takes the final decision. This pair 2 The Selection of Antidumping Cases for ITC ...econweb.rutgers.edu/prusa/cv/04 - selection of cases.pdf · 2009. 2. 11. · ITC will make its

takes the final decision. This pairme dependent variable would be]ecision (the only difference thenc) similar explanatory variables

contribution in this paper, mostory to test for the importance of~-ategy has problems, in large partwould be more persuasive if theyFinally, other pairs of policies

.ther commitment plays an impor-

2 The Selection of AntidumpingCases for ITC DeterminationThomas J. Prusa

2.1 Introduetlon

One of the most remarkable changes in international trade during the pasttwo decades has been the emergence of nontariff barriers (NTBs) as the newform of protectionism (Bhagwati 1988; Page 1987). Voluntary export re-straints, orderly marketing arrangements, countervailing duty complaints, andantidumping actions are the most commonly used weapons of the new protec-tionism. Of specific interest for this paper is the widespread popularity ofantidumping actions. While it is generally agreed that countervailing duty andantidumping laws have a legitimate role in maintaining "free and fair" trade,there is growing fear that such laws have been used to unfairly impede tradeand harass rival foreign suppliers.

Bhagwati (1988) argues that the design and implementation of antidumpinglaw have encouraged its strategic use. For instance, the Commerce Depart-ment’s procedures for calculating dumping margins can lead to positive mar-gins even though the average sales price across countries is the same. Domes-tic industries are not penalized for filing "frivolous" cases, and therefore aremore likely to file with the intent to harass their rivals. Most importantly, sincemany actions are settled (i.e., involve some type of price or quantity agree-ment), merely analyzing the incidence of antidumping duties greatly under-,rates the true trade distortion. Messerlin (1988, 1989) and Prusa (1988) have~hown that on average withdrawn cases are at least as restrictive as cases thatactually result in duties. This is an important result, implying that in practice

~h~mas J. Prusa is assistant professor of economics at State University of New York at Stony

the autO,or is grateful to John Caputo for excellent research assistance. He would like to thank!-;,b t~ald~in, Jim Brown, Wendy Hansen, Audrey Light, Tracy Murray, and Bob Stem for their~,t~ ~e and comments. Gary Hortick also provided valuable assistance. The American Council of~ ::,trned Societies plovided financial m~si~tance. All remaining errors are the author’s.

48 Thomas J. Prusa

antidumping law is far more protective than is typically measured. As we willdiscuss further in section 2.2, there are many reasons why a petition could bewithdrawn, some of which do not have any implication on the firms’ pricin~practices. But many others imply that prices for the goods involved will in-crease.

While economists are beginning to understand the implications of such pe-tition withdrawals (Prusa 1988; Staiger and Wolak 1988, 1991), there has notbeen a systematic analysis of what determines whether a case will be with-drawn or whether the International Trade Commission (ITC) will make thefinal decision. The purpose of this paper is to perform such an analysis. Sincemany (but not all) of the petition withdrawals have involved the steel industry,the simple answer is that cases involving the steel industry are withdrawn butothers proceed through to the ITC’s injury-determination process. Given thatthe steel industry has apparently entered a period of "managed" trade, wemight expect the settlement issue to fade away. However, a closer look at thedata suggests there are certain characteristics that seem to predict quitestrongly how a case will ultimately be resolved. Therefore, I believe it is bet-ter to think of the steel industry as a special case of a more general phenome-non. The analysis also reveals that the way in which an industry uses-anti-dumping law (i.e., filing patterns, countries involved, etc.) stronglyinfluences the outcome. Since many other industries also feature the samecharacteristics that apparently lead to settled outcomes, it is likely that we willcontinue to observe a large number of antidumping cases being withdrawn.

In particular, an antidumping investigation is best thought of as a two-stagedecision problem. At the first stage the industry (or the U.S. Trade Represent-ative) decides whether to settle the case. If the case is settled, the ITC neverhas to make its final injury decision. However, if the case is not settled, theITC will make its final decision. One advantage of this model is that we canpartially capture the dynamics of the first-stage decision problem. In particu-lar, if we find that the characteristics (e.g., high unemployment, low capacityutilization) which influence the ITC’s final decision also lead to settlements(suggesting the foreign industry settles because it realizes it will inevitablylose its case), then the welfare implications of such settlements may not beparticularly adverse. If on the other hand, we find that political pressure playsthe major role in determining whether a case is settled (independent of theinjury criteria), then the welfare implications of such agreements are muchmore unfavorable.

The paper proceeds as follows. Section 2.2 discusses the institutional back-ground of U.S. antidumping law and provides a broad description of its useduring the past decade. Section 2.3 discusses the political pressures and eco-nomic criteria that may affect how a case is resolved and examines the rela-tionship between political and economic variables and case outcomes. Section2.4 tests more formally these relationships, using a nested logit model to ana-

49 Selection of Antidumping Cases

tvze the decision problem. Section 2discussion.

2.2 Antidumping Law: The Statm

The purpose of this brief summaryrelevant aspects for this study. For a t:Vermulst (!989) provide an excellent ~those concerned with legal issues. N(Act of 1979 made significant changesion will focus only on actions during

Simply put, "dumping" describes tless than the price for which those pr,"normal" price). While such a simpleshould be aware that as applied, the c,cared. For instance, there are adjustrsales, adjustments if the country is asales are made below cost. For manyimportant considerations, but for thenient if we think of dumping in its sin

Typically an industry, or some gro~association or union), simultaneouslypartment’s International Trade Admitder for duties to be levied against the !that goods have been sold at less thdetermine that the domestic industryinjury by reason of dumped imports.against a single country. However, iferal countries have sold at ~xFv, it ~each country. A multiple petition filirtire injury decision because tfie ITC 1determining injury. The cumulation 1total volume of allegedly dumped i~example, if the do~nestic industry ill,countries, each of which has 5 perceits injury determination on the injurieven if the dumping margins differinvolved in the investigation, the gre:the greater is the likelihood of an1980, approximately 70 percent of a:multiple petition filing.

Once the petition has been filed,decisions. Within 45 days the ITC m~

typically measured. As we willreasons why a petition could be~nplication on the firms’ pricingfor the goods involved will in-

and the implications of such pe-’/olak 1988, 1991), there has not~es whether a case will be with-,~mmission (ITC) will make theperform such an analysis. Sincehave involved the steel industry,steel industry are withdrawn but::termination process. Given thatperiod of "managed" trade, we.~y. However, a closer look at the.tics that seem to predict quiteed. Therefore, I believe it is bet-case of a more general phenome-m which an industry uses anti-

atries involved, etc.) stronglyindustries also feature the sameoutcomes, it is likely that we willroping cases being withdrawn.~ is best thought of as a two-stagetry (or the U.S. Trade Represent-the case is settled, the ITC neverver, if the case is not settled, the

;~tage of this model is that we can’age decision problem. In particu-~igh unemployment, low capacitydecision also lead to settlementsause it realizes it will inevitably~ of such settlements may not bee find that political pressure plays~tse is settled (independent of the~ns of such agreements are much

2 discusses the institutional back-des a broad description of its use,:s the political pressures and eco-s resolved and examines the rela-:iables and case outcomes. Section


lyze the decision problem. Section 2.5 provides concluding comments anddiscussion.

2.2 Antidumping Law: The Statute and Recent Trends

The purpose of this brief summary of antidumping law is to highlight therelevant aspects for this study. For a more in depth presentation, Jackson andVermulst (1989) provide an excellent and up-to-date discussion, especially forthose concerned with legal issues. Note also that since the Trade AgreementAct of 1979 made significant changes to U.S. antidumping law, this discus-sion will focus only on actions during the 1980s.

Simply put, "dumping" describes the sale of products for export at a priceless than the price for which those products are sold in the home market (the"normal" price). While such a simple definition is adequate for this paper, oneshould be aware that as applied, the concept of dumping is extremely compli-cated. For instance, there are adjustments if there are too few home marketsales, adjustments if the country is a nonmarket economy, and adjustments ifsales are made below cost. For many studies these various complications areimportant considerations, but for the approach taken in this paper it is conve-nient if we think of dumping in its simplest terms.

Typically an industry, or some group representing an industry (e.g., a tradeassociation or union), simultaneously files a petition with the Commerce De-partment’s International Trade Administration (ITA) and with the ITC. In or-der for duties to be levied against the foreign industry, the ITA must determinethat goods have been sold at less than fair value (~rFV) and the ITC mustdetermine that the domestic industry has been or is threatened with materialinjury by reason of dumped imports. It is noteworthy that each petition is filedagainst a single country. However, if the domestic industry believes that sev-eral countries have sold at ~rrv, it will simultaneously file a petition againsteach country. A multiple petition filing increases the likelihood of an affirma-tive injury decision because the ITC follows the "cumulation principle" whendetermining injury. The cumulation principle allows the ITC to consider thetotal volume of allegedly dumped imports from all involved countries. Forexample, if the domestic industry files antidumping complaints against threecountries, each of which has 5 percent of the U.S. market, the ITC will baseits injury determination on the injurious effect of a 15 percent market share,even if the dumping margins differ by large amounts. The more countriesin\olved in the investigation, the greater is the volume of trade, and thereforeIhe greater is the likelihood of an affirmative injury determination. SinceI~80, approximately 70 percent of antidumping petitions have been part of arm~ltiple petition filing.

