WPS5223 Policy Research Working Paper 5223 Self-Enforcing Trade Agreements Evidence from Time-Varying Trade Policy Chad P. Bown Meredith A. Crowley The World Bank Development Research Group Trade and Integration Team March 2010 Policy Research Working Paper 5223 Abstract This paper estimates a model of a government making The analysis finds additional support for the theory that trade policy adjustments under a self-enforcing trade some US antidumping use is consistent with cooperative agreement in the presence of economic shocks. The behavior through a second empirical examination of empirical model is motivated by the formal theories of how trading partners responded to these new US tariffs. cooperative trade agreements. The authors find evidence Even after controlling for factors such as the expected that United States' use of its antidumping policy during cost and benefit to filing a WTO dispute or engaging in 1997­2006 is consistent with increases in time-varying antidumping retaliation, the analysis find that trading "cooperative" tariffs, where the likelihood of antidumping partners are less likely to challenge such "cooperative" US is increasing in the size of unexpected import surges, antidumping tariffs that were imposed under terms-of- decreasing in the volatility of imports, and decreasing trade pressure suggested by the theory. in the elasticities of import demand and export supply. This paper--a product of the Trade and Integration Team, Development Research Group--is part of a larger effort in the department to evaluate the impact that international institutions have on the market access. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at cbown@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 1 Introduction For decades, theorists of international commercial policy have demonstrated how purely economic incentives affect government tariffs and the decision to enter into trade agreements.1 Long thought to be more an intellectual curiosity than a determinant of how governments actually set their trade policies in practice, the terms-of-trade theory has recently become a focus of considerable empirical research. Two important contributions document how trade policy formation is determined by economic incentives in addition to the more generally accepted political economy or income redistribution a motives. Broda, Lim~o and Weinstein (2008) provide two pieces of evidence broadly consistent with the idea that countries exploit their market power in trade. First, they estimate disaggregate foreign export supply elasticities and find that countries that are not members of the World Trade Organization (WTO) systematically set higher tariffs on goods that are supplied inelastically. Sec- ond, for a WTO member like the US, they find that trade barriers on products not covered by the WTO agreement are significantly higher when the US has more market power. In a separate setting, Bagwell and Staiger (2009) focus on a set of countries newly acceding to the WTO between 1995 and 2005 in order to examine the role of market power in negotiating tariffs for new WTO members. They find evidence consistent with the theory of the importance of the terms-of-trade effect; the tariff to which a country negotiates is further below its non-cooperative level, the larger is its non-cooperative import volume. The current paper contributes to this empirical literature by estimating whether similar eco- nomic incentives explain government use of time-varying trade policies. In particular, we study how countries adjust their trade policies over time in response to economic shocks. The earlier empirical literature has examined the cross-sectional variation in a country's tariff level (Broda, a Lim~o and Weinstein, 2008) or the magnitude of a country's tariff reduction when moving from 1 Irwin (1996) provides a full account of the intellectual history of the terms-of-trade (a.k.a. optimal tariff) theory, which he finds dates back at least to Robert Torrens in the early nineteenth century. For more recent treatments, see the seminal work of Johnson (1953-1954) and for integration into the theory of trade agreements, see Bagwell and Staiger (1990, 1999). 1 a non-cooperative policy to a trade agreement (Bagwell and Staiger, 2009). We focus on a WTO member's time-varying tariff increases in the face of trade volume shocks whose influence may vary due to heterogeneity across import demand and export supply elasticities. To give our exercise context, it is important to note that while trade agreements like the WTO do require member countries to accept an upper limit on their permissible applied tariffs, there are exceptions to WTO rules which allow governments to raise their import-restricting policies above those upper tariff limits under certain conditions. The most prevalent exception used by industrialized countries and a number of middle income developing countries is the antidumping (AD) import-restricting policy, which is the policy forming the basis for our empirical analysis. One major contribution of this paper is to provide a first formal examination of the empirical relationship between the terms-of-trade theory and antidumping policy. The basic intuition behind our empirical approach follows the theoretical model of Bagwell and Staiger (1990), a repeated game in which two countries trade in the presence of unanticipated trade volume shocks. Bagwell and Staiger model a self-enforcing trade agreement as the set of cooperative tariffs that countries can support over time.2 They show that the incentive for a country participating in such a trade agreement to raise its tariff is greater when import volumes are larger than expected. In the presence of a positive import volume shock, a self-enforcing agreement to maintain low tariffs cannot be sustained in sectors with low import demand or export supply elasticities - either the cooperative tariff supported by the trade agreement will rise or one country will unilaterally defect from the agreement and impose its higher, noncooperative tariff. We empirically evaluate the Bagwell and Staiger (1990) model's ability to explain changes to US trade policy over the 1997-2006 period. In particular, we examine the extent to which US antidumping can be understood as an increase in a "cooperative" trade policy - i.e., increasing the tariff above a trade agreement's contractual level in order to manage import pressure so as to prevent the country from permanently defecting from the agreement. We find evidence from this period 2 To our knowledge, the only attempt to test implications of the Bagwell and Staiger (1990) approach is Prusa and Skeath (2002), which is much different from the exercise undertaken here, especially in that they do not look to formally model the impact of terms of trade motives for antidumping use. 2 that US antidumping use is consistent with increases in time-varying cooperative tariffs, where the likelihood of antidumping use is increasing in the size of unexpected import surges, decreasing in the volatility of imports, and decreasing in the elasticities of import demand and export supply. We thus provide new evidence that purely economic incentives are sizable determinants of US antidumping use, even after controlling for other political-economic determinants that have been the traditional focus of research. While a substantial literature examines a country's use of antidumping policies in response to macro-economic or industry-level economic fluctuations (e.g., Finger, Hall and Nelson, 1982; Feinberg, 1989; Knetter and Prusa, 2003; Irwin, 2005; Crowley, 2008), ours is the first to examine the relationship among time-varying trade policies, bilateral trade flows, micro trade elasticities, and characteristics of the importing country's domestic industries based on a theoretical framework in which countries participate in a self-enforcing trade agreement. While our estimates of US antidumping use are consistent with the influence of terms-of-trade motives in a cooperative trade agreement, we pause at this stage to more carefully interpret these results. First, we do not claim that the US imposes antidumping tariffs to take advantage of terms- of-trade gains, and we also do not claim that US antidumping duties are set as optimal tariffs. To better understand antidumping use as a theoretically-motivated cooperative tariff increase, consider the potential trade policy alternative were a similar economic shock to hit a country signed on to a trade agreement that did not provide access to a policy like antidumping. Because economic shocks to trade volumes may increase the incentive for governments to defect from the agreement and impose noncooperative (Nash) tariffs, we interpret some US antidumping use as a "necessary evil" that prevents the more draconian defection outcome and a potential systemic breakdown of the agreement.3 Do trading partners view US use of antidumping as such a necessary evil that is part of a cooperative, self-enforcing trade agreement? In the second half of the paper, we specifically address 3 This is consistent with the cooperative tariff interpretation of Bagwell and Staiger (1990, p. 780, emphasis added) which states "Countries can cooperatively utilize protection during periods of exceptionally high trade volume to mitigate the incentive of any country to unilaterally defect, and in so doing can avoid reversion to the Nash equilibrium. Thus, surges in the underlying trade volume lead to periods of "special" protection as countries attempt to maintain some level of international cooperation." 3 this question. In particular, suppose under the WTO regime that antidumping tariffs arise in sectors in which import demand and export supply elasticities are low and import volumes are unexpectedly high; i.e., where the use of antidumping policy is consistent with a cooperative increase in import protection to prevent defection from the agreement. How do the adversely affected foreign exporters respond to US imposition of new antidumping tariffs? The raw and anecdotal data on WTO trading partners' responses to US imposing new an- tidumping tariffs suggests that foreign countries certainly do not always view such tariff increases as cooperative. Out of roughly 130 US antidumping tariffs imposed between 1997 and 2006, trading partners formally challeged 27 for removal at WTO dispute settlement (Bown, 2009a).4 Further- more, trading partners have responded with dozens of other "extra-WTO" challenges to new US AD tariff increases with their own antidumping retaliation against US exports in the same sector. The second half of the paper exploits the substantial variation in the foreign trading partner response to the imposed US AD tariffs. The Bagwell and Staiger (1990) theory suggests that the response of WTO members to a new US antidumping tariff may vary according to the likelihood that it was imposed for "cooperative" reasons (motivated by economic shocks). More formally, our second step is to develop an empirical model to examine the consistency of the data with this interpretation of foreign challenges to US antidumping tariff increases between 1997-2006. Our results indicate that foreign trading partners are indeed less willing to challenge US antidumping tariff increases that are associated with the economic shocks that are significant determinants of the US trade policy changes. Our estimates indicate that the economic importance of this relationship is quite sizable, and it holds even once we introduce other control variables that capture the costs and benefits to the trading partner of WTO dispute settlement use, as motivated, for example, by the Maggi and Staiger (2008) theory of WTO dispute settlement.5 These results are robust to extensions of the model in which we extend the trading partner's policy response action space 4 While challenging an imposed AD tariff is quite a common subject for WTO trade disputes, the frequency of challenges to US tariffs is much larger than the instances facing any other WTO member country. 