How Do Countries Respond to Trade Disputes? Evidence from. Chinese Exporters

Transcription

1 How Do Countries Respond to Trade Disputes? Evidence from Chinese Exporters Tan Li and Ying Xue June 2017 Abstract This study examines China s trade response to the U.S.-China trade disputes from 2000 to 2006, using monthly product-level trade transaction data covering all Chinese exporters. It provides the first empirical evidence of the trade effects of all of the trade disputes occurring during the sample period, complementing the literature that is solely based on WTO disputes. We find a significant trade promoting effect on the export volumes of Chinese disputed products to the U.S. after China s initiation of trade disputes against the U.S. s WTO-inconsistent import restriction measures. We find that the export volumes to average non- U.S. markets significantly decrease, indicating a trade deflection effect. I also find that the export prices to the U.S. fall, whereas export prices to non-u.s. markets remain unchanged. Importantly, these trade effects result from disputes against the U.S. s safeguard measures. We find that the trade promotion effect is attributable to both an extensive margin (i.e., more exporters) and an intensive margin (i.e., higher export volumes of surviving exporters), whereas the trade deflection effect is attributable to an extensive margin (i.e., less exporters to non-u.s. markets). The main results are robust based on a series of checks using alternative specifications and various datasets. Keywords: Trade disputes; JEL Classifications: # Correspondence: Li: School of International Business, Southwestern University of Finance and Economics, China, litan@swufe.edu.cn. Xue, CCB International Securitires Ltd., xueying@ccbitl.com. Acknowledgement: i

2 1. Introduction Trade disputes arise when one country complains that another country is violating its commitment to bilateral (e.g. an FTA) or multilateral trade agreements (e.g. the GATT/WTO agreement). Trading partners frequently get involved in trade disputes. According to the statistics of the Dispute Settlement Body (DSB) of the WTO 1, member countries have reported more than 500 trade disputes to the DSB since its establishment in 1995 up until the end of Trade disputes in which the initiating governments choose to bring the cases to the WTO for consultation are generally known as WTO trade disputes. However, recent studies show that WTO trade disputes only constitute a small fraction of all of the trade disputes that have occurred between trading partners, which are known as primary trade disputes. As Horn and Mavroidis (2006) point out in their survey, WTO disputes are not just the tip, but the tip of the tip of the iceberg among all primary trade disputes. Using global data, Li and Qiu (2015) quantitatively show that WTO trade disputes only account for 30% of all of the primary trade disputes occurring between 1995 and We find a similar pattern for trade disputes between the U.S. and China, the top two trading nations in the world. Only 7 of the 46 primary U.S.-China trade disputes occurring between 2002 and are WTO disputes. Non-WTO trade disputes arise more frequently between trading partners than WTO disputes and differ from WTO disputes in their settlement approaches. For WTO trade disputes, the DSB of the WTO normally forms a consultation panel to review the case and then makes a final ruling 3 on whether the defendant have violated the WTO agreements cited by the complainants according to the evidence 1 WTO trade disputes database ( is the first year the WTO received U.S.-China trade disputes; 2007 is the latest year for which primary trade dispute data are available. 3 It is possible that some WTO disputes do not have final rulings, as these cases are dropped during consultation. Thus, formal settlement by the WTO is not required any more (see Chaudoin, Kucik, and Pelc, 2016). In addition, the review period before a final decision is made by the WTO may range from months to years depending on the complexity of the cases. 1

3 provided by both parties. However, non-wto trade disputes are not brought to the WTO for consultation, thus no WTO rulings are available. Countries involved in non-wto disputes may choose to resolve the dispute using alternative approaches. For example, if they belong to the same regional trade agreement (RTA), they can choose to bring the case to the Disputes Settlement Mechanism of the RTA. If they have no RTA partnership, they can also rely on bilateral negotiations to resolve disputes 4. Despite the rising recognition of the importance of non-wto disputes, almost all of the studies on various aspects of trade disputes are based on WTO disputes 5. Moreover, although the economic and political studies on the determinants of trade dispute initiations have proliferated dramatically 6, much less is known about the potential trade effects of trade disputes on complainant s export of products involved in dispute cases. Understanding whether and how exporters respond to trade disputes on earth may generate significant implications for their domestic industrial dynamics and economic growth. More importantly, as non-wto disputes account for the majority of primary trade disputes, and they differ from WTO cases in settlement procedures, it is especially worthwhile to study the trade responses to primary trade disputes. An empirical investigation of this issue would complement the literature, which primarily focuses on WTO trade disputes. To fill this void, we provide the first empirical investigation of the trade responses to primary trade disputes using monthly product-level (i.e., HS-4 digit) trade data. The first database of global primary trade disputes from 1995 to 2007 was constructed by Li and Qiu (2015) through keyword searches of trade dispute news reports in Factiva, one of the world s largest digital business archives. We use the primary trade disputes between the U.S. and China from 2000 to 2006 from the above database and Chinese 4 For non-wto cases, it s common for countries to fail to achieve any successful resolutions or to leave the case unresolved for years. 5 One exception is Li and Qiu (2015). The authors constructed the first primary trade disputes database for global economies from 1995 to 2007 and they explored the determinants of primary trade disputes. 6 Horn and Mavroidis (2006) provide a survey of the literature on the determinants of WTO dispute initiations before

4 exporters monthly product-level trade transactions over the same period from China Customs for empirical analysis. The analysis of China s trade responses to bilateral U.S.-China trade disputes is largely motivated by the fact that the Sino-U.S. trade relation 7 is one of the most prominent trading partnerships in the world, let alone the economic size of both countries. Furthermore, the U.S. has been the world s largest initiator and defendant of trade disputes, whereas China is a frequent target of trade disputes initiated by its trading partners. Sattler and Bernauer (2011) show that more economically powerful countries tend to initiate disproportionately more trade disputes against their trading partners. Moreover, countries that have higher trade volumes tend to get involved in more trade disputes (Horn et al., 1999; Bown, 2005; Li and Qiu, 2015). Hence, we observe that the U.S. has always been the largest initiator and defendant of primary trade disputes with China among all of its trading partners. Therefore, it is worthwhile to study the trade responses to Sino-U.S. trade disputes, especially from the perspective of how developing countries (e.g., China) cope with trade disputes with developed countries (e.g., the U.S.). Our choice of time period (i.e., ) is mainly due to the availability of Chinese Customs data. The difference-in-differences (DID) estimation strategy has been widely used in empirical trade studies to estimate the effects of various trade policies at either the country level (e.g., Egger et al., 2008; Foster et al., 2011), product level (e.g., Pierce and Schott, 2015), or firm level (e.g., Lu, Tao, and Zhang, 2013). Using the DID method, we compare a set of trade outcome variables (such as export volume, export price, number of exporters, and export volume of surviving exporters) for products at the four-digit Harmonized System (HS-4 digit level) involved in trade disputes with the same outcome variables for products in a control group (the first difference) before and after the initiation of trade disputes (the second difference) 8. Our control group consists of all HS-4 digit products not involved in any trade disputes but 7 The U.S. has already granted the Normal Trade Relation (a U.S. term for most-favored-nation ) status to China since 1980.The U.S. s Permanent Normal Trade Relations (PNTR) with China was also passed in Due to low NTR tariffs, the U.S. has become a major exporting destination for China since China opened up in Most of our primary trade disputes arise after the implementation of the temporary trade protection measures. While in this study, we mainly compare the export outcomes after the initiation of disputes to those before disputes initiation, 3