Once the petition has been filed, the ITA and ITC each make a series of.i~,~t,i~ms. Within 45 days the ITC makes a preliminary injury decision. If this

50 Thomas J. Prusa

decision is negative, the case is terminated. Review of ITC cases suggests thatthe material injury standard is lower at the preliminary stage than at the finaldecision, and therefore cases terminated at the preliminary injury stage do notappear to have merit (see Moore 1988 for statistical evidence). The purposeof the preliminary determination is to filter out these "frivolous" petitions. Ifthe ITC’s preliminary decision is affirmative, the ITA makes a preliminaryLTFV decision within 160 days and then a final ~-TFV decision within 75 daysof its preliminary decision. A final ~_TFV determination is made regardless ofthe preliminary decision (i.e., the ITA’s preliminary decision does not termi-nate the investigation). The chief purpose of the preliminary ~_TFV decision isto set a temporr-y bond rate that is in effect until the case is officially resolved.The bond grants the domestic industry temporary protection for the courseof the investigation. If the foreign industry is found to have dumped, it forfeitsthe bond. If the ITA’s final decision is affirmative, the ITC must make its finalinjury decision no longer than 75 days after the ITA’s final determination. Ifthe ITC’s final decision is affirmative, duties are collected for a period of noless than two years. If the ITC’s final decision is negative, the case is termi-nated and any bond paid during the course of the investigation is refunded.

At any time during the investigation, the party that filed the petition canwithdraw its petition. There are several reasons why this may happen. First, adisadvantageous decision or a piece of unfortunate evidence might be revealedwhich leads the party to withdraw its petition. For instance, the ITA maydecide to calculate dumping margins using a different set of adjustment pro-cedures than was proposed in the petition. Rather than proceed with a hope-less petition, the petition may be withdrawn. Second, the U.S. governmentmay arrange a quantity or price agreement with the foreign government/indus-try that eliminates the injury. Once an agreement is achieved, the U.S. govern-ment pressures the domestic industry to withdraw its petition. The 1982 and1984 steel arrangements are examples of this type of withdrawal. Third, thecase can be suspended if the foreign industry agrees to eliminate LTFV sales tothe U.S. market. The 1985 semiconductor agreement is an example of thistype of withdrawal. Fourth, the party may withdraw its petition on its ownaccord, without any cleat’ reason why it is doing so. It is possible that suchwithdrawals are based on some type of private agreement between the foreignand domestic parties. Prusa (1988) argues that such agreements are providedantitrust immunity by the Noerr-Pennington Doctrine.

As this discussion indicates, there are a number of ways a settlement can bereached, and one should not infer that a withdrawal is an indication of anunsuccessful petition. In fact, a review of the cases confirms that at least 80-90 percent of withdrawn cases involve some type of agreement. Unfortu-nately, given the way the U.S. government reports antidumping outcomes, itis sometimes impossible to know with certainty whether or not a withdrawalinvolves a settlement. To keep the analysis as simple as possible, we will treatall withdrawn cases as if they involved some type of settlement agreement,

5I Sclectio~ of Anlidumping ’

and hence we rise the terms in~Imping sections only dependsmaximize firm profits.

A look at recent trends in alarge number of cases (25 percmary of antidumping actions b5period. The broadly defined mclearly been the heaviest user omg lbr nearly two-thirds of theMMP industry has only had mo~bmining duties in approximatedeceptive since the MMP indu~ithdrawn cases, obtaining socases. All in all, no other induslaw as the MMP industry. The cantidumping law, while the agtneither of these industries displMMP industry.

Table 2.2 breaks down casesthe entire time period only twekvvv determination. This is co~the ITA’s calculation procedure:trast, showing material injury Irejected sixty-five cases at the !foreign industry’s has extremel,that the preliminary injury stan~dard. This suggests that manyalso terminated an additional fi~Although there have been peti~mately 60 percent occtirred in {rangements.

Table 2.3 provides a regionalcountries have been charged wifor nearly 50 percent of the caalleged to have dumped, althoufar more cases. The incidence,difference between Japan and ttJapanese firms have resulted infirms is only 18 percent. Howe~withdrawn, the incidence of sucterribly different. The fact that

I. Japan, \Vest German.,,’, Taiwan, Ira!often involved (in descending order)

Review of ITC cases suggests thatpreliminary stage than at the finalhe preliminary injury stage do notstatistical evidence). The purposeout these "frivolous" petitions. Ifre, the ITA makes a preliminaryinal LJFV decision within 75 daystermination is made regardless of:liminary decision does not termi-f the preliminary L’rFv decision ismtil the case is officially resolved.aporary protection for the course~ found to have dumped, it forfeits~ative, the ITC must make its finalr the ITA’s final determination. If,s are collected for a period of noion is negative, the case is termi-f the investigation is refunded.: party that filed the petition canons why this may happen. First, atunate evidence might be revealedtion. For instance, the ITA maya different set of adjustment pro-,Rather than proceed with a hope-n. Second, the U.S. governmentith the foreign governmentJindus-nent is achieved, the U.S. govern-hdraw its petition. The 1982 andis type of withdrawal. Third, the’ agrees to eliminate LTFV sales toagreement is an example of thiswithdraw its petition on its owndoing so. It is possible that suchte agreement between the foreigntat such agreements are providedDoctrine.tuber of ways a settlement can beithdrawat is an indication of ane cases coniirms that at least 80-~e type or agreement. Unfortu-eports antidumping outcomes, itintS~ whether or not a withdrawal, simple as possible, we will treat~e ~ype of settlement agreemeut.

Selection of Antidumping Cases

and hence we use the terms interchangeably. Further, the analysis in the fol-lowing sections only depends on the assumption that petition withdrawalsmaximize firm profits.

A look at recent trends in antidumping actions re4eals that a surprisinglylarge number of cases (25 percent) are withdrawn. Table 2.1 presents a sum-mary of antidumping actions by product description and year for the 1980-88period. The broadly defined metals and metal products (MMP) industry hasclearly been the heaviest user of antidumping law during the !980s, account-ing for nearly two-thirds of the 395 cases. Despite its heavy use of the law, theMMP industry has only had moderate success in (officially) winning its cases,obtaining duties in approximately one-third of its cases. However, this is a bitdeceptive since the MMP industry has accounted for over 90 percent of thewithdrawn cases, obtaining some type of relief in about ninety additionalcases. All in all, no other industry has had as much success with antidumpinglaw as the MMP industry. The chemicals industry is the second largest user ofantidumping law, while the agricultural products industry is third. However,neither of these industries display the same propensity for settlements as theMMP industry.

Table 2.2 breaks down cases by year. It is somewhat surprising that duringthe entire time period only twenty-four cases were terminated by a negative~_xvv determination. This is consistent with the belief that the rules governingthe ITA’s calculation procedures are biased against foreign producers. In con-trast, showing material injury has proven to be more difficult. The ITC hasrejected sixty-five cases at the preliminary stage, despite the fact that (i) theforeign industry’s has extremely limited time for defense preparation and (ii)that the preliminary injury standard is less exacting than the final injury stan-dard. This suggests that many frivolous cases are indeed filed. The ITC hasalso terminated an additional fifty-six cases at the final injury determination.Although there have been petition withdrawals nearly every year, approxi-mately 60 percent occurred in either 1982 or 1984, the years of the steel ar-rangements.

Table 2.3 provides a regional tabulation of cases. All told, more than fiftycountries have been charged with dumping, with seven countries accountingIor nearly 50 percent of the cases.~ Japan has been the country most oftenalleged to have dumped, although the EC (as a region) has been involved inl’ar more cases. The incidence of duties levied is probably the most strikingdifference between Japan and the EC: more than 50 percent of cases againstJapanese firms have resulted in duties, while the corresponding rate l:br ECtirms is only 18 percent. However, when one adjusts for the number of cases~ithdrawn, the incidence of successful cases against the EC and Japan is notterribly different. The fact that Japan is less inclined to settle antidumping

I Japan, West Germany. Ta wan. Italy, Canada, Brazil, and South Korea are the countries most,,lien ln’,olved (in descending orders.


=0

I;~hh’ 2.2 Antidumping Case Sumn

Dumping No Injury No~ ,..~ Order (Preliminary) (I

1 04 04 t35 2

t3 1819 78 2

27 1244 8

cases 135 65

’,,,to See appendix for data sources. Excludes~m~pmg cases initiated during 1978 and 1979 we

effect.

cases in a politically agreeable fashpensity for managed trade agreem¢surrounding official ITC decisionsJapan competes unfairly.

2.3 Political Pressure vs. Econo

The overview presented in the ~law has been used by a wide varletof countries. It also revealed that ~:Figure 2.1 depicts the hypothesizedomestic complainant and foreign :case is not settled/withdrawn, theFor simplicity we will assume thatITC’s preliminary determination arthe data presented in section 2.2, th

What factors influence whetherplore two hypotheses: self-selectio~hypothesis argues that cases are seSpecifically, if the domestic indusforeign industry defends itself) in owithdrawn only if the profits fromprofits from an otficial ITC decisi(foreign) industry’s profits from w(foreign) industry’s profits if the IT{tic (foreign) industry’s profits if the


Fable 2.2 Antidumping Case Summary, 1980-1988, by Year Initiated

Dumping No Injury No Injury No LTVF Withdrawal/ PetitionYear Order (Preliminary) (Final) Sales Agreement Dismissed Total

’~78979’~80

()82q83

t~85

987

\n of cases

1 0 0 0 0 0 14 0 3 1 0 0 84 13 1 1 10 0 295 2 1 1 4 2 15

13 18 5 3 23 3 6519 7 10 5 3 2 468 2 10 6 44 0 70

27 12 7 2 15 0 6344 8 9 3 7 0 718 2 3 0 2 0 152 I 7 2 0 0 12

135 65 56 24 108 7 395

\,,re. See appendix for data sources. Excludes 16 cases that have not been resolved. Nine of the antid-,roping cases initiated during 1978 and 1979 were not resolved until after the 1979 Trade Agreement Act

effect,

cases in a politically agreeable fashion is somewhat surprising, given its pro-pensity for managed trade agreements; in addition, the heightened publicitysurrounding official ITC decisions probably contributes to the feelings thatJapan competes unfairly.

