Embodied Environmental Emissions in U.S. International Trade, 1997

Jun 13, 2007 - Energy and Air Emissions Embodied in China−U.S. Trade: .... monitoring of the Swedish building and real estate management sector. Sus...
0 downloads 3 Views 190KB Size
Download PDF
Environ. Sci. Technol. 2007, 41, 4875-4881

Embodied Environmental Emissions in U.S. International Trade, 1997-2004 CHRISTOPHER L.WEBER* AND H. SCOTT MATTHEWS Department of Civil and Environmental Engineering and Department of Engineering and Public Policy, Carnegie Mellon University, Pittsburgh, Pennsylvania 15213

Significant recent attention has been given to quantifying the environmental impacts of international trade. However, the United States, despite being the world’s largest emitter of greenhouse gases and having large recent growth in international trade, has seen little analysis. This work uses a multi-country input-output model of the U.S. and its seven largest trading partners (Canada, China, Mexico, Japan, Germany, the UK, and Korea) to analyze the environmental effects of changes to U.S. trade structure and volume from 1997 to 2004. It is shown that increased import volume and shifting trade patterns during this time period led to a large increase in the U.S.’ embodied emissions in trade (EET) for CO2, SO2, and NOx. Methodological uncertainties, especially related to uncertainties of international currency conversion, lead to large differences in estimation of the total EET, but we estimate that the overall embodied CO2 in U.S. imports has grown from between 0.5 and 0.8 Gt of CO2 in 1997 to between 0.8 and 1.8 Gt of CO2 in 2004, representing between 9-14% and 13-30% of U.S. (2-4% to 3-7% of global) CO2 emissions in 1997 and 2004, respectively.

Introduction One of the most well-studied aspects of the life cycle assessment (LCA) and sustainability literature has been analyzing the energy and environmental emissions associated with household goods and services (1-3). While this research has been done since the 1970s, it remains of active interest due to changing energy, emissions, and consumption patterns around the world and increasingly refined and varied modeling techniques (3, 4). At the same time, there has been increasing interest in studying the effects of rapid globalization and escalating international trade on environmental impacts at the national level (5-7). Besides empirical evidence of increased global trade (8, 9) (see Figure 1a for the U.S. case) a large amount of literature exists theoretically explaining the motivation of international trade and firm-level location decisions. These decisions, driven largely by global differences in the prices of capital, labor, and materials, have profound consequences for the environment given international differences in economic efficiency, production methods, and energy/ emissions structures (10-12). In general, theory suggests that each nation will produce the goods or services it has a competitive advantage in making, potentially leading to highvalue goods (electronics, personal services) being made in * Corresponding author e-mail: [email protected]. 10.1021/es0629110 CCC: $37.00 Published on Web 06/13/2007

 2007 American Chemical Society

the developed world and low-value or pollution-intensive goods being made in developing nations. This theoretical background has driven substantial research on the issue of embodied emissions in trade (EET) and the potential for “carbon leakage” outside the regulatory control of the Kyoto Protocol, despite the many uncertainties (see below, Uncertainties section) in calculating exact values for these quantities (13-16). Recent empirical work has focused mostly on Europe; research on the United States, the world’s largest economy, has been either at a very aggregate level or has lacked data for U.S. trading partners (16, 17). Part of this may be due to complexity issuessthe U.S. trades with over 200 countries and as we suggest below, major uncertainties with aggregated data sources, international currency conversion, and foreign emissions data complicate the study of U.S. EET. However, given the importance of the United States to international environmental concerns and its tremendous import growth over the past decade (imports rose nominally 128% from 1996 to 2006 while exports rose only 65% (8)), it is a prime candidate for a detailed analysis of EET. This work analyzes the balance of embodied CO2, SO2, and NOx emissions in U.S. international trade using a multi-country input-output model of the U.S. and its seven top trading partners. The following section describes the methods and data utilized for the study, followed by sections describing results, uncertainties, and conclusions.

