Master Thesis
MSc. Economic Development and Globalization University of Groningen
Faculty of Economics and Business Submitted June 16, 2020
FOREIGN AND DOMESTIC DRIVERS OF CONSUMPTION-BASED
CO
2EMISSIONS IN DEVELOPED AND DEVELOPING COUNTRIES: A
COUNTRY-LEVEL STRUCTURAL DECOMPOSITION ANALYSIS
Author: Ulrike R. Zijlstra Student number: S2532492
Email: u.r.zijlstra.1@student.rug.nl
Supervisor: prof. dr. H.W.A. Dietzenbacher Co-assessor: prof. dr. B. Los
Abstract
Due to the international fragmentation of production processes, and the carbon leakage associated, the gap between consumption-based emissions and territorial emissions has increased. This study finds that all countries, except for China, Turkey and Indonesia, either leak more carbon or absorb less carbon in 2014 as compared to 2000. All the extra leakage is absorbed by China. This emphasizes the need for consumption-based accounting methods in environmental decomposition analyses. This study aims to identify the drivers of changes in consumption-based CO2 emissions in 13 countries and regions between 2000 and 2014, by applying a country-level structural decomposition analysis to global multiregional input-output data. Across all regions, the main upward drivers of consumption-based CO2 emissions are population growth and increasing consumption per capita. Technological development is the main downward driver. Changes in the domestic trade structure drive up consumption-based emissions in developed economies and in a number of emerging economies, which is due to the fact that a larger share of intermediate inputs and final goods are imported from China.
Keywords: structural decomposition analysis, multi-regional input-output analysis, carbon
1 1 Introduction
Carbon dioxide (CO2) is one of the main causes of human-induced global warming.
Global average temperatures have already increased by approximately 1°C above pre-industrial levels in 2017 and will increase by 0.2°C each decade due to the rising global CO2 emissions.
Thus, global CO2 emission levels will need to be quickly stabilized and reduced to ensure that
global average temperatures do not surpass 2°C above pre-industrial levels, as stipulated in the Paris Agreement (Peters et al., 2017; Le Quéré et al., 2019). Although CO2 emissions in
developing and emerging economies are continuously increasing, some developed countries are showing promising signs of slowing emissions. The USA and various European countries have had a consistent decrease in CO2 emissions since approximately 2005 (Feng et al., 2014; Peters
et al., 2017; and Le Quéré et al., 2019). Assessing the effectiveness of climate policies in these countries is important for the formation of new policies. Therefore, it is important to have an understanding of the factors that drive CO2 changes in emissions (Peters et al., 2017).
Country-level decompositions identify these drivers by breaking down changes in CO2 emissions into
changes in their constituent parts, which can reveal information about the underlying trends that result in changes in emissions.
A country’s CO2 emissions can be analyzed from two different perspectives: the
production-based accounting method and the consumption-based accounting method. Whereas the majority of structural decomposition analyses (SDAs) consider production-based emissions (e.g., Feng et al., 2014; Peters et al, 2017; and Le Quéré et al., 2019), this research conducts a country-level structural decomposition analysis of consumption-based emissions. The first accounting method considers territorial emissions, that is, production-based emissions; it measures the CO2 emitted by industries and households within the geographical territory of a
country. The second accounting method considers consumption-based emissions; it measures global CO2 emissions embodied in a country’s consumption of domestic and imported final
goods and services, and CO2 emissions by households. Therefore, consumption-based CO2
emissions measure a country’s carbon footprint beyond its territorial borders. The difference between the two accounting methods has become increasingly significant. As Peters et al. (2011) highlight, emissions transfers between developed and developing countries via international trade has grown significantly in recent decades, such that the territorial emissions in developed countries increased at a slower pace than their consumption-based emissions, and vice versa for developing countries. Thus, even if territorial emissions in some developed countries are steadily decreasing, this might not be the case for their consumption-based emissions (Peters et al., 2011; Kanemoto et al., 2014).
The difference between production-based accounting and consumption-based accounting can be traced to who is held responsible for CO2 emissions. Since
2
A comparison between the effects of the drivers of increasing consumption-based CO2
emissions and effects of the drivers of decreasing consumption-based CO2 emissions may
reveal information about differences in the underlying trends that drive changes in consumption-based emissions. Moreover, differentiating between domestic and foreign drivers reveals how changes abroad affect a country’s consumption-based emissions. Therefore, this study asks the following research questions: what are the main drivers of changes in consumption-based CO2 emissions, and how do the effects of these drivers differ between
countries with increasing consumption-based CO2 emissions and countries with decreasing CO2
emissions?
Using country-level SDAs, five drivers are used to explain the year-to-year changes in consumption-based CO2emissions between 2000 and 2014. These drivers are (i) technology,
(ii) population, (iii) consumption per capita, (iv) trade structure, and (v) consumption mix. Moreover, the changes in technology and the changes in trade structure will be split into domestic changes and changes abroad. Based on the existing literature, technology is expected to be the most significant downward driver of consumption-based CO2 emissions. Population
and consumption per capita are expected to be the most significant upward drivers of consumption-based CO2 emissions. Moreover, Peters et al. (2017) and Le Quéré et al. (2019)
hypothesize that low GDP growth is one of the main causes of decreasing CO2 emissions in
developed countries. Therefore, the effect of the changes in consumption per capita to changes in consumption-based CO2 emissions is expected to be smaller in developed countries with
decreasing consumption-based CO2 emissions than in countries with increasing CO2 emissions.
This research shows that consumption-based CO2 emissions in Europe and the USA decreased
between 2000 and 2014. Consumption-based emissions increased in all other countries.
This article aims to identify the foreign and domestic drivers of changes in consumption-based CO2 emissions and to identify differences in the effects of the drivers of increasing
consumption-based emissions and the effects of the drivers of decreasing consumption-based emissions. It is structured as follows. The next chapter will briefly discuss the existing literature on drivers of changes in CO2 emissions and the different decomposition techniques used in this
3 2 Literature review
2.1 Drivers of changes in CO2 emissions
Various environmental studies have identified the drivers of greenhouse gas emissions or, specifically, CO2 emissions. Over two decades ago, Dietz and Rosa (1997) suggested that
the anticipated growth in population and economic growth would drive a surge in global CO2
emissions. Indeed, most recent decomposition analyses have found that growth in population and growth in GDP per capita (or consumption per capita) are the main drivers of the growth in CO2 emissions in the past decades. However, technological improvements resulting in lower
CO2 emissions per unit of output and, to some extent, more efficient production structures, are
found to partially moderate the bolstering effects of growth in population and GDP per capita on CO2 (e.g., Raupach et al, 2007; Arto and Dietzenbacher, 2014; Feng et al., 2015). For
example, some technological improvements can create more efficient production processes that require less energy or use a “cleaner” energy source, such as solar energy, that results in fewer emissions.
Territorial CO2 emission in the USA and various countries in Europe have consistently
decreased since the mid-2000s (Feng et al., 2014; Peters et al, 2017; and Le Quéré et al., 2019). A decrease in energy use is the main driver of the decrease in CO2 emissions in these developed
countries. A decrease in energy use could be the result of climate policies promoting energy-efficient production processes. However, some research has found that a significant portion of the decrease in energy use can be attributed to low GDP growth (Peters et al, 2017; Le Quéré et al., 2019). Feng et al. (2014) found that the economic recession was the predominant driver of the sharp decline in CO2 emissions in the USA between 2007 and 2009 and that the following
stabilization of CO2 emissions between 2009 and 2013 was a result of the gradually recovering
economy. Whether the downward trend in territorial CO2 emissions can be sustained in times
of economic expansion partially depends on how the energy intensity of GDP – and thereby the CO2 intensity of GDP1 in these countries – develops. Peters et al. (2017, p. 119) state that “the
declines in energy intensity are an important long-term trend as economies develop, become more efficient, and shift to services.” If the CO2 effects of the improvements in the energy
intensity of GDP outweigh the CO2 effects of a potential acceleration of GDP growth, the
downward trend of territorial emissions may be sustained.
As well as low GDP growth, carbon leakage due to international trade, which has received merited attention in recent environmental studies, may be a driving force of the decrease in territorial emissions in developed countries. As production processes are becoming increasingly internationally fragmented, every product consumed has inputs, and thus embodies CO2 emissions, from all over the world (Timmer et al., 2014). In the first wave of globalization,
technological innovation and trade liberalization in the 19th century resulted in a geographical dispersion between the location of production and the location of consumption of final goods. Falling communication costs in the late 20th century triggered the second wave of globalization, which allowed for the geographical separation of different production stages (Baldwin, 2006).
