and production, they need to be changed. In this respect, it is imperative that nations take a leading role in the fight against climate change since they decide upon the business climate with rules and taxation, and thus the 'rules of the game'. The latest Conference of Parties (COP) by the United Nations in Paris, the COP21, mobilized 195 nations promising to combat global warming by limiting global temperature from increasing more than 2C compared to pre-industrial levels (UNFCCC 2015a). In 2012, the end of the first commitment period of the Kyoto Protocol, it already sat at 0.85C, meaning that world emissions would have to fall by between 40-70% by 2050 from current levels, to almost zero emissions in 2100 if the world wants to achieve this goal (IPCC 2013, UNFCCC 2015b). Additionally, at COP21 the second commitment of the Kyoto protocol was outlined, meant for the period 2020-2025. It will go into force when 55 parties, responsible for 55% percent of global emissions, sign the Paris

0 5,000 10,000 15,000 20,000 25,000 30,000 35,000 1810 1835 1860 1885 1910 1935 1960 1985 2010

(6)

6

agreement, open for signature from April 21st, 2016 until April 22nd, 2017. This is certainly a step in the right direction, since tackling a global problem such as climate change naturally requires global mobilization and coordination.

However, the first commitment period of the Kyoto protocol that ran from 2008-2012 brought to light several problems associated with such global coordination. One of them concerns the debate whether we should focus on producers or consumers when attributing emissions to a country. Because of the limited availability on all intermediate trade flows, the first commitment period of the protocol focused on producer emissions that could be derived from national accounts. In a closed economy, national output is also consumed within its borders so that the distinction between producer and consumer emissions does not matter. However, given that trade is a defining feature of today's economic order, it may give a very biased picture. China, for example, is still considered the 'world's factory'. Determining its emissions based on the producer approach gives a radically different outcome compared to the consumer approach, since many of the manufacturing products are destined for foreign consumers. Next to data availability, another reason for the producer approach could be the presence of asymmetric information. We cannot expect that the average consumer is a homo

economicus, who is fully aware of the negative externalities that come along with the products

it demands, while producers naturally have more information on these externalities. For this reason it could be argued that it is more relevant to focus on the supply side of emissions instead of the demand side.

Notwithstanding, in this paper we will take a consumer approach. From a policy perspective, it is more effective to change consumer behavior than producer behavior. It is conceivable that opportunities will arise elsewhere when a polluting activity is disincentivized in a certain country. As long as there is demand for a polluting product, production will continue. By focusing on the demand side we are focusing on the core of the problem. Furthermore, taking a producer approach is obscuring the trade and service flows that very much defines today's globalized world. Taking a consumer perspective brings us back to the original goal of determining how much is emitted within a country's borders. In the past a supply perspective sufficed, but nowadays a demand perspective is warranted, since consumers import their products from all over the world.

(7)

7

(think of China and India) in the agreement. With the exclusion of these enormous emitters the effectiveness of the agreement is severely diminished, despite the fact that these countries still made pledges to reduce emissions. From figure 2 we see that possibly due to the aforementioned reasons, the world has not seen much progress up until 2009. Furthermore, for the countries that are willing to commit themselves, the negotiations often feel like a political tug of war. While the UNFCCC stated that differences in the Parties' circumstances should be considered when setting emission targets, no specifics were mentioned. As a result, negotiations were dominated by parties each bringing up their own mitigating circumstances that were specifically favorable for their country. Phylipsen et al. (1998) identified the common themes in these circumstances and brought them under in three categories: Differences in standard of living, fuel mix, economic structure, and the competitiveness of international-oriented industries. Then, the economy was subdivided in three groups (the energy sector, internationally-oriented industries, and the residual, domestic industry), for which each of the categories reasonable emission allowances were estimated. With the exclusion of the competitiveness category, this approach was coined the 'Triptych approach' and was adopted by the EU to determine the targets for each of its members. Taking a consumer perspective however, it is expected these structural variables have a smaller explanatory power due to country demand not being confined to country borders with supply and demand being decoupled from each other.

In this paper, we will put forward different variables that can explain emissions from a demand perspective. Furthermore, we will test whether the original structural Triptych variables are still relevant from a consumer perspective, given the fact that still a significant share of domestic production is consumed domestically for many countries. With these variables as determinants for the variations in GHG emissions, we will introduce a benchmark that facilitates negotiations, thus answering the following question:

(8)

8

Figure 2. Worldwide CO2 emissions in kilotonnes between 1995-2009

To answer this question, we will first use the World Input-Output Database (WIOD) to determine each country's emissions via the consumer approach, a publicly available database constructed out of cooperation between institutions from seven European countries funded by the European Commission. The database is a set of harmonized supply and use tables, alongside with data on international trade in goods and services, for 41 countries (including Rest of the World) and 35 sectors (Timmer et al, 2015). Complemented by environmental satellite accounts, it is possible to track and attribute the emissions embedded in the flow of goods to its final destination. We will use structural decomposition analysis (SDA) to decompose the contributions of different factors that collectively determine total emissions. This technique enables us to make use of detailed sectoral data which is not obscured by aggregate country-level data methods (e.g. Kaya, 1990). After we established each of the 40 countries' emissions, we will run a panel-data regression analysis to find the average impact of each of our variables, thereby creating our benchmark. Finally, we will plot these predicted emissions for each country and their actual emissions and group them according to their performance.

The remainder of this paper is structured as follows: In Section 2 we will give an overview of how climate change awareness came to be. This will prove insightful in understanding why even today not everybody is fully devoted in bringing back emissions. Section 3 explains our data collection. Section 4 is used to outline the methodology needed to tackle our research question. Then, section 5 discusses the results of the regression analysis and establishes the benchmark. Section 6 will be devoted to some robustness checks, and section 7 ends the paper with a conclusion and discussion.

(9)

9

2. Climate Change Awareness

The evolution of climate change awareness can broadly be split up in awareness within three interacting groups: Scientific awareness, political awareness, and consumer awareness. These last two should be seen as especially reinforcing each other, since the political sphere represents the people, and will only implement pro-environmental policies when there is a demand from the people for such actions. In turn, the government can steer consumer preferences by altering the 'rules of the game'.

