The spread of antidumping regimes and the role of retaliation in filings.

Reynolds, Kara M.

1. Introduction and Previous Literature

Although business news in the developed world tends to emphasize trade policy enforcement by the two large economic powers, the European Union (EU) and the United States, the use of antidumping has become very widespread--39 other World Trade Organization (WTO) member countries (and possibly other nonmembers) initiated antidumping cases over the 1995-2003 period. From another perspective, U.S. exporters were subjected to 139 antidumping cases during this period, by enforcement agencies representing 20 countries (the EU being regarded for these purposes as a single country). In this context, it is natural to consider whether antidumping filings may be motivated as retaliation against similar measures imposed on a country's exporters. This is the focus of our study, though we also control for the bilateral export flows involved, exchange rate effects, and nonretaliatory impacts of past cases, dealing with other motivations--macroeconomic, industry, and political considerations--through industry, country, and year fixed effects. (1)

Antidumping duties are allowed under WTO rules when there is material injury or threatened injury to a competing domestic industry from sales by an exporter made at unfairly low prices (usually prices alleged to be below those charged in the home market, but often below costs). Each country establishes its own antidumping enforcement mechanism, and case filings are brought by domestic companies (as well as labor unions and trade associations) to their respective government enforcement agencies. In recent years the lines between this form of "administrative protection" and other forms of trade restrictions have been blurred--at least in the views of many observers. Therefore, in studying motivations for filing antidumping petitions, researchers have turned to considering not just case-specific factors, but also more general determinants of the demand and supply for protection against imports. However, until recently little attention has been given to strategic motivations for antidumping. (2)

But given the obvious spread of antidumping enforcement, economists have begun to consider the issue of retaliation in antidumping filings and the increasing globalization of this instrument of trade policy. Miranda, Torres, and Ruiz (1998) and the U.S. Congressional Budget Office (1998) were the first to document the dramatic growth in the number of countries joining the "antidumping club." Miranda, Torres, and Ruiz suggest that if the emergence of increased antidumping enforcement by developing countries was a quid pro quo for general trade liberalization, there may be welfare gains from this proliferation of antidumping filings, at least in a second-best sense. The CBO paper acknowledges this possibility as well, though their focus is more on whether U.S. exporters have been harmed and/or singled out for retaliation by new users of antidumping; on these latter issues they suggest minimal impact to that point, noting however that with continued growth in antidumping by developing countries U.S. exporters may begin to be more affected in the future.

Lindsay and Ikenson (2001) highlight the growing threat to U.S. interests posed by new antidumping users. They view these new users as following the "bad U.S. example" of protecting domestic industries from foreign competition under the banner of antidumping, agreeing with earlier authors that developing countries have been increasing the use of antidumping in part as an offset to lower negotiated tariffs.

Prusa (2001) focuses his analysis on the impact of U.S. antidumping on trade flows, but also discusses the data on the spread of antidumping enforcement in the developing world. In the latter context, he briefly discusses the strategic issues involved in a government's decision to adopt an antidumping policy---on the one hand, actions may be aimed at deterring other users of antidumping, but on the other hand, this deterrence may fail, with the result a prisoner's dilemma with retaliation occurring instead. Prusa and Skeath (2002) more fully develop this point, finding evidence consistent with strategic motivations behind antidumping filings.

More recent work, and closest to the focus of our empirical work below, is that by Blonigen and Bown (2003), Francois and Niels (2004), and Prusa and Skeath (2004). Blonigen and Bown develop a trigger price model, based on the reciprocal dumping framework, which allows for the threat of an antidumping action against a country to restrain that country's own antidumping activity, and find some evidence consistent with this prediction for the United States. Francois and Niels (2004) suggest that new users may be initiating antidumping actions to retaliate against countries taking antidumping action against their exports. They find that Mexican antidumping petitions were three times more likely to be successful when filed against countries that had initiated a case against Mexican exports in the previous year.

Prusa and Skeath (2004) address some of the issues considered in this paper, though their dataset covers an earlier period, 1980-1998, much of which was dominated by traditional users of antidumping. Their stated focus is to explore whether the increase in global use of antidumping was solely because of increased "unfair trading"--and they (not surprisingly--to anyone who has studied the subject) reject this hypothesis. They find that antidumping users are more likely to target other users of antidumping than those without such enforcement, and that countries are more likely to target exporting countries with a past history of bringing cases against them. The latter result Prusa and Skeath interpret as retaliation or tit for tat, but one would not generally view a 1995 case by India against the EU following an EU case against India 10 years, or even 3 years, earlier as strategic behavior; most game theoretic models suggest an immediacy of response in order to use retaliation as a means of establishing credibility of threat, or as an effective tit-for-tat mechanism.

Our analysis extends the previous work in two important ways: first, our time period for analysis is five years more recent (five years that were a period of tremendous growth in the membership in the antidumping "club"); second, we examine the potential for retaliation and]or threat at the industry level. We examine not merely the impact of threatened retaliation, but the actual patterns of retaliation that seem to have emerged over the past decade in the industry/country-target-specific antidumping actions of 40 countries. We attempt in our analysis to capture both retaliation motivations expanding the use of antidumping and possible threat impacts that may lessen this somewhat. We also allow for the possibility, as noted in a recent working paper by Bown and Crowley (2004), that past antidumping cases--through their trade-distorting effects, or what they refer to as "trade deflection"--can influence the use of import protection of all types, including antidumping.

