Economics and the International Trade Commission*

Transcription

1 Economics and the International Trade Commission* JAMES M. DEVAULT Lafayette College Easton, Pennsylvania I. Introduction U.S. trade policy over the last two decades has become increasingly reliant on the U.S. unfair trade laws. These laws have been used primarily to punish foreign firms whose exports to the U.S. are dumped or subsidized. Since the passage of the Trade Agreements Act of 1979, some 950 cases have been completed under the U.S. antidumping and antisubsidy laws, resulting in the imposition of more than 300 punitive duties. The U.S. antidumping and antisubsidy laws have also been the spawning ground for many other trade restrictions, including the steel VRAs negotiated during the 1980s and the U.S. semiconductor agreement with Japan. Because the U.S. antidumping and antisubsidy laws are now one of the most important means used to restrict imports, the administration of these laws has come under considerable scrutiny. Many observers are concerned that these laws are being administered in a way which favors U.S. industries. In this paper, I attempt to address some of these concerns by determining the extent to which the administration of the U.S. unfair trade laws is based on the economic criteria specified in the unfair trade legislation. In particular, this study examines the importance of economic criteria in the antidumping and antisubsidy decisions of the International Trade Commission (ITC). The ITC is one of the most important agencies responsible for administering the U.S. unfair trade laws and its role extends beyond these laws into other areas, including the escape clause. The administrative role of the ITC in unfair trade cases is to determine whether unfairly traded imports are a cause of material injury to a domestic industry. Many other authors have examined the basis for ITC decisions. The results of Baldwin [3] and Goldstein and Lenway [8] provide support for the hypothesis that the ITC uses at least some of the legislated economic criteria appropriately. Both studies examine the escape clause decisions of the ITC and find that industries with declining profitability are more likely to receive favorable ITC decisions. Goldstein and Lenway [8] also find that the ITC is immune to congressional pressures for protection. Both studies, however, focus solely on escape clause decisions, which represent only a small fraction of ITC decisions and do not involve allegations of unfair trade. Escape clause actions are also different from other ITC decisions in that the president must ultimately decide whether relief is warranted. Hansen [9], Prusa [16], and Finger, Hall and Nelson, [7] examine ITC decisions involving *I would like to thank James Hartigan, Elizabeth DeVault, and several anonymous referees for their helpful comments. All remaining errors are mine. 463

2 464 James M. DeVault unfair imports and find that economic criteria play a minor role in these decisions. This is not very surprising, however, since all three studies are based on aggregated economic data not actually employed in ITC decisions. These studies also provide support for the argument that ITC decisions are political. Hansen's results indicate that ITC decisions are influenced by pressure from influential members of Congress. Both Hansen [9] and Finger, Hall and Nelson [7] also find that ITC decisions are positively related to the size of the domestic industry, so that larger industries are more likely to obtain relief through the unfair trade laws. The studies which are based on the economic data available to the ITC find that some economic criteria are important in ITC unfair trade decisions. Herander and Schwartz [11] show that the probability of an affirmative ITC antidumping decision increases as the dumping margin rises and as employment falls. The results of Baldwin and Steagall [4] indicate that declining domestic shipments are a robust predictor of affirmative ITC antidumping and antisubsidy decisions. Moore [14], in what is probably the most thorough analysis of ITC decisions, finds that higher profit rates and increasing production and profitability decrease the probability of an affirmative ITC decision. All of these authors also find that non-economic factors influence ITC decisions. The most striking results here are obtained by Moore [14], who finds that ITC decisions are influenced by congressional pressure. Herander and Schwartz [11] find that ITC decisions are inversely related to the number of firms in the petitioning industry, suggesting that greater industry concentration makes it easier to present a more effective case. Both Moore [14] and Baldwin and Steagall [4] find that increases in overall imports or overall import penetration increase the likelihood of affirmative ITC decisions. What is to be made of all these results? The results indicate that economic criteria play at least a limited role in ITC decisions. Thus it appears that industries with low and declining profitability and declining shipments or employment are more likely to obtain relief. Two aspects of the results for the economic criteria are troubling, however. First, the studies identify different economic criteria as important. Second, the studies find little connection between the injury which is suffered and the imports supposedly responsible for this injury.' Indeed, the results show that increases in overall imports or overall import penetration are far better predictors of affirmative ITC decisions than increases in unfair imports or unfair import penetration. If valid, these findings reduce the U.S. unfair trade laws to a substitute for the escape clause, a practice which is clearly at odds with the GATT. Previous studies also indicate that non-economic criteria have a significant impact on ITC unfair trade decisions. The results of both Moore [14] and Hansen [9] support the conclusion that certain members of Congress can influence ITC decisions. The findings of Hansen [9] and Finger et al. [7] that industry size influences ITC decisions are also disturbing because they suggest that unfair trade cases are not treated equally, a conclusion which finds additional support in the results of Herander and Schwartz [11]. In summary, the available evidence indicates that both economic and non-economic criteria play a role in ITC decisions, implying that the ITC does not administer the unfair trade laws in a purely objective or apolitical fashion as it was designed to do. This study attempts to clarify 1. Moore [14] does find some evidence that increases in unfair import are positively related to ITC decisions, but this evidence is tenuous. His results show that increases in unfair import volumes are important in the preliminary decisions of the ITC but not in the final decisions. Baldwin and Steagall [4] find that increasing unfair imports are significant in some specifications but are not robust predictors of ITC decisions.