2.1 Scientific Awareness

Scientists have for long been interested in climate change. The earliest contribution can be dated back to 1824 when Fourier (1824) noted that the earth should actually be colder given its size and distance from the sun when purely considering solar radiation as a heating source. As an explanation for the discrepancy between observed and predicted temperatures, he hypothesized that the earth's atmosphere retains some of the heat. Almost four decades later, Tyndall (1863) tested this assumption by measuring the absorption of infrared radiation by carbon dioxide and concluded that it indeed raises the earth's temperature by the retention of radiation. The link with human activity remained absent however. It took until 1950, close to a century later, that people started to think about the possibility that human activities might have an effect on the climate by regulating the amount of carbon dioxide in the air. As Kellogg (1987, 113) rightfully says about the line of reasoning before that time:

"[...] One could easily show that the total power output of all human activities (about 8 x 1012W) was utterly insignificant (about 1/10,000) compared to the radiant heat absorbed from the sun. "

(10)

10

possibility to incorporate this data in complex climate models made it possible to model these factors and their interactions with other climatological variables and tentatively predict the impact on the future despite incomplete modeling of some factors, such as the parameterization of clouds (Kellogg, 1979) and the complex interactions of aerosols with

other variables (Brosset 1976, Viskanta et al., 1977). The link between climate change and the

anthropogenic effects became increasingly well understood, with the state of knowledge conveniently compounded by working groups in the IPCC assessment reports.

2.2 Political Awareness

However, while the topic of climate change was firmly present on the scientific agenda and knowledge kept expanding, for a long time policymakers did not pick up on the issue until the sixties and seventies. With computers becoming more powerful and able to run complex climate models that lead to a better understanding about the interactions between the biosphere, atmosphere, ocean, and the anthropogenic effects on each of those, the issue became increasingly present on the political agenda, despite the fact that the scientific community did not always unanimously agree on for example the incidence of climate extremes (Karl and Easterling, 1999).

(11)

11

targets, which led to the famous Kyoto-protocol drafted at the COP3 in 1997 that finally got ratified in 2005 after 55 countries responsible for 55% of greenhouse gasses committed to the protocol. The goal was for these industrialized countries (so-called Annex I countries) to bring back total emissions by 5.2% in the period 2008-2012 with varying targets per country (e.g. -21% for Germany, but no more than an increase of 27% for Portugal (UNFCCC 2011)). While this was a breakthrough insofar it was the first legally binding provision for countries to reduce GHG emissions, and on such a global scale, it was far from ideal. Some industrialized countries voiced their concerns about the exclusion of developing countries who are responsible for 29% of global CO2 (Morisette and Plantinga 1991), notably big emitters like India and China, and the consequent economic damage they would face as a result of unfair competition. For this reason, the US Congress did not ratify the protocol, weakening its effect since it was responsible for 21.6% of total carbon dioxide emissions in 2003 (Marland et al., 2003). Another developed country, Australia, did ratify the agreement but made use of windfall provisions, in which Australia benefitted from a clause that enabled it to gain emission credits for decreasing land clearances, with 1990 as the baseline, a record year in land clearing (Crowley, 2007). This is a dent in the effectiveness of the protocol, since Australia has the highest net greenhouse gas emissions per capita in the world. The author found it stood at 32% higher than the US calculated via the producer approach, and our calculations show it is 12.8% higher via the consumer approach. While Australia´s approach still brought along a drop in CO2 emissions, ideally we would like to see more structural efforts into reducing GHG emissions, not emission reductions by making use of these one-time provisions. Furthermore, the protocol was based on territorial emissions after which countries shifted their polluting industries to other countries in order to comply with the protocol (Serrano and Dietzenbacher, 2010). If these products are destined for final demand in the 'sending' country, in reality the emissions attributable to consumption in that country does not change by moving polluting activities elsewhere. Globally, assuming identical technologies, it is a zero-sum game. Understandably, the consumer generated emissions and producer-generated emission give a different picture, and in order to avoid the territorial emission shifting, consumer-generated emissions is preferred as it depicts a more accurate picture of the direct and indirect emissions needed to satisfy demand within a country's borders.

(12)

12

mostly tried to spell out the details of the protocol, but difficult decisions and points of dispute were continuously pushed forward to the next meeting. Inaction proves problematic with climate change, as it increases the risk of reaching a certain threshold that may trigger climate change ´surprises´ (Michaelowa and Rolfe, 2001). At COP18, preparations were made for the second legally binding term, to be adopted in 2015 for the post-2020 period (UNFCCC 2012, 2013). It faced the same problems as the Kyoto protocol did, with some large industrialized countries not participating, and with no legally binding targets for developing

COP21 last year in 2015 looked promising since the second commitment period of the Kyoto Protocol was finally drafted there. This second legally binding target, open for signature from April 22th 2016 until April 21st 2017, will enter into force when at least 55 parties accounting for at least 55% of greenhouse gas emissions signed up (UNFCCC, 2015).

After numerous conventions, one would believe that at least some progress has been made, either in stabilizing (the goal of earlier conventions) or bringing back the emissions of greenhouse gasses (the current goal). However, the worldwide emissions is steadily increasing (see figure 2) This might be due to the fact that the biggest emitters have only made pledges and not commitments. In order to understand the responses of governments, it is worth taking a look at how consumers perceive climate change, as governments ultimately represent their citizens.

2.3 Consumer awareness

(13)

13

with a general impression that the German economy came comparatively undisturbed through the world financial and economic crises, and that the labor market was relatively healthy

(Engels et al. 2013, p. 1025). However, Diekmann and Preisendorfer (1998) find that the link

between environmental awareness and pro-environmental behavior is rather weak, so that knowledge does not translate into action. This is why in today's climate debate, while the consumer ultimately is the end user, there is a crucial role for the government to steer consumption and production patterns.

Summarizing, it is clear that climate awareness has come a long way. The most important contribution to this is the scientific knowledge built up over the course of decades, aided by increased computing power. However, some difficulties persist, with uncertain consequences of the effect on a regional level. Politicians have picked up on the matter, but it is often of secondary importance, especially since not every country is participating in the legally-binding agreements, leading to unfair competition. This is the main reason why the US has not ratified the Kyoto Protocols as of yet. Furthermore, when parties are finally willing to negotiate, progress is slow due to numerous mitigating circumstances that each party is trying to put forward. Additionally, there must be demand from the public for bringing back GHGs. We have seen that there is widespread awareness from the consumer, but that this does not automatically translate into proactive behavior. When captured by economic downturn, consumers tend to think more about pecuniary matters and less about the environment. Now we will continue with our objective to establish a benchmark that can facilitate climate negotiations, for which we will first describe our data and methodology respectively in the next sections.

3. Data

3.1 Data for IO-analysis

3.1.1 WIOT

(14)

14

called 'Rest of the World' so that all national accounts are balanced, adding up to a total of 41 countries and 35 sectors. While the database has information on individual national IO tables, in this paper we use the world IO table which conveniently reports all domestic and international trade and service flows per sector. This data is presented in a 1435x1435 table, accompanied by end use split up in five final demand categories for each country. The WIOD tracks these flows from the year 1995 until 2011, but our samples include the tables until 2009, because the environmental accounts track emissions until the same year.

