EMISSIONS EMBODIED IN TRADE:
ACCOUNTING FOR FUNCTIONAL SPECIALIZATION
Linda Arfelt S3218074
l.s.arfelt@student.rug.nl
University of Groningen
Faculty of Economics & Business
2
EMISSIONS EMBODIED IN TRADE:
ACCOUNTING FOR FUNCTIONAL SPECIALIZATION
Abstract
This study is first to bridge the gap between research on functional specialization and emissions by evaluating the extent to which functional specialization determines the emissions embodied in trade by BRIC countries for 1999 and 2007. A positive and robust relationship between the emissions-intensity of production and fabrication is found, indicating that functions have a non-negligible effect on emissions embodied in trade. The proposed model provides diverging results from one accounting for technological and product differences, ultimately resulting in two distinct conclusions on the relationship BRIC countries have with the rest of the world with regard to trade in emissions.
Key Words
3
TABLE OF CONTENTS
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
1. INTRODUCTION ... 4
2. LITERATURE REVIEW... 7
2.1 Pollution Haven Hypothesis and Emissions Embodied in Trade ... 7
2.2 Global Value Chains and Functional Specialization ... 10
2.3 Relevance to Hypothesis ... 13
3. DATA & METHODOLOGY ... 14
3.1 Input-Output Analysis ... 15
3.2 Evaluating the Pollution Haven Hypothesis ... 19
3.3 Estimating the Emissions Caused by Functional Specialization ... 20
4
1. INTRODUCTION
The global race to prevent irreversible environmental damage has created a dichotomy between developing and advanced countries. The rapid pace of technological development from the early 1990s facilitated an uncoupling between the economy and ecology of individual countries, as the location of production processes could now be spatially divided from the consumption location of the final products (Duranton & Puga, 2005). This divide is expected to favor advanced countries, such that the emissions embodied in exports is smaller than that in imports, indicating that emission-intensive production has been offshored. The Pollution Haven Hypothesis (PHH) proposes that such an advanced country balance of emissions embodied in trade is due to their relatively strict environmental regulations, whereas developing regions have much more lenient regulations and are competing for the added income from the relocated industries (Cherniwchan, Copeland & Taylor, 2017). Yet, this balance can also stem from differences in human capital and factor inputs between countries and industries that determine specialization in a less-emitting production process (Grossman & Rossi-Hansberg, 2006). Climate research becomes exceedingly politically charged when considering the magnitude of these emissions embodied in trade that derives from advanced countries offshoring ‘dirty’ production processes.
This paper will adopt an alternative characterization of trade specialization to quantify the emissions embodied in trade for the BRIC countries. Since the onset of the 20th century, changes in international
production networks have driven countries to specialize in smaller portions of productions than the entire good or service. Rather than measuring value added in production of products, a growing share of academics prefer determining value added by groups of tasks, so-called functions, to capture the effects of this change in the nature of trade. Since countries perform different functions for the production of the same product, using functions as the basis for specialization provides a more realistic idea of comparative advantage. Functional specialization in trade has revolutionized how trade patterns are viewed, yet the traditional sectoral scope is still used for the study of emissions embodied in trade. With the context of this change in nature of global trade, this study gears to answer the question: to what extent does functional specialization determine the emissions embodied in trade by BRIC countries for 1999 and 2007? As some functions may be inherently more emission-intensive than others, a correlation between emissions and specialization patterns indicates that some industries and countries are ‘dirty’ for others to be ‘clean’.
5
the production of the final product. Using sectoral specialization as basis for emissions research therefore no longer suits the current nature of international trade, as this smaller dimension of specialization is overlooked by attributing emissions in trade to products exported.
To appropriately compare value-added across countries, Timmer, Miroudot and de Vries (2019) use functional specialization, namely the degree of specialization in ‘functions’ within the production process. Functions are classified into four categories (R&D, management, marketing, and fabrication) based on occupations with similar tasks. This division of labor into functions enables the study of occupational disparities across countries and industries. In practice, let’s take the classic example of quantifying the fragmentation of production via the iPod (Timmer, Erumban, Los, Stehrer & de Vries, 2014). Although both China and the US partake in the production of the iPod, therein exporting an electronic product, the importance of taking a functional approach to trade is illustrated by the different form of value added by each country. China specializes in fabrication, whereas the US in marketing, and therefore the quantity and types of factor inputs required for production in each country are different. By accounting for the functions used in production, differences in countries’ trade specializations can be evaluated without relying on products. Given the distinctive factor inputs required by functions, differences in emission coefficients and intermediate inputs can be related to the types of functions used extensively by each industry. The channels that determine non-zero emissions balances thus change to the relative shares of the four functions used in production, referred to as the functional mix, as well as differences in production technology, quantified as emissions coefficients and intermediate inputs.
6
proposed by the Heckscher-Ohlin model. No studies have explored the relationship between emissions and functions to date.
This study uses industry-specific fabrication shares for the BRIC countries in 1999 and 2007 as means to validate the Pollution Haven Hypothesis. The BRIC countries, namely Brazil, Russia, India, and China, were classified as an economic bloc in 2001 based on their size and growth potential, attributed primarily by agriculture and heavy industry (Groot, de Groot, Lejour & Möhlmann, 2011). In 2009, the BRIC countries amounted to approximately 35% of global CO2 emissions, and over 40% of global
population. Yet, in a study of emissions embodied in trade for India in 1991 and 1996, Dietzenbacher et al. (2007) find evidence of India gaining from trade and thus moving away from this expected classification as a pollution haven. Extending the methodology of Dietzenbacher et al. (2007) to account for functional specialization and technological differences between countries, this study expects to find evidence in support of the Pollution Haven Hypothesis for the BRIC countries. Assessing the validity of the Pollution Haven Hypothesis for these countries provides clarity on the level of influence international trade has on countries specialization in emission-intensive functions. The functional specialization database created by Timmer et al. (2019) classifies labor income by occupation to identify tasks carried out in exporting industries. This publication is the primary source for identifying functions by country and sector. The World Input-Output Database (WIOD) provides data on production linkages and emissions (Timmer, Dietzenbacher, Los, Stehrer & de Vries, 2015). Given clear evidence of the contribution of CO2 emissions to global warming (Zhou, Zhang & Li,
2013), this study only accounts for these emissions to air . This study measures emissions generated by fabrication from country-level CO2 emissions required by production, due to the concentration of
fabrication activities in developing countries, and the expectation that these are most emission-intensive of the four functions. The analysis of the Pollution Haven Hypothesis thereafter is then in context of this new division of trade specialization.