3 ECONOMICS AND THE INTERNATIONAL TRADE COMMISSION 465 the role of both economic and non-economic criteria in ITC decisions by carefully modeling the methods used by the ITC to evaluate economic criteria. Previous studies have ignored these methods, and as a result, they have not fully identified the role of economic and non-economic criteria in the decisions of the ITC. The remainder of the paper is comprised of six sections. The second section provides an overview of the ITC and its role in implementing the U.S. antidumping and antisubsidy laws. The third section describes the methodology used by the ITC in its antidumping and antisubsidy decisions. The fourth and fifth sections describe the data used in the study and the econometric model. The final two sections contain the results and conclusions of the paper. II. The ITC and the U.S. Antidumping and Antisubsidy Laws The ITC is an independent, quasi-judicial government agency which has several other functions in addition to its role as administrator of the U.S. unfair trade laws. Some of these functions include monitoring import levels, analyzing trade policy issues and providing recommendations to the president on certain trade policy questions. The ITC consists of 6 commissioners and a support staff of economists, lawyers and other personnel. The commissioners are appointed by the president and approved by Congress. Under normal circumstances, each commissioner serves one nine-year term which cannot be renewed. The commissioners are responsible for the antidumping and antisubsidy decisions of the ITC as well as for ITC decisions involving a number of other trade matters. Each ITC decision is based on a simple vote with the majority opinion prevailing.2 The role of the ITC in U.S. antidumping and antisubsidy investigations is to determine whether dumped or subsidized imports materially injure or threaten to materially injure a U.S. industry.3 Each antidumping or antisubsidy case involves a preliminary ruling by the ITC on whether there is a reasonable indication that dumped or subsidized imports have caused or threaten to cause material injury. If the ITC's preliminary ruling is negative, the case is dismissed; if the ruling is affirmative, the case continues. During the sample period considered here, approximately 15% of the ITC's preliminary decisions were negative. If the ITC makes an affirmative preliminary decision, the case is then passed over to the International Trade Administration (ITA), an agency of the Department of Commerce. The role of the ITA is to determine whether dumping or subsidization has occurred, and if so, to what extent. If the ITA finds that dumping or subsidization has taken place, then the ITC must render a final decision on the case. If this decision is affirmative, antidumping or countervailing duties will be levied. If the ITC's final decision is negative, the case is dismissed. The economic criteria upon which ITC decisions are supposed to be based vary depending on the nature of the injury facing the domestic industry. Different criteria exist for the case in which the domestic industry is experiencing material injury and for the case in which the domestic industry is threatened with material injury. In cases involving actual material injury, the commission is directed by law to consider changes in the following variables: output, sales, market share, profits, productivity, capacity utilization, investment, return on investment, prices, cash 2. In the case of a tie, the affirmative opinion prevails. 3. The U.S. antidumping and antisubsidy laws also provide relief to domestic industries whose establishment is materially retarded by dumped or subsidized imports. In practice, cases involving material retardation are rare, and as a result, material retardation is not considered further in this paper.

4 466 James M. DeVault flow, inventories, employment, wages, growth, and the industry's ability to raise capital. Also to be considered are changes in the volume and value of the unfair imports, the market penetration of the unfair imports, and the price of these imports. In addition, the law also allows each commissioner to examine any other "relevant economic factors." In cases involving the threat of material injury, the criteria set out by the unfair trade laws include, among other things, the nature of the subsidy (if a subsidy is involved), the potential for increases in foreign productive capacity, any unused foreign capacity, the likelihood of foreign product line shifts, and any substantial increase in importer's inventories. III. The Bifurcated Approach Although ITC commissioners analyze economic criteria using several different methods, the dominant approach over the last decade has been the bifurcated approach. Bifurcated decisions accounted for about 80% of the decisions reached during the sample period, and because of this, the bifurcated approach is the only approach considered further in this paper.4 The bifurcated approach involves a two stage decision making process which is illustrated in Figure 1. In the first stage, commissioners determine whether a domestic industry has been materially injured. In the second stage, commissioners determine whether unfair imports are a cause of material injury. If both decisions are affirmative, duties are levied. If there is no material injury to the domestic industry or if the unfair imports are not viewed as a cause of material injury, it must then be determined whether the unfair imports threaten material injury. If so, duties are levied; if not, the case is dismissed. This paper differs from all previous work in that it models bifurcation. To see the importance of modeling bifurcation, consider the following two cases. In the first case, suppose there is material injury to a domestic industry but the dumped or subsidized imports are not found to be a cause of that injury and are not found to threaten additional injury (so the ultimate decision in the case is negative.) If bifurcation is not modeled, the data on the status of the domestic industry (which indicate material injury) will seem to conflict with the final decision (which is negative). Thus unless bifurcation is modeled, it will appear that the economic criteria are being misused. In the second case, suppose a petition is dismissed because the domestic industry is not materially injured. Under these circumstances, no decision on whether the unfair imports are a cause of material injury is necessary since there is no material injury. Since there is no causation decision, no statement can be made about the variables which enter into this decision. If bifurcation is not accounted for, however, the variables which relate to causation are treated as if they affected the final negative decision. Both of these cases demonstrate how ignoring bifurcation can produce misleading results. This study also differs from other studies in that affirmative ITC decisions based on the threat 4. The most frequently used alternative to bifurcation is the unified approach, which attempts to determine what the condition of the domestic industry would have been in the absence of dumping or subsidization. Decisions are then based on a comparison of the current state of the domestic industry with its hypothetical state in the absence of dumping or subsidization. Thus an industry could.be healthy and still receive relief if the dumped or subsidized imports have made it substantially less healthy. Commissioners who use the unified approach reject the "trend analysis" which is standard under bifurcation and rely instead on computer simulations and/or elasticity estimates to reach their decisions. Since neither the results of the simulations nor the elasticity estimates are provided consistently in the ITC reports and since the number of individual votes involving the unified approach is small, it is difficult to empirically examine unified decisions. For more information on the unified approach, see Kaplan [12].