1 CO
2 intensity of GDP is calculated as the product of CO2 intensity of energy and energy intensity of GDP
4
Countries would specialize in a stage of the global value chain, and the trade of intermediate goods surged. Therefore, the location of production of final goods is not necessarily the same as the location of the production of intermediate inputs for that final good (Baldwin, 2006). The country where the production of intermediate or final goods, or both, takes place can no longer solely be held accountable for the environmental consequences of these products, as the country that consumes the final goods that are made with these intermediate products are also partly accountable (Dietzenbacher et al., 2012; Arto and Dietzenbacher, 2014). Developed countries have shifted a significant portion of their consumption-based emissions from domestic production to production in emerging economies (Peters et al., 2011; Yamakawa and Peters, 2011). Peters et al. (2011) found that from 1990 to 2008, the discrepancy between territorial emissions and consumption-based emissions gradually grew to the point where the two were very different. Therefore, decreasing territorial emissions may partially be the result of carbon leakage due to the international fragmentation of global value chains. Consumption-based CO2
emissions, on the other hand, comprise all direct and indirect emissions from the consumption of final goods, irrespective of where production takes place. Therefore, carbon leakage is accounted for in consumption-based CO2 emissions.
2.2 Decomposition techniques
There are several approaches to assessing the determinants or driving forces of environmental indicators or socio-economic indicators. A long-established method is the IPAT equation, devised by Ehrlich and Holdren in the early 1970s (Chertow, 2000). In the IPAT equation, environmental impact (I) is the product of three factors: population (P); affluence (A) defined as GDP per capita; and technology (T) defined as the environmental impact per unit of GDP (Chertow, 2000). A similar type of decomposition method is the Kaya identity, which also expresses global emissions as a product of population and GDP per capita, but decomposes the third factor, technology, into the energy intensity of GDP and carbon intensity of energy (Raupach et al., 2007; Arto and Dietzenbacher, 2014). A third often-used decomposition technique is index decomposition analysis (IDA). Index decomposition analysis uses aggregated sector information to explore the relationship between impact (e.g., employment, economic growth, or CO2 emission) and production levels.
5
indirect CO2 emissions of a final demand, which are crucial to consumption-based CO2
emissions. Therefore, a country-level SDA in a global multi-regional IO framework is used in this research to decompose year-to-year changes in consumption-based CO2 emissions between
2000 and 2014.
3 Data and methods
3.1 Structural decomposition analysis (SDA) of consumption-based CO2 emissions
The following section describes the methods used to answer the research question. Consumption-based CO2 emissions consist of emissions by industries and household emissions
(𝒉). The emissions by industries are estimated using a global multi-regional IO (GMRIO) framework and are a product of three determinants. First, the emissions coefficient vector 𝒆, where 𝑒𝑖𝑆 represents the emissions by industry 𝑖 in country 𝑆 for one unit of its output. Second, the Leontief matrix 𝑳 where element 𝑙𝑖𝑗𝑆𝑇represents the output by industry 𝑖 in country 𝑆 required for one unit of final demand for good 𝑗 produced in country 𝑇. Third, the final demand matrix 𝑭 where the element 𝑓𝑗𝑇𝑅 represents the final demand in country 𝑅 for good 𝑗 produced in country 𝑇. Consumption based CO2 emissions in country R are given by 𝑔𝑅 =
∑ ∑ ∑ ∑ 𝑒𝑆 𝑇 𝑖 𝑗 𝑖𝑆𝑙𝑖𝑗𝑆𝑇𝑓𝑗𝑇𝑅+ ℎ𝑅.
Changes in consumption-based CO2 emissions 𝒈 can be decomposed into the changes
in its determinants (𝒆, 𝑳, 𝑭, and 𝒉) using an SDA. Subsequently, the changes in 𝑳, 𝑭, and 𝒉 are further decomposed into the changes of their constituent parts. The final demand in country 𝑅 for good 𝑗 produced in country 𝑇 can be calculated as 𝑓𝑗𝑇𝑅 = 𝑡𝑗𝑇𝑅𝑐𝑗𝑅𝑦𝑅𝑝𝑅. From this, it follows that the changes in the final demand matrix 𝑭 are further decomposed into the changes of its four constituent parts. First, the matrix of trade coefficients of final goods (𝑻), where the element 𝑡𝑗𝑇𝑅 indicates the share of total final demand for good 𝑗 in country 𝑅 that is imported from country 𝑇. Second, the matrix of consumption mix (𝑪), of which the element 𝑐𝑗𝑅 indicates the share of total consumption per capita in country 𝑅 that is spent on good 𝑗, irrespective of its country of origin. Third, a vector of consumption per capita (𝒚), where the element 𝑦𝑅 indicates
the final demand per capita in country 𝑅. Finally, the vector (𝒑), of which the element 𝑝𝑅 indicates the total population in country 𝑅. In a similar manner, the changes in the Leontief matrix are further decomposed into intermediate trade coefficients (𝑸) and production structure coefficients (𝑩). Finally, household emissions are decomposed into direct household emissions per unit of final consumption (𝒅), consumption per capita (𝒚), and the population (𝒑). Using an SDA, the contribution of each of the nine determinants to changes in consumption-based CO2
emissions can be calculated. Full details of the decomposition are given in Appendix A. Here, a simplified model of an indicator with only two determinants is used to outline an SDA. In the example, 𝛼 is determined by 𝛽 and 𝛾, such that 𝛼 = 𝛽 ∗ 𝛾. The change in 𝛼 between period 0 and period 1 (∆𝛼1−0) is driven by changes in 𝛽 and 𝛾, such that
∆𝛼1−0 = 𝛼1− 𝛼0= (𝛽1−𝛽0)𝛾1+𝛽0(𝛾1−𝛾0) (1.1)
6 =1
2(∆𝛽1−0)(𝛾0+𝛾1) + 1
2(𝛽0+𝛽1)(∆𝛾1−0) (1.3)
The first term on the right-hand side of (1.3) indicates the contribution of ∆𝛽1−0 to ∆𝛼1−0, and
the second term indicates the contribution of ∆𝛾1−0 to ∆𝛼1−0. Since this simplified model includes two determinants, there are two decomposition forms (1.1 and 1.2), and the average (1.3) is considered. The number of decomposition forms increases as the number of determinants increases, such that the decomposition of an indicator with 𝑚 determinants results in 𝑚! unique decomposition forms. Dietzenbacher and Los (1998) address the issue of non-uniqueness of different decomposition forms. They conclude that the average of two so-called polar equations (as in 1.3) is strikingly close to the average of 𝑚! decompositions of a change in an indicator with 𝑚 determinants.
In this research, the year-to-year changes in consumption-based CO2 emissions are
decomposed, and the nine determinants are grouped into five drivers, namely
• Changes in the trade structure – includes changes in the trade structure of intermediate goods and services and changes in the trade structure of final goods and services. A change in the trade structure of final goods implies a change in the share of the total final consumption of good 𝑗 in country 𝑅 that is imported from country 𝑇. A change in the intermediate trade coefficient implies a change in how the input from sector 𝑖 for the production of good 𝑗 in country 𝑇 is divided over the countries of origin of that input from sector 𝑖.
• Changes in population;
• Changes in the consumption per capita;
• Changes in the consumption mix of final goods – implies which share of total consumption in country 𝑅 is spent on good 𝑗, irrespective of the country of origin. • Changes in technology –includes changes in industry emission coefficients, changes in
production structure coefficients, and changes in direct household emission coefficients. Production structure coefficients give the input of good 𝑖 (irrespective of the country of origin) that is required for one unit of output from sector 𝑗 in country 𝑇.
For a detailed explanation of the methodology, please refer to Appendix A.
3.2 Measuring emissions in the previous year’s prices
The World Input Output Database (WIOD) offers its World IO Tables (WIOT) both in current prices and the previous year’s prices. Measuring trade using the previous year’s prices allows for measuring changes in trade in terms of volume rather than the value of output (see Appendix A). Measuring trade in terms of volume is relevant when considering a trade in emissions, as is the case in consumption-based emissions accounting, because emissions per unit of output play a pivotal role.