3.1.2 Environmental accounts

Data on environmental accounts are also taken from the WIOD. It provides a wide range of environmental satellite accounts, but we are particularly interested in two accounts: The first one is the 'Energy use, Emission Relevant' account that contains information on the energy used in terajoule per sector per million USD, relevant for emissions. This thus excludes the non-energy use of energy commodities (e.g. asphalt for road building) and the input of energy commodities for transformation (e.g. coal being transformed into coke) (Genty, 2012).

The second account is the 'Emissions to air' account that holds information on emissions per sector and pollutant. All pollutants are denoted in metric tons, with the exception of CO2 which is in kilotonnes. A huge advantage is that these satellite accounts correspond to the WIOT in terms of industry classification, so that the data do not need to be synchronized. These accounts extend from the period 1995 until 2009 for all 41 countries.

3.2 Data for regression analysis

(15)

15

CO2 emissions per capita: This variable acts as our dependent variable and is

calculated via IO-analysis using the WIOT. In the next section we will explain how we ascribed emissions to each country.

GDP: We hypothesize that GDP per capita will have a positive effect on CO2

emissions since a higher affluence in a country increases its consumption needs, leading to higher emissions. For consistency we calculate GDP from the WIOT by subtracting total intermediate consumption from output at basic prices for each sector-country-year pair. We then add up the sector totals to arrive at total GDP in millions of USD, current prices.

Population: To arrive at per capita figures for GDP, we take information on

population from the World Bank and divide the GDP numbers for each country and year accordingly. Because the World Bank does not report statistics of Taiwan, we extracted data for this country from the Worldometer to complement the dataset.

Urban population share: With regard to energy consumption, emissions increase

(16)

16

Educational attainment: The rationale for including this variable is found in the paper

by Lee et al. (2015) who find that educational attainment is the single best predictor of climate change awareness. Moreover, Semenza et al (2008) find that in their analysis for Houston and Portland, climate change awareness is widespread but that educational attainment was not only a strong predictor for awareness but also for changing behavior as a response to climate change. As noted in section 2.3, some papers find a weak link from awareness to action, so that the relationship between this variable and emissions may either be negative or insignificant. To measure educational attainment, we use the percentage of the labor force that has enjoyed tertiary education, which is taken from the World Bank. A downside of this measure is that many countries did not consistently report this statistic over the years, so that there are several gaps in the dataset for this variable, or with some countries not reporting at all (e.g. Taiwan). As a check, we included another variable from the World Bank that is a proxy for educational attainment, namely the share of GDP of government expenditures on education. It ignores effectiveness of the spending, and unfortunately also suffers from data gaps. On the other hand, it might capture a broader aspect of education, instead of the narrowly defined tertiary education share of the labor force. Accepting the fact that this weakens the strength of our analysis, we will include it due to superiority over other measures like enrollment rates.

Green energy share: This variable is computed from the Energy use, Emission

Relevant' environmental account, provided by the WIOD. This structural variable is adopted from the original Triptych approach that captures the structure of the energy sector, and tells us how much of the total energy output is being generated by green energies. These green energies do not emit any carbon dioxide, and thus it is expected that a higher share of green energies has a negative impact of per capita emissions within a country. Note that the definition of green energies in this case is not the same as renewable energies. For example, we included nuclear energy as a green energy source, but whether it is a renewable energy source is open for discussion. For a list of the energies included in the variable, see the notes under table 1.

Industry share: This is another structural variable from the original approach and

(17)

17

though this share might be the same in two countries, their share of agriculture or services may vary widely, which has implications for per capita emissions. Even within the industry sector, countries may vary.

4. Methodology 4.1 IO-analysis

As noted above, in order to map emissions, we employ a multiregional IO analysis. It is an excellent quantitative economic technique dating back to Leontief (1953) who first created and made use of such IO tables. These tables provide insights into the interdependencies between sectors within an economy and contain detailed trade information regarding the in- and outflows of products and their final use category. In equational terms, it looks as follows:

𝑥𝑥𝑖𝑖 = 𝑧𝑧𝑖𝑖,1+⋯ + 𝑧𝑧𝑖𝑖,𝑗𝑗+ ⋯ + 𝑧𝑧𝑖𝑖,𝑛𝑛+ 𝑓𝑓𝑖𝑖 = � 𝑧𝑧𝑖𝑖,𝑗𝑗 𝑛𝑛

𝑗𝑗=1 + 𝑓𝑓𝑖𝑖

Where xi denotes output for sector i. On the right-hand side, the z coefficients denote the

output from sector i used in the production process of sector j, and thus represents the intermediate deliveries. Note that all numbers are in monetary units. The full set of equations looks like this:

𝑥𝑥1 = 𝑧𝑧1,1+ ⋯ + 𝑧𝑧1,𝑗𝑗 + ⋯ + 𝑧𝑧1,𝑛𝑛+ 𝑓𝑓1 𝑥𝑥𝑖𝑖 = 𝑧𝑧𝑖𝑖,1+ ⋯ + 𝑧𝑧𝑖𝑖,𝑗𝑗+ ⋯ + 𝑧𝑧𝑖𝑖,𝑛𝑛+ 𝑓𝑓𝑖𝑖 ⋮

𝑥𝑥𝑛𝑛 = 𝑧𝑧𝑛𝑛,1+ ⋯ + 𝑧𝑧𝑛𝑛,𝑗𝑗+ ⋯ + 𝑧𝑧𝑛𝑛,𝑛𝑛+ 𝑓𝑓𝑛𝑛

This set tells us all intermediate deliveries between each and every of the 1435 sectors (35 sectors for 41 countries), as well as the final demand for each sector. Let

x = � 𝑥𝑥1 ⋮ 𝑥𝑥𝑛𝑛 � , Z= � 𝑧𝑧1,1 ⋯ 𝑧𝑧1,𝑛𝑛 ⋮ ⋱ ⋮ 𝑧𝑧𝑛𝑛,1 ⋯ 𝑧𝑧𝑛𝑛,𝑛𝑛 �, and f = �𝑓𝑓⋮1 𝑓𝑓𝑛𝑛 �.

Then we can represent this set of equations in compact matrix form:

(1)

(2)

(18)

18

x = Zi + f

Thus, x is a column vector representing the total output per sector. Z is, as noted above, a matrix containing all intermediate deliveries between sectors (and countries). Post-multiplying this with i, a summation vector yields a column vector with the row sums added up. Finally, f is a column vector containing total final demand for each sector in each country. To facilitate computations and interpretations later on, we would like to have share coefficients instead of total value coefficients. Let us assume for example that industry 1 is

Agriculture, Hunting, Forestry and Fishing and industry 3 Food, Beverages and Tobacco.