5 ECONOMICS AND THE INTERNATIONAL TRADE COMMISSION 467 Material Injury to the Domestic Industry No Yes Threat of Material Injury by Material Injury Unfair Imports No Yes No Yes Case Terminated Duties Levied Threat of Duties Material Injury Levied No Yes Figure 1. The Bifurcated Approach Case Terminated Duties Levied of material injury are separated from those based on actual material injury. This separation is necessary because the economic criteria for the two decisions are quite different. Previous studies have treated all affirmative ITC decisions as if they involved actual material injury, and as a result, they have probably understated the role of economic criteria. In this work, affirmative ITC decisions based on the threat of material injury have been excluded. There are two reasons for this: first, the threat criteria are not reported in most cases; second, only about 15% of the affirmative ITC decisions in the sample involved a threat of material injury. IV. The Data All the economic data employed in the paper are taken from the final antidumping and antisubsidy reports of the ITC. The economic data from the ITC reports are used because they form the legal basis for ITC decisions. The sample includes only the final decisions of the ITC for two reasons. First, Moore [14] finds evidence that the ITC's final decisions are "more political", which he argues is to be expected since the industries involved have already met a minimum injury standard at the preliminary level. Second, Baldwin and Steagall [4] find evidence that the injury standard in final decisions is weaker than in preliminary decisions. The evidence provided by both studies suggests that the final decisions of the ITC are less rigorous than the preliminary decisions and are thus more likely to be influenced by non-economic criteria. The sample period runs from March 1985 through June This period was selected because it was in the spring of 1985 that bifurcation was formally recognized by the ITC commissioners. While many bifurcated decisions were offered in cases prior to March 1985, it was only after this period that the preferences of the individual ITC commissioners became distinguishable. A major weakness of the ITC data is that it is kept confidential in cases where its disclosure

6 468 James M. DeVault would involve the release of proprietary information. Typically, the ITC data is kept confidential in cases involving a single or small number of domestic or foreign firms. As a result of this confidentiality requirement, a substantial number of cases could not be included in the econometric analysis.5 In addition, a significant number of cases had to be dropped because they involved affirmative decisions based on the threat of material injury. The final sample included 63% of the cases completed between March 1985 and June This study also differs from previous work in that it controls for multi-petition cases involving imports of the same product from more than one country. These cases are commonplace and have involved imports from as many as ten different countries simultaneously. The problem with multi-petition cases is that the economic data in the individual petitions are roughly the same so that the decisions are not independent. Thus the data for material injury decisions are identical regardless of how many countries are involved in a petition because they concern only the status of the domestic industry. The data used in causation decisions are also similar since unfair imports from different countries are usually cumulated before their impact is addressed.7 The result of the shared data is that in almost all instances, duties are levied against all the countries involved in a particular petition or against none.8 Because of this, I have elected to collapse each multi-petition case into a single case which requires each commissioner to make only one decision that covers all the countries involved.9 Not collapsing multi-petition cases leads to a situation in which two or three large cases dominate the sample.'0 Since the study attempts to explain the voting patterns of ITC commissioners, the dependent variables represent the votes of individual commissioners. Because the paper examines only bifurcated decisions, there are potentially two dependent variables in each case (one for the material injury decision and one for the causation decision). Define MVOTE to be the vote of an individual commissioner on material injury (1 - affirmative, 0- negative) and CVOTE to be the vote of an individual commissioner on causation (1 - affirmative, 0 - negative). 5. It has been argued that ITC decisions are more likely to be political in cases where the economic data are confidential. To test this argument, I compared the percentage of affirmative decisions from the sample with economic data with this same percentage from a separate sample in which no economic data were publicly available. The new sample consisted of 78 observations drawn from the same sample period. The difference between the two means was not statistically significant. I also computed the same percentages from each sample conditional on some political representation, and again found no statistically significant difference. Finally, I ran a logit regression in the sample with no economic data using individual commissioner votes as the dependent variable and political variables and fixed effects as independent variables. None of the political variables were significant. While these results are far from conclusive, they suggest that ITC decisions are not more political in cases where the economic data are confidential. 6. Baldwin and Steagall [4] find evidence that the ITC uses different economic criteria in antidumping and antisubsidy cases. Unfortunately, the number of antisubsidy cases in my sample is too small for me to conduct a separate analysis of the two different types of cases. 7. Cumulation has been virtually mandatory since In only one multi-petition case in the sample did commissioners vote affirmatively for one country and negatively for another. In this case, imports from the two countries were sufficiently different so that cumulation was not mandatory. I treat these two cases as distinct in the sample because of this. It should be noted that commissioners have often split their votes on multi-petition cases when a threat of material injury is posed by one or more countries but not by others. As was discussed earlier, however, threat rulings have been excluded from the sample so this is not an issue. 9. Some differences do exist in the data used for individual countries in multi-petition cases. Most notably, the dumping or subsidy margins frequently differ significantly. To control for these differences, I take a weighted average of these margins, using the fraction of total unfair imports as weights. 10. Consider, for example, a recent case involving fresh cut flowers. In this case, which involved imports from ten different countries, three commissioners found seven different "like" products, each of which required separate votes. If all of these votes were included in the sample, this one petition would have accounted for 114 separate commissioner votes. Using my approach, this number is reduced to 23.