7
an industry has not changed, the emissions will likely stay constant as well. Consequently, if the emission coefficients are calculated by dividing the emissions of an industry by the output (measured in current prices) of an industry, the emissions coefficients in year t+1 will be lower than in year t. Dietzenbacher et al. (2020) found that correcting for price changes by using IO tables in the previous year’s prices resulted in very different outcomes of their SDA of changes in renewable energy use. The authors found that when using IO tables in current prices instead of IO tables in the previous year’s prices, the effects of the changes in the final demand per capita and of technology changes on renewable energy use were 125% and 175% larger, respectively. In other words, the effects of changes in the final demand per capita and in the technology on renewable energy use were significantly overestimated by not correcting for price changes.
3.3 Database
Analyzing consumption-based emissions requires GMRIO data that covers a broad set of countries and environmental accounts that are classified by the same countries and sectors as the GMRIO tables. While several databases include GMRIO tables and corresponding environmental accounts, this research uses the WIOD. The WIOD tables are provided in both current prices and the previous year’s prices, which is not the case for other IO databases, such as Exiobase, Eora, and the Organization for Economic Cooperation and Development (OECD) IO tables. The WIOT of the WIOD 2016 Release cover 28 EU countries and 15 other major countries, each divided into 56 industries, for the years 2000 to 2014 (Timmer et al., 2015). The European Commission released environmental accounts aligned to the WIOD 2016 Release (Corsatea et al., 2019). These environmental accounts include data on industry-level CO2
emissions, and the database covers the same countries and industries as the WIOD IO tables. This newly released database allows for country-level SDAs of the drivers of year-to-year changes in consumption-based emissions.
Data on the population sizes in the countries included and in the rest of the world between 2000 and 2014 were retrieved from the OECD Database of world indicators (OECD, 2020).
4. Results and discussion
The following sections will present and analyze the differences between in growth consumption-based emissions and territorial emissions, present the results of the SDAs of the changes in consumption-based emissions and discuss the underlying.
4.1 Production-based CO2 emissions and consumption-based CO2 emissions
Table 1 presents consumption-based CO2 emissions and territorial CO2 emissions in the
years 2000 and 2014 for 12 regions and countries and the countries labelled as the ‘rest of the world’ (RoW; see Appendix A.9 for the countries included). The last row presents the world totals. The first three columns present, in order, territorial CO2 emissions in the years 2000 and
2014 and the difference between these years. The following three columns display the same information for consumption-based CO2 emissions. The seventh column displays the difference
8
Table 1
Territorial emissions and consumption-based CO2 emissions between 2000-2014 (in Mt)
Territorial Emissions (PBA) Consumption-based
Emissions (CBA)
∆CBA-∆PBA CO2 trade balance 2000 2014 ∆ 2000 2014 ∆ USA 5854 5258 -596 6656 6496 -160 437 Importer EU 4625 3804 -822 5133 4517 -617 205 Importer Russia 1584 1699 115 902 1136 234 119 Exporter Australia 376 405 29 358 480 123 94 Importer Canada 527 577 50 457 555 98 49 Exporter Brazil 360 572 212 370 614 244 32 Importer Mexico 407 489 82 440 547 107 26 Importer EAS 2086 2237 151 2278 2450 172 21 Importer India 995 2193 1198 938 2142 1203 5 Exporter Indonesia 333 543 210 286 487 201 -8 Exporter Turkey 226 357 132 255 354 99 -32 Exporter China 3668 10529 6861 3223 8962 5739 -1122 Exporter RoW 4639 7835 3196 4386 7758 3372 176 Exporter Total 25680 39497 10817 25680 39497 10817
Finally, the last column indicates whether the country or region was a net importer or a net exporter of CO2 emissions in 2014. A country is a net exporter if territorial emissions are larger
than consumption-based emissions and vice versa.
As anticipated, the decrease in consumption-based emissions in Europe and the USA is lower than the decrease in production-based emissions. Accordingly, carbon leakage from Europe and the USA has increased, which implies that the already substantial gap between consumption-based emissions and territorial emissions in these countries grew larger between 2000 and 2014. Similarly, consumption-based emissions in Australia, Brazil, Mexico, and the East Asian region (EAS) have grown more than their territorial emissions, also increasing the gap between their consumption-based emissions and territorial emissions.
India, Canada, Russia, and the RoW countries remained net exporters of CO2 emissions
between 2000 and 2014. However, the gap between consumption-based emissions and territorial emissions decreased in these countries. This is because their consumption-based emissions grew to a greater extent than their territorial emissions, which implies that these countries absorbed less CO2 leakage in 2014 than in 2000.
As opposed to the countries mentioned above, China, Indonesia, and Turkey have lower absolute growth in consumption-based emissions than in territorial emissions. Turkey was a net importer of CO2 emissions in 2000, but its territorial emissions grew significantly faster than
its consumption-based emissions, resulting in a net export of CO2 emissions from Turkey in
2014. Indonesia remained a net exporter of emissions, increasing the gap between consumption-based emissions and territorial emissions between 2000 and 2014. However, by far the most notable results are for China. In 2014, China absorbed 1,122 Mt. more carbon leakage than it did in the year 2000, substantially increasing the gap between their consumption-based emissions and territorial emissions.
9
countries in the world, except Turkey, Indonesia, and China, either leak more carbon or absorb fewer leakages in 2014 than they did in 2000. As can be seen in Table 1, nearly all these extra leakages are absorbed by China. While decreasing emissions in Europe and in the USA suggest a positive development, these countries (and many other countries) increasingly off-shore CO2
emissions to China.
4.2 Decomposition results
Global CO2 emissions grew by 10.8 Gt. between 2000 and 2014. Figure 1 presents the
result of an SDA of year-to-year changes in global CO2 emissions between 2000 and 2014. The
green line indicates the total change in global CO2 emissions; the five remaining lines indicate
how each of the drivers contributes to the change in global emissions. Consumption per capita (FD per capita) is the main upward driver of global CO2 emissions. Similarly, population
growth drives up CO2 emissions steadily. Changes in technology, due to lower emission
coefficients and, to some extent, more efficient production structures, are the main downward driver of global CO2 emissions. Changes in the consumption mix have no significant effect on
global emissions. Finally, and most notably, there was a positive effect of changes in the trade structure on global CO2 emissions. As highlighted in Section 4.1, many countries have shifted
an increasing share of their consumption-based emissions to China. This implies that more intermediate inputs and final goods are imported from China, where the industry emissions coefficients are generally higher than those of developed countries.
Figure 1: Year-to-year decomposition of changes in global CO2 emissions between 2000 and 2014
4.2.1 Decomposition of decreasing consumption-based emissions in Europe and the USA
Figures 2 and 3 present the year-to-year decomposition results of changes in consumption-based CO2 emissions in Europe and the USA, respectively. As can be observed
in the graph in Figure 2, consumption-based emissions in Europe almost consistently decreased between 2006 and 2014 and in total decreased by 617 Mt. CO2. Consumption-based emissions
decreased steeply during the financial crisis of 2008-2009, which coincided with a drop in the effects of changes in consumption per capita and the consumption mix on changes in consumption-based emissions. The effect of consumption per capita stabilized after 2009 and consumption-based emissions in Europe decreased gradually due to technological
Population Trade; Consumption mix; FD per capita Technology Total (10817 Mt) -15000 -10000 -5000 0 5000 10000 15000 20000 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Cha n ge in CO2 (in Mt)
10
developments. A similar pattern is observed for the USA in Figure 3. However, the decrease in consumption-based CO2 emissions in the USA is less consistent and totals a decrease of 160
Mt. CO2 between 2000 and 2014.
An unexpected upward driver of consumption-based emissions in Europe and in the USA is the effect of the changes in the domestic trade structure, which includes the trade coefficients of intermediate goods and the trade coefficients of final goods. This effect is suspected to occur because more intermediate inputs and consumption goods are imported from China instead of from developed countries. The effect of changes in the domestic trade structure on consumption based emissions is most prevalent in countries that had increased carbon leakage between 2000 and 2014, such as the USA and Europe, and less so in countries that absorbed more leakages, especially China. Figure 4 presents the growth in Chinese imports and total foreign imports as shares of total intermediate input and total final consumption between 2000 and 2014 in Europe2 and the USA, in percentage points. As can be seen in the graph, between 2000 and 2014 imported final goods as share of total final consumption in Europe grew with 1.47 percentage point of which more than half, 0.78 percentage point, came from China. In the USA, the imported share of final consumption actually shrunk by 0.26 percentage points, while Chinese imports of final goods as share of total final consumption increased by 0.87 percentage points. A disproportionate share of the growth in the imported share of final consumption in Europe and in the USA comes from China. While less prominent, a similar pattern is seen for the Chinese share of the growth in the imported share of total intermediate input. The increased share of Chinese imports in total intermediate input and final consumption drives up consumption-based CO2 emissions because emission coefficients are generally
higher in China than in the developed Western economies, which implies that there are more emissions associated with Chinese imports than with locally sourced imports.