2. Empirical and Theoretical Motivation

Before discussing theoretical issues, it is instructive to examine the patterns of the global spread of antidumping, in terms both of cases brought and of number of active national enforcement agencies involved. Looking at Table 1, it may be surprising to some to see that the leading user of antidumping since 1995 has been India, with Argentina and South Africa among the top five. Turning to Figures 1 and 2, the same story is told from a different perspective. We see there that although there has been a significant increase in the number of antidumping cases brought worldwide, a more dramatic increase has occurred in the number of countries getting involved in bringing such cases, roughly a tripling of noncasual enforcers (defined as more than 2 cases brought in a year) between the late 1980s and today, with all of this growth brought about by new enforcement agencies in developing economies.

To motivate our empirical analysis, consider a model (built on Brander and Krugman 1983) referred to in Blonigen and Bown (2003) (and presented in Blonigen 2000). There are two quantity-setting firms, one from each of two countries, competing in the two markets (segmented by transport costs). Antidumping filings impose a duty [tau], with a probability of success [phi], requiring a fixed cost K. The probability of success is an increasing function of the foreign firm's quantity share of the domestic market. Blonigen and Bown then consider an infinitely repeated game--with 2 stages in each period-involving the choice of quantities and then the independent decision of each firm to file an antidumping petition or not. Using the trigger strategy to achieve the cooperative outcome of no antidumping filings (with the threat of antidumping infinitely into the future if the rival defects), they find the avoidance of antidumping is supported by sufficiently high punishment costs from the rival's antidumping actions.

However, if these threatened costs are relatively low (or nonexistent in the case of a rival without an antidumping enforcement apparatus), the filing of antidumping cases becomes more likely. (3) Furthermore, cost disadvantages by the domestic firm increases their gains from antidumping and the likelihood of a prisoner's dilemma result of antidumping by both countries. Not surprisingly for models of this sort, equilibrium outcomes are quite sensitive to the parameters of the model, and we can find the impact of both actual and threatened antidumping by one country against another to provoke either retaliation or deterrence. From an empirical perspective, it may be true both that in equilibrium the increased threat of antidumping by a rival leads to deterrence and that over the period we study a disequilibrium unraveling of retaliation may have occurred.

In what follows we examine the pattern of antidumping filings by particular industries in particular countries against particular target countries in response to past antidumping actions against that particular industry as well as more generally against the country more broadly. We also consider the role of threat, revealed through the target country's recent antidumping activity globally. Of course, there are other motivations for filing antidumping cases, and we control for exchange rate movements and import flows, and for other macroeconomic, industry-specific and political factors via fixed effects, as well as dealing with the concern (presented in Bown and Crowley 2004) that "trade deflection" caused by third-party antidumping may induce new antidumping filings.

3. Empirical Analysis

We have obtained WTO data from all member countries on their antidumping filings during 1995-2003 by target country and industry category (20 Harmonized System (HS) sections of which 19 were involved in at least one filing over this period). Counting the EU as one country, the dataset includes 40 importing countries filing at least one antidumping case against 72 exporting countries. (4) Limiting our analysis to the years 1996-2003, so we can observe a one-year lag in filings, we have 431,680 importing country/exporting country/industry sector/year observations as to whether an antidumping case was filed or not. Petitions were filed in 1,610 (or 0.37%) of these observations.

We seek to explain this filing decision using fixed effects in a probit binary choice model, with the primary explanatory variables of interest those that will determine whether the filing decision is motivated by the urge to retaliate against certain trading partners. (5) We include a dummy variable that indicates whether the exporting country filed an antidumping case against the importing country and industry category during the past year (CAT).

Unfortunately, the industry categories by which our data are organized, HS sections, are too broad for us to be certain that the same firms are involved in a bilateral exchange of cases between two countries in successive years, which would be the conventional notion of retaliation by firms involved. However, anecdotally, this does seem to occur; for example, a 2001 antidumping case brought by Canada against India in hot-rolled sheets was followed by a 2002 case by India against Canada in hot-rolled coils/sheets/strips/plates. Similarly a 2001 antidumping case filed by the United States against EU members in cold-rolled carbon steel fiat products was followed in 2002 by an EU case filed against the United States in that same narrowly-defined product.

But in general it is likely that a case against an industry category in a particular country the previous year involved a different group of firms than the subsequent case within the same industry category. This may be retaliation, but the mechanism through which it derives is less clear. (6) Especially in a small country there may be close business links between companies in different narrow product lines within the same broad category, and they may also be linked through unions, trade associations, or law firms in common. In addition, retaliation may in part be reflected at the country level--the government agency charged with enforcing antidumping statutes may be more likely to make an affirmative determination and impose larger dumping margins against those exporting countries that filed cases against the importing country in the previous year. If so, firms will anticipate higher expected benefits from filing cases against these countries, and will thus be more likely to file antidumping petitions against them.

