3.1.1 Article 9.3 DSU and Rule 20(2)(d)(i) of AB Working Procedures

Two preliminary procedural issues were raised on appeal and both were rejected by the AB. The EC alleged that the Panel acted inconsistently with DSU Article 9.3 by failing to harmonize the timetables of the proceedings in the Ecuador and US cases. The EC requested that the Panel's failure to harmonize the timetables ‘be reversed’.Footnote ¹⁷ The United States, on the other hand, as an appellee, contended that the EC's Notice of Appeal did not satisfy the requirements of Rule 20(2)(d)(i) of AB Working Procedures because the EC allegedly had failed to identify just which findings it deemed to be erroneous because it failed to mention the paragraph numbers of the US Panel Report to which the issues appealed related. The US requested that the AB dismiss the EC's appeal entirely.

In both cases, the AB continued its practice of rejecting procedural claims that are not supported by a showing of actual denial of due process. In the case of the EC's Article 9.3 claim, the AB held that ‘the mere possibility that due process rights of the European Communities could have been adversely affected by the Panel's decision to maintain separate timetables in these proceedings is not sufficient to establish that the due process rights of the European Communities have indeed been compromised’.Footnote ¹⁸ Similarly, in the United States's appeal, the AB ruled that ‘the deficiencies in the European Communities’ Notice of Appeal do not lead to dismissal of the European Communities' appeal', since even the US in the oral hearing had acknowledged that it had not been ‘prejudiced by the absence of paragraph numbers’.Footnote ¹⁹

3.1.2 The legal effect of the Understandings on Bananas and the use of estoppel in the WTO

The AB affirmed, for different reasons, the Panel's rejection of the EC's argument that the legal effect of the Understandings on Bananas was to preclude the complainants from initiating compliance proceedings. The dispute centered on the question whether the Understandings amounted to a ‘solution mutually acceptable to the parties to the dispute and consistent with the covered agreements’ within the meaning of Article 3.7 of the DSU, and, if so, whether such agreed solution could preclude the complainants from bringing their claims under Article 21.5 DSU.

The Panel had found that the complainants could be barred from initiating compliance proceedings only if the Understandings constituted a ‘positive solution and effective settlement of the dispute’. The AB disagreed with this legal test, but agreed with the Panel that the complainants were not barred from initiating compliance review.

Nothing in Article 3.7, the AB noted, ‘establishes a condition under which a party would be prevented from initiating compliance proceedings’.Footnote ²⁰ Members do not waive recourse to Article 21.5 review simply by entering into a mutually agreed solution absent ‘a clear indication in the agreement between the parties of a relinquishment of the right to have recourse to Article 21.5’.Footnote ²¹

In the course of its analysis, the AB expressed its disagreement with two intermediate conclusions of the Panel. The first was the Panel's reasoning that the conclusion of the Understandings after adoption of the original rulings and recommendations by the DSB was relevant to the question whether complainants were barred from bringing 21.5 review proceedings. ‘We see nothing in Article 3.7 or elsewhere in the DSU … that would preclude recourse to Article 21.5 proceedings after the adoption of recommendations and rulings by the DSB’, the AB said.Footnote ²² It went on to note that ‘the DSU itself clearly envisages the possibility of entering into mutually agreed solutions after recommendations and rulings are made by the DSB’.Footnote ²³ In addition, the AB disagreed with the Panel's reliance on ostensibly conflicting statements as to the legal nature of the Understandings made at the DSB after they were signed.Footnote ²⁴ Such statements, the AB said, may be taken into account where the interpretation of a document is not clear, but they have limited relevance when documents are unambiguous. ‘[E]x post communications of the parties concerning the Understandings have, at best, slight evidentiary value.’Footnote ²⁵

While the AB used several paragraphsFootnote ²⁶ to discuss the text of the two Understandings, its main point was that, in order for a party to be barred from recourse to Article 21.5, the mutually agreed solution must explicitly state that the complainant relinquishes that right.Footnote ²⁷ Thus, even though the Understanding involving Ecuador declared in paragraph G that ‘[t]he EC and Ecuador consider that this Understanding constitutes a mutually agreed solution to the banana dispute’, this statement did not entail a waiver of its Article 21.5 review rights by Ecuador.Footnote ²⁸

Finally, the AB agreed with the EC that the Panel had erred in its interpretation and application of the ‘good faith’ requirement of Article 3.10 DSU, but upheld the Panel's conclusion that Ecuador and the US had acted in good faith in bringing the review proceeding. According to the AB, its Report in US–Offset Act (Byrd Amendment) was misread by the Panel.Footnote ²⁹ In US–Byrd, the AB noted, it had addressed the principle of good faith as it relates to a substantive provision. In Bananas III, the case before it now, on the other hand, the Panel and the AB had to consider the principle as a procedural impediment for a Member to initiate review proceedings.Footnote ³⁰

In the course of its discussion on this point, the AB stated that, while the EC did not use the term, it was in fact advancing an estoppel argument.Footnote ³¹ It quoted from its Report in EC–Export Subsidies on Sugar,Footnote ³² that ‘even assuming arguendo that the principle of estoppel could apply in the WTO, its application would fall within these narrow parameters set out in the DSU’.Footnote ³³ The ‘narrow parameters’ referred to are those of DSU Article 3.10 regarding good faith in deciding to have recourse to dispute settlement.