7 ECONOMICS AND THE INTERNATIONAL TRADE COMMISSION 469 Table I. Summary of Independent Variables Material Injury Criteria: DPR SRDPR LRDPR DMS SRDMS LRDMS SRPRO LRPRO Causation Criteria: FMS SRFMS LRFMS MARGIN SRIM LRIM Non-Economic Criteria: VAL SN HN FAIRMS SRFAIR LRFAIR - Domestic profit rate. - One year percentage change in domestic profit rate. - Two year percentage change in domestic profit rate. - Domestic market share. - One year percentage change in domestic market share. - Two year percentage change in domestic market share. - One year percentage change in domestic production - Two year percentage change in domestic production - Market share of unfair imports. - One year percentage change in market share of unfair imports. - Two year percentage change in market share of unfair imports. Dumping or subsidy margin. - One year percentage change in volume of unfair imports. - Two year percentage change in volume of unfair imports. Value of domestic shipments in the most recent year. Number of active firms with production facilities located in states with senators on the Senate International Trade Subcommittee. Number of active firms with production facilities located in districts with representatives on the House Trade Subcommittee. Market share of fairly traded imports. One year change in the market share of fairly traded imports. Two year change in the market share of fairly traded imports. The independent variables used in the analysis are listed in table I and are described in greater detail in the appendix. The independent variables which reflect the economic criteria employed by the ITC commissioners include percentage changes in profit rates, domestic market share, and production, as well as percentage changes in the volume and market share of the unfair imports." Also included among the economic criteria are a number of variables not specifically mentioned in the legislation. These variables include the domestic industry's profit rate and market share, the dumping or subsidy margin, and the market share of unfair imports. These variables, while not cited in the legislation, are frequently mentioned in the opinions of a number of commissioners and one would be hard-pressed to argue that they are not "relevant economic factors" and thus consistent with the legislation. The non-economic criteria included in the study were selected to test three hypotheses which have found support in previous work. The first hypothesis is that industry size affects the amount of political pressure which can be applied to regulatory agencies. The results of Moore [14] and Herander and Schwartz [11] support the pressure group argument of Pincus [15], which is that smaller, more concentrated industries are better able to overcome coordination problems and hence can lobby more effectively. Hansen [9], Finger, Hall, and Nelson [7], and Baldwin and Steagall [4] find support for the argument of Caves [5], which is that larger industries with greater 11. Changes in domestic production and employment were highly correlated in the sample, giving rise to multicollinearity problems. In order to avoid these problems, I include only changes in domestic production.

8 470 James M. DeVault political and economic power are able to generate more pressure for relief. In order to test the hypothesis that size matters, I use the value of domestic shipments as a proxy for size. The second hypothesis tested here is provided by Weingast and Moran [18] and is referred to as "congressional dominance." This hypothesis argues that congressional oversight committees use their control over regulatory agencies to favorably influence decisions involving their constituents. In the case of the ITC, the two congressional bodies which have the most influence are the Senate and House Trade Subcommittees. Both subcommittees are involved in setting the budget for the ITC, and the Senate Trade Subcommittee plays a crucial role in the appointment of new ITC commissioners. Moore [14] and Hansen [9] both find evidence that members of these subcommittees are able to influence ITC decisions involving their constituents. To test this theory, I construct two variables. The first, SN, represents the number of "active" firms with production facilities located in states with senators on the Senate Trade Subcommittee while the second, HN, represents the number of active firms with production facilities located in districts with representatives on the House Trade Subcommittee. If congressional dominance exists, increases in both variables should raise the probability of affirmative ITC decisions. A firm is considered active in an investigation if it is a petitioner, if it serves as a witness for the domestic industry in an ITC hearing, or if it is listed as a supporter of the petition in the ITC final report. The two variables used here to test congressional dominance (SN and HN) differ from those in previous studies, which were based on the number of domestic firms represented regardless of whether these firms were active in an investigation. There are several reasons for adopting this new definition. First, it is the active firms which are most vocal and hence most visible to their political representatives and the ITC commissioners. Second, in many cases, some domestic firms oppose the investigation. Third, including all represented domestic firms may produce a downward bias because previous studies have shown that the greater the number of "inactive" firms in an industry, the less likely it is that the ITC will grant relief.12 For more information on the construction of SN and HN, see the appendix. The third hypothesis I test is that commissioners are more likely to provide relief to industries which face overall increases in imports rather than just increases in unfair imports. If correct, this hypothesis reduces antidumping and antisubsidy cases to substitutes for the escape clause. Such a reduction clearly violates the intent of the unfair trade laws. This hypothesis is strongly supported by the results of Moore [14], who finds that changes in overall import volumes are highly significant and positively related to the probability of an affirmative final ITC determination. The results of Baldwin and Steagall [4] also confirm the hypothesis that changes in overall import penetration are crucial to the ITC rather than changes in unfair import penetration. To test this hypothesis, I include the market share of fairly traded imports as well as the one and two year changes in this market share. To control for the fact that the data set pools the votes of different commissioners, commissioner-specific fixed effects are introduced. Discrepancies between the injury thresholds of different commissioners have been well documented in previous work, and these discrepancies are captured here by fixed effects. Commissioners Brunsdale, Eckes, Liebeler, Lodwick, Rohr, and Newquist are represented respectively by the fixed effects BRU, ECK, LIE, LOD, NEW, and ROH. 12. See Herander and Pupp [10] and Herander and Schwartz [11]. Preliminary results using all firms rather than just active firms support this argument. In particular, the coefficient on the "all firm" version of SN was found to be negative at very high significance levels for several specifications of the model.