To expand on the latter point, we also consider whether the exporting country filed a case against any other industry in the importing country in the past year (OTHER). Because broad industry categories may cause the CAT and OTHER variables to both pick up retaliation on the country level, in other specifications we instead include a single variable that indicates whether the exporting country filed at least one case against the importing country in the previous year (RETALIATION).

In the theoretical model described above, antidumping filings are avoided in the trigger strategy equilibrium when the punishment cost associated with filing is sufficiently high. We hypothesize that a country is more likely to be deterred from filing antidumping petitions against those countries with active antidumping laws, who are therefore more likely to retaliate. Moreover, we expect the potential threat of retaliation to be particularly intimidating when the potential target is an important export market for the country. (7) As a measure of the potential threat from the exporting country's own antidumping enforcement, we include the exporting country's total worldwide filings the previous year interacted with the importing country's total exports to that country in billions of dollars (DETER). (8) If countries are deterred from filing cases against their large export markets that have reputations for using antidumping enforcement, we would expect this variable to have a negative estimated coefficient.

In using these specific retaliation and deterrence variables, we are able to capture a number of strategic aspects of antidumping filings that were not captured in the Prusa and Skeath (2004) analysis. Prusa and Skeath measure retaliation using a "Tit-for-Tat" variable that is defined as whether the exporting country had ever filed an antidumping case against the importing country in any industry. Because antidumping statutes at the WTO specify that petitions must be filed by a specific industry, it is important to study the industry-level retaliation as measured in CAT. Moreover, as mentioned above, most game theory models emphasize the immediacy of retaliation (i.e., within the next year), which is more precisely measured in one-year lagged case filings rather than over a longer time period.

Prusa and Skeath (2004) also hypothesize that countries are more likely to file antidumping petitions against countries that have filed antidumping petitions in the past against any country, or those that have joined the "antidumping club." In contrast, we hypothesize that countries may actually be deterred from filing against heavy users of antidumping because of a fear of retaliation, as captured in DETER. However, we also attempt to control for possible "club" effects in the sense that we use a dummy variable that equals 1 when the exporting country is a "traditional" antidumping user, which includes Australia, Canada, the European Union, New Zealand and the United States (TRADITIONAL). In other specifications we add fixed effects for additional nontraditional users who have been the leading targets of petitions during the sample period: China, Korea, Taiwan, India, and Indonesia.

Although our primary interest is in the retaliation motivations, several control variables are included as well. The likelihood of filing a case clearly should depend on the volume of imports from the potential target, so we also include bilateral imports at the broad HS section level (IMPORTS) in the estimating equation. (9) In addition, Bown and Crowley (2004) have discussed the role the spread of antidumping has played in "trade deflection"--cases filed against one country may divert its trade flows elsewhere, leading to more import protection being sought by third countries, including antidumping filings. We therefore include a variable (DEFLECTION) that equals the number of global antidumping cases filed the previous year in the particular industry category, excluding those filed against the importer being considered.

In order to control for macroeconomic, political, and industry factors, we use year, industry category and importing country fixed effects in all specifications. As noted above, one explanation for the surge in antidumping petitions may be that developing countries have been increasing the use of antidumping in part as an offset to lower negotiated tariffs. Moreover, countries joining the WTO must subscribe to all aspects of the treaty, including the antidumping provisions, and this may encourage new members to develop and start using antidumping laws for the first time. However, because most of the countries in our sample joined the WTO at the same time, we expect that fixed importing country effects would capture most of these determinants of filings. (10) There may also be importing or exporting country macroeconomic effects that vary over time. We include lagged log bilateral real exchange rates (EXCHANGE) to control for some of these effects. (11)

In Table 2, we present our full sample results, where marginal effects on the dependent variable (likelihood of filing a case) rather than the actual probit coefficients are shown. A statistically significant positive retaliation effect is found in all three specifications, both the direct impact of a case the previous year in the same 2-digit HS sector (CAT) and the more indirect impact of a case filed the previous year against a different industry sector of the country (OTHER). These two effects are similar (both appear to increase the likelihood of filing a case by slightly more than 20%), suggesting that retaliation is often determined at the country level rather than simply by the industries directly involved. (12) We fail to find evidence that heavy use of antidumping laws can deter future filings against a country. In fact, countries are significantly more likely to file petitions against their important export markets that are heavy users of antidumping laws (DETER). If antidumping filings successfully deter future cases, the impact would be negative.

The volume of industry imports from the exporting country (IMPORTS) is, as expected, an important determinant in the decision to file an antidumping case. Our results suggest that a $1 billion increase in the sectoral volume of imports from the targeted country results in a 1-3% increase in the likelihood of filing an AD case. Like Knetter and Prusa (2003), we find that a real appreciation of the domestic currency increases the likelihood of filing an antidumping petition against the exporting country (EXCHANGE). Countries are slightly more likely to file petitions when there has been significant antidumping activity in the industry elsewhere in the world in the previous year (DEFLECT); this result is consistent with the view that antidumping cases deflect trade to third countries, thus increasing the likelihood that these third countries will seek some form of protection.