The estoppel allegations in both EC–Sugar and Bananas III involved the question of whether a complaining Member should be barred from proceeding with its complaint. However, in the context of working procedures for a dispute-settlement proceeding, the two Panels and the AB have employed the estoppel principle, but – as with the EC in Bananas III – without using the term.Footnote ³⁴ Thus, for example, in US–Steel Plate, the Panel denied India the right to pursue a claim that it had explicitly abandoned in its first written submission.Footnote ³⁵ Similarly, in Mexico–Corn Syrup (Article 21.5 – US), the AB held that failure to pursue an objection to the absence of consultations before the DSB prevents a Member from arguing that the Panel was improperly established. ‘A Member that fails to raise its objections in a timely manner, notwithstanding one or more opportunities to do so, may be deemed to have waived its right to have a panel consider such objections.’Footnote ³⁶ Even though words such as ‘waiver’ are used instead of ‘estoppel’, decisions of this kind amount to estoppel within the conduct of a case. This, of course, is a very different matter from holding that a Member is estopped from even initiating a proceeding in the first place.

3.1.3 ‘Repeal’ of the challenged measures

In the case of the United States's Report, the EC argued that the Panel was wrong in making findings with respect to a measure that had ceased to exist.Footnote ³⁷ The EC contended that the Panel had acted inconsistently with its obligation under Articles 3.4 and 3.7 of the DSU, which instruct panels to secure a positive solution to the dispute. In addition, the EC alleged that the Panel had provided a ‘concealed’ recommendation by stating that the original DSB recommendations and rulings ‘remained operative’.Footnote ³⁸

With respect to the Panel having made findings on an expired measure, the AB cited its earlier ruling in US–Upland Cotton, where it had held that ‘whether or not a measure is still in force is not dispositive of whether that measure is currently affecting the operation of any covered agreement’.Footnote ³⁹ The AB reminded that the present case is different from US–Upland Cotton in that the measure on appeal in these proceedings was still in full force at the time the Panel was established and expired only towards the end of the Panel proceedings.Footnote ⁴⁰ With respect to the EC's allegation that the Panel had made ‘concealed’ recommendations, the AB stated that it does ‘not believe that the Panel made a specific recommendation’, reminding that the Panel in its Report in the US 21.5 challenge (at para. 8.13) had stated that it ‘makes no new recommendation’ and that the ‘original DSB recommendations and rulings in this dispute remain operative’.Footnote ⁴¹ Consequently, the AB deemed it within the discretion of a panel to decide how it takes into account subsequent modifications or a repeal of the measures at issue.

The authors of this commentary believe that the AB was right in upholding Panel findings on the expired measure. If repealed policies could not be decided upon by WTO dispute panels, then Members would have an easy time engaging in strategic gamesmanship by entering into a repeal/amendment strategy, where the amended measure nominally replaces its ancestor yet shares most of its characteristics. A Member would never be able to challenge the measure at issue – simply because the respondent would make sure that it is not ‘at issue’ anymore, thanks to repeal and amendment. A decision with respect to an expired measure is also efficient, since it makes consecutive ‘measures taken to comply’ vulnerable in later 21.5 proceedings if it is too close to the original measure.

3.1.4 Legal effect of Panel suggestions and the relationship of Articles 19.1 and 21.5 of the DSU

In the Ecuador case, the EC requested that the AB dismiss Ecuador's claims under Article XIII, because the latter's ‘claims are in reality a challenge on the measures suggested by the [first Ecuador Article 21.5] Panel, rather than on the measures actually taken by the European Communities’.Footnote ⁴² The EC maintained that, once a Panel or AB report, containing suggestions made pursuant to Article 19.1 DSU, has been adopted, ‘the consistency of the measures suggested by the original panel with the covered agreement cannot be challenged by the complaining party before an Article 21.5 panel’, because parties are obliged to ‘unconditionally accept’ the DSB's recommendations, rulings, and suggestions, by virtue of Article 17.14 read in combination with Article 16.4 DSU.Footnote ⁴³ According to the EU, the compliance Panel should have first checked if the challenged measures were indeed the measures suggested by the original panel. And if the EC has effectively implemented any of the suggestions made by the first compliance Panel, the Panel should rule ‘without any further analysis’ that these measures are consistent with the covered agreements.

