Policy Research Working Paper 10855 Trade Policies Mix and Match Theory, Evidence and the EU-Sino Electric Vehicle Disputes Hiau Looi Kee Enze Xie Development Economics Development Research Group July 2024 Policy Research Working Paper 10855 Abstract This paper studies the factors affecting governments’ mixed trade agreements, and products with consumption external- use of tariffs and non-tariff measures (NTMs) as trade and ities. A terms-of-trade model with externalities rationalizes industrial policies. Results based on detailed bilateral-prod- the results. The model is further used to shed light on the uct-level ad valorem equivalent estimates for a wide range recent Sino-EU battery electric vehicle (BEV) disputes, of countries show that restrictive NTMs coexist with lower whereby the EU imposed NTMs on top of the tariffs on tariffs, particularly for high-income importing countries, China’s BEVs. low-income exporting countries, country pairs with deep This paper is a product of the Development Research Group, Development Economics. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at hlkee@worldbank.org and exie@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Trade Policies Mix and Match: Theory, Evidence and the Sino-EU Electric Vehicle Disputes* Hiau Looi KEE† Enze XIE‡ World Bank Zhejiang University and World Bank Originally published in the Policy Research Working Paper Series on July 2024. This version is updated on November 2024. To obtain the originally published version, please email prwp@worldbank.org. Keywords: Tariffs, non-tariff measures, ad valorem equivalent, trade policies, industrial policies, externalities, electric vehicles, Sino-EU trade disputes JEL Code: F13, F14, L52, O25 * We thank Joseph Stiglitz, Robert Koopman, David Hummels, Andreas Moxnes, Kaoru Nabeshima, Marcelo Olarreage, Anson Soderbery, Ariel Weinberger, Chong Xiang, Yoto Yotov, and participants at the EIIT at Purdue University, Jinan University, Singapore Management University, Sun Yat-Sen University, SIPA at Columbia University, UNCTAD-WB NTM workshops in Washington, DC and Geneva, Waseda University, World Bank Industrial Policy and the Role of Human Capital Workshop, Zhejiang University, and Zhongnan University of Economics and Law for valuable comments. Research for this paper has in part been supported by the World Bank’s Multi-donor Trust Fund for Trade and Development. The authors declare that they have no relevant or material financial interests that relate to the research described in this paper. The results and opinions presented in this paper do not represent the views of our institutions, the Executive Directors, or the countries they represent. † Development Research Group, The World Bank; 1818 H ST NW, Washington, DC 20433. Email: hlkee@worldbank.org. ‡ Corresponding author. School of Economics, Zhejiang University, China, and Development Research Group, The World Bank. Email: enzexie@zju.edu.cn. “The EU remains the global champion for open, fair and rules-based trade. We welcome competition, including in the electric vehicle sector, but it must be underpinned by fairness and a level playing field. By adopting these proportionate and targeted measures after a rigorous investigation, we’re standing up for fair market practices and for the European industrial base. In parallel, we remain open to a possible alternative solution that would be effective in addressing the problems identified and WTO compatible.” —Valdis Dombrovskis, Executive Vice-President and Commissioner for Trade - Oct 29th 2024 1 Introduction The recent EU’s decision to impose firm-specific countervailing duties (CVDs) on the import of battery electric vehicles (BEVs) from China, in addition to the existing 10% non-discriminative tariff, ushered a new era of governments mixing and matching trade policies to promote long-term industrial and climate goals (Harstad 2024a).¹ Other countries with similar decisions on Chinese BEVs include Canada, mixing tariffs with surtax, and the US, mixing tariffs with critical component restrictions.² While both CVDs and surtax are officially classified by UNCTAD (2015) as non-tariff measures (NTMs), the intentions of shielding and reshaping domestic industries with these NTMs make them industrial policies (Juhász et al. 2024). These import-related border NTMs, which also include sanitary and phytosanitary (SPS) measures and technical barriers to trade (TBT) measures, aiming primarily at protecting public health or the environment, have been more widely used in recent decades when tariffs fall and significantly limit the available policy tools of governments in shaping domestic industries (Figure 1). Do countries mixing tariffs with NTMs indicate that they are policy substitutes? In other words, facing historically low tariffs, do governments resort to imposing discriminatory and restrictive NTMs that are harder to identify, as alternatives to achieve policy agenda without violating the WTO rules? What factors affect governments’ mixed use of tariffs and NTMs as trade and industrial policies? Ultimately, are policies claimed to safeguard public health or the environment protectionism in disguise, intended instead to safeguard the domestic economy? The answers to these questions could broaden our understanding of why and how different countries use different trade policies on different products to address domestic or global objectives. To address these questions, this paper provides a comprehensive analysis of the trade policy landscape, taking into account the heterogeneity of nearly 50 importing countries, 100 plus exporting countries and around ¹Source: https://ec.europa.eu/commission/presscorner/detail/en/ip_24_5589 (Press release, the EU) ²Source: Surtax, https://www.cbsa-asfc.gc.ca/publications/cn-ad/cn24-32-eng.html (Custom notice, Govern- ment of Canada); Tariff, https://ustr.gov/about-us/policy-offices/press-office/press-releases/2024/september/ustr- finalizes-action-china-tariffs-following-statutory-four-year-review (Press release, the USTR); Component restriction, https://www.bis.gov/press-release/commerce-announces-proposed-rule-secure-connected-vehicle-supply-chains-foreign (Press release, the USDC) 1 5,000 products, further sheds light on the recent trade spats involving countries using discretionary or discriminatory trade policies as industrial policies.³ Many papers have contributed to this topic, with most papers focusing on specific countries, prod- ucts, or trade policies, mostly concluding that tariffs and NTMs are policy substitutes. For example, Conconi et al. (2018) find that rules of origin (ROO), a form of NTMs, in the North American Free Trade Agreement (NAFTA), leads to a sizable reduction in intermediate imports from third coun- tries relative to NAFTA partners, supporting that tariffs and NTMs are policy substitutes. However, anecdotal evidence suggests that the relationship between the two trade policy instruments is more nuanced. Tariffs and NTMs could be considered policy substitutes for the EU’s Everything But Arms (EBA) Initiative, which gives duty-free quota-free access to the EU market for products from the least developed countries that satisfy the stringent ROO requirements. On the contrary, since Jan- uary 1, 1994, virtually all semi-conductors and all local area network apparatus entering the US faced zero tariffs, regardless of country of origin, and no restrictive NTMs, which would imply tariffs and NTMs could be regarded as policy complements. India is the target of many WTO dispute settle- ments,⁴ with exporting countries such as Brazil, Australia, and Guatemala complaining about India’s NTMs on sugar,⁵ while Taiwan China, and the EU complaining about India’s tariff on information technology products.⁶ This could mean that for India, tariffs and NTMs are complementary policies. Overall, the anecdotal evidence suggests that the relationship between tariffs and NTMs depends on the characteristics of importing countries, exporting countries, products, and trade agreements. These observations motivate our empirical strategy, which accounts for country, product, and trade partner- ship heterogeneity by using interaction terms to examine how these factors influence the empirical relationship between tariffs and NTMs. This paper presents rich empirical evidence on factors affecting the correlation between tariffs and NTMs, based on the latest detailed estimates of product-level bilateral ad valorem equivalent (AVE) of import-related border NTMs from Kee and Nicita (2022).⁷ While the import-related border NTMs are mostly qualitative, the AVE estimates from Kee and Nicita (2022) capture the restrictiveness of these NTMs in affecting trade flows, converting them into quantitative terms comparable to tariffs.⁸ ³According to UNCTAD (2015), CVDs are contingent trade-protective NTMs (D2) that are firm-specific, while surtax is a customs surcharge type NTMs (F4). In addition, both CVDs and surtax are import-related border NTMs, levied solely on imported products in addition to tariffs to raise fiscal revenues and protect domestic industries. Section 6 presents a detailed discussion of why CVDs and surtax should be regarded as NTMs. ⁴India has been the respondent in 32 trade dispute cases and the third party in 182 trade dispute cases (Source: WTO Dispute Settlement Database. Last Update: Nov 3rd, 2024) ⁵Please refer to DS579, DS580 and DS581 for disputes between Brazil, Australia, Guatemala and India on non-tariff measures concerning sugar and sugarcane, respectively. ⁶Please refer to DS582, DS588 for disputes between the EU, Taiwan, China, and India on tariff treatment on certain goods in the information and communications technology sector. ⁷Section 2 provides detailed discussions on the concept of import-related border NTMs and the rationale for focusing on them. All empirical analyses of this paper used import-related border NTMs unless otherwise stated. ⁸Section 3.1 provides a detailed discussion on the merits of using AVE to measure the impact of the NTMs. 2 To capture various characteristics of importing countries, exporting countries, and products that may affect the relationship between tariffs and NTMs, relevant interaction terms and fixed-effects (FEs) are included in the regressions. The rich variations of AVEs across importers, exporters, and products allow this paper to study the empirical relationship between NTMs and tariffs in FEs instrumental variable regressions pooling across products, importing and exporting countries, and thus defined the point of departure of this paper from the existing literature. Our empirical results confirm that overall tariffs and NTMs are policy substitutes in the sense that governments impose more restrictive NTMs on products or trading partners with lower tariffs. How- ever, depending on the characteristics of the importing countries, exporting countries and products, governments also mix and match tariffs and NTMs, which may turn the relationship between them to be less substitutive and may even be complementary. Importing countries with higher income, more capital or skilled-labor abundant, are likely to have more liberal tariffs and restrictive NTMs. Like- wise, exporting countries that are labor abundant, or have lower income often face more liberal tariffs and restrictive NTMs. Lower tariffs coupled with restrictive NTMs are also found in country pairs with deep trade agreements, while engagement in traditional multilateral agreements such as WTO has no significant impact on the relationship between tariffs and NTMs. Policy substitution is further found in consumption products, agricultural products, and food and beverage products. In contrast, intermediate products and capital products which are part of supply chains, often face complementary trade policies. Finally, restrictive NTMs coexist with low tariffs for importing and exporting countries with better institution and governance. To rationalize the empirical findings that restrictive NTMs coexist with low tariffs in importing countries with higher income or better institution, or products with consumption externalities, this paper presents a simple terms-of-trade model built on Ederington (2001). In this model, the govern- ment chooses tariffs and NTMs to maximize social welfare. Instead of assuming a negative externality associated with the production of the imported product as in Ederington (2001), in this paper, there is a negative externality associated with the consumption of the product that is being imported, which can be reduced by the restrictive NTMs, similar to that of Copeland (1994) and Yue (2022). The effectiveness of NTMs in reducing externality depends on the governance and institutional quality of the importing country, the compliant capability of the exporting country, and the product character- istics. Jointly, both tariffs and NTMs create a wedge between the world price and the domestic price of imported goods.⁹ Similar to tariffs, NTMs reduce imports and depress the world price, leading to terms-of-trade gains. Furthermore, NTMs also improve social welfare directly by reducing the con- sumption externalities of imports through boosting public confidence. In equilibrium, countries with market power choose to impose NTM if tariff is below the optimal level, which gives rise to the policy substitution between the two trade policy instruments. The weight of the consumption externality in ⁹This paper focuses on import-related border NTMs, section 2 presents the detailed discussion. 