A positive and robust relationship between the emissions-intensity of production and fabrication is found, indicating that functions have a non-negligible effect on emissions embodied in trade. The method attributing emissions to fabrication provides diverging results from a model solely accounting for technological and product differences. Ultimately, these result in two distinct conclusions on the relationship BRIC countries have with the rest of the world with regard to trade in emissions. One diverging conclusion is the classification of Brazil as a pollution haven in both 1999 and 2007 when accounting for fabrication, as opposed to support of the hypothesis only in 1999 for the model traditionally adopted in literature. These different conclusions are central to enhancing the knowledge and methods surrounding the effects trade has on the environment, particularly in the case of developing countries. Further reiterations of the study could control for relative price level changes in order to
7
2. LITERATURE REVIEW
Differences in national environmental policies have long been a point of political disaccord, and one that has since extended to economic trade theory. Copeland and Taylor (1994) first referenced the PHH in examining the linkages between emissions, and national income for North-South trade. This sparked the recognition of trade liberalization as an environmental issue, where global trade between advanced and developing countries would inherently increase the level of emissions worldwide. Yet, the empirical validity of the relationship between international trade and the environment remains among the most challenged topics to date (Kellenberg, 2009). The number of different approaches utilized in existing research on the PHH calls for a careful study of the underlying mechanics of the hypothesis in order to accurately account for the applicability of the forthcoming addition of functional specialization.
2.1 Pollution Haven Hypothesis and Emissions Embodied in Trade
A substantial point of contention in climate research regards the validity of the PHH, which stipulates that trade liberalization drives emission-intensive industries from advanced countries with stringent environmental policies to developing regions with lax environmental regulation or enforcement (Cherniwchan et al., 2017). The PHH belongs to a larger field of climate research focused on emissions embodied in trade that quantifies the emissions created in the production of traded goods and services. This research effectively measures whether countries gain or lose from trade in terms of emissions, by evaluating the volume of emissions required to produce a country’s exports relative to those embodied in its imports. The PHH thus explicitly proposes that developing country specialization in emission-intensive exports is necessarily facilitated by advanced countries consuming more than they produce. This developmental divide in the international balance of emissions embodied in trade is critical for allocating international responsibility in environmental protection (Pan, Phillips & Chen, 2008). If the PHH holds, the lack of environmental regulation that attracts industries to developing countries could incite a ‘race to the bottom’ mentality in environmental standards (Millimet & Roy, 2016). Moreover, trade flows prompted by differences in environmental regulation influence the ecological quality of developing countries as well as the number of displaced jobs in advanced countries, both of which may have notable political and social ramifications (Taylor, 2004).
8
traditionally improves economic efficiency by means of factor income redistribution, this assumption may not apply when emissions are considered as a factor input. Specialization in emission-intensive products have unaccounted losses beyond the model, indicating that developing countries are unlikely to obtain the gains from trade proposed by Heckscher-Ohlin in alternative factor specialization. The evaluation of the PHH thus relates to the differences in national comparative advantages determined by the relative specialization in one of the three endowments: capital, labor, or emissions.
9
The difficulty in adequately measuring the PHH causes an assortment of results and approaches. Although developing countries may have higher emission-intensities in production, these are not necessarily caused by the level of environmental regulation, but rather by the level of economic development, or governmental intervention in sectors relevant to economic and energy security. Moreover, ‘dirty’ capital intensive industries are less likely to be influenced by environmental regulation than industries with comparatively low capital intensity, and thus are less likely to be offshored (Kellenberg, 2009). Some industries are less ‘offshorable’ due to large fixed costs, agglomeration economies, or transportation costs, making their organizational decisions less receptive to environmental policies (Ederington et al., 2005). For example, regional concentration of similar industries can streamline supply networks and generate productivity gains, creating specialized hubs such as Silicon Valley in the US for technological innovation or Frankfurt in Germany for financial services. Firm size and productivity, often used as flags for export-orientation, may also serve as factors influencing offshoring of dirty production processes (Cherniwchan et al., 2016). Accounting for the differences in emissions embodied in trade between countries based on differences in factor endowments and technological differences mitigates the endogeneity issues related to evaluating the PHH through a strict policy approach (Zeng & Zhao, 2009).
10
thus have diverging emissions-contents. If countries with similar specialization patterns in functions have a systematically positive emissions embodied in trade balance, the PHH relates to specialization in labor rather than capital or emission permits.
2.2 Global Value Chains and Functional Specialization
Understanding global value chains is crucial for grasping the origin of the PHH due to their relevance in driving regional development and the resulting interplay of regional comparative advantages. The share of foreign value added in the final value of a product has over the course of the last decade increased significantly (Timmer et al., 2014; Los, Timmer & de Vries, 2014). This increasing foreign value-added share demonstrates the growing trend of international fragmentation of production, implying a shift in trade specialization from products to a smaller dimension. Yet, traditional measurement of global value chains still persists. The PHH thus continues to be explained by the specialization in emission-intensive industries, stemming from the share of capital and labor used in production. By accounting for the different functions performed in different countries, however, the evaluation of the PHH relates to the specialization in different types of labor, measured by the functional mix used in production. Before accounting for the potential change caused by the functional mix as a determinant in the PHH, the emergence of global value chains and the need for using this different specialization pattern need to be addressed.
The fall in transportation and telecommunication costs caused by the rapid pace of technological change in the early 1990s enabled the fragmentation of business units (Duranton et al., 2005; Grossman et al., 2008). The internet revolution mitigated the transaction and transmission costs of relocating business units by improved information flows and ability for cross-border monitoring through developed data analysis and other managerial tools (Strange & Zucchella, 2017). Falling transmission costs meant that all activities no longer needed to be spatially integrated at the headquarters, because the means to effectively communicate within a multi-unit organizational structure were available (Duranton et al., 2005). The division into multiple units or business functions depends on the ability to offshore tasks as well as the type of comparative advantage held by the headquarters. Although technological improvements in transportation and telecommunication facilitates the offshoring of manufacturing and production tasks, these innovations cause a clustering of headquarter and business services (Duranton et al., 2005). Headquarter and business services tend to cluster in large cities, whereas smaller cities specialize in manufacturing and production. Innovations in transportation and telecommunication thus do not have a uniform effect on the probability of offshoring business units, but rather other factors may drive the geographical concentrations of occupations.
11
less developed regions. Similarly, the Heckscher-Ohlin model predicts that high-skilled labor and capital will concentrate in advanced countries while less-skilled activities will be offshored to developing economies. Yet, although evidence shows an increasing focus of advanced markets on activities carried out by high-skilled workers, emerging markets shift to capital-intensive activities with a decreasing share of low-skilled labor demands (Timmer et al., 2014). Fabrication activities require lower skill levels than the other functions identified, and as such, the abundance of less-skilled labor in developing countries will incite a sizable offshoring of fabrication activities to these locations. In contrast, advanced country abundance in higher-skilled labor will develop hubs of R&D functions due to the spillover effects from proximity to other units with similar processes. Autor et al. (2015) also propose factor intensity plays a role in determining location choice, whereas both Los et al. (2014) and Jiang & de Vries (2018) suggest that other factors are at play beyond factor intensity due to notable variation in the likelihood to offshore certain business functions at both an industry and sectoral level. This study views factor intensity of energy as endogenously determined by the nature of the function.