5 within the same HS-2 digit category as the products in the treatment group. To address unobservable product variations, we control for product fixed effects in the DID estimation specification. One possible source of variation is the level of domestic protection. Agricultural products normally enjoy a higher level of domestic protection, especially in developing countries, and thus are more frequently involved in trade disputes regarding high entry barriers. Another concern is that some products are more politically contentious. For example, steel products are found to be more vulnerable to tit-for-tat trade retaliations by the U.S. and Canada (Blonigen and Bown (2003)). Controlling for product fixed effects helps account for all possible product variations. We also control for the most-favored-nation (MFN) tariff rates of the U.S. and time fixed effects in our specification. Our empirical analysis yields a number of important findings. First, we find a significant trade promotion effect of primary trade disputes initiated by China against the U.S. The magnitude of the trade gains is found to be much larger than that found in previous studies on WTO disputes. China s trade disputes against the U.S. s safeguard measures play a dominant role in generating this effect. The the export quantity increase is a result of both an extensive margin (an 82.6% increase in the number of exporters) and an intensive margin (a 215% increase in the export volume of surviving exporters who export the disputed products to the U.S. throughout the entire period). Second, we find a substantial trade deflection on third countries. The export volumes of the same disputed products to non-u.s. markets decrease significantly. This trade deflection effect is attributable to the extensive margin rather than the intensive margin. The number of Chinese firms exporting disputed products to non-u.s. markets decreases by 12.5% after trade disputes, whereas the export volumes of surviving exporters remain unchanged. Third, the export prices of disputed products to the U.S. falls substantially in response to a more competitive U.S. export market. Export prices to non-u.s. markets demonstrate little change. but not to those before the imposition of temporary trade barriers. 4

6 Our main results remain robust after a series of checks using alternative empirical specifications and various data samples. First, we test the validity of our DID estimates by including product-specific time trends into the baseline specification to allow for different time trends across the treatment and control groups. Second, we use quarterly rather than monthly data to address the concern that monthly data is much more volatile, as firms may not export every month. Third, we exclude dispute cases occurring in the first and last years of our sample to have long enough pre- and post-dispute periods for the DID estimation. Last, we exclude all overlapping trade disputes that involve third countries other than the U.S. and China from our sample. Our main results remain robust, confirming the validity of our DID estimates (see Section 1.5 for more details). The present study makes several important contributions to the burgeoning literature on the trade effects of trade disputes. First, it provides the first empirical analysis of exporters trade responses to primary trade disputes. In the literature, the few empirical studies on the trade outcomes of trade disputes are all based on WTO trade disputes. Bown (2004) provides the first study of the economic outcomes of WTO trade disputes. He empirically examines the possible economic and political factors that may determine the level of trade liberalization achieved after the initiation of trade disputes. He finds that plaintiffs with greater trade retaliation power may more effectively make defendants remove illegal trade barriers after trade disputes. According to his calculations based on all formal GATT/WTO disputes from 1973 to 1998, the average trade effects of WTO disputes is 0.78%. That is, the export of disputed products (at the HS-6 digit level) from complainant to defendant only increase by 0.78% from 1 year before the trade disputes to 3 years after the end of the trade disputes, which is quite minimal. Although based on a new dataset of WTO trade disputes initiated from 1995 to 2011, Bown and Reynolds (2015) find that complainant countries increase their exports of disputed products to defendants by approximately 12% after the conclusion of WTO disputes. Bown (2004) finds a much larger effect using earlier WTO disputes. Moreover, Bechtel and Sattler (2015) estimate the trade effects of WTO trade disputes from 1995 to

7 using annual trade data at the one-digit SITC product level. By comparing export flows between countries involved in trade disputes with those between similar country pairs (measured by GDP and pre-dispute trade flows) not involved in any trade disputes, they find that complainants exports of disputed products to defendants increase by 23% in the 3 years after the final ruling on WTO disputes. Chaudoin, Kucik, and Pelc (2016) further empirically test whether all WTO members increase their exports of disputed products to defendant countries after the initiation of WTO disputes, given the multilateral nature of WTO agreements. Based on a sample of 293 WTO disputes arising from 1995 to 2010, the authors find no significant increase in the exports of all member countries to the defendants. Our approach differs from these studies in that we use all trade disputes (i.e., both WTO and non- WTO disputes). To further explore exporters responses to trade disputes we rely on the monthly productlevel export transaction data of Chinese firms, rather than the aggregated product-level data used in previous studies. As China has the largest number of trade disputes with the U.S. among all of its trading partners, we focus on trade disputes occurring between the U.S. and China. Moreover, relative to the estimation of trade effects using dispute dummy variables in previous works, our DID approach exploits both the cross-sectional variation (disputed products and control products) and time variation (before and after the initiation of trade disputes). Our DID estimates remain consistent when there are potential unobservable differences across the treated and control products. We separately test the quantity (i.e., export volume) and price changes after the imposition of trade disputes instead of simply examining the export value or volume changes as in previous studies. We also explore whether these trade responses emerging from the extensive margin (i.e., the number of exporters), the intensive margin (i.e., the export volumes of surviving exporters), or both. Thus, we complement the literature, which primarily address WTO disputes, by providing new evidence regarding the trade effects of all trade disputes (the majority of which are non-wto disputes) and their underlying mechanisms. Overall, we find a significant increase in the Chinese export of disputed 6

8 products to the U.S. after primary trade dispute initiation, which is generally consistent with the few studies on WTO disputes. However, the trade promotion effect is much more substantial for primary trade disputes. That is, trade disputes that are not brought to the WTO affect trade flows more. Second, we find positive trade gains for complainants that initiate trade disputes, a finding that also contributes to the political-economic literature of disputes settlement. Despite the incomplete contract nature of WTO agreements 9, consensus has been reached in the literature that trade disputes can significantly help promote international trade. Domestic firms actively pressure their governments to file trade disputes to improve their market access. Governments are also motivated to achieve more liberalized trading system and increased trade flow by lifting WTO-inconsistent trade restrictions from their trading partners through trade disputes litigation (Davis, 2011; Chaudoin et al., 2016). Complainant and defendant governments may rely on WTO panel consultation or bilateral negotiations outside of the WTO to reach settlements. If the final rulings are against the defendants, the complaints should achieve positive trade outcomes after the removal of illegal trade protection policies. However, it is possible that defendants may not comply with their obligations, as the WTO has no way to sanction them and their compliance with final rulings is largely self-enforced 10. Nevertheless, trade disputes may still facilitate the removal of WTO-inconsistent trade barriers through other channels. Political pressures imposed by disputes settlement may encourage the removal of trade barriers. Failure to fulfill this obligation would severely harm defendants international reputations and thus weaken their governments negotiation power in later trade disputes (Sattler, Spilker and Bernauer, 2014). Furthermore, the loss of domestic reputation is very harmful (Mansfield, Milner and Rosendorf, 2002; Rosendorff, 2005; Tomz, 2007; Fang, 2008). Trade disputes also incur various costs for defendants if they leave trade barriers in place (Maggi, 1999; Busch and Reinhardt, 2000; Bagwell and Staiger, 2005; 9 Recent theoretical studies include Maggi and Staiger (2011, 2013, 2015) and Staiger and Sykes (2013). 10 See, for example, Kono (2006) and Bown and Crowley (2013). 7

9 Bown, 2009; Maggi and Staiger, 2011). However, potential retaliatory measures 11 by complainants can help facilitate compliance. Bown (2004) finds that complainants with greater retaliation power enjoy higher levels of trade liberalization after trade disputes. Overall, if trade disputes effectively compel the removal of harmful trade restrictions, the complainants export of disputed products to defendants should be expected to increase after disputes. We therefore contribute to the political economic literature by presenting new evidence of complainants positive trade gains from trade disputes. Third, this is one of the few studies that explore the potential trade effect heterogeneity of trade disputes. Previous studies have examined several sources of trade dispute variation that may lead to heterogeneous trade responses. For example, Bown (2004) distinguishes trade disputes regarding tariff measures from those regarding escape clause measures (such as antidumping, safeguards, and countervailing measures). The author finds that trade disputes regarding tariff measures imply higher levels of trade promotion than those regarding escape clause measures (Bown, 2004). Moreover, Bown and Reynolds (2015) find that WTO trade disputes have different trade effects on complainants depending on whether the trade policies being challenged under the disputes are global policies (applicable to all member countries) or partial policies (imposed on specific countries). We contribute to this line of research by exploring another source of variation. That is, whether the alleged trade protection policies under disputes are safeguard or antidumping measures. The literature has compared the trade effects of these two types of temporary trade restriction measures. For example, Bown (2013) uses the U.S. productlevel steel import data to explore the trade effects of the U.S. safeguard measures applied in 2002 and finds that they are comparable to those of antidumping and other temporary trade restriction measures imposed from 1989 to However, the possible heterogeneous trade effects of trade disputes on these two trade barriers have rarely been examined. We find that trade disputes initiated by China against the U.S. s illegal 11 Retaliations have legal foundations under the WTO framework. These measures can be applied to defendants if they fail to comply with rulings. 8