2.3 Political Pressure vs. Economic Criteria

The overview presented in the previous section showed that antidumpinglaw has been used by a wide variety of industries and against a large numberof countries. It also revealed that a striking number of cases are withdrawn.Figure 2.1 depicts the hypothesized decision process. At the first stage thedomestic cotnplainant and foreign respondent decide whether to settle. If thecase is not settled/withdrawn, the ITC will proceed with its determination.For simplicity we will assume that frivolous cases have been eliminated at theITC’s preliminary determination and that the ITA has found dumping. Giventhe data presented in section 2.2, these are reasonable assumptions.

What factors influence whether a case is withdrawn? This paper will ex-plore two hypotheses: self-selection and political pressure. The self-selectionhypothesis argues that cases are settled to avoid the inevitable ITC outcome..’;pecifically, if the domestic industry files its antidumping petition (and theloreign industry defends itself) in order to increase profits, the petition will be;~ ithdrawn only if the profits from withdrawing are greater than the expectedprofits from tin oNcial ITC decision. Letting I!,,. (Er~,.) denote the domestici toreign) industry’s profits from withdrawing, f[~ (17[~) denote the domestic~l,,reign) industry’s profits it the ITC levies duties, 1I., (Fl.{,) denote the domes-I~ � tc¢cign) industrv’s profits if the [TC rejects the petition, and C (C*) denote

54 Thomas J. Prusa

Table 2.3 Antidumping Case Summary, 1980-1988, by Region

Dmnping No Injury No Injury No LTFV Withdrawal/ PetitionRegion/Country Order (Preliminary) (Finall Sales Agreement Dismissed

Selection of Antidumping Cas

EuropeanCommunity

NICsLatin AmericaJapanNonmarket

economiesCanadaOther EuropeAsiaOther

No. of cases

Note: See table 2.1.

23 36 13 6 43 5 12~26 5 12 6 5 0 5,:19 6 7 4 16 125 6 8 2 8 0

18 1 4 3 21 19 6 5 0 2 0 224 4 3 1 6 04 0 0 1 0 0 57 1 4 1 7 0

135 65 56 24 108 7 395 Fig. 2.1

Industry

With&a/A1

Estimated decision tree

the domestic (foreign) industry’s additional legal expenses if the case pro-ceeds through the injury decision, the petition will be withdrawn if

(1) 1Jw~ pglD + (1 - p)lJ,v - C,H:~, >-p’H; + (1 - p*)H,~., - C*,

where p (p*) denotes the domestic (foreign) industry’s subjective probabilityof an affirmative injury decision. Clearly, 1Jr, > Fl,v and gl~, > H~. That is,the domestic industry benefits and the foreign firm loses when duties are lev-ied. Further, we will assume IJo _> Hw _ II~,, and H~, _< I1;~, _< FI~,,. That is,the payoff or value of a withdrawn case falls somewhere between the payoffsunder the other two alternatives.2

As the ITC’s investigation proceeds, both parties will estimate p (p*). TheITC typically cites changes in capacity utilization, employment, and importmarket share, etc., in its reports, and thus these variables will be used to esti-mate p (p*). If the economic criteria indicate injury, then the parties will at-tempt to reach some agreement (i.e., p and p* close to 1). Under these circum-stances, withdrawn cases are self-selected based on their likely outcome. Ifself-selection describes the withdrawal process, then settlements have roughlythe same welfare consequences (for consumers) as duties. In particular, do-mestic consumers will pay higher prices for the foreign goods because therehas been a violation of fair trade principles.

In contrast, under the political pressure hypothesis, cases are withdrawn not

2. Prusa (1988) has shown that this ranking of payoffs need not always hold. For the presentanalysis, however, it is convenient to assume that the domestic industry’s payoff from withdrawalis less than the payoff when duties are levied.

because the inevitability of the ou~interest of the United States to arting action might become a politictries. The threat of duties beingelevates the issue into a major int,response to the wave of antidumpers, the EC threatened retaliationet al. 1988). Obviously neither ttwar, so they negotiated a settletxindustry files its petitions can incr,Under this political pressure hypoa settlement agreement to be actknow that it is very likely the casITC’s final decision (i.e., p = p*tually any settlement since H~v >only commit to a price undertakiforeign firm will not be willing toand thus it is not likely that case

However, under the political pr~be forced (by its own govemmenlnot injured the domestic industr3industry can force the U.S. and f

3. If p = O, the domestic industry maying) to avoid spending C.

Selection of Antidumping Cases

-1988, by Region

~o LTFV Withdrawal/ PetitionSales Agreement Dismissed Total

6 43 5 126

6 5 0 54

4 16 1 532 8 0 49

3 2l I 480 2 0 22

I 6 0 18

l 0 0 5

1 7 0 20

24 108 7 395 Fig. 2.1

Withdraw / ~ Continue with

D~nlA2 A3

Estimated decision tree

,U legal expenses if the case pro-on will be withdrawn if

;t) industry’s subjective probability[Io > [1.,~ and H;) > 1I,~. That is.ign firm loses when duties are lev-11x and [1~ ~< II~v -< H,~,. That is,

lls somewhere between the payoffs

th parties will estimate p (p*). The:ilizatiou, employment, and importthese variables will be used to esti-:ate injury, then the parties will at-: p* close to 1). Under these circum-I based on their likely outcome. If~cess, then settlements have roughlyumersl as duties. In particular, do-,-or the foreign goods because there

h.vpo~hesis, cases are withdrawn not

/101(I. l~’or tt/e pr~_,sel~l

from -9, {~idl;t~’, i,,~

because the inevitability of the outcome but rather because it is in the politicalinterest of the United States to arrange a negotiated settlement. An antidump-ing action might become a political issue if the filing involves multiple coun-tries. The threat of duties being levied against a large number of countrieselevates the issue into a major international dispute. In 1984, for instance, inresponse to the wave of antidumping petitions filed against EC steel produc-ers. the EC threatened retaliation against paper, beef, and fertilizers (Howellet al. 1988). Obviously neither the United States nor the EC wanted a tradewar, so they negotiated a settlement. In other words, the way in which an

’ ’ d"industry files its petitions can increase its chances for’ negotmte protection.Under this political pressure hypothesis, equation (1) need not be satisfied fora settlement agreement to be achieved. For example, suppose both partiesknow that it is very likely the case will be rejected if it proceeds through theITC’s final decision (i.e., p = p* ~ 0). The domestic firm will agree to vir-tually any settlement since Ilw >- I1,~, - C; however, the foreign firm willonly commit to a price undertaking if II{~ > 11" - C*. If C* is small, thet’oreign firm will not be willing to offer a price undertaking (since IIw -< llu)and thus it is not likely that case will be settled?

However, under the political pressure hypothesis, the foreign industry maybe forced (by its own government) to offer a price undertaking even if it hasnot injnred the domestic industry. By filing multiple petitions the domesticindustry can force the U.S. and foreign governments to deal with a sectoral

3 If p = t), the domestic industry l;qay prefer to withdraw its petition (without an.’,’ undertak-,p.~ ’, lo a\,fid spending C.

56 Thomas J, Prusa

problem in the broader context of overall U.S. foreign relations; the mereinvestigation and threat of duties threatens trade relations and can lead Io theimposition of a price/quota agreement even when it is unlikely that duties willbe levied?

The theory of "congressional dominance" provides another justification forthe political pressure hypothesis (Weingast 1981; 1984; Weingast and Moran1983). Under this theory Congress controls agency decisions via oversightcommittees and budget decisions. Members of Congress favor agencies who"serve their constituencies" and penalize those who do not. Baldwin (1985)argues that the structure of the ITC insulates the commissioners from unduecongressional pressure;5 however, Moore (t988) argues that because Congresshas complete control (no OMB oversight) over the ITC’s budget, Baldwinunderestimates Congress’s leverage.

Using different methodology and data sets, Moore (1988) and Hansen(1990) have both found that oversight committees exert significant influenceon ITC decisions. One advantage of Moore’s approach is that he uses datareported in the official ITC reports. However, this also severely limits the typeand quantity of data he can use, since only cases with ITC injury determina-tions enter his analysis. In addition, confidentiality prevents data from beingpublished for those cases involving a small number of firms. Because muchof the data he uses is unavailable for many cases, his data set coverage is muchmore limited than mine. Moreover, since the cases that do not enter Moore’sanalysis all have a similar industry structure, his results must be carefullyinterpreted. Hansen takes a different approach, estimating the ITC’s decisionfunction for antidumping, countervailing duty, and escape clause cases for the1975-84 period. Her data set is much broader than mine, but since the injurycriteria for an escape clause determination is quite different from antidumpingand countervailing duty determinations, her injury estimation does not have anatural interpretation. Furthermore, and most relevant for this analysis, nei-ther Hansen nor Moore consider the (firm’s) first-stage decision in their anal-ysis, and therefore they ignore a sizable number of cases. The chief advantageof the approach taken in this paper is that by also measuring the influenceexerted at the withdrawal stage, it permits a more precise characterization ofthe (potential) political pressure. For instance, congressional oversight com-~nittees could influence the likelihood of a settled case by (i) pressuring theU.S. Trade Representative (USTR) to negotiate a settlement and!or (ii) byraising the likelihood of an affirmative decision via pressure of the ITC. Fur-ther, there are a couple reasons why Congress would prefer protection granted

4. Howell et al. (1988) argue that steel quotas have been agreed to when it appeared that therewould be a "no injury" finding.

