Responsibility and Trade Emission Balances: An Evaluation of Approaches

This paper compares two concepts to evaluate the international responsibility of a country with respect to its emissions. Using a multi-regional input-output model, we show that the trade emission balance and the responsibility emission balance yield the same result. In practical work, however, a lack of data availability implies that the same technology assumption has been commonly adopted. In that case, also a third alternative exists, which simply evaluates the emissions embodied in the trade balance of the country. This third alternative yields the same results as the other two approaches at the aggregate level. At the level of individual products, however, the results are clearly different and it turns out that the third alternative answers a different question. That is, it is appropriate for measuring the emission content of the products that cross the border. In our empirical application, we consider Spain in 1995 and 2000, distinguishing nine different gases: CO2, CH4, N2O, SF6, HFCs, PFCs, SO2, NOx, and NH3.


Introduction
International trade links the consumption in one country to emissions generated in other countries. As a consequence, the emissions actually generated in one country do not need to be the same as the emissions that are (directly and indirectly) necessary for its consumption. The debate about the implications of international trade on the environment, although not new, has gained much importance in the last decade due to the Kyoto protocol, which determines emission ceilings on six specified greenhouse gases (GHG) for each ratifier country (United Nations, 1997).
These national targets, as well as the official data for monitoring countries' achievements, have been fixed on the basis of emissions generated by domestic production including emissions embodied in exports but not those embodied in imports. Such approach is particularly relevant to open economies and it has been suggested that international trade should be considered to establish equitable and feasible reduction targets (Munksgaard and Pedersen, 2001).
This situation leads to the question of how to evaluate the environmental responsibility of one country in global terms; that is, what would be the responsibility of one country regarding the rest of the world (RoW)? Traditionally, this aspect has been approached from two viewpoints. On the one hand, comparing the emissions embodied in exports with those embodied in imports; and on the other hand, confronting the producer and consumer responsibilities.
This paper aims at contributing theoretically and empirically to this issue by merging both approaches. We use an environmentally extended multi-regional inputoutput model to define and compare two concepts for the international (environmental) balances. These are, the trade emission balance (defined as the difference between the emissions embodied in exports and imports) and the responsibility emission balance (defined as the difference between the producer and consumer responsibility). Although both emission balances are derived from different approaches, we show that they are equivalent and thus yield the same result for the international responsibility of a country.
Due to a lack of data, many studies adopt the simplifying assumption that the domestic technological and the emission coefficients apply also abroad. For a 3 small open country, a third alternative emission balance is frequently used: namely, the emissions corresponding to the trade balance (i.e. the net exports) of a country. This third alternative is attractive because it requires much less data. We show that it equals the other two emission balances at the aggregate level (i.e. measuring total emissions). However, the detailed results for the emission balance of individual products are entirely different. We will argue that the third alternative answers a different type of question and is appropriate for measuring the emission content of the products that cross the border. The original two alternatives are more appropriate for analyzing the responsibilities as emissions are attributed to products of final use.
For the empirical analysis, we have applied the model to the Spanish economy in 1995 and 2000. We compute the results for nine different gases: the six GHG regulated by the Kyoto protocol-carbon dioxide (CO2), methane (CH4), nitrous oxide (N2O), sulfur hexafluoride (SF6), hydrofluorocarbons (HFCs), and perfluorocarbons (PFCs)-and three other gases related to local environmental pressures-sulfur dioxide (SO2), nitrogen oxides (NOx), and ammonia (NH3).
The remainder of the paper is as follows. Section 2 provides the background for this study. In Section 3, we develop an environmentally extended multi-regional input-output model to define the trade emission balance and the responsibility emission balance and we will prove their equivalence. Also we will discuss the third alternative (i.e. emissions of net exports) and show the equality with the other two balances at the aggregate level. In Section 4, we analyze the results for Spain in 1995 and 2000. They clearly show that the former equivalence at the aggregate level does not hold at the detailed product level. Finally, conclusions are presented in Section 5.

