Factors affecting carbon dioxide emissions embodied in trade

Kang, Jong Woo; Gapay, Joshua Anthony Working Paper Factors affecting carbon dioxide emissions embodied in trade ADB Economics Working Paper Series, No. 700 Provided in Cooperation with: Asian Development Bank (ADB), Manila Suggested Citation: Kang, Jong Woo; Gapay, Joshua Anthony (2023) : Factors affecting carbon dioxide emissions embodied in trade, ADB Economics Working Paper Series, No. 700, Asian Development Bank (ADB), Manila, https://doi.org/10.22617/WPS230479-2 This Version is available at: https://hdl.handle.net/10419/298146 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/3.0/igo/ ASIAN DEVELOPMENT BANK ASIAN DEVELOPMENT BANK 6 ADB Avenue, Mandaluyong City 1550 Metro Manila, Philippines www.adb.org FACTORS AFFECTING CARBON DIOXIDE EMISSIONS EMBODIED IN TRADE Jong Woo Kang and Joshua Anthony Gapay ADB ECONOMICS WORKING PAPER SERIES NO. 700 October 2023 Factors Affecting Carbon Dioxide Emissions Embodied in Trade This paper examines the impact of environmental regulation in exporter and importer economies on crossborder carbon flows. While stricter environmental regulations help reduce carbon dioxide (CO2) emissions from domestic production, leading to lower CO2 emissions embodied in exports, stricter regulations on the importing side lead to higher CO2 emissions embodied in imports. More importantly, stricter environmental regulations could encourage further outsourcing of intermediate inputs by exporters, prompting carbon leakages in the upstream segment of global value chains. About the Asian Development Bank ADB is committed to achieving a prosperous, inclusive, resilient, and sustainable Asia and the Pacific, while sustaining its efforts to eradicate extreme poverty. Established in 1966, it is owned by 68 members —49 from the region. Its main instruments for helping its developing member countries are policy dialogue, loans, equity investments, guarantees, grants, and technical assistance. ASIAN DEVELOPMENT BANK The ADB Economics Working Paper Series presents research in progress to elicit comments and encourage debate on development issues in Asia and the Pacific. The views expressed are those of the authors and do not necessarily reflect the views and policies of ADB or its Board of Governors or the governments they represent. ADB Economics Working Paper Series Factors Affecting Carbon Dioxide Emissions Embodied in Trade Jong Woo Kang and Joshua Anthony Gapay No. 700 | October 2023 Jong Woo Kang ([email protected]) is the director and Joshua Anthony Gapay (jgapay.consultant@adb. org) is a consultant in the Regional Cooperation and Integration Division, Economic Research and Development Impact Department, Asian Development Bank. Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) © 2023 Asian Development Bank 6 ADB Avenue, Mandaluyong City, 1550 Metro Manila, Philippines Tel +63 2 8632 4444; Fax +63 2 8636 2444 www.adb.org Some rights reserved. Published in 2023. ISSN 2313-6537 (print), 2313-6545 (electronic) Publication Stock No. WPS230479-2 DOI: http://dx.doi.org/10.22617/WPS230479-2 The views expressed in this publication are those of the authors and do not necessarily reflect the views and policies ofthe Asian Development Bank (ADB) or its Board of Governors or the governments they represent. ADB does not guarantee the accuracy of the data included in this publication and accepts no responsibility for any consequence of their use. The mention of specific companies or products of manufacturers does not imply that they are endorsed or recommended by ADB in preference to others of a similar nature that are not mentioned. By making any designation of or reference to a particular territory or geographic area, or by using the term “country” inthis publication, ADB does not intend to make any judgments as to the legal or other status of any territory or area. This publication is available under the Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) https://creativecommons.org/licenses/by/3.0/igo/. By using the content of this publication, you agree to be bound bytheterms of this license. For attribution, translations, adaptations, and permissions, please read the provisions andterms of use at https://www.adb.org/terms-use#openaccess. This CC license does not apply to non-ADB copyright materials in this publication. If the material is attributed toanother source, please contact the copyright owner or publisher of that source for permission to reproduce it. ADB cannot be held liable for any claims that arise as a result of your use of the material. Please contact [email protected] if you have questions or comments with respect to content, or if you wish toobtain copyright permission for your intended use that does not fall within these terms, or for permission to use theADB logo. Corrigenda to ADB publications may be found at http://www.adb.org/publications/corrigenda. Note: In this publication, “$” refers to United States dollars. ABSTRACT Trade encourages economic expansion and improves welfare based on international division of labor. However, trade also has an environmental footprint, particularly in the form of carbon dioxide (CO2) and other emissions. This paper examines the impact of environmental regulation in exporter and importer economies on cross-border carbon flows. It uses pooled estimation, random effects, fixed effects, fixed effects with instrumental variables, and Poisson pseudomaximum likelihood models to estimate the effect of more stringent environmental regulation, while controlling for scale, technique, and composition effects associated with CO2 emissions. While stricter environmental regulations help reduce CO2 emissions from domestic production, leading to lower CO2 emissions embodied in exports, stricter regulations on the importing side lead to higher CO2 emissions embodied in imports. More importantly, stricter environmental regulations could encourage further outsourcing of intermediate inputs by exporters, prompting carbon leakages in the upstream segment of global value chains. Keywords: CO2 emissions, carbon leakage, trade, global value chain JEL codes: F1, F14, F18 1. Introduction It is well recognized that trade promotes economic growth and helps increase welfare through job opportunities and income generation. International trade, however, entails environmental footprints, in particular due to the carbon dioxide (CO2) and other harmful emissions associated with the production of goods and services that are traded both within and across borders. A growing number of economies have recognized this dilemma and strengthened efforts to reduce the environmental degradation caused by trade. From 2000 to 2018, the CO2 emissions intensity of exports and imports declined in Asia, the European Union + United Kingdom (EU+UK), and North America. However, Asia still has the highest CO2 emissions intensity embodied in exports and imports, which leaves ample room for improvement. An inverse relationship between environmental regulations