Methods and Data Multi-Regional Input-Output (MRIO) Model. The method utilized is multi-regional input-output analysis. Inputoutput analysis (IOA) has seen growing use in the field of environmental assessment recently (3-6), but its origins date back to Leontief’s development from the 1930s to the 1960s (18). IOA has several advantages for use in LCA, such as being more time-efficient than process-based LCA as well as reducing cutoff error, one of the major drawbacks of processbased LCA (19). However, IOA has its drawbacks as well. For example, input-output tables are at the national level, and modeling usually assumes domestic production of imports, which can lead to significant errors in open economies. Aggregation in economic sectors is also a significant problem, but for analysis of large groups of products, detailed process models involve an impractical amount of work. The use of MRIO models theoretically solves the first major error type in IOA by using different technology for different regions (i.e., countries). The use of more disaggregated input-output tables, such as the 491 × 491 benchmark input-output model of the United States, helps to minimize aggregation error. Data Sources. Economic MRIO models require two major types of data: trade data between modeled regions and input-output tables for each region. Commodity trade data for the United States for approximately 18 000 commodities over the years 1997, 2002, and 2004 were obtained from the U.S. Census (9). This time frame was chosen because 1997 represents the most recent detailed input-output table for the U.S., and import levels between 1997 and 2002 and 2002 to 2004 displayed relatively fast growth. Data for U.S. trade in services were taken from two sources: service trading data published by the Bureau of Economic Analysis (BEA; 20) and the less-detailed annual input-output tables published by BEA. These tables are extrapolated versions of the benchmark tables, corrected for annual outputs and trade (21). While the number of countries to include is by definition somewhat arbitrary, seven (Canada, China, Mexico, Japan, VOL. 41, NO. 14, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4875

FIGURE 1. Total exports and imports for the United States by commodity (frame a) and embodied emissions in imports (I) and exports (E) for CO2 (frame b), SO2 (frame c), and NOx (frame d). EET is calculated using domestic-only and full models (denoted full) and MER and PPP (denoted p) exchange rates. The dotted line represents the best estimate for EEI and the shaded areas represent feasible ranges for EEE and EEI. Korea, the United Kingdom, and Germany) was chosen because after the top seven countries, the marginal increase in model coverage of imports dropped below 3% for each year (9). The rest of world (ROW) was modeled as the United States for data availability reasons and because the U.S. represents the most diverse economy in the data set. This modeling choice does not assume that any specific unmodeled trading partner is similar to the U.S., but rather that the average of the remaining partners is similar to the U.S. (see ref 6). This decision is of course debatable and is associated with significant uncertainty. Table 1 shows each country’s exports to the U.S. in each year, the year of each country’s input-output table, and the number of sectors in the table. Each country’s statistical agency was the source for the original input-output table, with the exception of Mexico, which was taken from source data for the Global Trade and Analysis Project (22-28). Due to sectoral classification issues between North America, which uses the North American Industrial Classification System (NAICS) and the rest of the world, which predominantly uses the International Standard Industrial Classification system (ISIC), a standard concordance between the systems was necessary and was developed from U.S. BEA and Statistics Canada data (29, 30). Because these systems are not completely intercomparable some uncertainty exists in concordances between them, but this uncertainty is largely handled by use of standardized concordances (29, 30). Model Development. As discussed in recent work, there are two main ways to define the “balance of embodied pollution in trade” (BEET), differing by allocation of emissions related to multi-directional trade (“through-trade”). Throughtrade is defined as imports used by one economy to produce its exports (31, 32). Here we use what past authors have called 4876

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 14, 2007

TABLE 1. Model Countries, Input-Output Data, and Total Imports and Exports in 1997, 2002, and 2004, in Current Dollars (billions) and Percentage of Total Yearly Tradea IO year sectors Canada China Germany Japan Korea Mexico United Kingdom rest-of-world United States total, imports Annex 1% total, exports deficit ($B)

1997 1997 2000 2000 1995 1989 2000 1997 1997

117 124 59 104 168 92 48 491 491

imp $B 1997

imp $B 2002

imp $B 2004

187 (18%) 96 (9%) 80 (8%) 147 (14%) 28 (3%) 38 (4%) 51 (5%) 393 (38%) -

232 (17%) 154 (11%) 72 (5%) 141 (10%) 42 (3%) 153 (11%) 45 (3%) 512 (38%) -

280 (16%) 243 (14%) 88 (5%) 151 (9%) 55 (3%) 174 (10%) 52 (3%) 669 (39%) -

1021 54% 919 102

1350 50% 926 424

1712 47% 1099 613

a Annex 1% refers to percentage of imports from ratified Annex 1 Kyoto parties. Deficit refers to U.S. dollar-valued trade deficit.

the “embodied emissions in trade” accounting (EET), roughly akin to Lenzen et al.’s “unidirectional trade model” as our base case and an approximation of the “embodied emissions in consumption” (EEC) or “multidirectional trade” accounting for an upper-bound estimate of embodied emissions in imports (EEI) and exports (EEE) (6, 32). The former (EET) allocates all embodied emissions to the direct trade partner, while the latter (EEC) attempts to distinguish between the multi-directional flows into and out of an economy. For instance, the EET formulation treats emissions from U.S. manufacture of semiconductors for export as allocated to the U.S., regardless of whether these devices end up in

computers made in China and imported to the U.S. Some previous studies have ignored this distinction (15-17). A detailed discussion of these concepts and the following model development is available in the Supporting Information, but is summarized here. The total output of an economy, x, can be expressed as the sum of intermediate consumption, Ax, and final consumption, y

x ) Ax + y

(1)

where A is the economy’s direct requirements matrix (33). This expression can be generalized for an open economy to include imports from other countries or regions