Background
As shown the comprehensive surveys of Jayadevappa and Chhatre (2000) and Wiedmann et al. (2007), since the 1970s there has been a growing interest on the interactions between trade and environment. The first study in analyzing the emissions contained in the international trade of the United States (U.S.) was Walter (1973). However, in the last years there has been an increase of studies that compare the emissions associated with exports and imports for different countries, such as Germany and the United Kingdom (U.K.) (Proops et al., 1993), Japan (Kondo et al., 1998), Denmark (Munksgaard and Perdersen, 2001), Brazil (Machado et al., 2001;Tolmasquim and Machado, 2003), Spain (Sánchez-Chóliz and Duarte, 2004), Italy (Mongelli et al., 2006), India (Mukhopadhyay and Chakrabory, 2005;Dietzenbacher and Mukhopadhyay, 2007), and Turkey (Tunç et al., 2007). Antweiler (1996) calculated an index of pollution terms of trade for 164 different countries.
Although there are methodological differences, all these studies have a common feature, i.e. they assume that the RoW has the same technology as the country analyzed. This assumption is frequently adopted due to the lack of data concerning the 'global' technology of the RoW. However, improvements in data availability and quality have made possible, in some cases, to take different technologies between regions into account. Thus, bilateral trade studies have been carried out between Japan and Canada (Hayami and Nakamura, 2002), Japan and South Korea (Rhee and Chung, 2006), Japan and the U.S. (Ackerman et al., 2007), and Canada and the U.S. (Norman et al. 2007). Other studies have considered more regions such as Wyckoff and Roop (1994) who estimated the embodied emissions in imports of six OECD countries (Canada, France, Germany, Japan, the U.K., and the U.S.). Ahmad and Wyckoff (2003) enlarged this study estimating the emissions embodied in international trade of goods of 24 OECD countries. Lenzen et al. (2004) calculated the trade balance taking into account five regions (Denmark, Germany, Sweden, Norway and the RoW); Nijdam et al. (2005) analyzed the impacts of Dutch household consumption considering the Netherlands and three different world regions; Peters andHertwich (2006a, 2006b) analyzed the environmental impacts of Norway's final demand aggregating all its trading partners into seven regions; and Weber and Matthews (2007) analyzed the environmental effects of changes to the structure and volume of U.S. trade with Canada, China, Mexico, Japan, Germany, the U.K. and Korea. The largest multi-regional studies consider 87 countries: Peters and Hertwich (2008a) analyzed CO2 emissions embodied in international trade; Wilting and Vringer (2009) also included CH4, N2O and land use; Hertwich and Peters (2009) compared carbon footprints; and Andrew et al. (2009) quantified errors induced by applying different assumptions. However, the principal drawback to this 5 approach is the difficulty of getting the necessary and detailed data on interregional transactions; and when this information is available it should be cautiously used because of the lack of consistency and accuracy of some databases (Peters and Hertwich, 2008a). Moreover, even in these multi-regional studies it is necessary to make some assumption about the technology of the RoW when it is considered as a region in the model. Nevertheless, all these studies are a sign of the importance of considering different technologies when estimating the emissions embodied in trade.
On the other hand, the international environmental responsibility of a country can also be defined from a different perspective. Since the place where the production of goods and services takes place does not need to be the same as the place where these products are 'consumed', we can define the responsibility of a country from two sides: the producer or the 'consumer' standpoint1 (Proops et al., 1993;Steenge, 1999;Munksgaard and Pedersen, 2001). The former establishes that any country is responsible for the emissions within the country associated with its domestic production regardless where it is going to be 'consumed'. Whereas the latter determines the country's responsibility depending on its 'consumption', i.e. a country is responsible for emissions generated globally in order to satisfy its domestic final demand regardless where it has been produced. This second perspective coincides with the philosophy of the ecological footprint (Rees, 1992;Rees and Wackernagel, 1994;Wackernagel and Rees, 1996), which reflects total (real and virtual) land that a country would need to absorb the impact of its residents on the earth. This concept can be appropriately calculated by applying multi-regional input-output models (Bicknell et al., 1998;Wiedmann et al., 2006;Turner et al., 2007).
The distinction between producer and 'consumer' responsibility has also generated other interesting contributions closely related to our study, but not investigated in this article. On the one hand, the methodological debate about sharing responsibility that deals with the question of how to assign the environmental responsibility (Gallego and Lenzen, 2005;Rodrigues et al., 2006;Lenzen et al., 6 2007;Rodrigues and Domingos, 2008); and on the other hand, the analysis of the appropriateness of production-based versus consumption-based national emission inventories (Peters and Hertwich, 2008b;Peters, 2008).