5. Commissioners cannot be reappointed, they are nominated by the executive branch but ap-proved by the legislative branch, and no more than three commissioners can be from the sameparty.

+,+ petition withdraws rather tha:(+,+t|gress are risk averse, they wott,, the uncertain IT(-" outcome. Semc,tsures of protectionism. Thus,¯ =re,s may prefer protection grante,~ ~eil for trade restrictions.

Beti+re proceeding to the formaol insight into the data by simply.,hips among variables. For instmli~med the data set into those cas(,tares (districts) which had a repr,Subcommittee or House Trade Suthat have direct oversight respons~and all of the following) I focus ~t~ere subject to a final ITC decisic.to drop eighteen cases, most of wtdetailed discussion of the metho(this paper can be found in the appt

If the null hypothesis is that the~e would expect industries withmore often than those industries ’top panel (Senate), it is clear thaaffect on the withdrawal decisionsented by at least two senators,senators. Virtually the identical poversight subcommittee. Since ~protection, these numbers sugges~sentation are more likely to be pro11988) and Hansen (1990) studie.,tees do not exert undue influence,t 75 cases that proceeded to the IT~we control for congressional rep~cases without congressional repr~raining protection than those withconditional probabilities are not ~do suggest that if Congress doeseffect is on the settlement process.

The data in table 2.1 revealed ttmetal products and thus it is nattable 2.5, I partition the cases by ’(as expected) most of the withdravthe second-stage decision appear,,dustry is involved. This is a bit

. foreign relations; the mererelations and can lead to the

a it is unlikely that duties will

vides another justification forl; 1984; Weingast and Moran:ency decisions via oversightCongress favor agencies whowho do not. Baldwin (1985)e commissioners from undueargues that because Congressr the ITC’s budget, Baldwin

, Moore (1988) and Hansenees exert significant influenceapproach is that he uses datais also severely limits the typees with ITC injury determina-ality prevents data from beingtuber of firms. Because much, his data set coverage is muchases that do not enter Moore’shis results must be carefullyestimating the ITC’s decision

md escape clause cases for thethan mine, but since the injuryrite different from antidumpingury estimation does not have arelevant for this analysis, nei-:st-stage decision in their anal-r of cases. The chief advantage’ also measuring the influence~ore precise characterization of. congressional oversight corn-fled case by (i) pressuring theate a settlement and/or (ii) byn via pressure of the ITC. Fur-,vould prefer protection granted

n agreed to when it appeared that there

~inated by the executive branch but ap-


via petition withdraws rather than via ITC decisions. First, if members ofCongress are risk averse, they would prefer the certainty of a settled outcometo the uncertain ITC outcome. Second, settled cases do not enter the officialmeasures of protectionism. Thus, in the context of trade negotiations, Con-gress may prefer protection granted via unofficial settlements since it providesa veil for trade restrictions.

Before proceeding to the formal econometric test, one can gain quite a bitof insight into the data by simply examining simple cross-tabulation relation-ships among variables. For instance, to examine political pressure, I parti-tioned the data set into those cases with and without production facilities instates (districts) which had a representative on the Senate International tradeSubcommittee or House Trade Subcommittee (table 2.4), the subcommitteesthat have direct oversight responsibilities for the ITC. Note that in this table(and all of the following) I focus only on those cases that were withdrawn orwere subject to a final ITC decision. Also due to data limitations I was forcedto drop eighteen cases, most of which involved agricultural products. A moredetailed discussion of the methods used to construct the data series used inthis paper can be found in the appendix.

If the null hypothesis is that these subcommittees influence decisions, thenwe would expect industries with oversight representation to gain protectionmore often than those industries without representation. Looking first at thetop panel (Senate), it is clear that congressional representation has a strongaffect on the withdrawal decision. Of the 106 withdrawals, 104 were repre-sented by at least two senators, while 99 were represented by at least fivesenators. Virtually the identical pattern is found when we look at the Houseoversight subcommittee. Since withdrawals typically involve some type ofprotection, these numbers suggest that industries with strong oversight repre-sentation are more likely to be protected. However, in contrast with the Moore11988) and Hansen (1990) studies, this table suggests the oversight commit-tees do not exert undue influence on the ITC’s injury decision. Overall, of the175 cases that proceeded to the ITC, 126 (72 percent) resulted in duties. Whenwe control for congressional representation, we find rather surprisingly thatcases without congressional representation have a higher probability of ob-taining protection than those with representation. Although the differences inconditional probabilities are not as dramatic as those for the first stage, theydo suggest that if Congress does influence antidumping outcomes, the majoreffect is on the settlement process.

The data in table 2.1 revealed that the majority of withdrawn cases involvedmetal products and thus it is natural to explore this relationship further. Intable 2.5, 1 partition the cases by whether they involve steel products and findtas expected) most of the withdrawn cases (87 percent) involve steel; however,the second-stage decision appears to be independent of whether the steel in-,tustry is involved. This is a bit surprising since a number of steel analysts

58 Thomas J. Prusa

1hble 2.4 Congressional Oversight

Petition Withdrawn/Agreement Dumping Order No Injury (Final)

Senate International Trade Subcommittee

Selection of Antidumping

lahle 2,6 Number of Cow

PetitionAgre

Senators with production facilities in their state:Cases with < 2 members 2 17 (81%t 4 (19%) 23Cases with _> 2 members 104 109 (71%) 45 (29%1Cases with < 5 members 7 62 (77%) 19 (23%)Cases with _> 5 members 99 64 (68%) 30 (32%) 193Overall 106 126 (72%1 49 (28%) 281

House of Representatives Trade Subcommittee

Members of Congress with production facilities in their state:Cases with < 2 members 5 27 (75%) 9 (25%) 41Cases with _> 2 members 101 99 (71%) 40 (29%) 240Cases with < 4 members 10 53 (67%) 26 (33%) 89Cases with ~ 4 members 96 73 (76%) 23 (24%) 192Overall 106 126 (72%) 49 (28%1 281

Note: Numbers in parentheses are conditional probabilities (i.e., probability of affirmative [negative] IT("decision, given that case proceeds to final decision). See appendix for data sources.

Table 2.5 Steel Industry Effect

Petition Withdrawn/Agreement Dumping Order No Injury (Final) Total

Cases not involving steel 14 59 (69%) 26 (31%) 99productsCases involving steel 92 67 (74%) 23 (26%) 182productsOverall 106 126 (72%) 49 (28%) 281Note: See table 2.4

have argued quite strongly that foreign dumping has greatly injured the U.S.steel industry.6 However, if cases are being withdrawn because of self-selection, we would expect this result. For instance, if it was known that thesteel industry received preferential treatment from the ITC, rational partieswould take this into account when forming their expectations of p (9"). Sincethe withdrawal decision will adjust for the industry-specific preferential treat-ment at the ITC, the steel cases that proceed to the ITC will have the samechance of obtaining protection as nonsteel cases, even though on their objec-tive merits, steel cases could be weaker than nonsteel cases. Therefore, the

6. See Howell et al. (19881 for numerous references.

l.thngx with 1 CountryFilings with > I Country

~-~lings ~ith -> 2 CountriesI:thng~, with > 2 Countries

tqlmgs with ~< 4 Countries[:dings with > 4 Countries

Overall

.Vote: See table 2.4.

independence of the second-stmerely reflect that agents are act

It was argued above that the pquite standard. Filing multiplethe dumping allegations, increa~,verifies this. Over 80 percent olcountries. It is also argued that fiity of injury via the "cumulatiorbility of an affirmative ITC decnumber of countries involved, q"protective" as originally thougl~lhe self-selection process.