3 the social welfare function and the severity of externality depends on the characteristics of importing countries and products, influence how the welfare-maximizing government may mix and match the two policies. Results based on structural estimations of the model parameters lend credence to the theory and collaborate with the reduced-form findings by showing that the weight of the consumption externality increases with the income level of the importing countries, and the marginal externalities of agriculture products are higher. Finally, this simple model is used to analyze the recent Sino-EU trade disputes, whereby the EU imposed NTMs on top of tariffs on imported Chinese BEVs. This model shows that such a policy mix by imposing additional NTMs is optimal to reduce imports facing a lower world price and capture terms-of-trade gains, when EU’s existing tariff is not sufficient to do so. The results of this paper is further used to shed light on the different policy responses of the EU, the US and Canada in the recent trade spats involving Chinese BEVs, whereby discretionary or discriminatory trade policies are used as industrial policies. This paper relates to both the theoretical and empirical literature on trade policy determinations, especially regarding the relationship between the use of tariffs and NTMs. Even though many papers have contributed to this topic, there is no clear consensus. On the empirical front, while the earlier evidence indicates that tariffs and NTMs are policy complements (Lee and Swagel 1997), the more recent evidence since 2000 suggests the opposite (Beverelli et al. 2019, Bown and Tovar 2011, Chen et al. 2022, Conconi et al. 2018, Feinberg and Reynolds 2007, Herghelegiu 2018, Kee et al. 2009, Ketterer 2016, Limão and Tovar 2011, Moore and Zanardi 2011, Niu et al. 2018; 2020, Orefice 2017).¹⁰ Some other works support that tariffs and NTMs are complementary or the relationship between them is overall substitutive but contingent, influenced by the government’s bargaining power to special interest groups (Limão and Tovar 2011), product type (Heo and Choi 2023), countries development stages or growth rates (Beverelli et al. 2019, Heo and Choi 2023, Niu et al. 2018; 2020). Conconi et al. (2018) focus on NAFTA, by exploiting cross-product and cross-country variation in treatment over time, they show that NAFTA ROO reduced imports of intermediate goods from third countries relative to NAFTA partners. Our finding that the substitution between tariffs and NTMs is higher for bilateral country pair with a deep trade agreement is consistent with this result, even though we used a completely different approach covering different countries, products and measures. Our highly disaggregated AVE estimates enable us to detect richer determinants of the tariffs and NTMs relationships, to provide a comprehensive overview.¹¹ The relationships between tariffs and NTMs are also not settled on the theoretical front. In the ¹⁰These recent studies are heterogeneous in types and empirical measurements of NTMs, tariffs, and sample coverage. Beyond those, Kuenzel (2020) and Beshkar et al. (2015) find that the overhang in tariff and NTMs are substitutes. The overhang in tariff is the difference between WTO members’ bound tariff rates and applied tariff rates, namely water in the tariff, which reflects the government’s flexibility in adjusting tariffs under the WTO regulation. ¹¹The limitation of cross section data on NTMs refrains us from investigating the relationship between NTMs and tariffs over time. However, our empirical results uncover several novel and important determinants of trade policy, such as global value chain (GVC) participation, governance, and engagement in regional trade agreements. 4 classic paper of Grossman and Helpman (1994), the government endogenously chooses the combi- nation of trade policy instruments considering the political support from the interest groups, which leads to policy substitutions.¹² Similarly, Yu (2000) shows that the degree of substitution between NTMs and tariffs increases with the government’s valuation of political contribution. Limão and To- var (2011) emphasize that the improved bargaining position of the government relative to interest groups brought by international cooperation commitments motivates the government to use less ef- ficient NTMs.¹³ Tariffs and NTMs are complementary in reducing production misallocation in the recent work of Macedoni and Weinberger (2024), because lower tariffs imply less misallocation which requires smaller regulations to correct. This paper contributes to this set of literature by showing that the degree of substitution between tariffs and NTMs depends on the weight of consumption external- ity in the government’s social welfare function, as well as the effectiveness of NTMs in reducing the consumption externality, with collaborating structural estimations of the parameters of the model. This paper is also related to a new strand of literature focusing on industrial policies, given that governments use discretionary trade policies, particularly some NTMs to affect the domestic industrial structure to promote economic or climate goals. According to Juhász et al. (2024), industrial policies are those government policies that explicitly target the transformation of the economic structure in pursuit of some public goal, such as structural transformation, re-industrialization, decarbonization, good jobs creation, supply chain resilience, and national security. Juhász et al. (2022) further shows that industrial policies are used by high-income importing countries more. Our paper provides collab- orative evidence consistent with this result, that the substitution between tariffs and NTMs is higher for the high-income importing countries which, facing low tariffs, have to resort to NTMs to achieve domestic industrial goals. In addition, our results that countries with better institutions and gover- nance used more NTMs, are particularly relevant, when we consider NTMs as industrial policies that may encounter constraints on state or bureaucratic capacity and the need to minimize rent-seeking behavior to successfully implement these policies. Atkin et al. (2024) analyze the market power in international trade through studying NTMs as discretionary trade policies targeting specific firms and products. Focusing on the import licensing requirements of Argentina, they find that high-income exporting countries with larger market power confront less restrictive NTMs. Our results based on nearly 50 importing countries showing that, for a given importer and product, there is less substitution between tariffs and AVEs for high-income exporting countries, collaborate well with their findings, suggesting that this result has broad supports beyond Argentina. Finally, this paper emphasizes the role of NTMs in reducing consumption externality, in addition to achieving terms-of-trade gains, which serves as a rationale for governments to substitute tariffs with NTMs as industrial policies. ¹²Bown (2014) provides a comprehensive review of political-economic research on international trade policy. ¹³Empirical evidence in India finds that anti-dumping and safeguard are used to replace tariffs to protect domestic market . Similarly, Ruckteschler et al. (2022) find that politically connected firms receive higher-level NTM protection after the trade agreement enrollment, highlighting the importance of institutional factors in determining trade policy. 5 This paper also connects to two growing areas of research. The first focuses on the role of climate change related NTMs in trade agreements (Cruz and Rossi-Hansberg 2024, Harstad 2024a;b). In these papers, the presence of climate change related NTMs may act as a source of comparative advan- tage which increases the trade of climate-intensive products, broadly encompassing energy-intensive, carbon-intensive, or emission-intensive products; or trade may lead to the deterioration of local cli- mate, which could be addressed with tariffs or NTMs. In particular, Harstad (2024a) shows that with negative externalities, a trade agreement that mixes tariffs with NTMs reaches the first best outcome without deteriorating climate. Our findings that governments may use more NTMs to regulate trade of climate-intensive goods such as forestry products, which are included in agriculture, provide empir- ical support for Harstad (2024a). The second growing area studies the recent trade disputes between the major economies (Fajgelbaum et al. 2019, Fajgelbaum et al. 2024). These papers mainly focus on how the US government raised tariffs to protect the domestic market from Chinese imports. By analyzing the EU government mixing tariffs and NTMs in the Sino-EU BEVs trade disputes from the lens of theoretical model, our paper highlights the importance of examining both policy instruments. This paper proceeds as follows. Section 2 discusses the backdrop of trade policy landscape change and data used in the analysis. Empirical strategies and reduced-form empirical results are shown in Sections 3 and 4, respectively. Section 5 presents a simple terms-of-trade model to rationalize the empirical findings, together with structural estimations of the model’s parameters that related back to the reduced-form findings. Section 6 applies our model to analyze the case that the EU imposes NTMs on top of tariffs on imported Chinese BEVs, and discuss the difference in BEV trade policy across the EU, the US and Canada based on our model and empirics. Section 7 concludes the paper. 2 Change in Landscape of Trade Policy and Data The landscape of trade policy has changed dramatically in the recent decades, as shown in Figure 1. Under the effort of multilateral negotiations, the world average applied tariff has been falling steadily. However, deadlocks in multilateral negotiations since the 1990s gave rise to regional trade agreement or PTAs, with most of them covering CVDs, anti-dumping (AD) and other NTM-related provisions, deviated more from the WTO mandate (Figure B1).¹⁴ Overall, tariffs and NTMs could be considered trade policy substitutes based on the time-series pattern presented in Figure 1, with PTAs enabling more NTMs to coexist with low tariffs. Together with the anecdotal evidence showing that the relationship between them depends on the charac- teristics of importing countries, exporting countries, product and the participation of regional trade agreements, our empirical analysis provide a comprehensive analysis on factors that affect the tendency for governments to use NTMs to replace tariffs. ¹⁴The usage of CVDs and AD notified by member countries to the WTO has been increased as well. 6 Data Source The highly disaggregated tariffs and AVEs of the import-related border NTMs data at importer-exporter-products (HS 6-digit) level comes from Kee and Nicita (2022), reflecting the cross-sectional trade policy pattern of 2018. There are 44 importing countries and 117 exporting countries in our sample, including both developed and developing countries, as well as 5,124 HS-6 digit products. To be clear, due to the lack of comparable NTMs data across countries and over time, this paper uses cross-sectional variations to study the relationship between tariffs and NTMs. Despite missing the temporal variation, the broad coverage of our data across countries and products, coupled with a special focus on border NTMs, enable us to examine the factors influencing the relationship between tariffs and NTMs in a comprehensive approach.¹⁵ Furthermore, most NTMs are regulations and laws which by nature seldom change from year to year, thereby reducing the impact of missing time-series variation on our identification strategy.¹⁶ Figure 1: Overall Trend of the the Tariffs, NTMs and Proliferation of Preferential Trade Agreements Data source: WTO data set on the content of preferential trade agreements (Hofmann et al. 2017), the WTO NTMs Notification Database (Integrated Trade Intelligence Portal, I-TIP) and World Bank Development Indicators. Other data used in this paper are the updated WTO data set on the content of preferential trade agreements (Hofmann et al. 2017), the WTO NTMs Notification Database (Integrated Trade In- telligence Portal, I-TIP), the World Bank World Development Indicators (WDI), GVC Indicators ¹⁵Please refer to WTO (2012) and UNCTAD (2013) for broad surveys of available data in NTMs. ¹⁶Furthermore, some NTMs are implemented for a period, thereby also reducing the impact of missing temporal variations on our identification. For instance, the EU imposed the CVDs on Chinese BEV producers for five years. 7 (Fernandes et al. 2022), CEPII Gravity Database (Conte et al. 2022) and World Governance Indicator (Kaufmann et al. 2011). Appendix A Table A.1 presents the detailed definitions of the variables used in this paper and the corresponding data sources. Import-Related Border NTM Previous literature on investigating the relationship between tariffs and NTMs differs substantially in which type of NTMs studied. Despite of rich heterogeneity, these choices are mostly based on the data-collecting based classification of NTMs (UNCTAD 2015), which encompass mixed impacts on customers, domestic producers and importers that hard to disentangle. In contrast, based on the impact on domestic and international prices, NTMs can be grouped into four groups theoretically according to Ederington and Ruta (2016): customs regulations (alterna- tively, border management policies or border NTMs), process regulations, customer regulations and producer regulations. In this paper, we focus on the import-related border NTMs for three reasons: First, border NTMs only influence foreign producers directly but leave domestic producers unaffected, other than through some general equilibrium effects. These discriminatory NTMs protect domestic industries from import competition, making it reasonable to view them as industrial policies. Second, border NTMs are applied at the customs and may drive a wedge between domestic price and inter- national price. These two features make border NTMs the most comparable alternative trade policy instruments to tariffs. Third, border NTMs are widely used in practice, the CVDs and surtax applied on Chinese BEVs import in the Sino-EU & Canada BEVs disputes both belong to border NTMs. Empirically, the import-related border NTMs mainly consist of SPS measures, TBT measures, CVDs, price control measures and so on. Table B.1 in the Appendix provides a brief composition of these NTMs.