5 The second part of the paper fits into a related empirical literature examining determinants of WTO dispute set- tlement use. Consistent with the results we provide, much of the systematic differences can be explained by expected political-economic determinants. See, for example, Horn, Mavroidis and Nordstrom (2005), Bown (2004a,b; 2005a,b) and Busch, Reinhardt, and Shaffer (2008). Nevertheless, one shortcoming of this literature that the current paper 4 to include an additional "extra-WTO" challenge to the US antidumping tariff increase via the country's own antidumping retaliation against US exporters in the same sector. The rest of this paper proceeds as follows. Section 2 briefly reviews the Bagwell and Staiger (1990) theory that we use to motivate antidumping tariffs modeled as a trade policy change in a cooperative agreement. There we highlight the theoretical predictions regarding how economic shocks to trade volumes, in the presence of low export supply and import demand elasticities, af- fect time-varying trade policies. Section 2 also introduces our empirical model of US antidumping formation and a discussion of the underlying data used in this portion of the estimation. Section 3 presents our first estimates of the relationship between the terms-of-trade theory and US use of antidumping over the 1997-2006 period. In section 4 we extend the analysis by examining determi- nants of whether WTO trading partners view US AD tariffs as acts of "cooperative" adjustment to trade policy. There we estimate a model of foreign trading partners challenging US AD both inside and outside the WTO. Finally, section 5 concludes. 2 A Theoretical and Empirical Model of Antidumping Tariffs un- der a Cooperative Trade Agreement 2.1 The Bagwell and Staiger (1990) theory Bagwell and Staiger (1990) develop a model in which two countries play a repeated game char- acterized by a terms-of-trade driven prisoner's dilemma.6 These countries can support a welfare- improving equilibrium of lower "cooperative" tariffs with the threat of infinite reversion to the Nash equilibrium of the stage game. Bagwell and Staiger showed that the lowest cooperative tariff that can be supported at any time fluctuates in response to unexpected changes in import volumes: in response to a positive import volume shock, the lowest cooperative tariff that can be supported increases. Because the importing country's terms-of-trade gains from deviating from the agreement ultimately seeks to remediate is that the formation of the policies that are actually challenged may be endogenously affected by the dispute settlement process. 6 Bagwell and Staiger (2003) consider additional theoretical extensions to this model. 5 are higher when import volumes are larger, it becomes infeasible to support the tariff originally negotiated under the trade agreement. In these circumstances, the cooperative tariff supported by the trade agreement rises for the duration of the import surge. In developing an empirical model of US antidumping tariff formation within the framework of US participation in the WTO, we are guided by the predictions of the Bagwell and Staiger model. Start from the tariffs in the cooperative trade agreement equilibrium. Then according to the Bagwell and Staiger (1990) model's equation (28), the incentive to raise tariffs is increasing in the size of the import volume shock and is decreasing in the export supply and import demand elasticities. The model thus provides an explanation for intertemporal and cross-sectional variation in trade policy changes by a member of the trade agreement that we ultimately take to the data. With regard to intertemporal variation, a given sector may have low export supply and import demand elasticities, but the model of self-enforcing trade agreements predicts that we should only observe in- creases in the tariff when the volume of imports in that sector rises unexpectedly. Cross-sectionally, if two sectors simultaneously experience unexpected import surges of the same size, the model pre- dicts that the tariff increase will be larger in the sector with less elastic export supply and import demand. Intuitively, when export supply is more inelastic, the terms of trade gain from a tariff is larger. When import demand is less elastic, the domestic efficiency costs of the tariff are smaller. Furthermore, temporary increases in trade protection will be larger when import surges are more "unusual." Equations (19) and (20) of Bagwell and Staiger (1990) together imply that the magni- tude of the tariff increase is greater for sectors in which import surges are uncommon. For import surges of the same size in two different sectors, the magnitude of the tariff increase will be larger in the sector with the lower variance of imports. To investigate whether the US uses its AD tariff formation policy to respond to these economic shocks, we use Bagwell and Staiger (1990) to develop the following expression for changes in its tariff.7 Consider an increase to the "cooperative" US tariff, given by the following expression: 7 In particular, see equations (19), (20), and (28). 6 1 ikt = f Mikt - E(Mikt ) (1) M ik (xk + mk ) where ikt is the change in the tariff against imports from i in industry k, M ik is the standard deviation of imports from country i in industry k, xk is the elasticity of export supply facing the US (importer) in industry k, mk is the elasticity of US import demand in industry k, Mikt is imports from country i in industry k at time t, and E(Mikt ) is the expected value of imports from country i in industry k at time t. To summarize the implications of Bagwell and Staiger (1990) for our approach, in a cooperative trade agreement equilibrium, the incentive for the US to temporarily raise its tariff is higher if (i) imports are larger than expected, (ii) the volatility (standard deviation) of imports is low, (iii) the export supply elasticity is low, and (iv) the import demand elasticity is low. We estimate empirical models of (1) on a panel of US imports to determine if the US's use of antidumping policy is broadly consistent with the US raising its tariff as in a cooperative trade agreement in which it was confronted by these (measurable) economic shocks. In section 2.2 we present the empirical model, in section 2.3 we describe the underlying data sources, and in section 3 we turn to the results. 2.2 Empirical model of US antidumping tariff formation In this section, we present our empirical models of US antidumping tariff formation to determine if the US is using antidumping policy in a manner consistent with Bagwell and Staiger (1990). We log linearize (1) and estimate a binary choice (probit) model of whether the US imposed an antidumping tariff on imports from country i in industry k in year t. In the binary model, the US government makes a decision of whether to impose an antidumping tariff according to a latent measure of the potential benefits to the US of imposing a tariff against country i in industry k. The latent measure d is unobserved, but takes the form d = xikt +ikt ikt ikt where i denotes the foreign country accused of dumping, k denotes the industry in which dumping is alleged to occur, and t denotes the time period in which the complaint is filed by the US industry. 7 The variables included in xikt are a measure of the import growth into the US from foreign country i in industry k in year t - 1, the standard deviation of imports from country i in industry k over the sample period, a dummy indicating if the export supply elasticity facing the US in industry k is relatively inelastic, a dummy indicating if the US import demand elasticity in industry k is relatively inelastic, the bilateral real exchange rate between the US and country i at t-1 (measured in foreign currency over US dollars) and, in some specifications, characteristics of the US industry at time t - 1. The US imposes an antidumping tariff according to the rule: 1 if d > 0 ikt dikt = (2) 0 if d 0 ikt Assuming ikt N (0, 1), then the likelihood is dikt 1-dikt L = ( xikt ) 1 - ( xikt ) (3) where is the standard normal cdf. Marginal effects derived from coefficient estimates obtained from maximizing the log of the likelihood (3) are reported in tables 2 - 8. 2.3 Data used to estimate US antidumping tariff formation We estimate the empirical model of US antidumping tariff formation on a panel dataset con- structed from several primary data sources: (1) trade policy data for the US come from the Global Antidumping Database (Bown, 2009a), (2) US bilateral imports at the industry-level come from the US International Trade Commission's DataWeb, (3) industry-level foreign export supply elas- a ticities facing the US come from Broda, Lim~o, and Weinstein (2008), (4) industry-level US import demand elasticities come from Broda, Greenfield, and Weinstein (2006), (5) variables describing the characteristics of US domestic industries come from the US Census Bureau, and finally, (6) annual 8 bilateral real exchange rates in foreign currency per US dollar come from the USDA Economic Research Service. Summary statistics for all variables in the dataset are reported in table 1. The ikt panel includes 49 countries denoted i, 283 NAICS 2002 manufacturing industries k at 5 or 6 digits of aggregation, depending upon availability, for the years (t) 1997 through 2006.8 The US trade policy datasets in the Global Antidumping Database provide detailed informa- tion on the date a petition was filed, the identity of the country accused of dumping, tariff line information on the products involved, the outcome of the investigation and the magnitude of any final antidumping tariff imposed by the US against country i. All tariff-line level (HS10 or HS8 digit) trade policy data were concorded to 238 NAICS 2002 5 and 6 digit US industries to merge into the ikt, foreign country-industry-year panel. Industry-level foreign export supply elasticities facing the US at the 4 digit HS level were concorded to 2002 NAICS 5 and 6 digit industries. Because multiple 4-digit HS sectors map into each NAICS industry, we record the minimum, median, and maximum HS4 digit export supply elasticity that maps into each NAICS industry and check the robustness of our results to the range of products produced by each industry. Similarly, import demand elasticities at the 3 digit HS level were concorded to NAICS industries with the minimum, median and maximum elasticity in each industry recorded to check the robustness of any results. The WTO's Agreement on Antidumping specifies empirical criteria that must be satisfied in order for a country to impose an antidumping tariff. We included two measures of domestic industry health to capture the WTO's injury criteria - the growth of US industry employment and the ratio of inventories to shipments. Other industry characteristics are frequently used in the literature to control for political eco- nomic determinants of an industry's propensity to obtain import protection through antidumping tariffs. Because a free-rider problem must be overcome in filing an petition on behalf of the in- 8 Data on US manufacturing industries are available at the 5 digit level over the entire sample period. For some larger 5 digit industries, data is also available at the 6 digit level over the entire sample period. When the more disaggregated 6 digit industry data were available for all 6 digit industries within a 5 digit industry, we replaced the more aggregated 5 digit industry data with the less aggregated 6 digit industry data. 9 dustry, more concentrated industries are thought to have a higher propensity to seek antidumping protection. Thus, we include both the 4-firm concentration ratio (the shipments of the 4 largest shippers relative to total industry shipments) and the Herfindahl index of industry concentration. Further, we include a measure of industry size, total employment, because large industries may be better able to assume the large legal fixed cost of filing a petition. Similarly, we include total production employment as a measure of an industry's political importance. The vertical structure of an industry may matter; industries that are further downstream may file more petitions because they are more sensitive to industry price changes. We proxy for the vertical structure of an industry with the value-added to output ratio. Because the current values of industry specific variables and the choice of whether to petition for protection may be endogenous, we use lagged values of these variables in estimating the model. 