10 safeguard measures play a dominant role in facilitating exports, whereas disputes against the U.S. s antidumping measures have no significant effects. Thus, our finding of the different trade effects of these two types of disputes on the exports from China (i.e., the complainant) to the U.S. (i.e., the defendant) verifies the existence of heterogeneous responses. Our last contribution to the literature lies in the empirical identification of possible changes in firms exporting strategies (e.g., export destination and export price) in response to trade disputes. Compared to the aggregated annual product-level trade data used in previous studies, our Chinese Customs data allow us to explore firms export-destination choices and pricing behavior. Specifically, beyond the direct effect of trade disputes on the exports of the complainant country (i.e., China) to the defendant country (i.e., the U.S.), we also examine the indirect effect on exports to non-defendant (i.e., non-us) countries, which is barely addressed in the literature. The trade diversion or deflection effects of trade policies have been extensively studied. Prusa (1997) documents the trade diversion of US antidumping cases initiated between 1980 and 1988 and finds that antidumping largely divert U.S. imports from subject countries to non-subject countries. Prusa (2001) and Carter and Gunning-Trant (2010) obtain robust trade diversion results by estimating a dynamic model of panel data of U.S. antidumping measures. Subsequent studies also find trade diversion effects of antidumping cases for the E.U. (Lasagni, 2000; Konings, Vandenbussche, and Springael, 2001) and China (Park, 2009). Different from these trade diversion studies for the initiators of antidumping actions, Bown and Crowley (2007) further examine the trade deflection effects on the named countries. They theoretically prove that import-restricting policies, such as antidumping and safeguard measures deflect foreign countries exports to third countries. Their theoretical analysis is fully supported by the empirical evidence of Japan s export deflection from the U.S. to third countries when subject to the U.S. s antidumping measures. The present study is similar to that by Bown and Crowley (2007) in that it focuses on the trade deflection of the target countries rather than the initiating countries of these import restriction policies. 9

11 However, we fill the void in the literature by investigating the trade deflection effect of trade disputes against temporary import barriers, especially safeguard measures. In addition, we find changes in the pricing strategies of Chinese exporters in response to trade disputes, which have important implications for market competition and terms-of-trade studies 12. The rest of the paper is organized as follows. Section 1.2 describes the data. Section 1.3 introduces the empirical strategy. Section 1.4 presents the main empirical findings. Section 1.5 discusses some robustness checks. Section 1.6 concludes the paper. 2. Data Our empirical analysis of Chinese exporters response to trade disputes with U.S. requires both the trade transaction data of Chinese exporters and product-level information on the trade dispute cases between the U.S. and China. The sample period of our study is from 2000 to 2006, when the monthly product-level trade transaction data are available 13. Chinese firm-level trade data are from China Customs. This comprehensive data set covers the monthly export and import transactions (at the HS-8 digit level) of all Chinese exporters and importers over the period of 2000 to It provides detailed information for each trade transaction, such as the product information, trade volume, trade value, unit price, export destination or importing countries, the mode of trade (e.g. ordinary trade or processing trade), name and identity of Chinese exporters or importers, and form of the exporting or importing companies (e.g. stated-owned or privately owned). As we focus on China s export responses to trade disputes, we collect all of the trade transactions of Chinese exporters. We use two sets of trade disputes data. The first set is WTO trade disputes, which arise when one 12 Studies include, for example, Bagwell and Staiger (2002), Broda et al. (2008), Bagwell and Staiger (2011), Bown and Crowley (2013), and Ludema and Mayda (2013). 13 China Customs began to release the monthly product-level trade transactions of Chinese exporters in 2000; the latest year of observations available to the author is

12 member government of the WTO complains that another member government is violating its commitment to WTO agreements and thus brings the case to the WTO s DSB for consultation. Full records of each of the WTO s trade dispute cases can be found in the WTO s trade dispute database 14. In each case, the WTO reports the complainant and respondent member countries, the date the dispute was brought to the WTO, the agreement cited by the complainant requesting WTO consultation, and the summary of the dispute to date (products involved in the dispute are usually discussed here). In total, the WTO reports 25 trade dispute cases between the U.S. and China, from 2002 to However, during our sample period (i.e., ), there are only three trade disputes. Two cases were brought by the U.S. to the WTO in 2004 and 2006 respectively, with the former complaining of China s high value-added tax (VAT) for U.S. products and the latter complaining of China s unfair import policy on U.S. automobile parts. China also brought one case against the U.S. to the WTO in March 2002, complaining about U.S. s WTO-inconsistent safeguard measures toward China s steel products 15. As we want to explore how disputes affect Chinese exports, we exclude the two U.S.-initiated cases in which U.S. exports are the main targets. Thus, only one WTO trade dispute case is left for our study period (i.e., ). The second set of trade dispute data concern primary trade disputes, which are defined as all of the trade dispute cases that have actually occurred between trading partners. Thus, by definition, the WTO trade disputes are a subset of the primary trade disputes. The remaining disputes in the primary dispute dataset are those not brought to the WTO by complainants. Primary trade dispute data are not available from the WTO or any other trade database. Li and Qiu (2015) constructed the first primary trade dispute dataset by searching, collecting and identifying trade dispute cases from first-hand news reports provided by the world s largest digital business archives - Factiva 16. Their primary trade disputes dataset ranges from 14 The WTO Trade Disputes Database ( 15 WTO finally ruled in November 2003 that US violated the GATT agreements regarding the safeguard measures. 16 Factiva is a comprehensive news report database, that covers more than 36,000 news sources in the forms of 11

13 January 1, 1995 to December 31, 2007, a 13-year period starting with the WTO s official establishment (replacing the GATT) and ending before the 2008 global financial crisis that caused dramatic collapse of world trade and might have further affected trade disputes. In total, Li and Qiu (2015) obtained 23,149 news articles relevant to trade disputes from Factiva based on some search criteria 17. After carefully screening each of these articles, they successfully identified 1,130 trade dispute cases 18 that are in accord with the WTO s trade dispute definition. These disputes involve more than 110 countries. Of these primary disputes, 369 cases are WTO trade disputes and the remaining 761 cases are not. They also converted the multiple-country trade disputes into 6,228 bilateral trade disputes at the country-pair level. For each primary trade dispute, they recorded its key information, such as the date (i.e., year and month) the dispute occurred, the complainant and defendant countries involved in the dispute, the issue of the dispute (e.g., dumping, export subsidies, and import quota), the industry or products involved, and the proposed settlement approach (if available). Some may raise the concern that news reports may be biased in terms of country coverage, language and ideology. However, such media biases should not affect our study for three reasons. First, the media may focus more on larger countries than smaller, less-developed ones, implying that there should be more reports for the high-income countries covered by Factiva. As we mainly focus on trade disputes between China and the U.S. the largest and second-largest countries in the world 19 potential bias toward larger countries, if any, should not interfere with our sample or the issues we study. Second, although Factiva s news reports are all in English, many major news agencies in China and other non-english speaking newspapers, magazines, newswires, televisions, audio transcripts, and web and social media in over 200 countries. It contains global, regional and local news reported by numerous domestic and global news agencies, such as Reuters and the Associated Press. Its wide country coverage and rich news sources allow it to track almost all of the trade disputes that have happened in the world. 17 Their search approach is simple and direct. They use the key words trade disputes to search for news reports in Factiva. Other similar key words, such as trade dispute, trade war, and trade conflicts, were also used but none of them returns more outcomes than trade disputes. 18 The same trade dispute cases may arise several times over different years. They are treated as independent cases in each year that they arise. 19 The country size of the U.S. and China is measured using the nominal GDP (PPP based) of