9 ECONOMICS AND THE INTERNATIONAL TRADE COMMISSION 471 V. Econometric Modeling Because the dependent variable are qualitative and because bifurcated decisions involve two stages, a sequential logit model was selected for the analysis. A sequential logit model is appropriate only if the decisions concerning material injury and causation are independent. This appears to be the case since material injury decisions seem to be based on data for the domestic industry while causation decisions seem to be based on data for the unfair imports. Let Xij represent the (k x 1) vector of independent variables which summarizes the condition of the domestic industry in case i for commissioner j (including fixed effects) and let Zij represent the (1 x 1) vector of independent variables used to address the question of causation in case i for commissioner j (again including fixed effects). The probability of an affirmative material injury decision by commissioner j in case i (denoted AMIDij) is then given by P(AMIDijXij) = 1/(1 + e- x) (1) while the probability of an affirmative causation decision by commissioner j in case i (denoted ACDij) can be written as P(ACDijXij,Zij) = [1/(1 + e a Zij)lP(AMIDij). (2) Given the independence between the material injury and causation decisions, the (k x 1) and (1 x 1) parameter vectors /3 and a can be estimated by maximizing the likelihood functions of two successive dichotomous logit models (See Amemiya [2] or Maddala [13] for some examples). In the first case, represented by (1), the parameters relevant to the material injury decision are estimated using the entire data set. In the second case, represented by (2), the parameters relevant to the causation decision are estimated using the subset of cases in which material injury is present. VI. Results The econometric results are presented in Tables II through V. Tables II and III contain the results for the first half of the bifurcated decision (which relates to material injury) while Tables IV and V contain the results for the second half of the bifurcated decision (which relates to causation). Consider Table II first, which contains the logit coefficient estimates for five different specifications of the material injury decision. If the economic variables in the material injury decision are used appropriately, all the coefficients on these variables should be negative since all the economic variables are positively related to the health of the domestic industry and hence negatively related to the probability of an affirmative material injury decision. Increases in the number of active firms represented should raise the probability of an affirmative material injury decision if congressional dominance holds; hence the coefficients on SN and HN should be positive if this hypothesis is valid. Since there are conflicting theories about how industry size affects ITC decisions, the VAL coefficient does not have an expected sign. Examining the results for the economic criteria in Table II first, note that 18 of the 20 coefficients in models 2 through 5 have negative signs and that all the statistically significant coefficients have negative signs. The economic variables which are significant include the domestic indus-

10 472 James M. DeVault Table II. Results for the Material Injury Decision (standard errors in parentheses) Fixed Effects Model 1 Model 2 Model 3 Model 4 Model 5 BRU *** 2.00** 3.94*** 2.31* (0.46) (0.85) (0.85) (1.08) (1.06) ECK 1.46*** 4.31*** 3.73*** 5.86*** 4.21*** (0.35) (0.85) (0.85) (1.09) (1.06) LIE 1.39*** 3.44*** 2.93*** 4.78*** 3.16*** (0.42) (0.83) (0.83) (1.06) (1.07) LOD 1. 14*** 4.04*** 3.48*** 5.50** 3.90*** (0.29) (0.82) (0.80) (1.06) (1.00) NEW *** 2.90*** 4.44*** 2.82** (0.46) (0.99) (0.93) (1.21) (1.15) ROH 1. 10*** 3.71*** 3.17*** 5.03*** 3.42*** (0.30) (0.80) (0.78) (1.03) (1.00) Economic Factors DPR -0.21*** -0.20*** -0.30*** -0.26*** (0.04) (0.03) (0.05) (0.04) SRDPR (0.0012) (0.001) LRDPR (0.0004) (0.001) DMS -0.03*** -0.03*** -0.05k* 0.3* (0.01) (0.01) (0.01) (0.01) SRDMS -0.05* **- (0.02) (0.03) LRDMS * ** (0.02) (0.03) SRPRO -0.02*** ***- (0. 01) (0.01) LRPRO (0. 01) (0.01) Non-economic Factors SN ** (0.12) (0.1I1) HN (0.41) (0.38) VAL *** 0.002*** (0.0005) (0.0005) Summary Statistics Log likelihood % of correct predictions McFadden's R Observations (% Affirmative) (74.70) (74.70) (74.70) (74.70) (74.70) ****Significant at 10%, 5 %, and 1 % levels respectively.