12
Figure 1: Illustration of a Simplified and Exemplary Global Value Chain from the Product Specialization
(adopted from Los et al., 2015) and Functional Specialization Perspectives
Note! The functions illustrated as part of the production process are likely to vary by firm, industry, and country.
R&D is not exclusively a pre-production process, and marketing is not exclusively a post-production process. The illustration in Figure 1 is thus solely used as an example to provide clarification to the distinction between perspectives.
Timmer, Miroudot and de Vries (2019) define functions as a series of similar occupations classified into one of four groups (R&D, management, marketing or fabrication), ranging from production and assembly tasks categorized as fabrication to research and technological development defined as R&D. The relative share of functions used in production is then defined as the functional mix. Functions differ in their relative factor input requirements, and are thus largely reliant on the proximity of other similar production networks. Fabrication is classified by comparatively low-skilled labor and high share of value added due to the intensive use of intermediate inputs, especially relating to energy. As such, fabrication is likely to be more polluting than other functions, which are expected to have systematically lower value added and intermediate energy inputs. These different factor requirements for functions are thus prone to regional clusters based on the availability of labor. Regional development thus plays a determining role in not only the cost advantage of offshoring but also in the occupational polarization due to different levels of skilled labor required. The pervasiveness of the functional mix, as discussed by Timmer et al. (2019), indicates that tasks are likely to remain
Marketing Country 1 Capital and Labor Intermediate Good R&D Country 2 Capital and Labor Domestic Intermediate Goods Intermediate Good Fabrication Management Country 3 Capital and Labor Domestic Intermediate
Goods Final Good
for domestic and foreign
demand
Management
13
unchanged for extensive periods of time as supply networks and comparative advantages take time to change.
When accounting for differences in the functional mix between countries, the PHH regards the emission-intensity of the different factor inputs used by each function. Fabrication, relative to other functions, requires higher volumes of intermediate energy inputs due to the nature of the tasks used in production, and thus is likely to also be more emission-intensive. Given the pervasiveness of the functional mix and this higher emission-intensity, a region that specializes in fabrication is expected to have an abundance of emission permits due to lax environmental standards or poor monitoring. As high fabrication shares are generally found in developing countries (Timmer et al., 2019), their functional specialization is likely to cause their persistent classification as pollution havens.
2.3 Relevance to Hypothesis
Our understanding of the PHH rests entirely on literature using sectoral specialization as the basis for international trade. This product perspective on trade assumes that differences between countries’ emission trade balances derive from two sources, namely differences in the product mix and differences in technologies used. Yet, literature on the nature of international trade concede that the emergence of global value chains commands for a more careful study of value added than traditionally seen in this sectoral scope. Using functional specialization for study of emissions assumes a relationship between the production and emissions beyond industry classification. It is no longer correct to evaluate emissions embodied in trade via the product mix or technological channels, as recognizing the smaller dimension of specialization excludes the product mix of exports and imports as an independent channel of emissions determination in trade. Analyzing the differences in production technologies and the functional mix of exported products a provides more accurate and up-to-date evaluation of emissions embodied in trade. This relationship between functions and emissions has not been analyzed in existing literature.
The shift of capital-intensive and less-skilled tasks to developing countries, and significant variation in industry emissions are elements which point to a potential relationship between functions and emissions. Assuming that fabrication has higher emission coefficients and requires more intermediate energy inputs, imports and exports which have higher fabrication shares are likely to exhibit higher emission embodied in trade than other functions. The extent to which functional specialization causes developing countries to lose from trade in terms of emissions, or classify as pollution havens, therefore depends on the extent to which the production of exported products involves fabrication. I expect this share of fabrication used in production to remain roughly constant due to the pervasiveness of functional specialization. The balance of emissions embodied in trade between advanced and developing countries are thus also expected to remain categorically the same.
14
I expect that accounting for fabrication as a channel of emissions embodied in trade will increase the emissions embodied in trade balance of developing countries and decrease the balance of advanced countries due to the differing shares of fabrication, relative to a study solely relating to emissions to the product mix.
3. DATA & METHODOLOGY
The aim of this study is to evaluate the PHH by accounting for functional specialization in trade. Specifically, the extent to which functional specialization determines the emissions embodied in trade by BRIC countries for 1999 and 2007 is measured. Emissions embodied in trade have traditionally been evaluated based on differences in production technologies and the product mix of exports and imports. However, this approach neglects to account for the shift in specialization from products to functions characteristic of current trade. In order to account for this shift, I revise the methodology for assessing the emissions embodied in trade by attributing emissions required for the production of exports to the fabrication share of labor income used by each industry. By quantifying the emission-intensity of exports generated by fabrication, this study expects to find evidence of greater emissions embodied in exports by the BRIC countries relative to exports by the rest of the world. Furthermore, I expect these losses from trade in terms of emissions to be greater in 2007 than 1999, indicating a growing share of emission-intensive fabrication in BRIC country exports.
This paper begins by measuring the differences in production technologies through emission coefficients and intermediate inputs by industry and country. Quantifying the volume of emissions directly and indirectly required for each unit of output, or the emission coefficients of production, requires use of input-output analysis. When combined with exports and a production multiplier, these emission coefficients indicate the volume of emissions required for an extra unit of exports or the emission-intensity of each industry. Data on emissions is linked to the inputs used in production, as for example, the production of electricity does not generate a large volume of emissions but its use in production of other products does. As the PHH relates to the offshoring of ‘dirty’ industries from advanced to developing countries, data on emission-intensity by industry and country distinguishes ‘dirty’ from ‘clean’ industries. However, this classification is not enough to validate the PHH as the limited perspective of domestic industry inputs provides no indication of the influence of trade on emissions nor the significance of functions. Advancing to a two-region level requires a framework for measuring trade and the emissions embodied in its production.
15
The incorporation of functional specialization by industry and country into the current approach utilizes data on functions as shares of employment and the technological differences between countries. I begin by estimating the relation between emissions and fabrication to evaluate the assumption that fabrication functions are the most emission-intensive. The resulting regression coefficients in conjunction with fabrication are used to quantify the emissions caused by fabrication required for exports. Accounting for functional specialization in the emission requirements for BRIC countries and the rest of the world facilitates the evaluation of the PHH with a more inclusive measure of production technologies.