14 countries publish English-version news reports. For example, the English news reports of China s mainstream media, Xinhua News Agency, are fully covered in Factiva. Thus, the potential for language bias in news reports is low. Third, even if media reports are biased in ideology, the countries and products involved in trade dispute are all objective facts that are unlikely to be affected. [Table 1 Inserted Here] Table 1.1 reports the top 20 countries as either complainant or defendant in trade disputes over the whole sample of bilateral primary trade disputes. The U.S. is both the biggest complainant and biggest defendant of trade disputes in the world. U.S.-initiated disputes against other countries account for nearly one fourth of the total disputes. Dispute cases that target the U.S. account for more than 16% of all of the disputes. China initiated 171 trade disputes as a complainant between 1995 and 2007, ranking 18 th among 110 countries. However, China is the third largest country in the world as a defendant of trade disputes, involved in 444 cases. During our sample period (i.e., ), China had 525 trade disputes with 46 countries. Table 1.2 presents the distribution of these 525 bilateral trade disputes by complainant and defendant countries. Of the 46 countries, the U.S. is both the largest complainant, initiating 24 dispute cases against China, and also the largest defendant, receiving 15 complaints from China. To summarize, there are 39 primary trade disputes between the U.S. and China from 2000 to As we focus on the response of Chinese exports to trade disputes with the U.S., we exclude seven cases that target U.S. exports. Table 1.3 displays examples of trade disputes that target the exports of either the U.S. or China. The first example is the U.S. s initiation of a trade dispute against China in 2004, complaining of China s high VAT on semiconductors imported from the U.S. The second case is China s complaint against the U.S. s dumping of PVC. Both of these cases target U.S. s exports, regardless of whether initiated by China or the U.S. Trade disputes that target U.S. exports are likely to affect U.S. rather than Chinese exports, and are therefore excluded from our sample. We exclude 3 ( one from 2002 and two from 2004) of the remaining 32 cases because they focus 13

15 on trade regulations or practices rather than specific merchandise exports. For example, the 2002 case was initiated by China against the U.S. for seeking recognition of its market economy status and implementation of farm agreement. The remaining two cases were initiated by the U.S., complaining that the devaluation of China s currency and its violation of worker rights make its exports more competitive. Thus, we have a total of 29 primary trade disputes between the U.S. and China from 2000 to Twelve of these disputes are China-initiated (i.e., China is the complainant) and 17 cases are U.S.-initiated (i.e., China is the defendant). [Table 2 & 3 Inserted Here] According to the trade policies being challenged in trade disputes, we further classify our 29 primary cases into two types. Type I cases are defined as trade disputes regarding the U.S. s (potential) safeguard measures to temporarily restrict the imports of certain products from China. These actions normally occur in the form of temporary tariff increases or other non-tariff measures, such as import bans and import quotas. Type I trade disputes are generally initiated by China against the U.S., arguing that the U.S. s trade protections are inconsistent with the WTO s pre-requisites for imposing safeguard measures under the GATT/WTO agreements 20. Type II cases generally refer to the primary trade disputes concerning the (potential) anti-dumping measures of the U.S. against certain products from China. This type of dispute may be raised by either the U.S. (i.e., complaining that China is dumping into its market) or by China (i.e., defending itself or protesting against the U.S. s (potential) anti-dumping actions). Of the 12 trade disputes initiated by China, 7 are Type I and 5 are Type II. Of the 17 cases in which China is the defendant, 7 are Type I and 10 are Type II. 20 Uruguay Round of the WTO transformed the GATT Article XIX, which regulates the utilization of safeguards, into the WTO s Agreement on Safeguards. This WTO safeguard agreement introduces some changes, such as the prohibition of gray area measures for import restrictions (e.g. bilateral voluntary export restraints, orderly marketing agreements, and similar measures). Moreover, the WTO safeguard provisions are written into various articles, for example, the Agreement on Textiles and Clothing (Article VI), the Agreement on Agriculture (Article V), and the General Agreement on Trade in Services (Article X). See Bown (2002) for a detailed discussion on the changes in Safeguards Agreements under the WTO. 14

16 Next, we classify the products involved in each of these 29 U.S.-China primary trade disputes at a Chinese HS-4 digit level, which is the most disaggregated level at which almost all products involved in these disputes may be identified. In sum, between 2000 and 2006, we obtain 113 monthly trade dispute observations at an HS-4 product level. To be comparable to the trade dispute data, we also aggregate the monthly export data (e.g., export volume, export price, and number of exporters) of Chinese exporters to the U.S. from the HS-8 digit level to the HS-4 digit level. For Chinese exports to non-u.s. markets, we first generate their total export volume, numbers of exporters, and trade- weighted export prices at the HS-8 level and then use the same method to aggregate these data to the HS-4 level. Our final dataset was constructed by matching the monthly product level US-China trade disputes with Chinese exporters monthly trade transaction data at the HS-4 digit level. Products involved in trade disputes belong to the treatment group, whereas unaffected products under the same HS-2 digit level as the affected products belong to the control group. We also use the U.S. s MFN tariff rates at the HS-4 product level as control variables in our estimation. Feenstra, Romalis and Schott (2002) provide the U.S. s ad valorem equivalent MFN tariff rates at THE HS-8 digit level from 1989 to We aggregate the MFN rates of the U.S. from the HS-8 level to the HS-4 level using the U.S. s import flows from China at the corresponding levels as weights. The U.S. s product-level import data are drawn from the UN Comtrade database. The matched dataset of China s exports to the U.S. contains 40,514 product-month level observations. To examine the possible effects of U.S.-China trade disputes on China s exports to non-u.s. markets, we further match China s monthly export transactions to non-u.s. markets at the HS-4 level and obtain 41,822 observations. For the single WTO trade dispute in our sample, we use the same method as for the primary trade disputes to classify the products involved in the WTO dispute at the HS-4 digit level and then match 21 The tariff rates data of Feenstra et al. (2002) are widely used in empirical trade literature because of their larger coverage of industries. However, their tariff rates are not available after 2001, so we follow Pierce and Schott (2015) to assume them constant in the years afterward. 15

17 these monthly product-level WTO dispute observations to China s HS-4 level export data to U.S. and non- U.S. export data. The matched sample of China s exports to the U.S. contains 3,117 observations; that of exports to non-us markets has 3,647 observations. 3. Empirical Specification To explore China s possible trade responses to U.S.-China trade disputes, we use the difference-indifferences (DID) specification, which examines whether the export performance of products that are involved in trade disputes is significantly different from that of unaffected products after the initiation of trade disputes. The DID approach is usually adopted in studies that attempt to exploit both the crosssectional variations (between the treatment and control groups) and time variations (before and after an event). Another merit of the DID approach is that it can directly eliminate potential time-invariant unobservable differences between products in the treatment and control groups. In our study, the treatment group consists of products (at the HS-4 level) involved in the U.S.-China trade disputes and the control group consists of all unaffected products within the same HS-2 level to which the affected products belong. The products in the control group are considered good counterfactuals of products in the treatment group because they are in the same broad product category and thus have similar applied tariff rates, export levels, and probabilities of involvement in trade disputes. Our sample includes monthly product level observations from 2000 to Thus, the DID estimation specification at product level takes the following form: y it = β 0 + β 1 TREAT i + β 2 POST it + β 3 TREAT_POST it +β 4 ln(mfn) it + θ i + δ t + ε it (1) where y it is the outcome variables for product i in year-month t. We examine four outcome variables using Eq To estimate the possible export responses (i.e., export volume and export price) to primary trade 16

18 disputes, we use the outcome variables ln(export Volume) it and ln(export Price) it (i.e., the natural logarithm of export volume and the natural logarithm of export price at product level). For further analysis of the extensive and intensive margins of Chinese exports, the outcome variables in Eq. 1.1 are replaced by ln(exporters) it and ln(export_survive) it, which respectively represent the natural logarithm of the number of exporters, and the natural logarithm of the export volumes of all surviving exporters (i.e., exporters that continue to export after the initiation of trade disputes). TREAT i is a dummy variable that takes a value of 1 for products in the treatment group and 0 for products in the control group. POST it is a time dummy variable, constructed based on the following equation: POST it = { 1, if t t i0 0, if t < t i0 (2) where t i0 is the date of the initiation of primary trade disputes for product i; POST it equals 1 in postdispute months and 0 in pre-disputes months; TREAT_POST it is the DID term, an interaction of the variable TREAT and POST. The coefficient estimate of TREAT_POST it, β 3, is the DID estimator, measuring the average effect of trade disputes on products involved in dispute cases. ln(mfn) it is the natural logarithm of the U.S. s trade-weighted MFN tariff rates at the HS-4 digit product level; θ i represents product fixed effects, capturing all of the unobservable time-invariant product characteristics; δ t represents the yearmonth fixed effects, accounting for the shocks that are common to all products in time t; β 0 is a constant; and ε it is an error term. We cluster the standard errors at the HS-4 product level to address the potential heteroskedasticity and serial correlations. To estimate the effects of primary U.S.-China primary trade disputes on China s exports to the U.S. and to non-u.s. markets, we constructed the aforementioned four outcome variables for U.S. and non-u.s. markets. Using Eq. 1.1 we separately test their responses to trade disputes. Moreover, our estimation of the possible heterogeneous responses of these four outcome variables to different types of trade disputes is also based on Eq Specifically, we split the sample of trade disputes in which China is the complainant 17