11 ECONOMICS AND THE INTERNATIONAL TRADE COMMISSION 473 try's profit rate, its market share, and changes in its market share and its production. Changes in profitability are not significant. The explanatory power of the economic variables is quite high. A log likelihood ratio test of the hypothesis that the economic coefficients in models 2 or 3 are all zero is easily rejected in both cases at the 0.5% significance level. Furthermore, McFadden's R2, which provides a rough measure of fit, jumps from 0.02 to about 0.30, indicating that the inclusion of the economic criteria improves the fit of both models substantially. Finally, the percentage of correct predictions rises from 74.7% to about 83% after the economic criteria are included. Clearly economic criteria do influence material injury decisions. The results for the non-economic criteria (see models 4 and 5 in Table II) also support the hypothesis that non-economic criteria influence ITC material injury determinations. A log likelihood ratio test that all the coefficients on the non-economic factors are zero can be easily rejected at the 0.5% significance level in both cases. The most important non-economic factor is industry size, as measured by VAL. The coefficient on this variable is positive and highly significant, indicating that larger industries are more likely to receive affirmative material injury decisions. The coefficients on the variables used to test congressional dominance (SN and HN) either have the wrong sign or are insignificant, so ITC material injury decisions do not appear to be influenced by congressional pressure.13 To determine the relative importance of economic and non-economic factors in ITC material injury decisions, elasticity estimates are computed for the influential variables in Table II. The results, which are presented in Table III, indicate that the domestic industry's profit rate and its market share have the greatest effect on ITC material injury decisions. The elasticities associated with the other variables are relatively small, indicating that these variables, while significant, are of lesser importance in ITC material injury decisions. In summary, both economic and non-economic variables influence ITC material injury determinations. Two economic variables, the domestic industry's profit rate and its market share, appear to be the most important determinants of ITC material injury decisions. Other economic variables which are important include the rate of change in domestic market share and the rate of change in domestic production. All of these variables enter into the material injury decisions of the ITC in a way which is consistent with the unfair trade legislation. The only non-economic factor which is important is industry size, which is positively related to the probability of an affirmative material injury decision. No support for congressional dominance is found. Tables IV and V contain the results for the second stage of the bifurcated decision, which determines whether the unfair imports are a cause of material injury to the domestic industry. If the ITC uses economic criteria appropriately in its causation decisions, increases in the volume and market share of the unfair imports should increase the probability of an affirmative decision and hence the coefficients on these variables should be positive. Greater unfair import penetration and higher dumping or subsidy margins should also increase the likelihood of an affirmative decision and hence the coefficients on these variables should be positive as well. If the antidumping and antisubsidy laws are being used by the ITC as substitutes for the escape clause, then signifi- 13. This result is not a product of the way SN and HN are constructed. Using the same dummy variables as Moore [14], I find no evidence of congressional dominance in either the material injury or causation decisions. Using the same variables as Hansen [9], I find that her version of HN is significant and positive, but only in the material injury decision. The impact of HN on ITC material injury decisions (as measured by its elasticity) is, however, quite small. Furthermore, defining SN in the same way, I find that this variable is highly significant and negative in the material injury decision, a result that is completely contrary to congressional dominance.

12 474 James M. DeVault Table III. Elasticity Estimates for Statistically Significant Material Injury Variables in Table II Variable Mean Standard Deviation Maximum Minimum Model 4 DPR DMS SRDMS SRPRO SN VAL Model 5 DPR DMS LRDMS SN VAL Notes: The formula for the elasticities computed here is given by (IXijkl/Pij)(aPij /axijk) = 1k Xijk /(1 + ep x ) where Xijk gives the value of the kth independent variable in case i for commissionerj and Pij gives the probability of an affirmative material injury decision in case i for commissionerj. cant and increasing fair import penetration should raise the probability of an affirmative decision, so FAIRMS, SRFAIR, and LRFAIR should have positive coefficients if this hypothesis is correct. Examination of Table IV reveals that two of the economic variables have a significant impact on causation decisions. The first is the dumping or subsidy margin and the second is the market share of unfair imports. The signs of the coefficients on both of these variables are positive, implying that cases involving higher dumping or subsidy margins and greater unfair import penetration are more likely to be affirmed, a result which is consistent with the unfair trade legislation.'4 Neither changes in unfair import penetration nor changes in the volume of unfair imports appear to be important in ITC causation decisions. When the economic criteria alone are used to explain the causation decision (see models 2 and 3 in Table IV), they raise McFadden's R2 from 0.2 to 0.5 and thus significantly improve the fit of both equations. Adding the economic criteria also increases the percentage of correct predictions from 76% to 86%. Thus economic criteria appear to be important in ITC causation decisions. This conclusion is reinforced by the elasticity estimates provided in Table V. These estimates indicate that a one percent increase in the dumping margin or the market share of the unfair imports increases the probability of an affirmative causation decision by roughly one quarter of one percent. Non-economic criteria appear to have little impact on the causation decision (see models 4 and 5 in Table IV.) Using a log likelihood ratio test, the null hypothesis that all of the coefficients on the non-economic variables equal zero cannot be rejected at the 10% significance level. Perhaps the most surprising result in light of previous work is that fairly traded imports appear 14. The significance of the dumping or subsidy margin is quite surprising given the fact that several commissioners claim not to use margins in their decisions. This result holds even when the principle proponents of margins (Brunsdale and Liebeler) are eliminated from the sample. One explanation for this result is that the dumping margin serves as a proxy for the amount by which the unfair imports undersell domestic producers. The amount of underselling is repeatedly cited as important in the opinions of commissioners who do not favor the use of margins.