3 Empirical Results: US Antidumping Tariff Formation The empirical results reported in tables 2-8 suggest that that US imposes antidumping tariffs in a manner consistent with the theoretical model of Bagwell and Staiger (1990). Table 2 reports that US antidumping tariffs are more likely to be imposed when past import growth has been strong, when import growth is less variable, and when import demand and export supply are relatively inelastic. Specifically, table 2 presents estimates of the binary model of the US government's decision to impose a final antidumping tariff against country i in industry k in year t. Estimates are presented as marginal effects in which a one-unit increase in a variable is associated with an incremental increase in the probability that the US will impose an antidumping tariff. Column (1) presents results for the basic specification of the model. A one-unit increase in the growth of bilateral imports from country i in industry k in the year before an antidumping petition is filed is associated with an increase in the probability of an antidumping tariff of 0.04%. This positive relationship between imports and antidumping tariffs is consistent with the terms-of-trade theory that hypothesizes that the pressure to raise a cooperative tariff is increasing in the size of 10 the import surge. Proceeding down column 1, we find that antidumping tariffs are less likely when import flows are highly volatile. A one-unit increase in the standard deviation of imports from country i in sector k is associated with a fall in the probability of an antidumping tariff of 0.17%. Turning to the indicator variables for the elasticity of import demand and export supply, a discrete change for an industry from a relatively elastic to inelastic import demand is associated with a 0.07% increase in the probability of an antidumping tariff, while a discrete change from a relatively elastic to inelastic export supply elasticity increases the probability of an antidumping tariff by 0.06%. These results are consistent with the Bagwell and Staiger (1990) theory. Column (2) presents estimates from a model which uses alternative measures of the inverse export supply and import demand elasticities. Because a wide range of products are produced by each NAICS industry, several elasticity estimates are available for each US industry. For 5 and 6 digit NAICS industries, some point estimates for the import demand elasticities associated with these industries ranged from slightly more than 1 to over 30. For export supply elasticities, the range was even broader with some 5 and 6 digit NAICS industries encompassing point estimates for export supply elasticities that ranged from close to zero to as high as 150, 185 or almost 300. In specification (2), we use a dummy variable derived from the median point estimate of the export supply and of the import demand elasticities to check the robustness of our main result in column 1. We find that the results regarding the export supply elasticity are highly sensitive to which measure is used. When we substitute an indicator for the median export supply elasticity, the sign on the marginal effect changes; the estimated marginal effect is negative suggesting that antidumping tariffs are more likely in industries facing relatively elastic export supply. Taking the results from columns (1) and (2) together, it is difficult to definitively conclude what role the export supply elasticity plays in the imposition of US antidumping tariffs. Antidumping cases frequently target only a subset of countries that export a particular product to the US (see Crowley, 2008) or a subset of products of an industry. Unfortunately, the estimated elasticities of a export supply from Broda, Lim~o, and Weinstein are averaged across both multiple products and all sources of a product to the US market. The lack of precision in a key variable of interest is likely to muddy some of the results. One possible explanation is that the US government imposes 11 antidumping tariffs against the individual countries with the least elastic export supply or the products (within an industry) with the least elastic export supplies, and this is not captured by the average export supply elasticity variable which does not have variation across countries or products. The signs of all other variables in column (2) are consistent with column (1). Specification (3) adds additional industry controls to the basic specification in column (1). Previous studies (Staiger and Wolak, 1994; Crowley, 2008) have found that domestic industry characteristics are important determinants of antidumping activity. We find that antidumping impositions are more likely in more concentrated industries and larger industries.9 Results for all other variables are robust to the inclusion of these additional controls. In column (4), we substitute the industry's level of production employment for the industry's level of total employment while keeping the industry concentration ratio. Because a high number of production workers may be associated with greater political power, one might expect this measure to be a more important determinant of trade protection than overall employment, but the two measures have similar positive effects on the likelihood of an antidumping tariff. Column (5) adds two additional industry characteristics to the specification in column (3) - the value-added to output ratio and the inventories to shipments ratio. Consistent with the findings of Crowley (2008) and Staiger and Wolak (1994), a higher value-added to output ratio is associated with a lower likelihood of receiving an antidumping tariff. A high level of inventories to output in the period before an antidumping tariff is filed is also associated with a higher probability of protection. This is consistent with the WTO legal criterion regarding injury to the domestic industry; rising inventory levels are often used by government agencies as a proxy for injury. Again, other coefficient estimates are robust with regard to these additional control variables. Finally, specification (6) uses an alternative measure of the import surge facing the US. The unexpected import growth variable captures a shock to the import growth rate. It is the import growth rate at time t - 1 less the average of the import growth rates at t - 2 and t - 3. The sign and magnitude of this variable are in line with the estimates in specifications (1)-(5). However, with the 9 These results are robust to the use of a Herfindahl index of industry concentration in lieu of the four firm concentration ratio. 12 smaller sample size necessitated by the use of the t - 3 import growth rate, the coefficient estimate is no longer statistically significant. Estimates for other variables are consistent with specification (1). Lastly, in specification (1), a real appreciation of the US dollar is associated with a higher antidumping tariff. Quantitatively, a one standard deviation increase in the variable increases the magnitude of the antidumping tariff by 1.14%. This estimate is in line with previous work by Knetter and Prusa (2003), Feinberg (2005), and Crowley (2008) all of whom find that the probability of an antidumping tariff is higher when the real dollar appreciates. Table 3 quantifies the key results of table 2 in terms of their economic significance. In table 3, we report the predicted probability that the US will impose an antidumping tariff in response to a one standard deviation increase in each of the continuous explanatory variables. For the dummy variables, we report that predicted probability of an antidumping tariff if the industry accused of dumping faces relatively inelastic import demand or export supply. We begin by observing that antidumping tariffs are rarely used. The last two rows of table 3 present statistics on the infrequent use of antidumping tariffs in the estimation sample for each specification. The last row reports the number of observations in the estimation sample while the second to last row reports the probability of an antidumping tariff being imposed in that sample. For all specifications in table 2, antidumping tariffs were imposed against only sixteen hundredths of one percent of country- industry-year observations. How would a change in the magnitude of an explanatory variable affect the probability of a tariff being imposed? Beginning with specification (1), a one standard deviation increase in the growth of lagged bilateral imports would increase the probability of an antidumping tariff by 25%, to 21 hundredths of one percent. Recall that Bagwell and Staiger predict that eruptions of trade protection are less likely in industries with more volatile imports. Consistent with the Bagwell and Staiger theory, a one standard deviation increase in the volatility of imports is associated with a very low probability of an antidumping tariff of only 5 hundredths of one percent. This is a substantial reduction of 69% in the likelihood of an antidumping tariff relative to its baseline level in the sample. Continuing with the baseline specification, a discrete change from a relatively 13 elastic to inelastic import demand increases the probability of a tariff by 44%. For the export supply elasticity dummy, a discrete change from relatively elastic to relatively inelastic increases the probability of a antidumping tariff by 37.5%. As discussed earlier, column (2) documents how the export supply elasticity results are sensitive to the exact measure used. When the elasticity dummies are constructed from a set of point estimates that include some very large estimates of the export supply elasticity, the results conflict with the terms-of-trade theory. Proceeding across columns (3) through (6), the results are quantitatively similar to those in column (1). Interestingly, the variables that are predicted by Bagwell and Staiger (1990) to be important determinants of cooperative tariff increases - import growth, import volatility, and the trade elasticities - all have roughly the same or greater economic significance as the variables tradi- tionally considered to be important determinants of antidumping activity - industry concentration, employment, and inventory levels. A one standard deviation increase in the 4-firm concentration ratio increases the probability of a tariff by about 6% in specifications (3) and (4), while a one standard deviation increase in the logged level of employment increases the probability of a tariff by 38% and a one standard deviation increase in the inventories to shipments ratio increases the likelihood of an antidumping tariff by 19%. Table 4 introduces a new explanatory variable to proxy for the unexpected import surge in the Bagwell and Staiger model. Some of the literature on the terms-of-trade theory of trade agreements emphasizes the importance of the magnitude of the negotiated market access implied by a given tariff concession rather than the tariff concession itself. In mapping the repeated static (i.e., no growth) environment of Bagwell and Staiger (1990) to an empirical environment characterized by underlying domestic economic and trade growth, we interpret negotiated market access as a commitment to a specified share of the importing country's market. From this, we define an import surge at t - 1 as an increase in country i's share of the US's market for k between t - 2 and t - 1. Moreover, we include an additional explanatory variable on market share to help us understand how competition among exporters in the US market for k affects US trade policy decisions. The 14 variable, the change in market share for the Rest of World (ROW) at time t - 1, is the change in the cumulative US market share for all countries that export goods in industry k to the US except for country i. We include this measure to help disentangle the effects of an aggregate import surge of k from a country-specific import surge of k from i. The first column reports our basic specification using the market share variables in lieu of import growth measures. The results are consistent with those using import growth as the measure of a bilateral import surge in table 2. A one-unit increase in a country's change in market share at time t - 1 increases the probability of an antidumping tariff by 4.5%. Interestingly, the marginal effect of a change in the cumulative market share of country i's competitors is negative. A one-unit increase in the change in the cumulative market share of all countries except country i is associated with a 1.2 percentage point reduction in the likelihood that country i will face on antidumping tariff. Although an increase in country i's market share increases the likelihood that country i will face an antidumping tariff, this result suggests that it is not unjustifiably grouped with competing exporters to the US. All else equal, if competing exporters increase their market shares, country i is less likely to face an antidumping tariff. Proceeding across columns (2) - (5) of table 4, the results line up with those of table 2. Table 5 quantifies the results from table 4. The magnitudes of the effects of the explanatory variables on the likelihood of an antidumping tariff are similar to those reported in table 2. However, the quantitative effect of a one standard deviation increase in the change in a country's market share is smaller than that of import growth, increasing the likelihood of an antidumping tariff by 6.3%. An interesting result is that a one standard deviation increase in the change in the cumulative market share of competing exporters reduces the probability of an antidumping tariff by 20%. Table 6 focuses on two industries that are frequent recipients of antidumping protection in