3.1 Input-Output Analysis
This study uses input-output analysis to calculate emission-intensity by industry for each of the BRIC countries in order to identify ‘dirty’ and ‘clean’ industries for the basis of the PHH. This analysis, first introduced by Leontief (1986), quantifies the direct and indirect linkages between different sectors of an economy. Input-output analysis is frequently extended to the study of economic-environmental relationships, as the structure of the analysis is suitable to, for example, uncover economic drivers of emissions. Before extending this analysis to the scope of this study, however, the structure of the input-output table and its components is discussed for an understanding of the tools used for the forthcoming methodology.
Input-output tables illustrate the supply and use relationships between producers and consumers by quantifying the flows of final and intermediate products between industries within a year. For example, these tables capture the monetary value of intermediate products from agriculture (such as grain) required by the food and beverage industry to produce its final product (such as bread). Data on the value of imported and exported intermediate or final products required in production are also included in tables for a complete account of the trade flows within an economy. The network of products moving from suppliers, in this example the agriculture industry, to users, the food and beverage industry, provides an understanding of the inter-industry production structure. National Input-Output Tables (NIOTs) quantify these interdependencies within the industries of a singular country. The structure of World Input-Output Tables (WIOT) will be discussed at a later stage. For simplicity, matrixes are henceforth conveyed by bold and italic capitals, vectors are denoted in bold and italic lowercase letters, and italic letters denote scalars.
Input-output tables divide production data into intermediate use and production for final use, which add up to total output as illustrated in Figure 2. Intermediate use represents the flows of intermediate products between industries, denoted as matrix Z. Each element of matrix Z indicates the value of intermediate inputs demanded from industry i by industry j, noted as 𝑧𝑖𝑗. Industries indicated by the
16
Industry
(user) Final Use Total Use
Industry (supplier) Intermediate Use
Z
Domestic Final Demand F Exports ex Total Output x Imports im Value Added vTotal Use Total Input x’
Figure 2: An Illustration of a Simplified National Input-Output Table
In contrast to the intermediate use of matrix Z, final use matrix F includes the demand for final products from each industry. This demand is not for products used as inputs for the production of other products, but rather by households, governments, investment and as exports. The final demand vector f is produced by summing over columns of final demand matrix F to create a vector with domestic final demand. Finally, the last category of production data is total use, denoted as vector x, which is the sum of all products used (input) or supplied (output) by each industry i. By definition, total output is equal to total input, such that the first element in the total output row is equal to the first element of the total input column. As such, total input is denoted as x’, to indicate it is identical to vector x but transposed. In short, input-output tables formulate total output of industry i (𝒙𝒊) as
equal to monetary flows from all other industries and its total final demand (𝒇𝒊)
𝑥𝑖 = 𝑧𝑖1+ 𝑧𝑖2+ ⋯ + 𝑧𝑖𝑗 + 𝑓𝑖 (1)
With an economy of n industries, the input-output table represents Equation (1) for all n industries, each equation disaggregating output into intermediate flows between industries and final demand. Calculating fixed ratios of this distribution specifies the amount of input required from each industry to produce total output of industry i. By dividing each cell in matrix Z by its corresponding total output in vector x, the new matrix A denotes the input of industry i required to produce one dollar of output in industry j: 𝑎𝑖𝑗 = 𝑧𝑖𝑗/𝑥𝑗. The input coefficient matrix transforms Equation (1) into
𝑥𝑖 = 𝑎𝑖1𝑥𝑖+ 𝑎𝑖2𝑥𝑖+ ⋯ + 𝑎𝑖𝑗𝑥𝑖+ 𝑓𝑖 (2)
As seen in the notation of Equation (2), total output x by the intermediate output Ax and final use, which is the sum of vector f and vector ex. The basis of input-output models is thus denoted as 𝒙 = 𝑨𝒙 + (𝒇 + 𝒆𝒙) . Vector ex is a vector of exports by industry. In order to solve for total output x, this equation can be rewritten as
𝒙 = (𝑰 − 𝑨)−1× (𝒇 + 𝒆𝒙) (3)
where (𝑰 − 𝑨)−1 is the Leontief Inverse (L), and I denotes an identity matrix with the same
17
al., 2007). Moreover, accounting for domestic final demand (f) as part of final use is irrelevant to this study as the aim of this paper is to quantify emissions embodied in trade. Therefore, the discussion hereafter uses exports instead of final use, as to focus exclusively on the inputs required for production of exports. Emissions are defined as
𝒆𝒊 = 𝑒1
⋮ 𝑒𝑛
(4)
where each value in vector e quantifies kilotons (kt) of CO2 emissions by industry. Vector e requires
a conversion into fixed ratios of emissions to output to be comparable to the input-output variables produced. Each element of vector e is divided by its respective total output, calculated as 𝒄𝒊= 𝒆𝒊/𝒙𝒊 to create vector c 𝒄𝒊= 𝑐1 ⋮ 𝑐𝑛 (5)
Vector c is composed of emission coefficients by industry, where each element denotes the required emissions per dollar of output for that industry. This vector is sufficient for comparison of emissions content of production by industry and country but does not account for the element of trade that is central to this study. To measure the emissions required per unit of exports, the emissions coefficient vector c is supplemented into Equation (3) to derive a model (Wang et al., 2019) for finding the emissions required to satisfy exports
𝒑 = 𝒄′𝑳 × 𝒆𝒙 (6)
18
from the final use matrix F, which is now classified by country as final demand of households, investment or governments. Total use vector x is the total output of each country-industry pair, which is equal to the total input vector x’. Thus, for a world composed of m countries and n industries, the RoW economy consists of (𝑚 − 1) × 𝑛 industries. For simplicity, assume Brazil is Country 1 in
Figure 3, the RoW economy is composed of Countries 2 to M. This method follows for each of the
BRIC countries. The Leontief Inverse is based on the intermediate use matrix that details all counties (excluding the BRIC country evaluated), and therefore is a matrix of (𝑚 − 1) × 𝑛 by (𝑚 − 1) × 𝑛 industries. The organization of country-industry emissions is in a similar fashion, such that the emissions vector is (𝑚 − 1) × 𝑛 in length.
Assuming that the world consists of two regions, a BRIC country and the RoW, imports for one region equal the exports for the other, and vice versa. Brazil, denoted by B, exemplifies the analysis carried out for each of the BRIC countries. Relating back to Equation (6), 𝒄𝑩′𝑳𝑩(𝒆𝒙𝑩) quantifies the emissions required for production of Brazil’s exports and 𝒄𝑹𝒐𝑾′𝑳𝑹𝒐𝑾(𝒆𝒙𝑹𝒐𝑾) for exports by the RoW economy. This method allows for a distinction of different production technologies used in Brazil and the RoW, namely by the separate emission coefficients (c) and intermediate inputs (L) used for each. The validity of the PHH is signaled by larger emission embodied in exports by BRIC countries than that embodied in exports by RoW.