The AB assessed the legal nature of the Panel or AB suggestions pursuant to Article 19.1 DSU and found that suggestions, while potentially providing useful guidance, following a Panel suggestion does not guarantee substantive compliance with the rulings and recommendations by the DSB. Interpreting the language of Articles 19.1 and 21.5 of the DSU, the AB ruled that suggestions are not themselves the subject of review and a compliance Panel is limited to the assessment of the existence or consistency with the covered agreements of the measures taken to comply with the rulings and recommendations, but not of previous suggestions made. Hence, the AB upheld the Panel's finding that the adoption by a Member of a measure that directly implements a suggestion made by a Panel pursuant to Article 19.1 DSU does not prevent a complaining party from challenging the consistency of that measure with the covered agreements in an Article 21.5 proceeding.

Article 19.1 DSU authorizes Panels and the AB to ‘suggest ways in which the Member concerned’ may come into compliance. This rarely used authority was utilized by the first Article 21.5 Panel requested by Ecuador.Footnote ⁴⁴ Based on this suggestion, the EC argued to the AB that Ecuador's claim under Article XIII GATT should be dismissed. According to the EC, the WTO consistency of measures suggested by a Panel cannot be challenged in a compliance review. The EC reasoned that adoption of reports under Articles 16.4 and 17.14 of the DSU covers all the recommendations and rulings in the report, including any suggestions as to compliance.Footnote ⁴⁵

The EC's argument was rejected by the AB. Article 19.1 DSU, it noted, neither directly addresses the legal status of suggestions, nor are they subject to review.Footnote ⁴⁶ However, ‘“measures taken to comply with the recommendations and rulings”, but not “measures taken to comply with suggestions”’ are subject to Article 21.5 review, regardless of whether those measures were suggested.Footnote ⁴⁷ ‘The adoption of a panel or Appellate Body report by the DSB makes the recommendations and rulings therein binding upon the parties’, the AB held. But ‘adoption by the DSB does not make suggestions for implementation binding upon the parties’.Footnote ⁴⁸ ‘Even if the measure taken to comply conformed to a suggestion made, this would not bar [a Member] from bringing Article 21.5 proceedings to determine whether the implementing measure achieves full compliance with the DSB recommendations and rulings.’Footnote ⁴⁹ Further, the Appellate Body noted that ‘the fact that a Member has chosen to follow a suggestion does not create a presumption of compliance in Article 21.5 proceedings’.Footnote ⁵⁰

We believe the AB's reasoning is sound. Full compliance with DSB rulings cannot simply be presumed just because the respondent declares that its measure taken to comply conforms to the Panel's suggestions under Article 19.1 DSU.Footnote ⁵¹ This would unduly change the nature of the obligation from one of complying with the rulings and recommendations, based on certain WTO provision(s) to implementing compliance suggestions. Also, the parties see compliance suggestions for the first time when they receive the Panel's report. They have no opportunity to address – much less to challenge – a proposed suggestion. Yet, a suggestion that might seem to result in compliance may, upon further examination, fail to do so.Footnote ⁵² A complaining Member should not be denied ‘its day in court’ to challenge directly the question of whether a measure taken to comply in fact does so, whether or not that measure has been suggested.

Petros Mavroidis suggests that, irrespective of the legal force of a suggestion, a WTO Member that has accepted a suggestion should be presumed to be in compliance.Footnote ⁵³ Provided this is a rebuttable presumption, we would agree. But such a presumption would seem to be of little, if any, benefit to the Member concerned, since a complaining Member challenging a measure taken to comply effectively already has the burden of proving non-compliance in any event. True, if a compliance Panel is composed of the same individuals who made the suggestion, a complaining Member might, as a practical matter, face an uphill challenge. At the AB stage, the complaining Member might face an even more difficult challenge because there is no further appeal. This suggests that the AB has been wise not to make suggestions, in part because of its short timetables, when it has not had the opportunity to benefit from briefings and arguments from the parties on any particular suggestion.

An irrebuttable presumption would deny complaining Members a fair opportunity to challenge the substance of a measure taken to comply that follows a suggestion and clearly does not seem to be justified by the terms of Article 19.1.

3.1.5 Article 3.8 DSU and nullification or impairment suffered by the US and Ecuador

As a final procedural issue, the EC requested that the AB reverse the Panel's legal interpretations that the ACP preference in the Banana Import Regime had caused nullification or impairment (NoI) to Ecuador or the United States. In the case of Ecuador, the EC argued that the quantity limitation on ACP imports entering at zero-tariff was actually a ‘benefit’ to Ecuador, because it put a ‘cap’ on the preferential treatment of ACP countries and consequently limited the amount of duty-free ACP banana imports. In addition, the EC argued that the Panel had failed to explain how the measure at issue had caused NoI over and above that already found under GATT Article I (a finding that the EC had not appealed). According to the EC, the Panel engaged in ‘double counting’ of NoI by adding to the negative effects caused by the ACP preference under Article I GATT the ‘limitation of these negative effects through the quantity limit’.Footnote ⁵⁴

The AB dismissed the EC's claims, stating that both infringements of Article I and Article XIII GATT trigger the presumption of NoI. Nullification or impairment resulting from inconsistencies with Articles I and XIII GATT may thereby ‘coincide’ or ‘overlap’.Footnote ⁵⁵ However, such coincidence and overlap is only relevant to the calculation of the total level of NoI suffered for the purposes of an arbitration pursuant to DSU Article 22.6. According to the AB, it is not for the AB to decide on the extent of such overlap, and a demonstration by the EC that NoI may overlap is not sufficient to rebut the presumption in Article 3.8 DSU that any of these infringements constitutes a prima facie case of NoI.