¹⁷ By estimating the AVEs of the border NTMs, Kee and Nicita (2022) quantify and convert the trade impacts of border NTMs into AVE terms, making these two theoretically most com- parable policy instruments quantitatively comparable, further allow this paper to study the relationship between tariffs and border NTMs in greater details with precision. 3 Empirical Strategies Equation (1) specifies the baseline empirical model to study the overall relationship between the NTMs and tariffs: ∑ tijn = β1 AV E ijn + δk + εijn , (1) k where tijn and AV E ijn are the effectively applied tariffs and the AVE of border NTMs imposed by im- porting country i on product n from exporting country j , respectively. The highly disaggregated AVE ¹⁷Appendix Table A.1 in Ederington and Ruta (2016) provides the conversion between the NTM classification used by data collecting agencies (UNCTAD 2015) and the theory-based classification proposed themselves in more detail. 8 estimates allow us to control various multi-dimensional FEs, which not only enables us to eliminate omitted variable concerns to the largest extent but also allow us to analyze the relationship between ∑ tariffs and NTMs using different level of variations. k δk denotes different combinations of the FEs to control for different sets of omitted variables, including (1) δi , δj , δn , which denote importer FEs, exporter FEs, and product FEs, respectively; (2) δin , δj , which are importer-product FEs, and exporter FEs; (3) δij , δn , which denote importer-exporter FEs, and product FEs; and (4) δjn , δi , which are exporter-product FEs and importer FEs. εijn is an independent and identically distributed (i.i.d) error term, and the standard errors are clustered at importer-product level.¹⁸ The coefficient of interest is β1 , and tariffs and NTMs are considered policy substitutes if β1 < 0. Conversely, tariffs and NTMs are considered policy complements if β1 > 0.¹⁹ Moreover, to capture the contingent relationship between NTMs and tariffs, which may depend on factors related to importers, exporters and products, as well as any bilateral and multilateral agreements, interaction terms between the AVE and the determinants, Φ, are included in the following empirical specification: ∑ tijn = β2 AV E ijn + β3 AV E ijn × Φijn + δk + εijn , (2) k where Φijn denotes the determinants that may affect the relationship between NTMs and tariffs, including importer characteristics (Φi ), exporter characteristics (Φj ), product characteristics (Φn ), and bilateral characteristics (Φij ). Other variables’ definitions are the same as the baseline regression in equation (1). In this specification, β3 is the coefficient of interest, with a negative value indicating that the interacted variable increases the degree of substitution between tariffs and NTMs, while the converse is true for a positive value.²⁰ ¹⁸There is another specification to study the complementarity or substitutability of tariffs and NTMs which is through a gravity regression, regressing bilateral imports on tariffs, NTMs and their interaction terms. The coefficient on the interaction term captures the moderating effect of AVE in affecting the effectiveness of tariffs in reducing bilateral imports. If the coefficient is positive, it shows that the effectiveness of tariffs in reducing trade flows is tempered by the presence of NTMs, and vice versa. However, in this setting, both tariffs and AVEs are endogeneous variables that each needs an separate IV. In addition, detailed bilateral gravity regressions at HS 6-digit level often run into the problems of missing trade or large presence of zeros, which brings econometric complexity and needs to be addressed carefully (Chen and Roth 2024, Silva and Tenreyro 2006). Finally, while the coefficient on the interaction term captures the effect of NTMs in affecting the effectiveness of tariffs in reducing trade, it does not necessarily show how governments use one policy to complement or substitute the other, which is the void this paper wants to fill. ¹⁹One consideration to have tariffs on the left hand side (LHS) and AVEs on the right hand side (RHS) is that tariffs are actual data while AVEs are estimates with potential measurement errors. Standard errors are clustered to address correlations of tariffs, while measurement errors for AVEs is addressed by using instrument variables. There may not be easy ways to address these problems if AVEs are the LHS variable and tariffs are on the RHS. ²⁰As a robustness check, we also add indicators of other types of NTMs (i.e., process regulations, customer regulations and producer regulations) as additional control variables in the regression. The results, available among request, remain qualitatively the same and have no significant differences in quantitative magnitude. Nevertheless, Kee and Nicita (2022) already take the impact of other types of NTM into consideration when estimate the AVE of the border NTM. As a result, the AVE is not correlated with other NTMs even if the latter is in the error term, therefore avoiding omitted variable bias. 9 3.1 Ad-Valorem Equivalent To study the relationship between tariffs and NTMs, previous literature have used NTM indicator variables, importer-product level AVE estimates, and various NTM incidence indicators including coverage ratio (the share of imports subject to NTMs), and frequency ratio (the percentage of traded products subject to NTMs). Compared to the literature, the highly disaggregated AVEs at importer-exporter-HS 6-digit prod- uct level used in this paper from Kee and Nicita (2022) have the following advantages: First, they better capture the restrictiveness of NTMs in curbing import, which reduces measurement errors, without assuming that NTMs affect all exporting countries equally. Specifically, given an importing country and product, the NTM incidence indicators assume that all NTMs are equally binding in affecting trade flows across all exporting countries. However, due to the lack of bureaucratic capacity or gov- ernance of the importing countries, the presence of some NTMs may not have significant impact on imports, which may result in low AVE estimates despite the large NTM incidence. In addition, while some NTMs may be importing country and product specific, the restrictiveness on import may depend on the compliance costs and capability of the exporting countries, leading to variations of AVEs by exporting countries.²¹ Second, by converting the qualitative NTMs into the quantitative AVEs, not only is the impact of NTMs comparable across different types of NTMs within the category of border NTMs, it is also comparable to tariffs, both in terms of measurement as well as a trade policy to curb imports. Third, the resulting AVE estimates indeed show a wide range of variations, suitable to study the empirical relationship between border NTMs and tariffs. 3.2 Instrumental Variables We use instrumental variables (IVs) to address omitted variable bias and measurement error of AVEs. Finding a suitable IV at the highly disaggregated level is challenging. Following Kee and Nicita (2022), AV E ijn and AV E ijn Xijn are instrumented using the average AVE of exporting country j of the product n in non-i markets AV E −ijn and AV E −ijn Xijn , respectively. To justify that the AVE faced by the same exporter and product in the other market is a valid IV for the AVE of the importing country i, we need to show that: First, the average AVE of other markets is correlated with the AVE of the importing country, given exporter and product. Second, the average AVE of the other markets is not correlated with the tariffs of the importing country, given the same exporter and product. Specifically, given the same exporter and product, the AVE in the other markets is likely to be correlated with the AVE of the importing country. This is because some NTMs are influenced by ²¹Importer-product specific AVE estimates assumed away the exporter variation which may limit the effectiveness of such variable in empirical analysis. By interacting market sizes of importing countries and exporting countries with the border NTM dummy variable in product-level gravity regressions, Kee and Nicita (2022) estimate and quantify trade impacts of NTMs into AVE terms with variation across importer-exporter-HS 6-digit product 10 both exporter factors and product characteristics, and are not importing country specific. This means that for certain products exported by specific countries, NTMs are necessary to be applied irrespective of the importers, which makes the AVE of the other markets not a weak instrument. In addition, AV E −ijn and tijn are not the trade policies of the same country, which implies that the AVE of other countries may not correlate a priori with the tariffs of the importing country, thereby making the former justifies the exclusive restriction of the instrumental variable. Nevertheless, even though we use the instrumental variables approach to solve the potential measurement errors of the AVEs and omitted bias, the relationship revealed from our empirical setup is correlations instead of causality. 4 Empirical Results 4.1 Baseline Results: The Overall Relationship between Tariffs and NTMs Table 1 presents the both first-stage and the second-stage baseline instrumental variable regression results for the overall relationship between the restrictiveness of NTMs and tariffs, as specified in equation (1).²² In Column (1), tariffs is regressed on the instrumented AVEs, controlling for the importer FEs, exporter FEs and product FEs. The negative and significant coefficient suggests that, the restrictiveness of NTMs and tariffs are negatively correlated, implying that, overall, NTMs and tariffs are policy substitutes. While the magnitude of the coefficient is not as important as the sign, it suggests that a 10 percentage point increase in the restrictiveness of NTMs (AVE) is associated with a 0.8 percentage point reduction in tariffs. The high first-stage F-statistic further suggests that the AVE of the other markets faced by the same exporter and product has enough explanatory power and is not a weak instrument for the AVE of the importing country. Column (2) presents the regression result when we regress tariffs on the instrumented AVEs, controlling for the importer-product FEs and exporter FEs. The coefficient on AVE is still significantly negative, which implies that, given the same importer and product, the exporters that enjoy lower tariffs tend to face more restrictive NTMs. This again suggests that the NTMs and tariffs are trade policy substitutes. This result is consistent with the observation that when an importing country gives a tariff reduction on a product to an exporting country, the preferential access may come with restrictive NTMs. One example of this is the EBA of the EU, where restrictive rules of origin requirements are necessary in exchange for the duty-free access to the EU market by the least developed countries, while other countries do not have such an arrangement and face higher tariffs. Column (3) presents the regression result controlling for the importer-exporter FEs and product FEs. The coefficient of AVEs remains negative, implying that, given two trading countries, products that face lower tariffs tend to have restrictive NTMs. This would be the case in a trade agreement, such ²²In an alternative specification where we investigate the relationship between overhang and AVE of the border NTM, the results suggest that importing countries having flexibility in using tariffs tend to imply less restrictive NTMs, which are in line with Kuenzel (2020) and Beshkar et al. (2015). The results are available upon request. 11 as the US-Peru Free Trade Agreement, where US tariffs on agricultural products will be eliminated provided that NTMs on the agricultural sector, such as the environmental protections are meet. Column (4) shows the regression result when we control for the exporter-product FEs and im- porter FEs. The significantly positive coefficient of AVEs suggests that, when comparing trade policy stances across importing countries, for a specific product from a specific exporting country, importing countries that have higher tariffs tend to impose more restrictive NTMs, while importing countries that have lower tariffs have less restrictive NTMs. This positive relationship reflects the overall trade policy environment and stance of the importing countries which determines the usage of both tariffs and NTMs. Countries that are in favor of trade protectionism are likely to impose both restrictive NTMs and high tariffs simultaneously relative to those more open countries that have lower tariffs and NTMs. One example of this scenario is India, which has high tariffs and restrictive NTMs on all products from all countries, compared to Singapore, which has no tariffs and less restrictive NTMs.