None of these results seem to be driven by any particular industry or correlation with unobserved exporting country effects; we continue to find a positive and significant retaliation effect on the likelihood of filing antidumping when we allow for differential retaliation effects in the leading user industry (metals) and when we control for cases against traditional users of antidumping laws and the leading targets of antidumping petitions. In fact, the average retaliation effect associated with a case filed in the same industry sector is nearly three times higher when we allow this result to differ in the metals industry, because the metals industry is less likely to retaliate against a case filed against it than other industry sectors. In contrast, the results suggest that cases are more likely to be filed in the metals industry to retaliate against a case filed in the previous year against some other industry. (13)

Cases do seem more likely to be brought against traditional users of antidumping, which could be interpreted as a type of retaliation for past cases or a "club effect" as defined by Prusa and Skeath (2002, 2004). Although the magnitudes of the marginal effects are smaller when we control for the leading nontraditional targets of antidumping actions presented in column 3, the effects of CAT, OTHER, DETER, IMPORTS, EXCHANGE, and DEFLECTION continue to be positive and significant.

The large marginal effects associated with exporters China, India, Korea, and Indonesia suggest that highly competitive, low-cost countries such as these are targeted more often than predicted by our other variables. Marginal effects associated with the importing country fixed effects, as presented in Table 3, confirm the summary statistics described above: whereas traditional users continue to be heavy users of antidumping, nontraditional users such as India and Argentina have grown in importance. The industry fixed effects show that on average more antidumping petitions are filed in the Base Metals and Plastic & Rubber industries than any others.

Because it is likely that both CAT and OTHER are capturing retaliation on a country level rather than on a narrowly-defined industry basis, in Table 4 we combine the two effects into a single variable, RETALIATION. The results are similar to those found in Table 2. The retaliation variable is positive and statistically significant; estimates suggest that the likelihood of a country filing a case is more than 7% higher against those countries that targeted it in the previous year. There is no indication that the metals industry retaliates more than other industries, reflecting the results above that whereas the metals industry is less likely to retaliate against same-industry cases, it is more likely to be used to retaliate against cases filed against other industries. Estimates associated with other variables, including DEFLECT and DETER, are virtually identical to those discussed in Table 2.

In Table 5, we create several subsamples for analysis. We examine separately the filing decisions by traditional users and new users, both those filed against all countries and those filed only against traditional users. The results from these subsamples are slightly different. When we consider cases brought by either new users or traditional users against all countries, the retaliation effect continues to be positive and significant. When we consider cases brought by traditional users against just traditional users, the retaliation effect is no longer statistically significant (though still positive), although there is still a greater likelihood of filing against heavier users of antidumping. This suggests that whereas both new and traditional users believe they may be able to deter future antidumping actions by new users through retaliation, those entrenched in the antidumping club, or the traditional users, know that little can be gained from retaliating against other members of this club.

The subsample results confirm that countries are more likely to file against important trading partners that are heavy users of antidumping (DETER), and users target traditional members of the antidumping club more often (TRADITIONAL) even after controlling for the level of antidumping use by these countries. The estimates associated with the trade deflection impact vary across subsamples. Traditional users seem to be somewhat more likely to file against new users in order to protect themselves from deflected trade, whereas new users are more likely to file against traditional users to protect themselves from trade surges that occur because of increased antidumping activity. This result undoubtedly reflects differing trade patterns between the two groups of users.

One limitation of the sample we have been analyzing to this point is that many of the observations are characterized by no bilateral import flows, and for these observations there is no reason to expect an antidumping filing. In Table 6, we replicate the probit specification reported in Table 4 on a more limited sample, one that excludes zero bilateral import observations. (14) The estimated magnitude of the retaliation effect remains statistically significant using this subsample; countries are 12-30% more likely to file a case against a country that targeted it in the previous year. Other results are unchanged from earlier specifications; DETER, IMPORTS, DEFLECT, TRADITIONAL, and EXCHANGE are significant and of the same sign and approximately the same magnitudes as those reported earlier.

Except when we limit our analysis to cases by traditional users against traditional users of antidumping, we find strong support for a retaliation motive in filing antidumping petitions. No support is found for a deterrent effect, though we cannot reject that as a possibility. Note that deterrence in the Blonigen/Bown model, as in the standard trigger price result, is an equilibrium concept. There is no reason to think that the period we are observing--looking again at the explosion of active antidumping enforcers around the world during the late 1990s--represents an equilibrium in which deterrence may be supported. It may well be that we now are or soon will be in such an equilibrium; a future study may perhaps capture that relationship.

4. Conclusion

In recent years, many observers have begun to note the proliferation of antidumping regimes and the possibility that established users of this trade policy instrument are being retaliated against. Others have suggested that at some point (if not quite yet) an equilibrium could be reached in which the threat of antidumping provides a deterrent to further cases. In this paper we have used a unique WTO dataset to provide the strongest evidence to date that a significant share of antidumping filings worldwide can be interpreted as retaliation.