In the case of the United States, the EC claimed that the Panel had confused the notion of ‘nullification or impairment’ with that of ‘legal standing’ in that the US had never been deprived of any competitive opportunities to export bananas. The EC pointed out that the US had always been a net importer of bananas and had never exported bananas to any country in the world, let alone the EU.

The AB agreed with the EU that ‘standing’ is a broader concept than NoI, but concluded that the Panel did not confuse those two issues. It ruled that it was insufficient for the EC to allege that the Panel had failed to explain the effect on the US internal market, noting that the burden of rebutting the presumption resides with the EC. Citing US–Superfund, the AB supported the Panel's finding that ‘the arguments advanced by the European Communities on the alleged lack of nullification or impairment have not rendered irrelevant the considerations made by the panel and by the Appellate Body in the course of the original proceedings, regarding the actual and potential trade interests of the United States in this dispute’. The AB agreed with the earlier panels that the US is a producer of bananas and that a potential export interest by the United States thus cannot be excluded. It also held that ‘the internal market of the United States for bananas could be affected by the EC bananas regime and by its effects on world supplies and world prices of bananas’.Footnote ⁵⁶

This conclusion is consistent with earlier jurisprudence of both the AB and Panels. In the initial Bananas III proceedings, for example, both the original panel and the AB had found that it is not necessary for a Member to have a ‘legal interest’ as a prerequisite to initiating dispute settlement.Footnote ⁵⁷ The Panel in Korea–Dairy interpreted ‘legal interest’ to mean ‘economic interest’ and concluded that ‘under the DSU there is no requirement that parties have an economic interest’.Footnote ⁵⁸

We end this section with a general observation from the AB's examination of procedural issues: at various instances, the AB digressed and examined issues that were not pertinent for reaching its conclusions.Footnote ⁵⁹ The question whether a tribunal's ruling should be narrowly focused on the facts of the case before it, or whether it should be broader than the minimum necessary to decide the case, has long been a matter of jurisprudential dispute. There are advantages and disadvantages to each approach.

A judicially modest, narrow ruling has the advantage of not anticipating problems that have not yet arisen and not foreclosing argument that might arise when those problems are squarely presented. It allows fine distinctions to be drawn between different factual situations, and reduces risk of unintended and unanticipated consequences. On the other hand, it can lead to uncertainty by leaving parties in the dark as to how a tribunal might rule in future cases with slightly different factual situations. It therefore can result in further, extended, litigation to settle outstanding questions.

A broad ruling, on the other hand, has the advantage of providing guidance to interested parties, thus reducing the need for further litigation to resolve outstanding questions. But it risks making unanticipated errors as the tribunal moves into areas that were not thoroughly aired during the course of the dispute.

In Bananas III, the AB adopted both approaches. When considering whether the Understanding on Bananas was a ‘measure taken to comply’, the AB noted that ‘strictly speaking, we would not be required to assess whether the EC Bananas Import Regime fell within the scope of Article 21.5 because of a “particularly close relationship” to the declared measure taken to comply’.Footnote ⁶⁰ However, because the Panel made findings on this question, and because the parties argued it before the AB, the AB addressed the question and made additional findings.Footnote ⁶¹

At a later point in its opinion, discussed above, the AB found that it did not need to decide whether nullification and impairment resulting from an inconsistency in quota allocation under Article XIII may coincide or overlap with that resulting from an inconsistency of a tariff preference with Article I:1.Footnote ⁶²

The AB was probably correct in both instances. Rather than adopt an iron-clad rule, it seems wiser to approach this question as the AB did, with an eye to the specific situation it faces. When dealing with a question that had been dealt with by the Panel and argued by the parties on appeal, it went ahead and offered its views on the issue. But when dealing with an issue implicating aspects of Article I, Article XIII, and tariff preferences that it did not have to decide, it wisely took a narrow, cautious approach.

3.2.1 Non-discriminatory administration of tariff quotas under Article XIII

On substance, the EC appealed the Panel's findings that the duty-free tariff quota reserved for ACP banana imports falls within the scope of Article XIII GATT and was inconsistent with Articles XIII:1, XIII:2, and XIII:2(d), which regulate the non-discriminatory application of quantitative restrictions and tariff quotas.

The EC essentially brought forth three arguments, all of which the AB rejected.