²³ Therefore, the positive coefficient in Column (4) is not in conflict with the first three columns in Table 1. Table 1: Baseline Results Panel A: Second Stage (1) (2) (3) (4) Dependent Variables Tariff Tariff Tariff Tariff AVE of Border NTM -0.086*** -0.058*** -0.046** 0.050*** (0.020) (0.008) (0.019) (0.009) Importer Fixed Effect Yes Yes Exporter Fixed Effect Yes Yes Product Fixed Effect Yes Yes Importer-Product Fixed Effects Yes Importer-Exporter Fixed Effects Yes Exporter-Product Fixed Effects Yes Panel B: First Stage Other AVE 0.553*** 0.560*** 0.555*** -27.608*** (0.018) (0.017) (0.018) (0.587) F Statistics 958.12 1148.18 969.09 2221.86 Observations 1,003,854 991,995 1,003,626 923,096 Note: Robust standard errors in parentheses are clustered by importer-product. *, ** and *** indicate that coefficients are significant at 90%, 95% and 99%, respectively. AVEnij is instrumented using the average AVE of exporter j of the product n in non-i markets. ²³The effectively applied tariff weighted average (customs duty) for Singapore is 0% and the most favored nation (MFN) weighted average tariff is 0.16%. (Data source: https://wits.worldbank.org/CountryProfile/en/SGP) 12 4.2 Further Analysis: What Factors Determine the Relationship between NTMs and Tariffs As aforementioned, the highly disaggregated data at importer-exporter-HS 6-digit product level enables us to control different combinations of FEs, and to identify the effects of various moderating factors on the relationship between tariffs and NTMs. In this section, we focus on the last three FE combinations in Table 1. 4.2.1 Given Exporter-Product, Which Importer Characteristics Matter? Figure 2 shows the results of regressions specification (2), controlling for exporter-product FEs and importer FEs. Each row represents an individual regression, showing the coefficients and confidence intervals (90%) of AVE and its interaction with the variables on the left-hand side, which are the individual importer characteristics introduced in a separate, consecutive manner. Figure 2: Given Exporter-Product, Which Importer’s Macroeconomic Characteristics Matter? Notes: Each row represents a regression, showing the coefficients and confidence intervals (90%) of AVE and its interaction with the variable on the left-hand side, controlling for exporter-product FEs and importer FEs. The results show that, given the same exporter-product, the degree of substitution between tariffs and NTMs increases with the GDP per capita, capital-labor ratio or skilled-labor ratio of the import- ing countries. This could be because these high-income developed countries have lower tariffs and overhang, and at the same time value public health, safety and the environment to a larger extent, as well as to protect and reshape domestic industries.²⁴ To promote these domestic objectives, or to ²⁴High-income developed countries usually have lower overhang compared to developing countries, which restrict their flexibility of raising tariffs without retaliation or violating WTO regulations (Lorz and Thede 2024). 13 minimize the negative externalities of imports, or to disguise trade protective or industrial motives, these countries may resort to more restrictive NTMs, leading to a stronger substitution between tariffs and NTMs. We will test the hypothesis that high-income countries value public health, safety and the environment more than low-income countries in Section 5.2. These results are also consistent with the finding of Juhász et al. (2022) that high-income countries use more industrial policies to restructure domestic industries. The determination of trade policies could also depend on the political-economy factors and gov- ernance of the importing countries (Grossman and Helpman 1994). As emphasized by Ruckteschler et al. (2022), since NTMs are more complicated and less tangible compared to tariffs, the effect of NTMs is highly dependent on the institutional quality and implementation efficiency of the gov- ernment administration. Similar set up as Figure 2, results presented in Figure 3 show that, given exporter-product, the degree of substitution between tariffs and NTMs is higher for importers with better control of corruption, more effective government, better rule of law and higher regulatory quality. This could be because importing countries with these characteristics are better in enforcing NTMs, given that NTMs are more complicated and harder to put into force than tariffs. Figure 3: Given Exporter-Product, Which Importer Institutional Characteristics Matter? Notes: Each row represents a regression, showing the coefficients and confidence intervals (90%) of AVE and its interaction with the variable on the left-hand side, controlling for exporter-product FEs and importer FEs. Figure 4 explores the impacts of the GVC participation of the importing countries on the rela- tionship between tariffs and NTMs. Specifically, we use three variables from Fernandes et al. (2022) 14 to measure the importing countries’ GVC characteristics: backward GVC participation, forward GVC participation and the total distance to the three GVC hubs (i.e., the US, Germany, and China). The first two measurements capture the import content in the importing countries’ exports and the domestic value added in importing countries’ exports, respectively. The results show that, for importing countries that are deeply integrated into the GVC, the relationship between tariffs and NTMs tends to be more complementary, which means both tariffs and NTMs are lower relative to other countries that are less engaging in GVCs. These results are consistent with the later finding in Section 4.2.3 that intermediate products and capital products generally face more liberal trade policies since these products are widely embedded in GVC trade. Figure 4: Given Exporter-Product, Which Importer’s GVC Characteristics Matter? Notes: Each row represents a regression, showing the coefficients and confidence intervals (90%) of AVE and its interaction with the variable on the left-hand side, controlling for exporter-product FEs and importer FEs. 4.2.2 Given Product, Which Importer-Exporter Characteristics Matter? Similar to Figure 2, Figure 5 shows the empirical results of regression specification (2) controlling for importer-exporter FEs and product FEs, with the interaction terms capturing the impact of bilateral factors on the relationship between the restrictiveness of NTMs and tariffs. The first two rows of the Figure 5 show that whether both importing countries and exporting countries are WTO members or not has no significant impact on the relationship between the NTMs and tariffs. In sharp contrast, when the two countries are engaged in the deep trade agreement (DTA), the relationship between NTMs and tariffs become more substitutive. As pointed out by Ederington and Ruta (2016), the GATT and the following WTO restrict countries to negotiate over the NTMs 15 to prevent from trade policy substitution. However, the DTAs contain both tariff reduction and the increasing coverage of NTMs, including the prohibition, implementation, or harmonization of NTMs (WTO 2023). In addition, the third and fourth rows of Figure 5 further investigate the impact of DTAs’ horizontal depth, measured by the number of provisions and the number of legally enforceable provisions contained in a DTA, following Hofmann et al. (2017). The results are in line with the second row result.²⁵ Overall, Figure 5 suggests that, relative to the GATT or WTO, the degree of substitution between tariffs and NTMs is higher under deeper bilateral or multilateral trade agreements. Figure 5: Given Product, Which Importer-Exporter Characteristics Matter? Notes: Each row represents a regression, showing the coefficients and confidence intervals (90%) of AVE and its interaction with the variable on the left-hand side, controlling for importer-export FEs and product FEs. 4.2.3 Given Importer-Exporter, Which Product Characteristics Matter? Figure 6 shows the empirical results of regression specification (2) controlling for importer- exporter FEs and product FEs, when we focus on the impact of product characteristics on the re- lationship between the tariffs and NTMs for the same importer-exporter pair. The results show that tariffs and NTMs are policy complements for capital goods and interme- diate products. These products are more deeply embedded in supply chains, and generally face more complementary trade policies overall: On the one hand, with lower tariffs and less restrictive NTMs to promote supply chains allocate across the national borders. The aforementioned anecdotal evidence ²⁵The magnitude of the coefficients of the interaction term between DTA, DTA depth and AVE of the border NTM differ because, deep trade agreement is a dummy variable which takes the value of one if the importing country and exporting country are engaged in a trade agreement while the DTA depth is the count of the number of provisions. 16 that the semi-conductors and all local area network apparatus entering the US face no tariff and no restrictive NTM would be an example of the complementary relationship of the trade policy due to the embeddedness of products in supply chain. On the other hand, the recent trade disputes that the US imposing both tariffs on BEV’s component such as battery and semi-conductors and restrictive NTMs on critical components from China or Russia, stands on the opposite. To develop domestic supply chains and enhance national security on critical products, higher tariffs and restrictive NTMs are use concurrently as trade policy complements. On the contrary, the degree of substitution between tariffs and NTMs is higher for consump- tion products, agricultural products²⁶ and food products.²⁷ One possible explanation could be that consumption products, agricultural products and food products may have consumption externalities, which could be addressed by more restrictive NTMs, given constrained tariffs, especially for high- income countries. Note that some agricultural products, such as forestry, are also climate intensive goods. This finding suggests that governments may use NTMs to regulate the trade of climate in- tensive goods which may contribute to the global climate goals. This result collaborates well with the finding of Harstad (2024a). We will test the hypothesis that agricultural and food products have larger consumption externalities in Section 5.2. Figure 6: Given Importer-Exporter, Which Product Characteristics Matter? Notes: Each row represents a regression, showing the coefficients and confidence intervals (90%) of AVE and its interaction with the variable on the left-hand side, controlling for importer-exporter FEs and product FEs. ²⁶See https://unstats.un.org/wiki/display/comtrade/HS+2002+Classification+by+Section for the UN industry classifi- cation. Section 1-4 (HS 2-digit: 1-24), are defined as agriculture sectors. ²⁷The identification of food and beverage products follows the definition of U.S. Department of Agriculture (USDA). Please see https://www.ers.usda.gov/data-products/u-s-food-imports/documentation/ for the detailed HS code list. 17 4.2.4 Given Importer-Product, Which Exporter Characteristics Matter? Figure 7 displays the empirical results of regression specification (2), controlling for importer- product FEs and exporter FEs, when we focus on the impact of the exporter’s macroeconomic char- acteristics on the relationship between the tariffs and NTMs. The first two rows of Figure 7 show that, the lower GDP per capita and capital-labor ratio of the exporters are associated with a higher degree of substitution between tariffs and the NTMs. This could be because many low-income labor abundant exporting countries have preferential tariffs in the high income destination markets. In order to protect domestic market from import competition driven by their low-wage exports, importing countries may impose restrictive NTMs, which leads to the substi- tution between tariffs and NTMs. Conversely, this result also confirms that high-income exporting countries will face lower AVEs for a given importer-product market. This finding is consistent with Atkin et al. (2024) showing that high-income exporting countries hold market power which leads to less restrictive NTMs. Finally, the skilled labor ratio of the exporting countries has no impact. Figure 7: Given Importer-Product, Which Exporter’s Macroeconomic Characteristics Matter? Notes: Each row represents a regression, showing the coefficients and confidence intervals (90%) of AVE and its interaction with the variable on the left-hand side, controlling for importer-product FEs and exporter FEs. Similarly, results in Figure 8 show that, given importer-product, the degree of substitution be- tween tariffs and NTMs is higher for exporters with better control of corruption, more effective gov- ernment, better rule of law and higher regulatory quality. This could be because exporting countries with these characteristics are better in complying NTMs with supporting domestic regulations, given that NTMs are more complicated and harder to put into force than tariffs. 18 Figure 8: Given Importer-Product, Which Exporter’s Institutional Characteristics Matter? Notes: Each row represents a regression, showing the coefficients and confidence intervals (90%) of AVE and its interaction with the variable on the left-hand side, controlling for importer-product FEs and exporter FEs. Likewise, the results in Figure 9 show that, for exporting countries that are deeply integrated into the GVC, the relationship between tariffs and NTMs tends to be more complementary, which means both tariffs and NTMs are lower relative to other exporting countries that are less engaging in GVCs. Together with the findings that GVC products such as capital goods and intermediate goods and importing countries deeply embedded in GVCs, face both lower tariffs and less restrictive NTMs, our results sugguest that GVC is associated with liberalization in both tariffs and NTMs. Overall, the reduced-form regression results suggest that the degree of substitution between tar- iffs and NTMs depends on the characteristics of the importing countries, exporting countries and products. For the rest of the paper, we will develop a simple terms-of-trade model to explain how dif- ferent characteristics may affect the substitution between tariffs and NTMs. Furthermore, structural estimation of the model parameters will allow us to relate the model to some of the main findings of the reduced-form regressions. Next, we will apply this simple model to analyze the Sino-EU trade disputes, whereby the EU imposes NTMs on top of tariffs to reduce the imports of Chinese BEVs. After that, we will discuss the possible policy considerations and rationale behind the different trade policy action in response to import competition from Chinese BEV sector. 