Then intermediate deliveries from sector 1 to sector 3 is denoted as:

𝑎𝑎1,3 =𝑧𝑧_𝑥𝑥1,3 3 =

𝑉𝑉𝑎𝑎𝑉𝑉𝑉𝑉𝑉𝑉 𝑜𝑜𝑓𝑓 𝑝𝑝𝑝𝑝𝑜𝑜𝑝𝑝𝑉𝑉𝑝𝑝𝑝𝑝𝑝𝑝 𝑓𝑓𝑝𝑝𝑜𝑜𝑓𝑓 𝑝𝑝𝑉𝑉𝑝𝑝𝑝𝑝𝑜𝑜𝑝𝑝 1 𝑉𝑉𝑝𝑝𝑉𝑉𝑝𝑝 𝑖𝑖𝑖𝑖 𝑝𝑝ℎ𝑉𝑉 𝑝𝑝𝑝𝑝𝑜𝑜𝑝𝑝𝑉𝑉𝑝𝑝𝑝𝑝𝑖𝑖𝑜𝑜𝑖𝑖 𝑜𝑜𝑓𝑓 𝑝𝑝𝑉𝑉𝑝𝑝𝑝𝑝𝑜𝑜𝑝𝑝 3 𝑉𝑉𝑎𝑎𝑉𝑉𝑉𝑉𝑉𝑉 𝑜𝑜𝑓𝑓 𝑝𝑝𝑜𝑜𝑝𝑝𝑎𝑎𝑉𝑉 𝑜𝑜𝑉𝑉𝑝𝑝𝑝𝑝𝑉𝑉𝑝𝑝𝑜𝑜𝑓𝑓 𝑝𝑝𝑉𝑉𝑝𝑝𝑝𝑝𝑜𝑜𝑝𝑝 3

(19)

19

Where x-1 represents the diagonal n × n matrix with the x column vector elements on the

diagonal. This ensures every element of the intermediate deliveries Z matrix is divided by the total output element of the sector that receives the input.

This set of equations depict the interdependencies of each sector on every other sector originating from each of the 41 countries. However, these equations do not show the indirect intermediate deliveries yet. For example, how much does the output increase of the agricultural sector increase when final demand in Rubber and Plastics increase? This sector uses direct inputs from the remaining 34 sectors, which in turn might use inputs from the agricultural sector. The agricultural sector is thus directly and indirectly involved in the production in the Rubber and Plastics sector. In order to answer such questions, we need to perform an additional step. In equation (5), bring all input coefficients to the left-hand sight and group together all total output coefficients x1,xi,...xn, to arrive at:

�1 − 𝑎𝑎1,1�𝑥𝑥1− ⋯ − 𝑎𝑎1,𝑗𝑗𝑥𝑥𝑗𝑗− ⋯ − 𝑎𝑎1,𝑛𝑛𝑥𝑥𝑛𝑛 = 𝑓𝑓1 −𝑎𝑎𝑖𝑖,1𝑥𝑥1− ⋯ − (1 − 𝑎𝑎1,𝑗𝑗)𝑥𝑥𝑗𝑗− ⋯ − 𝑎𝑎1,𝑛𝑛𝑥𝑥𝑛𝑛 = 𝑓𝑓𝑖𝑖 ⋮

−𝑎𝑎𝑛𝑛,1𝑥𝑥1− ⋯ − 𝑎𝑎𝑛𝑛,𝑗𝑗𝑥𝑥𝑗𝑗− ⋯ (1 − 𝑎𝑎𝑛𝑛,𝑛𝑛)𝑥𝑥𝑛𝑛 = 𝑓𝑓𝑛𝑛 This can be written as

(I-A)x = f

In which I is a identity matrix consisting of only ones on the diagonal. To solve for x, we divide by (I-A) to arrive at our final equation:

x = (I - A)-1f = Mf, with M = (I-A)-1

M is known as the Leontief inverse. Equation (9) tells us how much output of a certain sector x is needed for a certain final demand vector f, both directly and indirectly. For example, let

us assume that element m1,3 = 0.06. This tells us that USD 60,000 is needed from sector 1,

Agriculture, Hunting, Forestry and Fishing, to satisfy USD 1 mln of final demand for sector 3, Food, Beverages and Tobacco.

(7)

(8)

(20)

20

Now it is easy to allow for the inclusion of environmental accounts. If we have information on the total amount of emissions E and divide it by total output for each sector, we have the emission coefficients matrix e. We can pre-multiply the Leontief inverse M with matrix e to arrive at our implied emissions per unit of final demand:

CO2 = eMf, with e = Ex-1

which tells us the amount of CO2 in Kilotonnes (Gg) embedded in the direct and indirect inputs needed to satisfy USD 1 mln of final demand for a product. Furthermore, it becomes possible to decompose emission vector e into vector k (corresponding to the account 'Energy use, Emission Relevant') that explains the composition of the fuel mix used in the production process and into vector g (corresponding to the account 'Emissions to air' ) that explains the energy intensity per unit of output:

e=kg', with k=eF-1 and g= Fx-1

This gives us all the information regarding the factors determining CO2 emissions, resulting in the following aggregate equation:

CO2=kg'Mf + HH

where HH stands for household emissions. This equation will be used for our Structural Decomposition Analysis.

4.2 Structural Decomposition Analysis (SDA)

Equipped with equation (12), it is possible to decompose total CO2 emissions into four factors outlined above: the fuel mix k; the energy intensity g; the technology input matrix M; and final demand f. A complication is that there is no unique method of structurally decomposing the terms (see Dietzenbacher and Los (1998) for a more technical discussion). In our case, 4 terms lead to 4!=24 unique decomposition methods. In the same paper, the authors show that there is considerable variation in the contributions when calculating each of the 24 terms, so that the outcome crucially depends on the method used. In this paper, we

(10)

(11)

(21)

21

perform SDA based on the average of two 'polar equations', a method frequently used in the academic literature. To decompose the change in emissions between 1995 and 2009 one of such polar equation looks like:

∆CO2= CO2(09)-CO2(95)

= k(09)g'(09)M(09)f(09) - k(95)g'(95)M(95)f(95) = (∆k)g'(09)M(09)f(09) + k(95)( ∆g')M(09)f(09) + k(95)g'(95) (∆M)f(09 ) + k(95)g'(95)(M(95) (∆f)

which is the first polar equation. The second polar equation looks the same, but then with the years switched around in each of the four terms. We will use the average of these two polar equations for determining each factor's contribution towards the percentage change between 1995-2009, with 1995 as the base year.