This is not to say that macroeconomic, political, and industry-specific factors are not important-of course they are. Moreover, it is possible that some of the growth in antidumping filings over the past decade can be explained by a quid pro quo effect associated with tariff concessions made by countries during the Uruguay Round, and we explore this possibility in ongoing work. But even after controlling many other possible determinants (in part through fixed effects) and for trade-related rationales for filings, we have shown that retaliation clearly plays a role. This suggests, among other things, that industries considering bringing an antidumping petition may wish to add in to their calculations the possible costs to their exports of future antidumping cases against them (and perhaps that antidumping authorities could internalize these "externalities" in their decision-making process).

We have used the term "retaliation" throughout this paper--is it possible that what we are observing is simply "learning"? The theoretical motivations are quite different; retaliation is motivated by the need to maintain credibility in attempting to deter future antidumping (or part of a disequilibrium movement to the prisoners dilemma outcome), whereas learning simply reflects a changed awareness of the relative costs and benefits of bringing a case. It seems unlikely that learning how to file antidumping cases (or to create an antidumping authority) requires that a prior case be brought against the country (or that country A learns of the wisdom of filing against country B only when country B has filed against A the year before). Our empirical results seem more consistent with some variant of retaliation, though future research should try to disentangle these motivations.

References

Aggarwal, Aradhna. 2004. Macro economic determinants of antidumping: A comparative analysis of developed and developing countries. Worm Development 32:1043-57.

Blonigen, Bruce A. 2000. U.S. antidumping filings and the threat of retaliation. Unpublished paper, University of Oregon.

Blonigen, Bruce A., and Chad P. Bown. 2003. Antidumping and retaliation threats. Journal of International Economics 60: 249-73.

Blonigen, Bruce A., and Thomas J. Prusa. 2003. Antidumping. In Handbook of international trade, edited by E.K. Choi and J. Harrigan. Oxford, U.K. and Cambridge, MA: Blackwell Publishers, pp. 251-84.

Bown, Chad P., and Meredith A. Crowley. February 2005. Trade deflection and trade depression. Federal Reserve Bank of Chicago Working Paper (WP-2003-26).

Brander, James, and Paul Krugman. 1983. A "reciprocal dumping" model of international trade. Journal of International Economics 15:313-21.

Brink, Lindsey, and Dan Ikenson. 2001. "Coming home to roost: Proliferating antidumping laws and the growing threat to U.S. exports." Cato Institute, July 30, 2001. Available http://www.freetrade.org/pubs/pas/tpa-014es.html.

Feinberg, Robert M., and Barry T. Hirsch. 1989. Industry rent seeking and the filing of "unfair trade" complaints. International Journal of Industrial Organization 7:325-40.

Finger, J. Michael. 1981. The industry-country incidence of less than fair value cases in the U.S. import trade. Quarterly Review of Economics and Business 21:260-79.

Francois, Joseph F., and Gunnar Niels. 2004. Political influence in a new antidumping regime: Evidence from Mexico. Timbergen Institute Working Paper No. TI 2004-011/2.

Greene, William. 2003. Fixed effects and bias due to the incidental parameters problem in the Tobit model. Economic Reviews 2004.

Knetter, Michael M., and Thomas J. Prusa. 2003. Macroeconomic factors and antidumping filings: Evidence from four countries. Journal of International Economics 61:1-17.

Miranda, Jorge, Raul A. Torres, and Mario Ruiz. 1998. The international use of antidumping: 1987-1997. Journal of World Trade 32:5-71.

Prusa, Thomas J. 1992. Why are so many antidumping petitions withdrawn? Journal of International Economics 33:1-20.

Prusa, Thomas J. 2001. On the spread and impact of antidumping. Canadian Journal of Economics 34:591-611.

Prusa, Thomas J., and Susan Skeath. 2002. The economic and strategic motives for antidumping filings. Weltwirtschaftliches Arhiv 138:389-413.

Prusa, Thomas J., and Susan Skeath. 2004. Modern commercial policy: Managed trade or retaliation? In The handbook of international trade, Volume 11: Economic and legal analyses of trade policy and institutions, edited by E. K. Choi and James Hartigan. Malden, MA and Oxford, UK: Blackwell Publishers. pp. 358-82. Forthcoming.

Staiger, Robert W., and Frank A. Wolak. 1994. Measuring industry specific protection: Antidumping in the United States. Brookings Papers on Economic Activity: Microeconomics volume 1; pp. 51-118.

U.S. Congressional Budget Office. 1998. Antidumping action in the United States and around the world: An analysis of international data. Washington, DC: U.S. Congressional Budget Office.

World Trade Organization. Antidumping. Accessed 8 June 2005. Available at http://www.wto.org/english/tratop_e/adp_e/ adp_e.htm.

Zanardi, Maurizio. 2004. Antidumping: What are the numbers to discuss at Doha? The World Economy 27:403-33.

(1) There is a large literature explaining antidumping filings by these industry-specific, political, and macroeconomic factors, starting with Finger (1981), and including work by Feinberg and Hirsch (1989), Knetter and Prusa (2003), and Aggarwal (2004). For an excellent survey see Blonigen and Prusa (2003).