The Multi-regional Input-Output Framework
We consider a world economy consisting of two regions (r, s = 1, 2) that may differ in production technology and/or pollution patterns. Each region is composed of n sectors, which produce one product that might be used by other sectors as intermediate input (either at home or abroad) or consumed or invested (at home or abroad) as final product by final user categories such as households and the government. The model is given by , or in its partitioned form as2 indicates the Leontief inverse. In partitioned form, this reads as follows (2) 7 In order to estimate emissions associated with the production in each region, we define the matrix of atmospheric emission coefficients r W , whose element r kj w indicates the domestic emission of pollutant k per unit of industry j's output in region r. The emissions (or pollution) generated in each region are given by (3) Note that element rs kj p of matrix rs P gives the total amount of (extra) emissions of pollutant k in region r, due to the delivery of one (extra) unit of product j from region s to its final users (either at home or abroad).

Trade and Responsibility Emission Balances
The trade emission balance for region 2 may be obtained in the same way. In this two-region case, we then have . Note that the term 21 12 12 21 y P y P + is both in (4) and in (5). It reflects emissions that are first imported and then exported again. In the balance in (6) they cancel each other out.
The international responsibility of a region is defined from a different perspective. That is, by comparing the emissions produced inside the region with the emissions required by its domestic final users. The producer's responsibility in region 1 covers all emissions generated by the region's production, i.e.
The responsibility emission balance for region 1 (i.e. 1 reb ) is defined as the difference between the emissions due to the producer responsibility and the 'consumer' responsibility. That is, equation (7) minus equation (8). It is readily seen that this yields equation (6), which proves that 1 1 reb teb = . 9 If the balances have a positive sign it implies that the emissions embodied in exports are higher than those embodied in imports, i.e. the region is a "net importer" of emissions or, alternatively, the producer responsibility is larger than the 'consumer' responsibility. 4 Since official statistics reflect territorial data (i.e. emissions generated by domestic production) this region would actually be less responsible for the environmental pollution than is reported. In terms of ecological footprints, a positive balance would reflect an ecological creditor region, which has enough ecological resources within its own territory and, hence, it would be 'ceding' land to the other region.
The analysis in this section can be readily extended to the multi-region case and the results remain the same. The only difficulty is that the equality 1 2 teb teb − = for the two-region case does not hold for the multi-region case. Instead, we have 'consistency' of the balances in the sense that 0 ...
The intuition behind the equality of the responsibility and the trade emission balance has been widely discussed at the aggregate level in a single country setting and has led to simple graphical representations (see e.g. Weidema et al., 2006).
Whether it holds at a sectoral level within a multi-regional setting depends on the definitions of emissions in exports, emissions in imports, producer responsibility, and 'consumer' responsibility. In this paper, we have started from plausible definitions (4), (5), (7), and (8), and proved the equivalence. In contrast, e.g. Peters and Hertwich (2008a), did the opposite and imposed the equality r r reb teb = to define 'consumer' responsibility in a single country case with multiple countries, which corresponds to setting 0 21 12 = = A A in our equation (1).