and CO2 emissions embodied in exports and imports suggests that environmental policies have been effective in reducing environmental degradation from trade. The effect, however, is not straightforward and could have rather undesirable side effects by propagating carbon leakage: firms relocating production capacity from an economy with stringent environmental regulations to economies with laxer policies to capitalize on regulatory arbitrage. The positive relationship between environmental policies and CO2 emissions embodied in imports and the negative relationship with respect to CO2 emissions embodied in exports suggest that carbon leakage may be occurring. North America and EU+UK, with regulations more stringent than Asia, demonstrate higher CO2 emissions embodied in their imports compared to exports, while Asia displays higher CO2 emissions embodied in exports, which implies carbon leakages could be flowing from advanced to developing economies. This paper aims to investigate the impact of environmental regulation on cross-border carbon flows. This is crucially related to the potential for carbon leakage through the impact of environmental regulations of both exporters and importers on CO2 emissions embodied in trade. Previous literature has yielded mixed evidence in describing the effect of regulation on carbon leakage. This paper builds on the theoretical model of Copeland and Taylor (1994). It uses pooled estimation, random effects, fixed effects, fixed effects with instrumental variables, and Poisson pseudo-maximum likelihood models to estimate the effect of an environmental policy on CO2 emissions embodied in exports, while controlling for scale, technique, and composition effects associated with CO2 emissions. Based on Fixed Effect and Poisson pseudo-maximum likelihood (PPML) regressions, results show that an increase in a partner (importer) economy’s environmental policy stringency is associated with a reporter (exporter) economy’s increase in domestic, foreign, and overall CO2 emissions embodied in gross exports. This implies that importers with stricter environmental regulations might be relegating the production capacity of dirtier industries to exporters so as to import such products from their trade counterparts. Moreover, an increase in the reporter economy’s environmental policy stringency is associated with a decrease in its domestic and overall CO2 embodied in its gross exports and CO2 embodied in its final goods exports. This supports the argument that stricter environmental regulations for different sectors help to green production procedures. It is notable that the reporter’s foreign CO2 embodied in gross exports increases as its environmental policy becomes more stringent. This result suggests potential carbon leakage at 2 upstream production segments and in downstream functions across borders because economies are inherently incentivized to outsource unclean production in both ways—upstream for exporters and downstream for importers—in the face of environmental regulations. This points to crucial drawbacks in the Carbon Border Adjustment Mechanism being contemplated by the European Union and other advanced economies in influencing exporters’ production patterns through regulatory disciplines should the focus be on the final outcome of carbon leakage embodied in traded goods and services, without first considering the original source of emission productions embodied in the foreign value-added components of final exports. Likewise, the regulatory authorities need to consider the potential carbon leakages at the upstream of value chains in case the carbon intensity of intermediate products are not taken into account with environmental regulations. A discussion and review of related literature on the relationship between trade, environment, and environmental policies is presented in Section 2, while Sections 3 and 4 present data and variables, and the paper’s empirical strategy in testing its hypothesis. The results of the empirical analysis are discussed in Section 5, while Section 6 concludes the paper. 2. International Trade and the Environment A. Background International trade brings welfare improvements by supporting specialization of cross-border production and plugging gaps between domestic production and gross consumption at the economy level. However, these economic activities require, to varying degrees, significant fuel consumption and pursuant CO2 emissions, which aggravate climate change and its adverse impacts. As an economy grows, it tends to emit more emissions because economic activities expand. Progress in economic growth and trade tends to increase carbon emissions (scale factor). From 2000 to 2018, Asia’s imports and exports quadrupled based on the TECO2 data from the Organisation for Economic Co-operation and Development (OECD), while its carbon emissions embodied in trade only doubled. Meanwhile, although trade of EU+UK and North America doubled during the same period, their carbon emissions in trade stayed at the same level (Figure 1, panels a and b). Technical development could promote low carbon emissions for a given level of production (technology factor). The trend was of a decline in CO2 emissions intensity (measured as tonnes of CO2 per million US dollars of exports and imports) across all regions from 2000 to 2018. However, Asia still recorded higher CO2 emissions intensity than EU+UK and North America in both exports and imports (Figure 1, panels c and d). This indicates although the region has progressed significantly in limiting CO2 emissions from economic activities, it still has a lot of catching up to do with other parts of the world. 3 Figure 1: Carbon Emissions and International Trade over Time (a) Exports versus Carbon Emissions Embodied in Exports (2000 = 100) (b) Imports versus Carbon Emissions Embodied in Imports (2000 = 100) (c) Tonnes of CO2 per $ million of Exports (d) Tonnes of CO2 per $ million of Imports CO 2 = carbon emissions, EU = European Union, EXGR_T = gross exports ($), EXGR_TCO2 = Tonnes of carbon emissions embodied in gross exports, EXGR_TCO2int = Tonnes of carbon emissions per $ million exports, IMGR_T = gross imports ($), IMGR_TCO2 = Tonnes of carbon emissions embodied in gross imports, IMGR_TCO2int = Tonnes of carbon emissions per $ million imports, UK = United Kingdom. Source: Authors’ estimates. The OECD’s TECO2 data covers 68 industry groups with varying aggregation levels. A total of 14 non-overlapping industry groups were picked to cover all industries when aggregated. Carbon intensity of production is estimated according to these industrial groupings. This estimation suggests the most carbon-intensive traded products are in industries such as (1) electricity, gas, steam and air conditioning and water supply; (2) basic metals and fabricated metal products; (3) Furniture, repair and installation, other manufacturing; (4) chemicals and nonmetallic mineral products. The least carbon-intensive traded products are in (1) construction; (2) 0 50 100 150 