We also expect that the econ~injury to influence both the first-2.8 present data on employmen~tios; a review of ITC ease report.,tant role in ITC decisions. A dration may reflect injury.8 The coappropriability of the benefits oItion ratios will not suffer from thcompetitive industries and thus ~protection. Note, however, thatexports, and inventory are also stunately the unavailability of datain this study. More will be said or

7. Priest and Klein (19841 formally dev8. The ITC actually determines injury

the overall economy. Therefore, one woulchanges in employment and capacity utilinents. If the time series existed as montlGNP to remove the cyclical component.variables.

ping Order No Injury (Final) Total

:tional Trade Subcommittee

17 (81%) 4 (19%) 23

)9 (71%) 45 (29%) 258

"52 (77%) 19 (23%) 88

54 (68%) 30 (32%) 193

26 {72%) 49 (28%) 281

,entatives Trade Subcommittee

27 (75%) 9 (25%) 41

99 (7 t %) 40 (29%) 240

53 (67%) 26 (33%) 89

73 (76%,) 23 (24%) 192

!26 t72%) 49 (28%) 281

e., probability of affirmative [negative] IT(?endix for data sources.

roping Order No Injury (Final) Total

59 (69%) 26 (31%) 99

67 174c;4 ~ 23 (26%) 182

126 (72q- ~ 49 (28%) 281

ping has greatly injured the U.S.ng withdrawn because of self-,,nstartce, if it was known that thett from the ITC, rational partiesheir expectations of p (p*). Sinceldustry-specific preferential treat-~d to the ITC will have the same;tse,< even though on their ob ec-,m mm:,tee! cases. Therefore, the


Table 2.6 Number of Countries Involved in Filing

Petition Withdrawn/Agreement Dumping Order No Injury (Final) Total

Filings with 1 Country 15 51 (74%) 18 (26%) 84

Filings with > 1 Country 91 75 (71%) 31 (29%) 197

Filings with >- 2 Countries 22 75 (74%) 26 (26%) 123

Filings with > 2 Countries 84 51 (69%) 23 (31%) 158

Filings with -< 4 Countries 43 112 (76%) 35 (24%) 190

Filings with > 4 Countries 63 14 (50%) 14 (50%) 91

Overall 106 126 (72%) 49 (28%) 281

Note: See table 2.4.

independence of the second-stage decision from known variables (may)merely reflect that agents are acting optimally at the first stage.7

It was argued above that the practice of filing multiple petitions has becomequite standard. Filing multiple petitions raises the international visibility ofthe dumping allegations, increasing the likelihood of a settlement. Table 2.6verifies this. Over 80 percent of the withdrawn cases involved at least threecountries. It is also argued that filing multiple petitions increases the probabil-ity of injury via the "cumulation principle." However, the conditional proba-bility of an attirmative ITC decision does not appear to be affected by thenumber of countries involved. This suggests either that cumulation is not as"’protective" as originally thought or that cumulation affects the outcome via

the self-selection process.We also expect that the economic criteria that the ITC uses to determine

injury to influence both the first- and second-stage decisions. Tables 2.7 and2.8 present data on etnployment, capacity utilization, and concentration ra-tios; a review of ITC case reports suggests that these variables play an impor-tant role in ITC decisions. A dramatic fall in employment or capacity utiliza-tion may reflect injury.~ The concentration ratio serves as a proxy for theappropriability of the benefits of protection. Industries with high concentra-tion ratios will not suffer from the free-rider problem that might plague morecompetitive industries and thus might be able to more effectively lobby forprotection. Note, however, that other criteria such as import market share,exports, and inventory are also sometimes cited as important factors; unfor-tunately the unavailability of data prevented these variables from being usedia this ~tudy. More will be said on this subject in section 2.5.

Priest and Klein ( 1984~ formally develop this selection argument for litigation cases.The ITC actually determines in_ury after controlling for changes due to cyclical changes in

o\erat! economy. "Therefore. one would actually like to examine the relationship between the. !aanges m ernployl~ent aad capacity utilization after purging these variables of cyclical compo-:tent’s. If the time series existed as monthly data then one could regress capacity utilization on;NP to rem~ve the ~,,<:lica! component. Because l only have annual data+ t do not detrend the

60 Thomas J. Prusa

Table 2.7 Economic Criteria--Distributions

PetitionWithdrawn/ No InjuryAgreement Dumping Order (Final) Overall

Mean

61 Selection of Antidumping I

lable 2.8 Economic Crile

Petitio~

Capacily Utilization (year t,,) 59.5 66.7 66.4 64.5%& capacity utilization - I 1.0% 2.0% 2.4% - 2.9%

(between year t, and t_ ~)%A capacity utilization ~6.0% -0.9% 5.8% 4.4%

(between year t_ ~ andEmployment (thou sands) 234.5 108.3 115.2 148.4%k employment (between -8.1% -5.6% -4.0% -6.4%

year to and t ,)%A employment (between -7.1% -4.0% -5.4% -5.2%

year t_~ and t__,)Final ITA Duty Margin . . . 28.98% 30.32% . . .

Median

Capacity Utilization (year to) 59 66 66 66%A capacity utilization - 10.6% - 2.6% 0% - 5.6%

(between year t. and t t)%A capacity utilization 0% - 3.6% 0ca - 3.6%

(between year t_~ and t_~,)Employment (thousands) 247.3 73.3 43.7 90.6%A employment (between -2.6% -5.2% -2.6% -4.3%

year t. and t ,)%A employment (between -3.3% -3.Vk -1.2% - 3.1~/c

year t_~ and t__,)Final ITA Duty Margin . . . 14.71% 17.58% . . .

Table 2.7 presents the mean and median of these criteria, conditional onoutcome. A number of insights emerge from this table. First, withdrawn caseshave the lowest capacity utilization; however, if the case proceeds to the ITC,capacity utilization does not appear to be an important predictor of the finaloutcome. This pattern is consistent with the self-selection hypothesis. With-drawn cases have also had the greatest fall in capacity utilization (during theyear prior to the petition), which also suggests that cases are being self-selected on the basis of injury criteria.Second, withdrawn cases tend to befrom "large" industries, having employment about twice as great as thosecases that proceed to the ITC. This is consistent with the adding-machinemodel of Caves (1976) and suggests that industries with many employees aremore effective lobbyers. However, emp!oyment does not appear to have anysignificant affect on the ITC’s decision, which is consistent with Baldwin’sconjecture that the ITC is insulated from lobbying pressure. It is also consist-ent with the argument that any measurable employment effects are eliminated

¢’apacit3 utilization <(’,lpacit3 utilization ~- 66

tm~ployment < 137t{mployment ~ 137

Concentration < 42(onccnlration ~ 42

()~eralt

.\’ot~,. Nunlbers in parentheses are conan~e] ITC decision, given that case pThe parlitioning cutoffs are the distrib,

via self-selection. Withdrawn ~ment (during the year prior tosuggest that the ITC ignores ~since average duties are highersions.

Table 2.8 partitions the dat~the median value of capacity ~tio. The capacity utilization rlself-selection hypothesis. Thethat large concentrated industrismall competitive industries. 1"predictors," none of these varITC’s final decision.

2.4 Estimating the Decisim

The hypothesized two-sta~nested-logit analysis. In the fi~decision whether to settle/witpected ITC decision. In thewhether the domestic industryconditional on the case not htheory is well established m3Fadden (1978) and Hausman ~tions of the theory; my prest(1983).

At the first stage, we will agiven by


Table 2.8 Economic Criteria Influence

Petition Withdrawn~ Dumping No InjuryCases with Agreement Order (Final) TotalNo Injury

ping Order IFinal) Overall

Mean

6.7 66.4 64.52,0% 2.4% - 2.9%

0.9% 5.8% 4.4%

8.3 115.2 148.45.6% -4.0% -6.4%

4.0% -5,4% -5.2%

8.98% 30.32% . . .

Median

,6 66 662.6qc 0% - 5.6%

3.6% 0% -3.6%

’3.3 43.7 90.65.2~ 2.6% - 4.3c/r

3.1~ 1.2cA -3.1%

14.71~c 17.58% . . .

of these criteria, conditional onthis table. First, withdrawn casesr, if the case proceeds to the ITC,a important predictor of the final

self-selection hypothesis. With-m capacity utilization (during the_,,gests that cases are heing self-rod. withdrawn cases teud to bent about twice as great as those~sistent with the adding-machinetustries with many employees arenent does not appear to have anyaich is consistent with Baldwin’s".hying pre’~sure. It is also consist-:mployment effects are eliminated

Capacity utilization < 66 33 84 (74%) 29 (26%) 146Capacity utilization -> 66 73 42 (68%) 20 (32%) 135

Employment < 137 18 91 (75%) 31 (25%) 140Employment -> 137 88 35 (66%) 18 (34%) 141

Concentration < 42 17 79 (77%) 24 (23%) 120Concentration ->- 42 89 47 (65%) 25 (35%) 161

Overall 106 126 (72%) 49 (28%) 281

Note: Numbers in parentheses are conditional probabilities (i.e., probability of atfirmative [neg-ative] ITC decision, given that case proceeds to final decision). See appendix for data sources,The partitioning cutoffs are the distribution medians. Employment is in thousands.

via self-selection. Withdrawn cases also have had the greatest fall in employ-ment (during the year prior to the petition). Finally, the results in Table 2.7suggest that the ITC ignores the dumping margin when determining injurysince average duties are higher for "no injury" decisions than for injury deci-sions.

Table 2.8 partitions the data set into those cases that fall above and belowthe median value of capacity utilization, employment, and concentration ra-tio. The capacity utilization results continue to be quite consistent with theself-selection hypothesis. The employment and concentration results suggestthat large concentrated industries are more able to negotiate a withdrawal thansmall competitive industries. However, as has been the case for all the other"predictors," none of these variables appear to be a significant predictor of theITC’s final decision.

2.4 Estimating the Decision Tree

The hypothesized two-stage decision problem lends itself naturally tonested-logit analysis. In the first stage the industry (or USTR) makes a binarydecision whether to settle/withdraw (choice A,), taking into account the ex-pected [TC decision. In the second stage the ITC makes a binary choicewhether the dotnestic indttstry has been injured (choice A_,) or not (choicec~mditional on the case not having been withdrawn. Since the econometric1heory is well established my presentation of the model will be brief. Mc-Fadden (1978) and Hausman and McFadden (1984) are the seminal presenta-tion,; of the theory; my presentation most closely follows that of Maddala

At the fit’~t stage, ~ve will assume that the ith case’s withdrawal decision isgiven by

62 Thomas J. Prusa Selection of Antidumpin~

(2)

where 3’7 is a latent continuous measure of the likelihood that the petition iswithdrawn. However, we do not observe y; but rather an indicator function y,defined by

1, if 3’7 > o,(3)Y’ = 0, otherwise.