()(

A11 x1 A x2 ) · 21 ·· ·· · xm Am1

A12 A22 ·· · Am2

· · · A1m ··· A2m · · · ··· · · · Amm

)( ) (

x1 y11 + x2 y + · 21 ·· ·· · xm ym1

∑

j*1

y1j

)

(2)

which shows the relation between total final demand in country 1, y, and output in each country, xj, to meet this final demand. Each yj1 represents imports from country j to final demand in country 1 and each y1j represents country 1’s exports to final demand in other countries (34). As the data requirements for this general case are immense (720 000 data required for the full matrix in this study), usually a series of simplifications are appropriate. Our base case (EET accounting) is shown in eq 3. It assumes uni-directional trade by definition. By replacing the diagonal Aii matrices with total (domestic and imported) matrices, Ai ) Aii + ∑i*jAij, an approximation of multidirectional trade is achieved where imports into each trading partner are assumed to be produced similarly to domestically produced goods. This model will be denoted as “full accounting,” since the full A matrix is utilized. Two advantages of these models are the ability to use each country’s inputoutput table without aggregation, which is necessary for full multi-directional trade models (6), and the ability to use unaltered bilateral trade data (31).

( )(

X1 A11 X2 0 ) · ·· ·· · Xm 0

0 A22 ·· · 0

··· ··· ··· ···

0 0 ·· · Amm

)( ) (

x1 y11 + x2 y + · 21 ·· ·· · xm ym1

∑

j*1

y1j

)

(3)

A simplified multi-directional model (of the form in eq 2) was also developed with only two regionssthe U.S. and a U.S.-modeled ROWsto approximate the level of throughtrade for the U.S. in each year. This quantity, which can only be approximated without the use of a full multi-country multidirectional model (see Supporting Information), represents the amounts of U.S. EEE that end up in U.S. imports and U.S. EEI that end up in U.S. exports. As this model was only used sparingly, its development is left to the Supporting Information. For both models, the U.S. import matrix, Am ) ∑i*jAij was taken directly from the Bureau of Economic Analysis’s supplementary input-output files (35). Where import matrices were not available with other countries’ input-output data, the standard import penetration assumption was used to derive them (see Supporting Information). All prices were deflated to 1997 dollars using chained gross output price indices at the 491-sector level (36). Because these models attempted to discover changes due to import penetration and volume only, environmental data were assumed constant with time. Thus, uncertainty in calculated emissions increases moving forward in time since efficiency improvements are not accounted for.

Environmental Data. Environmental data were taken from a variety of sources. Due to data availability issues for several countries, only three pollutants (CO2, SO2, and NOx) and energy data were collected. The EIO-LCA web model derived at Carnegie Mellon University (37, 38) was the basis for the United States’ data, and these data are originally taken from U.S. governmental sources. Data for China, available online, were based on national-level data sources and have been fully described previously (39). Data for the UK and Germany were taken from national sources (40, 41). Canadian data were taken from the Canadian EIO-LCA project and national sources (42-43). Japanese data were taken from the 3EID project at the National Institute for Environmental Studies (44). Limited national environmental data were available for Korea, so data from the International Energy Agency (45) were used for energy and CO2 data, and air pollutant data were approximated by the closest related country (Japan) and checked with totals from the EDGAR 3.2 database (46). Mexican data for CO2 were also taken from the IEA, and air pollution data were taken from the recently completed Mexican National Emissions Inventory (47). It should be stressed that the IEA and EDGAR data sources are less reliable than the national sources used for other countries. Where process-related CO2 emissions were not included in national data, they were taken from ref 48. Where multiple years of data were available, preference was given first to the base year for the country’s inputoutput table, then to data from 1997 (the base year for U.S. input-output data), and finally to the most recent data. When data were not from 1997, each country’s consumer price index was used to bring monetary values to 1997. A more detailed method would utilize sector-specific price levels. Following conversion to 1997, average market exchange rates (MER) from the Penn World Table (49) were used to bring each country’s environmental data into the common unit (mt CO2/SO2/NOx as NO2) per $M U.S. in 1997 dollars. While the alternative choice of using purchasing power parity (PPP) conversion rates has been suggested by some authors (11, 34), market exchange rates were used for this analysis with PPP values utilized for uncertainty analysis, following ref 11. This choice is partially justified by the fact that trade data used for this analysis are exchanged using daily MER values when imports are denoted in foreign currency (9). In the interest of transparency, all environmental data, aggregated to a common ISIC-based sectoral system, are shown in the Supporting Information.