19 Figure 9: Given Importer-Product, Which Exporter’s GVC Characteristics Matter? Notes: Each row represents a regression, showing the coefficients and confidence intervals (90%) of AVE and its interaction with the variable on the left-hand side, controlling for importer-product FEs and exporter FEs. 5 Model and Structural Estimation This section presents a general equilibrium terms-of-trade model to show how different country and product characteristics may determine the policy relationship between tariffs and NTMs, to ra- tionalize the empirical findings that high-income importing countries tend to substitute tariffs with restrictive NTMs, particularly on products with consumption externalities. The model is also used to provide guidance for simple structural estimations to extract some deep parameters in order to shed lights on some of the results of the reduced-form regressions. 5.1 Model Framework and Trade Policy Determination This model is based on the insights of Ederington (2001), which has two goods (X and Y ) and two countries (Home and Foreign(*)) with welfare maximizing governments. In Ederington (2001), governments choose tariffs and domestic production taxes to extract terms-of-trade gains and to reduce the negative production externality of good X , such as pollution. In our model, the governments impose tariffs and AVEs of NTMs to extract terms-of-trade gains and to lower the consumption externality of X , such as the public health crisis due to pesticide residuals in food products and the worsening quality of lives due to congestion with more cars on the road. Additionally, AVEs also enter the social welfare function directly through the negative consumption externality function, E (·) > 0, through boosting public confidence that the government is combating negative externalities. Finally, 20 the consumption externality function enters the social welfare function with a positive weight, θ, that depends on the characteristics of the countries.²⁸ Specifically, both X and Y are produced in Home and Foreign, with strictly concave and downward sloping production possibility frontier (PPF): Y = F (X ); Y ∗ = F ∗ (X ∗ ). (3) For each good n = {X, Y }, domestic consumption, Cn is the sum of domestic production n, and net import, Mn .²⁹ Assuming that each country has identical citizens and the representative citizen’s welfare function, which is also the social welfare function faced by the government, is quasi-linear with respect to the quantity of each good consumed and the negative consumption externality, E : W = CY + U (CX ) − θE (CX , AV E ), (4) ∗ W ∗ = CY ∗ + U ∗ ( CX ∗ ) − θ ∗ E ∗ ( CX , AV E ∗ ), where θ > 0 is the weight of the negative consumption externality in the social welfare, or the marginal welfare impact of the externality. A large θ indicates that the representative citizen or the government cares more about reducing negative consumption externality to improve the social welfare. The consumption externality function, E (·), depends positively on CX , and negatively on AV E . Let’s define: ∂E λ ≡ > 0, (5) ∂CX ∂E ϕ ≡ − > 0, (6) ∂AV E where λ is the marginal externality from consuming X , and ϕ is the effectiveness of the AVE of NTMs on directly reducing the negative consumption externality. The reason AVE may directly reduce the negative consumption externality and improve social welfare is that the public confidence is boosted when the government actively use AVEs to address consumption externalities.³⁰ The magnitude of ϕ depends on the governance and institutional quality of the importing countries in enforcing the rules and regulations related to NTMs. It also depends on the capability of the exporting countries in complying with NTMs. Finally, products with stronger consumption externalities, such as food or agricultural products, may have higher λ and require more restrictive NTMs. ²⁸E (·) < 0 could depict positive externality that discussed in the case of the Sino-EU BEVs dispute in Section 6. ²⁹For simplicity, we use X and Y to denote the domestic production of product X and Y , respectively. ³⁰Disdier et al. (2015) made a similar assumption with the utility of the representative agent affected by a negative externality function, which depends on standard-like NTM policies. Similarly, Costinot (2008), Essaji (2010) and Fischer and Serra (2000) also present a model that consumption is associated with negative externality (pollutants, specifically) while the use of NTMs (product standard, specifically) can reduce the negative externality. 21 Let pw and pd denote the relative world and domestic prices of X , respectively. And let X be the natural imported good for Home. The Home government imposes tariff, t and AVE on X , which drive a wedge between pw and pd . Likewise for the Foreign government on Y : pd = (1 + t + AV E )pw , (7) pd∗ = pw /(1 + t∗ + AV E ∗ ). (8) In equilibrium, profit maximization will ensure that the relative domestic price of X equals the marginal rate of transformation between the two goods, which is the absolute value of the slope of the production possibility frontier (PPF), while utility maximization will also equate relative domestic price of X to the marginal rate of substitution between the two goods: pd = − F ′ ( X ) = U ′ ( CX ) . (9) Tariff revenue in Home collected from the net import of X is lump-sum redistributed back to the representative consumer, such that the equilibrium CX = X + MX is a function of pw , t and AV E . The relative world price, pw , is determined by the market-clearing condition that net imports of the home country of each good are equal to foreign country’s net exports: ∗ Mn = − Mn . (10) Let m denote direct lump-sum transfers (in terms of the numeraire good), then balance of payment requires that for any world price: MY + p w MX + m = 0 , (11) ∗ ∗ MY + p w MX − m = 0. (12) Equation (4) can be rewritten all in terms of X : W = F (X ) − pw MX − m + U (X + MX ) − θE (X + MX , AV E ). (13) Home government chooses t and AV E to maximize social welfare as specified in equation (13), which leads to the following two first-order conditions: 22 ( ) ( ) ∂W ′ ∂X∂pw w ∂MX ′ ∂X ∂MX ∂X ∂MX = F − MX − p +U + − θλ + = 0, (14) ∂t ∂t ∂t ∂t ∂t ∂t ∂t ∂t ( ) ∂W ′ ∂X ∂pw w ∂MX ′ ∂X ∂MX = F − MX − p +U + ∂AV E ∂AV E ∂AV E ∂AV E ∂AV E ∂AV E ( ) ∂X ∂MX −θλ + + θϕ = 0. (15) ∂AV E ∂AV E Substituting equations (7), (8), (9) and (10) into equation (14), we have the following:³¹ ( ) 1 θλ ∂X /∂t t + AV E = + +1 , (16) ϵ pw ∂MX /∂t ∗ ∗ ∂ ( − MX ) pw pw ∂MX /∂t ϵ ≡ ∗ = ∗ >0 (17) ∂p w ( − MX ) w MX ∂p /∂t where ϵ is the elasticity of Foreign country’s supply of net exports.³² Consider when there is no consumption externality (λ = 0) or when the country does not care about the externality (θ = 0). In this case, the welfare-maximizing optimal tariff of Home is 1/ϵ when there are no restrictive NTMs. This is the standard result based on the terms-of-trade effect of tariff: facing an upward-sloping foreign export supply curve, the optimal tariff is positive given that tariffs reduce world prices. However, if the tariff is restricted to be less than 1/ϵ due to trade agreements, then the welfare maximizing government will impose restrictive NTMs, such that the sum of tariff and AVE equals 1/ϵ. This shows that overall tariffs and AVEs are substitutes, even in the absence of externality and the relationship is completely driven by the terms-of-trade effects of tariffs and NTMs. Conversely, when there is consumption externality (λ > 0) and the country does care about the externality (θ > 0), the optimal tariff will be larger than 1/ϵ when there are no restrictive NTMs. This is because, in addition to the terms-of-trade effect of suppressing world price, the higher tariff is also used to curb the negative externality through depressing domestic consumption.³³. If tariff is lower than the optimal level, then welfare maximizing government will impose restrictive NTMs with positive AVEs which indicates policy substitutions. The degree of substitution between NTMs and tariffs depends on θ, λ and ∂CX /∂t. Similarly, substituting equations (7), (8), (9) and (10) into equation (15), we have the following relationship depicting the optimal AVE: ³¹The derivations are available upon request. ∗ ³²To see this, we start from the market equilibrium condition, equation (10): MX (t) = −MX ( pw ) > 0 ∗ ∗ ∗ ∗ ∂ (− MX ) ∂ (− MX ) ∂pw ∂ (− MX ) ∂ (− MX )/∂t ∴ ∂t = ∂t = ∂pw ∂t ⇒ ∂pw = ∂pw /∂t . ∂MX ∂CX ∂pd ³³By definition of CX , we have ∂CX ∂t = ∂X ∂t + ∂MX ∂t ⇒ ∂pd ∂t = ∂X ∂t + ∂MX ∂t ⇒ ∂CX w ∂pd p = ∂X ∂t + ∂MX ∂t <0⇒ ∂X /∂t ∂MX /∂t + 1 > 0. 23 ( ) 1 θλ ∂X /∂AV E θϕ t + AV E = + w +1 − w . (18) ϵ p ∂MX /∂AV E p ∂MX /∂AV E Note that, by definition of AV E , ϵ in equation (18) is same as the export supply elasticity of ∂MX ∗ /∂t ∂MX∗ /∂AV E Foreign in equation (17), since ∂pw /∂t = ∂pw /∂AV E .³⁴ If the government does not care about reducing externality (θ = 0), or if NTMs are completely ineffective in curbing externality (ϕ = 0) and there is no externality (λ = 0), then the optimal AV E is the reciprocal of foreign export supply elasticity due to the terms-of-trade effect of NTMs if tariff is zero. If λ, θ and ϕ are all positive, the optimal AV E will be higher than 1/ϵ, and it is also higher than the optimal tariff with externality as in equation (16). This is because in addition to reducing world price (terms-of-trade effect), and reducing consumption (curbing externality effect), AV E also directly improves social welfare through ϕ (boosting public confidence). So, the optimal level of AV E will be higher than tariff. The degree of substitution between tariffs and NTMs depends on θ, which is the weight of the consumption externality in social welfare. A higher θ indicates the importing country values the re- duction of consumption externality more, and therefore will have more incentives to impose restrictive NTMs. These are likely to be the high income developed countries, which tend to be more capital and skilled-labor abundant. Conversely, developing countries which have lower income and are more labor abundant, will be less inclined to impose restrictive NTMs. In addition, the degree of substitution between tariffs and NTMs also depends on λ and ϕ, which are the marginal externality from consuming product X , and the effectiveness of NTMs in reducing consumption externalities and boosting public confidence. Consumption products and agricultural products are more affected by NTMs and consumption externalities, which causes their λ to be larger and leads to higher trade policy substitution. Importing countries with good governance or better institutions will be more able to enforce the rules and regulations related to NTMs. For these countries, their ϕ will be larger, which implies that there will be more use of restrictive NTMs. On the contrary, small developing exporting countries that are not capable of complying with NTMs will have smaller ϕ, which may reduce the substitution between tariffs and NTMs. Equations (16) and (18) jointly imply that ∂X /∂AV E ∂X /∂t ϕ/λ = + . (19) ∂MX /∂AV E ∂MX /∂t ∂MX /∂AV E In words, equation (19) suggests that for the optimal tariff and AVEs to be both positive, it is necessary that tariffs and NTMs are imperfect substitutes, since (ϕ/λ)/(∂MX /∂AV E ) < 0:³⁵ ∂M ∗ w pw ∂M ∗ /∂AV E pw ∂M ∗ /∂AV E pw ∂M ∗ /∂t ³⁴By definition, we have ϵ ≡ ∂pwX p MX∗ = M ∗ ∂pw /∂AV E , therefore we have M ∗ ∂pw /∂AV E = M ∗ ∂pw /∂t . X X X X X X ³⁵Please note that tariffs and NTMs are imperfect policy substitutes is the result of our model. This is unlike the existing papers, such as Ederington (2001) which assumes trade and domestic policy are imperfect substitutes. 24 ∂MX /∂AV E ∂X /∂AV E < . (20) ∂MX /∂t ∂X /∂t Thus, given the marginal consumption externality, λ > 0, and the effectiveness of AV E in reducing externality, ϕ > 0, for the optimal tariff and AVE to be both positive, it is necessary that NTMs are more effective in protecting domestic production while tariffs are more effective in curbing imports. In addition, equation (19) implies that it will be optimal to use higher tariff to curb imports ∂X /∂t if ϕ is higher, since ∂M X /∂t will be higher. Conversely, it will be optimal to have more restrictive ∂X /∂AV E NTMs if λ is higher, since ∂MX /∂AV E will be higher. For any given tariff level, t, the optimal AV E , according to (18) would be: ( ) 1 θλ ∂X /∂AV E θϕ AV E = −t+ w +1 − w , (21) ϵ p ∂MX /∂AV E p ∂MX /∂AV E ( ) θλ ∂X /∂AV E θϕ 1 = w +1 − w if t = . (22) p ∂MX /∂AV E p ∂MX /∂AV E ϵ Equation (21) depicts the substitutive relationship between t and AV E in the equilibrium. In the absence of any externality (λ = 0 and ϕ = 0), or if the government does not care about reducing consumption externality (θ = 0), equation (22) shows that the optimal AV E is zero when t = 1/ϵ. In contrast, in the presence of the negative consumption externality (λ > 0), the optimal AV E is positive, if the government cares about reducing such externality (θ > 0). The level of AV E will be higher if AV E is very effective in reducing consumption externality (ϕ > 0). This is true even if the marginal externality of consumption is zero (λ = 0) as long as θ > 0 and ϕ > 0 , as higher AV E boosts public confidence in the government in addressing externality.