13 ECONOMICS AND THE INTERNATIONAL TRADE COMMISSION 475 Table IV. Results for the Causation Decision (standard errors in parentheses) Model 1 Model 2 Model 3 Model 4 Model 5 Fixed Effects BRU ECK LIE LOD NEW ROH Economic Factors MARGIN FMS SRFMS LRFMS SRIM LRIM Non-economic Factors SN HN VAL FAIRMS SRFAIR LRFAIR Summary Statistics (0.61) 1.82*** (0.44) -1.30*** (0.46) 1.05*** (0.32) 2.40** (1.04) 0.90*** (0.33) -3.31*** (1.07) 0.17 (0.65) -5.19** (1.13) -1.30** (0.58) -2.90** (1.43) *** (0.62) 0.04*** (0.014) 0.17*** (0.038) (0.0006) 2.5E-5 (0.0002) Log likelihood % of correct predictions McFadden's R Observations (% Affirmative) (67.70) (67.70) *, **, *** - Significant at 10%, 5% and 1% levels respectively *** (1.07) 0.11 (0.64) -5.16*** (1.13) -1.31** (0.57) -2.90** (1.43) *** (0.62) 0.04*** (0.013) 0.17*** (0.038) 3.6E-5 (0.0002) (0.0002) (67.70) -3.29** (1.37) 0.39 (1.03) -5.18** (1.39) (1.00) (1.82) (1.05) 0.05*** (0.015) 0.18*** (0.043) (0.0004) 8.8E-6 (0.0002) (0.15) 0.42 (0.44) (0.0003) (0.014) (0.008) (67.70) -3.14** (1.35) 0.57 (0.98) -4.99*** (1.35) (0.949) (1.84) (0.99) 0.05*** (0.015) 0.17*** (0.042) 6.1E-5 (0.0002) 5.5E-7 (0.0002) (0.15) 0.49 (0.45) (0.003) * (0.013) -3.4E-8 (3.1E-7) (67.70)

14 476 James M. DeVault Table V. Elasticity Estimates for Statistically Significant Causation Variables in Table IV Variable Mean Standard Deviation Maximum Minimum Model 4 MARGIN FMS Model 5 MARGIN FMS Notes: The formula for the elasticities computed here is given by (Zijk/Pij)(aPij/Zijk) = akzijk/(l + ea j) where Zijk gives the value of the kth independent variable in case i for commissioner j and Pij gives the probability of an affirmative causation decision in case i for commissionerj. to have little impact on the causation decision. Indeed, even when the coefficient on the market share of fairly traded imports is significant (and it is only marginally so), it has the wrong sign. This is in sharp contrast to the results of previous studies. Another interesting feature of the causation results is that VAL (the value of domestic shipments) is no longer significant. The coefficient on this variable is positive and highly significant in the material injury decision, indicating that larger industries are more likely to receive favorable material injury decisions. One explanation for the loss of significance is that industries which make it through the material injury stage have the ability to lobby the ITC effectively. Since larger industries have greater access to resources, they will in general be able to lobby the ITC more effectively and hence are more likely to receive affirmative material injury decisions. Larger industries do not appear to have an advantage in the causation decision, however, because all industries (large or small) that make it past the material injury decision have demonstrated an ability to lobby the ITC effectively. This result is in keeping with the results of Herander and Pupp [10], who find that the ability to lobby the ITC effectively is important in determining who receives relief in steel unfair trade cases. It also suggests that some smaller industries may have been denied relief because they lacked the resources to adequately present their complaints. In summary, the evidence presented here indicates that at least some of the economic criteria are used appropriately in the causation decisions of the ITC commissioners. The two variables which are most significant are the market share of unfair imports and the dumping or subsidy margin. Perhaps equally important, non-economic criteria do not appear to influence ITC commissioners at this stage of the decision-making process. VII. Conclusion This study attempts to determine the importance of economic and non-economic criteria in the antidumping and antisubsidy decisions of the ITC. The study is novel in that it models the methods used by the ITC commissioners to reach their decisions. These methods have been ignored in previous work. The econometric results strongly support the hypothesis that economic criteria are used appropriately by the ITC commissioners. Key explanatory variables include the profit rate and market share of the domestic industry, the change in domestic production, the dumping or subsidy

15 ECONOMICS AND THE INTERNATIONAL TRADE COMMISSION 477 margin, and the market penetration of unfair imports. One of the most important findings of the paper is that domestic firms involved in unfair trade cases must demonstrate both material injury and a causal link between this injury and the unfair imports. Previous studies have not demonstrated a clear link between the material injury suffered by the domestic industry and the unfair imports. The study finds some evidence that non-economic criteria influence ITC decisions. In particular, it appears that ITC decisions are positively influenced by the size of the domestic industry. This may be because larger industries have the resources to present more effective cases to the ITC. No support is found for the hypothesis that ITC decisions are influenced by congressional pressure. The study also finds that fairly traded imports have little impact on ITC decisions. Thus there is no evidence to support the contention that antidumping or antisubsidy cases are being used as a substitute for the escape clause. Appendix: Description of the Data The economic data for this paper were drawn entirely from the final antidumping and antisubsidy reports of the ITC. In what follows, I briefly describe the data and some of the problems associated with it. Before I discuss the variables themselves, I provide some background on the commissioners who served on the ITC during the sample period. During the period under investigation (March 1985 through June 1992), 11 different commissioners served on the ITC. Two of the commissioners (Lodwick and Rohr) served on the commission over the entire period. Three commissioners (Stem, Liebeler and Eckes) began the period on the commission but eventually left. Commissioner Ster left in February 1987, Commissioner Liebeler left in December 1988, and Commissioner Eckes left in June During the observation period, six new commissioners (Brunsdale, Cass, Newquist, Watson, Crawford, and Nuzum) were appointed. Commissioner Brunsdale was appointed in January 1986, Commissioner Cass was appointed in January 1988, and Commissioner Newquist was appointed in October Commissioner Cass resigned in June Commissioners Watson, Crawford, and Nuzum were all appointed in December During the sample period, Commissioners Eckes, Lodwick, Newquist, and Rohr consistently rendered bifurcated opinions while Commissioners Stem and Cass consistently rendered unified opinions (see footnote 5 for more information on the unified approach). Commissioners Liebeler and Brunsdale switched their approach from bifurcated to unified in May I have been unable to determine the reasons for this switch, but the addition of Commissioner Cass to the ITC in January 1988 probably played an important role. As of this writing, it appears that Commissioners Crawford, Nuzum, and Watson have all adopted the unified approach. The dependent variables were all taken from the opinions of the commissioners, which are included in the final report. As was mentioned in the text, the dependent variables represent the votes of the commissioners on issues pertaining only to actual material injury caused by the dumped or subsidized imports and not to any threat of material injury. While considerable care was exercised in collecting the votes, some ambiguities did arise and these should be mentioned. In a number of cases, commissioners offered bifurcated opinions in which they ruled negatively on both material injury and causation. These rulings appear to be inconsistent with bifurcation because the absence of material injury to the domestic industry makes any ruling on causation unnecessary. This inconsistency is cleared up by noting that the causation analysis is included for the sake of argument or for the sake of completeness. The commissioners themselves state repeatedly that the causation analysis is offered "assuming arguendo" that the domestic industry is materially injured. In some cases, commissioners who normally used bifurcation were unable to come to any conclusion concerning the status of the domestic industry. In these cases, their votes were dropped from the sample. Finally, it is not necessarily true that each commissioner cast only one vote (or one sequence of votes) in each case or that each commissioner casts the same number of votes in each case. In cases in which a commissioner finds more than one "like product", that commissioner must vote on each of the like products. As an example, consider a case involving flowers from Colombia. In that case, Commissioners Eckes,