Figure 3: An Illustration of a Simplified World Input-Output Table
The formulation of RoW presents some limitations. Primarily, reducing global trade to a two-region setting disregards trade at a bilateral level. An aggregated variable of trade flows to the BRIC country hides patterns that could be important for a more detailed study of the PHH, such as the specific countries and industries that contribute to or gain from the BRIC country specializing in emission-intensive functions. BRIC countries are a significant economic force, and although they may be characterized as pollution havens relative to more advanced countries, it is also possible that the emission-intensive functions may have been further offshored to relatively less developed countries.
User Final Use
Total Use Country 1 … Country M Country 1 … Country M Industry 1 … Industry N … Industry 1 … Industry N Supplier Country 1 Industry 1 Intermediate Use Z Final Use F Total Output x … Industry N … Country M Industry 1 … Industry N Value Added va
19
However, despite its shortcomings, the RoW variable allows for an attempt to be made at quantifying the global balance of emissions supposed by the methodology, following the logic of the PHH. The formulation of emission coefficients and intermediate inputs quantifies the differences in production technology between the BRIC countries and the RoW. These emission coefficients (c) combined with the Leontief Matrix (L) account for the direct and indirect emissions required for exports in Equation (6). For now, the analysis remains grounded in the emissions-content of the product mix, due to the use of imports and exports without accounting for the country-industry specific functions used for production. To introduce the systematic framework for studying the PHH, the product mix remains an independent channel of emissions embodied in trade for the time being.
3.2 Evaluating the Pollution Haven Hypothesis
The PHH suggests that advanced countries gain and developing countries lose from trade in terms of emissions. In simple terms, comparing the emissions content of exports and imports between a BRIC country and the RoW allows for the hypothesis to be tested. The congruency between exports and imports allows for the extension of Equation (6) from a single-country export perspective to a two-region trade setting which considers both imports and exports. This methodology aligns with that used by Dietzenbacher et al. (2007) in their evaluation of the PHH for India for an earlier time period. Their model and assumptions are discussed below.
Measuring the industry shares of 1 billion total exports allows for comparison of emissions embodied in trade for Brazil and the RoW. Calculating sectoral shares relative to the real total export volume of each year, and multiplying these shares by 1 billion provides comparable export shares by industry and country. Although 1 billion is an arbitrary figure and a small share of the total exports of the BRIC countries, it provides an easy and consistent scale for comparison of industry export shares across time and space. This method also controls for the scale effect as a channel of emissions determinance.
The total effect on emissions relies on the balance between emissions from exports and imports. Imports have a negative effect on the balance of emissions embodied in trade for Brazil, 𝜋𝐵 = 𝒄𝑩′𝑳𝑩(𝒆𝒙𝑩− 𝒊𝒎𝑩), as they mitigate the emissions domestically required for production. In contrast, exports have a positive influence on the emissions balance. If the PHH holds, the balance of emissions embodied in trade is positive for Brazil, 𝜋𝐵 > 0, as a consequence of the larger volume of emissions embodied in exports relative to imports. Moreover, emissions embodied in trade for the RoW, 𝜋𝑅𝑂𝑊 = 𝒄𝑹𝒐𝑾′𝑳𝑹𝒐𝑾(𝒆𝒙𝑹𝒐𝑾− 𝒊𝒎𝑹𝒐𝑾) are negative for the hypothesis to hold, 𝜋𝑅𝑂𝑊 < 0, as result
of the two-region setting. As such, the authors present four possible outcomes for emissions embodied in trade:
1: 𝜋𝑅𝑂𝑊 < 0 & 𝜋𝐵< 0
20
2: 𝜋𝑅𝑂𝑊 > 0 & 𝜋𝐵 > 0
Each region specializes in exports that would be less emission-intensive in the other region in terms of production technology. The emissions content of exports is higher than that of imports, indicating that each region would be better off producing the product supposed in the potential outcome first presented. Both regions lose from trade in terms of emissions. Dietzenbacher et al. (2007) classify this scenario as highly unlikely.
3: 𝜋𝑅𝑂𝑊 > 0 & 𝜋𝐵< 0
A role reversal of the PHH where Brazil gains from trade in terms of emissions, and the RoW loses. The emissions content of exports for the RoW is higher than of its imports, meaning that RoW specializes in ‘dirty’ production technology.
4: 𝜋𝑅𝑂𝑊 < 0 & 𝜋𝐵> 0
A scenario which validates the PHH is where RoW gains from trade in terms of emissions, and Brazil loses. The emissions content of exports for Brazil is higher than of its imports, meaning Brazil specializes in ‘dirty’ production technology. This outcome is what I assume to find evidence of in the case of the BRIC countries.
As mentioned, the basis of validating the PHH rests on a two-region perspective of global trade. In order to reach a sound conclusion, both conditions of each possible outcome need to be fulfilled. However, the lack of available data at the time of publication for the formulation of a RoW economy limits Dietzenbacher et al. (2007) to a one-sided study of the PHH. Given the current availability of input-output data, the estimation of the corresponding import and export emissions-intensity for a RoW variable is possible. The methodology Dietzenbacher et al. (2007) use to assess the PHH assumes that the differences in emissions embodied in trade are caused by differences in the product mix of exports and imports. The study overlooks differences between production technologies due to lack of available data, and assumes that countries have identical emissions requirements. As such, the inherent differences in production processes between industries were not accounted. The distinction between production technologies for the BRIC countries and RoW corrects for this inaccuracy in the presented method.
Recognizing the shift from products to functions in trade specialization allows for the product mix channel to be relaxed. Technological differences between countries can be accounted for with WIOT, but this will still not account for the distinctive difference in output of identical industries across countries. As literature has shown, functional specialization patterns are relatively heterogeneous, and play a distinctive role in the type of labor and resources used for production. Utilizing emissions-intensity figures from Equation (6) and shares of business functions by industry to estimate the magnitude of emissions related to functions overcomes this methodological limitation.
3.3 Estimating the Emissions Caused by Functional Specialization
21
hypothesis has been detailed, including the measurement of emission coefficients and intermediate energy inputs. Modifying this methodology to account for the differences in the use of fabrication across industries and countries accommodates for the change in nature of international trade from sectoral to functional specialization in trade. This study attributes the emissions-intensity of exports to fabrication to show that countries with high shares of fabrication subsequently have high emissions embodied in exports, and thus qualify as pollution havens.
The research method in its current form quantifies emissions relative to exported and imported products. Yet, Timmer et al. (2019) argue that this trade metric is an incomplete measure of trade specialization, due to the lack of information on the nature of tasks involved in the creation of value added. The incorporation of functional specialization in trade provides a more complete understanding of the regional position of BRIC countries in global production networks. Particularly in the context of emissions embodied in trade, specialization in ‘dirty’ functions holds significance for domestic developmental potential with regard to human capital and pollution mitigation.