the US - steel and chemicals. While only 1.3% of the observations in our dataset are for the steel industry, 26.7% of the antidumping tariffs recorded in the dataset are in the steel sector. Similarly, while only 2.3% of the observations in the dataset are of chemical sectors, 11.8% of the antidumping policies in the dataset are against chemical exporters. These facts raise the question of whether the results in tables 2 - 5 driven by antidumping activity in these sectors, and whether 15 US antidumping use treats these sectors differently than other sectors based on the underlying economic fundamentals? To address these questions, the specifications in table 6 utilize industry dummies and interaction terms. Beginning with the first column, the basic specification (1) from table 2 is augmented to include a dummy for steel industries. The positive coefficient on the steel dummy indicates a strong bias in favor of antidumping tariffs against steel exporters that goes beyond the basic economic variables of the Bagwell and Staiger (1990) model; a discrete change from a non-steel to steel-industry increases the probability of an antidumping tariff by 2.4%. However, even after controlling for this bias, the results on the other coefficients are largely unchanged suggesting that the basic results in table 2 are not driven by observations on the steel industry. The second column, which includes a dummy for chemical sectors, shows a smaller bias against chemical exporters. A discrete change from non-chemical to chemical products is associated with a 0.43% increase in the likelihood of an antidumping tariff after controlling for the economic variables of the Bagwell and Staiger (1990) model. Columns (3) and (4) of table 6 include terms that interact the volatility of imports with steel and chemical dummies, respectively. Here the predictions of the Bagwell and Staiger model are reversed in the steel industry - volatile import growth increases the probability of a steel tariff. In contrast, for the chemical sectors, more volatile import growth is associated with a reduction in the probability of an antidumping tariff, as predicted by Bagwell and Staiger. However, the magnitude of this effect in chemical sectors is much smaller than that reported in our basic specification of the average across all industries reported in table 2. Finally, the last two columns of table 6 add domestic industry characteristics - industry concentration, employment, the value-added to output ratio, and the inventories to shipments ratio - to specifications (1) and (2). After accounting for these general industry characteristics, the extent of the "bias" against exporters of steel and chemicals to the US is smaller, but it still economically and statistically significant. Table 7 presents an analysis of the special status of China in US antidumping investigations. While China accounts for only 2.5% of the observations in our dataset, it is the target of 42% of US antidumping tariffs in our sample. As with the steel and chemical industries, we examine whether 16 US imports from China are a special case by augmenting our basic specification from Table 2 with China dummies and interaction terms. Beginning with column (1) of table 7, the inclusion of the China dummy in the basic specification shows a much higher likelihood of antidumping in US imports from China relative to imports from all other foreign sources; the likelihood of an antidumping tariff increases by 1.79 percentage points for products imported from China. Interestingly, after including the dummy for China, the effect of import growth on the probability of an antidumping tariff is smaller and no longer statistically significant. Column (2) attempts to identify the source of the higher likelihood of US antidumping against China by examining the volatility of imports from China through an interaction of the standard deviation of import growth with a China dummy. The results are similar to those for the steel industry. While more variable US imports are associated with a lower probability of protection in general, the pattern is reversed for imports from China; i.e., an antidumping tariff against imports from China is more likely when imports from China are highly volatile. This result for China contradicts the predictions of the Bagwell and Staiger model and suggests one possible economic reason for why antidumping investigations against China often end in new tariffs. Column (3) focuses on the import demand and export supply elasticities for goods from China. Here the predictions of the Bagwell and Staiger model are quantitatively much stronger for imports from China than for imports from other countries. A discrete change from being a non-China exporter in a sector with inelastic import demand to being a China exporter in the same sector is associated with a 0.6 percentage point increase in the probability of an antidumping tariff. This is an increase of 62.5% over the probability of an antidumping tariff in the estimation sample. Similarly, a discrete change from a non-Chinese to a Chinese exporter in a sector with relatively inelastic export supply increases the probability of an antidumping tariff by 0.33 percentage points. Column (4) augments specification (1) by including domestic industry characteristics. The increased probability of US antidumping decisions against China is smaller after taking into account features of the domestic US industry, but it is still quantitatively significant. After controlling for industry characteristics, if a case involves imports from China, the likelihood of an antidumping 17 tariff increases 1.61 percentage points. Column (5) replicates the specification of column (2) but adds details on the domestic US industry. A similar pattern is observed; the China-specific effect on the standard deviation of imports is muted when industry details are taken into account, but the China effect is still substantial. Finally, column (6) adds industry variables to the analysis of China's relationship to the import demand and export supply elasticities. The results are essentially the same as those in column (3), but the estimated magnitudes of the effects are smaller. The final table of results, table 8, examines changes in market shares for exporters in the steel and chemical industries and for Chinese exporters. The first three columns of table 8 introduce dummy variables for steel, chemicals and China to the basic market share specification found in column (1) of table 4. In all three specifications, the bias toward imposing antidumping tariffs for these types of cases is present when we use changes in market share rather than import growth as the measure of an import surge. More interestingly, columns (4), (5) and (6) interact the relevant industry or country dummy with the two market share variables. In column (3), the marginal effect of being a steel exporter and having a change in market share is not statistically significant. A steel exporter with an increased market share will be more likely to face a US antidumping tariff, but the size of this effect is the same as in all other industries. However, steel exporters are treated differently when it comes to competition. For a steel exporter, if competing exporters in other countries experience an increase in their US market share, all else equal, the steel exporter will be more likely to face an antidumping tariff. Thus, it appears that steel is an area in which antidumping tariffs penalize some of the exporters who are not responsible for a US import surge. Column (4) reveals a difference between the policy treatment of steel and chemical products. If a chemical exporter experiences an increase in market share, the increased likelihood of an antidumping tariff is much larger than it is for other industries. However, unlike in steel, exporters in chemical sectors are not grouped with their competitors if their competitors experience an increase in US market share. Finally, column (6) reports results for the interaction of China cases with market share variables. First, when Chinese exporters increase market share, the likelihood of an antidumping tariff is greater than for non-Chinese exports. Second, China experiences a small negative policy externality when 18 its competitors increase their share of the US market. An increase in the cumulative market share of non-China countries increases the likelihood that China will face an antidumping tariff. To summarize, the basic results on market share reported in table (4) are robust to the inclusion of a steel, a chemical or a China dummy, but there are interesting variations in the effects of market share and of competitor's market share for each. To conclude this section, a large literature has explored the determinants of the antidumping tariff policy. Our paper is the first to develop an empirical model of antidumping tariffs as a cooperative tariff increase in a self-enforcing trade agreement subject to unexpected import surges. We find evidence consistent with the Bagwell and Staiger (1990) theory through our results that US antidumping tariffs are more likely the larger is lagged import growth, the greater the increase in the exporter's share of the US market, the lower the volatility of imports, and the less elastic are US import demand and foreign export supply. 4 The Trading Partner Response to US Antidumping: Are the Tariffs Viewed as Cooperative? Our results on the United States' use of antidumping indicate substantial evidence consistent with the Bagwell and Staiger (1990) theory of self-enforcing trade agreements. The main innovation of our research is to highlight the extent to which US use of antidumping trade policy is associated with the changing value of the economic incentive to respond to trade volume shocks whose importance may vary because of heterogeneity across import demand and export supply elasticities (Broda, a Lim~o and Weinstein, 2008; Bagwell and Staiger, 2009). This particular economic rationale for changing a US tariff is distinct from the political economy considerations that have been the focus of much of the previous literature on antidumping as a time-varying trade policy. Nevertheless, these results introduce a new and related question: if US antidumping policy use is consistent with the norms of a cooperative and self-enforcing trade agreement, then how can we explain the variety of foreign policy responses to US tariff increases? Indeed, there are at least two observable (or measurable) ways by which trading partners have actively responded to a 19 US antidumping tariff increase with their own policy change. First, nine different WTO members (counting EU states as one WTO member) used the WTO's dispute settlement procedures to formally challenge 27 out of the roughly 130 different US antidumping tariffs imposed against them between 1997 and 2006.10 Second, six WTO members responded to an additional 23 US antidumping tariffs imposed during this period by initiating their own antidumping activity against US exporters within the same 4-digit industry; and one interpretation is that such actions were simply direct antidumping retaliation.11 Combined, this implies that almost 40% of the US tariff increases under antidumping between 1997-2006 were challenged by its trading partners either formally at the WTO or informally through unilateral retaliatory action. On its face, such foreign policy responses seem at odds with the results of the first section of the paper that US antidumping use is consistent with the theory of cooperative trade agreements. Our next goal is to attempt to explain the variation in the foreign response to the US raising its tariffs under its antidumping policy. The first step is to build from a related literature (e.g., Bown, 2005a; Busch, Reinhardt and Shaffer, 2008) and examine determinants of whether the foreign exporter responds to the US tariff increase by initiating a formal WTO dispute settlement challenge that seeks its removal.12 Specif- ically, we begin by estimating determinants of a binary choice model of whether an exporter hit with a new US antidumping tariff between 1997-2006 decides to challenge it by initiating a formal WTO dispute. If the WTO is a cooperative trade agreement in the sense of Bagwell and Staiger (1990), the theory suggests that formal disputes are not likely to be filed against US trade policy changes that are imposed because of underlying economic shocks to trade volumes in sectors with 10 These WTO members were Brazil, Canada, Ecuador, European Union (on behalf of separate US actions against exporters from Belgium, Finland, France, Germany, Italy, Netherlands, Spain, Sweden, and the UK), India, Japan, Korea, Mexico and Thailand. 