Figure 2: Year-to-year decomposition of changes in consumption based CO2 emissions in Europe
2 The imported share of intermediate inputs and final consumption in Europe is calculated as the weighted
average of the import shares of the countries in Europe, such that final goods from Germany that are consumed in the Netherlands are seen as imports.
11
Figure 3: Year-to-year decomposition of changes in consumption-based CO2 emissions in the USA
Figure 4: Growth in imported intermediate goods as share of total intermediate inputs and imported final goods as share of total final consumption between 2000 and 2014, in percentage points
The effect of changes in technology is by far the largest downward driver of consumption-based emissions in Europe and the USA. Domestic and foreign changes in technology have a similar decreasing effect on changes in consumption-based emissions in Europe. In the USA, however, the effect of domestic changes in technology is far greater than the effect of foreign changes in technology. The relatively large contribution of foreign changes in technology to changes in consumption-based emissions in Europe may be because economies in Europe are more open than in the USA. Figure 5 presents the imported share of total intermediate inputs and the imported share of total final consumption for Europe and the USA in 2000 and 2014. As can be seen in the graph, imported input shares and imported consumption shares in Europe are higher than in the USA both in 2000 and 2014. Consequently, the effect of foreign changes in technology on changes in consumption-based CO2 emissions was larger
in Europe than in the USA. The weighted average of the industry emission coefficients in Trade (d) Trade (f) Population Consumption mix FD per capita Technology (d) Technology (f) Total (-160 Mt) -1500 -1000 -500 0 500 1000 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Cha n ge in CO2 (in Mt)
USA
0,83% 0,78% 0,78% 0,87% 5,08% 1,47% 2,35% -0,26% -1,00% 0,00% 1,00% 2,00% 3,00% 4,00% 5,00% 6,00%Intermediate input Final consumption Intermediate input Final consumption
EUR EUR USA USA
Growth of Chinese import shares and total import shares of
intermediate input and final consumption
Chinese input as share of total input
12
Europe and the USA decreased by approximately the same amount between 2000 and 2014. Therefore, a significant difference in technological development between Europe and the USA does not seem like a viable alternative explanation for the low effect of foreign changes in technology to consumption-based emissions in the USA (refer to Figure 17 in Appendix B.3 for detailed data).
Finally, whereas changes in the consumption mix are generally found to have no significant impact on emissions, changes in the consumption mix in Europe and the USA seem to consistently negatively contribute to consumption-based emissions, especially after 2008. This implies that consumers in Europe and the USA consume fewer CO2-intensive goods. This
might be due to the growing awareness of climate change among consumers in these countries. An alternative explanation could be that consumers are making different consumption choices due to low GDP growth, as is reflected in the moderate growth in consumption per capita in Europe and the USA.
Figure 5: Imported intermediate inputs and imported final goods as share of total intermediate inputs and final consumption in Europe and the USA in 2000 and 2014
4.2.2 Decomposition of increasing consumption-based emissions
The effects of the different drivers of increasing consumption-based emissions differ across countries. First, the decomposition results for various countries are relatively similar to those of Europe and the USA. Figure 6 presents the decomposition results of increasing consumption-based CO2 emissions in Canada, which is a stand-in for Australia, the EAS region,
Mexico, and Turkey (see Appendix B.1 for decomposition results of these countries). Similarly to Europe and the USA, the change in the final demand per capita is the largest contributor to increasing consumption-based emissions between 2000 and 2014 in Canada. However, the drop in the effect of changes in final demand per capita in Canada between 2008 and 2009 due to the financial crisis is not nearly as significant as in Europe and the USA. The effect of final demand per capita increased rapidly after 2009, and the consumption-based emission path returned to its upward trend. Moreover, the changes in the consumption mix do not have as significant a downward effect on consumption-based emissions in Canada as they do in Europe and the USA. Similarly to Europe and the USA, changes in domestic trade coefficients drive up emissions and foreign and domestic changes in technology are the main downward drivers of consumption-based CO2 emissions in these countries mentioned above. The significant effect
19,8% 12,0% 7,9% 5,9% 24,8% 13,4% 10,3% 5,7% 0% 5% 10% 15% 20% 25% 30%
Intermediate input Final consumption Intermediate input Final consumption
EUR EUR USA USA
Import shares of intermediate input and final consumption
13
of foreign changes in technology on changes in consumption-based emissions in Canada can be explained by the fact that Canada is a very open economy with a large share of imports from the USA. In 2014, 6.2 percent of final consumption and 12.1 of intermediate inputs in Canada were imported from the USA.
Figure 6: Structural decomposition analysis of changes in consumption based CO2 emissions in Canada
Figure 7: Structural decomposition analysis of changes in consumption based CO2 emissions in China
Second, China, Indonesia, India, Russia, and the (RoW) countries have different decomposition results than Europe and the USA. Figure 7 presents the year-to-year decomposition results of changes in consumption-based emissions in China, which is taken as a stand-in for these countries (see Appendix B.2 for SDAs for Indonesia, India, Russia, and the RoW). The first main difference is that the global financial crisis of 2008-2009 does not seem to have affected consumption-based emissions in China. Secondly, the main downward driver is the domestic change in technology, whereas foreign changes in technology have little to no effect on changes in consumption-based emissions in China and the countries mentioned above (except for the RoW). In China, this is due to the fact that its economy is large and relatively closed. Between 2000 and 2014, the imported share of intermediate inputs decreased from 7.8
14
percent to 6.4 percent and the imported share of final consumption decreased from 5.2 percent to 4.8 percent. Moreover, the weighted average of Chinese industry-emissions coefficients decreased by 70 percent between 2000 and 2014, which also explains the large effect of domestic changes in technology on changes in consumption-based emissions (please refer to Figure 17 in Appendix B.3 for detailed data). The remaining countries mentioned above are also large economies3 that are less dependent on imported goods, similar to the USA and
differently to the countries in Europe, which results in a small effect of foreign changes in technology on consumption-based emissions. Finally, the effect of changes in the domestic trade structure, which is important for Europe and the USA, seems to contribute very little to the consumption-based emissions in China and the countries mentioned above.
In conclusion, the effects of the drivers of decreasing consumption-based CO2 emissions
differ across countries and regions. However, consumption per capita, population growth, and domestic changes in technology are the main drivers of decreasing consumption-based emissions across all countries and regions. The differences are mainly found in the contributions of domestic changes in the trade structure and foreign changes in technology to changes in consumption-based emissions.
5. Conclusion and recommendations
In this study, changes in global CO2 emissions and country-level changes in
consumption-based CO2 emissions for 13 countries and regions between 2000 and 2014 are
decomposed into the changes of 5 (foreign and domestic) drivers. Moreover, the differences between changes in consumption-based emissions and territorial emissions are analyzed. There are four main conclusions to be drawn from the results and analyses presented in Section 4.
First, decreasing territorial emissions in Europe and the USA can be partially explained by increased carbon leakage between 2000 and 2014. The decrease in consumption-based emissions is significantly smaller than the decrease in territorial emissions. Therefore, the gap between consumption-based emissions and territorial emissions grew in almost all countries between 2000 and 2014. Moreover, almost all countries leaked more carbon or absorbed fewer leakages in 2014 than they did in 2000. Most of this extra carbon leakage is absorbed by China, which has increasingly become the ‘carbon haven’ of the world. The increasing gap between based emissions and territorial emissions emphasizes the need for consumption-based accounting methods in decomposition analyses.
The second conclusion answers the question of whether the effects of the drivers of increasing consumption-based emissions differ from the effects of the drivers of decreasing consumption-based emissions. The effects of the changes in consumption per capita are smaller in countries with decreasing consumption-based emissions. The low effect of changes in consumption per capita around and after 2008 partially explains the decrease in consumption-based emissions in Europe and the USA between 2000 and 2014. Whereas many countries experienced a trough in consumption per capita (and consumption-based CO2 emissions,
accordingly) due to the global financial crisis, only in Europe and the USA did these effects persist after 2009. The changes in consumption per capita barely brought the based emissions to their pre-crisis level in Europe and the USA, whereas the
consumption-3 The countries grouped under RoW act as one country in the GMRIO framework, such that international trade
15
based emissions quickly grew in other countries after 2009. Moreover, the downward effects of the changes in the consumption mix and changes in technology are larger in countries with decreasing consumption-based emissions than in countries with increasing consumption-based emissions.