Results Aggregate results are shown in Figure 1. In the upper-left quadrant is a breakdown of U.S. imports and exports by commodity group in constant dollars. The other quadrants show the total trade balance of emissions (BEET) calculated using EET and “full” accounting methods for 1997, 2002, and 2004 with both market exchange rates and PPP rates (p). We consider the full accounting MER values and the domesticonly PPP values to represent a feasible range for EEI. The reasons for such a wide feasible range are discussed below. It is shown that the value of BEET has grown for all three pollutants between 1997 and 2004, and that the rate of growth was faster between the 2 year period of 2002 to 2004 than for the 5 year period of 1997 to 2002. It is also clear that these are very large masses of emissions: for example, bestestimates for CO2 embodied in U.S. imports doubled from 0.6 to 1.3 Gt between 1997 and 2004, which represents 3-5% of world CO2 emissions in each respective year. While these numbers may seem high at first glance, they are within the range of previous estimations for embodied CO2 in U.S. trade for 1997 only (11) and similar to results obtained from a less-detailed study of U.S.-China interactions (16). VOL. 41, NO. 14, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4877

FIGURE 2. Embodied CO2 in imports by major trading partner (above) and commodity group (below). Countries are shown on the primary axis in terms of total embodied CO2, and the share of total embodied CO2 emissions by ratified Annex 1 parties to the Kyoto Protocol is shown on the alternate axis. Commodity groups are listed in full in the Supporting Information. The aggregate results can be broken down to gain useful insights into potential causation: commodity group and country are two obvious choices. We focus on CO2, and results for SO2 and NOx are shown in the Supporting Information. Figure 2 shows best-estimate EEI for CO2 in terms of trading partner and major commodity group. The growth of U.S.-China trade is clear here, as is the importance of the “rest-of-world” country group. The Supporting Information shows a financial breakdown of these ROW countries. Interestingly, results for EEI from China using MER were close to a previous study which used PPP, perhaps due to this study ignoring Chinese imports by using a full U.S.based production matrix (16). The top seven trading partners, accounting for approximately 60% of dollar-valued trade in each year (9), account for 65-80% of embodied CO2, 70-86% of embodied SO2, and 64-79% of embodied NOx emissions in the time series. Detailed results for SO2 and NOx are provided in the Supporting Information. NOx in general follows a pattern similar to that of CO2, while SO2 emissions are more region-specific, due to their dependence on fuel mix, sulfur content of fuels, and prevalence of control technology. Figure 2 also breaks down EEI of CO2 into emissions from ratified Annex 1 parties to the Kyoto Protocol and from nonAnnex 1 parties (the developing world and Australia). The share of EEI from non-Annex 1 parties is seen to be rising, from approximately 50% of EEI in 1997 to over 70% in 2004, mostly due to the massive growth in trade between the U.S. and China between these years, but also from other trading partners, such as energy-exporting nations like Saudi Arabia, Nigeria, and Trinidad and Tobago, and others, including India, Brazil, Vietnam, and Turkey. The commodity breakdown lends other interesting insights. The monetary breakdown of exports and imports in 4878

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 14, 2007

Figure 1 shows that U.S. export structure contains a relatively low-energy mix of commodities, such as services and advanced chemicals such as pharmaceuticals, while import structure contains many energy-intensive goods such as raw materials, energy goods, electronics, and several types of machinery. Large portions of the growth in EEI (Figure 2) are due to increased imports of electric and electronic goods and components (growing from 11 to 18% of total), machinery and equipment (16 to 19%), miscellaneous manufacturing (4 to 8%) and textiles (6 to 9%). Most other commodity groups decreased in relative share of EEI, but increased in volume of emissions, including primary metals and raw materials, chemicals and plastics, and transportation equipment. The remaining industry groups demonstrate stable or decreasing EEI. Given changes in trade patterns (i.e., import partners), commodity mix, and trade volume, it is relevant to ask the contributions of each of these factors to overall growth in EEI. A 3-factor decomposition analysis was performed for each pollutant to determine each factor’s relative contribution to EEI growth. For CO2, the decomposed contributions were 50%, -8%, and 58% respectively, meaning that trade volume was the largest driver for growth, closely followed by trade patterns, and that the import commodity mix actually contributed negatively to EEI growth. Similar results were seen for NOx, (50%, -7%, and 57%) while SO2 (57%, -7%, and 50%) showed more dependence on trade patterns, which is intuitive given its dependence on fuel mix and control technology. Further explanation and results are available in the Supporting Information. Uncertainties. Several types of uncertainty are important in this analysis. Past work (33, 50, 51) has detailed the uncertainties inherent in IOA, and several of these error types are likely important here but difficult to quantifysthis list includes temporal and spatial variability, allocation uncertainty, and aggregation uncertainty (51). Allocation and aggregation uncertainty are especially important for analyses of trade because different firms in a country may produce only for export or for the domestic market, and the exportproducers may be more efficient than the national average due to foreign investment. For example, the steel industry in China, particularly before WTO accession, has consisted of a mix of relatively new and highly efficient plants and older, highly inefficient plants, and it is likely that the newer plants produce more exports than the older ones. Since the aggregate mix of U.S. imports and exports contains both high-value and low-value goods, and more and less environmentally intensive commodities, the cumulative effect of this uncertainty is unclear. There are additional uncertainties related to MRIO modeling that are not present in national studies. The uncertainty associated with the treatment of rest-of-world (ROW) is likely high, but is difficult to quantify without the inclusion of more countries. Comparison to “world averages” cited in ref 6 shows the U.S. is probably a reasonable choice to model ROW. Another major uncertainty is in the definition of the BEET itself. Here BEET was calculated with the domestic-only tables and the full tables (marked full), where the full matrix assumes imports to be produced with technology similar to that of the producing country. This assumption will clearly be more appropriate for some countries than others. As previously discussed (6, 31), a true consumption perspective of EET must consider throughtrade, since U.S. exports which are reimported into the U.S. should be allocated to the U.S., not to the exporting economy. We estimate through-trade to be relatively small but not ignorable; for example, CO2 EEE which return to the U.S. for consumption and CO2 EEI which leave the U.S. in exports were estimated to be 21-30 Mt CO2 and 56-79 Mt CO2, respectively, between 1997 and 2004, representing 5-6% of