Special Case: A Small Country Using the Same Technology
The special case of a small country is discussed in this subsection. Using the model of the previous subsection, the country is region 1 and the RoW is region 2. Although the exports of the country are non-zero, for a small country they will typically be 10 negligible when compared to the RoW's outputs. Therefore it seems plausible to assume that 0 12 = A .
The second assumption that we will make is quite common in the literature.
Although technology will differ across countries, actual data for the RoW are typically lacking. Therefore it is assumed that the production technology and the emission intensities are the same for the country and the RoW. Note that the technology is given by the structure of the inputs, no matter whether domestically produced or imported. That is, assuming the same technology implies We can now simplify the notation. Writing for the technical input matrix (that holds for both countries), equation (1) becomes Instead of equation (3) we now have The emission balance (no matter whether trade or responsibility) in (6) now becomes Usually, a much simpler expression is used when the assumption of the same technology is adopted. That is, calculating the emissions corresponding to the trade balance (i.e. exports minus imports). So, the alternative expression is given by 11 ) ( ) (  (12) has one major advantage in terms of data requirements. Observe that the application of (11) requires that the input-output data are such that the domestically produced inputs are separated from the imported inputs. Such data are not always available and for many countries only the sum of the two matrices (i.e. the technical input matrix M A + ) is available. In addition, equation (11) requires the vectors of imports and exports separately, while equation (12) only requires the vector of net exports.
Appendix A shows that the two expressions in (11) and (12) (13) and (14) are not the same, and we will discuss the differences.

Empirical Results
This section presents the empirical results for the models presented above. The basic data sources of this paper are the 1995 and 2000 Spanish NAMEA for air emissions (INE 2005(INE , 2006. Appendix B discusses the data set and the assumptions and procedures adopted in this paper.

Aggregate Results
Tables 1 and 2 present the aggregate emission balances for Spain in 1995 and 2000. responsibility for the GHG (in CO2 equivalent) is 9% (resp. 15%) larger than producer responsibility in 1995 (resp. 2000), showing that from the ecological footprint viewpoint Spain is a debtor country that is 'usurping' land.

INSERT TABLE 2
Both emission balances can be interpreted as an indicator for the international environmental responsibility of a country. The implications of a negative emission balance, either trade or responsibility, may vary depending on the kind of pollutant considered. In the case of local and regional gases, such as SO2, NOx, and NH3, it means that Spain shifts environmental costs to other countries. For global pollutants, such as the GHG, it means that the responsibility of the country is greater than it appears from official reports.
Considering the changes in environmental responsibility over time, it follows from the last column in Table 2 that the international responsibility of Spain increased considerably between 1995 and 2000. Note that a positive (negative) percentage change indicates an increase (decrease) of the deficit on the balance. It appears that the deficit has roughly doubled for the synthetic GHG and for two of the three gases related to energy products, i.e. CO2 and NOx. For SO2 the increase was much more moderate. In contrast, the evolution of the international responsibility for the gases connected with agricultural and food activities, i.e. CH4, N2O, and NH3, was just the opposite; their deficits have decreased and Spain has even become an 'importer' of pollution in terms of NH3 emissions. The different behavior between both groups of gases mainly reflects changes in final demand structure during this period (Roca and Serrano, 2007). With respect to climate change, also the deficit for the total amount of GHG (measured in CO2 equivalent) has almost doubled. The 14 producer responsibility has increased by 17% between 1995 and 2000, whereas the 'consumer' responsibility has increased by no less than 24%.

Detailed Results at Product Level
In Section 3.3 we showed that under the assumption of the same technology, the emission balance (no matter whether trade or responsibility) at the aggregate level (i.e. for a country) can be calculated by applying either expression (11) or the simpler shortcut expression (12), which was used e.g. in Machado et al. (2001) or Mongelli et al. (2006). However, when we calculate the emission balance at the detailed level of products, the counterpart expressions (13) and (14) lead to different outcomes.
The results are given in Table 3 for GHG.7 The totals in the bottom row of Table 3 coincide with the emission balances for GHG in Tables 1 and 2. At the aggregate level, expressions (13) and (14) yield the same result. For separate products, however, the emission balances differ largely.
In fact, each expression answers a different research question. The general expression (13) gives the difference between producer and 'consumer' responsibility, whereas the shortcut expression (14) evaluates the emissions embodied in the products that cross the border.