Third, changes in trade structures had a positive effect on global CO2 emissions and
domestic CO2 emissions of several developed and emerging countries. Whereas Arto and
Dietzenbacher (2014) found that changes in the trade structure had no significant effect on global emissions between 1995 and 2010, this research finds that changes in the trade structure alone drove up global CO2 emissions by 2.25 Gt. between 2000 and 2014. This is because more
production activities are shifted from developed countries to China, where emission coefficients are generally higher than in developed countries.
Finally, in large, relatively closed economies, domestic changes in technology contribute significantly more to consumption-based emissions than foreign changes in technology do. Consumption-based emissions in smaller, more open economies are more sensitive to foreign changes in technology. Moreover, foreign changes in trade structures have no significant effect on changes in consumption-based emissions, whereas domestic changes in trade structures do have a significant upward effect on consumption-based emissions in various countries. The fact that domestic changes overall have a larger effect on consumption-based emissions than foreign changes do supports the statement that the responsibility of consumption-based emissions lies with the government and especially the consumers of a country.
A limitation of this research is that not all underlying trends of the drivers of changes in consumption-based CO2 emissions have been identified. The causes of the larger effects of
changes in technology and changes in the consumption mix on increasing consumption-based emissions can be identified by further decomposing these drivers into industry-level changes. By disentangling the changes in technology, one can identify in which industries the technological change was made. This allows for a more detailed assessment of the effectiveness of climate policies. Similarly, disentangling changes in the consumption mix can reveal whether consumers indeed make different consumption choices due to awareness of climate change.
The decomposition analysis of global CO2 emissions as presented in Section 4.2 shows
that global emissions increased by 10.8 Gt. CO2 between 2000 and 2014. While
consumption-based emissions in Europe and the USA have decreased since the mid-2000s, the total decrease of 0.78 Gt. CO2 does not counter the aggregated growth in CO2 emissions in the rest of the
world. Forecasting consumption-based CO2 emissions in both developed and developing
countries may reveal information on whether the current climate policy efforts are sufficient to maintain the downward trend in developed countries, and – if so – whether the downward trend is sufficient to counterbalance the upward trend in developing economies. While some studies create forecasting scenarios of a specific indicator based on SDA results (e.g., Hoekstra and Van den Bergh, 2006; Guan et al., 2008),little research has been done to project future changes in consumption-based CO2 emissions. This is beyond the scope of this research but is potentially
16 References
Arto, I., and Dietzenbacher, E. (2014). Drivers of the Growth in Global Greenhouse Gas Emissions. Environmental Science & Technology, 48(10), 5388-5394.
Baldwin, R. (2006). Globalisation: The Great Unbundling(s). in: Globalisation Challenges for
Europe and Finland.
Cansino, J., Román, R., and Ordóñez, M. (2016). Main Drivers of Changes in CO2 Emissions in the Spanish Economy: A Structural Decomposition Analysis. Energy Policy, 89, 150-159. Corsatea, T., Lindner, S., Arto, I., Román, M., Rueda-Cantuche, J., Velázques Afonso, A., Amores, A., and Neuwahl, F. (2019). World Input-Output Database Environmental Accounts. Update 2000-2016. Publications of the European Union, ISBN 978-92-79-64439-9,
doi:10.2791/947252, JRC116234.
Dietz, T., and Rosa, E. (1997). Effects of Population and Affluence on CO2 Emissions.
Proceedings of the National Academy of Sciences of the United States of America, 94(1),
175-179.
Dietzenbacher, E., Kulionis, V., and Capurro, F. (2020) Measuring the effects of energy transition: A structural decomposition analysis of the change in renewable energy use between 2000 and 2014. Applied Energy, 258, 114040.
Dietzenbacher, E., and Los, B. (1998). Structural Decomposition Techniques: Sense and Sensitivity. Economic Systems Research, 10(4), 307-324.
Dietzenbacher, E., Pei, J., and Yang, C. (2012). Trade, Production Fragmentation, and China's Carbon Dioxide Emissions. Journal of Environmental Economics and Management, 64(1), 88-101.
Guan, D., Hubacek, K., Weber, C., Peters, G., and Reiner, D. (2008). The Drivers of Chinese CO2 Emissions from 1980 to 2030. Global Environmental Change, 18, 626-634.
Hoekstra, R., and Van den Bergh, J. (2006). The Impact of Structural Change on Physical Flows in the Economy: Forecasting and Backcasting using Structural Decomposition Analysis. Land Economics, 82(4), 582-601.
Hoekstra, R., and Van den Bergh, J. (2003). Comparing structural and index decomposition analysis. Energy Economics, 25(1), 39-64.
Kanemoto, K., Moran, D., Lenzen, M., and Geschke, A. (2014). International Trade
Undermines National Emission Reduction Targets: New Evidence from Air Pollution. Global
Environmental Change, 24, 52-59.
Le Quéré, C., Korsbakken, J., Wilson, C., Tosun, J., Andrew, R., Andres, R., Canadell, J., Jordan, A., Peters, G. and van Vuuren, D. (2019). Drivers of declining CO2 emissions in 18 developed economies. Nature Climate Change, 9(3), 213-217.
17
Peters, G., Andrews, R., Canadell, J., Fuss, S., Jackson, R., Korsbakken, J., Le Quéré, C., and Nakicenovic, N. (2017). Key Indicators to Track Current Progress and Future Ambition of the Paris Agreement. Nature Climate Change, 7, 118-123.
Peters, G., Minx, J., Weber, C., and Edenhofer, O. (2011). Growth in emission transfers via international trade from 1990 to 2008. Proceedings of The National Academy of Sciences,
108(21), 8903-8908
Raupach, M., Marland, G., Ciais, P., Le Quéré, C., Canadell, J., Klepper, G. and Field, C. (2007). Global and Regional Drivers of Accelerating CO2 Emissions. PNAS, 104(24), 10288-10293.
Rose, A. and Caslers, S. (1996). Input–Output Structural Decomposition Analysis: A Critical Appraisal. Economic Systems Research, 8(1), 33-62.
Timmer, M., Dietzenbacher, E., Los, B., Stehrer, R. and de Vries, G. (2015). An Illustrated User Guide to the World Input–Output Database: The Case of Global Automotive Production.
Review of International Economics., 23, 575–605
Timmer, M., Erumban, A., Los, B., Stehrer, R., and de Vries, G. (2014). Slicing Up Global Value Chains. Journal Of Economic Perspectives, 28(2), 99-118.