³⁶ However, if some existing trade agreements restrict tariffs such that t < 1/ϵ (i.e. lower than the optimal tariff level), then the optimal AV E is shown in equation (21), which is higher than equation (22). Governments have to resort to NTMs to capture some fraction of the terms-of-trade gains that are not realized when setting t < 1/ϵ. Likewise, if AV E is restricted below the optimal level due to some provisions of a trade agreement, then the government will have an incentive to raise tariff higher than the optimal tariff level, in order to reduce consumption externality. As a result, both scenarios will generate a substitutive relationship between tariffs and NTMs. ³⁶The preconditions of θ > 0 and ϕ > 0 have two implications: (1) the public cares about the consumption externality; (2) the enforceability and effectiveness of NTMs regulated by the government boost public confidence and raise the welfare as well. For instance, think of a situation when regular inspections at customs are applied to all imported products, some of them may not have negative consumption externality. However, the inspection of these products boost confidence of consumers and serve as a official endorsement for the quality and safety of the product. 25 5.2 Structural Estimation To estimate the model structurally, we start with equation (16), which is one of the first-order condition. Rearrange the terms and using the definitions of ϵX and ϵM , we will have: ( ) 1 θλ θλϵX X (t + AV E ) = + + (23) ϵ pw p w MX ϵ M ∂M ∗ w d d where ϵ ≡ ∂pw X p MX∗ ,ϵ X ≡ ∂p ∂X p d X,ϵ M ≡ ∂M X p ∂pd MX . Note that θ is the weight of externality in the social welfare function, which is assumed to be importing country specific. The marginal externality of consumption, λ, depends on the product and externality in social welfare function, so we assume it to be importer-product specific. The domestic supply elasticity of X , ϵX , is importer-product specific. We can therefore estimate equation (23) based on FEs regressions, regressing (t + AV E ) on 1/pw , where pw is the relative unit value of import, and 1/(pw MX ϵM ), with pw MX equals the value of imports of X and ϵM is the import demand elasticity from Kee and Nicita (2022). Table 2 provides the summary of variables and parameter used in the following structural estimation.³⁷ Table 2: Summary of the Variables and Parameter used in the Structural Estimations Variable Definition Level t Tariff Importer-Exporter-Product AVE Ad Volrem Equivalent Tariff of the Border NTMs Importer-Exporter-Product pw Relative Price (Trade Unit Value) Importer-Product p w MX Import Value Importer-Exporter-Product ϵM Import Demand Elasticity Importer-Exporter-Product Parameter Definition Level ϵ (Foreign) Export Supply Elasticity Exporter-Product θ The Weight of Negative Externality in the Social Welfare Importer λ The Marginal Consumption Externality Importer-Product ϵX X (Home) Domestic Supply Elasticity Multiply with Domestic Production Importer-Product With these variables in hand, we will be able to retrieve θ, and λ from the estimation of β ’s, based on the following specifications: ( ) [ ] 1 1 1 tijn + AV Eijn = + βin w + γin + εijn (24) ϵjn pn (pw MX )ijn ϵM ijn where i, j, n denote the same as in Section 3, ε is the error term, and ³⁷The unit of measurement is adjusted to be the same within each product. The unreasonable trade unit values are dropped following the criteria proposed by Kee and Nicita (2022). 26 βin = θi λin (25) γin = θi λin ϵX in Xin . (26) ( ) Equation (24) shows that ϵjn can be absorbed by exporter-product FEs, and βin p1 w can be n proxied by importer-product FEs. In addition, γin can be obtained by interacting importer-product FEs with the inverse of the product of import value and import elasticity (pw MX )ijn ϵM ijn , which is available in Kee and Nicita (2022). [ ] 1 tijn + AV Eijn = αjn + αin + γin + εijn , (27) (pw MX )ijn ϵM ijn 1 αjn = , (28) ϵjn αin pw n = θi λin , (29) γin /(αin pw X n ) = ϵin Xin . (30) To obtain θi and λin , we regress the log of αin pw n on a full set of importer FEs, and the exponents of the importer FEs are θi : ln(αin pw n ) = ln(θi ) + ln(λin ) (31) ln(αin pw n ) = αi + ϵin , (32) θi = exp(αi ), (33) λin = αin pw n /exp(αi ). (34) Table 3 presents the regression result for the structural estimation, with the appropriate sets of FEs according to equation (27). The coefficient of the right-hand side variable is positive and highly significant, which is consistent with the model. Table 3: Structural Estimation VARIABLES Tariff + AVE Inverse of Imports × Elasticity 0.058** (0.025) Importer-Product Fixed Effects Yes Exporter-Product Fixed Effects Yes Observations 244,380 Adjusted R2 0.792 Note: Robust standard errors in parentheses are clustered by importer-product. *, ** and *** indicate that coefficients are significant at 90%, 95% and 99%, respectively. 27 Using the estimated results, θi and λin are constructed based on equations (33) and (34). All the estimated θi and λin are positive. 5.3 Relating Structural Estimation to Reduced-Form Results The applicability of our model can be verified by using the estimated parameters to explain the previous reduced-form results. One of the main results in Section 4 is that the degree of substitution between tariffs and NTMs increases with the income level of the importing countries (see Figure 2). On the other hand, our model suggests that the degree of substitution between tariff and NTMs increases with θ, which is the weight of consumption externalities in the social welfare function. For the model to be consistent with the reduced-form results, it is necessary that θ increases with the income level of the importing countries. Column (1) of Table 4 shows the regression result when we regress the estimated θi on the GDP per capita of the importing country. The positive and statistically significant coefficient suggests that θi increases with the income level of the importing country, controlling for country size and the share of import duty in total revenue. This is consistent with our previous reduced-form regression result. Figure 10 further presents the positive and significant partial correlation between the estimated θi and GDP per capita of the importing countries, confirming both the theoretical and empirical findings. Reduced-form regression results from Section 4 also show that the degree of substitution between tariffs and NTMs is higher for agricultural and food products (see Figure 6). In the model, product characteristics that may affect the substitution between tariff and NTMs are captured by λin , which is the marginal consumption externalities. For the model to be consistent with the reduced-form results, λin should be higher for these products, particularly if there are fewer regulations in the importing countries to reduce consumption externalities. Columns (2) and (3) of Table 4 show the regression results when we regress the estimated λin on product characteristics, controlling for importer FEs. The results show that agricultural and food products have statistically larger λin , suggesting higher marginal consumption externalities for these products. Given that forestry products are included in agricultural products, this result is consistent with Harstad (2024a), highlighting how governments may mix tariffs and NTMs to regulate the trading of products with environmental externality to achieve the first-best goal of combating climate change. Furthermore, the negative coefficients on the interaction terms between GDP per capita of the importing countries and agricultural or food products indicators, suggest that the marginal consump- tion externalities of agriculture and food products decrease with the income level of the importing countries. This could be because high-income countries have better regulations and safeguards in place that already reduce the consumption externalities of these products. Overall, the results based on the structural estimation of the theoretical model presented in Table 4 are consistent with the previous reduced-form empirical findings. Together, these results show that 28 while tariffs and NTMs are policy substitutes, the degree of substitution depends on the characteristics of the importing countries, exporting countries, and products. Table 4: The Relationships between Importing Country, Product Characteristics and Parameters (1) (2) (3) VARIABLES ln(θi ) ln(λin ) ln(λin ) ln(GDP per capita) 0.442** (0.146) ln(GDP) 0.290** (0.104) Duty Share in Revenue 0.071*** (0.016) Agricultural 2.311*** (0.638) ln(GDP per capita) × Agricultural -0.401*** (0.073) Food 2.106*** (0.647) ln(GDP per capita)i × Food -0.377*** (0.074) Importer Fixed Effects No Yes Yes Observations 15 10,981 10,981 Adjusted R2 0.333 0.049 0.047 Note: Robust standard errors in parentheses are clustered by importer in the first column. *, ** and *** indicate that coefficients are significant at 90%, 95% and 99%, respectively. Figure 10: Partial Correlation between θ and GDP per capita of Importing Countries 29 6 Policy Analysis: Recent Trade Spats Involving Chinese BEVs This section aims to analyze the economic rational behind EU’s policy actions regarding imports of Chinese BEVs from a neutral and academic perspective, without delving into the complex reasons behind these actions. We also provide a short discussion to understand the difference in policy rationale and response between the EU, Canada and US on Chinese BEVs import. 6.1 The Sino-EU BEVs Trade Disputes Consider the recent trade tension represented by the EU imposing CVDs on the imports of Chi- nese BEVs, on top of the existing 10% tariff. According to UNCTAD (2015), though ad-valorem and quantitative in nature, CVDs are a type of contingent trade-protective NTMs instead of tariffs, designed and implemented to counteract particular adverse effects of imports in the importing coun- try’s market, contingent upon the fulfillment of specific procedural and substantive requirements.³⁸ Applying CVDs after the fulfillment of procedures requested is WTO-compatible. The extra duties were announced on July 4, 2024, targeting various Chinese vehicle manufacturers that EU claims to received “unfair subsidisation” from the Chinese government, which is purportedly causing a threat of economic injury to the EU BEV producers due to the lower world price. These duties include 17.4% for BYD, 19.9% for Geely, and 37.6% for SAIC.³⁹ The Chinese government and the vehicle producers had publicly denied these accusations.⁴⁰ On August 9, 2024, China lodged a complaint by bringing the case to the WTO’s dispute settlement mechanism over EU’s CVDs on the import of Chinese BEVs, further complicating the trade dispute.⁴¹ On August 20, 2024, the European Commission disclosed the draft decision to impose CVDs on the imports of BEVs from China, with slight adjustments of the previous proposed duties. These duties include 17.0% for BYD, 19.3% for Geely, 36.3% for SAIC, with Tesla as an exception regulated by a low rate at 9%.⁴² On Oct 28th, the EU finalized the decision to impose CVDs on import from Chinese BEVs producers for five years with slightly downward duty rates. Without taking a stance on this issue, this section focuses only on shedding light on understanding the policy actions of the EU in response to the lower world price of the BEVs. The underlying reasons of these trade policy actions are likely far more complex than any economic models can capture. The ³⁸CVDs are also included in border NTMs studied in this paper, as noted by Ederington and Ruta (2016). Compared with traditional tariffs, CVDs can be applied to the individual producers. It is important to emphasize that CVDs are NTMs instead of tariffs, given that several well-known main stream media miscall EU’s policy action as raising tariffs, such as Reuters, BBC, The Wall Street Journal, New York Times and so on. ³⁹Source:https://ec.europa.eu/commission/presscorner/detail/en/ip_24_3630 (Press release, the EU). ⁴⁰Source:http://english.mofcom.gov.cn/article/newsrelease/press/202407/20240703522821.shtml (Press release, the Ministry of Commerce, People’s Republic of China) ⁴¹Source:https://www.bnnbloomberg.ca/business/international/2024/08/09/china-takes-europes-ev-tariffs-to-wto- as-trade-tensions-rise/ (BNN Bloomberg) ⁴²Source: https://ec.europa.eu/commission/presscorner/detail/en/ip_24_4301 (Press release, the EU). 30 reasons may involve balancing competing objectives, such as developing domestic BEV production capacity in the longer run and addressing current climate issues or pollution. Additionally, domestic political factors as well as the international geopolitical factors are inevitably considered. In the context of this paper, we purposely abstract from all these factors and solely focusing on how the EU govern- ment may use NTMs to promote social welfare on a product that may have consumption externality, given that tariffs are already in place. Such a simplification is necessary to distill any policy lessons we can learn from analyzing this ongoing real world dispute. To analyze this trade dispute through the lens of our model, we look at the goods market equi- librium condition, equation (10), which states that the import of Home (EU) equals the net export of Foreign (China) for good X (BEV), which determines the equilibrium world price, as shown in Figure 11. Any factors Z that shift the net export curve to the right, lead to a lower world price and an increase in the imports of BEVs. One such factors could be the state subsidies of the Chinese govern- ment, as argued by the EU. Other possible factors include technological advancement or productivity gains in China, which could also shift the net export curve to the right, resulting in a lower world price. For the purpose of this analysis, the reasons behind the shift of the net export curve are not as important and do not affect our results, thereby we abstract from examining them. In addition to the decrease in the price of the BEVs, the rightward shift in the net export curve also increases the import of BEVs, from M0 to M1 . This may lead to an increase in the overall BEV consumption in the EU, causing negative externalities due to more cars on the road, such as heavy congestion, the deterioration of the public health and the environment.