19 country into two subsamples. One subsample consists of products involved in Type I disputes and their control group of unaffected products. The other contains the treated and control products under Type II disputes. The sample of trade disputes in which China is the defendant is also separated into Type I and Type II subsamples in the same way. Next, we test the responses of our four outcome variables to each type of trade disputes using Eq. 1.1 and then compare them. The consistent estimation of the DID terms in Eq. 1.1 and Eq. 1.3 hinges upon the assumption that the error terms of the treatment and control groups follow common trends. That is, the difference of the error terms between pre- and post-dispute periods for the products in the treatment group should equal the corresponding difference for products in the control group: E[ ε it TREAT i = 1] = E[ ε it TREAT i = 0]. (3) To check the robustness of our results, we follow Imbens and Wooldridge (2009) and include a time trend term in our baseline specification (Eq. 1.1): y it = β 0 + β 1 TREAT i + β 2 POST it + β 3 TREAT_POST it + β 4 ln(mfn) it + θ i + θ i t + δ t + ε it (4) This allows for possible differences in time trends between the treatment group and control groups. We also examine the robustness of our findings to data frequency. We re-estimate our baseline equation using quarterly data. The estimation equation is similar to Eq. 1.1, with only a change in the time subscript, t, standing for quarter rather than month. More details are discussed in Section Empirical Results In this section, we empirically investigate the possible responses (in terms of the export volumes, export prices, and extensive and intensive margins) of Chinese exporters to trade disputes with the U.S. from 2000 to Our baseline findings on primary trade disputes are presented in Sub-section to

20 We also compare these results to those of WTO disputes in Sub-section Export Volume Responses In this sub-section, we examine the response of export volumes to primary trade disputes at the product level. Before presenting the empirical results, we first plot the time trends of the export volume of products in the treatment and control groups to both the U.S. and non-u.s. markets over the pre- and post-trade dispute time periods in Figure 1.1. Specifically, the left side of Figure 1.1a shows the time trends of export volumes to the U.S. and the right side of Figure 1.1b shows time trends of export volumes to non-u.s. markets. The dashed vertical line in the figure points to the month in which the primary trade disputes were initiated (month = 0). These two figures demonstrate several clear patterns. First, the export volumes of products in both the treatment and control groups show similar upward trends before the initiation of the trade disputes in both figures, indicating that in general China s export volumes to the U.S. and to non-u.s. markets are increasing, consistent with China s export pattern over this period. Second, the export volumes of treated products to the U.S. show substantial increases after the initiation of trade disputes (month > 0) compared to the products in the control group. Third, for export volumes to non-u.s. markets, products in the treatment group do not seem to increase as much as those in the control group. [Figure 1 Inserted Here] The results of the empirical estimations of export volume responses to trade disputes at the product level by Eq. 1 are reported in Table 4. Specifically, Column (1) reports the effects of all primary trade disputes between the U.S. and China from 2000 to 2006 on the export volumes of the disputed Chinese products to the U.S. market. The coefficient of TREAT_POST it, β 3, is the DID estimate, measuring the average effect of trade disputes on the affected products. As shown in the first column, β 3 is positive and significant at 5%, indicating that the Chinese products involved in primary trade disputes are exported to 19

21 the U.S. market more in post-dispute periods. This is consistent with the time trends of the export volumes of both treated and control products shown in Figure 1.1. In terms of magnitude, the DID estimate of 0.34 indicates a 40.5% [=100*(e )] increase in the export volumes of disputed products to the U.S. after the initiation of U.S.-China trade disputes. The coefficient of ln(mfn) it is positive but insignificant. Columns (2) and (3) of Table 1.4 report the export volume responses to trade disputes initiated by China (i.e., China is the complainant) and to trade disputes initiated by the U.S. (i.e., China is the defendant), respectively. β 3 is significantly positive in Column (2) but negative and insignificant in Column (3). Thus, the significant trade promotion effects of all trade disputes found in Column (1) are mainly attributable to the trade disputes in which China is the complainant country. On average, export volumes to the U.S. increase by 60.5% [=100*(e )] after China initiates trade disputes against the U.S. Similarly, Columns (4)-(6) of Table 1.4 report the effects of the U.S.-China trade disputes on Chinese export volumes to countries other than the U.S. We find significant negative effects of all trade disputes on China s export volumes of disputed products to non-u.s. markets in Column (4). These negative effects are also caused by trade disputes initiated by China. As shown in Column (5), β 3 is found to be and significant at 1%, indicating that the average effect of the trade disputes initiated by China against the U.S. on China s export volumes to non-u.s. markets is -26.5% [=100*(e )]. Column (6) shows that the DID estimate for trade disputes initiated by the U.S. against China is positive and insignificant. The results in Column (2) and (5) together reveal that, after China initiates trade disputes against the U.S. s various trade restriction measures, Chinese exporters increase their export volumes of involved products to the U.S., but largely reduce their export volumes of the same products to countries other than the U.S., implying trade deflection. [Table 4 Inserted Here] 20

22 4.2 Export Price Responses The time trends of the export prices of products in the treatment and control groups over the pre- and posttrade dispute periods are plotted in Figure 2. The left side of Figure 1.2a shows the price trends of products exported to the U.S. The export prices of treated and control products generally remain stable and do not exhibit any differential time trends before the initiation of trade disputes. However, the export prices of products involved in disputes decrease sharply after the initiation of trade disputes, whereas the prices of products in the control group remain stable. The price trends of exports to non-u.s. markets are presented on the right side of Figure 1.2b. In pre-dispute periods, the export prices of both the treated and control products exhibit similar time trends. After trade disputes are initiated, there is a pronounced upward trend in the export prices of both the treatment and control group products. The increases in the prices of disputed products are larger than that of the control products. [Figure 2 Inserted Here] The empirical results of the export price responses to trade disputes estimated by Eq. 1 are presented in Columns (1)-(3) and (4)-(6) of Table 1.5, where the dependent variable of the (logarithm of) export prices to the U.S. and to non-u.s. markets are used, respectively. As shown in Column (1), the DID estimate for all trade disputes is negative and significant at 1%. Importantly, we find that the export prices of Chinese products to the U.S. only respond to trade disputes initiated by China, as shown in Column (2). In terms of magnitude, the export prices of disputed Chinese products to the U.S. decreases by 19.7% [=100*(e )] on average after the initiation of trade disputes initiated by China against the U.S. This finding is generally consistent with the corresponding export volume responses shown in Column (2) of Table 1.4 in the previous section. After China complains about the U.S. s WTO-inconsistent safeguard measures, the export quantities of disputed Chinese products to the U.S. largely increases. As the export prices to the U.S. are normally higher than those to non-u.s. markets (see Figure 1.2), firms are more likely 21

23 to divert their exports from non-u.s. markets to the U.S. after illegal trade barriers are lifted. As more firms enter the U.S. market and expand their export volumes to the U.S., the competition in the U.S. export market definitely increases. Hence, Chinese exporters may set lower prices to reach a larger market when facing more competition. With the result of such a price-setting strategy, we observe a substantial export price drop after trade disputes against the U.S. However, as shown in Column (3), trade disputes initiated by the U.S. against China do not significantly affect Chinese export prices to the U.S. As the U.S. has always complained about China s dumping into the U.S. market, Chinese exporters have no reason to further drop their export prices in response to the U.S. trade disputes. They are more likely to maintain their prices or even slightly increase them to avoid any potential anti-dumping actions by the U.S. Furthermore, the export prices to non-u.s. markets generally do not change in response to U.S.-China trade disputes, regardless of whether the cases are initiated by China or the U.S. The DID estimates for all primary trade disputes, trade disputes initiated by China, and disputes initiated by the U.S. are all positive and insignificant, as shown in Columns (4)-(6) of Table 5. This indicates that Chinese exporters do not raise the export prices of disputed products to non-u.s. markets in post-dispute periods, although they greatly reduce the quantity of their exports to non-u.s. markets. [Table 5 Inserted Here] 4.3 Heterogeneous Responses: Different Types of Trade Disputes As mentioned in Section 1.2, we classify primary trade disputes between the U.S. and China into two types. Type I cases generally refer to trade disputes regarding the U.S. s (potential) safeguard measures in the forms of either temporary tariff increases or other non-tariff measures, such as import bans and import quotas, to restrict the import of certain Chinese products. The majority of these disputes are initiated by China, arguing that such trade protection measures from the U.S. are inconsistent with the WTO s pre- 22