4.3 Regression analysis

For our regression analysis we use a simple panel log-linear OLS model in per capita terms with both country and year fixed effects to account for heterogeneity within countries and years. The Hausman test shows that fixed effects are more appropriate than random effects for our dataset. Furthermore, the Levin-Lin-Chu unit root test shows that all variables are stationary so that we can continue with our analysis. In these tests, we demean to mitigate the impact of cross-sectional dependencies as suggested in Levin et al. (2002). We work with 39 countries by removing Taiwan because of limited data availability for this country of our relevant variables, with a timespan of 15 years from 1995 until 2009. We scale per capita GDP so that it is interpreted in 100's. Furthermore, we include a quadratic term for per capita GDP to capture the idea of the environmental Kuznet's curve, which states that per capita emissions first increase as a country grows richer but after certain point decreases again. Our main specification thus is:

ln(CO2 per capita) = αit+ β1it GDP per capita + β2it GDP per capita2

+ β3iturbanization + β4it education + β5it green energy share

+ β6it economic structure + εit

(22)

22

5. Results 5.1 IO analysis

Figure 3 summarizes our main findings of our IO-analysis, for which we used the SDA outlined above. We see that CO2 emissions (excluding households) saw an increase of 33.33% from 1995 to 2009, which drops to 11.69% when taking into account population growth (not depicted in the graph). SDA makes it possible to decompose the CO2 emission increase into its four concomitant components. The table is interpreted such that the component percentages add up to the CO2 percentage increase. The fuel mix used into the production process saw a very slight decrease over the 15-year period by -3.97%. This means that ceteris

paribus, in the aggregate we have not seen a substitution away from polluting fossil fuels

towards greener alternatives. The energy intensity however saw a significant decline over the period by -140.03%. This indicates that even though the world did not shift away from polluting energy sources, it shifted away from polluting activities. All things equal, there is a lot less energy used in generating the same amount of GDP. These two factors provide a positive outlook for the future. However, the two remaining factors more than offset these positive developments. Final demand saw an increase of 132.57%, Almost entirely voiding the progress in reducing energy intensity. Furthermore, the production structure of the world also contributed to the increased CO2 emissions by 45.31%, reflecting the fact that production processes has become less efficient. This might be a consequence of adopting a producer-approach in the past for the abatement of CO2 emissions, in which countries shift polluting activities to other locations with less stringent emission legislation, resulting in even more pollution.

From the period up till 1999 we see modest changes in each factor, with an annual increase in CO2 by around one percent. After that they greatly accelerated with an annual increase of 2.35% between 1999 and 2009. An explanation for this acceleration after 1999 could be China's integration in the world market by its admission to the WTO in 2001 and the anticipation of this event. To substantiate this claim, we will turn to figure 4 which shows the CO2 emissions of some large emitters. The developed world, roughly represented by the US and the EU27, did not see any significant increases in CO2 over the period. China on the other hand saw more than a doubling of its CO2 emissions from 2001 until 2009. Additionally, unlike the developed world, it did not experience a fall during the financial crisis of 2008.

(23)

23

Figure 3. Percentage change in factors compared to 1995. Percentages add up to the change in CO2 emissions.

Due to unavailability of data of the environmental accounts after 2009 and due to the incomplete recovery of economies from the crisis as of yet, it is hard to tell whether this fall is sustained in the long-run or whether it will return to its previous trajectory when the economies pick up again. It is important to note that this result is established via the consumer-approach in attributing CO2 emissions to countries, so that the argument of China as the world's factory is not valid here since emissions are being attributed to the location of the product's final destination.

Still, from figure 2 and 4 it is striking how much worldwide CO2 emissions moves in tandem with the emissions in China. In fact, The increase in China makes up 58.9% of the increase in worldwide emissions. Of course, this is percentage obscuring the finer details of the picture since there are many countries who experienced a decrease in their emissions, but the conclusion remains the same. However, the question whether China is emitting too much is an ethical one. It appears alarming in a graph, but China is the largest country in the world. It is a remarkable feat to increase welfare (and emissions) for so many people, but emissions per capita stands at just one-third of that of the US. Should China be allowed to have the same per capita emissions? Or should the US perhaps lower its emissions to China's level? The sheer subjectivity surrounding this question has produced stasis during climate change debates since there is no reference point or an objective view of fairness. In this paper we therefore stress the need for a new benchmark taking a consumer perspective, to which we will turn next. -175.00% -125.00% -75.00% -25.00% 25.00% 75.00% 125.00%

(24)

24 Figure 4. Emissions of CO2 in kilotonnes for big emitters.

5.2 Regression Analysis

Before we start benchmarking the emissions, we first present relevant factors that can explain the consumer-generated emissions for each country. Using the coefficients of each factor we can create a benchmark and compare each country to this benchmark. Because our focus is on countries, we excluded RoW in our regression analysis. Table 1 shows the results for our regression analysis. Column (1) lists the results for our specification including year or country effects. All variables are insignificant, except for our measure of education, which is significant at the 10% level. A reason for this insignificance could be either a limited amount of observations since there are quite some data gaps for the measurement of education, or due to a high correlation between this measure and the other variables (see table A.2 in the appendix). The variable has a negative sign, meaning that a 1 unit increase (in this case percentage point) of government expenditure on education leads to an almost 0.02% decrease in CO2 emissions per capita. We will exclude the variable in the next specifications since it allows us to work with more observations. Moving to column (2) we see that excluding the previous measure renders the remaining variables significant again, all at the 5 percent level and the quadratic term at the 1 percent level. Furthermore, the urbanization measure is quantitatively important since a 1-unit increase (for urbanization, one unit is one percentage point) leads to a 2.39% increase in CO2 emissions per capita. This is an important finding since it indicates a more than one-for-one relationship between urbanization and CO2 emissions (and thus indirectly consumption), pointing towards important findings of

(25)

25

agglomeration as mentioned in section 3.2. Furthermore, the fact that both GDP per capita and urbanization are significant indicating that urbanization still plays a significant role in explaining emissions despite the fact that incomes are generally higher than rural areas.

Up till this point we have not included the structural variables used in the Triptych method. To test their significance in a consumer-perspective, we first include the share of green energies in column (3). It is highly significant at the 1% level, with a negative sign that tells us that all else equal, a 1-unit increase (in this case percentage) in the share of green energies brings about a 0.15% decrease in the per capita emissions. The significance of urbanization decreases somewhat but is still significant at the 10% level. We also note that the goodness of fit and the adjusted R-squared further increases, indicating a better predictability of our model. In column (4) we add the other structural variable to our specification that tells us something about the economic structure of the country. Admittedly, it is hard to capture this structure in just one variable. Countries with the same percentage of value added being earned by the industry sector can still differ widely in their agricultural or service sector and even within the industry sector since it covers 36 industries according to the ISIC classification. Nonetheless, column (4) tells us that this variable is also significant at the 1% level. Moreover, it does not negatively affect the significance of the remaining variables.