11 The WTO members were Canada, China, European Union (on behalf of exporters from Italy and Spain), India, Mexico and Poland. 12 See also Horn, Mavroidis and Nordstrom (2005), Bown (2005b), and the literature on determinants of the use of WTO dispute settlement surveyed in Bown (2009b, chapter 4). By foreign exporter, of course, we refer to the combined ability of the exporting industry and its government to coordinate the mounting a formal WTO dispute legal challenge since only governments have standing in WTO disputes. 20 low import demand and export supply elasticities that create an incentive to raise the cooperative tariff. On the other hand, US tariff increases that appear to be outliers with respect to the esti- mated rule governing the cooperative trade agreement may be good targets for the foreign trading partner to mount a formal WTO challenge. In addition to a number of additional determinants that the previous literature has indicated are likely to affect the decision of a trade partner to formally challenge the US use of antidumping, our main innovation is to test specifically whether the predicted likelihood of US antidumping (generated from the first empirical model of the pa- per) is itself a significant determinant of the trading partner's policy response to the new US tariff increase. As a second step, in section 4.3 we consider a multinomial extension to this first model in which we allow the foreign trading partner an additional policy instrument with which to respond to the US tariff increase. In this extension, we consider the possibility that some trading partners may bypass WTO dispute settlement in favor of more directly using their own antidumping policy to retaliate against US exporters in the same sector in which their exporters were facing a tariff increase in the US. In the multinomial model, we give the foreign trading partner three choices of how to respond to the increased US tariff: challenge it with a WTO dispute, directly retaliate by using its own antidumping against US exporters in the same sector, or do nothing. Before proceeding to the formal empirical approach and results, the next section introduces the data and variable construction used to estimate the models. 4.1 Data used to estimate determinants of the foreign trade policy response The imposed US antidumping policies result in tariff increases that undoubtedly hurt the foreign exporter, at least in the short run. What determines whether the foreign exporter will challenge the US tariff and seek its removal? One potential determinant has already been proposed by the Bagwell and Staiger (1990) theory; i.e., if the US tariff on h is not a response to an economic shock that changed the level of the cooperative tariff in the self-enforcing trade agreement. We capture this possibility through use of the predicted likelihood of the US antidumping tariff on h generated from model specification 1 of table 2 for each of the antidumping tariffs imposed in our data set 21 during the 1997-2006 period. To guide our selection of additional potential determinants of the foreign response to the new US tariff, we draw from the Maggi and Staiger (2008) theoretical model of WTO dispute settlement. Consider again the perspective of a foreign exporter of product h adversely affected by a new US antidumping tariff, and the expected costs and benefits that are likely to influence its decision of whether to formally challenge the new US tariff with a formal WTO dispute. When the costs of a dispute fall disproportionately on the exporter relative to the importer, one prediction that comes out of the Maggi and Staiger model is that the importer will be more likely to engage opportunistically in the first place and thus impose WTO-inconsistent tariffs that will lead to a dispute. Since 20% of the US antidumping tariffs imposed during this period resulted in WTO challenges, it is likely that exporter litigation costs are therefore sufficiently high in this sample of data. In our cross-sectional empirical analysis with a common (US) importer, it is reasonable to assume that the litigation cost to the US government defending a challenge is similar across all AD tariffs it imposed. An implication of the Maggi and Staiger theory is that equilibrium WTO disputes will arise in instances in which the benefits to the exporter from successful resolution to a dispute - put differently, the opportunity cost of failing to challenge it at the WTO - are also high. In our empirical model, we use two different variables to capture the opportunity costs to the exporter for failing to initiate a dispute that would remove the US antidumping tariff. Each variable is derived from data on the underlying foreign exports of the product h that is the target of the US antidumping tariff; the variables are constructed to reflect the size of those exports to different markets. The first variable is the value of access to the US import market in product h, which is to proxy for the amount of trade that foreign is likely to lose because of the new US tariff.13 The larger the amount of pre-antidumping exports to the US market that foreign has at risk, the higher the opportunity cost of not challenging the US tariff change, and the more likely the exporter will overcome the fixed cost of filing a WTO dispute. 13 Specifically we measure this as the log of foreign's AD-affected product exports of h to US in t - 1, the year before the antidumping investigation (that ultimately resulted in the tariff increase) is initiated. 22 The opportunity cost to the exporter for failing to challenge the US tariff may also involve a second variable that we refer to as the "terms of trade spillover" to third markets in the rest of the world (ROW). I.e., exporters that ship to multiple markets are likely concerned that the US tariff will negatively impact their profits associated with sales of product h to ROW, which could occur through at least two channels associated with a negative terms-of-trade externality. First, for an unchanged level of import demand in ROW for product h, a shift in foreign's export supply toward this market (due to sales shut out of the US market caused by the US AD) will reduce the exporter's terms-of-trade in those markets as well, through a "trade deflection" result (Bown and Crowley, 2006, 2007; Chang and Winters, 2002). A second and reinforcing channel may be through ROW governments engaging in policy learning. Because WTO rules mandate that the US must notify a new antidumping tariff to the world (via the WTO), this new US tariff in product h reduces the fixed cost to ROW governments collecting information necessary to ultimately impose their own new antidumping import restriction on product h from foreign. This would also create a negative terms-of-trade externality by shifting in the ROW import demand curve.14 The reinforcing nature of these two effects indicates that the larger is the share of foreign's total exports of h that it sends to the ROW, the more likely it will challenge the US tariff increase of h to seek its removal. In WTO dispute settlement, the expected benefit to a successful outcome to the dispute is also determined by the likelihood that the new US tariff is removed. While the WTO is a rules-based organization, a necessary condition for one trading partner to enforce another's WTO commitments - e.g., in this case remove or revise a US imposed antidumping tariff that the WTO may determine was imposed in a way inconsistent with the cooperative trade agreement's rules - is that foreign has the capacity to retaliate by threatening new import restrictions against US exporting industries. One important theory motivating this empirical regularity stems from the underlying interpretation of the WTO's guiding principle of reciprocity provided by Bagwell and Staiger (1999). Prior evidence from other samples of data stemming from different trade policy decisions (e.g., Bown 2004a,b) related to WTO dispute settlement finds evidence that foreign retaliation capacity matters. Here we measure foreign's retaliation capacity as simply the log of the value of total US exports 14 Maggi (1999) explores alternative implications of the WTO performing an information dissemination role as a multilateral institution. 23 sent to the foreign country. We expect a positive relationship that larger US exports to foreign give foreign the necessary capacity to make retaliatory threats to induce the US to comply with the WTO legal process, increasing the likelihood that foreign will enjoy the benefit of removal of the US tariff (and restoration of US market access). In section 4.3 we extend the basic model to also given foreign the policy choice to respond with a more direct retaliation against US exports through its own antidumping action in the same sector. In that model we include additional determinants to assess foreign's capacity to retaliate directly against the US exporter in the same sector k in which the US imposed the antidumping tariff. We capture foreign's AD retaliation capacity as the log of US exports of k sent to foreign.15 We expect the higher US exports of k, the more likely is foreign to respond to the US imposition of antidumping in k with its own antidumping against US exports of k. In this second model, we also redefine the retaliation capacity variable intended to influence foreign's use of WTO dispute settlement. There we focus on US exports of all other goods (= k), and we specifically define this variable as the share of US total exports of these all other goods that are destined for the foreign market. To conclude this section, it is important to highlight the important difference in the data used for this portion of the estimation. Since many of the variables in this section will be defined at the level of the 6-digit Harmonized System (HS) product h at issue in the antidumping investigation, the data is much more disaggregated than what we used for the estimation of the antidumping tariff formation equation, which required access to industry-level data at 5- or 6-digit NAICS industry level k.16 The product h 6-digit HS data for foreign exports is taken from Comtrade. The data on US antidumping, WTO challenges, and foreign retaliatory antidumping is again taken from the Global Antidumping Database (Bown, 2009a). 15 One item worth noting is that in this section of the paper the industry k is defined at the 4-digit HS level, which is different from the first section of the paper in which k was a 5- or 6-digit NAICS industry. Since we do not rely on industry (production) data in this part of the estimation exercise, we do not bother to concord the HS 6-digit to the NAICS data, which does have the benefit of reducing some measurement error. 16 US antidumping investigations are typically carried out at the 8- or 10-digit level, the 6-digit HS level is the most disaggregated level of trade that is consistently defined across countries, as is required for the variable construction that we undertake below. 24 Table 9 presents summary statistics of the variables used in the estimation. 4.2 Estimating an empirical model of WTO dispute settlement Table 10 presents estimates of the marginal effects of the binomial probit model in which the dependent variable takes on a value of one if the foreign exporter issues a formal legal challenge at the WTO to the US antidumping tariff imposed against imports from foreign of product h. Column (1) of table 10 presents our simplest specification that focuses solely on the predicted likelihood of a US antidumping tariff, where this determinant is generated from the model presented in the last section. There is evidence of a negative bivariate relationship between these variables, consistent with the interpretation of the Bagwell and Staiger (1990) framework that US antidumping tariffs associated with a higher predicted likelihood of a tariff being imposed are less likely to be challenged through a WTO dispute. By itself this effect is also sizable - a one standard deviation increase from the mean value of the predicted likelihood of an antidumping tariff decreases the probability of foreign responding with a WTO dispute by roughly 50%, from 0.186 to 0.108. While the magnitude of the estimated marginal effect coefficient shown in the table varies slightly in the additional specifications that we consider and discuss below, the estimated size of this impact on the probability of a challenge does not. Column (2) of table 10 is our preferred specification which includes additional variables that control for other aspects of the expected costs and benefits likely to determine the foreign decision of whether to challenge the US tariff with a WTO dispute.17 The estimates of these other determinants all have the theoretically-predicted sign, and only the variable capturing the potential impact of the US tariff on foreign exports to the ROW (via the "terms of trade spillover") is not statistically significant at conventional levels. Importantly, the size and statistical significance of the impact of the predicted likelihood of the US antidumping tariff is virtually unchanged from specification (1). Consider next the estimated marginal effects estimates of the other determinants of the foreign WTO dispute decision reported in specification (2). First, the size of US market access threat- 17 The results of this section are consistent with earlier estimates of the impacts of these determinants on other contexts. See again Bown (2005a) and the survey presented in Bown (2009b, chapter 4). 