24 requisites for imposing safeguard measures. Type II cases mainly refer to the trade disputes concerning the U.S. s (potential) anti-dumping actions toward certain Chinese products. These disputes are raised either by the U.S., complaining about China s dumping into the U.S., or by China, defending itself or protesting against the U.S. s (potential) anti-dumping actions. The main difference between these two types of disputes lies in the trade policies that are being challenged by the complainant. For Type I cases, the WTO agreements that are cited by China in its complaints against the U.S. are the safeguard measures under Article XIX of the GATT. In Type II disputes, complainant countries all cite the WTO agreement of the imposition of anti-dumping measures. This essential difference probably implies different trade responses of Chinese exporters. In this section, we explore the possible heterogeneous trade responses of Chinese exporters to different types of primary disputes. Here, we focus on trade disputes initiated by China against the U.S., as these disputes play a dominant role in affecting the trade outcomes of disputed Chinese products, as shown in previous sections. Table 6 reports the effects of trade disputes on China s export volumes and prices to the U.S. for both Type I disputes, as shown in Column (1) and (2), and Type II disputes, as shown in Column (3) and (4)). For Type I cases, we find a significant positive effect on export volume and a significant negative effect on export price, which is consistent with the overall effects of trade disputes initiated by China as shown in the second columns of Tables 1.4 and Table 1.5. The DID estimate of presented in Column (1) of Table 1.6 implies that on average the export quantities of disputed products to the U.S. increase by 75.4% [=100*(e )] after the initiation of Type I disputes by China. The export prices to the U.S. decrease by 19.3% [=100*(e )], as shown in Column (2) of Table 1.6. Moreover, as shown in Column (2), the coefficient of the U.S. MFN tariff rates is positive and significant at 1% when the logarithm of the export price is used as the dependent variable. Thus, higher tariff rates from the U.S. are associated with higher export prices of Chinese products. For Type II disputes, the DID estimate of export volume, shown in Column (3), is negative but 23

25 insignificant. This is contrary to the positive and significant estimate for Type I disputes. As shown in Column (4), the coefficient of export price is positive and insignificant, which contrast the negative and significant coefficient of Type I cases. However, the direction of the trade effects of Type II disputes (initiated by China against the U.S. s anti-dumping measures) is generally in line with the finding of lower export volumes and higher export prices of foreign exporters as the result of anti-dumping duties in the literature (see the survey by Blonigen and Prusa, 2015). [Table 6 Inserted Here] The effects of Type I and Type II disputes on export volume and prices to non-us markets are reported in Table 7. In Column (1), the DID estimate of export volume for Type I disputes is significantly negative (-0.332), indicating that the export volumes of disputed Chinese products to non-u.s. decrease by 28.2% after Type I trade disputes are raised by China. The estimated β 3 of export price is positive but insignificant; it is shown in Column (2). These findings are consistent with the trade responses to all trade disputes initiated by China against the U.S. shown in the fifth columns of Tables 1.4 and 1.5. That is, although the export volumes of the involved products to non-u.s. markets drop, the export prices barely changed. However, the trade responses to Type II disputes seem to be different from the responses to Type I disputes. Regarding the export volume to non-u.s. markets, the coefficient of the DID term for Type II disputes, shown in Column (3) of Table 1.7, is positive and insignificant, which is contrary to the negative and significant coefficient for Type I cases. Furthermore, as shown in Column (4) of Table 1.7, Type II disputes positively affect the export price to non-u.s. markets, such that they increase by 15.1% after the initiation of Type II disputes. [Table 7 Inserted Here] In summary, we find heterogeneous trade responses of Chinese exporters to these two types of disputes in terms of their export volumes and prices to both the U.S. and non-u.s. markets. More importantly, the overall positive trade effects found in the previous two sections are largely attributable to 24

26 disputes regarding the U.S. s safeguard measures (Type I). However, the trade effects of trade disputes on anti-dumping measures (Type II) are all opposite in sign and largely insignificant. As non-wto trade disputes all relying on bilateral negotiations to research resolutions, the different trade outcomes of these two types of disputes should not be due to the different settlement procedures. It s more likely that different settlement results lead to different trade responses. The U.S. and China may successfully work out a solution for Type I disputes such that after the U.S. complies by removing illegal safeguard barriers, the export quantities of disputed Chinese disputed products to the U.S. increases substantially. In contrast, China and the U.S. probably fail to resolve their disputes on anti-dumping measures, such that the export quantities of these products experience no significant changes. One fundamental difference between safeguard and anti-dumping measures may account for the different settlement results of these two types of disputes. The imposition of anti-dumping measures requires the finding of unfair trade practice while that of safeguard measures does not. Therefore, it may be easier for both parties to negotiate a resolution for disputes on safeguard measures which are imposed on fair trade. 4.4 Extensive and Intensive Margins We find a significant promoting effect of primary trade disputes initiated by China against the U.S., especially Type I disputes, on the export volumes of involved Chinese products to the U.S. We also find a significant dampening effect on the export volumes of said products to non-us markets. In this section, we further investigate the mechanisms of these trade effects by examining the extensive margin (i.e., the number of exporters to the U.S. and to non-u.s. markets) and the intensive margin (i.e., the export volumes of surviving exporters to the U.S. and non-u.s. markets) of Chinese exports. We define surviving exporters as those firms that exported their products (at the HS-4 level) to the U.S. (or to non-u.s. markets) in pre-dispute periods and continue to export the same products in post- 25

27 disputes periods. Thus, Chinese firms that exit the export market or enter the export market after trade disputes are initiated are all excluded from the intensive margin sample. In our full dataset of export volumes to the U.S. from 2000 to 2006, there are a total of 74,864 Chinese exporters, 24,739 of which survive throughout the entire period. We keep these surviving exporters and then aggregate their export volumes at the HS-4 level. [Figure 3 Inserted Here] To examine the extensive margin of Chinese exports, we first plot the time trends of the number of Chinese exporters to the U.S. and non-u.s. in Figure 1.3. The left side of Figure 1.3a shows the number of exporters to the U.S. for both the treatment and control groups over the pre- and post-dispute periods. The number of exporters shows an upward trend for both product groups. However, after the initiation of trade disputes, the exporters of the treated products seem to increase more than those of the control products. In contrast, for exports to non-u.s. markets, the number of exporters does not show much different trends between the treatment and control groups as shown on the right side of Figure 3b. The time trends of surviving exporters export volumes to the U.S. and non-u.s. markets are shown in Figure 4. We first show the results for all trade disputes in Table 8. The extensive and intensive margin effects of Chinese exports to the U.S. are reported in Column (1) and (2). The DID estimates of the number of exporters (extensive margin) and the export volumes of surviving exporters (intensive margin) to the U.S. market are both positive and significant at 1%. The corresponding estimates for non-u.s. markets are insignificant, as shown in Column (3) and (4) of Table 8. We further examine trade disputes initiated by China and find that the effects on the extensive and intensive margins of Chinese exports to the U.S. and to non-u.s. markets (see Table 9) are all consistent with the results for all trade disputes. However, trade disputes initiated by the U.S. against China have no significant effects on either margin. Thus, of all bilateral U.S.-China trade disputes, Chinese exporters only respond to disputes initiated by China against the U.S. Specifically, the substantial trade promoting effect on export volume to the U.S. is driven by two forces 26