16 478 James M. DeVault Lodwick and Rohr found that there were seven like products, and thus each cast seven votes, one for each like product. Commissioners Brunsdale and Liebeler found only one like product and thus voted only once on this product. Consider next the independent variables. Domestic profit rates were obtained by taking the ratio of operating income to net sales. Both domestic and foreign market share data are based on the volume of shipments wherever possible. When data on the volume of shipments was not available, market shares were based on the value of shipments. The dumping and subsidy margins were taken directly from the reports. In several cases, a specific rather than ad valorem margin was given. These cases were dropped. The market share of dumped or subsidized imports (FMS) as well as any changes in import volumes were adjusted to allow for the cumulation of imports. Cumulation takes place in cases involving dumped or subsidized imports from more than one country. In these cases, if the imports are sufficiently similar, the ITC must cumulatively assess the impact of the dumped imports on the domestic industry. This means that for the purposes of addressing causation, the import shares must be combined across some or all of the countries involved in the investigation. Both the variables HN and SN were constructed using a detailed atlas, Ward's Business Directory [17], Congressional Districts in the 1980s [6], and the Almanac of American Politics [1] in conjunction with the location of "active" firms provided in the ITC reports. In a few cases, the number of domestic producers was so large that they were not listed individually. In these cases, only those domestic producers who served as witnesses for the petitioner at the ITC hearings or who were identified as supporting the petition in the ITC report were used in the construction of HN or SN. References 1. Almanac of American Politics. New York: E. P. Dutton. 2. Amemiya, Takeshi. Advanced Econometrics. Cambridge: Harvard University Press, Baldwin, Robert E. The Political Economy of U.S. Import Policy. Cambridge: The MIT Press, and Jeffrey W. Steagall. "An Analysis of Factors Influencing ITC Decisions in Antidumping, Countervailing Duty, and Safeguards Cases." Presented at the University of Wisconsin-Free University of Brussels Conference in Washington, D.C., May Caves, Richard, "Economic Models of Political Choice: Canada's Tariff Structure." Canadian Journal of Economics, May 1986, Congressional Districts in the 1980s. Washington: Congressional Quarterly, Inc., Finger, J. Michael, H. Keith Hall, and Douglas Nelson, "The Political Economy of Administered Protection." American Economic Review, June 1982, Goldstein, Judith and Stefanie Lenway, "Interests or Institutions: An Inquiry into Congressional-ITC Relations." International Studies Quarterly, September 1989, Hansen, Wendy L., "The ITC and the Politics of Protection." American Political Science Review, March 1990, Herander, Mark G. and Roger L. Pupp, "Firm Participation in Steel Industry Lobbying." Economic Inquiry, January 1991, and J. B. Schwartz, "An Empirical Test of the Impact of the Threat of U.S. Trade Policy: The Case of Antidumping Duties." Southern Economic Journal, July 1984, Kaplan, Seth. "Injury and Causation in USITC Antidumping Determinations: Five Recent Approaches," in Policy Implications of Antidumping Measures, edited by P. K. M. Tharakan. Amsterdam: North Holland, 1991, pp Maddala, G. S. Limited Dependent and Qualitative Variables in Econometrics. Cambridge: Cambridge University Press, Moore, Michael O., "Rules or Politics?: An Empirical Analysis of ITC Antidumping Decisions." Economic Inquiry, July 1992, Pincus, Jonathan, "Pressure Groups and the Pattern of Tariffs." Journal of Political Economy, August 1975, Prusa, Thomas J. "The Selection of Antidumping Cases for ITC Determination." Presented at the Conference on Empirical Studies of Commercial Policy in Cambridge, Massachusetts, March Ward's Business Directory of U.S. Private and Public Companies. Detroit: Gale Research Inc., Weingast, Barry and Mark Moran, "Bureaucratic Discretion or Congressional Control?: Regulatory Policymaking by the Federal Trade Commission." Journal of Political Economy, October 1983,