The first argument the EC made was that Article XIII does not apply at all in the case of tariff quotas, since Article I GATT prevailed over Article XIII in case of an overlap of the two provisions. The EC claimed that the Panel had developed a flawed theory pursuant to which a lower tariff offered to one Member automatically becomes a quantitative restriction on all other Members as soon as it was accompanied by a quantitative cap. According to the EC, such an interpretation would deprive the MFN principle of Article I:1 GATT of any value and would render WTO-inconsistent limitations on trade preferences offered by developed countries to developing countries under the ‘Enabling Clause’.

In response to the EC's claim, the AB assessed the relationship between Articles I and XIII GATT. According to the AB, tariff quotas thus must comply with both Articles I:1 and XIII. This protects Members against differential in-quota duties (thanks to Article I GATT), against discriminatory access to quota shares (Article XIII:1), and against discriminatory administration of quota allocation (Article XIII:2). In the absence of Article XIII, Article I would not provide specific guidance on how to administer tariff quotas in a non-discriminatory fashion in the allocation of quota shares. Hence, the AB determined that with respect to tariff quotas the application of Article I does not prejudge the application of Article XIII.

As its second argument, the EC challenged the applicability of Article XIII:1 because of the identity of the party affected by the ‘restriction’ in place. According to the EU, the Panel had misinterpreted the notion of ‘quantitative restriction’ contained in Article XIII:1 because the limitation on the tariff preference for ACP suppliers did not constitute a restriction imposed on the imports of the ‘aggrieved Member’ (i.e. on imports originating in the complainants' territory), but on imports from third countries (namely the ACP countries). As a consequence, the EC contended that there was ‘no basis that would allow an examination of whether “similar” quantitative restrictions are also imposed on all other countries’.Footnote ⁶³

In addressing this question, the AB read Article XIII:1 with reference to a tariff quota and stated: ‘Article XIII:1 is rendered thus: no tariff quota shall be applied by a Member on the importation of any product … of any other Member, unless the importation of the like product of all third countries is similarly made subject to the tariff quota.’ Taking a broad interpretation of the term ‘restriction’, the AB ruled that imports of like products of all third countries must have access to, and be given an opportunity for, participation.Footnote ⁶⁴ Applying Article XIII:1 to the facts of the case, the AB ruled that since the zero-tariff quota reserved for ACP countries plainly excluded non-ACP countries, the tariff quota did not equally apply to, or similarly restrict imports of, like products from MFN countries, thus offending the principle of access to, and participation in, a tariff quota enshrined in Article XIII:1.Footnote ⁶⁵

The third argument brought by the EU was that Article XIII:2 on non-discriminatory distribution of tariff quotas did not apply to the EC import regime vis-à-vis Ecuador and the US, because both countries were not even subject to the Banana Import Regime in question. Since the contentious EC tariff quota only concerned imports from ACP countries, imports from the US and Ecuador were only subject to a simple tariff – an issue that should be dealt with under Article I of the GATT.

The AB explained that Article XIII:2 regulates the distribution of tariff quotas among Members. It noted that ‘the chapeau of Article XIII:2 requires that the tariff quota be distributed so as to serve the aim of a distribution of trade approaching as closely as possible the shares that various Members may be expected to obtain in the absence of the tariff quota’.Footnote ⁶⁶ The AB continued that Article XIII:2(d) on historical allocation of quota shares is a permissive safe haven: compliance with the requirements of Article XIII:2(d) is presumed to lead to a distribution of trade as foreseen in the chapeau of Article XIII:2. Examining the EU Banana Import Regime under these provisions, the AB stated that ACP preference failed to meet the requirements regarding the distribution and allocation of shares in Article XIII:2. It noted that the exclusion of non-ACP suppliers from the tariff quota is not aimed ‘at a distribution of trade … approaching as closely as possible the shares which the various Members might be expected to obtain in the absence of restrictions’, as required by Article XIII:2. It further ruled that ‘the exclusion of non-ACP suppliers [does not] respect the “safe harbour” allocation requirements in Article XIII:2(d) based upon the representative proportions of Members having a substantial interest in the supply of bananas to the European Communities’.Footnote ⁶⁷

The authors of this paper have four brief comments on the AB's examination of Article XIII GATT. First, the AB was right in arguing that tariff quotas, being quotas and tariffs at the same time, must comply with the requirements of Article I:1, as well as with the non-discrimination obligation of Article XIII. Only if these provisions act in concert can abusive tariff quotas be effectively avoided: Article XIII:1 safeguards equal access to, or participation in, a certain tariff-quota regime. Once all MFN countries have equal access to a tariff quota, equal in-quota treatment must be ensured. This consists of uniform in-quota tariffs, and of equitable quota distribution among exporting countries. Article I:1 GATT safeguards the former; Article XIII:2 the latter. Hence, Articles I and XIII apply to different elements of an import measure.