total EEE and 11-13% of total EEI in these years. These numbers should be seen only as first estimates, though the fact that they are similar to ratios seen previously (6) is encouraging. More detail in available in the Supporting Information. Perhaps most important to overall uncertainty are issues related to prices. Each country’s input-output table is derived in local currency, and further, due to the time lag of publishing IO data, it is often impossible to find several tables from the same year. The use of a simple consumer price index to convert monetary values between years is biased toward consumer goods inside each country, as opposed to intermediate and capital goods, which are a substantial portion of imports to the U.S. (see Figure 1). However, the use of simple deflation methods such as a CPI is common to limit data requirements (5). Future work should analyze the importance of this assumption. Ideally, one would build input-output tables in physical units to avoid issues in differential pricing between countries. However, data rarely allow for this possibility outside of a few homogeneous commodities. Thus, price variability must be accounted for, and price levels (measured as PPP) can vary substantially between countries. For most developed countries, the difference between MER and PPP is relatively small, reflecting similar price levels. However, the difference between MER and PPP can be much higher for developing countriessa factor of about 2 for Mexico and 4 for China in 1997 (49) which explains the large difference in EEI seen in Figure 1. It is likely that the true value of EEI falls somewhere between the values calculated using MER and PPP and that the mix varies by commodity, as each commodity’s output in each country includes a mix of exports and domestically consumed goods, and the exports are usually valued higher per unit than domestically consumed goods. However, in the absence of physical unit data for thousands of commodities, this uncertainty is difficult to reduce.

Discussion This analysis adds to the growing literature showing the importance of considering trade in the pursuit of global environmental objectives (14, 16, 17, 31, 52). We estimate that strong import growth, relatively slow export growth, and shifting trade patterns has produced large and widening imbalances in the U.S.’ BEET. Several factors are likely underlying causes of these shifts, including large population size and growth, low savings rates, increased trade, and changes in U.S. economic structure. Analyses of EET are important for policy on at least two levels: equity concerns over local- and regional-scale pollution and global policy effectiveness concerns for globalscale pollution. Local-scale pollutants such as SO2 and NOx which are emitted in one country in order to produce goods for export to another country usually affect only people in the exporting country or surrounding area. A large body of literature has examined such equity concerns and issues of whether the positive impacts of increased trade outweigh potential negative environmental impacts (53, 54). Potentially more policy-relevant today is the issue of carbon leakage and international responsibility for climatechange-related emissions. The U.S. has come under increasing criticism and pressure to return to international climate change negotiations, and movement at state and local levels indicate that such a return may be likely within the next decade. This would benefit global climate policy, though this work and others (11, 16) show the importance of bringing all parties to the table. For instance, although the U.S. is responsible for producing 22% of estimated world fossil fuel CO2 emissions in 2004 (55), the amount the U.S. is responsible for consuming, based on numbers from Figure 1, is between 25 and 26% (where the consumed amount is the produced