(2) Papers by Prusa (1992) and Staiger and Wolak (1994) discuss strategic motivations for filings from a somewhat different perspective than the more recent work discussed below. Their emphasis was on antidumping as a mechanism to promote tacit collusion between domestic and foreign firms on the price/quantity dimension, rather than focusing on the interdependencies between antidumping enforcement policies in different countries.

(3) This might be referred to as the penalty phase of the game.

(4) Because members are the only countries required to report their filings to the WTO, our dataset may underestimate the number of petitions filed by new WTO members prior to joining, notably Taiwan, which used its antidumping law extensively in that period, between 1995 and 1999. See Zanardi (2004) for more information about antidumping use by non-WTO members. We therefore exclude Taiwan from the sample, in addition to nine exporting countries (Bahrain, Bosnia-Herzegovina, Macau, Cuba, Georgia, North Korea, Libya, Oman, and Qatar) that had extremely limited numbers of antidumping cases filed against them during this period and for which we were unable to obtain trade and exchange rate data. Preliminary estimates of our primary variables of interest using a sample with limited control variables but including these countries were not qualitatively different from the results presented below.

(5) Some studies suggest that the probit model does not lend itself well to fixed effects because the parameter estimates are biased and inconsistent when the length of the panel is small and fixed (the "incidental parameter problem"). Greene (2003) found that although the upward bias is persistent, it drops off dramatically as the number of time periods increases beyond three. He concludes that a fixed-effects model may be preferred if the alternative is a misspecified random effects model or a pooled estimator that neglects the cross-unit heterogeneity. Our preliminary estimates from random-effects probit models were not qualitatively different from those presented below.

(6) As suggested by a referee, the following simple game motivates the rationale for using industry-level filings to study retaliation motivations. Assume that the probability that an antidumping petition will be successful is higher when any antidumping action was taken against the industry by the exporting country in the previous period. Now consider two firms, A and B, in the same industry that produces a nonoverlapping set of products. If an antidumping petition is filed against firm A in period 1, firm B is more likely to file an antidumping petition in period 2 knowing that the likelihood of its success is higher. Thus retaliation does not need to be at the firm level if the antidumping authority is likely to take past filings against the industry into account when making decisions. Some recent empirical evidence supports this view. As noted above, Francois and Niels (2004) found that Mexican antidumping petitions were more likely to be successful when filed against countries that had initiated a case against Mexican exports in the previous year.

(7) We thank a referee for suggesting this measure.

(8) Specifically, we use the importing country's total exports to the potential target of the antidumping investigation in a single midsample year, 1998, because consistent trade data were not available for all years in our sample. It should be noted, however, that the results from estimations using annual export data un a sample from 1996 to 2001 were not significantly different from those presented here.

(9) These data are obtained from the NBER-United Nations Trade Data, 1962-2000. We use a single midsample observation, 1998, for this variable as consistent data were not available for all years in our sample and our primary rationale for including this variable was to capture cross-sectional variation. However, results from estimations on a sample from 1996 to 2001 using annual trade data and including the annual growth in industry imports were virtually identical to those presented here.

(10) As noted by a referee, the impact of tariff concessions made by the importing country under the Uruguay Round may not be uniform across all exporting countries. Thus, the importing country fixed effects will not fully account for the effect of tariff concessions on the likelihood of the country filing an antidumping petition against a particular country. In ongoing research, we are attempting to further analyze the "quid pro quo" issue; however, we prefer here to focus on the potential retaliation/ threat response to past cases. It should be noted that tariff liberalization by an importing country is unlikely to be correlated with past cases brought by an exporter, suggesting that omitting this factor should not bias our estimates of the retaliation coefficient.

(11) Specifically, we calculate lagged log bilateral real exchange rates using nominal exchange rate and consumer price index data from the International Monetary Fund's International Financial Statistics. We normalize each series by dividing by its sample mean prior to taking logs. Other macro variables that have found to be significant in research such as Knetter and Prusa (2003) include real GDP growth. Because of data limitations, we exclude variables such as these in the specification presented here. However, results from estimations using a sample from 1996 to 2001 that include the exporting and importing countries GDP growth were virtually identical to those presented here.

(12) The similarity of effects may also suggest that our industry classifications are too broad to effectively capture true retaliation by the same parties targeted by a past antidumping case.

(13) One explanation for this is the well-developed legal/consulting infrastructure built up in metals to bring antidumping cases, at least in the United States.

(14) We also replicated the results of Tables 2 and 5, with a similar outcome. In addition, we considered two smaller samples, one of which includes only exporters with active antidumping enforcement agencies and a second that includes only those industry categories in which more than 50 cases were filed during the sample period. Results from both subsamples are similar to those reported here. In focusing our attention on the fuller sample, we axe likely to be conservative in our estimates of the retaliation effect (because cases against nonusers of antidumping obviously cannot be explained by retaliation). Similarly, because there are other mechanisms available for retaliation (both through trade laws and political pressures), our estimates may understate the opportunities for retaliation. Of course, we would not pick up retaliation by domestic petitioners against exporters who have brought other types of trade cases--for example, escape clause--or have had a dumping duty continued after a sunset review.