⁴³ Such an externality will decrease the welfare of the EU. To counteract the lower pw , our model shows that the welfare-maximizing response from the EU will be to increase the AV E . To see this, from equation (21), set t = 10%: [ ] [ ( ) ( )] ∗ ∂AV E ( − MX ) ∂pw /∂t θ ∂X ϕ = − w 2 ∗ − w 2 λ ∂AV E + 1 − ∂MX (35) ∂pw (p ) ∂ ( − MX )/∂t (p ) ∂MX ∂AV E ∂AV E ∂X ∂MX ϕ ∂CX ϕ < 0 if + < or < . (36) ∂AV E ∂AV E λ ∂AV E λ The first term of equation (35) captures the effect the world price on the export supply elasticity, w ϵ, which is determined by the terms of trade gains due to the tariff, ∂p ∂t , and the size of Chinese BEV ∗ imports, −MX . It is negative indicating, without considering any externalities, it is always optimal for the EU to raise AVE when facing a lower pw , given fixed tariffs. More Chinese BEV imports will lead to large increases in AVE. The second term of equation (35) captures the effect of consumption externalities. If the condition ⁴³For simplicity, the substitution of gasoline vehicles with BEVs is ignored here. The case with positive externalities will be discussed later. 31 in equation (36) holds, then the item in the square brackets will be positive. This implies that the more the EU government cares about externalities, with θ > 0, the more they will raise the AVE. This occurs when the sum of the positive effect of AV E on the domestic production of BEVs and the negative effect of AV E on imported BEVs, is less than the ratio of the effectiveness of NTMs in reducing the externalities directly (ϕ) to the marginal externality of BEVs (λ). If the effectiveness of NTMs is very high (large ϕ) because of public confidence boost, or if the marginal externality of BEVs is very low (small λ) because BEVs are relatively environmental friendly products compared to the gasoline vehicles, such that ϕ λ is very high, then the condition in equation (36) will likely hold, implying that it is optimal to increase AV E in response to a decrease in pw . This is particularly the case if the AVEs do not cause an overall increase in the EU’s consumption of BEVs, that is ∂AV ∂CX E <0< ϕ λ . Thus, the optimal response of the EU government in the case with externalities will be larger than that without externalities. In the extreme case, if the Chinese export supply of BEVs is completely not responsive to world price changes, such that ϵ is a constant, which is like to be true in the short run, it is still optimal for the EU to raise AVE to reduce imports and thus consumption externalities. Figure 11: Sino-EU Battery Electric Vehicle Disputes Figure 11 nicely illustrates the Sino-EU BEV disputes, assuming that equation (36) holds. Given the initial downward-sloping import demand curve, MX , and the initial upward-sloping export supply ∗ curve, −MX , of BEVs, the initial market equilibrium is at point A, with the world price equals pw 0 and the import of BEVs equals M0 . A change in the supply-side factor, Z , which could be the states subsidies of the Chinese government, or technological advancement and productivity gains in China, ∗ shifts the −MX w curve to the right, leading to a lower world price, p1 , and an increased imports, M1 , at point B. Confronting the lower world price and the larger imports, EU’s existing 10% tariff on imports of BEVs from China shifts the MX curve leftward and reduces the import from M1 to M2 , at point C. On top of this tariff, imposing NTMs further shifts the MX curve leftward, leading to an additional decrease in the import of BEVs, from M2 to M3 (M3 = M0 ), at point D, to fully offset the increase in imports due to the change in factor Z . The use of NTMs in addition to tariffs to further reduce imports nicely demonstrates that tariffs and NTMs are substitutive trade policies. Overall, the 32 mixed use of tariffs and NTMs could capture the terms-of-trade gains and cancel out the impacts on BEV imports due to any supply-side factors such as the Chinese state subsidies or productivity gains. Now consider the case if BEVs generate positive consumption externalities on the social welfare of the EU. Without changing the model and the government objective function (equation 4), this can be done by making the following adjustments: (1) E (·) < 0 is the positive externality function, such that an increase in E is a decrease in positive externality; (2) λ = ∂E /∂CX < 0, i.e. an increase in the consumption of BEVs increases positive externality; (3) ϕ = −∂E /∂AV E < 0, which is that an increase in AVEs reduces positive externality directly, because of losing public confidence that the government is restricting environmental friendly products like BEVs. Equation (35) implies that the sufficient condition for raising AVEs when pw falls is when: ∂AV E ∂X ∂MX ϕ ∂CX ϕ w < 0 if + > > 0 , or > > 0 (∵ ϕ < 0 & λ < 0). (37) ∂p ∂AV E ∂AV E λ ∂AV E λ Thus, if BEVs generate positive externalities, then when facing a lower pw , it is rational for the EU to raise AV E if higher AV E leads to a substantial increase in X (EU’s domestic BEV production), which more than offsets the decrease in MX (the import of BEVs from China), such that CX (the consumption of BEVs) increases, which leads to an overall welfare gains for the EU. This scenario is more likely when restricting imports using AVEs only causes a small loss in public confidence (small ϕ), and the consumption of BEVs generates significant positive externalitites (large λ). In summary, without considering consumption externalities, facing an increase in imports of BEVs from China due to the falling world price, it is optimal for the EU to raise AVEs given fixed tariffs, if the export supply of Chinese BEVs is sensitive to changes in the world price. This could be the case for a long-run equilibrium. However, even in the short-run when the export supply of Chinese BEVs is likely fixed or does not responsive to the changes in the world price, it is still optimal to raise AVEs given fixed tariffs. If the consumption externality is negative due to congestion, raising AVEs not only reduces imports and promotes domestic production of BEVs, but also boosts public confidence that the government is addressing a pressing issue with actions. If the consumption externality is positive due to the substitution with gasoline vehicles, raising AVEs is optimal as well when the AVEs promote domestic production of BEVs so much that it more than offsets the reduction in imports, leading to an increase in the overall consumption of BEVs. This could be because the public will always buy BEVs so that the increase in positive externalities due to more consumption of BEVs in total outweighs the decrease in positive externalities due to the loss in public confidence because of the restrictive AVEs on environmental friendly products.⁴⁴ ⁴⁴Considering the short-run case when the export supply elasticity of Chinese BEVs is fixed and not responsive to world price change, then the first term in equation (35) equals to zero. If we assume away consumption externality (θ = 0), then ∂AV E /∂pw equals to zero. As a result, adding externality in our model is necessary for explaining the policy action of the EU. 33 6.2 Difference in the Policy Responses of the EU, Canada and the US To counter the import competition from low-price Chinese BEVs, the EU, Canada, and the US have implemented distinct combinations of trade policies in 2024. The EU imposed CVDs on Chinese BEV producers on top of existing 10% MFN tariffs, Canada imposing 100% surtax in addition to the existing 6.1% MFN tariffs, and the US raising on the existing tariffs on EVs from 25% to 100%, together with critical component requirements and increasing tariffs on BEV’s component.⁴⁵ The common features of these policy reactions are the objective to develop a domestic BEV industry which is under stress of low-cost Chinese imports, making these trade policies industrial policies. The differences lie in that the EU and Canada combine discriminatory NTMs with the existing MFN tariffs. In contrast, the US invoke the domestic section 301 clause to implement discriminatory tariff unilaterally. Moreover, while CVDs of the EU are firm-specific, surtax and section 301 tariff are not. The analysis of this paper could shed light on the different policy responses of these trade spats. First, our theoretical model shows that when the existing tariff is bound by trade agreement below the optimal level, governments will resort to imposing NTMs. The existing EV tariffs of the EU and Canada are MFN tariffs, which are regulated by the WTO agreement and could not be raised over the bound tariffs, while the US tariff on Chinese BEVs is section 301 discriminatory tariff, can be raised unilaterally and arbitrarily. Thus the differences in flexibility of changing existing tariffs results in their different policy responses. Second, our theoretical model also shows that given the existing tariffs, the optimal AVEs increases with how much the government care about reducing externalities (θ), and decreases with the foreign country’s export supply elasticity (ϵ), which is negatively correlated with the market size (See in equation 17). Given that the EU and Canada are both signatories of the Paris Climate Agreement and care more about externality while the US is less committed, our model would suggest that the optimal AVEs for the EU and Canada could be higher. In addition, the BEV market for the US is far smaller relative to the EU. The differences in BEV market size also imply that it is optimal for the EU to impose restrictive AVEs. Third, our empirical result shows that products that are part of national supply chains have more complementary trade policies. The US considers China’s BEVs as threats to the US technology supply chains and national security, providing the rationale to impose complementary high tariffs on EVs and their components, and critical component restrictions. Finally, the EU, as a rule-based customs union committed to promoting open, fair, and rule-based trade, would generally adhere to WTO-compatible measures to boost public confidence and trust. Violating WTO rules would undermine the EU’s own foundational principles and raison d’être. On the contrary, the US might switch to a more radical, unilateral, and protective overall trade policy stance against China, due to precautionary motives and racing for leadership or dominance in critical industries for national security. More detailed analysis awaits future research. ⁴⁵Source: White House https://www.whitehouse.gov/briefing-room/statements-releases/2024/05/14/fact-sheet- president-biden-takes-action-to-protect-american-workers-and-businesses-from-chinas-unfair-trade-practices/ 34 7 Conclusions This paper studies the relationship between tariffs and non-tariff to answer these questions. “Do countries mixing tariffs with NTMs indicate that they are policy substitutes?” Or “Facing historically low tariffs, do governments resort to imposing discriminatory and restrictive NTMs that are harder to identify, as alternatives to achieve policy agenda without violating the WTO rules?” The answer is yes. Based on latest and detailed product-level tariff and AVE data with bilateral variations, this paper shows that more restrictive NTMs coexist with lower tariffs, making them policy substitutes. “What factors affect governments’ mixed use of tariffs and NTMs as trade and industrial policies?” The answer is that, it depends on the characteristics of importing countries, exporting countries, and prod- ucts. The degree of substitution between tariffs and NTMs increases with the importing countries’ GDP per capita, capital-labor ratio, skilled-labor ratio. The reverse is true facing the exporting coun- tries, except that skilled-labor ratio has no significant impacts. Better institution and governance, and engagement in a deep trade agreement for both importing and exporting countries make NTMs good substitutes for tariffs, while GVC participations turn the relationship between tariffs and NTMs more complementary. Tariffs and NTMs are also more substitutive for consumption, agricultural, and food products, while the opposite holds for intermediate products and capital goods. Some of the empirical findings of this paper can be further explain in a general equilibrium terms- of-trade model whereby welfare-maximizing governments choose tariffs and NTMs to capture terms- of-trade gains and reduce negative consumption externality. In addition, restrictive NTMs also de- crease the negative consumption externalities directly by boosting public confidence. In equilibrium, tariffs and NTMs are policy substitutes, but the degree of substitution depends on the weight of neg- ative consumption externality in social welfare, the effectiveness in enforcing NTMs and the nature of consumption externality. Characteristics of the importing countries, exporting countries and prod- ucts may affect the weight, effectiveness and the nature of externality, thereby providing economic explanations for the empirical findings. Structural estimation of the model collaborate the empir- ical findings that high-income countries use tariffs and NTMs to reduce consumption externalities due to imports, while agricultural and food products, including forestry products, tend to have larger marginal externalities in consumption, which leads to higher substitution between tariffs and NTMs. “Are policies claimed to to safeguard public health or the environment actually protectionism in disguise, intended instead to safeguard the domestic economy?” The answer is they are increasingly so. The recently announced definitive CVDs, a type of border NTMs, imposed by the EU on Chinese BEV exporters, with the stated goal of protecting EU’s BEV producers from unfair threats, on top of the existing 10% tariff, nicely encapsulate how governments mix trade policies as industrial policies to achieve domestic agenda, without violating existing international agreements, ushered a new era of policy mixing to promote long-term industrial and climate goals, further highlight the close link between trade and industrial policies. 