amount + EEI less EEE). This shows the issue of carbon leakage on a grand scale: though the U.S.’ share of production-based CO2 emissions shrank between 1997 and 2004, the share of consumption-based CO2 emissions increased, due to increased trade volume and to shifting toward more carbon intensive trading partners. If these trends continue unaltered, and the U.S. continues toward a more servicebased economy (56), it is conceivable that EEI could exceed domestic production-related emissions within 20 years. Further, shifting to less CO2-intensive energy in the U.S. could be negated by increased trade with CO2-intensive economies. Furthermore, it should be noted that this analysis has, to some degree, neglected the energy and emissions associated with international freight transport of goods, and thus, the difference between produced and consumed CO2 emissions may be even larger. Although imports of transportation services (purchases by U.S. residents or businesses of foreignowned travel or freight transport) were modeled in the analysis, the majority of such emissions related to transportation service imports are due to personal purchases of international flights on non-U.S. carriers, and thus freight transport is likely undercounted. Further, IOA-based emissions factors derived for national transportation industries likely underestimate actual emissions in international transport for several reasons, including price differences and differences in technology. The authors plan to model these impacts more thoroughly, but previous work (57, 58) and data on international bunker fuel use (45) show that CO2 emissions due to international freight transport are unlikely to increase the totals in Figures 1 and 2 by more than 10%. However, vessel shipping does use high-sulfur fuel and is thus associated with rather large SO2 emissions. Wang et al. have estimated total SO2 emissions associated with ships entering and leaving North America to be 2.4 Mt of SO2 in 2002, which is nearly 40% of best-estimate EEI for that year (59). Despite the wealth of recent work (6, 11, 15), significant uncertainties remain in the estimation of energy and emissions embodied in trade. Given the importance of the issue, projects such as the Global Trade and Analysis Project (22) and its recently completed climate extension are important contributions. On the other hand, resolution is an important issue in such models, and using more disaggregated models is important as well; aggregation error and errors associated with uncertainties in trade data are likely significant for such projects (31). Future work should, where possible, focus on the construction of international physicalunit models to solve issues related to currency conversion and pricing.

Acknowledgments This work was funded by an EPA Science to Achieve Results Fellowship to C.L.W. and was partially funded by the National Science Foundation (NSF) MUSES grant 06-28232. The opinions expressed herein are those of the authors and not of the NSF. We gratefully acknowledge helpful discussion with Glen Peters, Eric Williams, and James Corbett.

Supporting Information Available Detailed discussion of model development and methods, emissions factors, and more figures and data. This material is available free of charge via the Internet at http:// pubs.acs.org.

Literature Cited (1) Bullard, C. W.; Herendeen, R. A. Energy Cost of Goods and Services. Energy Policy 1975, 3, 268-278. (2) Reinders, A.; Vringer, K.; Blok, K. The direct and indirect energy requirement of households in the European Union. Energy Policy 2003, 31, 139-153. VOL. 41, NO. 14, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4879