200 250 300 350 400 450 Asia EXGR_T Asia EXGR_TCO2 EU+UK EXGR_T EU+UK EXGR_TCO2 North America EXGR_T North America EXGR_TCO2 0 50 100 150 200 250 300 350 400 450 Asia IMGR_T Asia IMGR_TCO2 EU+UK IMGR_T EU+UK IMGR_TCO2 North America IMGR_T North America IMGR_TCO2 - 200 400 600 800 1,000 1,200 1,400 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 Asia EXGR_TCO2int EU+UK EXGR_TCO2int North America EXGR_TCO2int - 200 400 600 800 1,000 1,200 1,400 Asia EXGR_TCO2int EU+UK EXGR_TCO2int North America EXGR_TCO2int 4 food products, beverages and tobacco; (3) total services; (4) transport equipment; and (5) agriculture, hunting, forestry and fishing (Figure 2, panels a and c). Overall, the high CO2 intensity in Asian exports and imports might in part be due to the high shares of traded products coming from carbon intensive industries (the composition factor). In 2018, the share of carbon-intensive exports in Asia was 62%, while it was 40.2% in EU+UK and 37.3% in North America. Meanwhile, the share of carbon-intensive imports in Asia, at 58.4%, which is also higher than the shares of EU+UK and North America (Figure 2, panels b and d). Figure 2: CO2 Emissions Intensity per Industries and Trade Shares per Region (a) CO 2 Intensity per Industry, Exports (2018) (b) Industry Shares in Exports, 2018 (%) (c) CO 2 Intensity per Industry, Imports (2018) (b) Industry Shares in Imports, 2018 (%) n.e.c. = not elsewhere classified, RoW = rest of the world. Source: Authors’ estimates. 0500 1000 1500 2000 2500 Construction Food products, beverages and tobacco Total services Transport equipment Agriculture, hunting, forestry and fishing Wood and paper products and printing Machinery and equipment, n.e.c. Textiles, textile products, leather and footwear Computer, electronic and optical products;… Mining and quarrying Chemicals and non-metallic mineral products Furniture; other manufacturing; repair and… Basic metals and fabricated metal products Electricity, gas, steam and air conditioning… All industries Less carbonintensive More carbonintensive Most carbonintensive Tonnes CO2 emissions per $ million exports 38.0 59.8 62.7 49.0 38.4 17.3 18.4 28.2 23.6 22.9 18.9 22.7 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0 Asia EU+UK North America RoW Most carbon-intensive industries More carbon-intensive industries Less carbon-intensive industries 0500 1000 1500 2000 2500 Construction Total services Transport equipment Food products, beverages and tobacco Agriculture, hunting, forestry and fishing Machinery and equipment, n.e.c. Mining and quarrying Wood and paper products and printing Textiles, textile products, leather and footwear Computer, electronic and optical products;… Chemicals and non-metallic mineral products Furniture; other manufacturing; repair and… Basic metals and fabricated metal products Electricity, gas, steam and air conditioning and… All industries Less carbonintensive More carbonintensive Most carbonintensive Tonnes CO2 emissions per $ million imports 41.6 58.8 47.1 51.1 37.7 21.5 30.4 23.4 20.7 19.7 22.5 25.5 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0 Asia EU+UK North America RoW Most carbon-intensive industries More carbon-intensive industries Less carbon-intensive industries 11 4. Identification Strategy Appendix A.1 summarizes papers that empirically analyzed the effects of environmental regulations on trade and CO2 embodied in trade. Papers mostly used imports or exports (and their corresponding CO2 emissions) as dependent variables. Others utilized the location decision of firms as dependent variable. Various models, both parametric and nonparametric, were employed to identify the determinants of CO2 emissions and examine potential carbon leakages. This paper follows the approaches of Ederington, Paraschiv, and Zanardi (2018), Assogbavi and Dees (2021), and Misch and Wingender (2021). It utilizes CO2 emissions embodied in exports as a dependent variable and runs various models such as ordinary least squares (OLS), fixed effects, random effects, fixed effect with instrumental variables, and PPML models to determine the effects on CO2 emissions of environmental policy and of scale, composition, and technique effects. Time-invariant fixed effects such as variables for distance, common language, and contiguity are added in the analysis, similar to the gravity model approach of Assogbavi and Dees (2021). (1) Pooled OLS Model Estimation Following Assogbavi and Dees (2021), this paper estimates the model initially using pooled OLS. One key new feature is the attempt to explicitly control scale, composition, and technique effects, which are believed to affect carbon footprints of production and the international trade thereof based on theories and literature. ln 𝐶𝐶𝐶𝐶2 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 =𝛽𝛽 0 +𝛽𝛽 1 ln 𝐺𝐺𝐺𝐺𝐺𝐺 𝑖𝑖𝑖𝑖 +𝛽𝛽 2 ln 𝐺𝐺𝐺𝐺𝐺𝐺 𝑖𝑖𝑖𝑖 +𝛽𝛽 3 ln 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 𝑖𝑖𝑖𝑖 +𝛽𝛽 4 ln 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 𝑖𝑖𝑖𝑖 +𝛽𝛽 5 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 𝑖𝑖𝑖𝑖 +𝛽𝛽6𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖+𝜌𝜌1𝑠𝑠𝑠𝑠𝑐𝑐𝑖𝑖𝑖𝑖+𝜌𝜌2𝑠𝑠𝑠𝑠𝑐𝑐𝑖𝑖𝑖𝑖+𝜌𝜌3𝑝𝑝𝑠𝑠𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖+𝜇𝜇1ln 𝑑𝑑𝑑𝑑𝑠𝑠𝑠𝑠𝑖𝑖𝑖𝑖 +𝜇𝜇2𝑐𝑐𝑐𝑐𝑐𝑐𝑠𝑠𝑖𝑖𝑖𝑖 +𝜇𝜇3𝑐𝑐𝑐𝑐𝑐𝑐𝑔𝑔𝑖𝑖𝑖𝑖 +𝜇𝜇4𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖 +𝜖𝜖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 (4.1) Where 𝐶𝐶𝐶𝐶2𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 can either be: total, domestic, or foreign CO2 emissions embodied in gross exports; or CO2 emissions embodied in intermediate or final goods exports depending upon model specifications. 𝐺𝐺𝐺𝐺𝐺𝐺 is the gross domestic product of an economy, 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 is CO2 emissions embodied in production per capita, 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 is the share of the top cleanest industries to total production, 𝑠𝑠𝑠𝑠𝑐𝑐 is environmental policy stringency index, and 𝑝𝑝𝑠𝑠𝑐𝑐 is the presence of preferential trade agreement between reporter and partner economies. Time invariant variables are also included: 𝑑𝑑𝑑𝑑𝑠𝑠𝑠𝑠 is the distance between reporter and partner, 𝑐𝑐𝑐𝑐𝑐𝑐𝑠𝑠 identifies if reporter and partner share a border, 𝑐𝑐𝑐𝑐𝑐𝑐𝑔𝑔 identifies if reporter and partner share a common language, and 𝑐𝑐𝑐𝑐𝑐𝑐 identifies if reporter and partner have colonial ties. (2) Random Effects Model Estimation Random effects model assumes that there is unobserved heterogeneity across economy pairs and sector captured by 𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖. ln 𝐶𝐶𝐶𝐶2𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 =𝛽𝛽0+𝛽𝛽1ln 𝐺𝐺𝐺𝐺𝐺𝐺𝑖𝑖𝑖𝑖+𝛽𝛽2ln 𝐺𝐺𝐺𝐺𝐺𝐺𝑖𝑖𝑖𝑖+𝛽𝛽3ln 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖+𝛽𝛽4ln 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖 +𝛽𝛽5𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖 +𝛽𝛽6𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖+𝜌𝜌1𝑠𝑠𝑠𝑠𝑐𝑐𝑖𝑖𝑖𝑖+𝜌𝜌2𝑠𝑠𝑠𝑠𝑐𝑐𝑖𝑖𝑖𝑖 +𝜌𝜌3𝑝𝑝𝑠𝑠𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖+𝜇𝜇1ln 𝑑𝑑𝑑𝑑𝑠𝑠𝑠𝑠𝑖𝑖𝑖𝑖 +𝜇𝜇2𝑐𝑐𝑐𝑐𝑐𝑐𝑠𝑠𝑖𝑖𝑖𝑖+𝜇𝜇3𝑐𝑐𝑐𝑐𝑐𝑐𝑔𝑔𝑖𝑖𝑖𝑖 +𝜇𝜇4𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖 +𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖 +𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 (4.2) 12 (3) Fixed Effects Model Estimation The fixed effects model is less restricted than the random effects model as it allows the economypair-sector-specific effects 𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖 to be correlated with the regressors. ln 𝐶𝐶𝐶𝐶2𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 =𝛽𝛽0+𝛽𝛽1ln 𝐺𝐺𝐺𝐺𝐺𝐺𝑖𝑖𝑖𝑖+𝛽𝛽2ln 𝐺𝐺𝐺𝐺𝐺𝐺𝑖𝑖𝑖𝑖+𝛽𝛽3ln 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖+𝛽𝛽4ln 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖 +𝛽𝛽5𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖 +𝛽𝛽6𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖 +𝜌𝜌1𝑠𝑠𝑠𝑠𝑐𝑐𝑖𝑖𝑖𝑖+𝜌𝜌2𝑠𝑠𝑠𝑠𝑐𝑐𝑖𝑖𝑖𝑖+𝜌𝜌3𝑝𝑝𝑠𝑠𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖 +𝜇𝜇1ln 𝑑𝑑𝑑𝑑𝑠𝑠𝑠𝑠𝑖𝑖𝑖𝑖 +𝜇𝜇2𝑐𝑐𝑐𝑐𝑐𝑐𝑠𝑠𝑖𝑖𝑖𝑖 +𝜇𝜇3𝑐𝑐𝑐𝑐𝑐𝑐𝑔𝑔𝑖𝑖𝑖𝑖 +𝜇𝜇4𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖 +∑ ∑ ∑ 𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑖𝑖𝑖𝑖𝑖𝑖 𝑆𝑆 𝑖𝑖 𝐽𝐽 𝑖𝑖 𝐼𝐼𝑖𝑖+𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 (4.3) (4) Fixed Effects Model Estimation with Instrumental Variables To minimize the potential endogeneity with the variables of industrial composition, 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐, and policy stringency, 𝑠𝑠𝑠𝑠𝑐𝑐, we use R&D share to GDP, 𝑐𝑐𝑐𝑐𝑠𝑠, and number of international environmental agreements, 𝑑𝑑𝑐𝑐𝑠𝑠𝑐𝑐𝑐𝑐 as instruments respectively. R&D may not directly affect the CO2 emissions embodied in exports while it is correlated with industrial structure of an economy and international environmental agreements may have a direct impact on environmental policy stringency while it may not directly affect CO2 emissions embodied in exports. The validity of these instruments are described through various statistics in the Appendix. IV test results show rejection of null hypothesis that the independent variables, “clean” and “str” are exogenous. Moreover, the null hypothesis that the instrument variables are under-identified and weakly identified were rejected in all estimates. The first stage of this model can be written as: 𝑑𝑑𝑐𝑐𝑠𝑠𝑠𝑠𝑐𝑐𝑑𝑑=𝜂𝜂0+𝜂𝜂1𝑐𝑐𝑐𝑐𝑠𝑠𝑖𝑖𝑖𝑖 +𝜂𝜂2𝑐𝑐𝑐𝑐𝑠𝑠𝑖𝑖𝑖𝑖 +𝜂𝜂3𝑑𝑑𝑐𝑐𝑠𝑠𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖 +𝜂𝜂4𝑑𝑑𝑐𝑐𝑠𝑠𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖+𝛾𝛾1ln 𝐺𝐺𝐺𝐺𝐺𝐺𝑖𝑖𝑖𝑖 +𝛾𝛾2ln 𝐺𝐺𝐺𝐺𝐺𝐺𝑖𝑖𝑖𝑖+𝛾𝛾3ln 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖+𝛾𝛾4ln 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖 +𝛾𝛾5𝑝𝑝𝑠𝑠𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖 +𝛾𝛾6𝑑𝑑𝑑𝑑𝑠𝑠𝑠𝑠𝑖𝑖𝑖𝑖 +𝛾𝛾7𝑐𝑐𝑐𝑐𝑐𝑐𝑠𝑠𝑖𝑖𝑖𝑖 +𝛾𝛾8𝑐𝑐𝑐𝑐𝑐𝑐𝑔𝑔𝑖𝑖𝑖𝑖 +𝛾𝛾9𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖 +∑ ∑ ∑ 𝛾𝛾𝑖𝑖𝑖𝑖𝑖𝑖𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑖𝑖𝑖𝑖𝑖𝑖 𝑆𝑆 𝑖𝑖 𝐽𝐽 𝑖𝑖 𝐼𝐼𝑖𝑖+𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖 (4.4) Where the vector 𝑑𝑑𝑐𝑐𝑠𝑠𝑠𝑠𝑐𝑐𝑑𝑑′=�𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖 𝑠𝑠𝑠𝑠𝑐𝑐𝑖𝑖𝑖𝑖 𝑠𝑠𝑠𝑠𝑐𝑐𝑖𝑖𝑖𝑖�. We then conduct the second stage using the estimates 𝚤𝚤𝑐𝑐𝑠𝑠𝑠𝑠𝑐𝑐𝑑𝑑 �′=�𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 �𝑖𝑖𝑖𝑖 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 �𝑖𝑖𝑖𝑖 𝑠𝑠𝑠𝑠𝑐𝑐 �𝑖𝑖𝑖𝑖 𝑠𝑠𝑠𝑠𝑐𝑐 �𝑖𝑖𝑖𝑖�. ln 𝐶𝐶𝐶𝐶2𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 =𝛽𝛽0+𝛽𝛽1ln 𝐺𝐺𝐺𝐺𝐺𝐺𝑖𝑖𝑖𝑖+𝛽𝛽2ln 𝐺𝐺𝐺𝐺𝐺𝐺𝑖𝑖𝑖𝑖+𝛽𝛽3ln 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖+𝛽𝛽4ln 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖 +𝛽𝛽5𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 �𝑖𝑖𝑖𝑖+𝛽𝛽6𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 �𝑖𝑖𝑖𝑖+𝜌𝜌1𝑠𝑠𝑠𝑠𝑐𝑐 �𝑖𝑖𝑖𝑖+𝜌𝜌2𝑠𝑠𝑠𝑠𝑐𝑐 �𝑖𝑖𝑖𝑖+𝜌𝜌3𝑝𝑝𝑠𝑠𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖+𝜇𝜇1ln 𝑑𝑑𝑑𝑑𝑠𝑠𝑠𝑠𝑖𝑖𝑖𝑖 +𝜇𝜇2𝑐𝑐𝑐𝑐𝑐𝑐𝑠𝑠𝑖𝑖𝑖𝑖+𝜇𝜇3𝑐𝑐𝑐𝑐𝑐𝑐𝑔𝑔𝑖𝑖𝑖𝑖 +𝜇𝜇4𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖 +∑ ∑ ∑ 𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑖𝑖𝑖𝑖𝑖𝑖 𝑆𝑆 𝑖𝑖 𝐽𝐽 𝑖𝑖 𝐼𝐼𝑖𝑖+𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 (4.5) (5) PPML Model Estimation Poisson pseudo-maximum likelihood (PPML) estimation allows observations with zero values. This paper uses the PPML regression with multiple high-dimensional fixed effects (HDFE) by Correia et al. (2020). 𝐶𝐶𝐶𝐶2𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 =exp ⎝ ⎜ ⎛ 𝛽𝛽1𝐺𝐺𝐺𝐺𝐺𝐺𝑖𝑖𝑖𝑖 +𝛽𝛽2𝐺𝐺𝐺𝐺𝐺𝐺𝑖𝑖𝑖𝑖 +𝛽𝛽3𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖+𝛽𝛽4𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖 +𝛽𝛽5𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖 +𝛽𝛽6𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖+𝜌𝜌1𝑠𝑠𝑠𝑠𝑐𝑐𝑖𝑖𝑖𝑖+𝜌𝜌2𝑠𝑠𝑠𝑠𝑐𝑐𝑖𝑖𝑖𝑖 +𝜌𝜌3𝑝𝑝𝑠𝑠𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖 +𝜇𝜇1𝑑𝑑𝑑𝑑𝑠𝑠𝑠𝑠𝑖𝑖𝑖𝑖 +𝜇𝜇2𝑐𝑐𝑐𝑐𝑐𝑐𝑠𝑠𝑖𝑖𝑖𝑖 +𝜇𝜇3𝑐𝑐𝑐𝑐𝑐𝑐𝑔𝑔𝑖𝑖𝑖𝑖 +𝜇𝜇4𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖 +𝜃𝜃𝑖𝑖+𝜃𝜃𝑖𝑖+𝜃𝜃𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 ⎠ ⎟ ⎞ +𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 (4.6) 13 Where 𝜃𝜃𝑖𝑖 denotes sector fixed effects, 𝜃𝜃𝑖𝑖 denotes time fixed effects, and 𝜃𝜃𝑖𝑖𝑖𝑖𝑖𝑖 denotes economypair-sector fixed effects. 5. Results Results show that FE and PPML regressions are consistent across the board with the impact of reporters’ environmental policies exerting negative impact on CO2 emissions embodied in gross exports and a generally positive impact of partners’ policies on CO2 emissions embodied in gross imports at economy bilateral, sectoral level. The former suggests that the stricter the environmental regulations, the fewer CO2 emissions are embodied in production, which leads to equally less dirty exports at the bilateral sectoral level. The latter could imply the existence of carbon leakage by letting others produce carbon intensive products and import those under stricter environmental regulations. The results on the reporter’s policy impact were further investigated by using domestic and foreign CO2 emissions embodied in gross exports. The model for foreign CO2 emissions also shows a positive sign from PPML while the domestic CO2 emissions model has the same results as that of the total CO2 emissions model. It is likely that that the negative impact on total CO2 emissions embodied in gross exports in FE and part of PPML models stems from the regulatory impact of domestic CO2 production, and at the same time economies might be simply bypassing foreign produced CO2 emissions to others through exports. A. CO2 Emissions Embodied in Total Exports Economic size works in the direction of raising CO2 emissions embodied in both exports and imports, corroborating the conventional theory that the larger the economy, the greater CO2 emissions produced, that are then embodied in both exports and imports. Technique effect points to a positive impact of CO2 emissions on production per capita for both exporters and importers, except in the FE model with instruments for the importers. This implies that less carbon efficient production procedures lead to greater CO2 emissions embodied in exports and imports. Composition effect also largely points to the expected direction except for the FE models. Looking at the policy variables when using a pooled model, which has the most restrictive assumptions, the higher stringency index entails lower CO2 emissions embodied in trade for both the reporter and partner. However, using the random effects estimator, the coefficients’ magnitude decreases while partners’ policy stringency now entails higher CO2 emissions embodied in imports. Moreover, using the fixed effects model after rejecting the null hypothesis in the Hausman test3, both reporter and partners’ stringency index gained positive coefficients. The signs of the coefficients in the fixed effects model change depending on the instrument variables used. On one hand, if the stringency of environmental regulations for both reporter and partner were instrumented using the number of international environmental agreements, both reporter and 3 Hausman Test determines whether random effects or fixed effects estimation fits the model better. Meanwhile Breusch-Pagan Lagrange Multiplier Test checks the significance of random effects in the panel data model. 