28 one is more Chinese firms entering the U.S. export market (extensive margin) and the other is the increased export quantities of surviving exporters (intensive margin). On average, trade disputes initiated by China generate an effect of 66% on the number of exporters to the U.S. market, as shown in Column (1) of Table 9, and an effect of 167% on the export quantities of surviving exporters to the U.S., as shown in Column (2). [Table 8 & 9 Inserted Here] As we document in Section 4.3, the trade effects of trade disputes initiated by China are mainly attributable to Type I disputes (regarding the U.S. s safeguard import restrictions), whereas Type II disputes (regarding the U.S. s antidumping measures) have no significant effects. We further report the extensive and intensive margin effects of Type I and Type II disputes in Tables 10 and 11 respectively. Columns (1) and (2) of Table 1.10 present the results for Chinese export to the U.S. β 3 of both the extensive and intensive margins are positive and significant at 1%, implying that the number of Chinese exporters and the export volumes of surviving exporters to the U.S. increase by 82.6% and 215.5%, respectively, after China s initiation of Type I disputes against the U.S. The results of Type I disputes on exporters to non- U.S. markets are shown in Column (3) and (4) of Table 10. Column (3) shows that the DID estimate of the extensive margin is negative and significant but that of the intensive margin is positive and insignificant, as shown in Column (4). The number of exporters of disputed products to non-u.s. markets decreases by 12.5% in post-dispute period. In contrast, we find no significant effect of Type II disputes on the extensive or intensive margins of Chinese exports to the U.S. and to non-u.s. markets in Table 11. This indicates that neither the number of exporters nor the export quantities of surviving exporters are affected by Type II disputes. This explains why the overall effects of Type II disputes on export volume are insignificant. [Table 10 & 11 Inserted Here] These results support our finding of the dominant effects of Type I disputes on China s exports among all trade disputes in the previous section. In summary, the export promoting effect of Type I disputes on 27

29 the export of disputed Chinese products to the U.S. occurs as a result of more Chinese exporters entering the U.S. market (extensive margin) and higher export volumes of surviving exporters to the U.S. (intensive margin). The export deteriorating effect on the export of the same products to non-u.s. markets occurs through the extensive margin rather than the intensive margin. Specifically, less Chinese firms export to non-u.s. markets but more enter into the U.S. market after Type I dispute initiation, reflecting a notable trade deflection. However, the export volume of surviving exporters to non-u.s. markets remain unchanged in post-dispute periods. 5 Robustness Checks In this section, we report the empirical results of a series of robustness checks. Our main results of the trade responses to primary trade disputes remain robust when additionally controlling for time trends, using quarterly data, excluding the first and last years dispute cases, and excluding all overlapping trade disputes with third countries. 5.1 Controlling for Product-Specific Time Trends To address the concern that products in the treatment and control groups follow different time trends, we include a product-specific time trend term, θ i t, in the baseline estimation Eq. 1. The results are reported in Table A.1 of the Appendix. The trade disputes still exert positive and significant effects on the export volumes of Chinese products to the U.S., as shown in Column (1) of Table A.1, and negative and significant effects on export volumes to non-u.s. markets, as shown in Column (3). Moreover, the export prices to the U.S. significantly decrease after trade disputes initiations, as shown in column (2), whereas the export prices to non-u.s. markets demonstrate little change, as shown in Column (4). These effects are also similar in magnitude to our baseline effects in Tables 4 and 5. Therefore, our main findings remain robust even when product-specific time trends are controlled, confirming the validity of our DID 28

30 estimations. [Table A.1 Inserted Here] 5.2 Use of Quarterly Data We further test the robustness of our main results by using quarterly rather than monthly data to address the possible concern that monthly data are much more volatile, as firms may not export to foreign markets every month. We aggregate the trade transaction data of Chinese exporters from the monthly to the quarterly level. The total number of observations for exports to the U.S. and non-u.s. is 13,769 and 13,969, respectively. The regression results using quarterly data are reported in Table A.2 of the Appendix. As shown in Columns (1)-(2), trade disputes positively and significantly affect China s export volumes to the U.S., but negatively and significantly affect export prices to the U.S., consistent with our baseline results from Sub-section 4.1 and 4.2. The average effects on export volumes and prices to the U.S. are found to be 38.5% and -16.9%, respectively, which are similar to our baseline findings. The DID estimate of export volume to non-u.s. markets is still negative but insignificant. Thus, our main results of Chinese exporters trade response to trade disputes remain robust when using quarterly data. [Table A.2 Inserted Here] 5.3 Exclusion of Trade Disputes in the First and Last Years of Our Sample To alleviate the concern that the pre- or post-disputes periods of cases occurring in the very beginning or end of our sample period (i.e ) may not be long enough to carry out DID estimations, we exclude dispute cases from 2000 to 2006 from our sample. The observation of U.S. and non-u.s. samples are thus reduced to 39,086 and 40,323, respectively. The regression results of the remaining disputes are reported in Table A.3 of the Appendix. Our main findings remain robust, as the direction and magnitude 29

31 of the trade dispute effects on export volumes and prices to the U.S. and non-u.s. markets are similar to those from the full sample, implying that our DID estimates are valid. [Table A.3 Inserted Here] 5.4 Exclusion of Overlapping Trade Disputes with Third Countries As we discussed in Section 4.5, other countries may also initiate the same trade disputes (i.e., regarding the same products in the same time) as our primary U.S.-China disputes. The existence of third countries may confound the effects of disputes on Chinese exports to the U.S. or to non-u.s. markets. To address this problem, we drop two such overlapping dispute cases (one from early 2001 and the other from late 2004, from our sample. The regression results for the remaining bilateral U.S.-China disputes are shown in Table A.4 of the Appendix. The effects of primary trade disputes on Chinese exporters trade volumes and prices to the U.S. and non-u.s. markets remain robust in this sample. More importantly, the magnitudes of these effects turn out to be much larger. For instance, the average effect of trade disputes on export volume to the U.S. is 44% (40% in the full sample) and the average effect on export volumes to non-u.s. markets is % (-15.2% in the full sample). Moreover, the resulting export prices to the U.S. are also more profound, showing an average effect of -20% (-18% in the full sample). [Table A.4 Inserted Here] 6 Conclusions Trading partners are frequently involved in trade disputes. Although WTO trade disputes receive the most attention in the literature, they only constitute a tiny fraction of all occurring trade disputes (primary trade disputes). We provide the first attempt to empirical examination of the trade effects of primary trade disputes. We focus on all U.S.-China dispute cases between 2000 and 2006 and utilize the monthly 30

32 product-level (HS-4 digit) export transaction data of Chinese firms over the same period to estimate the trade responses. Using DID estimation, we find a significant trade-facilitating effect on Chinese exports to the U.S. after China s initiation of trade disputes against the U.S. s WTO-inconsistent trade protection policies. This result is generally consistent with the literature on WTO disputes. Chinese exports to non- U.S. markets decrease, indicating a substantial trade deflection. In addition, export prices to the U.S. fall substantially, whereas those to non-u.s. markets remain unchanged. Importantly, disputes regarding the illegal safeguard measures of the U.S. play a dominant role in generating these trade effects. We find that the trade promoting effect results from more exporters entering the U.S. (extensive margin) and from increasing export quantities of surviving exporters (intensive margin); the trade deflection effect is largely attributable to extensive margin. That is, some Chinese firms shift their exporting destination from non- U.S. markets to the U.S. in response to trade disputes. As previous studies have focused on WTO trade disputes, our investigations of exporters trade responses to primary trade disputes (i.e., both WTO and non-wto disputes) fill the void by presenting a more complete picture of the trade effects of trade disputes. The examination of various trade outcomes (i.e., export quantity, export price, number of exporters, export volumes of surviving exporters, and trade deflection) give us a more profound and comprehensive understanding of the effects of primary trade disputes, especially non-wto disputes. However, the lack of information on the durations and settlement processes of the primary trade disputes in our sample largely limits our examination of exporters responses to trade disputes in different stages of disputes settlement. Tracking and collecting such information is challenging and left for further work. With the rising recognition of the importance of non-wto disputes, future research on trade or other economic and political effects resulting from non-wto disputes would be beneficial. The comparison of the different initiation processes or settlement approaches of WTO and non-wto disputes may also generate meaningful policy implications. Why are some trade protection policies raised by complainants 31