(3) Tukker, A.; Jansen, B. Environment impacts of products - A detailed review of studies. J. Ind. Ecol. 2006, 10, 159-182. (4) Lenzen, M.; Wier, M.; Cohen, C.; Hayami, H.; Pachauri, S.; Schaeffer, R. A comparative multivariate analysis of household energy requirements in Australia, Brazil, Denmark, India and Japan. Energy 2006, 31, 181-207. (5) Peters, G. P.; Hertwich, E. G. The Importance of Imports for Household Environmental Impacts. J. Ind. Ecol. 2006, 10, 89109. (6) Lenzen, M.; Lise-Lotte, P.; Munksgaard, J. CO2 Multipliers in Multi-region Input-Output Models. Econ. Syst. Res. 2004, 16, 391-412. (7) Streets, D.; Yu, C.; Bergin, M.; Wang, X.; Carmichael, G. Modeling Study of Air Pollution Due to the Manufacture of Export Goods in China’s Pearl River Delta. Environ. Sci. Technol. 2006, 40, 2099-2107. (8) IMF. World Economic Outlook 2005; International Monetary Fund; http://www.imf.org/external/pubs/ft/weo/2005/ (Accessed 1/18/06). (9) U.S. Census. U.S. Imports and Exports of Merchandise Monthly; DVD-ROM; U.S. Census Foreign Trade Division: Washington, DC, 2005. (10) Antweiler, W.; Copeland, B. R.; Taylor, M. S. Is Free Trade Good for the Environment? Am. Econ. Rev. 2001, 91, 877-908. (11) Ahmed, N.; Wyckoff, A. Carbon Dioxide Emissions Embodied in International Trade of Goods; OECD Working Papers: Paris, 2003. (12) Cole, M. A.; Elliott, J. R. FDI and the Capital Intensity of “Dirty” Sectors: A Missing Piece of the Pollution Haven Puzzle. Rev. Dev. Econ. 2005, 9, 530-548. (13) Babiker, M.; Reilly, J. M.; Jacoby, H. D. The Kyoto Protocol and developing countries. Energy Policy 2000, 28, 525-536. (14) Babiker, M. H. Climate Change Policy, Market Structure, and Carbon Leakage. J. Int. Econ. 2005, 65, 421-445. (15) Maenpaa, I.; Siikavirta, H. Greenhouse Gases Embodied in the International Trade and Final Consumption of Finland: An Input-Output Analysis. Energy Policy 2007, 35, 128-143. (16) Bin, S.; Harriss, R. The Role of CO2 Embodiment in US-China Trade. Energy Policy 2006, 34, 4063-4068. (17) Muradian, R.; O’Connor, M.; Martinez-Alier, J. Embodied Pollution in Trade: Estimating the ‘Environmental Load Displacement’ of Industrialised Countries. Ecol. Econ. 2002, 41, 51-67. (18) Leontief, W. Environmental Repercussions and the Economic Structure: An Input-Output Approach. Rev. Econ. Stat. 1970, 52, 262-271. (19) Hendrickson, C. T.; Lave, L. B.; Matthews, H. S. Environmental Life Cycle Assessment of Goods and Services: An Input-Output Approach, 1st ed.; RFF Press: Washington, DC, 2005. (20) Bureau of Economic Analysis. International Services: Crossborder Trade 1986-2005; U.S. BEA; http://www.bea.gov/bea/ di/intlserv.htm (Accessed July 2005). (21) Bureau of Economic Analysis. 1998-2004 Supplementary Make and Use Tables After Redefinitions at the Summary Level; U.S. BEA; http://www.bea.gov/bea/dn2/i-o_annual.htm (Accessed March 2006). (22) Dimaranan, B. V. Global Trade, Assistance, and Production: the GTAP 6 Database; Center for Global Trade Analysis, Purdue University; https://www.gtap.agecon.purdue.edu/databases/ v6/v6_doco.asp (Accessed July 2006). (23) Bjorn, A.; Declercq-Lopez, L.; Spatari, S.; MacLean, H. L. Decision support for sustainable development using a Canadian economic input-output life cycle assessment model. Can. J. Civil Eng. 2005, 32, 16-29. (24) Director General for Policy Planning. 2000 Input-Output Tables for Japan; Japan Ministry of Internal Affairs and Communications; http://www.stat.go.jp/English/data/io/index.htm (Accessed July 2006). (25) Bank of Korea. 2000 Bench Mark Input-Output Table; http:// ecos.bok.or.kr/EIndex_en.jsp (Accessed June 2006. (26) Statistical Offices of the European Communities. Eurostat; http:// ec.europa.eu/eurostat/ (Accessed July 2006). (27) Bureau of Economic Analysis. Benchmark Input-Output Accounts of the United States, 1997; Bureau of Economic Analysis: Washington, DC, 2002. (28) China National Statistical Bureau. 1997 Input-Output Tables; China Statistics Press: Beijing, 1999. (29) Bureau of Economic Analysis. 1997 Benchmark Input-Output Accounts; U.S. BEA; http://www.bea.doc.gov/bea/dn2/home/ annual_industry.htm (Accessed June 2005). 4880