18 Appendix A
A.1 GMRIO Framework
The following section outlines a global multiregional input output (GMRIO) framework with 𝑁 countries, each with 𝑛 industries. Capital bold letters indicate a matrix, bold letters indicate vectors, and scalars are indicated by regular cursive letters. The (𝑁𝑛 × 𝑁𝑛) matrix 𝒁 is a matrix of intermediate deliveries, the (𝑁𝑛 × 𝑁) matrix 𝑭 is a matrix of final demand, and the 𝑁𝑛 vector 𝒙 is a vector of total output which is given by 𝒙 = 𝒁𝒖 + 𝑭𝒖, where 𝒖 is a summation vector of ones of the appropriate length. The 𝑁𝑛 vector 𝑭𝒖 is a vector of total final demand. The 𝑁𝑛 × 𝑁𝑛 matrix 𝒁, the 𝑁𝑛 × 𝑁 matrix 𝑭, and the 𝑁𝑛 × 1 vector 𝒙 are defined as follows
𝒁 = [ 𝑍11 ⋯ 𝑍1𝑅 ⋮ ⋱ ⋮ 𝑍𝑅1 ⋯ 𝑍𝑅𝑅 ⋯ 𝑍1𝑁 ⋯ ⋮ ⋯ 𝑍𝑅𝑁 ⋮ ⋯ ⋮ 𝑍𝑁1 ⋯ 𝑍𝑁𝑅 ⋱ ⋮ ⋯ 𝑍𝑁𝑁] 𝑭 = [ 𝒇11 ⋯ 𝒇1𝑅 ⋮ ⋱ ⋮ 𝒇𝑅1 ⋯ 𝒇𝑅𝑅 ⋯ 𝒇1𝑁 ⋯ ⋮ ⋯ 𝒇𝑅𝑁 ⋮ ⋯ ⋮ 𝒇𝑁1 ⋯ 𝒇𝑁𝑅 ⋱ ⋮ ⋯ 𝒇𝑁𝑁] 𝒙 = [ 𝒙1 ⋮ 𝒙𝑆 ⋮ 𝒙𝑁] (2)
The element 𝑧𝑖𝑗𝑅𝑆of the (𝑛 × 𝑛) matrix 𝒁𝑅𝑆 indicates the output from sector 𝑖 in country 𝑅 used for the output of sector 𝑗 in country 𝑆 (in millions of dollars). The element 𝑓𝑗𝑅𝑆of the 𝑛 vector 𝒇𝑅𝑆 indicates the final demand in country 𝑆 for good 𝑗 from country 𝑅. The element 𝑥
𝑗𝑆 indicates
total output by sector 𝑗 in country 𝑆. The (𝑁𝑛 × 𝑁𝑛) matrix 𝑨 is a matrix of input coefficients is given by 𝑨 = 𝒁𝒙̂−𝟏, 𝑨 = [ 𝐴11 ⋯ 𝐴1𝑅 ⋮ ⋱ ⋮ 𝐴𝑅1 ⋯ 𝐴𝑅𝑅 ⋯ 𝐴1𝑁 ⋯ ⋮ ⋯ 𝐴𝑅𝑁 ⋮ ⋯ ⋮ 𝐴𝑁1 ⋯ 𝐴𝑁𝑅 ⋱ ⋮ ⋯ 𝐴𝑁𝑁] (3)
where the element 𝑎𝑖𝑗𝑅𝑆of the (𝑛 × 𝑛) matrix 𝑨𝑅𝑆 indicates the input from sector 𝑖 in country 𝑅
used for one unit of output from sector 𝑗 in country 𝑆. From this follows 𝒙 = (𝑨 − 𝑰)−𝟏𝑭𝒖 =
𝑳𝑭𝒖. The (𝑁𝑛 × 𝑁𝑛) matrix L is the Leontief inverse
19
where the element 𝑙𝑖𝑗𝑅𝑆of the (𝑛 × 𝑛) matrix 𝑳𝑅𝑆 indicates the output from sector 𝑖 in country 𝑅
required (directly and indirectly) for one unit of final demand for good 𝑗 from country 𝑆. Industry CO2 emissions per are given by the 𝑁𝑛 vector 𝒘,
𝒘 = [ 𝒘1 ⋮ 𝒘𝑆 ⋮ 𝒘𝑁] (5)
where the element 𝑤𝑗𝑆indicates total emissions by sector 𝑗 in country 𝑆. The 𝑁𝑛 vector of emission coefficients 𝒆 is given by 𝒆 = 𝒘𝒙̂−𝟏,
𝒆 = [ 𝒆1 ⋮ 𝒆𝑆 ⋮ 𝒆𝑁] (6)
where the element 𝑒𝑗𝑆 indicates the emissions per unit of output by sector 𝑗 in country 𝑆. Household emissions are represented by the 𝑁 vector 𝒉,
𝒉 = [ ℎ1 ⋮ ℎ𝑆 ⋮ ℎ𝑁] (7)
where element ℎ𝑆 indicates total household emissions in country 𝑆.
A.2 Consumption based CO2 emissions
The 𝑁-element row vector of consumption based emissions 𝒈 is given by
𝒈 = 𝒆′𝑳𝑭 + 𝒉′ (8)
𝒈 = [𝑔1 ⋯ 𝑔𝑆 ⋯ 𝑔𝑁] (9)
where the element 𝑔𝑆 indicates the total consumption based emissions in country 𝑆.
A.3 Final demand decomposed
Each element 𝑓𝑗𝑅𝑆 of the (𝑁𝑛 × 𝑁) final demand matrix 𝑭 is the product of four components, 𝑓𝑗𝑅𝑆 = 𝑓𝑗 𝑅𝑆 ∑ 𝑓𝑅 𝑗𝑅𝑆 ∑ 𝑓𝑅 𝑗𝑅𝑆 ∑ ∑ 𝑓𝑗 𝑅 𝑗𝑅𝑆 ∑ ∑ 𝑓𝑗 𝑅 𝑗𝑅𝑆 𝑝𝑆 𝑝𝑆 = 𝑡𝑗𝑅𝑆𝑐𝑗𝑆𝑦𝑆𝑝𝑆 (10)
Final demand in country 𝑆 for good 𝑗 from country 𝑅 (𝑓𝑗𝑅𝑆) is split into (1) population in country 𝑆 (𝑝𝑆); (2) consumption per capita in country 𝑆 (𝑦𝑆 = ∑ ∑ 𝑓𝑗 𝑅 𝑗𝑅𝑆⁄𝑝𝑆) ; (3) share of total consumption in country 𝑆 spend on good 𝑗, (𝑐𝑗𝑆= ∑ 𝑓𝑅 𝑗𝑅𝑆 ∑ ∑ 𝑓𝑗𝑅𝑆
𝑅
𝑗 )
⁄ ; and (4) the share of total
consumption of good 𝑗 in country 𝑆 that is imported from country 𝑅 (𝑡𝑗𝑅𝑆 = 𝑓𝑗𝑅𝑆 ∑ 𝑓𝑗𝑅𝑆
𝑅 )
⁄ . In
20 𝑭 = [ 𝒇11 ⋯ 𝒇1𝑅 ⋮ ⋱ ⋮ 𝒇𝑅1 ⋯ 𝒇𝑅𝑅 ⋯ 𝒇1𝑁 ⋯ ⋮ ⋯ 𝒇𝑅𝑁 ⋮ ⋯ ⋮ 𝒇𝑁1 ⋯ 𝒇𝑁𝑅 ⋱ ⋮ ⋯ 𝒇𝑁𝑁] = ([ 𝒕11 ⋯ 𝒕1𝑅 ⋮ ⋱ ⋮ 𝒕𝑅1 ⋯ 𝒕𝑅𝑅 ⋯ 𝒕1𝑁 ⋯ ⋮ ⋯ 𝒕𝑅𝑁 ⋮ ⋯ ⋮ 𝒕𝑁1 ⋯ 𝒕𝑁𝑅 ⋱ ⋮ ⋯ 𝒕𝑁𝑁] ⊗ [ 𝒄1 ⋯ 𝒄𝑅 ⋮ ⋱ ⋮ 𝒄1 ⋯ 𝒄𝑅 ⋯ 𝒄𝑁 ⋯ ⋮ ⋯ 𝒄𝑁 ⋮ ⋯ ⋮ 𝒄1 ⋯ 𝒄𝑅 ⋱ ⋮ ⋯ 𝒄𝑁] ) × [ 𝑦1 ⋯ 0 ⋯ 0 ⋮ ⋱ ⋮ ⋯ ⋮ 0 ⋯ 𝑦𝑅 ⋯ 0 ⋮ ⋯ ⋮ ⋱ ⋮ 0 ⋯ 0 ⋯ 𝑦𝑁] × [ 𝑝1 ⋯ 0 ⋯ 0 ⋮ ⋱ ⋮ ⋯ ⋮ 0 ⋯ 𝑝𝑅 ⋯ 0 ⋮ ⋯ ⋮ ⋱ ⋮ 0 ⋯ 0 ⋯ 𝑝𝑁] Such that 𝑭 = (𝑻 ⊗ 𝑪)𝒚̂𝒑̂ (11)
Where ⊗ indicates the elementwise multiplication of two matrices, also known as the Hadamard product.