From the table it becomes clear that even with a consumer perspective, structural variables used in the original Triptych method are still relevant. With such a perspective, production and consumption are split up but since still a significant part of production is consumed domestically even in today's globalized world, the structure of a country remains important in explaining domestic emissions. Even more, while we found that our education measure was not significant in our specification since it was correlated too much with the other variables, our measure of urbanization does add more explanatory power to the model. Removing the measure decreases our adjusted R-squared by 16.4% to 0.316 (not included in the table).

5.3 Benchmarking the emissions

(26)

26

degree line so it is easy to see which countries are below and which countries are above the benchmark. Furthermore, we plotted the values of 1995 (in blue) and 2009 (in red) so we can

Table 1. Robust standard errors are in parentheses: *** p<0.01, ** p<0.05, * p<0.1. GDP per capita is denominated in 100's. The green energies included in the variable greenshare include biogasoline, biodiesel biogas, heat, nuclear, hydro energy, geothermal, solar and wind energy, next to 'other renewables' and 'other sources', according to the WIOD definition. The industries included in the

indper variable are ISIC 10-45.

track developments over time. Although China (CHN) emitted below the expected amount in 1995 given our specification at a low level, in 2009 it is almost crossing the benchmark line meaning that even taking into account our benchmark variables, it emitted more than predicted. Also, Mexico (MEX), below the benchmark, did not see a large increase in its per capita emissions (the vertical distance between its red and blue dot), while the underlying variables predicted that it would emit more than its actual emissions so that it positioned itself further below the benchmark. The same has happened in the Netherlands (NLD), which is above the benchmark in 1995. Because of this horizontal shift, it is close to crossing the line, meaning that it would actually be below the benchmark, a welcome development.

To structure our results, we choose to list and group the countries according to their performance in table A.3. We see that relatively few countries were able to pass through the benchmark. For the countries that were above the benchmark in 1995, just 10% of our sample moved below it, namely France and Luxembourg.

(1) (2) (3) (4)

ln (CO2 per capita) - - -

GDP per capita 0.000455 0.000485** 0.000733*** 0.000932***

(0.000349) (0.000191) (0.000172) (0.000173)

GDP per capita2 -6.83e-07 -9.00e-07*** -9.60e-07*** -1.09e-06***

(6.98e-07) (1.20e-07) (1.03e-07) (1.10e-07)

Urbanization share 0.0128 0.0239** 0.0150* 0.0176**

(0.00776) (0.00938) (0.00748) (0.00671)

Education expenditure -0.000196*

(0.000110)

Green energy share -0.00149*** -0.00133***

(0.000349) (0.000339)

Industry share value added 0.000664***

(0.000114)

Constant -6.001*** -6.821*** -5.791*** -6.272***

(0.538) (0.649) (0.540) (0.496)

Observations 445 600 600 571

# of countries 39 40 40 39

Year Effects YES YES YES YES

Country effects YES YES YES YES

F-statistic 2.130 34.20 51.45 47.41

R-Squared 0.316 0.358 0.427 0.458

(27)

27

Figure 5. Emission benchmark. The 45o degree line divides the countries into a green zone and a red zone. Countries in the green zone are doing better than expected, given the values of our benchmark variables, while countries in the red zone are doing worse than expected. The blue dots depict emissions in 1995, while the red depict emissions in 2009.

This observation may obscure other details in the graph however. It is therefore important to read figure 5 and table A.3 together since the table may give the illusion of low dynamics in emissions. Looking at the table, we see that many countries moved either up, down, left, or right. In the aggregate the countries in the category 'consistent high emitters' moved 0.00068Kt emissions per capita away from the benchmark in 2009 if we compare to 1995. Even taking into account rising emissions as a country grows more developed over time, the countries in this category managed to emit more than expected given the values of our benchmark variables. In the same way, the countries in the category 'consistent low emitters' moved closer toward the benchmark by 0.0041Kt. While this is expected for developing countries, it is worrisome that the countries with the highest emissions did not see a absolute decrease, let alone a relative decrease in aggregate emissions.

6. Robustness

To check whether our results are not driven by choice of variables, we perform several robustness checks.

(28)

28

only have one value, we convert CH4 and N2O according to their global warming potential, as outlined in Myhre et al. (2013). The results can be found in table A.3, column (1). While all variables remain significant with their expected signs, urbanization turns insignificant with our alternative dependent variable. An explanation for this might be the fact that the GHGs included in our dependent variable are driven by different developments. While CO2 is generally associated with manufacturing, CH4 is associated with agriculture. When a country develops, usually it attracts labor from the agricultural sector to be put to use in the manufacturing sector, so that people move from the countryside towards urban areas. Therefore, the two gasses might balance out, explaining the insignificance of our urbanization measure.

Also, we chose different variables to depict the economic structure of the country. Instead of the share of the industry sector in value added we respectively chose the agricultural, manufacturing, and services share in columns (2), (3), and (4). The first two are insignificant, albeit with a correct sign, while the third one is significant. Out of a total of four measures used, jut half of it is significant, pointing towards the difficulty of capturing the structure of the economy in just one variable.

In the last column, we excluded the urbanization share to see if it really is a good addition to our model. If it were not, we might as well use the original Triptych method with only structural variables for future emission abatement negotiations. When comparing the adjusted R-squared of column (5) of table A.3 with that of column (4) of table 1, we see that the variable improves the model by looking at the adjusted R-squared, raising it from 0.316 to 0.378.

(29)

29

7. Conclusion and discussion

This paper first started with the observation that despite many conventions, not everybody is willing to bring back emissions. In the business world, automobile manufacturers got caught recently trying to cheat emissions during performance tests. Also, in section 2 we read about consumers who tend to forget the severity of problem when in economic downturn. Scientists are still debating about different aspects of climate change and their complex interactions, despite the wealth of knowledge available. Maybe this wealth is exactly the problem of a lack of consensus, and one of UNFCCC's objectives was to bring together this knowledge in their assessment reports. Already in 1997 it was recognized that negotiations during these binding agreements on conventions was in need of a structural approach that can guide the process. From this the Triptych approach was born. While initially taking a producer approach in the negotiations, the consumer-approach has been getting more attention recently. For this reason, this paper was set to establish a new benchmark with new variables. It was found that the original structural Triptych variables were still relevant taking such a perspective. It also found that a new variable, the degree of urbanization, had significant explanatory power in explaining per capita emissions. Our hypothesized variable of education was found significant as well with the correct sign, however, caused considerable noise in our regressions, which we removed as a result. With the information on the impacts of our variables we established a benchmark and regressed the fitted values on the actual values. We found that in the time period 1995-2009, not many countries crossed the benchmark, so that only two countries decreased their emissions sufficiently, positioning themselves below the benchmark while initially above it. Likewise, two countries positioned themselves above the benchmark as well, while initially below it. It can be argued that therefore the world as not seen any progress. When focusing on changes in emissions instead of cross-overs, we see that in the aggregate the countries above the benchmark emitted 0.00068Kt emissions per capita more relative to the benchmark. This thus means a higher overall increase since the benchmark is rising, for the reason that as countries develop, they will have higher values for our relevant variables. Therefore this benchmark is not an indication of how well a country is doing, absolutely speaking. It gives a description of how well countries did over the time period relative to the benchmark.