Subsequent estimations assume that fabrication functions are relatively more emission-intensive than marketing, management, and R&D functions. Given the negative correlation between GDP per capita and fabrication (Timmer et al., 2019), it is feasible to assume that the abundance of emission permits in developing countries drives specialization in fabrication. If fabrication is offshored to developing countries based on the comparatively high emission-intensity, this pattern would substantiate the PHH for this subset of countries. In order to test this assumption, the relationship between emissions and specialization in fabrication is estimated using a regression equation,
𝑐′𝐿𝑖𝑗 = 𝛽0+ 𝛽1𝐹𝑆(𝑣𝑎
𝑥 ∗ 𝐿)𝑖𝑗 + 𝜀𝑖𝑗 (7)
where 𝑐′𝐿𝑖𝑗 is the amount of extra emissions required for a unit increase in final demand by industry
i and country j, defined by Equation (6). These extra emissions are dependent on the fabrication share,
𝐹𝑆𝑖𝑗, which is a vector of country-industry values of the fabrication share of labor income used in production of exports, multiplied by the product of value added per unit of output and the Leontief Inverse. The interaction term between value added per unit of output and the Leontief Inverse is included to account for the indirect use of electricity, which is assumed to drive the higher emission-intensity of fabrication. Fabrication has relatively higher value added than other functions due to its greater factor intensity of production. Subsequently, industries where domestic value added is high are likely to produce a higher amount of emissions. Using value added per unit of output as a weight for the Leontief Inverse harmonizes the dependent and independent variables. By combining the interaction term with the fabrication share, the independent variable denotes the portion of value added generated in producing a specific final product that is attributable to fabrication. Everything included in 𝑐′𝐿𝑖𝑗 not attributed to 𝐹𝑆(
𝑣𝑎
𝑥 ∗ 𝐿)𝑖𝑗 is assumed to derive from technological differences.
22 variables, 𝛽1 and 𝐹𝑆(𝑣𝑎
𝑥 ∗ 𝐿)𝑖𝑗, determines by how much emissions increase given the share of
fabrication in the value added generated by the respective industry. The interaction between these is defined such that
𝛾𝑖𝑗 = 𝛽0+ 𝛽1× 𝐹𝑆( 𝑣𝑎
𝑥 ∗ 𝐿)𝑖𝑗 (8)
where 𝛾𝑖𝑗 is the emissions generated by fabrication required per unit of final demand. Equation (8)
serves as the basis for calculation of emission requirements for both the RoW and the BRIC countries, to determine the relative share of fabrication generated emissions in exports. The classification of exports and imports within the two-region setting requires further discussion. The fabrication shares of the RoW and the BRIC countries are expected to accurately describe the share of fabrication used in their exports. The functional shares for imports, however, are derived based on the congruency which states that imports for RoW are equal to exports for Brazil. Emissions embodied in trade, 𝜋∗,
is defined individually for the RoW and Brazil as
𝜋𝐵∗ = 𝜸𝑩(𝒆𝒙𝑩− 𝒊𝒎𝑩) (9)
𝜋𝑅𝑜𝑊∗ = 𝜸𝑹𝒐𝑾(𝒆𝒙𝑹𝒐𝑾− 𝒊𝒎𝑹𝒐𝑾)
where 𝜸𝑩is composed of the emissions required in final demand considering fabrication of Brazil and
𝜸𝑹𝒐𝑾 of the RoW. Equation (9) serves as the empirical model used to evaluate the PHH for the BRIC countries between the years 1999 and 2007. The change in emissions embodied in trade quantified by Equation (9) for both the RoW and Brazil serve as the basis for analysis of the four possible outcomes outlined for the study of the PHH. These possible outcomes now account for the inherent differences in functions used for production by including fabrication in the estimation of technological differences. The product mix is thus excluded as an independent channel of emission determination, and is rather linked to the technological differences through the fabrication shares used by industries to produce the exported products.
Existing literature determines emissions embodied in trade as driven by technological differences in production, measured by emission coefficients and intermediate inputs, as well as the product mix of exports and imports. By recognizing for the change in nature of international trade, these three channels no longer serve as independent determinants of emissions. Countries no longer hold comparative advantages in the production of a product, but rather in the functions that make up the production process. In accounting for the share of fabrication used in production, technological differences between countries now also account for the specialization in trade previously indicated by the product mix. The extent to which functional specialization determines the emissions embodied in trade by BRIC countries is evaluated based on the difference between import and export content of emissions caused by fabrication used in production.
3.4 Data
23
is used in this study as the ISIC Rev 3.1 industry classification it uses matches that of the environmental accounts and Timmer et al. (2019) functional specialization database. If the 2016 Release were to be used, there is a risk of losing accuracy in the industry conversion between ISIC Rev 3 and 4.
Input-Output Tables are in current prices, expressed in millions of US dollars. As the dependent variable 𝒄′𝑳 is denoted as kilotons per million USD, changes in relative prices driven by inflation, for example, may account for variance in the variable. This may cause comparisons of metrics derived from the dependent variable, such as 𝜸𝑩 and 𝜋𝐵∗ to provide inaccurate conclusions when compared across time. However, this should not have a significant influence on the final results of the study as the PHH is evaluated in each year separately, and therefore the influence of price level differences is mitigated.
The NIOTs and WIOT published by the WIOD are used. For each of the BRIC countries, the NIOT is used as exports and imports are clearly distinguished in the publication. As no NIOT is published for the RoW exclusively, data from WIOT is utilized. Complementary to the input-output tables used, the WIOD environmental accounts references ‘Emissions to Air’ data by Gentry, Arto & Neuwahl (2012),which provides a country-industry breakdown of CO2 emissions. The WIOT and Emissions
to Air harmonize data from 35 industries for 40 countries, including all 27 EU Member States (Members as of 01 Jan 2007, including thus the UK and not Croatia), and 13 other major countries, including the BRIC. The WIOT also includes an aggregated variable for all countries not included individually, denoted as ‘OTHER’.
The final dataset used is the functional specialization in trade published by Timmer et al. (2019). Compiling survey and census data on occupational labor and wages, Timmer et al. (2019) trace the domestic value added in exports by different occupations and use labor income shares rather than the number of workers to determine the labor input. The classification of occupations into one of four functions, namely R&D, marketing, management, or fabrication, occurs based on similarities in the tasks performed of each occupation. For example, R&D consists of engineers and computing professionals, fabrication of assemblers and machine operators, marketing of sales persons, and customer service representatives, and management functions of general and financial managers, as well as human resources. The database organizes functions by country and industry for the 1999 to 2011 period. The country selection and industry division are identical to the WIOD 2013 Release and environmental accounts used. As part of the replication files, Timmer et al. (2019) provide each of the four business functions as shares of labor income for 1999 to 2011. These shares, organized by industry and country, are used to classify the relative importance of fabrication in the production processes of each industry and country. The lack of data on the functional mix of value added elicits the assumption that the labor share of fabrication is identical to its share in value added.