14 partner coefficients become negative. On the other hand, when all instruments mentioned are used simultaneously, the reporter’s policy stringency coefficient turns negative while partner’s stays positive. Now using the PPML model, the reporter’s increase in policy stringency would be associated with decreasing CO2 emissions embodied in its total exports. Meanwhile, the partner’s increase in policy stringency would be associated with increasing CO2 emissions embodied in total imports at bilateral economy, sectoral level. Results show that a 1-percentage-point increase in the reporter’s stringency of environmental regulation is associated with 4% decrease in CO2 emissions embodied in its total exports. And then a 1-percentage-point increase in partner’s stringency of environmental regulation is associated with 15% increase in CO2 emissions embodied in its total imports.4 Adding sector and year fixed effects in the PPML does not significantly change the results (Table 2). 4 Using PPML estimation, the effect on the dependent variable is (EXP(coefficient) – 1) × 100. 15 Table 2: CO2 Emissions Embodied in Total Exports Dependent Variable: Natural log of CO2 emissions embodied in Total Exports except for PPML model Pooled OLS RE Panel Regression FE Panel Regression with economypair-sector FE Fixed Effects IV Regression PPML Regression Instrumented: Stringency of Environmental regulation Instrumented: Industrial composition Instrumented: Both No fixed effects with fixed effects: sector, year, sector-year with fixed effects: sector, year, sectoryear, economy-pairsector Independent Variables: (1) (2) (3) (4) (5) (6) (7) (8) (9) Scale Effect Natural log of reporter economy's real GDP (base year = 2015) 0.576*** 0.408*** -0.225*** -0.200*** -0.736*** -0.763*** (0.00270) (0.00632) (0.0152) (0.0193) (0.0688) (0.0815) Reporter's Real GDP (base year = 2015) 0.000175*** 0.000177*** 1.96e-05*** (2.55e-06) (2.43e-06) (6.05e-06) Natural log of partner economy's real GDP (base year = 2015) 0.566*** 0.506*** 0.803*** 0.776*** 1.414*** 1.452*** (0.00266) (0.00621) (0.0152) (0.0182) (0.0749) (0.0814) Partner's Real GDP (base year = 2015) 0.000189*** 0.000191*** 8.59e-05*** (2.84e-06) (2.66e-06) (6.00e-06) Technique Effect Natural log of reporter economy's CO2 embodied in production per capita 0.219*** 0.374*** 0.770*** 0.891*** 1.282*** 1.553*** (0.00645) (0.00949) (0.0131) (0.0213) (0.0717) (0.103) Reporter's CO 2 embodied in production per capita 0.0507*** 0.0491*** 0.0577*** (0.00327) (0.00319) (0.00422) Natural log of partner economy's CO2 embodied in production per capita 0.0284*** 0.223*** 0.232*** 0.280*** -0.210*** -0.478*** (0.00626) (0.00922) (0.0126) (0.0214) (0.0703) (0.0912) Partner's CO 2 embodied in production per capita 0.0106*** 0.00898*** 0.0403*** (0.00339) (0.00320) (0.00459) Composition Effect Reporter's share of top five cleanest industries to total production -0.00973*** -0.00721*** 0.00120*** 0.000106 0.0466*** 0.0480*** -0.0273*** -0.0269*** -0.00792*** (0.000274) (0.000300) (0.000351) (0.000473) (0.00556) (0.00658) (0.00101) (0.000951) (0.00131) Partner's share of top five cleanest industries to total production -0.00349*** -0.00189*** 0.00144*** -0.00104** -0.0428*** -0.0518*** -0.0158*** -0.0154*** -0.00150 (0.000268) (0.000281) (0.000327) (0.000453) (0.00565) (0.00590) (0.000892) (0.000833) (0.00104) 16 Dependent Variable: Natural log of CO2 emissions embodied in Total Exports except for PPML model Pooled OLS RE Panel Regression FE Panel Regression with economypair-sector FE Fixed Effects IV Regression PPML Regression Instrumented: Stringency of Environmental regulation Instrumented: Industrial composition Instrumented: Both No fixed effects with fixed effects: sector, year, sector-year with fixed effects: sector, year, sectoryear, economy-pairsector Policy Variables Reporter's stringency of environmental regulation (7 = most stringent) -0.223*** -0.0115*** 0.0188*** -0.498*** -0.00523 -1.052*** -0.0437*** -0.0412*** -0.0204* (0.00553) (0.00406) (0.00419) (0.0669) (0.00657) (0.103) (0.0149) (0.0139) (0.0110) Partner's stringency of environmental regulation (7 = most stringent) -0.0775*** 0.00601 0.0200*** -0.269*** 0.0271*** 0.470*** 0.135*** 0.136*** -0.00628 (0.00565) (0.00407) (0.00420) (0.0691) (0.00629) (0.0971) (0.0157) (0.0144) (0.0117) 1 = Preferential trade agreement between economies is present 0.131*** -0.0261*** -0.00557 0.0384*** -0.00693 0.00986 -0.124** -0.0979** -0.0107 (0.0161) (0.00602) (0.00612) (0.00794) (0.00794) (0.0111) (0.0483) (0.0486) (0.0174) Timeinvariant variables Natural log of distance -0.428*** -0.290*** (0.00412) (0.0110) distance -0.0875*** -0.0870*** (0.00376) (0.00357) Contiguity 0.961*** 1.336*** 1.143*** 1.140*** (0.0147) (0.0440) (0.0356) (0.0328) Common Language 0.213*** 0.169*** 0.152*** 0.153*** (0.0134) (0.0401) (0.0367) (0.0334) Same colony 0.0489** 0.101 0.0348 0.0405 (0.0204) (0.0626) (0.0359) (0.0290) Constant -30.04*** -27.32*** -21.68*** -17.91*** -24.46*** -21.55*** -2.079*** -1.583*** -0.258** (0.103) (0.212) (0.324) (0.519) (2.329) (2.922) (0.113) (0.106) (0.104) Observations 229,535 229,535 229,535 229,535 190,245 190,245 371,840 371,840 264,754 R-squared 0.329 0.051 Number of id 24,797 24,797 24,797 23,672 23,672 CO2 = carbon dioxide; FE = fixed effects; GDP = gross domestic product; IV = instrumental variables; OLS = ordinary least squares; PPML = Poisson pseudo-maximum likelihood; RE = random effects. Note: Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Sources: Authors’ estimates. 