33 as formal WTO trade disputes and other trade barriers left for bilateral or multilateral negotiations outside of formal litigations? What determines complainants course of action? Which strategy is more effective in resolving disputes? These important issues are still unexplored yet and left for future research. References Bagwell, K. and R. W. Staiger. (2002).The Economics of the World Trading System. Cambridge: MIT Press. Bagwell, K. and R. W. Staiger. (2005). Enforcement, Private Political Pressure, and the General Agreement on Tariffs and Trade/World Trade Organization Escape Clause. Journal of Legal Studies, 34(2), Bagwell, K. and R. W. Staiger. (2011). What Do Trade Negotiators Negotiate About? Empirical Evidence from the World Trade Organization. American Economic Review, 101(4), Bechtel, M.M. and T. Sattler. (2015). What is Litigation in the World Trade Organization Worth? International Organization, 69(2), Blonigen, B.A. and C. P. Bown. (2003). Antidumping and Retaliation Threats. Journal of International Economics, 60(2), Blonigen, B. A. and T. J. Prusa. (2015). Dumping and Antidumping Duties. NBER Working Paper Series, No Bown, C. P. (2002). Why are Safeguards under the WTO So Unpopular? World Trade Review, 1(1), Bown, C. P. (2004). On the Economic Success of GATT/WTO Dispute Settlement. The Review of Economics and Statistics, 86(3), Bown, C. P. (2005). Participation in WTO Dispute Settlement: Complainants, Interested Parties and Free Riders. World Bank Economic Review, Bown, C. P. (2009). Self-Enforcing Trade: Developing Countries and WTO Dispute Settlement. Washington, DC: Brookings Institution Press. Bown, C. P. and M. A. Crowley. (2007). Trade Deflection and Trade Depression. Journal of International Economics, 72(1), Bown, C. P. and M. A. Crowley. (2013). Self-Enforcing Trade Agreements: Evidence from Time-Varying 32

34 Trade Policy. American Economic Review, 103(2), Bown, C.P. and K. M. Reynolds. (2015). Trade Flows and Trade Disputes. Review of International Organizations, 10(2), Broda, C., Limão, N. and D. E. Weinstein. (2008). Optimal Tariffs and Market Power: The Evidence. American Economic Review, 98(5), Carter, C. A. and C. Gunning-Trant. (2010). U.S. Trade Remedy Law and Agriculture: Trade Diversion and Investigation Effects. Canadian Journal of Economics, 43(1), Chaudoin, S., Kucik, J., and K. Pelc. (2016). Do WTO Disputes Actually Increase Trade? International Studies Quarterly, 0, Davis, C. (2011). Why Adjudicate? Enforcing Trade Rules. Princeton: Princeton University Press. Egger, H., Egger, P., and D. Greenaway, (2008). The trade structure effects of endogenous regional trade agreements. Journal of International Economics, 74(2), Fang, S. (2008). The Informational Role of International Institutions and Domestic Politics. American Journal of Political Science, 52(2), Feenstra, R. C., Romalis, J. and P. K. Schott. (2002). U.S. Imports, Exports and Tariff Data, NBER Working Paper Series, No Foster, N., Poeschl, J., and R. Stehrer, (2011). The impact of Preferential Trade Agreements on the margins of international trade. Economic Systems, 35(1), Horn, H. and P.C. Mavroidis. (2006). A Survey of the Literature on the WTO Dispute Settlement System. IFN Working Paper No. 684, Research Institute of Industrial Economics, Sweden. Horn, H., Mavroidis, P.C. and H. Nordstrom. (1999). Is the Use of the WTO Dispute Settlement System Biased? CEPR Discussion Paper Series No Imbens, G. W. and J. M. Wooldridge. (2009). Recent Developments in the Econometrics of Program Evaluation. Journal of Economic Literature, 47(1), Konings, J., Vandenbussche, H. and L. Springael. (2001). Import Diversion under European Antidumping Policy. Journal of Industry, Competition and Trade, 1(3), Kono, D. (2006). Optimal Obfuscation: Democracy and Trade Policy Transparency. American Political Science Review, 100(3), Lasagni, A. (2000). Does Country targeted Anti dumping Policy by the EU Create Trade Diversion? Journal of World Trade, 34(4), Li, T. and L. D. Qiu. (2015). Beyond Trade Creation: Free Trade Agreements and Trade Disputes, Working Paper. Lu, Y., Z. Tao, and Y. Zhang. (2013). How do exporters respond to antidumping investigations? Journal of International Economics, 91(2), Ludema, R. and A. M. Mayda. (2013). Do Terms-of-Trade Effects Matter for Trade Agreements? Theory and Evidence from WTO Countries. Quarterly Journal of Economics, 128(4),

35 Maggi, G. (1999). The Role of Multilateral Institutions in International Trade Cooperation. American Economic Review, 89(1), Maggi, G. and Robert W. Staiger. (2011). The Role of Dispute Settlement Procedures in International Trade Agreements. The Quarterly Journal of Economics, 126(1), Maggi, G. and R. W. Staiger. (2013). Trade Disputes and Settlement. University of Wisconsin Manuscript. Maggi, G. and R. W. Staiger. (2015). Optimal Design of Trade Agreements in the Presence of Renegotiation. American Economic Journal: Microeconomics, 7(1), Mansfield, E.D., Milner, H. V. and B. P. Rosendorf. (2002). Why Democracies Cooperate More: Electoral Control and International Trade Agreements. International Organization, 56(3), Park S. (2009). The Trade Depressing and Trade Diversion Effects of Antidumping Actions: The Case of China. China Economic Review, 20(3), Peritz, L. (2015). When are International Institutions Effective? UCLA Working Paper. Pierce, J. R. and P. K. Schott. (2015). The Surprisingly Swift Decline of U.S. Manufacturing Employment. American Economic Review, 106(7), Prusa, T. J. (1997). The Trade Effects of U.S. Antidumping Actions. In Robert C. Feenstra (ed.), The Effects of U.S. Trade Protection and Promotion Policies. Chicago, IL: University of Chicago Press, Prusa, T.J. (2001). On the Spread and Impact of Antidumping. Canadian Journal of Economics, 34(3), Rosendorff, B. P. (2005). Stability and Rigidity: Politics and Design of the WTO s Dispute Settlement Procedure. American Political Science Review, 99(3), Sattler, T. and T. Bernauer. (2011). Gravitation or Discrimination? Determinants of Litigation in the World Trade Organization. European Journal of Political Research, 50(2), Sattler, T. Spilker, G. and T. Bernauer. (2014). Does WTO Dispute Settlement Enforce or Inform? British Journal of Political Science, 44(4), Staiger, R. W. and A. O. Sykes. (2013). How Important can the Non-Violation Clause be for the GATT/WTO? NBER Working Paper Series, No Tomz, M Domestic Audience Costs in International Relations: An Experimental Approach. International Organization, 61(4),

36 Figure 1: Time Trends of Export Volume at HS-4 Digit Product Level Figure 1a Export Volumes to the U.S. Figure 1b Export Volumes to non-u.s. Markets Note: The left figure reports the time trends of China s export volume to US for the treatment and control groups; The right figure reports the time trends of China s export volume to non-us countries on average for the treatment and control groups; The horizontal axis is month, with positive figures representing the number of post-disputes months and negative figures representing the pre-disputes ones; The vertical dashed lines point to the initiation month of the trade disputes; The vertical axis represents the coefficient estimates of month dummies for treatment and control groups. 35

37 Figure 2: Time Trends of Export Price at HS-4 Digit Product Level Figure 2a Export Prices to the U.S. Figure 2b Export Prices to non-u.s. Markets Note: The left figure reports the time trends of the price of China s exports to US for the treatment and control groups; The right figure reports the time trends of the price of China s exports to non-us countries on average for the treatment and control groups; The horizontal axis is month, with positive figures representing the number of post-disputes months and negative figures representing the pre-disputes ones; The vertical dashed lines point to the initiation month of the trade disputes; The vertical axis represents the coefficient estimates of month dummies for treatment and control groups. 36

38 Figure 3: Time Trends of the Number of Exporters at HS-4 Digit Product Level Figure 3a: Number of Exporters to the U.S. Figure 3b: Number of Exporters to non-u.s. markets Note: The left figure reports the time trends of the number of Chinese exporters to US for the treatment and control groups; The right figure reports the time trends of the number of Chinese exporters to non-us countries for the treatment and control groups; The horizontal axis is month, with positive figures representing the number of post-disputes months and negative figures representing the pre-disputes ones; The vertical dashed lines point to the initiation month of the trade disputes; The vertical axis represents the coefficient estimates of month dummies for treatment and control groups. 37