35 References Atkin, D., Blaum, J., Fajgelbaum, P., and Ospital, A. (2024). Trade Barriers and Market Power: Evidence from Argentina’s Discretionary Import Restrictions. NBER Working Paper, No. 32037. Beshkar, M., Bond, E. W., and Rho, Y. (2015). Tariff Binding and Overhang: Theory and Evidence. Journal of International Economics, 97(1):1–13. Beverelli, C., Boffa, M., and Keck, A. (2019). Trade Policy Substitution: Theory and Evidence. Review of World Economics, 155:755–783. Bown, C. P. (2014). Trade Policy Instruments Over Time. The Political Economy of International Trade, page 57. Bown, C. P. and Tovar, P. (2011). Trade Liberalization, Antidumping, and Safeguards: Evidence from India’s Tariff Reform. Journal of Development Economics, 96(1):115–125. Chen, J. and Roth, J. (2024). Logs with Zeros? Some Problems and Solutions. The Quarterly Journal of Economics, 139(2):891–936. Chen, T., Hsieh, C.-T., and Song, Z. M. (2022). Non-Tariff Barriers in the US-China Trade War. NBER Working Paper, No. 30318. Conconi, P., García-Santana, M., Puccio, L., and Venturini, R. (2018). From Final Goods to Inputs: The Protectionist Effect of Rules of Origin. American Economic Review, 108(8):2335–65. Conte, M., Cotterlaz, P., Mayer, T., et al. (2022). The CEPII Gravity Database. CEPII. Copeland, B. R. (1994). International Trade and the Environment: Policy Reform in a Polluted Small Open Economy. Journal of Environmental Economics and Management, 26(1):44–65. Costinot, A. (2008). A Comparative Institutional Analysis of Agreements on Product Standards. Journal of International Economics, 75(1):197–213. Cruz, J.-L. and Rossi-Hansberg, E. (2024). The Economic Geography of Global Warming. The Review of Economic Studies, 91(2):899–939. Disdier, A.-C., Fontagné, L., and Mimouni, M. (2015). Trade Restrictiveness Indices in the Presence of Externalities: An Application to Non-Tariff Measures. The Canadian Journal of Economics, 48(4):1513–1536. Ederington, J. (2001). International Coordination of Trade and Domestic Policies. American Economic Review, 91(5):1580–1593. Ederington, J. and Ruta, M. (2016). Nontariff Measures and the World Trading System. Handbook of Commercial Policy, 1:211–277. Essaji, A. (2010). Trade Liberalization, Standards and Protection. The BE Journal of Economic Analysis & Policy, 10(1). 36 Fajgelbaum, P., Goldberg, P., Kennedy, P., Khandelwal, A., and Taglioni, D. (2024). The US-China Trade War and Global Reallocations. American Economic Review: Insights, 6(2):295–312. Fajgelbaum, P. D., Goldberg, P. K., Kennedy, P. J., and Khandelwal, A. K. (2019). The Return to Protectionism. The Quarterly Journal of Economics, 135(1):1–55. Feinberg, R. M. and Reynolds, K. M. (2007). Tariff Liberalisation and Increased Administrative Protection: Is There A Quid Pro Quo? World Economy, 30(6):948–961. Fernandes, A. M., Kee, H. L., and Winkler, D. (2022). Determinants of Global Value Chain Partici- pation: Cross-Country Evidence. The World Bank Economic Review, 36(2):329–360. Fischer, R. and Serra, P. (2000). Standards and Protection. Journal of International Economics, 52(2):377–400. Grossman, G. M. and Helpman, E. (1994). Protection for Sale. The American Economic Review, 84(4):833. Harstad, B. (2024a). Trade and Trees. American Economic Review: Insights, 6(2):155–75. Harstad, B. (2024b). Contingent Trade Agreements. Manuscript. Heo, I. and Choi, B.-Y. (2023). The Global Financial Crisis and Protectionism: Substitution from Tariffs to Non-Tariff Measures. Bulletin of Economic Research, 75(4):880–894. Herghelegiu, C. (2018). The Political Economy of Non-Tariff Measures. The World Economy, 41(1):262–286. Hofmann, C., Osnago, A., and Ruta, M. (2017). Horizontal Depth: A New Database on the Content of Preferential Trade Agreements. World Bank Policy Research Working Paper, (7981). Horn, H., Mavroidis, P. C., and Sapir, A. (2010). Beyond the WTO? An Anatomy of EU and US Preferential Trade Agreements. The World Economy, 33(11):1565–1588. Juhász, R., Lane, N., Oehlsen, E., and Pérez, V. C. (2022). The Who, What, When, and How of Industrial Policy: A Text-Based Approach. SocArXiv Papers. Juhász, R., Lane, N., and Rodrik, D. (2024). The New Economics of Industrial Policy. Annual Review of Economic, 16:213–42. Kaufmann, D., Kraay, A., and Mastruzzi, M. (2011). The Worldwide Governance Indicators: Method- ology and Analytical Issues. Hague Journal on the Rule of Law, 3(2):220–246. Kee, H. L. and Nicita, A. (2022). Trade Fraud and Non-Tariff Measures. Journal of International Economics, 139:103682. Kee, H. L., Nicita, A., and Olarreaga, M. (2009). Estimating Trade Restrictiveness Indices. The Economic Journal, 119(534):172–199. Ketterer, T. D. (2016). EU Anti-Dumping and Tariff Cuts: Trade Policy Substitution? The World Economy, 39(5):576–596. 37 Kuenzel, D. J. (2020). WTO Tariff Commitments and Temporary Protection: Complements or Sub- stitutes? European Economic Review, 121:103344. Lee, J.-W. and Swagel, P. (1997). Trade Barriers and Trade Flows across Countries and Industries. Review of Economics and Statistics, 79(3):372–382. Limão, N. and Tovar, P. (2011). Policy Choice: Theory and Evidence from Commitment via Inter- national Trade Agreements. Journal of International Economics, 85(2):186–205. Lorz, O. and Thede, S. (2024). Tariff Overhang and Aid: Theory and Empirics. Journal of Development Economics, 166:103209. Macedoni, L. and Weinberger, A. (2024). International Spillovers of Quality Regulations. SSRN Working Paper No. 4565981. Moore, M. O. and Zanardi, M. (2011). Trade Liberalization and Antidumping: Is There a Substitution Effect? Review of Development Economics, 15(4):601–619. Niu, Z., Liu, C., Gunessee, S., and Milner, C. (2018). Non-Tariff and Overall Protection: Evidence across Countries and over Time. Review of World Economics, 154:675–703. Niu, Z., Milner, C., Gunessee, S., and Liu, C. (2020). Are Nontariff Measures and Tariffs Substitutes? Some Panel Data Evidence. Review of International Economics, 28(2):408–428. Orefice, G. (2017). Non-Tariff Measures, Specific Trade Concerns and Tariff Reduction. The World Economy, 40(9):1807–1835. Ruckteschler, C., Malik, A., and Eibl, F. (2022). Politics of Trade Protection in an Autocracy: Evidence from an EU Tariff Liberalization in Morocco. European Journal of Political Economy, 71:102063. Silva, J. S. and Tenreyro, S. (2006). The Log of Gravity. The Review of Economics and Statistics, pages 641–658. UNCTAD (2015). International Classification of Non-Tariff Measures. United Nations Publication. UNCTAD, G. (2013). Non-tariff Measures to Trade: Economic and Policy Issues for Developing Countries. UNCTAD. WTO (2012). World Trade Report 2012: Trade and Public Policies: A Closer Look at Non-Tariff Measures in the 21st Century. WTO (2023). RTA Provisions Glossary. World Trade Organization. Yu, Z. (2000). A Model of Substitution of Non-Tariff Barriers for Tariffs. Canadian Journal of Economics, 33(4):1069–1090. Yue, K. (2022). Non-Tariff Measures, Product Quality and Import Demand. Economic Inquiry, (2):870–900. 38 Appendix A Variables Definitions and Data Sources Table A.1: Variables Definitions and Data Sources Variable Name Definitions Source Tariff Effectively applied tariff rate at importer-exporter-HS 6 digit product level UNCTAD TRAINS AVE Ad valorem equivalent (tariff ) of the border NTM Kee and Nicita (2022) GDP per capita GDP per capita (current US$) World Bank WDI Capital / Labor (log) Capital stock (at constant 2017 national prices in mil.2017 US $) devided by total employment Peen World Tables 10.01 High Skilled Labor Share Labor force with advanced education (% of total working-age population with advanced education) World Bank WDI Duty Share in Revenue Customs and other import duties (% of tax revenue) World Bank WDI Control of corruption the extent to which public power is exercised for private gain, including both petty and grand forms of corruption, as well as World Governance Indicators ”capture” of the state by elites and private interests. Government Effectiveness the quality of public services, the quality of the civil service and the degree of its independence from political pressures, the quality World Governance Indicators of policy formulation and implementation, and the credibility of the government’s commitment to such policies 39 Rule of Law the extent to which agents have confidence in and abide by the rules of society, and in particular the quality of contract enforcement, World Governance Indicators property rights, the police, and the courts, as well as the likelihood of crime and violence Regulatory Quality the ability of the government to formulate and implement sound policies and regulations that permit and promote private sector World Governance Indicators development. Intermediate Products HS 6-digits Products falling into the group of intermediate products Broad Economic Classifications Consumption Products HS 6-digits Products falling into the group of consumption products Broad Economic Classifications Capital Products HS 6-digits Products falling into the group of capital products Broad Economic Classifications Agricultural Products HS 6-digits products falling into the chapters 1-24 of the Harmonized System WTO Agreement on Agriculture Food and Beverages HS 6-digits products, edible products (food and beverages) USDA Economic Research Service WTO members The indicator that taking value of one if both importer and exporters are WTO members and zero otherwise CEPII Gravity Database Deep trade agreement The indicator that taking value of one if importer and exporter are engaged in a trade agreement and zero otherwise Hofmann et al. (2017) DTA depth The horizontal depth of deep trade agreement, measured by the number of provisions covered in the DTA Hofmann et al. (2017) DTA depth LE The horizontal depth of deep trade agreement, measured by the number of legally enforceable provisions covered in the DTA Hofmann et al. (2017) Backward GVC participation (log) The import content of country’s exports, the intensity of GVC participation Fernandes et al. (2022) Forward GVC participation (log) The domestic value-added in exports that is used by the country’s bilateral partner countries for export production Fernandes et al. (2022) RTA number The number of Regional trade agreements currently in force WTO RTA database NTM notifications The number of NTM notifications from the WTO member countries to the WTO WTO NTM database Distance to GVC hubs (log) Logarithm of sum of distance to China, Germany, and the United States (capital city to capital city) CEPII Gravity Database Appendix B Additional Tables and Figures Table B.1: Conversion Table Between Theory-Based and Data Collecting-Based NTMs Classification Theory-Based Data Collecting-Based Chapter A-Sanitary and Phytosanitary (SPS) Chapter B-Technical Barriers to Trade (TBT) Chapter C-Pre-shipment Inspections and Other Formalities Border NTM Chapter D-Contingent Trade-Protective Measures (safeguard provisions, escape clauses, anti-dumping and countervailing duties) Chapter E-Non-automatic licensing, quotas, prohibitions and quantity-control measures Chapter F-Price-control measures, including additional taxes and charges Chapter P-Export Subsidies Product NTM Chapter A, Chapter B, Chapter E Process NTM Chapter A, Chapter B Customer NTM Chapter E, Chapter F The Horizontal Depth of PTA and Deviation from WTO The recent trade agreements cover more policy areas beyond the WTO mandate. Horn et al. (2010) classify the provisions covered by the PTAs into two groups: ‘WTO+ (WTO Plus)’ and ‘WTO-X (WTO Extra)’, which contains 14 and 38 provisions, respectively. The former refers to provisions that are also governed by the current mandate of the WTO, but the PTAs contain the same or more stringent commitments. The latter refers to provisions of the PTAs that go beyond the regulation of the WTO. Using the updated data on the content of the PTAs provided by Hofmann et al. (2017), Figure B1 shows the number of provisions falling into the two categories for each PTA. First, after the end of the Uruguay Round negotiation in 1994, there is an increasing number of PTAs. Second, the number of provisions falling in the two categories both increase, which reflects an increase in the horizontal depth of the PTAs. Third, more provisions go beyond the mandate of the WTO agreement, which suggests the incremental distinction in content between global trade negotiations and multilateral or bilateral trade negotiations. Figure B1: The Horizontal Depth of Preferential Trade Agreements 40