24
4. RESULTS
Before progressing into measuring the emissions embodied in trade, the importance of accounting for differences in technology between the BRIC countries and the RoW is assessed by comparing the distinct 𝒄𝑩𝑹𝑰𝑪′𝑳𝑩𝑹𝑰𝑪 and 𝒄𝑹𝒐𝑾′𝑳𝑹𝒐𝑾 production technology variables of each. Dietzenbacher et al. (2007) utilize a common variable based on the production technology of their developing country due to a lack of data. Utilizing a common variable is inaccurate if clear differences between the two 𝒄′𝑳 variables exist. To demonstrate this relationship, examples from agricultural, manufacturing, and services sectors for 1999 and 2007 are considered. Industries are with high export shares for 1999 that also accounted for a large share of other BRIC country exports were selected. Specifically, Agriculture, Forestry, and Fishing (‘Agriculture’), Textiles and Textile Products (‘Textiles’), Basic Metals and Fabricated Metal (‘Metals’), and Renting and Other Business Activities (‘Business’) are considered. To reach a single comparable value of industrial emissions multipliers for RoW, the average of each country included is taken.
Table 1: Detecting differences in production technologies between BRIC countries and RoW
Brazil Russia India China
1999 2007 1999 2007 1999 2007 1999 2007 𝒄𝑩𝑹𝑰𝑪′𝑳𝑩𝑹𝑰𝑪 Agriculture 0.92 0.40 6.22 0.93 2.65 1.42 4.22 1.12 Textiles 0.31 0.14 1.09 0.14 4.20 1.63 2.94 1.59 Metals 1.43 0.72 16.59 3.37 5.81 5.30 11.18 5.78 Business 0.83 0.35 2.86 0.72 0.47 0.36 1.81 0.80 𝒄𝑹𝑶𝑾′𝑳𝑹𝑶𝑾 Agriculture 1.13 0.88 0.99 0.87 1.09 0.86 1.05 0.86 Textiles 0.66 0.54 0.64 0.54 0.56 0.50 0.60 0.50 Metals 2.52 2.02 2.13 1.96 2.40 1.91 2.27 1.89 Business 0.61 0.53 0.56 0.52 0.62 0.53 0.59 0.52
Note: Values are expressed in kilotons by million USD. For example, agriculture in Brazil 1999 produced 0.92 kt
of carbon dioxide emissions per million USD of exports.
Differences in technologies used in production between the BRIC countries and RoW are significant, as seen in Table 1. These findings verify the importance of differences in production technologies between countries as a channel of emissions embodied in trade. By overlooking these differences and assuming the RoW has identical production technologies to a developing country, the conclusions Dietzenbacher et al. (2007) derive with regard to the PHH are likely to be biased. Given that developing countries have higher emissions requirements than advanced countries, as illustrated by differences in emission multipliers in Table 1, the assumption of identical production technologies leads to the measurement of emissions embodied by the RoW (𝜋𝑅𝑂𝑊) at a higher level than in reality. Consequently, this positively biased RoW emissions balance lowers the likelihood of finding evidence in support of the PHH as the emissions balance is less likely to be negative.
25
emissions for Agriculture and Textiles are found between the BRIC countries. In contrast, emissions by output (c) indicate a comprehensive decrease. The distortion of the true volume of emissions caused by a higher nominal output in 1999 than 2007 potentially drives the notable reduction of emission multipliers found. Fluctuations in price levels rather than emissions may mask changes in technologies especially for industries that use intermediate energy inputs with volatile prices, as is the case with the Metal industry for Russia and China. For the RoW economy, the averaged emissions multipliers indicated smooth the individual countries that are significantly affected by price volatility of energy inputs.
To have a point of comparison between the conclusions drawn by the methodology adopted by current literature and the forthcoming adjustment to account for functional specialization, the results drawn from the method used by Dietzenbacher et al. (2007) are first evaluated. Their sectoral perspective on trade assumes that emissions embodied in trade are determined by differences in the product mix of exports and imports for each country. However, given the current availability of data and the clear variation in production technologies between countries, 𝒄𝑩𝑹𝑰𝑪′𝑳𝑩𝑹𝑰𝑪 and 𝒄𝑹𝒐𝑾′𝑳𝑹𝒐𝑾 are used to
account for differences in production technologies and to provide a point of comparison. The emissions embodied in trade balances of BRIC countries and their respective RoW economy are evaluated based on the four possible outcomes outlined in Section 3.2.
Table 2: Emissions Embodied in Trade Based on Sectoral Specialization
Brazil Russia India China
1999 2007 1999 2007 1999 2007 1999 2007
𝜋𝐵𝑅𝐼𝐶 -53.92 4.34 1 953.7 505.61 -765.07 -1 343.5 -901.54 -513.01
𝜋𝑅𝑂𝑊 -566.61 -1 445.2 392.72 115.03 1 664.4 873.56 3.81 413.43
Note: Values are expressed in kilotons (kt). For example, the balance of emissions embodied in trade (related to 1
billion USD of total exports and imports) for Brazil in 1999 is roughly 54 kt of carbon dioxide.
The PHH is decomposed into emissions embodied in imports and exports for a two-region setting, where a positive emissions balance for BRIC countries, 𝜋𝐵𝑅𝐼𝐶 > 0, combined with a negative balance for the RoW, 𝜋𝑅𝑂𝑊 < 0, signifies that BRIC countries are characterized as a pollution haven. While Brazil and the RoW initially both gained from trade in terms of emissions in 1999, this shifted to a pollution haven classification for Brazil in 2007. A distinct difference may be the rise of mining in Brazil’s exports in 2007, an industry with relatively high emissions by output. A quite dissimilar result is in turn found for India and China, as both countries gain whereas the RoW loses from trade in terms of emissions for both periods. Table 2 corroborates the results found for India for 1991 and 1996 by Dietzenbacher et al. (2007). India is not classified as a pollution haven, and is rather moving away by the decrease in its emissions embodied in trade. China, in contrast, increased its balance of emissions embodied in trade despite still gaining relative to the RoW. As expressed based on Table
1, relative changes in price levels may bias the true changes in emissions for China.
26
to stem from the reliance on emission-intensive oil in, and as, an exported product. The relative price changes of oil and other energy inputs, however, may skew results particularly for 1999 in comparison to 2007. The sizable drop in magnitude of emissions embodied in trade for Russia suggests that these inflationary pressures may significantly distort the true changes in emissions for the period.