INSERT TABLE 3
For example, for 'Pulp, paper and paper products' (sector 13, for short: paper), the emission balance in 1995 for GHG is roughly 0.6 million tonnes of CO2 equivalent. The difference between producer and 'consumer' responsibility captures: the Spanish emissions involved in producing the exports of Spanish paper to foreign 'consumers'; minus the foreign emissions involved in producing the imports of foreign paper by Spanish 'consumers'; minus the foreign emissions involved in producing the imported inputs that are required in Spain to produce paper for Spanish 'consumers'. From the first two columns in Table 3, it follows that Spain imported almost twice as much paper than it exported (for 2706 and 1367 million euros, respectively) in 1995. However, what matters in the emission balance in equation (13) is the final use or 'consumption' of paper. In the input-output framework the ultimate goal of all production is 'consumption'. Emissions are therefore attributed to or embodied in 'consumption'. In the case of paper, for example, the exports of Spanish paper to foreign 'consumers' equal 1367 million euros, whereas the Spanish imports of paper by 'consumers' amount only to 119 million euros. This means for instance that a very large part of the imports of paper (i.e. 2587 million euros) is as an intermediate input for the production of (mainly) other goods than paper. The foreign emissions involved in producing this imported paper will be accounted for in the 'consumption' of these other goods.
The shortcut expression (14) has a different focus. The vector 1 imp gives all the imports into Spain. This includes the foreign paper that goes to Spanish 'consumers', but also the imports of foreign paper by Spanish producers to make other goods (for 'consumption' in Spain or abroad). In this last case, expression (14) does not attribute emissions to the product of final use, as was the case in (13).

Instead, expression (14) reflects the viewpoint of measuring trade at the border,
where customs simply measure the goods that cross the border, no matter whether they are for final use or for further production. For example, take the exports of Spanish paper to foreign 'consumers', for which Spain imports foreign wood products as inputs. In expression (13), only the Spanish emissions for paper production are taken into account, not the foreign emissions. These foreign emissions (part of which stem from the wood products that go into Spanish paper) have been imported and then exported again, and thus cancel out. In expression (14), all emissions (Spanish as well as foreign emissions involved in producing inputs such as wood products) are included the paper exports. The emissions involved in the foreign production of wood products that are imported by Spain are not listed (i.e. subtracted) under 'paper', but under 'wood products'.
The shortcut expression (14) measures all trade as if it were a final product.
Consequently, as long as the net exports of a product are positive (negative) the emission balance will show a positive (negative) outcome. For example, in the case of paper we find an emission balance of -1.3 million tonnes of CO2 equivalent (which contrasts the +0.6 million tonnes when expression (13) is used) in 1995. Note 16 that summing over the products yields the same result for (13) and (14). This implies that at the product level, the emissions have been redistributed. In expression (13) the emissions are attributed to the product of final use, whereas in expression (14) the emissions are attributed to the products that actually cross the border.
Sector 44 'Health and social work' illustrates another consequence of the difference between the two expressions. The two first columns in Table 3 indicate that Spain neither exported nor imported any 'health services' in 1995. The emission balance thus equals zero when the shortcut expression (14) is used, because no 'health services' cross the border. Using the general expression (13), it follows from Table 3 that the emission balance was negative for all gases. The reason is that this sector used some inputs produced abroad such as chemical products (sector 16) or medical and precision instruments (sector 25). The corresponding foreign emissions are attributed to the Spanish 'consumption' of health services.
Summarizing the discussion about the contents of and the difference between the general expression (13) and the shortcut expression (14), the first seems to be more appropriate for measuring responsibilities, which rely heavily on the principle of embodied emissions. The shortcut expression comes closer to measuring the emission content of the products that cross the border.
In order to further analyze the difference between expression (13) and (14), we derive the most important players in terms of the emission balance. Table 4 shows the products that contribute the most (in either direction) to the difference between the two responsibilities, which might be relevant also from a policy perspective.