17 B. Domestic CO2 Emissions Embodied in Gross Exports Results using domestic CO2 emissions embodied in gross exports are similar to CO2 embodied in total exports when using pooled and random effects estimation. However, since this model rejects the null hypothesis in both Breusch-Pagan and Hausman tests, the fixed effects models are estimated. Compared to the previous result, the reporter’s stringency of environmental regulation becomes not significant, while the partner’s policy stringency remains positive and significant. Adding instrumental variables would change the signs of the coefficients. When policy stringency is instrumented for both reporter and partner, the reporter coefficient becomes negative and significant while the partner coefficient become negative but not significant. In contrast, when industrial composition is instrumented, the partner coefficient becomes positive and significant while the reporter coefficient remains negative and significant. When all instrumental variables were used, the reporter coefficient remains negative while that for the partner remains positive; in both cases, the coefficients were significant and larger in magnitude. Using PPML, which accounts for observations with zero values, the results are similar to CO2 embodied in total exports: an increase in the reporter’s policy stringency is associated with decreasing domestic CO2 embodied in its gross exports, while an increase in the partner’s policy stringency entails increasing domestic CO2 embodied in the reporter’s gross exports. The difference is that reporter coefficient is higher in magnitude than in the previous result, while partner coefficient is lower. Results show that a 1-percentage-point increase in the reporter’s stringency of environmental regulation is associated with 7% decrease in domestic CO2 emissions embodied in its total exports. And then a 1-percentage-point increase in partner’s stringency of environmental regulation is associated with 9% increase in domestic CO2 emissions embodied in its total imports (see footnote 4). This suggests that the production procedures of domestic valueadded are more directly and greatly influenced by environmental regulations. Adding sector and year fixed effects preserves the results. 18 Table 3: Domestic CO2 Emissions Embodied in Gross Exports Dependent Variable: Natural Log of Domestic CO2 emissions embodied in Gross Exports except for PPML model Pooled OLS RE Panel Regression FE Panel Regression with economypair-sector FE Fixed Effects IV Regression PPML Regression Instrumented: Stringency of Environmental regulation Instrumented: Industrial composition Instrumented: Both No fixed effects with fixed effects: sector, year, sectoryear with fixed effects: sector, year, sectoryear, economypair-sector Variable Description (1) (2) (3) (4) (5) (6) (7) (8) (9) Scale Effect Natural log of reporter economy's real GDP (base year = 2015) 0.552*** 0.395*** -0.230*** -0.182*** -0.728*** -0.770*** (0.00280) (0.00648) (0.0157) (0.0219) (0.0722) (0.0917) Reporter's Real GDP (base year = 2015) 0.000189*** 0.000190*** 9.50e-06 (2.95e-06) (2.83e-06) (6.66e-06) Natural log of partner economy's real GDP (base year = 2015) 0.525*** 0.477*** 0.810*** 0.771*** 1.396*** 1.440*** (0.00276) (0.00636) (0.0157) (0.0202) (0.0755) (0.0923) Partner's Real GDP (base year = 2015) 0.000187*** 0.000189*** 9.37e-05*** (3.01e-06) (2.71e-06) (6.38e-06) Technique Effect Natural log of reporter economy's CO2 embodied in production per capita 0.222*** 0.391*** 0.828*** 0.925*** 1.318*** 1.706*** (0.00671) (0.00985) (0.0138) (0.0221) (0.0816) (0.133) Reporter's CO 2 embodied in production per capita 0.0457*** 0.0441*** 0.0722*** (0.00383) (0.00373) (0.00513) Natural log of partner economy's CO2 embodied in production per capita 0.0302*** 0.195*** 0.179*** 0.190*** -0.318*** -0.688*** (0.00648) (0.00948) (0.0131) (0.0223) (0.0751) (0.117) Partner's CO 2 embodied in production per capita 0.0193*** 0.0180*** 0.0348*** (0.00382) (0.00359) (0.00539) Composition Effect Reporter's share of top five cleanest industries to total production -0.0138*** -0.00984*** -0.000974*** -0.00109** 0.0476*** 0.0591*** -0.0380*** -0.0376*** -0.00790*** (0.000290) (0.000321) (0.000375) (0.000515) (0.00643) (0.00874) (0.00135) (0.00128) (0.00177) Partner's share of top five cleanest industries to total production -0.00295*** -0.00202*** 0.00107*** -0.000859* -0.0481*** -0.0604*** -0.0159*** -0.0155*** -0.00159 (0.000279) (0.000296) (0.000347) (0.000492) (0.00609) (0.00754) (0.00103) (0.000962) (0.00125) 19 Dependent Variable: Natural Log of Domestic CO2 emissions embodied in Gross Exports except for PPML model Pooled OLS RE Panel Regression FE Panel Regression with economypair-sector FE Fixed Effects IV Regression PPML Regression Instrumented: Stringency of Environmental regulation Instrumented: Industrial composition Instrumented: Both No fixed effects with fixed effects: sector, year, sectoryear with fixed effects: sector, year, sectoryear, economypair-sector Policy Variables Reporter's stringency of environmental regulation (7 = most stringent) -0.270*** -0.0351*** 0.000748 -0.402*** -0.0288*** -1.133*** -0.0722*** -0.0683*** -0.0249** (0.00570) (0.00414) (0.00427) (0.0716) (0.00681) (0.133) (0.0166) (0.0155) (0.0127) Partner's stringency of environmental regulation (7 = most stringent) -0.0857*** 0.00957** 0.0244*** -0.0575 0.0382*** 0.802*** 0.0850*** 0.0869*** -0.00499 (0.00582) (0.00420) (0.00434) (0.0723) (0.00672) (0.124) (0.0181) (0.0166) (0.0141) 1 = Preferential trade agreement between economies is present 0.118*** -0.0194*** -0.00491 0.0191** -0.00895 -0.00357 -0.169*** -0.146*** 0.0223 (0.0168) (0.00630) (0.00639) (0.00793) (0.00825) (0.0123) (0.0525) (0.0533) (0.0173) Time - invariant variables Natural log of distance -0.340*** -0.202*** (0.00424) (0.0112) distance -0.0846*** -0.0841*** (0.00427) (0.00400) Contiguity 0.932*** 1.285*** 1.110*** 1.107*** (0.0147) (0.0435) (0.0402) (0.0371) Common Language 0.203*** 0.130*** 0.223*** 0.223*** (0.0136) (0.0402) (0.0390) (0.0356) Same colony 0.0363* 0.0858 0.0667 0.0737** (0.0205) (0.0621) (0.0412) (0.0336) Constant -28.95*** -27.01*** -21.88*** -19.93*** -24.13*** -22.49*** -1.812*** -1.293*** -0.506*** (0.107) (0.218) (0.331) (0.499) (2.404) (3.316) (0.125) (0.116) (0.123) Observations 206,191 206,191 206,191 206,191 171,001 171,001 371,840 371,840 241,905 R-squared 0.307 0.055 Number of id 22,655 22,655 22,655 21,604 21,604 CO2 = carbon dioxide; FE = fixed effects; GDP = gross domestic product; IV = instrumental variables; OLS = ordinary least squares; PPML = Poisson pseudo-maximum likelihood; RE = random effects. Note: Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Source: Authors’ estimates. 