Second, we believe that the AB was right in interpreting the non-discrimination tenet for quota allocation as a most-favored nation (MFN) treatment principle.Footnote ⁶⁸ In the context of tariff quotas, non-discrimination could theoretically be interpreted as: (i) granting equitable shares to all (current and potential) exporters; (ii) allotting the global quota in proportion to current suppliers; (iii) first-come, first-served regime of quota filling; or (iv) as an MFN regime according to the free-trade counterfactual (see Skully, Reference Skully2001).Footnote ⁶⁹ However, we believe that only the MFN interpretation can satisfy the free-trade counterfactual enshrined in the chapeau of Article XIII:2, which states that ‘[i]n applying import restrictions … contracting parties shall aim at a distribution of trade in such product approaching as closely as possible the shares which the various contracting parties might be expected to obtain in the absence of such restrictions’. Also, the language used in the Agreement on Safeguards, particularly in Article 5.2 on the application of safeguard measures, is nearly identical to certain passages in Article XIII. As many authors have argued (e.g., Jackson, Reference Jackson1997), the Agreement on Safeguards mandates MFN treatment of all exporters.Footnote ⁷⁰ Finally, if the wording of Article XIII was meant to be an explicit departure from the fundamental WTO (and economic) principle of non-discrimination among a Member's trading partners, the founding fathers of the WTO should have made such deviation more explicit. However, nothing in the preparatory work of the WTO is suggestive of that (Hudec, Reference Hudec, Jackson and Sykes1997).

Our third comment is that despite this de jure MFN obligation for quota allocation in Article XIII:2, Article XIII:2(d) allows for de facto discrimination of certain suppliers and therefore is in striking contrast to the principles laid out in the chapeau of Article XIII:2. From an economic perspective we have to state: an alleged MFN tenet under the chapeau of Article XIII:2 cannot be reconciled with the ‘safe haven’ of historical quota allocation of Article XIII:2(d). This will be shown in more detail below (Section 4.1.3).

Finally, the AB was correct in rebutting the EC's peculiar notion of ‘discriminatory’ vs. ‘preferential’ quantitative restrictions (QR), which implies that an import regime falls under the purview of Article XIII, only if the complainant is a direct subject of a QR and therefore ‘aggrieved’. From a political-economic perspective, such view is baseless. Given that imports under any QR are a zero-sum game among competing exporters, it is obvious that a ‘preferential’ QR to one set of countries – here ACP suppliers – necessarily goes to the detriment of all other countries – here MFN suppliers. Members excluded from a preferential QR necessarily are ‘restricted’ and therefore suffer losses relative to a situation without QR.

3.2.2 Doha Article I Waiver as ‘subsequent agreement’

The EC Banana Import Regime provided for an MFN tariff of €176/mt on all banana imports that did not fall under the ACP tariff quota of 775,000 mt at zero tariff. In the compliance proceedings, Ecuador had requested the Panel to find WTO-inconsistency with that provision. According to Ecuador, this provision was inconsistent with the EC's tariff schedule, that is its obligations under Article II GATT, which contained an MFN tariff quota of 2.2 million mt with an in-quota tariff rate of only €75/mt. The compliance Panel had found that while the MFN tariff quota and the accompanying tariff rate of €75 had undeniably expired on 31 December 2002 (under the terms of paragraph 9 of the Banana Framework Agreement; see above), the tariff quota nevertheless had been extended through ‘common intention’ of all WTO Members. According to the Panel, such common intension had manifested itself in the Article I Waiver. The Panel had held that the Article I Waiver modified the Schedule of the EC, and therefore was to be considered a ‘subsequent agreement regarding the interpretation of the original treaty’ as mentioned in Article 31(3)(a) of the Vienna Convention on the Law of Treaties (Vienna Convention).Footnote ⁷¹

On appeal, the EC requested the AB to reverse this finding. It claimed that the Panel erred in determining that the Doha Article I Waiver was a subsequent agreement on the interpretation of the EC Tariff Schedule. The EC brought forth two arguments. First, even by the Panel's own standards, the Waiver did not interpret, but modify the EC Schedule by extending the tariff-quota concession beyond 31 December 2002. In addition, the Waiver did not contain any such ‘common intent’.Footnote ⁷²

In examining this issue, the AB set off with a general discussion on the ‘roles and functions’ of four basic methods that Members may use to interpret or modify WTO law: (i) waivers (Article IX:3 of the WTO Agreement); (ii) multilateral interpretations (Article IX:2); (iii) multilateral amendments (Article XI); (iv) tariff renegotiations (Article XXVIII GATT).

The AB found that waivers are specific and exceptional instruments. Their purpose is not to modify existing provisions or to create new law or add or amend obligations under a covered agreement, but to relieve certain WTO Members from obligations temporarily, subject to very tight timelines and notification requirements. Thus, the AB concluded that the Panel had erred in finding that the Article I Waiver was a subsequent agreement to the EC schedule, especially because the WTO has in stock a particular instrument dealing with tariff modification in the form of Article XXVIII.