A.4 Direct household emissions
Direct household emissions ℎ𝑆 can be disintegrated into three components
ℎ𝑆 =∑ ∑ 𝑓ℎ𝑆 𝑗𝑅𝑆 𝑅 𝑗 ∑ ∑ 𝑓𝑗 𝑅 𝑗𝑅𝑆 𝑝𝑆 𝑝 𝑆 = 𝑑𝑆𝑦𝑆𝑝𝑆 (12)
Namely (1) direct household emissions per unit of final consumption in country 𝑆 (𝑢𝑆 = ℎ𝑆 ∑ ∑ 𝑓𝑗𝑅𝑆
𝑅
𝑗 )
⁄ ; (2) consumption per capita in country 𝑆 (𝑦𝑆); and (3) population in country 𝑆 (𝑝𝑆), which can be written as
𝒉 = [𝑑1⋯𝑑𝑆⋯𝑑𝑁] × [ 𝑦1 ⋯ 0 ⋯ 0 ⋮ ⋱ ⋮ ⋯ ⋮ 0 ⋯ 𝑦𝑅 ⋯ 0 ⋮ ⋯ ⋮ ⋱ ⋮ 0 ⋯ 0 ⋯ 𝑦𝑁] × [ 𝑝1 ⋯ 0 ⋯ 0 ⋮ ⋱ ⋮ ⋯ ⋮ 0 ⋯ 𝑝𝑅 ⋯ 0 ⋮ ⋯ ⋮ ⋱ ⋮ 0 ⋯ 0 ⋯ 𝑝𝑁] = 𝒅′𝒚̂ 𝒑̂ (13)
From this follows that 𝒈 = 𝒆′𝑳𝑭 + 𝒉′= 𝒆′𝑳(𝑻 ⊗ 𝑪)𝒚̂ 𝒑̂ + 𝒅′𝒚̂ 𝒑̂ (14)
A.5 Decomposition of changes in consumption based emissions
A structural decomposition analysis disaggregates a change in an indicator into contributions of changes in the constituent components of that indicator. In this research, year-to-year changes in consumption based CO2 emissions ∆𝒈 are determined by changes in the industry
emissions for final consumption 𝒆′𝑳(𝑻 ⊗ 𝑪)𝒚̂ 𝒑̂ and the direct household emissions 𝒅′𝒚̂ 𝒑̂. An
21
determinants results in as many as 𝑚! unique decomposition forms. Dietzenbacher and Los (1998) conclude that the average of two so-called polar equations is strikingly close to the average of 𝑚! decompositions of a change in an indicator with 𝑚 determinants.
In this research, year to year changes in 𝒈 can be denoted as the difference between period 1 and period 0, that is ∆𝒈 = 𝒈𝟏− 𝒈𝟎, where 𝒈 = 𝒆′𝑳(𝑻 ⊗ 𝑪)𝒚̂ 𝒑̂ + 𝒅′𝒚̂ 𝒑̂. The first polar equation (marked by the letter 𝑎) is given by
∆𝒈𝑎 = ∆𝒆′𝑳1(𝑻1⊗ 𝑪1)𝒚̂1𝒑̂1+ 𝒆′0∆𝑳(𝑻1⊗ 𝑪1)𝒚̂1𝒑̂1+ 𝒆′0𝑳0(∆𝑻 ⊗ 𝑪1)𝒚̂1𝒑̂1
+ 𝒆′0𝑳0(𝑻0⊗ ∆𝑪)𝒚̂1𝒑̂1+ 𝒆′0𝑳0(𝑻0⊗ 𝑪0)∆𝒚̂ 𝒑̂1+ 𝒆′0𝑳0(𝑻0⊗ 𝑪0)𝒚̂0∆𝒑̂
+ ∆𝒅′𝒚̂
1𝒑̂1+ 𝒅′0∆𝒚̂ 𝒑̂1 + 𝒅′0𝒚̂0∆𝒑̂ (15)
The second polar equation (marked by the letter 𝑏) is given by ∆𝒈𝑏 = ∆𝒆′𝑳
0(𝑻0⊗ 𝑪0)𝒚̂0𝒑̂0+ 𝒆′1∆𝑳(𝑻0⊗ 𝑪0)𝒚̂0𝒑̂0+ 𝒆′1𝑳1(∆𝑻 ⊗ 𝑪0)𝒚̂0𝒑̂0
+ 𝒆′1𝑳1(𝑻1⊗ ∆𝑪)𝒚̂0𝒑̂0+ 𝒆′1𝑳1(𝑻1⊗ 𝑪1)∆𝒚̂ 𝒑̂0+ 𝒆′1𝑳1(𝑻1⊗ 𝑪1)𝒚̂1∆𝒑̂
+ ∆𝒅′𝒚̂0𝒑̂
0+ 𝒅′1∆𝒚̂ 𝒑̂0 + 𝒅′1𝒚̂1∆𝒑̂ (16)
The change in 𝒈 between period 1 and period 0 is obtained by taking the average of the two polar equations
∆𝒈 = 𝒈𝟏− 𝒈𝟎 = 1
2(∆𝒈𝒂+ ∆𝒈𝒃) (17)
A.6 Decomposing changes in the Leontief Matrix
The changes in the Leontief matrix (∆𝑳) reflect changes in the input coefficients. Consequently, ∆𝑳 can be rewritten in terms of ∆𝑨 as follows:
∆𝑳 = 𝑳0(∆𝑨)𝑳1 = 𝑳1(∆𝑨)𝑳0 (18)
The second term of (15) is now 𝒆′0𝑳0(∆𝑨)𝑳1(𝑻1⊗ 𝑪1)𝒚̂1𝒑̂1, and the second term of (16) is
now 𝒆′1𝑳1(∆𝑨)𝑳0(𝑻0⊗ 𝑪0)𝒚̂0𝒑̂0.
The input coefficient 𝑎𝑖𝑗𝑅𝑆 can be split into two components 𝑎𝑖𝑗𝑅𝑆= 𝑎𝑖𝑗
𝑅𝑆
∑ 𝑎𝑅 𝑖𝑗𝑅𝑆∑ 𝑎𝑖𝑗 𝑅𝑆
𝑅 = 𝑞𝑖𝑗𝑅𝑆× 𝑏𝑖𝑗∙𝑆 (19)
where the technology coefficient 𝑏𝑖𝑗∙𝑆 = ∑ 𝑎𝑅 𝑖𝑗𝑅𝑆 indicates the output from sector 𝑖 (irrespective of its country of origin) that is required for one unit of output from sector 𝑗 in country 𝑆. The intermediate trade coefficient 𝑞𝑖𝑗𝑅𝑆 = 𝑎𝑖𝑗𝑅𝑆⁄∑ 𝑎𝑅 𝑖𝑗𝑅𝑆 indicates which share of the total output from
sector 𝑖 that is required for one unit of output from sector 𝑗 in country 𝑆 originates from country 𝑅. The (𝑁𝑛 × 𝑁𝑛) matrix of technology coefficients is given by
22
The (𝑁𝑛 × 𝑁𝑛) matrix of intermediate trade coefficients is given by
𝑸 = [ 𝑸11 ⋯ 𝑸1𝑅 ⋮ ⋱ ⋮ 𝑸𝑅1 ⋯ 𝑸𝑅𝑅 ⋯ 𝑸1𝑁 ⋯ ⋮ ⋯ 𝑸𝑅𝑁 ⋮ ⋯ ⋮ 𝑸𝑁1 ⋯ 𝑸𝑁𝑅 ⋱ ⋮ ⋯ 𝑸𝑁𝑁] (21)
Note that 𝑎𝑖𝑗𝑅𝑆= 𝑞𝑖𝑗𝑅𝑆× 𝑏𝑖𝑗∙𝑆 , from which follows that the 𝑨 matrix is given by
𝑨 = [ 𝑸11 ⋯ 𝑸1𝑅 ⋮ ⋱ ⋮ 𝑸𝑅1 ⋯ 𝑸𝑅𝑅 ⋯ 𝑸1𝑁 ⋯ ⋮ ⋯ 𝑸𝑅𝑁 ⋮ ⋯ ⋮ 𝑸𝑁1 ⋯ 𝑸𝑁𝑅 ⋱ ⋮ ⋯ 𝑸𝑁𝑁] ⊗ [ 𝑩1 ⋯ 𝑩𝑅 ⋮ ⋱ ⋮ 𝑩1 ⋯ 𝑩𝑅 ⋯ 𝑩𝑁 ⋯ ⋮ ⋯ 𝑩𝑁 ⋮ ⋯ ⋮ 𝑩1 ⋯ 𝑩𝑅 ⋱ ⋮ ⋯ 𝑩𝑁] = 𝑸 ⊗ 𝑩 (22)
Changes in the input coefficients (∆𝑨) can now be decomposed into changes in the production structure (technology coefficients) and changes in the trade structure of intermediate goods (trade coefficients), such that
∆𝑨 =1
2(∆𝑸) ⊗ (𝑩1+ 𝑩0) + 1
2(𝑸1+ 𝑸0) ⊗ (∆𝑩) (23)
In this research, the change in consumption based CO2 emissions is disintegrated into changes
in its constituent components. Following Xu and Dietzenbacher (2014), changes in these components can be further disentangled into domestic changes and changes abroad. This is done for the changes in emission coefficients, in technology (or production structure), and in the trade structure. Hence, for the country 𝑅(= 1, … , 𝑁) we define
𝒆 = 𝒆(−𝑹)+ 𝒆(𝑹) (24)
𝑩 = 𝑩(−𝑹)+ 𝑩(𝑹) (25)
𝑸 = 𝑸(−𝑹)+ 𝑸(𝑹) (26
The 𝑁𝑛-element vector 𝒆(−𝑹) constitutes of the elements of the emissions coefficients of
country 𝑅 and zero’s for all other elements, such that
𝒆(𝑅) = [ 0 ⋮ 0 𝒆𝑅 0 ⋮ 0 ] and 𝒆(−𝑅)= [ 𝒆1 ⋮ 𝒆𝑅−1 0 𝒆𝑅+1 ⋮ 𝒆𝑁 ] (27)
Similarly, the (𝑁𝑛 × 𝑁𝑛) matrix of trade coefficients 𝑩(𝑹) is defined as
23
The (𝑁𝑛 × 𝑁𝑛) matrix of intermediate trade coefficients 𝑸(𝑹) constitutes of all elements that
indicate import by country 𝑅, such that
𝑸(𝑹) = [ 0 ⋯ 𝑸1𝑅 ⋯ 0 ⋮ ⋱ ⋮ ⋯ ⋮ 0 ⋯ 𝑸𝑅𝑅 ⋯ 0 ⋮ ⋯ ⋮ ⋱ ⋮ 0 ⋯ 𝑸𝑁𝑅 ⋯ 0] and 𝑸(−𝑹)= 𝑸 − 𝑸(𝑹) (29)
Thus far, eight components of consumption based emissions have been identified: • The 𝑁𝑛-element vector of emission coefficients 𝒆 (= 𝒆(−𝑹)+ 𝒆(𝑹))
• The 𝑁𝑛 × 𝑁𝑛 matrix of intermediate trade coefficients 𝑸 (= 𝑸(−𝑹)+ 𝑸(𝑹))
• The 𝑁𝑛 × 𝑁𝑛 matrix of technology coefficients 𝑩(= 𝑩(−𝑹)+ 𝑩(𝑹))