20 C. Foreign CO2 Emissions Embodied in Gross Exports Similar to CO2 embodied in total exports and domestic CO2 emissions embodied in exports, the pooled OLS and random effects models reject the null hypothesis in both Breusch-Pagan and Hausman tests, which prompts the use of the fixed effects model. The results are somewhat different from those for total exports and the domestic portion of CO2 emissions in gross exports. When the stringency of environmental regulation for both the reporter and partner are instrumented in particular, their coefficient signs flip from positive to negative. Whereas the negative signs of the partners’ environmental policy are largely preserved, the sign of reporters’ environmental policy regulation becomes positive and the magnitude of the positive coefficients for partners’ environmental policy regulation becomes smaller. Estimating the model using PPML, estimates now differ from domestic CO2 embodied in exports; the reporter’s increase in regulatory stringency is associated with an increase in foreign CO2 embodied in gross exports, while partner’s increase in regulatory stringency is also associated with an increase in foreign CO2 embodied in gross imports. Results show that a 1-percentage point increase in reporter’s stringency is associated with 6% increase in foreign CO2 emissions embodied in its total exports (see footnote 4). Adding sector and year fixed effects preserves the results although both the coefficients become insignificant when economy-pair-sector fixed effects are used. The results indicate stricter environmental regulations in exporting economies might lead to carbon leakages in the upstream segment of global value chains, through economies outsourcing the dirtier segments of productions to other economies. 27 A.2: IV Test Results for CO2 Embodied in Gross Exports Table A.2.1: Each Instrumented Variable Underidentification test Weak identification test Anderson-Rubin Wald test SandersonWindmeijer SandersonWindmeijer Variable F(4,16656 4) P-val Chisq(1) P-val F stat (1,166564) Reporter's share of top five cleanest industries to total production 449.71 0.0000 391.69 0.0000 391.67 Partner’s share of top five cleanest industries to total production 383.20 0.0000 482.79 0.0000 482.76 Reporter's stringency of environmental regulation (7 = most stringent) 293.62 0.0000 329.11 0.0000 329.09 Partner’s stringency of environmental regulation (7 = most stringent) 291.50 0.0000 363.48 0.0000 363.46 Source: Authors’ estimates. Table A.2.2: Overall Results Under-identification test Ho: under-identified Anderson canon. corr. LM statistic Chi-sq(1)=275.51 P-val=0.0000 Weak identification test Ho: equation is weakly identified Cragg-Donald Wald F statistic 68.99 Tests of joint significance of endogenous regressors B1 in main equation Ho: B1=0 and orthogonality conditions are valid Anderson-Rubin Wald test F(4,166564)= 59.37 Pval=0.0000 Chi-sq(4)= 237.51 P-val=0.0000 Stock-Wright LM S statistic Chi-sq(4)= 237.17 P-val=0.0000 Over-identification test of all instruments Sargan statistic 0.000 (equation exactly identified) Source: Authors’ estimates. A.3: IV Test Results for CO2 Embodied in Domestic Value-added Exports Table A.3.1: Each Instrumented Variable Underidentification test Weak identification test Anderson-Rubin Wald test SandersonWindmeijer SandersonWindmeijer Variable F(4,16656 4) P-val Chisq(1) P-val F stat (1,166564) Reporter's share of top five cleanest industries to total production 383.09 0.0000 232.79 0.0000 232.78 Partner’s share of top five cleanest industries to total production 347.98 0.0000 282.13 0.0000 282.12 Reporter's stringency of environmental regulation (7 = most stringent) 254.54 0.0000 199.41 0.0000 199.39 Partner’s stringency of environmental regulation (7 = most stringent) 259.03 0.0000 218.86 0.0000 218.85 Source: Authors’ estimates. 28 Table A.3.2: Overall Results Under-identification test Ho: under-identified Anderson canon. corr. LM statistic Chi-sq(1) = 171.60 P-val = 0.0000 Weak identification test Ho: equation is weakly identified Cragg-Donald Wald F statistic 42.95 Weak-instrument-robust inference tests of joint significance of endogenous regressors B1 in main equation Ho: B1=0 and orthogonality conditions are valid Anderson-Rubin Wald test F(4,149388)= 45.26 Pval=0.0000 Chi-sq(4)= 181.07 P-val=0.0000 Stock-Wright LM S statistic Chi-sq(4)= 180.85 P-val=0.0000 overidentification test of all instruments Sargan statistic 0.000 (equation exactly identified) Source: Authors’ estimates. A.4: IV Test Results for CO2 Embodied in Foreign Value-added Exports Table A.4.1: Each Instrumented Variable Underidentification test Weak identification test Anderson-Rubin Wald test SandersonWindmeijer SandersonWindmeijer Variable F(4,16656 4) P-val Chisq(1) P-val F stat (1,166564) Reporter's share of top five cleanest industries to total production 442.79 0.0000 234.82 0.0000 234.81 Partner’s share of top five cleanest industries to total production 354.13 0.0000 242.93 0.0000 242.91 Reporter's stringency of environmental regulation (7 = most stringent) 266.52 0.0000 195.67 0.0000 195.66 Partner’s stringency of environmental regulation (7 = most stringent) 248.93 0.0000 202.80 0.0000 202.79 Source: Authors’ calculations. Table A.4.2: Overall Results Under-identification test Ho: under-identified Anderson canon. corr. LM statistic Chi-sq(1)= 164.72 P-val=0.0000 Weak identification test Ho: equation is weakly identified Cragg-Donald Wald F statistic 41.23 Weak-instrument-robust inference Tests of joint significance of endogenous regressors B1 in main equation Ho: B1=0 and orthogonality conditions are valid Anderson-Rubin Wald test F(4,138570)= 66.47 Pval=0.0000 Chi-sq(4)= 265.91 P-val=0.0000 Stock-Wright LM S statistic Chi-sq(4)= 265.40 P-val=0.0000 overidentification test of all instruments Sargan statistic 0.000 (equation exactly identified) Source: Authors’ estimates. 29 A.5: Stock-Yogo Weak ID F Test: Critical Values for Single Endogenous Regressor 5% maximal IV relative bias 16.85 10% maximal IV relative bias 10.27 20% maximal IV relative bias 6.71 30% maximal IV relative bias 5.34 10% maximal IV size 24.58 15% maximal IV size 13.96 20% maximal IV size 10.26 25% maximal IV size 8.31 Note: Critical values are for Sanderson-Windmeijer F statistic. Source: Stock-Yogo (2005). Reproduced by permission. 30 References Aichele, Rahel, and Gabriel J. Felbermayr. 2011. “Kyoto and Carbon Leakage: An Empirical Analysis of the Carbon Content of Bilateral Trade.” CESifo Working Paper, No. 3661,Center for Economic Studies and ifo Institute (CESifo), Munich. Aldy, Joseph E, and William A Pizer. 2015. “The competitiveness impacts of climate change mitigation policies.” J. Assoc. Environ. Res. Econ. 2 (4) 565-595. Arroyo-Curras, Tabare, Nico Bauer, Elmar Kriegler, Valeria Jana Schwanitz, Gunnar Luderer, Tino Aboumahboub, Anastasis Giannousakis, and Jerome Hilaire. 2013. “Carbon leakage in a fragmented climate regime: the dynamic response of global energy markets.” Technol. Forecast. Soc. Change 90 (A) 192–203. Assogbavi, Koutchogna Kokou, and Stephane Dees. 2021. “Environmental policy and the CO2 emissions embodied in international trade.” Bordeaux Economics Working Papers, BxWP2021-07. Babiker, Mustafa H, and Thomas F Rutherford. 2005. “The economic effects of border measures in sub-global climate agreements.” Energy J. 26 (4) 99-125. Barrett, Scott. 1994. “Self-enforcing international environmental agreements.” Oxf. Econ. Pap. 46 878-894. Bauer, Nico, Ioanna Mouratiadou, Gunnar Luderer, Lavinia Baumstark, Robert J Brecha, Ottmar Edenhofer, and Elmar Kriegler. 2016. “Global fossil energy markets and climate change mitigation–an analysis with REMIND.” Clim. Change 136 (1) 69–82. Baylis, Kathy, Don Fullerton, and Daniel H Karney. 2013. “Leakage, welfare, and costeffectiveness of carbon policy.” Am. Econ. Rev. 103 (3) 332-337. Bohringer, Christoph, Edward J Balistreri, and Thomas F Rutherford. 