Although it was not necessary to do so, the AB went on to examine whether there really existed a ‘common intent’ among parties in extending the tariff-quota concession in the EC schedule by means of the Doha Waiver.Footnote ⁷³ The AB did not see how the language of the Article I Waiver had expressly extended the MFN tariff quota of 2.2 million mt, the in-quota tariff rate of €75/mt, or the terms of the BFA. It noted that the ‘Doha Article I Waiver is concerned with the zero-duty preference for ACP suppliers, not with the tariff quota concession for MFN suppliers specified in the European Communities’ Schedule'.Footnote ⁷⁴ On this basis, the AB reversed the Panel's findings that the Doha Article I Waiver constituted a subsequent agreement reflecting the common intention of WTO Members that the EC tariff-quota concession would continue to constitute its scheduled commitments regarding bananas, pending the completion of the Article XXVIII negotiations.

We believe that the AB was right in finding that the Doha Article I Waiver was not a subsequent agreement on the interpretation of treaty provisions in the sense of Article 31(3) of the Vienna Convention. An extension of an MFN tariff quota clearly constitutes a temporary modification of the EC's Schedule of Concessions, and therewith by virtue of Article II:7 GATT, a modification of the covered agreements. However, this episode in Bananas III unveils a more systemic problem of the WTO, namely that temporary opt-out mechanisms are missing in the WTO (Schropp, Reference Schropp2009a).

The WTO is an incomplete contract among sovereign countries, where previously unforeseen contingencies or ‘shocks’ occur with a high probability (Horn et al., Reference Horn, Maggi and Staiger2009; Maggi and Staiger, Reference Maggi and Staiger2008). Unforeseen shocks can give rise to ‘regret contingencies’ (Goetz and Scott, Reference Goetz and Scott1981), and all parties to the contract could benefit from letting the affected party temporarily ‘escape’, or ‘opt out’ of, the agreement, subject to an adequate compensation payment. The WTO knows several de jure emergency escape mechanisms, such as GATT Articles XVIII, XIX, and XX. However, enactment of those emergency relief mechanisms is exacerbated by significant levels of conditionality, which greatly reduces their usability.Footnote ⁷⁵

In Bananas III, the AB held the view that ‘if the duration of the tariff quota concession in the European Communities’ Schedule had to be modified or extended, negotiations under Article XXVIII would be the proper procedure to follow'.Footnote ⁷⁶ However, given the elaborate and extensive procedure entailed in Article XXVIII, the tool of tariff renegotiation is evidently not designed for urgent temporary, but for permanent deviations:Footnote ⁷⁷ first, the procedure is excruciatingly complicated and thus renders emergency and temporary protection virtually impossible.Footnote ⁷⁸ Second, tariff renegotiations are an untargeted, non-discriminatory measure. Tariff concessions must be offered by the demanding Member on an MFN basis, thus reducing its value as an efficient emergency relief. Third, strategic gamesmanship on the part of Members affected by the proposed tariff change make speedy conclusions rather unlikely. The affected Members hold a substantial bargaining power vis-à-vis the requesting Member, if they know that the latter is under a sizeable pressure to temporarily opt out of its previously made concessions.Footnote ⁷⁹

In sum, whenever Members strive for temporary, transient, deviations from previously agreed concessions, the WTO legal framework proves slack and burdensome. In order to avoid that, Members make use of various de facto opt-out mechanisms (anti-dumping, anti-subsidy measures, breach of the agreement, voluntary export restraints, etc.); more flexible and user-friendly ways of compensated contractual escape should be found (Mahlstein and Schropp, Reference Mahlstein and Schropp2007; Schropp, Reference Schropp2009a).

3.2.3 Interpretation of the EC market-access commitments and expiry of the EC's ‘MFN’ tariff quota

Given that the AB reversed the Panel's findings on the interpretation of the legal nature of the Article I Waiver, the AB was asked to consider Ecuador's conditional appeal on the interpretation of the EC's market-access commitments and the expiry of its MFN tariff-quota concession beyond 31 December 2001. Ecuador submitted that the Panel had erred in finding that the expiry of the Bananas Framework Agreement had automatically implied the termination of the EC's tariff-quota concession vis-à-vis all MFN exporters under the terms of its Schedule. Ecuador held that paragraph 9 of the BFA (‘This agreement shall apply until 31 December 2001’) did not establish an expiration date for the entire concession with respect to bananas. According to Ecuador, paragraph 9 rather established expiry only of those terms and conditions in the BFA related to the renegotiation of quota shares (as Article XIII:2 requests Members to conduct). It thus argued that the EC's MFN tariff-quota concession remained in effect, and requested the AB to uphold the Panel's ultimate conclusion that ‘the tariff applied by the European Communities to MFN imports of bananas set at €176/mt … without consideration of the tariff quota for 2.2 million mt bound at an in-quota tariff rate of €75/mt, [is] in excess of that set forth … in the European Communities’ Schedule' and thus inconsistent with Article II:1(b) of the GATT.Footnote ⁸⁰

In addressing Ecuador's appeal, the AB held as a preliminary issue that there are ‘limits to the terms, conditions, or qualifications that may be incorporated in a Member's Schedule of concessions’, but that the scheduling of temporal limitations to a tariff concession in the BFA was not incompatible with prior dispute-settlement findings. To determine whether the EC's tariff-quota concession vis-à-vis all non-ACP countries had really expired on 31 December 2001, the AB examined all relevant parts of the EC Schedule, including the column entries in Part I, Section I-B of the EU's Schedule and the BFA annexed to it under ‘other terms and conditions’.