The underlying assumption in relating fabrication to emissions embodied in trade is that developing countries have pervasive, high shares of fabrication compared to advanced countries. The values of the independent variable summarized in Table 3 for the sub-section of industries are unitless, and thus do not provide anything beyond relative sizes and changes. The underlying assumption thus holds as BRIC countries have systemically higher values for the independent variable relative to two advanced countries included in the RoW economy, and these high levels tend to continue through the period in question. The pervasiveness of fabrication is not clearly indicated in Table 3, however, as most values tend to decrease between 1999 and 2007. As such, it is plausible that fabrication functions have started to be offshored from the BRIC countries to less developed regions, or that the volume of value added produced by fabrication has decreased due to an increase in another function or due to relative price changes. Returning to the iPod example, this possible drop in value added across time may also be explained by a shift from general production of parts in 1999 to the assembly of a single module in 2007. Similar results can be found in comparison to the other countries included in the RoW.
Table 3: Detecting Differences in the Independent Variable for BRIC Countries Relative to RoW
Brazil Russia India China
1999 2007 1999 2007 1999 2007 1999 2007 Agriculture 0.96 0.95 1.20 0.92 1.42 1.24 1.33 0.92 Textiles 0.53 0.40 0.35 0.40 0.80 0.57 0.72 0.69 Metals 0.54 0.47 0.84 0.79 0.92 0.73 1.22 1.15 Business 0.09 0.04 0.19 0.19 0.06 0.03 0.15 0.15 USA Netherlands 1999 2007 1999 2007 Agriculture 0.39 0.39 0.20 0.17 Textiles 0.32 0.32 0.21 0.20 Metals 0.47 0.39 0.27 0.24 Business 0.08 0.08 0.06 0.05
Note: Values are unitless due to the composition of the independent variable. No economic meaning exists beyond
the relative comparison across time and space.
27
number of outliers may drive the linear relationship for Metals, as most data points are concentrated at low levels of fabrication and emissions. The heterogeneity present may indicate that a linear relationship is not best fit to describe the two variables for the countries selected. The dispersal of data points may signify that the regression coefficients are driven by outlying countries rather than be representative of the true relationship.
Figure 4: Relationship between dependent and independent variable for manufacturing sub-sector Metals
Observations correspond to the Basic Metals and Fabricated Metal industry of 40 countries included in the study. Each data point plots the amount of c’L (in kilotons (kt) by million USD) changes for a single unit increase in 𝐹𝑆(𝑣𝑎
𝑥 ∗ 𝐿)𝑖𝑗 (unitless). Lines correspond to the line-of-best-fit to indicate the approximated trend incorporated into the regression coefficients.
Similar results for Metals are presented by the results of the regression analysis in Table 4 for the sub-set of important export industries, and in Table A1 for the entire set of industries. The coefficient for the independent variable (𝛽1) is valid at a 1% confidence level, with the exception of the metal and business industries in 2007 (Table 4). The outliers found in Figure 4 do thus influence a suboptimal line of best fit. However, the statistically significant and positive coefficients verify that fabrication has a positive effect on emissions. Significant variation across industries in the magnitude of this effect exists, for example, a unit increase in the independent variable had a much more notable increase in emissions by 2.45 kilotons per million USD in 1999 for Agriculture, whereas for Metals this is 7.86 kilotons per million USD. The smaller coefficients in 2007 relative to 1999 indicate that this effect has diminished but is still relevant. Since the independent variable is unlikely to take a zero value and emissions cannot be negative, the statistically insignificant intercept coefficient (𝛽0) for most industries can be disregarded.
28
Table 4: Linear Regression Results for Equation (8) for a sub-section of industries
1999 2007 𝛽0 𝛽1 𝑅2 𝛽0 𝛽1 𝑅2 Agriculture -0.66 2.45*** 0.36 0.27 0.97*** 0.17 Textiles -0.79** 3.76*** 0.31 -0.33 2.49*** 0.22 Metals -1.41 7.86*** 0.27 0.38 3.70** 0.16 Business 0.14 6.60*** 0.33 0.45*** 1.22 0.03 Significance level *** 1% ** 5% * 10%
Number of Observations per industry and year: 40
Decomposition of the dependent variable into the share explained by differences in fabrication and that explained by differences in production technology allows for the relative importance of these distinct channels to be evaluated. Based on the R-squared, the percent of variance explained by fabrication shares ranges from 3% to 36%. The share of fabrication thus has a non-negligible effect on emissions, which verifies the importance of accounting for this scope of trade specialization in future research. Differences in production technology still account for majority of emissions. These low values for R-squared, however, also indicate a weak linear relationship between the variables. The independent variable thus does not explain a significant portion of variation in production emissions. This variation can be detected as the non-normal pattern of residuals in the right tail in the distribution of residuals for all industries. These outliers may be influenced by non-linear effects, or may stem from deviations in factor intensity at the occupation level. The forthcoming robustness analysis corrects for these outliers.
As indicated by Timmer et al. (2019), functional specialization is relatively pervasive. Figure 5 illustrates the emissions required in final demand given the fabrication shares used in production of exports, 𝛾𝑖𝑗, of each of the forty countries included in the WIOT. The 1400 data points of each
29
Figure 5: Variation in fabrication share across BRIC countries and countries composing of RoW
Observations correspond to 35 industries in each of the 40 countries included in the study. Each data point plots the amount of c’L (in kilotons (kt) by million USD) changes for a single unit increase in 𝐹𝑆(𝑣𝑎
𝑥 ∗ 𝐿)𝑖𝑗 (unitless). The line corresponds to 45° to indicate an outcome where values in 1999 and 2007 are identical.
Attributing emission embodied in trade to the share of fabrication and production technologies provides the more up-to-date methodology in terms of the nature of trade for analyzing the PHH. The results for the main Equation (9) are summarized in Table 5. For most BRIC countries and their respective RoW economies, accounting for fabrication reduces the magnitude of emissions embodied in trade as well as the differences between 1999 and 2007. Smaller figures are expected as emissions are only attributed to one of four functions. The smoothing effect does not necessarily stem from smaller balances of trade emissions, as for half these balances have increased relative to the other model, but likely rather due to the pervasiveness of fabrication. The hypothesized increase between 1999 and 2007 in the magnitude of positive emissions balances is not supported by the data. The conclusions based on the four possible outcomes are largely similar when using sectoral and functional specialization.
Table 5: Emissions Embodied in Trade Based on Functional Specialization
Brazil Russia India China
1999 2007 1999 2007 1999 2007 1999 2007
𝜋𝐵𝑅𝐼𝐶∗ 64.44 50.13 497.26 284.40 13.29 -639.13 -409.49 -90.74
𝜋𝑅𝑂𝑊∗ -530.08 -1 037.9 1 038.9 761.93 472.59 648.70 -34.79 226.87
Note: Values are expressed in kilotons (kt). For example, the balance of emissions embodied in trade (related to 1
billion USD of total exports and imports attributed to fabrication) for Brazil in 1999 is roughly 64 kt of carbon dioxide.