2012a. “The role of border carbon adjustment in unilateral climate policy: overview of an energy modeling forum study (EMF 29).” Energy Econ. 34 97-110. Bohringer, Christoph, Jared C Carbone, and Thomas F Rutherford. 2012b. “Unilateral climate policy design: efficiency and equity implications of alternative instruments to reduce carbon leakage.” Energy Econ. 34 (December) S208-S217. Brandt, Urs Steiner. 2004. “Unilateral actions, the case of international environmental problems.” Resour. Energy Econ. 26 (4) 373-391. Branger, Frederic, and Philippe Quirion. 2014. “Climate policy and the ‘carbon haven’ effect.” Wiley Interdiscipl. Rev. Clim. Change 5 (1) 53-71. Buchholz, Wolfgang, and Kai A Konrad. 1994. “Global environmental problems and the strategic choice of technology.” J. Econ. 60 (3) 299-321. Carbone, Jared C. 2013. “Linking numerical and analytical models of carbon leakage.” Am. Econ. Rev. 103 (3) 326-331. 31 Carraro, Carlo, and Domenico Siniscalco. 1993. “Strategies for the international protection of the environment.” J. Publ. Econ. 52 (3) 309-328. Chichilnisky, Graciela. 1994. “North-south trade and the global environment.” Am. Econ. Rev. 84 (4) 851-874. Copeland, Brian R., and M. Scott Taylor. 1994. “North-South Trade and the Environment.” The Quarterly Journal of Economics, Vol. 109, No. 3. (Aug., 1994) 755-787. Correia, Sergio, Paulo Guimaraes, and Tom Zylkin. 2020. “Fast Poisson estimation with highdimensional fixed effects.” The Stata Journal (2020) 20, Number 1 95-115. Edenhofer, Ottmar, and Matthias Kalkuhl. 2011. “When do increasing carbon taxes accelerate global warming? A Note on the green paradox.” Energy Policy 39 (4) 2208-2212. Ederington, Josh, and Jenny Minier. 2003. “Is Environmental Policy a Secondary Trade Barrier? An Empirical Analysis.” Canadian Journal of Economics, Vol. 36, No. 1. Ederington, Josh, Arik Levinson, and Jenny Minier. 2003. “Footloose and Pollution-Free.” NBER Working Paper Series. Ederington, Josh, Mihai Paraschiv, and Maurizio Zanardi. 2018. “The Short and Long-Run Effects of International Environmental Agreements on Trade.” Lancaster University Management School. Economics Working Paper Series. 2018/015. Elliott, Joshua, Ian Foster, Samuel Kortum, Todd Munson, Fernando Perez Cervantes, and David Weisbach. 2010. “Trade and carbon taxes.” Am. Econ. Rev. 100 (2) 465–469. Felder, Stefan, and Thomas F Rutherford. 1993. “Unilateral CO2 reductions and carbon leakage: the consequences of international trade in oil and basic materials.” J. Environ. Econ. Manag. 25 (2) 162–176. Jakob, Michael. 2021a. “Climate policy and international trade - A critical appraisal of the literature.” Energy Policy 156. _____. 2021b. “Why carbon leakage matters and what can be done against it.” One Earth 4, May 21, 2021 609-614. Jakob, Michael, and Kai Lessmann. 2012. “Signaling in international environmental agreements: the case of early and delayed action.” Int. Environ. Agreements Polit. Law Econ. 12 (4) 309-325. Jensen, Svenn, Kristina Mohlin, Karen Pittel, and Thomas Sterner. 2015. “An introduction to the green paradox: the unintended consequences of climate policies.” Rev. Environ. Econ. Pol. 9 (2) 246-265. Levinson, Arik. 2009. “Technology, International Trade, and Pollution from US Manufacturing.” The American Economic Review, Vol. 99, No. 5 (Dec., 2009) 2177-2192. Levinson, Arik, and M. Scott Taylor. 2004. “Unmasking the Pollution Haven Effect.” NBER Working Paper Series. 32 List, John A., Daniel L. Millimet, Per G. Fredriksson, and W. Warren McHone. 2003. “Effects of environmental regulations on manufacturing plant births: Evidence from a propensity score matching estimator.” Faculty Bibliography 2000s. Markusen, J.R. 1975. “International externalities and optimal tax structures.” J. Int. Econ. 5 1529. Marschinski, Robert, Christian Flachsland, and Michael Jakob. 2012. “Sectoral linking of carbon markets: a trade-theory analysis.” Resour. Energy Econ. 34 (4) 585-606. Martin, Ralf, Laure B de Preux, and Ulrich J Wagner. 2014. “The impact of a carbon tax on manufacturing: evidence from microdata.” J. Publ. Econ. 117 (September) 1-14. Misch, Florian, and Philippe Wingender. 2021. “Revisiting Carbon Leakage.” IMF Working Paper. OECD. 2021. OECDstat: Carbon dioxide emissions embodied in international trade (TECO2) data set. https://stats.oecd.org (accessed December 2021). Sato, Misato, Karsten Neuhoff, Verena Graichen, Katja Schumacher, and Felix Matthes. 2015. “Sectors under scrutiny: evaluation of indicators to assess the risk of carbon leakage in the UK and Germany.” Environ. Resour. Econ. 60 (1) 99-124. Shahbaz, Muhammad, Giray Gozgor, Philip Kofi Adom, and Shawkat Hammoudeh. 2019. “The Technical Decomposition of Carbon Emissions and the Concerns about FDI and Trade Openness Effects in the United States.” Munich Personal RePEc Archive. Shapiro, Joseph S, and Reed Walker. 2018. “Why is Pollution from U.S. Manufacturing Declining? The Roles of Environmental Regulation, Productivity, and Trade.” NBER Working Paper Series. Stock, James H, and Motohiro Yogo. 2005. “Testing for weak instruments in linear IV regression.” Identification and Inference for Econometric Models: Essays in Honor of Thomas Rothenberg. Cambridge: Cambridge University Press. Chapter 5 80-108. Winchester, Niven, and Sebastian Rausch. 2013. “. A numerical investigation of the potential for negative emissions leakage.” Am. Econ. Rev. 103 (3) 320-325. Yamano, Norihiko, and Joaquim Guilhoto. 2020. “CO2 emissions embodied in international trade and domestic final demand: Methodology and results using the OECD InterCountry Input-Output Database.” OECD Science, Technology and Industry Working Papers 2020/11. ASIAN DEVELOPMENT BANK ASIAN DEVELOPMENT BANK 6 ADB Avenue, Mandaluyong City 1550 Metro Manila, Philippines www.adb.org FACTORS AFFECTING CARBON DIOXIDE EMISSIONS EMBODIED IN TRADE Jong Woo Kang and Joshua Anthony Gapay ADB ECONOMICS WORKING PAPER SERIES NO. 700 October 2023 Factors Affecting Carbon Dioxide Emissions Embodied in Trade This paper examines the impact of environmental regulation in exporter and importer economies on crossborder carbon flows. While stricter environmental regulations help reduce carbon dioxide (CO2) emissions from domestic production, leading to lower CO2 emissions embodied in exports, stricter regulations on the importing side lead to higher CO2 emissions embodied in imports. Moreover, stricter environmental regulations could encourage further outsourcing of intermediate inputs by exporters, prompting carbon leakages in the upstream segment of global value chains. About the Asian Development Bank ADB is committed to achieving a prosperous, inclusive, resilient, and sustainable Asia and the Pacific, while sustaining its efforts to eradicate extreme poverty. Established in 1966, it is owned by 68 members —49 from the region. Its main instruments for helping its developing member countries are policy dialogue, loans, equity investments, guarantees, grants, and technical assistance.