25 ened by the new US antidumping tariff clearly matters. Foreign is more likely to challenge US antidumping tariffs imposed on larger (t - 1) US imports of h. Market access is also a very impor- tant economic determinant in the estimation, as the estimated marginal effect (0.068) implies that a one standard deviation increase in the amount of US imports at stake more than doubles (from 0.136 to 0.288) the predicted probability of foreign responding with a WTO dispute. Second, while it is not statistically significant in specification (2), the larger is the foreign exporter's affected product h exports to ROW as a share of its total exports of h, the more likely is foreign to challenge the US antidumping tariff on h at the WTO. The estimated impact is not as important as the (first order) impact of the direct US market access variable, but it is still sizable, as the estimated marginal effect (0.170) implies that a one standard deviation increase in the share of foreign exports of h it sends to ROW leads to a 28.0% increase (from 0.136 to 0.180) in the predicted probability of foreign responding to the US tariff with a WTO dispute. Third, the capacity of foreign to make credible retaliatory threats against US exports is also estimated to positively affect the decision to challenge the US tariff with a WTO dispute. And the estimated impact (0.057) is also sizable, as a one standard deviation increase in the amount of US exports sent to foreign leads to a 56.8% increase (from 0.136 to 0.240) in the predicted probability of foreign responding with a WTO dispute. Columns (3) and (4) of table 10 present additional robustness checks to the estimates. Spec- ification (3) is identical to specification (2) except it drops the predicted likelihood of the US antidumping tariff. The estimated marginal effect estimates on the other determinants are rel- atively unaffected. Finally, specification 4 includes two additional control variables. The first is simply the size of the US antidumping tariff, and this is included to address a potential concern that WTO challenges may be more likely when new US antidumping tariffs are low.18 The estimated 18 Including the size of the US tariff controls for two different phenomenon. First, consider the many WTO challenges to the US use of "zeroing" in antidumping investigations. Foreign may be more likely to challenge low tariffs when there is zeroing, under the expectation that the WTO dispute may result in the US revising the dumping calculation without using the zeroing methodology and this calculation could lead to a new antidumping tariff that is below the de minimus threshold resulting in a complete removal of the antidumping tariff and not simply a downward revision. For a discussion of zeroing, see Crowley and Howse (2010). Second, the US imposed much higher antidumping tariffs 26 effect is negative, and it is statistically and economically significant as lower US antidumping tariffs are more likely to be challenged by WTO dispute settlement. Finally, specification (4) also includes an indicator for whether the foreign target of US antidumping is in a preferential trade agreement (PTA) with the US. While the estimate is negative, it is not statistically different from zero. Most importantly, inclusion of these additional determinants does not affect the estimated effects of the main variables of interest. Even the reduced size of the estimated coefficients of the marginal effects is of little consequence, as the estimated impact on the predicted probability of a foreign WTO dispute is virtually unchanged from the levels of economic significance discussed above. 4.3 Estimating an empirical model of WTO disputes, AD industry retaliation, or no response As a final exercise, in this section we present the results of an expanded empirical model in which we allow foreign one additional policy response to a newly imposed US antidumping tariff. Here we expand the foreign exporter's choice set so that it also includes responding with a retaliatory antidumping policy against US exports in the same 4-digit industry k. The motivation for this exercise stems from the results and conjecture of Bown (2005a), that a contributor to why some trading partners do not use the WTO to challenge US antidumping is because they have an alter- native and relatively direct channel to obtain redress - i.e., their own use of antidumping against the US in the same sector.19 Table 11 presents our empirical results of the marginal effects estimates of the three choice multinomial probit model. We present three different model specifications; the left column of each on imports from non-market economies such as China during this time period (Bown, 2010), and after its WTO accession in 2001 China did not initiate a single WTO dispute over US antidumping through 2006. Thus this variable may also partially capture a China-specific effect. However, as we explore in the next section, China did actively use its own antidumping policy against US exporters during this time period. 19 Since Bown (2005a) did not have access to data on foreign use of antidumping by industry against US firms, that paper could not explicitly test this hypothesis. The indirect evidence in that paper was the US industries with larger exports to the foreign country were less likely to have their AD tariffs targeted for a WTO dispute. The resulting conjecture, that we confirm with evidence below, was that the foreign country was engaging in "vigilante justice" by simply using reciprocal antidumping instead of WTO dispute settlement where it had the capacity to do so. 27 specification presents the estimates of the determinants of the exporter's choice to challenge the imposition of a new US antidumping import restriction through a formal WTO dispute. The middle column presents estimates of the determinants of the choice to respond via subsequently imposing a same-industry antidumping action of its own within the next two years after the US imposed AD in year t. Finally, the right column presents estimates of the determinants of the choice not to respond with either a WTO dispute challenge or an antidumping retaliation. The determinants are the same variables used in the binomial probit estimates (see table 10) of the last section, with the one exception described earlier. In specifications (2) and (3) of the multinomial probit model we disentangle two components of the foreign retaliation capacity. First, we allow for the exporter's capacity to retaliate with a same-industry antidumping action to affect its choice, and we capture this industry-level retaliation capacity with the log of US exports of 4-digit industry k sent to foreign in t - 1. The second retaliation capacity variable is US exports of all other goods (= k) sent to foreign, which we define as the share of US exports of all other goods sent to all markets.20 Specification (1) of table 11 estimates the foreign choice solely as a function of the predicted likelihood of the US antidumping tariff from the model of the first section of the paper. The predicted probabilities of foreign's three choices based on the estimates in table 11 are 0.181 for a WTO dispute, 0.182 for antidumping retaliation, and 0.637 for no policy response.21 Just as in the binomial probit model in the last section, there is evidence of a negative relationship between the predicted likelihood of this US antidumping tariff and the foreign decision to file a WTO challenge. Furthermore, the effect is sizable; ceteris paribus, a one standard deviation increase in the predicted likelihood of AD formation above the sample mean decreases the predicted probability of a WTO dispute by almost half from 0.181 to 0.098. Interestingly, specification (1) of the three choice model also suggests that the failure of foreign not to respond by initiating WTO disputes against instances of high predicted probabilities of US 20 These two variables are highly positively correlated in the levels, which is one reason why we define the sec- ond variable as a share. Nevertheless, the high correlation also makes the estimation results relatively sensitive to specification choice, as columns 2 and 3 of table 11 illustrate. 21 This compares to the sample of raw data of outcomes which were 0.195 for a WTO dispute, 0.186 for antidumping retaliation, and 0.619 for no policy response. 28 antidumping tariffs should not be mistaken for a decision to "do nothing." Indeed, the estimated impact of the predicted probability that this determinant has on foreign engaging in direct an- tidumping retaliation in the same sector is positive (48.953) and statistically significant. The same one standard deviation increase in the predicted likelihood of a US antidumping tariff that led to a decrease in the predicted probability of a WTO dispute (from 0.181 to 0.098) leads to a substantial increase in the predicted probability of foreign antidumping retaliation (from 0.182 to 0.255). Specification (2) is our preferred specification of the multinomial probit model as it includes additional control variables. When we include variables controlling for the importance of access to the US market, the possibility of the US antidumping tariff spilling over to impact foreign's terms of trade in ROW, and foreign's retaliatory capacity, the estimates' signs on the explanatory variables are generally consistent with the theory. First, the size and estimated impact of the negative relationship between the predicted likelihood of the US antidumping tariff and the WTO dispute choice continues to hold - the higher the likelihood that the first-stage model predicted for a US antidumping tariff, the less likely that foreign responds with a WTO dispute. However, while the estimated impact of this determinant on the antidumping retaliation choice in specification 2 is still positive, unlike specification 1, it is no longer statistically significant at conventional levels. Next consider again the determinants associated with the Maggi and Staiger (2008) theory of the expected costs and benefits to foreign pursuing WTO dispute settlement. First, the US market access variable is estimated to have a positive and significant impact on the decision to seek removal of the US antidumping tariff through WTO dispute settlement. Furthermore, unlike the binomial probit model, the impact on the WTO dispute choice of the "terms-of-trade spillover" variable in specification (2) of table 11 is also statistically significant and positive, as predicted by the theory. The retaliation capacity variables of specification 2 also have the theoretically predicted sign, and they are statistically significant. Increased US exports of k to foreign increase the likelihood that foreign will respond to the US antidumping tariff with a retaliatory antidumping action in k of its own (0.053). Furthermore, the higher the share of total US exports of all other goods (= k) sent to foreign, the greater is foreign's capacity to retaliate under the WTO in some other sector, and 29 the higher the likelihood that foreign responds to the US antidumping tariff with a WTO dispute challenge. In terms of the economic significance of the effects, a one standard deviation increase in foreign's antidumping retaliation capacity increases the probability of an AD retaliation from 0.186 to 0.335 with virtually no statistically significant impact on the probability of an WTO dispute. On the other hand, a one standard deviation increase in WTO retaliation capacity increases the probability of a WTO dispute from 0.122 to 0.211. Combined these results suggest that retaliation capacity threats work in predictable ways - large (same) industry k US exports to the foreign market and relatively small overall US exports to foreign increases the probability of foreign AD as a response instead of a WTO dispute. Furthermore, this evidence is consistent with the conjecture of Bown (2005a) that one important determinant for why foreign countries do not use WTO dispute settlement to challenge AD is because they have access to a separate policy to address the issue - reciprocal antidumping in the same sector. Finally, specification (3) of table 11 again includes the additional control variables (the size of the US antidumping tariff and a PTA indicator) as a final robustness check. While the statistical significance of some of the estimates changes, for the most part the qualitative pattern to the results is largely unaltered. 