Equations (2) and (3) imply that the probability of observing choice A~ can bewritten as

(4)P(y, = 1) = P(3’; > 0 IX,)

= POx. > -Xil3)= 1 - F(-Xi~),

where F(.) is the cumulative distribution function of tx,rSimilarly, at the second stage, we wil! assume the ith case’s final decision

is given by

(5) v*, = Zi7 + tx2,,where v7 is a latent continuous measure of the likelihood that duties are levied.Once again, we do not observe v; but rather an indicator function vi definedby

1, if v7 > 0,(6) v~ = 0, otherwise.

Equations (5) and (6) imply that the probability of observing choice A2 canbe written as

(7)1) = P(v7 >

= P(la,2i > -Ziy)= 1 --

where G(.) is the cumulative distribution function ofX is a vector of exogenous variables that determine the withdrawal decision

while Z is a vector of exogenous variables that determine the final decision.The #., and P,2~ are residuals that capture unmeasured variables, case idiosyn-crasies, etc. In this case, the observed values of 3’~ and v, are just realizationsof a binomial process given by equations (4) and (7), varying from case tocase (i.e., depending on Xi and Z,). If we assume that tx,, and #.2, are indepen-dently distributed we can write the likelihood function as

(8) L = l~ [l - F(-X,{B)] 1~ F(-X, I3)xYt = I yi=O

I’he functional forms for/~assumptions made about ix~:\~uming that g~, and ~:~ arebution allows us to write the C

P(A, ~ Z, X; ~

P(A~ I Z, X," ~

P(A~ I Z, X; ~,

Given these probabilities, the

LOGL(f3. y) =

where D~, = 1 if the nth caseIn many applications the ch{

In the current application mo.paper we will use the economsection 2.3. Further discussior.

Given our model, the first-sration of the second-stage decical and economic factors. To,drawal decision directly or in~will first present a variety of spaltering the second-stage spechas been adequately characteri.the injury decision without alsense the econometric analysisthe determinants of the withdraof the injury decision. It is iminvolves reestimating the enti~presents a variety of specificatifirst set of regressions (modelswhile the second set of regressi,decision.

The analysis of the first-stagemp!oyment and the number ocharacteristics of a case that p

9, lu McFadden’s (19781 original for:sire value that weights the second-stagespecific, the inclusive value is not idenlditional Iogit model,

Lip

dm likelihood that the petition ismt rather an indicator function Yi

ity of observing choice A~ can be

> - X,13)"(

:tion of ~.,ume the ith case’s final decision

likelihood that duties are levied.an indicator function v, defined

bility of observing choice A, can

,olz,)"> - Z,T);(-Z,y),

ction ofetermine the withdrawal decisionhat determine the final decision.neasured variables, case idiosyn-s of y, and v, are just realizations~1 and {7) varying from case to,ume that ~,~ and g_,~ are in@pen-] [’unctiofl as


The functional forms for F(.) and G(.) in equation (8) will depend on theassumptions made about t% and 1*2~ in equations (2) and (5), respectively.Assuming that tx~,. and I*~, are i.i.d, with the generalized extreme-value distri-bution allows us to write the discrete choice probabilities as9

P(A, ] Z, X; 13, y) = p, =+ + 1P(Az I Z, X; [3, y) = Pz = (1 - pt)eZ.~

+ 1P(A3 I Z, X; 13, V) = p3 = (1 - pl)eZv 1

+1

Given these probabilities, the log likelihood can be written as

N 3

(9) LOGL(13, -,/) = ~] ~ D~. log P(A~IZ, X; 13, 3’),n=l j=l

where D~,, = 1 if the nth case is resolved via outcome A~ and zero otherwise.In many applications the choice of X and Z is restricted by economic theory.

In the current application model specific,ation is more open-ended. For thispaper we will use the economic and political pressure variables discussed insection 2.3. Further discussion of this issue is found in section 2.5.

Given our model, the first-stage decision will be influenced by the expec-tation of the second-stage decision and possibly more directly by other politi-cal and economic factors. To distinguish whether a variable affects the with-drawal decision directly or indirectly via its affect on the injury decision, Iwill first present a variety of specifications of the withdrawal decision withoutaltering the second-stage specification. Then, after the withdrawal decisionhas been adequately characterized, I will present a variety of specifications of~lae injury decision without altering the first-stage specification. Thus, in asense the econometric analysis proceeds in two steps, since we first analyzethe determinants of the withdrawal decision and then analyze the determinants

t.)t" the injury decision. It is important to stress that each model specificationrevolves reestimating the entire decision tree (i.e., both stages). Table 2.9presents a variety of specifications that help clarify the decision process. Thefirst set of regressions (models 1-6) concentrates on the first-stage decision,while the second set of regressions (models 7-11) focuses on the second-stagedecision.

The analysis of the first-stage withdrawal decision reveals that the level ofemployment and the number of countries involved in the filing are the keycharacteristics of a case that predict whether a petition will be withdrawn,

In McFadden’s t 19781 origma! formulation of the nested-logit mode!, he estimates an inclu-due that ~eights ~hc second-stage decision. In ~his application, since the Z vector is chooser-

~fic. Ihc inchJsive value is not identified. [n this case. the nested Iogit is equivalent to a con-:’,t’~,!l.~i klglt model.

Table 2,9 Nested Logit Estimates

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

First stage:Constant - 0.696 - t .702 - 1.763 - 1.401 - 2.136

( - 2.195) (-4.534) (-4.622) ( - 1o774) ( - 2.838- 1.710

( - 4.395)Steel dummy (= I if steel) 1.005 0.180

(2.342) (0.383)No. countries in Filing (= t if>2) I. t04 !.333 1.281 1.387 1.336 1,240

(2.851) (3.930) (3. 126) (4.315) (3.926) (3 277)Employmem 0.009 0.009 0.009 0.009

(6.317) (5.670) (6.000) (5,448)Senate oversight ( = I if--2) 1,338 0.482

(1.712) (0.638)%A capacity util. (t,, and t ~)

-2879

%A capacity util. (t ~ and t ,)( -- 1.929j

-- 0.301(0.262)

%A employment (t,, and t. ~) - 0.378(-0.111)

%,3 employment (t ~ and t ..,) -0.237

Second stage:( - 0.088)

Constant 0,825 0,744 0.731 0.904 0,739 0.714(3.425) (3.058) (3.006) (3.755) (3.026) (2.756)

%,_’.’X capacity util. (t, and t_ ,) 2.037 1.466 1.482 1.884 1.456 - 0.135(2.349) ( 1.673) ( 1.682/ (2.127) ( 1.635) ( - 0.126)

%& capacity util. (t_ , and t :) -1.458 -1.763 -1.726 -I.578 -1.715 -1.587( - 1.679) ( - 1.952) ( - 1.857) ( - 1,866) (- 1.888) ( - 1.634~

%A employment (t, and t ,)

%A employment (t_, and t :)

Final ITA duty margin

Log likelihood% correct predictions

First stageSecond stage

-0.241 - 1.778 - 1.9{)4 ¢/.683 1.759( -0. 138) { -0.997) ( - 1.0591 (0.399) ( - 0.985)4.366 3.734 3.626 4.892 3.662 3.480(2.219) (1.825) (1,726) (2.535) (1,769) ~ 1.31810.019 0.016 0.015 {}.021 0.016 0.015(4.093) (3.258) (3.224) {4.343) {3.253) (3.0~}%- 243.257 - 224.034 - 223.941 - 245. 162 - 223.849 - 22(/. 195

80.43% 85.41% 85.41% 78.65 % 85.41% 82.21 cA72.57% 73.14% 73.14% 73.14% 73.14% 73.71~

0.825(3.425)2.037

t2.349)- 1.458- 1.679)

(/.30110.262)

-0.378(-0.111)-0.237

( - 0.088)

0.744 0.731 0.904 0.739 0.714

(3.058) (3.006) (3.755) (3.026) (2.756)

1.466 1.482 1.884 1.456 - O. 135

( 1.673) ( 1.682) (2.127) ( 1.635) ( - O. 126)

- 1.763 - 1.726 - 1.578 - 1.715 - 1.587

- 1.952) (- 1.857) (- 1.866) (- 1.888) (- 1.634)

Log likelihood’)’~ concct predictions

Second stage

First. stage:Constant

- 0.241- 0. 138)

4.366(2.219)0.019

(4.093)

- 243.257

- 1.778-0.997)

3.734(1.825)0.016

(3.2581

-- 224.034

- t.904 0.683 -- 1.759 --2.031( - 1.059) (0.399) ( -- 0.985) ( -- 0.946)

3.626 4.892 3.662 3.480(1.726) (2.535) (1.769) (1.318)0.015 0.021 0.016 0.015

(3.224) (4.343) (3.253) (3.009)

-- 223.941 - 245.162 -- 223.849 -- 220.195

80.43% 85.41% 85.41% 78.65% 85.41% 82.21%72.57cA 73.14% 73.14% 73.14% 73.14% 73.71%

Model 7 Model 8 Model 9 Model 10 Model 11

- 1.683 - 1.608 - 1.740 - 1.677( - 4.397) ( - 4.040) ( - 4.493) ( - 4.372)

No. countries in filing ( = I if >2)

Kmployment

9{,A capacity util. (t0 and I. ,)

Second stage:Constant

%A capacity util. (t,~ and t_ ,)

%A capacity util. (t_ ~ and t. :)

(continued)

1.263 1.081 1.341 1.263(3.643) (2.672) (3.692) (3.637)0.009 0.009 0.009 0.009

(5.993) (6.041) (6.137) (5.897)