(30)

30

how the next round of emission reductions can be distributed over countries. It can also be a source for further research on which variables are relevant for following negotiations. As the world is becoming more globalized and integrated, consumption and production will be more decoupled from each other, and it can be argued that therefore the structural variables will have less and less explanatory power. By that time, negotiations will have to be based on variables explaining demand within a country, such as the urbanization variable found in this paper.

This paper grouped all years together and found the average impact of the benchmark variables. However, it is reasonable to expect that impacts of certain variables will change over the years, with some becoming more relevant while others become less relevant. This paper does not differentiate in this and may be a topic for further research. In the same way, the variables may have different impacts depending on the country, which we did not take into account as well. Then again, while scientific knowledge is ever-expanding, it is up to political sphere to do something about it. And as of yet, most are doing less than what is needed for a sustainable environment.

(31)

31

8. References

Ackerman, T., Baker, M. B. 1 977. Shortwave radiative effects of unactivated aerosol particles in clouds. Journal of Applied Meteorology. 16 : 63-69

Badoe, D.A., Miller, E.J. 2000. Transportation - land-use interaction: Empirical findings in North America, and their implications for modeling. Transportation Research Part D, 5: 235-263.

Brosset, C. 1976. Airborne particles : Black and white episodes. Ambio. 5 : 157-163

Cook, J., Nuccitelli, D.,Green, S.A., , Richardson, M.W, B., Painting, R., Way, R., Jacobs, P., Skuce, A. 2013. Quantifying the consensus on anthropogenic global warming in the scientific literature. Environmental Research Letters, 8(2).

Ciccone, A., Hall, R. 1996. Productivity and the density of economic activity. American

Economic Review, 86: 54-70.

Crowley, K. 2007. Is Australia Faking It? The Kyoto Protocol and the Greenhouse Policy Challenge. Global Environmental Politics, 7(4): 118-139

Diekmann, A., Preisendorfer, P. 1998. Environmental behavior: discrepancies between aspirations and reality. Rationality and Society, 10: 79–102.

Dietzenbacher, E., Los, B. 1998. Structural Decomposition Techniques: Sense and Sensitivity.

Economic Systems Research, 10(4): 307-324.

Fourier, J. B. J. 1824. Remarques Générales Sur Les Températures Du Globe Terrestre Et Des Espaces Planétaires. Annales de Chimie et de Physique, 27: 136-167.

Friedl, B., Getzner, M. 2003. Determinants of CO2 emissions in a small open economy.

(32)

32

Genty, A., (ed) 2012. Final Database of Environmental Satallite Accounts: Technical Report on their Compilation, WIOD Deliverable 4.6, Documentation, available at:

http://www.wiod.org/publications/source_docs/Environmental_Sources.pdf

Gibbard, S., Caldeira, K., Bala, G., Phillips, T.J., Wickett, M. 2005. Climate effects of global land cover change. Geophysical research letters, 32(23): 1-4.

Heinonen, J., Jalas, M., Juntunen, J.K., Ala-Mantila, S., Junnila, S. 2013. Situated lifestyles: I. How lifestyles change along with the level of urbanization and what the greenhouse gas implications are - a study of Finland. Environmental Research Letters, 8(2).

Idso, S.B. 1982. Carbon dioxide - friend or foe? An inquiry into the climatic and

agricultural consequences of the rapidly rising CO2 content of Earth's atmosphere.

Tempe (AZ): 1BR Press.

IPCC, 2007. Summary for Policymakers. In: Climate Change 2007: The Physical Science

Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press,

Cambridge, United Kingdom and New York, USA.

IPCC, 2012. Summary for Policymakers. In: Managing the Risks of Extreme Events and

Disasters to Advance Climate Change Adaptation. Cambridge University Press,

Cambridge, UK, and New York, NY, USA.

IPCC, 2013. Summary for Policymakers. In: Climate Change 2013: The Physical Science

Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press,

Cambridge, United Kingdom and New York, USA.

(33)

33

Kaya, Y., 1990: Impact of Carbon Dioxide Emission Control on GNP Growth:

Interpretation of Proposed Scenarios. Paper presented to the IPCC Energy and

Industry Subgroup, Response Strategies Working Group, Paris.

Keeling, C. D., Carter, A. F., Mook, W. G. 1984. Seasonal, Latitudinal, and Secular Variations in the Abundance and Isotope Ratios of Atmospheric CO2. Journal of

Geophyisical Research, 89: 4615-4628.

Kellogg, W.W. 1979. Influences of mankind on climate. Annual Reviews Earth Planet

Science, 7: 63-92.

Kellogg, W.W. 1987. Mankind's impact on climate: The evolution of an awareness. Climate

Change, 10: 113-136

Lee, T.M., Markowitz, E.M., Howe, P.D., Ko, C., Leiserowitz, A.A. 2015. Predictors of public climate change awareness and risk perception around the world. Nature

Climate Change, 5(11): 1014-1020

Leontief, W. 1953. Structural change, in: W. Leontief et al. (eds). Studies in the Structure of

the American Economy, pp. 17-52. Oxford University Press, New York, USA.

Levin, A., Lin, C., Chu, C.J. 2002. Unit root tests in panel data: asymptotic and finite-sample properties. Journal of Econometrics, 108: 1-24.

Liddle, B., Lung, S. 2010. Age-structure, urbanization, and climate change in developed countries: revisiting STIRPAT for disaggregated population and consumption-related environmental impacts. Population and Environment, 31(5): 317-343.

(34)

34

Michaelowa, A., Rolfe, C. 2001. Early Action to Reduce Greenhouse Gas Emissions Before the Commitment Period of the Kyoto Protocol: Advantages and Disadvantages.

Environmental Management, 28(3): 281-292

Morrisette, P.M., Platinga, A.J. 1991. Global Warming: A Policy Review. Policy Studies

Journal, 19(2): 163-172.