• The 𝑁𝑛 × 𝑁 matrix of final goods trade coefficients 𝑻 (= 𝑻(−𝑹)+ 𝑻(𝑹))
• The 𝑁𝑛 × 𝑁 matrix of consumption mix 𝑪
• The 𝑁-element vector of consumption per capita 𝒚 • The 𝑁-element vector of population 𝒑
• The 𝑁-element vector of direct household emissions per unit of final consumption 𝒅
A.7 Final decomposition form
Substituting ∆𝑳 for its decomposition as defined above gives the final decomposition for changes in consumption based CO2 emissions, where ∆𝒈 = 𝒈𝟏− 𝒈𝟎=
24 +1 2𝒆 ′ 0𝑳0(∆𝑻 ⊗ 𝑪1)𝒚̂1𝒑̂1+ 1 2𝒆 ′ 1𝑳1(∆𝑻 ⊗ 𝑪0)𝒚̂0𝒑̂0 (30.7) +1 2𝒆 ′ 0𝑳0(𝑻0⊗ ∆𝑪)𝒚̂1𝒑̂1+ 1 2𝒆 ′ 1𝑳1(𝑻1⊗ ∆𝑪)𝒚̂0𝒑̂0 (30.8) +1 2𝒆 ′ 0𝑳0(𝑻0⊗ 𝑪0)∆𝒚̂ 𝒑̂1+ 1 2𝒆 ′ 1𝑳1(𝑻1⊗ 𝑪1)∆𝒚̂ 𝒑̂0 (30.9) +1 2𝒆 ′ 0𝑳0(𝑻0⊗ 𝑪0)𝒚̂0∆𝒑̂ + 1 2𝒆 ′ 1𝑳1(𝑻1⊗ 𝑪1)𝒚̂1∆𝒑̂ (30.10) +1 2∆𝒅′𝒚̂1 𝒑̂1+ 1 2∆𝒅′𝒚̂0 𝒑̂0 (30.11) +1 2𝒅′0∆𝒚̂ 𝒑̂1+ 1 2𝒅′1∆𝒚̂ 𝒑̂0 (30.12) +1 2𝒅′0𝒚̂0∆𝒑̂ + 1 2𝒅′1𝒚̂1∆𝒑̂ (30.13)
Each of the components of the decomposition analysis as presented above attributes to one of the five drivers of changes in consumption based CO2 emissions. Firstly, (30.1) and (30.5)
represent the effect that changes in technology abroad has on emissions, whereas (30.2), (30.6) and (30.12) represent the effect that domestic changes in technology has on emissions. Secondly, the effect of changes in the trade structure abroad on emissions is represented by (30.3). The effect of domestic changes in the trade structure is captured by (30.4) and (30.7). Thirdly, the effect of changes in the consumption mix is represented by the outcome of (30.8). Fourthly, the contribution of changes in consumption per capita to emissions is captured by (30.9) and (30.11). Finally, the effect of population growth on emissions is captured by (30.10) and (30.13).
A.8 Measuring changes in terms of volumes
The WIOD offers GMRIO tables in current prices (CUR) as well as in previous year’s prices (PYP). Following Arto and Dietzenbacher (2014), the annual results should be chained. The volume change of variable 𝛼 in year 𝑡 is obtained by measuring the difference between 𝛼𝑡−1 in
current prices and 𝛼𝑡 in previous year’s prices, such that
∆𝛼𝑡1−𝑡0 = 𝛼𝑡1𝑃𝑌𝑃− 𝛼𝑡0𝐶𝑈𝑅 (31)
Since variable 𝛼 is measured in the same price in both time 𝑡1 and time 𝑡0, the change in variable 𝛼 is measured only in terms of volume. Consequently, the change between time 𝑡2 and time 𝑡1, is measured as
∆𝛼𝑡2−𝑡1 = 𝛼𝑡2𝑃𝑌𝑃− 𝛼𝑡1𝐶𝑈𝑅 (32)
25
∆𝛼𝑡2−𝑡0 = ∆𝛼𝑡2−𝑡1+ ∆𝛼𝑡1−𝑡0 = (𝛼𝑡2𝑃𝑌𝑃 − 𝛼𝑡1𝐶𝑈𝑅) + (𝛼𝑡1𝑃𝑌𝑃− 𝛼𝑡0𝐶𝑈𝑅) (33) The changes in the determinants are all calculated in this manner, except for changes in population size ∆𝑝.
A.9 Countries included in database
EU Region Austria, Belgium, Bulgaria, Cyprus, Czech Republic, Germany,
Denmark, Spain, Estonia, Finland, France, Great Britain, Greece, Hungary, Croatia, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, the Netherlands, Norway, Poland, Portugal, Romania, Slovakia, Slovenia, Sweden, Switzerland.
EAS Japan, South Korea, Taiwan
Australia Brazil Canada China India Indonesia Mexico Russia Turkey
26 Appendix B
B.1 Structural decomposition analyses of year-to-year changes in consumption-based emissions
Figure 8: Structural decomposition analysis of changes in consumption based CO2 emissions in Australia
Figure 9: Structural decomposition analysis of changes in consumption based CO2 emissions in Brazil
27
Figure 10: Structural decomposition analysis of changes in consumption based CO2 emissions in EAS
Figure 11: Structural decomposition analysis of changes in consumption based CO2 emissions in Mexico
28
Figure 12: Structural decomposition analysis of changes in consumption based CO2 emissions in Turkey
29
B.2 Structural decomposition analyses of year-to-year changes in consumption-based emissions
Figure 13: Structural decomposition analysis of changes in consumption based CO2 emissions in India
Figure 14: Structural decomposition analysis of changes in consumption based CO2 emissions in Indonesia
30
Figure 15: Structural decomposition analysis of changes in consumption based CO2 emissions in Russia
Figure 16: Structural decomposition analysis of changes in consumption based CO2 emissions in the rest of the world (RoW)
Trade (d) Trade (f) Population Consumption mix FD per capita Technology (d) Technology (f) Total (234 Mt) -600 -400 -200 0 200 400 600 800 1000 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 C h an ge in CO2 (i n Mt)
Russia
Trade (d) Trade (f) Population Consumption mix FD per capita Technology (d) Technology (f) Total (3372 Mt) -1500 -1000 -500 0 500 1000 1500 2000 2500 3000 3500 4000 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Cha n ge in CO2 (in Mt)31
B.3 The weighted average of emission coefficients
Figure 17 Weighted average of industry-emission coefficients in the USA, Europe and China in 2000 and 2014
0,272 0,220 1,038 0,140 0,081 0,313 0,000 0,200 0,400 0,600 0,800 1,000 1,200 USA EUR CHN CO2 e m is sion p er u n it o f o u tp u t (m easured in m ill ion s o f US$)