5 Conclusion This paper empirically examines how governments make trade policy adjustments under a self- enforcing trade agreement in the presence of economic shocks. Using data on US antidumping (AD) policy formation between 1997-2006, we find that US antidumping tariffs are consistent with an increase in "cooperative" tariffs of a Bagwell and Staiger (1990) model of self-enforcing trade agreements. These results provide empirical support for models of trade agreements that emphasize the importance of the terms-of-trade motive in tariff setting. We find that both the likelihood of a US AD as well as the size of the imposed AD tariff are negatively related to the foreign export supply and the US import demand elasticities. The results are consistent with the optimal tariffs a theory as well as other empirical research (Broda, Lim~o, and Weinstein 2008; Bagwell and Staiger, 30 2009) in the literature from other settings that the terms of trade affects trade policy formation. Furthermore, in light of this interpretation of US AD use as a cooperative increase in trade policy, we empirically investigate determinants of which of these US AD actions are challenged by trading partners both inside (through dispute settlement) and outside (through antidumping retaliation) the WTO. After controlling for factors such as a trading partner's expected cost and benefit to challenging US antiumping, we find that trading partners are less likely to challenge such "cooperative" US antidumping tariffs that were imposed under terms-of-trade pressure suggested by the theory. 31 Table 1: Summary Statistics: US Antidumping Tariff Formation Standard Mean Variable Deviation Dependent Variables AD imposed by the US against country i industry k in year t(=1) or not 0.0016 0.0395 AD duty against country i industry k in year t 0.7535 0.8791 Explanatory Variables Measures of bilateral import surge Growth of imports_ikt-1 0.1021 0.9482 Change in Market Share_ikt-1 0.0002 0.0033 Change in Market Share_Rest of World, kt-1 0.0099 0.0217 Std. deviation of import growth_ik 0.7246 0.6610 Trade elasticities High Inverse Import Demand Elasticity = 1 0.7273 0.4453 High Inverse Export Supply Elasticity = 1 0.7516 0.4321 Ln (Real Exchange Rate)_it 1 1.9332 2.5510 Domestic Industry Variables Ln(Four firm concentration ratio)_k 3.4673 0.6080 Ln(US employment_kt-1) 10.3918 0.9730 Ln(US production employment_kt-1) 10.0156 0.9888 Value-added/Shipments_kt-1 0.5127 0.1179 Inventories/Shipments_kt-1 0.1290 0.0630 Notes: 82320 observations of country i exporting goods in industry k to the US between 1997 and 2006. 32 Table 2: US Antidumping Tariff Imposition: Marginal Effects from a Binary Model: Import Growth Substitute Add industry Substitute Add more Substitute Basic alternative size and political industry "unexpected" specification elasticity concentration strength characteristics import growth measures to (1) for size (1) (2) (3) (4) (5) (6) Measures of bilateral import surge Growth of imports_ikt 1 0.0004** 0.0004** 0.0003** 0.0003** 0.0003** (0.0002) (0.0002) (0.0001) (0.0001) (0.0001) "Unexpected" Import Growth_ ikt 1 0.0002 (0.0002) Std. deviation of import growth_ik 0.0017*** 0.0017*** 0.0014*** 0.0013*** 0.0012*** 0.0019*** (0.0002) (0.0002) (0.0002) (0.0002) (0.0002) (0.0003) Trade elasticity dummies High Inverse Import Demand Elasticity 0.0007*** 0.0006*** 0.0006*** 0.0005*** 0.0006*** (0.0002) (0.0002) (0.0002) (0.0002) (0.0002) High Inverse Export Supply Elasticity 0.0006*** 0.0004* 0.0004** 0.0004** 0.0004 (0.0002) (0.0002) (0.0002) (0.0002) (0.0002) Alternative High Inv. Imp Dem Elast 0.0011*** (0.0002) Alternative High Inv. Exp. Sup. Elast 0.0009*** (0.0002) Ln (Real Exchange Rate)_it 1 0.0001*** 0.0001*** 0.0001*** 0.0001*** 0.0001*** 0.0001*** (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) Domestic Industry Variables Ln(Four firm concentration ratio)_k 0.0002** 0.0003** 0.0001 (0.0001) (0.0001) (0.0001) Ln(US employment_kt 1) 0.0006*** 0.0005*** (0.0001) (0.0001) Ln(US production employment_kt 1) 0.0006*** (0.0001) Value added/Shipments_kt 1 0.0034*** (0.0007) Inventories/Shipments_kt 1 0.0042*** (0.0008) Number of Observations 82320 82320 81990 81902 81922 58964 Log likelihood 951.361 936.79 927.896 924.089 910.436 654.388 Notes: Huber-White robust std errors in parentheses with ***,**, and * indicating statistical significance at the 1%, 5% and 10% levels. 33 Table 3: Predicted Probability of a US AD Tariff for a One S.D. Increase in... Substitute Add industry Substitute Substitute Add more Basic alternative size and political "unexpected" industry specification elasticity concentration strength import characteristics measures to (1) for size growth (1) (2) (3) (4) (5) (6) Measure of bilateral import surge Growth of imports_ikt 1 0.0020 0.0020 0.0019 0.0019 0.0019 "Unexpected" growth of imports_ikt 1 0.0000 Std. Dev. of import growth_ik 0.0005 0.0005 0.0007 0.0007 0.0008 0.0004 Trade elasticity Indicators discrete change Inelastic import demand 0.0023 0.0022 0.0022 0.0021 0.0022 Inelastic export supply 0.0022 0.0020 0.0020 0.0020 0.0020 Alternative inelastic import demand 0.0027 Alternative inelastic export supply 0.0007 Domestic Industry Characteristics Ln(4 firm concentration ratio_k) 0.0017 0.0018 0.0000 Ln(Total employment_kt 1) 0.0022 0.0021 Ln(Production employment_kt 1) 0.0022 Value added/Shipments_kt 1 0.0011 Inventories/Shipments_kt 1 0.0019 Prob of a US antidumping duty in estimation sample 0.0016 0.0016 0.0016 0.0016 0.0016 0.0016 Number of Observations in estimation sample 82320 82320 81990 81902 81922 58964 34 Table 4: US Antidumping Tariff Imposition: Marginal Effects from a Binary Model: Change in Market Share Substitute Add industry Substitute Add more Basic alternative size and political industry specification elasticity concentration strength for characteristics measures to (1) size (1) (2) (3) (4) (5) Measures of bilateral import surge Change in MS_ikt 1 0.0448*** 0.0436*** 0.0439*** 0.0441*** 0.0356*** (0.0106) (0.0104) (0.0099) (0.0099) (0.0093) Change in MS_ROWkt 1 0.0124*** 0.0100*** 0.0106*** 0.0104*** 0.0096*** (0.0034) (0.0033) (0.0035) (0.0035) (0.0030) Std. Dev. of import growth_ik 0.0013*** 0.0013*** 0.0010*** 0.0009*** 0.0008*** (0.0002) (0.0002) (0.0002) (0.0002) (0.0002) Trade elasticities High Inverse Import Demand Elasticity 0.0006*** 0.0005*** 0.0005*** 0.0005*** (0.0002) (0.0002) (0.0002) (0.0001) High Inverse Export Supply Elasticity 0.0005*** 0.0004** 0.0003** 0.0003** (0.0002) (0.0002) (0.0002) (0.0001) Alternative High Inv. Imp Dem Elast 0.0009*** (0.0002) Alternative High Inv. Exp. Sup. Elast 0.0009*** (0.0002) Ln (Real Exchange Rate)_it 1 0.0001*** 0.0001*** 0.0001*** 0.0001*** 0.0001*** (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) Domestic Industry Characteristics Ln(4 firm concentration ratio_k) 0.0002** 0.0003*** 0.0001 (0.0001) (0.0001) (0.0001) Ln(Total employment_kt 1) 0.0006*** 0.0005*** (0.0001) (0.0001) Ln(Production employment_kt 1) 0.0006*** (0.0001) Value added/Shipments_kt 1 0.0034*** (0.0006) Inventories/Shipments_kt 1 0.0030*** (0.0007) Number of Observations 86400 86400 85992 85901 85918 Log likelihood 917.824 904.823 891.546 886.427 871.126 Notes: Huber-White robust std errors in parentheses with ***,**, and * indicating statistical significance at the 1%, 5% and 10% levels. 35 Table 5: Predicted Probability of a US AD Tariff for a One S.D. Increase in... Substitute Add industry Substitute Add more Basic alternative size and political industry specification elasticity concentration strength characteristics measures to (1) for size (1) (2) (3) (4) (5) Measure of bilateral import surge Change in Market Share_ikt 1 0.0016 0.0016 0.0016 0.0016 0.0016 Change in Market Share_ROW,kt 1 0.0012 0.0013 0.0013 0.0013 0.0013 Std. Dev. of import growth_ik 0.0005 0.0006 0.0008 0.0008 0.0009 Trade elasticity Indicators discrete change Inelastic import demand 0.0021 0.0020 0.0020 0.0020 Inelastic export supply 0.0020 0.0018 0.0018 0.0018 Alternative inelastic import demand 0.0024 Alternative inelastic export supply 0.0007 Domestic Industry Characteristics Ln(4 firm concentration ratio_k) 0.0016 0.0017 0.0000 Ln(Total employment_kt 1) 0.0020 0.0019 Ln(Production employment_kt 1) 0.0021 Value added/Shipments_kt 1 0.0011 Inventories/Shipments_kt 1 0.0015 Prob of a US antidumping duty in estimation sample 0.0015 0.0015 0.0015 0.0015 0.0015 Number of Observations in estimation sample 86400 86400 85992 85901 85918 36 Table 6: US Antidumping Tariff Imposition: Import Growth and Industry Effects Basic with Basic Basic Basic with steel Industry chem & Industry specification specification & std. dev. of characteristics std.dev. of characteristics & with steel with chemical imports & chemical imports steel dummy dummy dummy interaction dummy interaction (1) (2) (3) (4) (5) (6) Measures of bilateral import growth Growth of imports_ikt 1 0.0003** 0.0004** 0.0003** 0.0004** 0.0003** 0.0003** (0.0001) (0.0002) (0.0002) (0.0002) (0.0001) (0.0001) Measures of Import Volatility Std. deviation of import growth_ik 0.0014*** 0.0016*** 0.0018*** 0.0018*** 0.0012*** 0.0011*** (0.0002) (0.0002) (0.0002) (0.0002) (0.0002) (0.0002) Steel*Std.dev of imp. growth_ik 0.0027*** (0.0004) Chemicals*Std.dev of imp growth_ik 0.0015*** (0.0004) Trade elasticity indicators High inverse import demand elast_k 0.0003*** 0.0006*** 0.0004** 0.0006*** 0.0002 0.0004*** (0.0002) (0.0002) (0.0002) (0.0002) (0.0002) (0.0002) High inverse export supply elast_k 0.0005*** 0.0005*** 0.0004** 0.0006*** 0.0004*** 0.0003* (0.0002) (0.0002) (0.0002) (0.0002) (0.0002) (0.0002) Ln (Real Exchange Rate)_it 1 0.0001*** 0.0001*** 0.0001*** 0.0001*** 0.0001*** 0.0001*** (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) Domestic Industry Variables Ln(Four firm concentration ratio)_k 0.0000 0.0001 (0.0001) (0.0001) Ln(US employment_kt 1) 0.0003*** 0.0006*** (0.0001) (0.0001) Value added/Shipments_kt 1 0.0020*** 0.0030*** (0.0006) (0.0006) Inventories/Shipments_kt 1 0.0016** 0.0042*** (0.0008) (0.0008) Industry indicators Steel dummy 0.0240*** 0.0139*** (0.0048) (0.0033) Chemical dummy 0.0043*** 0.0040*** (0.0014) (0.0014) Number of Observations 82320 82320 82320 82320 81922 81922 Log likelihood 872.212 872.212 911.152 948.324 857.884 897.39 Notes: Huber White robust std errors in parentheses with ***,**, and * indicating statistical significance at the 1%, 5% and 10% levels. 37 Table 7: US Antidumping Tariff Imposition: Import Growth and China Effects Industry Industry Basic Basic with Industry characteristics Basic with China characteristics specification China & std. characteristics with China & & elasticity with China & with China dev.of imports & China std.dev of import interactions elasticity dummy interaction dummy growth interactions interaction (1) (2) (3) (4) (5) (6) Measures of bilateral import growth Growth of imports_ikt 1 0.0001 0.0003* 0.0001 0.0001 0.0002* 0.0001 (0.0001) (0.0001) (0.0001) (0.0001) (0.0001) (0.0001) Measures of Import Volatility Std. deviation of import growth_ik 0.0008*** 0.0016*** 0.0009*** 0.0004*** 0.0011*** 0.0005*** (0.0002) (0.0002) (0.0002) (0.0001) (0.0002) (0.0001) China*Std.dev of imp. growth_ik 0.0031*** 0.0023*** (0.0005) (0.0005) Trade elasticity indicators High inverse import demand elast_k 0.0006*** 0.0005*** 0.0003* 0.0004*** 0.0004*** 0.0002 (0.0002) (0.0002) (0.0002) (0.0001) (0.0001) (0.0001) High inverse export supply elast_k 0.0005*** 0.0005*** 0.0003* 0.0003** 0.0003** 0.0001 (0.0002) (0.0002) (0.0002) (0.0001) (0.0001) (0.0001) China*High Inv Imp Dem Elast_k 0.0062*** 0.0041** (0.0047) (0.0036) China*High Inv Exp Sup Elast_k 0.0033** 0.0033** (0.0030) (0.0032) Ln (Real Exchange Rate)_it 1 0.0001*** 0.0001*** 0.0001*** 0.0001*** 0.0001*** 0.0001*** 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 Domestic Industry Variables Ln(Four firm concentration ratio)_k 0.0001 0.0001 0.0001 (0.0001) (0.0001) (0.0001) Ln(US employment_kt 1) 0.0004*** 0.0004*** 0.0005*** (0.0001) (0.0001) (0.0001) Value added/Shipments_kt 1 0.0025*** 0.0021*** 0.0029*** (0.0005) (0.0006) (0.0005) Inventories/Shipments_kt 1 0.0032*** 0.0029*** 0.0035*** (0.0005) (0.0007) (0.0006) China indicator 0.0179*** 0.0161*** (0.0030) (0.0031) Number of Observations 82320 82320 82320 81922 81922 81922 Log likelihood 847.766 909.913 854.43 796.305 863.119 804.495 Notes: Huber White robust std errors in parentheses with ***,**, and * indicating statistical significance at the 1%, 5% and 10% levels. 