INSERT TABLE 4
The lower part of Table 4 (for the shortcut expression) is relatively simple to interpret. It corresponds to the balance of the emissions generated for the products that cross the border. If the exports are much larger (smaller) than the imports and if producing the good or service generates a large amount of emissions, it will be a key product with a positive (negative) balance. The upper part of the table (for the general expression) is less straightforward to interpret. Roughly speaking, however, we find that products that are in the initial stages of the production process have a positive balance. The reason is that these products require relatively few imported inputs. In contrast, products that are in the final stages of the production process typically rely heavily on intermediate inputs. A part of them will be imported which often leads to a negative balance.
Note that in some cases the key products coincide. On the one hand, this happens when Spain is a net exporter of a certain 'primary' product, which results in a positive balance. In 1995, this occurs for 'Other non-metallic mineral products' (sector 18) and 'Water transport' (sector 36). On the other hand, negative balances are found when Spain is a net importer of 'final' products, such as 'Food products, beverages, and tobacco' (sector 8) and 'Machinery and equipment' (sector 21).
At least as interesting are the key products for which the two expressions yield opposite outcomes. In 1995, this is observed for two products, i.e. 'Basic metals' (sector 19) and 'Chemicals and chemical products' (sector 16). In both cases, the imports are substantially larger than the exports, which explains why they show a negative balance for the shortcut expression (14). From the perspective of expression (13), however, basic metals and chemicals are typical 'primary' products that are used by many other industries as an intermediate input. The Spanish imports of these products (and their foreign emissions) will therefore be taken into account by (i.e. attributed to the 'consumption' of) other products and not by basic metals or chemicals.
The results for 2000 show that the cases where key products have a coinciding balance decreased, whereas those with opposite balances increased. We also find a key product ('Wholesale and retail trade', sector 33) for which the exports exceed the imports (and thus a positive balance in the lower part of the table), whereas expression (13) shows a negative balance (in the upper part of the table).
Although this sector is one of the largest exporters, its services include a relatively small amount of Spanish emissions, whereas its production relies heavily on imported intermediate inputs (and their corresponding emissions). 8 Also it is interesting to observe that 'Air transport' (sector 37) has taken over the role of 18 'Water transport' (sector 36) and that the imports of 'Crude petroleum and natural gas' (sector 5) have almost tripled (with an obvious effect when using the shortcut expression) between 1995 and 2000.

Conclusions
On the basis of a multi-regional input-output model we defined and compared two concepts to evaluate the international responsibility of emissions generated by one country. We defined the trade emission balance as the difference between the emissions embodied in a country's exports and imports; and the responsibility emission balance as the difference between the responsibility of one country as a producer and its responsibility as a 'consumer'. Both emission balances were shown to be equivalent in the sense that they yield the same result. So, a negative (positive) balance would indicate that the country is 'exporting' ('importing') pollution and it would be more (less) responsible. The implications of these results may vary depending on the kind of pollutant considered: when analyzing GHG, the country's responsibility on global environmental pressures would be greater than the official statistics report; when analyzing local gases, the country would be displacing environmental costs to other territories.
Although it has been widely recognized in the last decade that trade in emissions can be studied properly only if multi-regional input-output tables are used, such tables are rare.9 Therefore, several studies have applied the same technology assumption (both in terms of production and emissions). We showed that-under this assumption-the two emission balances above are equivalent to a shortcut alternative.
This third approach evaluates the emissions corresponding to the trade balance of the country and has a major advantage in terms of data requirements. The equivalence, 9 An early exception is the series of tables for a set of European countries (in current and constant prices, for the years 1970, 1975, 1980, and 1985) that can be downloaded at http://www.regroningen.nl/irios_tables.shtml. Recently, however, several projects have started that focus on the construction of multi-country input-output tables. For instance, the EU-funded project, EXIOPOL, aims at constructing a multi-country for 43 countries with detailed information for environmental aspects (see http://www. feem-project.net/exiopol/, Tukker et al., 2009). Another EUfunded project, WIOD, aims at constructing a time series of annual inter-country input-output tables (in current and constant prices) for 40 countries, including socio-economic and environmental satellite accounts (see http://www.wiod.org). however, applies only at the aggregate level where responsibilities are evaluated for an entire country. The equivalence does not hold at the level of individual products and provides a true alternative. The first two types of emission balance are in particular appropriate for evaluating responsibilities at the product level, because these approaches attribute emissions to the 'consumption' of products. The alternative approach (which gives the emission of the trade balance) is more appropriate for evaluating the emissions that are embodied in the products that cross the border.

Appendix A. Equivalence of expressions for the emission balances.
Our starting point is equation (13) Next consider the sum of the two matrices in the bracketed term  (14).