- 3.021 - 3.152 - 3.059 - 3. 008- 2.474) ( - 2.582) ( - 2.464) ( - 2.435)

0.737 0.845 0.558 1.001(2.980) (2.979) (1.916) (1.947)

- 0.204 - 0.260 - 0.336 - 0.214

- 0.206) ( - 0.260) ( - 0.332) ( - 0.217)- 1.740 - 1.637 -2,018 - 1.699

- 1.920) (- 1.760) (- 2.240) (- 1.863)

- 1.683(-4.122)

1.282(3.671)0.009

(4.66O)- 2.995

(-2.441)

0.215(0.174)

- 0.540( - 0.485)- 1.831

(- 1.957)

Selec6on of Antidumpin~

When regressed without empoversight dummy are positivespectively). ,0 A positive coeffioversight dummies increase ttemployment is added to thecance, being dominated bylively). Moreover, when emphoversight dummies, the resultsthe dummies included (modelployment and the steel durumindustry has accounted for mosithat influences the withdrawal difications the number of countri~ence on the withdrawal decisiolstrategy can significantly incresome type of protection (in th~interpret this as implying that thpressures the countries into arneconomic criteria (which are kmfor their influence on the withdrathe percentage change in capacprior to the petition filing. The epacity utilization increases the ctcoefficient on this variable at thenegative, providing only weak e~self-selection. ~

The second set of regressionsstage decision, controlling forcountries dummy, employment,during the prior year. The analys(1) the economic criteria do relatiSenate oversight effect is not signis an important predictor of the inj~

Consider first the set of economcentage change in employment antof the coefficients are insignificatv(1990) and Moore (1988) studies, sstate that these economic criteria a~importance. In contrast, in all the s

10. I also tested for the significance of ttvirtually identical to the Senate dummy and ;

I I. I also ran other specifications to test £level of capacity utilization and the concentnavailable on request.


When regressed without employment, the steel industry dummy and senateoversight dummy are positive .and s!gnifi.cant (see model 1 and model 4, re-spectively), m A positive coefficient implies that the steel industry and senateoversight dummies increase the likelihood of a withdrawal. However, whenemployment is added to the specification both variables lose their signifi-cance, being dominated by the employment effect (models 3 and 5, respec-tively). Moreover, when employment is regressed without either the steel oroversight dummies, the results are virtually the same as the specifications withthe dummies included (model 2). This-may reflect collinearity between em-ployment and the steel dummy, but it also suggests that although the steelindustry has accounted for most of the withdrawals, the more general attributethat influences the withdrawal decision is an industry’s size. Also, in all spec-ifications the number of countries involved has a positive and significant influ-ence on the withdrawal decision, which suggests that the "multiple petition"strategy can significantly increase the likelihood the industry will receivesome type of protection (in the form of some price/quantity agreement). Iinterpret this as implying that the tension created by a multiple petition filingpressures the countries into arranging a settlement. Finally, in model 6 theeconomic criteria (which are known to the parties at the first stage) are testedfor their influence on the withdrawal decision. The only significant variable isthe percentage change in capacity utilization during the year immediatelyprior to the petition filing. The estimated coefficient implies that a fall in ca-pacity utilization increases the chances of the petition being withdrawn. Thecoetficient on this variable at the second stage, although insignificant, is alsonegative, providing only weak evidence that cases are withdrawn because ofself-selection. ~

The second set of regressions (models 7-11) concentrate on the secondstage decision, controlling for the first stage decision using the number ofcotintries dummy, employment, and percentage change in capacity utilizationduring the prior year. The analysis of the second-stage decision reveals thatI I) the economic criteria do relatively poorly predicting the decision, (2) theSenate oversight effect is not significant, and (3) the ITA’s final duty marginis an important predictor of the injury decision.

Consider first the set of economic criteria. In all the specifications the per-centage change in employment and capacity utilization perform poorly. Most~l the coe~cients are insignificant. This is consistent with both the HansenI1990) and Moore (1988) studies, suggesting that although the commissionersstate that these economic criteria are important, it is difficult to measure theirm~portance. In contrast, in all the specifications the final dumping margin has

als~ tested for the significance of the House oversight committee dummy. Results wereidentical to the Senate dummy and are available on request.

~b,o ran other specificat>ns to test for the importance of the other economic cdteria (thecapaciu utilization and the concentrat~oa ratio), but they were not significant, Results are

68 Thomas J. Prusa

a positive and significant effect on the outcome. This is a bit surprising sincethe simple cross-tabulation results (table 2.7) suggested that there should belittle effect. Note further that the final dumping margin is the one key variablenot known during most of the investigatory period. In model 8 the number ofcountries involved in the filing is included in the specification. It is negativebut insignificant. A negative coefficient implies that, conditional on the casenot being withdrawn, multiple petitions decrease the chance of protection.Model 9 estimates the steel industry’s effect, while model 10 estimates theSenate oversight effect; neither are significant. In contrast with the Hansen(1990) and Moore (1988) studies, I do not find that the Senate oversight com-mittee exerts influence on the ITC’s final decision. This result, along with theresults of models 4 and 5, suggests that political pressure is most significantat the first stage. This is consistent with Baldwin’s (1985) conjecture that theITC’s structure insulates it from political pressure. Finally, in model 11 I findthat the levels of capacity utilization, employment, and concentration ratio donot affect the outcome.

2.5 Concluding Comments and Interpretation of Results

In this paper I have argued that the history of U.S. antidumping usage re-quires we model the process as a two-stage problem. Given that approxi-mately 25 percent of petitions are withdrawn (usually with some type of pro-tection), a true understanding demands that we analyze the first-stagewithdrawal decision. Because protection granted via a price/quota agreementappears to be so desirable, we must analyze whether the failure to arrange anagreement is a signal of the eventual ITC decision.

The key insight gained from the analysis in section 2.4 is that industry sizeand the number of petitions filed are the key determinants of the first-stagedecision. Although the simple cross-tabulations (section 2.3) suggested thateconomic criteria influence the withdrawal decision, I found very little econ-ometric evidence that the economic criteria influence the withdrawal decision.This suggests that the withdrawal decision is chiefly influenced by politicalpressure. Moreover, (1) the steel industry dummy, (2) number of countriesinvolved in the petition, and (3) industry employment are all significant deter-minants of the withdrawal decision but are not significant determinants of theinjury decision. This, too, is consistent with the political pressure hypothesisbut runs counter to the self-selection hypothesis.

Furthermore, the analysis in section 2.4 also suggests that the economiccriteria are not significant predictors of the injury decision. Although this re-sult is a bit paradoxical, especially in light of the arguments and discussionfound in the ITC’s case reports, it is consistent with both Moore’s (1988) andHansen’s (1990) findings, which suggests that it is difficult to measure theeconomic criteria (if any) that are truly related to injury. As discussed earlier,there are most likely other variables such as import market share, exports, and

Seleclion of Antidumping Cas

inventory that are used by the ITClhcse variables precluded their iwould surely benefit if these variail~ of these variables also makes tl~hesc "’missing" economic criteriacriteria, then this result, that eco~x~ill not be altered by the inclusi~However, if the excluded variabhalready used, then this result may ~

The fact that I find no congmssision is an important and interesti~Moore (1988) and Hansen (1990).the data sets--Moore’s analysis ~cases while Hansen’s analysis cornactions with antidumping cases--general insights about the influenc~

If cases are withdrawn becau~selection, then the welfare implic;Antidumping law was conceived atential threat of predatory pricing.stantial size can strategically mani1to gain protection when none maymust be shown can be circumventeicant.

This analysis, along with the earPrusa (1988), suggests that the filiras strategically motivated as it isfindings will encourage others to ftof, such behavior.

Data Appendix

1. Basic case information such ascountry, and the number (and type)Guide to Antidumping Findings an,seven-digit TSUSA code. TSUSA cnotices that accompany each petitioto the TSUSA code can be found inGary Horlick also provided assistartions involved officially sanctioned ~

2. Capacity utilization (practical


][’his is a bit surprising sinceggested that there should be,argin is the one key variable,d. In model 8 the number of~ specification. It is negativethat, conditional on the casee the chance of protection.i~ile model 10 estimates the[n contrast with the Hansen,at the Senate oversight com-~. This result, along with the! pressure is most significant~’s (1985) conjecture that the,:. Finally, in model 11 I findtt, and concentration ratio do

m of Results

U.S. antidumping usage re-,~blem. Given that approxi-;tally with some type of pro-we analyze the first-stagevia a price/quota agreement

’ her the failure to arrange an

tion 2.4 is that industry size’.erminants of the first-stagesection 2.3) suggested that

~on, 1 found very little econ-~ce the withdrawal decision.:icily influenced by politically, (2) number of countries

~ent are all significant deter-nificant determinants of the,olitical pressure hypothesis

suggests that the economicdecision. Although this re-

: arguments and discussionth both Moore’s (1988) andis difficult to measure the

iniur.v As discussed earlier,~ market share, exports, and

inventory that are used by the ITC in determining injury. The unavailability ofthese variables precluded their inclusion in the analysis. Future analysiswould surely benefit if these variables could be constructed. The unavailabil-ity of these variables also makes the results a bit more difficult to interpret. Ifthese "missing" economic criteria are well proxied by the "known" economiccriteria, then this result, that economic criteria have little predictive power,will not be altered by the inclusion of these additional economic variables.However, if the excluded variables are not related to the economic criteriaalready used, then this result may not be robust.