Robert M. Feinberg * and Kara M. Reynolds ([dagger])

* American University and U.S. International Trade Commission, Department of Economics, Washington, DC 20016-8029 USA; E-mail [email protected]; corresponding author.

([dagger]) American University, Department of Economics, Washington, DC 20016-8029 USA; E-mail [email protected].

Earlier versions of this paper were presented at the U.S. International Trade Commission and American University, and we thank seminar participants for their comments. We particularly thank two anonymous referees and Chad Bown for helpful suggestions and Alan Fox and Raul Tortes for assistance in obtaining data. All views expressed (and any errors or omissions) are those of The authors, and do not represent the views of the U.S. International Trade Commission or any individual Commissioners.

Received November 2004; accepted June 2005.

Table 1. Antidumping Cases Filed by 15 Leading Users, 1995-2003

India 379
United States 329
European Union 274
Argentina 180
South Africa 166
Australia 163
Canada 122
Brazil 109
Mexico 73
China 72
Turkey 61
Korea 59
Indonesia 54
Peru 48
New Zealand 42

Source: Miranda, Torres, and Ruiz (1998) and
World Trade Organization (2005).

Table 2. Marginal Effects of Types of Retaliation on Probability of
Filing a Petition

Variable (1) (2)

CAT 0.001087 * 0.000832 *
 (0.000323) (0.000267)
CAT*METALS
OTHER 0.001483 * 0.000945 *
 (0.000222) (0.000165)
OTHER*METALS
DETER 0.0000003 * 0.0000002 *
IMPORTS (4.10 x [10.sup.-8]) (3.68 x [10.sup.-8])
 (in billions) 0.000101 * 0.000089 *
 (0.000012) (0.000012)
DEFLECT 0.000002 * 0.000002 *
 (0.000001) (0.000001)
EXCHANGE 0.000299 * 0.000297 *
 (0.000064) (0.000062)
TRADITIONAL 0.000703 *
 (0.000122)
CHINA

INDIA

INDONESIA

KOREA

Year effects Yes Yes
Category effects Yes Yes
Importer effects Yes Yes
Observed
 probability 0.0037 0.0037
Observations 431,680 431,680

Variable (3) (4)

CAT 0.000235 * 0.001874 *
 (0.000105) (0.000544)
CAT*METALS -0.000352 *
 (0.000041)
OTHER 0.000230 * 0.000793 *
 (0.000059) (0.000156)
OTHER*METALS 0.000362 *
 (0.000220)
DETER 0.0000001 * 0.0000002 *
IMPORTS (2.08 x [10.sup.-8]) (3.72 x [10.sup.-8])
 (in billions) 0.000042 * 0.000089 *
 (0.000006) (0.000012)
DEFLECT 0.000001 * 0.000002 *
 (5.61 x [10.sup.-7]) (0.000001)
EXCHANGE 0.000183 * 0.000296 *
 (0.000037) (0.000062)
TRADITIONAL 0.001070 * 0.000749 *
 (0.000156) (0.000081)
CHINA 0.012974 *
 (0.001468)
INDIA 0.002205 *
 (0.000454)
INDONESIA 0.002454 *
 (0.000475)
KOREA 0.004642 *
 (0.000732)
Year effects Yes Yes
Category effects Yes Yes
Importer effects Yes Yes
Observed
 probability 0.0037 0.0037
Observations 431,680 431,680

Standard errors are in parentheses. * denotes those marginal effects
significant at the 1% level.

Table 3. Marginal Effects of Strongest Category and Importing Country
Fixed Effectsa

 Marginal Effect
Industry Categories
 Base metals and articles thereof (XV) 0.00515 * (0.00124)
 Chemical products 0.00283 * (0.00065)
 Plastics, rubber, and articles thereof (VII) 0.00195 * (0.00047)
 Textiles (XI) 0.00099 * 0.00029)
 Machinery and mechanical appliances (XVI) 0.00077 * (0.00025)
Importing countries
 India 0.00688 * (0.00175)
 United States 0.00394 * (0.00113)
 European Community 0.00387 * (0.00110)
 Australia 0.00367 * (0.00107)
 Argentina 0.00332 * (0.00099)
Observed probability 0.0037
Observations 431,680

Standard errors are in parentheses. * denotes those marginal effects
significant at the 1% level. (a) Selected fixed effect estimates
associated with the first specification presented in Table 2.

Table 4. Marginal Effect of Any Retaliation on probability of Filing a
Petition

Variable (1) (2)

RETALIATION 0.001778 * 0.001135 *
 (0.000245) (0.000183)
RETALIATION *
 METALS
DETER 0.0000002 * 0.0000002 *
IMPORTS (4.02 x [10.sup.-8]) (3.62 x [10.sup.-8])
 (in billions) 0.000099 * 0.000087 *
 (0.000012) (0.000001)
DEFLECT 0.000002 * 0.000002 *
 (1.04 x [10.sup.-6]) (9.91 x [10.sup.7])
EXCHANGE 0.000296 * 0.000293 *
 (0.000062) (0.000061)
TRADITIONAL 0.000673 *
CHINA
INDIA
INDONESIA
KOREA
Year effects Yes Yes
Category effects Yes Yes
Importer effects Yes Yes
Observed
 probability 0.0037 0.0037
Observations 431,680 431,680