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 14, 2007

(30) Statistics Canada. Concordance: NAICS 1997 - ISIC Rev. 3; http://www.statcan.ca/english/Subjects/Standard/concordances/ naics97-to-isicrev3.htm (Accessed 6/15/05). (31) Peters, G. P.; Hertwich, E. G. Opportunities and Challenges for Environmental MRIO Modeling: Illustrations with the GTAP Database. In 16th International Input-Output Conference; International Input-Output Association: Instanbul, 2007. (32) Peters, G. P.; Hertwich, E. G. Anthropogenic Carbon Flows in International Trade: Implications for Climate Change Policy. in preparation. (33) Miller, R. E.; Blair, P. D. Input-Output Analysis: Foundations and Extensions; Prentice Hall: Englewood Cliffs, NJ, 1985. (34) Peters, G. P.; Hertwich, E. G. Production Factors Embodied in Trade: Theoretical Development; Norwegian University of Science and Technology Working Papers: Trondheim, Norway, 2004. (35) Bureau of Economic Analysis. Import Matrix from the 1997 Benchmark Input-Output Accounts; U.S. BEA; http://www. bea.gov/bea/dn2/i-o_benchmark_1997.htm (Accessed March 2006). (36) Bureau of Economic Analysis. GDP by indusry Accounts: Gross Ouput by Industry in Current Dollars and Price Indexes by industry; U.S. BEA; http://www.bea.gov/bea/dn2/gdpbyind_ data.htm (Accessed September 2006). (37) Green Design Institute. Economic Input-Output Life Cycle Assessment (EIOLCA); Green Design Institute, Carneie Mellon University; http://www.eiolca.net (Accessed March 2006). (38) Lave, L. B.; Cobasflores, E.; Hendrickson, C. T.; McMichael, F. Using Input-Output Analysis to Estimate Economy-Wide Discharges. Environ. Sci. Technol. 1995, 29, A420-A426. (39) Peters, G. P.; Weber, C. L.; Liu, J. Construction of Chinese energy and emissions inventory; Norwegian University of Science and Technology Report: Trondheim, Norway, 2006; Available online at http://www.indecol.ntnu.no/indecolwebnew/publications/ reports/reports06.htm. (40) Harris, R. UK Atmospheric Emissions and Energy Use Accounts, 1990-1999; National Statistics UK; http://www.statistics.gov. uk/CCI/article.asp?ID)117&Pos)3&ColRank)2&Rank) 448 (Accessed July 2006). (41) Statistisches Bundesamt. Tabellen zu den Umweltokonomischen Gesamtrechnungen 2003; Statistisches Bundesamt: Berlin, 2003. (42) MacLean, H. L. Canadian Energy and Emissions data (unpublished work); University of Toronto: Toronto, 2006. (43) Environment Canada. National Pollutant Release Inventory; http://www.ec.gc.ca/pdb/npri/npri_home_e.cfm (Accessed August 2006). (44) Nansai, K.; Moriguchi, Y.; Tohno, S. Embodied Energy and Emission Intensity Data for Japan Using Input-Output Tables; Center for Global Environmental Research; http:// www-cger.nies.go.jp/cger-e/db/d031_e/3eid_e.html (Accessed June 2006). (45) International Energy Agency. CO2 Emissions from Fuel Combusion Database; http://data.iea.org/ieastore/statslisting (Accessed July 2006). (46) Olivier, J. G. J.; Berdowski, J. J. Emissions Database for Global Atmospheric Research (EDGAR) V3.2; http://www.mnp.nl/edgar/ (Accessed 4/9/07). (47) Eastern Research Group. Mexico National Emissions Inventory; National Insitute of Ecology of Mexico: Sacramento, CA, 2005. (48) Marland, G.; Boden, T. A.; Andres, R. J. Global, Regional, and National Annual CO2 Emissions from Fossil-Fuel Burning, Cement Manufacture, and Gas Flaring: 1751-2003; Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory: Oak Ridge, TN; http://cdiac.esd.ornl.gov/ftp/ ndp030/ (Accessed 6/20/06). (49) Heston, A.; Summers, R.; Aten, B. Penn World Table 6.2; Center for International Comparisons of Production, Income, and Prices, University of Pennsylvania; http://pwt.econ.upenn.edu/ php_site/pwt_index.php (Accessed September 2006). (50) Lenzen, M. Errors in Conventional and Input-Output based Life-Cycle Inventories. J. Ind. Ecol. 2000, 4, 127-148. (51) Suh, S.; Lenzen, M.; Treloar, G. J.; Hondo, H.; Horvath, A.; Huppes, G.; Jolliet, O.; Klann, U.; Krewitt, W.; Moriguchi, Y.; Munksgaard, J.; Norris, G. System boundary selection in lifecycle inventories using hybrid approaches. Environ. Sci. Technol. 2004, 38, 657-664. (52) Munksgaard, J.; Pedersen, K. A. CO2 Accounts for Open Economies: Producer or Consumer Responsibility? Energy Policy 2001, 29, 327-334.

(53) Muradian, R.; Martinez-Alier, J. Trade and the environment: from a ‘Southern’ perspective. Ecol. Econ. 2001, 36, 281-297. (54) Giljum, S. G.; Eisenmenger, N. North-South Trade and the Distribution of Environmental Goods and Burdens: a Biophysical Perspective. J. Environ. Dev. 2004, 13, 73-100. (55) U.S. Energy Information Agency. International Energy Annual 2004; http://www.eia.doe.gov/iea/carbon.html (Accessed September 2006). (56) Suh, S. Are Services Better for Climate Change? Environ. Sci. Technol. 2006, 40, 6555-6560. (57) Corbett, J. J.; Koehler, H. W. Updated Emissions from Ocean Shipping. J. Geophys. Res. 2003, 108, 1-13.

(58) Munksgaard, J.; Lenzen, M.; Jensen, T. C.; Pade, L. L. Transport Energy Embodied in Consumer Goods: a Hybrid Life-Cycle Analysis. Energy Environ. 2005, 16, 283-301. (59) Wang, C.; Corbett, J. J.; Firestone, J. Modeling Energy Use and Emissions from North American Shipping: Application of the Ship Traffic, Energy, and Environment Model. Environ. Sci. Technol. 2007, 41 (9), 3226-3232.

Received for review December 7, 2006. Revised manuscript received April 27, 2007. Accepted May 10, 2007. ES0629110

VOL. 41, NO. 14, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4881