The fact that I find no congressional oversight influence on the injury deci-sion is an important and interesting difference between this work and that ofMoore (1988) and Han, sen (1990). This is most likely due to the differences inthe data sets--Moore s analysis includes only a subset of all antidumpingcases while Hansen’s analysis combines countervailing duty and escape clauseactions with antidumping cases--which suggests that it is too early for anygeneral insights about the influence of congressional oversight committees.

If cases are withdrawn because of political pressure rather than self-selection, then the welfare implications of such withdrawals are disturbing.Antidumping law was conceived as a protective measure to eliminate the po-tential threat of predatory pricing. However, it appears that industries of sub-stantial size can strategically manipulate the law (by filing multiple petitions)to gain protection when none may be warranted. The requirement that injurymust be shown can be circumvented if the potential political fall-out is signif-icant.

This analysis, along with the earlier findings of Messerlin (1988, 1989) andPrusa (1988), suggests that the filing of an antidumping petition may often beas strategically motivated as it is economically motivated. Hopefully thesefindings will encourage others to further study the reasons for, and the effectsoL such behavior.

Data Appendixt. Basic case information such as the outcome, date of initiation, subject,country, and the number (and type) of petitioners was found in the Fed-TrackGtdde to Antidumping Findings and Orders. Products are identified by theirseven-digit TSUSA code. TSUSA codes can be found in the Federal Registernotices that accompany each petition. The four-digit SIC code correspondingto the TSUSA code can be found in U.S. Foreign Trade Statistics, Schedule 6.Gary Horlick also provided assistance in determining which withdrawn peti-tions involved officially sanctioned agreements.

2. Capacity utilization (practical rate) at the four-digit SIC level by year

70 Thomas J. Prusa

was obtained from U.S. Bureau of the Census Current Industrial Reports,Survey of Plant Capacio:

3. Total employment at the tour-digit SIC level by year was obtained fromU.S. Bureau of the Census, Census of Manufactures, Subject Series.

4. Four-firm concentration ratios at the four-digit SIC level was obtainedfrom U.S. Bureau of the Census, Census of Manufactures, Industry Series.

5. Congressional influence was measured by matching (four-digit SIC) in-dustry location with congressional districts. Typically, each product (identi-fied by SIC code) is produced in a number of regions in the country. If aproduct is produced in a district whose congressional representative (House ofRepresentatives or Senate) is a member of the Trade Subcommittee of theHouse Ways and Means Committee or the International Trade Subcommitteeof the Sena’:e Finance Committee, then I considered there to be a potentialpolitical pressure from that congressman. The Almanac of American Politicswas used to determine subcommittee membership. Data for industry locationby district and year at the four-digit SIC level were obtained from the Censusof Manufactures, Geographic Area Series. An industry had to employ at leasta thousand people in a district to be considered as potentially exerting politicalinfluence.

References

Almanac of American Politics. 1988. Ahnanac of American politics. Crawfordsville,Ind.: National Journal, Inc.

Baldwin, R. E. 1985. The political economy ofU.S, import policy. Cambridge, Mass.:MIT Press. "

Bhagwati, J. 1988. Protectionism. Cambridge, Mass.: MIT Press.Caves, R. A. 1976. Economic models of political choice: Canada’s tariff structure.

Canadian Journal of Economics 9(2): 278-300.

Fed-Track. 1988. Fed-Track guide to antidumping findings and orders. Washington,D.C.: Fed-Track.

Hansen, W. L. 1990. The International Trade Commission and the politics of protec-tionism. American Political Science Review 84( 1): 21-46.

Hausman, J., and D. McFadden. 1984. Specification tests for the multinomial logitmodel. Econotnetrica 52:1219-40.

Howell, T. R., W. A. Noellert, J. G. Kreier, and A. W. Wolff. 1988. Steel and thestate. Boulder, Colo.: Westview Press.

Jackson, J. H., and E. A. Vermulst. 1989. Antidumping law and practice. Ann Arbor:University of Michigan Press.

McFadden, D. 1978. Modelling the choice of residential location. In Spatial interac-tion theory and residential location, ed. A. Karlquist, 75-96. Amsterdam: North-Holland.

Maddala, G. S. 1983. Limited-dependent and qualitative variables in econometrics.Cambridge: Cambridge University Press.

71 Selection " ’ "ol Ant|dumpi~

Mc~serlin. P. A. 1988. The EC~98(1-85. World Bank, Wash.

--- 1989. GATT-inconsiste~evolation qf the EC antidumpt

Moore, M. 1988. Rules or polilsions. Department of Econom

Page. S. 1987. The rise in prote3t 1 ): 37-51.

Priest, G. L., and B. Klein. 19~Legal Studies 13:1-55.

Prusa, T. J. 1988. Why are so mEconomics, State University o

Slaiger, R. W., and F. A. Wolal,[ernational collusion. Typescri

--- 1991. The strategic effetacit foreign collusion. Journa~

U.S. Bureau of the Census. 19~U.S. Department of Commerct

U.S. Bureau of the Census. 1983Department of Commerce.

U.S. Bureau of the Census. 198"Department of Commerce.

U.S. International Trade CommiD.C.: U.S. International Trade

Weingast, B. R. 1981. Regulatio~dations of agency clientele relal

~ 1984. The congressionalwith applications to the SEC. P

Weingast, B. R., and M. J. Mo~control? Regulatory policym~Political Economy 91:765-800.

Comment Robert M.

Let me begin by saying that I liespecially important aspect ofclearly, organizes the data wellsible manner.

Most of my remarks are desiland policy issues involved in asome questions about Prusa’s ninterpretation of results.

In order to set Prusa’s paperconstitute by far the most intens

Robert M. Stern is professor of econand Institute of Public Policy Studies at

Current Industrial Reports,

~’1 by year was obtained from~res, Subject Series.digit SIC level was obtained~tfactures, Industry Series.matching (four-digit SIC) in-fically, each product (identi-regions in the country. If a

onal representative (House ofTrade Subcommittee of the

aational Trade Subcommitteedered there to be a potential\hnanac of American Politicsip. Data for industry locationere obtained from the Census~dustry had to employ at leasts potentially exerting political

~terican politics. Crawfordsville,

report policy. Cambridge, Mass.:

. : MIT Press.choice: Canada’s tariff structure.

~Mings and orders. Washington,

,~ission and the politics of protec-21-46.m tests for the multinomial logit

\. W. \Volff. 1988. Steel and the

-,rag law and practice. Ann Arbor:

:ntial location, in Spatial imerac-!uist. 75-96. Amsterdam: North-

,.tative v~¢riables in econometrics.


Messerlin, P. A. 1988. The EC antidumping regulations: A first economic appraisal:1980-85. World Bank, Washington, D.C. Typescript.

~ 1989. GATl’-inconsistent outcomes of GATT-consistent laws: The long-termevolution of the EC antidumping laws. World Bank, Washington, D.C. Typescript.

Moore, M. 1988. Rules or politics? An empirical analysis of ITC antidumping deci-sions. Department of Economics, George Washington University. Typescript.

Page, S. 1987. The rise in protection since 1974. Oxford Review of Economic Policy,3(1): 3%51.

Priest, G. L., and B. Klein. 1984. The selection of disputes for litigation. Journal ofLegal Studies 13:1-55.

Prusa, T. J. 1988. Why are so many antidumping petitions withdrawn? Department ofEconomics, State University of New York, Stony Brook. Typescript.

Staiger, R. W., and F. A. Wolak. 1988. Antidumping law as enforcement of tacit in-ternational collusion. Typescript.

-- 1991. The strategic effects of domestic antidumping law in the presence oftacit foreign collusion. Journal of International Economics.

U.S. Bureau of the Census. 1980. U.S. foreign trade statistics. Washington, D.C.:U.S. Department of Commerce.

U.S. Bureau of the Census. 1983. Current industrial reports. Washington, D.C.: U.S.Department of Commerce.

U.S. Bureau of the Census. 1987. Census of manufactures. Washington, D.C.: U.S.Department of Commerce.

U.S. International Trade Commission. various issues. Annual Report. Washington,D.C.: U.S. International Trade Commission.

Weingast, B. R. 1981. Regulation, reregulation, and deregulation: The political foun-dations of agency clientele relations. Law and Contemporary Problems 44:147-77.

-- 1984. The congressional-bureaucratic system: A principal-agent perspectivewith applications to the SEC. Public Choice 41:147-92.

Weingast, B. R., and M. J. Moran. 1983. Bureaucratic discretion or congressionalcontrol? Regulatory policymaking by the Federal Trade Commission. Journal ofPolitical Economy 91:765-800.

Comment Robert M. Stem

Let me begin by saying that I liked this paper very much. It is dealing with anespecially important aspect of U.S. trade policies. Prusa sets up his analysisclearly, organizes the data well, and interprets his empirical results in a sen-sible manner.

Most of my remarks are designed to elaborate further some of the an.alyticaland policy issues involved in antidumping (AD) actions, but I will also raisesome questions about Prusa’s modeling decisions, selection of variables, andinterpretation of results.

In order to set Prusa’s paper in context, it is worth noting that AD actions,:onstitute by far the most intensively used of the available instruments of trade

R~hcvt M. Stern is professor of economics and public policy in the Department of EconomicsInstitute o( Public Policy Studies at the University of Michigan,

Documents

takes the final decision. This pair 2 The Selection of Antidumping Cases for ITC ...econweb.rutgers.edu/prusa/cv/04 - selection of cases.pdf · 2009. 2. 11. · ITC will make its