Myhre, G., D. Shindell, F.M. Bréon, W. Collins, J. Fuglestvedt, J. Huang, D. Koch, J.-F. Lamarque, D. Lee, B. Mendoza, T. Nakajima, et al. 2013: Anthropogenic and Natural Radiative Forcing. In: Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Stocker, T.F., D. Qin, G.-K. Plattner, M. Tignor, S.K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex and P.M. Midgley (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.

NOAA: Greenhouse gas benchmark reached. 2025. Silver Spring (MD): National Oceanic and Atmospheric Administration, accessed 2016 March 15th. Available at:

http://research.noaa.gov/News/NewsArchive/LatestNews/TabId/684/ArtMID/1768/Art icleID/11153/Greenhouse-gas-benchmark-reached-.aspx

O'Brien, K.L. 1996. Tropical Deforestation and Climate Change. Progress in Physical

Geography, 20(3): 311-335.

Phylipsen, G.J.M., Bode, J.W., Blok, K. 1998. A Triptych sectoral approach to burden differentation; GHG emisssions in the European bubble. Energy Policy, 26(12): 929-943.

Ratter, G., Philipp, K.H.I., Storch, H.v. 2012. Between hype and decline: recent trends in public perception of climate change. Environmental Science and Policy, 18: 3–8.

(35)

35

Semenza, J.C., Hall, D.E., Wilson, D.J., Bontempo, B.D., Sailor, D.J., George, L.A. 2008. Public Perception of Climate Change: Voluntary Mitigation and Barriers to Behavior Change. American Journal of Preventive Medicine, 35(5): 479-487.

Serrano, M., Dietzenbacher, E. 2010. Responsibility and trade emission balances: An evaluation of approaches. Ecological Economics, 69: 2224-2232.

Timmer, M.P., Dietzenbacher, E., Los, B., Stehrer, R., 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: 575-605.

Tol, R.S.J. 2014. Quantifying the consensus on anthropogenic global warming in the literature: A re-analysis. Energy Policy, 73(27): 701-705.

Tyndall, J. 1863. On Radiation through the Earth's Atmosphere. Philosophy Magazine, 4: 200-207.

UNSTATS 2016. Demographic yearbook, accessed 2016 April 9th. Available at: http://unstats.un.org/unsd/demographic/products/dyb/dyb2005.htm

UNFCCC 2011. Fact sheet: The Kyoto Protocol. United Nations Framework Convention on Climate Change. Available at:

https://unfccc.int/files/press/backgrounders/application/pdf/fact_sheet_the_kyoto_prot ocol.pdf

UNFCCC 2012. Report of the Conference of the Parties on its seventeenth session, held in Durban from 28 November to 11 December 2011. Part Two: Action taken by the Conference of the Parties at its seventeenth session. Available at:

http://unfccc.int/resource/docs/2011/cop17/eng/09a01.pdf

(36)

36

http://unfccc.int/resource/docs/2012/cop18/eng/08a01.pdf

UNFCCC 2015a. Report of the Conference of the Parties on its twenty-first session, held in Paris from 30 November to 13 December 2015 . Part two: Action taken by the Conference of the Parties at its twenty-firstsession. Available at:

http://unfccc.int/resource/docs/2015/cop21/eng/10.pdf

UNFCCC 2015b. Synthesis report on the aggregate effect of the intended nationally determined contributions. Available at:

http://unfccc.int/resource/docs/2015/cop21/eng/07.pdf

Viskanta, R., Bergstrom, R. W., Johnson, R . O. 1977. Radiative transfer in a polluted urban planetary layer. Journal of the Atmospheric Sciences. 34 : 1091-1103.

(37)

37

9. Appendix

EU27 countries Non-EU countries

AUT Austria IRL Ireland AUS Australia

BEL Belgium ITA Italy BRA Brazil

BGR Bulgaria LVA Latvia CAN Canada

CYP Cyprus LTU Lithuania CHN China

CZE Czech Republic LUX Luxembourg IDN Indonesia

DEU Germany MLT Malta IND India

DNK Denmark NLD Netherlands JPN Japan

ESP Spain POL Poland KOR Korea

EST Estonia PRT Portugal MEX Mexico

FIN Finland ROU Romania ROW Rest of the World

FRA France SVK Slovak Republic RUS Russia

GBR United Kingdom SVN Slovenia TWN Taiwan

GRC Greece SWE Sweden TUR Turkey

HUN Hungary USA United States

Table A.1. Country abbreviations.

Table A.2. Correlation table between the regression variables.

(38)

38

Table A.3. Countries grouped based on their performance relative to the benchmark in figure (5). For the meaning of the abbreviations, see table A.1. For USA and CAN either one of the years was missing, so we could not establish their development over time. However, they were sufficiently high above the benchmark that we did not expect them to fall below it in 2009, so that they are both classified as high emitters.

Consistent low emitters Positive crossover emitters Consistent high emitters Negative crossover emitters

BRA GRC AUS FRA

CHN KOR AUT LUX

(39)

39

(1) (2) (3) (4) (5)

ln(GWP per capita) ln(CO2 per capita) - - -

GDP per capita 0.000644*** 0.000683*** 0.000841*** 0.00101*** 0.00116***

(0.000175) (0.000162) (0.000186) (0.000193) (0.000190)

GDP per capita2 -9.80e-07*** -9.21e-07*** -1.09e-06*** -1.16e-06*** -1.15e-06***

(1.21e-07) (9.59e-08) (1.57e-07) (1.31e-07) (1.21e-07)

Urbanization share 0.00869 0.0140* 0.0152* 0.0166**

(0.00624) (0.00777) (0.00771) (0.00743)

Green energy share -0.00109*** -0.00148*** -0.00141*** -0.00140*** -0.00177***

(0.000309) (0.000346) (0.000344) (0.000352) (0.000524)

Industry share value added

0.000578*** 0.000530***

(0.000106) (0.000111)

Agriculture share value added -8.65e-05 (7.77e-05) Manufacturing share value added 0.000150 (0.000133) Services share value

added -0.000455*** (0.000129) Constant -5.286*** -5.725*** -5.935*** -5.873*** -4.903*** (0.460) (0.552) (0.567) (0.526) (0.153) Observations 571 571 549 571 571 # of countries 39 39 38 39 39

Year Effects YES YES YES YES YES

Country effects YES YES YES YES YES

F-statistic 54.42 43.55 42.10 31.84 50.68

R-Squared 0.303 0.420 0.466 0.426 0.269

Adjusted R-squared 0.309 0.321 0.331 0.341 0.316

Future climate change negotiations: Expanding the Triptych approach taking a demand perspective