38 Table 8: US Antidumping Tariff Imposition: Change in Market Share and Industry and China Effects Market share Market share Market share Market share Market share Market share specification specification specification with steel with chemical with China with steel with chemical with China interactions interactions interactions dummy dummy dummy (1) (2) (3) (4) (5) (6) Measures of bilateral import growth Change in Market Share_ikt 1 0.0358*** 0.0443*** 0.0034 0.0396*** 0.0419*** 0.0010 (0.0093) (0.0103) (0.0089) (0.0100) (0.0104) (0.0242) Change in Market Share_ROW,kt 1 0.0085*** 0.0110*** 0.0067** 0.0164*** 0.0121*** 0.0141*** (0.0028) (0.0034) (0.0029) (0.0030) (0.0034) (0.0031) Interactions with Market Share of Country i Steel*Change in MS_ikt 1 0.0995 (0.2676) Chemicals*Change in MS_ikt 1 0.4941*** (0.1170) China*Change in MS_ikt 1 0.0620** (0.0277) Interactions with Market Share of Rest of World Steel*Change in MS_ROW,kt 1 0.1025*** (0.0256) Chemicals*Change in MS_ROW,kt 1 0.0003 (0.0246) China*Change in MS_ROW,kt 1 0.0273* (0.0162) Measures of Import Volatility Std. deviation of import growth_ik 0.0011*** 0.0013*** 0.0007*** 0.0012*** 0.0013*** 0.0013*** (0.0001) (0.0002) (0.0002) (0.0002) (0.0002) (0.0002) Trade elasticity indicators High inverse import demand elast_k 0.0002*** 0.0005*** 0.0005*** 0.0005*** 0.0006*** 0.0006*** (0.0002) (0.0002) (0.0001) (0.0002) (0.0002) (0.0001) High inverse export supply elast_k 0.0005*** 0.0005*** 0.0005*** 0.0005*** 0.0005*** 0.0005** (0.0001) 0.0002*** (0.0001) (0.0002) (0.0002) (0.0001) Ln (Real Exchange Rate)_it 1 0.0001*** 0.0001*** 0.0001*** 0.0001*** 0.0001*** 0.0001*** (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) Indicators Steel dummy 0.0218*** (0.0044) Chemical dummy 0.0036*** (0.0012) China dummy 0.0161*** (0.0030) Number of Observations 86400 86400 86400 86400 86400 86400 Log likelihood 839.509 906.301 826.732 893.333 909.026 826.619 Notes: Huber White robust std errors in parentheses with ***,**, and * indicating statistical significance at the 1%, 5% and 10% levels. 39 Table 9: Summary Statistics: Foreign Response to US AD Variable Expected Mean Standard Sign Deviation Dependent Variables WTO dispute (=1) or not 0.195 0.398 WTO dispute (=2), AD industry k retaliation (=1), do nothing (=0) 0.576 0.800 Explanatory Variables Likelihood of US antidumping tariff: [] 0.003 0.001 predicted probability of US AD on imports of h from Foreign in t US market access: [+] 10.188 1.729 log of US imports of (AD product) h from Foreign in t ­ 1 Terms of trade spillover: [+] 0.644 0.236 share of Foreign's total US (AD product) exports h to ROW in t 1 Foreign retaliation capacity: [+] 16.992 1.498 log of total US exports to Foreign in t ­ 1 Foreign AD retaliation capacity: [+] 8.655 2.515 log of US exports of k to Foreign in t ­ 1 Foreign DSU retaliation capacity: [+] 0.073 0.079 share of total US exports of all other goods ( k) to Foreign in t ­ 1 Size of US antidumping duty: [] 0.292 0.290 log of 1+US imposed AD duty on h PTA: [] 0.085 0.280 indicator for whether Foreign is in a US preferential trade agreement Notes: 120 observations of US antidumping duties imposed between 1997 and 2006 against WTO members. Prediction from model of table 2, specification 1. Variable h is the 6 digit HS import subject to the US antidumping tariff, k is the 4 digit HS industry, and t is the year of the beginning of the US antidumping investigation that resulted in the newly imposed tariff. 40 Table 10: Foreign's Binomial Probit Model Policy Response to US AD: Marginal Effects Estimates of Filing a WTO Dispute Dependent variable = 1 if Foreign target of US antidumping formally challenges through a WTO dispute Explanatory variables (1) (2) (3) (4) Likelihood of US antidumping tariff: predicted probability of US AD on imports of h from Foreign in t 64.452*** 62.925*** 42.024* (24.569) (20.780) (21.749) US market access: log of US imports of (AD product) h from Foreign in t ­ 1 0.068*** 0.061*** 0.050** (0.021) (0.021) (0.020) Terms of trade spillover: share of Foreign's total US (AD product) exports h to ROW in t 1 0.170 0.167 0.190 (0.148) (0.150) (0.170) Foreign retaliation capacity: log of total US exports to Foreign in t ­ 1 0.057** 0.066** 0.044* (0.026) (0.028) (0.023) Size of US antidumping duty: log of 1+US imposed AD duty on h 0.395*** (0.139) PTA: indicator for whether Foreign is in a US preferential trade agreement 0.022 (0.108) Predicted probability of a WTO dispute (at mean values of the data) 0.186 0.136 0.147 0.107 Observations (number of US AD duties imposed against WTO members) 120 118 118 116 Pseudo R2 0.054 0.217 0.148 0.282 Log pseudolikelihood 56.798 45.591 49.577 41.460 Notes: Marginal effects evaluated at the means of the data. Also estimated with a constant term which is suppressed. Huber White robust standard errors in parentheses with ***, **, and * indicating statistical significance at the 1, 5 and 10% level, respectively. Prediction from model of table 2, specification 1. Variable h is the 6 digit HS import subject to the US antidumping tariff, k is the 4 digit HS industry, and t is the year of the beginning of the US antidumping investigation that resulted in the newly imposed tariff. 41 Table 11: Foreign's Multinomial Probit Model Policy Response to US AD: Marginal Effects Esti- mates of Filing a WTO Dispute, AD Retaliation, or Nothing Dependent variable: Dependent variable: Dependent variable: Foreign's choice of... Foreign's choice of... Foreign's choice of... ...AD ...AD ...AD ...WTO retaliation ...Do ...WTO retaliation ...Do ...WTO retaliation ...Do Explanatory variables Dispute (=2) in k (=1) nothing (=0) Dispute (=2) in k (=1) nothing (=0) Dispute (=2) in k (=1) nothing (=0) (1) (2) (3) Likelihood of US antidumping tariff: 68.938*** 48.953* 19.985 57.826** 25.256 32.569 44.529** 20.754 23.775 predicted probability of US AD on (25.172) (25.166) (30.845) (23.360) (27.401) (33.547) (22.096) (29.943) (33.234) imports of h from Foreign in t US market access: 0.062*** 0.043* 0.018 0.040** 0.035 0.004 log of US imports of (AD product) h (0.023) (0.022) (0.028) (0.019) (0.023) (0.027) from Foreign in t ­ 1 Terms of trade spillover: 0.268* 0.161 0.107 0.197 0.117 0.080 share of Foreign's total US (AD (0.153) (0.178) (0.211) (0.162) (0.210) (0.243) product) exports h to ROW in t 1 Foreign AD retaliation capacity: 0.029 0.053** 0.081*** 0.033** 0.047** 0.081*** log of US exports of k to Foreign in t ­ 1 (0.017) (0.022) (0.024) (0.016) (0.022) (0.024) Foreign DSU retaliation capacity: 0.986** 1.166* 0.180 0.469 1.029 0.561 share of total US exports of all other (0.482) (0.655) (0.759) (0.413) (0.732) (0.778) goods ( k) to Foreign in t ­ 1 Size of US antidumping duty: 0.346*** 0.268** 0.078 log of 1+US imposed AD duty on h (0.131) (0.132) (0.163) PTA: 0.053 0.156 0.103 indicator for whether Foreign is in a US (0.066) (0.208) (0.215) preferential trade agreement Predicted probability of outcome (at mean 0.181 0.182 0.637 0.122 0.186 0.693 0.084 0.191 0.724 values of the data) Observations (number of US AD duties 118 116 114 imposed against WTO members) Log pseudolikelihood 105.081 85.857 80.430 Notes: Marginal effects evaluated at the means of the data. Also estimated with a constant term which is suppressed. Huber White robust standard errors in parentheses with ***, **, and * indicating statistical significance at the 1, 5 and 10% level, respectively. Prediction from model of table 2, specification 1. Variable h is the 6 digit HS import subject to the US antidumping tariff, k is the 4 digit HS industry, and t is the year of the beginning of the US antidumping investigation that resulted in the newly imposed tariff. 42 References [1] Bagwell, Kyle and Robert W. Staiger. 1990. "A Theory of Managed Trade," American Economic Review 80(4): 779-795. [2] Bagwell, Kyle and Robert W. Staiger. 1999. "An Economic Theory of GATT," American Economic Review 89(1): 215-248. [3] Bagwell, Kyle and Robert W. Staiger. 2003. "Protection and the Business Cycle," Ad- vances in Economic Analysis & Policy, Berkeley Electronic Press 3(1): article 3. [4] Bagwell, Kyle and Robert W. Staiger. 2009. "What Do Trade Negotiators Negotiate About? Empirical Evidence from the World Trade Organization," July. Revised version of NBER Working Paper No. 12727. [5] Bown, Chad P. 2010. "China's WTO Entry: Antidumping, Safeguards, and Dispute Settle- ment," in Robert C. Feenstra and Shang-Jin Wei, eds. China's Growing Role in World Trade. Chicago, IL: University of Chicago Press. [6] Bown, Chad P. 2009a. "Global Antidumping Database" Version 5.0, available on-line at http://www.brandeis.edu/~cbown/global ad/ , July. [7] Bown, Chad P. 2009b. Self-Enforcing Trade: Developing Countries and WTO Dispute Set- tlement. Washington, DC: Brookings Institution Press. [8] Bown, Chad P. 2005a. "Trade Remedies and World Trade Organization Dispute Settlement: Why Are So Few Challenged?" Journal of Legal Studies 34(2): 515-555. [9] Bown, Chad P. 2005b. "Participation in WTO Dispute Settlement: Complainants, Interested Parties and Free Riders" World Bank Economic Review 19(2): 287-310. [10] Bown, Chad P. 2004a. "On the Economic Success of GATT/WTO Dispute Settlement," The Review of Economics and Statistics 86(3): 811-823. [11] Bown, Chad P. 2004b. "Trade Disputes and the Implementation of Protection under the GATT: An Empirical Assessment," Journal of International Economics 62(2): 263-294. 43 [12] Bown, Chad P. and Meredith A. Crowley. 2007. "Trade Deflection and Trade Depres- sion," Journal of International Economics 72(1): 176-201. [13] Bown, Chad P. and Meredith A. Crowley. 2006. "Policy Externalities: How US An- tidumping Affects Japanese Exports to the E.U.," European Journal of Political Economy 22(3): 696-714. [14] Broda, Christian, Joshua Greenfield and David E. Weinstein. 2006. "From Ground- nuts to Globalization: A Structural Estimate of Trade and Growth," NBER Working Paper No. 12512, September. [15] Broda, Christian, Nuno Lim~o and David E. Weinstein. 2008. "Optimal Tariffs and a Market Power: The Evidence," American Economic Review 98(5): 203265. [16] Busch, Marc L., Eric Reinhardt, and Gregory Shaffer. 2008. "Does Legal Capacity Matter? Explaining Patterns of Protectionism in the Shadow of WTO Litigation," Georgetown University manuscript. [17] Chang, Won and L. Alan Winters. 2002. "How Regional Blocs Affect Excluded Countries: The Price Effects of MERCOSUR," American Economic Review 92(4): 889-904. [18] Crowley, Meredith A. 2008. "Cyclical Dumping and US Antidumping Protection: 1980- 2001," Federal Reserve Bank of Chicago Working Paper # 2007-21. [19] Crowley, Meredith A. 2007. "Split Decisions in Antidumping Cases," Federal Reserve Bank of Chicago mimeo. [20] Crowley, Meredith A. and Robert Howse 2010. "US - Stainless Steel (Mexico)," World Trade Review 9(1): ... [21] Feinberg, Robert M. 1989. "Exchange Rates and Unfair Trade," The Review of Economics and Statistics, 71(4): 704-707. [22] Feinberg, Robert M. 2005. "US Antidumping Enforcement and Macroeconomic Indicators Revisited: Do Petitioners Learn?" Review of World Economics, 141: 612-622. 44 [23] Feinberg, Robert M. and Kara M. Reynolds. 2006. "The Spread of Antidumping Regimes and the Role of Retaliation in Filings," Southern Economic Journal 72(4): 877-890. [24] Finger, J. Michael, H. Keith Hall and Douglas R. Nelson. 1982. "The Political Econ- omy of Administered Protection," American Economic Review 72(3): 452-466. [25] Irwin, Douglas A. 2005. "The Rise of US Anti-dumping Activity in Historical Perspective," The World Economy (28)5: 651-668. [26] Irwin, Douglas A. 1996. Against the Tide: An Intellectual History of Free Trade. Princeton, NJ: Princeton University Press. [27] Johnson, Harry G. 1953-54. "Optimum Tariffs and Retaliation," Review of Economic Stud- ies 21(2): 142-53. [28] Knetter, Michael M. and Thomas J. Prusa. 2003. "Macroeconomic Factors and An- tidumping Filings: Evidence from Four Countries," Journal of International Economics, 61:1- 17. [29] Maggi, Giovanni. 1999. "The Role of Multilateral Institutions in International Trade Coop- eration," American Economic Review 89(1): 190-214. [30] Maggi, Giovanni and Robert W. Staiger. 2008. "On the Role and Design of Dispute Settlement Procedures in International Trade Agreements," NBER Working Paper No. 14067, June. [31] Prusa, Thomas J. and Susan Skeath. 2002. "The Economic and Strategic Motives for Antidumping Filings." Review of World Economics, 138: 389-413. [32] Staiger, Robert W. and Frank A. Wolak. 1994. "Measuring Industry-Specific Protection: Antidumping in the United States." Brookings Papers on Economic Activity: Microeconomics, 51-118. 45