Variable (3) (4)

RETALIATION 0.000271 * 0.001143 *
 (0.000064) (0.000193)
RETALIATION * -0.000015
 METALS (0.000096)
DETER 0.0000001 * 0.0000002 *
IMPORTS (2.06 x [10.sup.-8]) (3.63 x [10.sup.-8])
 (in billions) 0.000042 * 0.000088 *
 (0.000006) (0.000012)
DEFLECT 0.000001 * 0.000002 *
 (5.55 x [10.sup.-7]) (9.90 x [10.sup.7])
EXCHANGE 0.000182 * 0.000293 *
 (0.000037) (0.000061)
TRADITIONAL 0.001050 * 0.000672 *
 (0.000155) (0.000119)
CHINA 0.012914 *
 (0.001463)
INDIA 0.002242 *
 (0.000458)
INDONESIA 0.002394 *
 (0.000466)
KOREA 0.004587 *
 (0.000725)
Year effects Yes Yes
Category effects Yes Yes
Importer effects Yes Yes
Observed
 probability 0.0037 0.0037
Observations 431,680 431,680

Standard errors are in parentheses. * denotes those marginal effects
significant at the 1% level.

Table 5. Marginal Effect of Retaliation on Probability of Filing a
Petition: SubSamples

Variable (1) (2) (3)

Importing country Traditional New Traditional
Exporting country All All Traditional
RETALIATION 0.0057046 * 0.0009395 * 0.010575
 (0.001276) (0.000194) (0.006191)
DETER 0.0000007 * 0.0000004 * 0.000005 *
IMPORTS (0.000000) (7.44 x [10.sup.-8]) (0.000001)
 (in billions) 0.0005110 * 0.000290 * 0.000938 *
 (0.000074) (0.000040) (0.000429)
DEFLECT 0.0000375 * 4.34 x [10.sup.-7] 0.000176
 (0.000012) (1.03 x [10.sup.6]) (0.000133)
EXCHANGE 0.0010292 0.000299 * 0.030361
 (0.000775) (0.000062) (0.016159)
TRADITIONAL 0.0028187 * 0.000429 *
 (0.001034) (0.000104)
Year effects Yes Yes Yes
Category effects Yes Yes Yes
Importer effects Yes Yes Yes
Observed
 probability 0.0111 0.00282 0.0319
Observations 53,960 357,840 2,720

Variable (4)

Importing country New
Exporting country Traditional
RETALIATION 0.003234 *
 (0.001070)
DETER 0.000004 *
IMPORTS (8.18 x [10.sup.-7])
 (in billions) 0.001790 *
 (0.000412)
DEFLECT 0.000044 *
 (0.000019)
EXCHANGE 0.004825 *
 (0.001521)
TRADITIONAL

Year effects Yes
Category effects Yes
Importer effects Yes
Observed
 probability 0.0148
Observations 15,360

Standard errors are in parentheses. * denotes those marginal effects
significant at the 1% level.

Table 6. Marginal Effect of Retaliation on Probability of Filing a
Petition (a)

Variable (1) (2)

RETALIATION 0.003826 * 0.003513 *
RETALIATION * (0.000581) (0.000575)
 METALS

DETER 0.000001 * 0.000001 *
BILATERAL (2.59 x [10.sup.-7]) (2.64 x [10.sup.7])
 IMPORTS
 (in billions) 0.000529 * 0.000519 *
 (0.000056) (0.000056)
DEFLECT 0.000017 * 0.000017 *
 (0.000006) (0.000006)
EXCHANGE 0.002212 * 0.002223 *
 (0.000403) (0.000404)
TRADITIONAL 0.000584 *
 (0.000320)
CHINA
INDIA
INDONESIA
KOREA
Year effects Yes Yes
Category effects Yes Yes
Importer Yes Yes
 effects
Observed
 probability 0.0111 0.0111
Observations 133,552 133,552

Variable (3) (4)

RETALIATION 0.001359 * 0.003556 *
RETALIATION * (0.000331) (0.000621)
 METALS -0.000115
 (0.000579)
DETER 0.0000008 * 0.000001 *
BILATERAL (0.0000002) (2.64 x [10.sup.7])
 IMPORTS
 (in billions) 0.000321 * 0.000518 *
 (0.000040) (0.000056)
DEFLECT 0.000012 * 0.000017 *
 (0.000004) (0.000006)
EXCHANGE 0.001704 * 0.002221 *
 (0.000303) (0.000404)
TRADITIONAL 0.002401 * 0.000584 *
 (0.000421) (0.000320)
CHINA 0.029427 *
 (0.002919)
INDIA 0.007224 *
 (0.001431)
INDONESIA 0.006485 *
 (0.001301)
KOREA 0.012604 *
 (0.001848)
Year effects Yes Yes
Category effects Yes Yes
Importer Yes Yes
 effects
Observed
 probability 0.0111 0.0111
Observations 133,552 133,552

Standard errors are in parentheses. * denotes those marginal effects
significant at the 1% level. Sample excludes those observations in
which imports from the potential target were zero in 1998.