In Part I, Section I-B (entitled ‘Tariff Quotas’), replicated above, the EC Schedule contains a tariff-quota concession of 2.2 million mt at €75 in Column 4 (‘final quota quantity’). The BFA, referred to in Column 7 and annexed to the EC Schedule, repeats the in-quota amount and in-quota tariff (paragraphs 1 and 7), but states in paragraph 9: ‘This agreement shall apply until 31 December 2002.’ The AB claimed to give full meaning and effect to both the final tariff-quota concession in Column 4 and the temporary nature of the BFA referred to in paragraph 9, when it ruled that ‘the Bananas Framework Agreement constitutes an agreement on the allocation, management and reallocation of country-specific shares within the “global basic tariff quota” referred to in paragraph 1. Therefore, we are of the view that the sentence “[t]his agreement shall apply until 31 December 2002” in paragraph 9 refers to an agreement on the allocation of shares.’Footnote ⁸¹

The AB then turned to supplementary means of interpretation of the EC Schedule, specifically to the negotiating history of the EC tariff concessions and the BFA, the Uruguay Round ‘Modalities Paper’, the Doha Article I Waiver, and the EC request for tariff renegotiations under Article XXVIII. The AB largely found confirmation in its interpretation and concluded that the tariff-quota concession for 2.2 million mt bound at the in-quota rate of €75/mt remained in force beyond 31 December 2002 until the rebinding of the EC's Schedule of Concessions for bananas. Accordingly, the AB upheld the Panel's initial findings – albeit for different reasons – that the tariff applied by the EC to MFN imports on bananas, set at €176/mt, without consideration of the tariff quota of 2.2 million mt bound at an in-quota tariff of €75/mt was inconsistent with Article II:1(b) of the GATT.

The Modalities Papers used during the Uruguay Round negotiations expressly provide that they ‘shall not be used as a basis for dispute settlement proceedings’.Footnote ⁸² The EC, accordingly, challenged the Panel's use of the Uruguay Round Modalities Paper for agricultural market access as a means of supplementary interpretation. The AB disagreed with the EC.

The prohibition on the use of the Modalities Paper simply means, the Appellate Body said, that ‘it does not in itself confer on WTO Members’ rights and obligations enforceable in dispute settlement. However, this does not preclude references to the Modalities Paper when interpreting the WTO agreements and Members' Schedules of Concessions that were prepared in accordance with these modalities.'Footnote ⁸³

This statement was consistent with the Panel's conclusion in EC–Export Subsidies on Sugar in which the EC had referred to that Modalities Paper as a supplementary means of interpretation.Footnote ⁸⁴ However, in the Sugar dispute the AB had not found it necessary to rule on the relevance of the Modalities Paper.Footnote ⁸⁵ In the present Bananas III dispute, however, the AB quoted with seeming approval the statement of the Panel in Sugar:

This decision seems correct, but it does raise an issue concerning its possible impact on the – already tedious – negotiating process. Now that it is clear that preparatory documents, such as the Uruguay Round Modalities Papers, may be used as supplementary means of interpretation, despite their ‘shall not be used as a basis for dispute settlement proceedings’ language, agreement on modalities in current and future negotiations might become even more difficult to achieve.

4.1.1 A brief economic assessment of tariff quotas

A tariff-rate quota, or tariff quota, is a two-tiered tariff, consisting of a lower in-quota tariff (t), applied to the first Q units of imports, and a higher over-quota tariff (T).Footnote ⁸⁷ Tariff quotas are not considered quantitative restrictions, because technically they do not limit import quantities. Therefore, ‘tariff quotas do not fall under the prohibition of Article XI:1 and are in principle lawful under the GATT 1994’,Footnote ⁸⁸ since exporters may always import by paying the over-quota tariff. The precondition, however, is that the tariff rates are applied consistently with the MFN tenet of Article I GATT, that in-quota and out-of-quota tariffs do not exceed bound tariff concessions (Article II GATT), and that the quota administration is in accordance with the disciplines of Article XIII GATT.Footnote ⁸⁹ Yet, if the over-quota tariff makes it prohibitively expensive to import, the tariff quota yields the same import volume as a traditional quota.

To see that, consider Figure 1, which illustrates the operation of a tariff quota. Imports are plotted on the x-axis, prices on the y-axis. For simplicity, the effective supply curve of exports (S) into the importing country consists of two horizontal lines. The lower line represents the in-quota imports and extends from zero to point Q at the price P _t=w+t, whereby w is the prevailing world price and t is the in-quota tariff rate. At Q the supply curve makes a vertical jump to connect in-quota with over-quota imports. The upper part of the supply curve represents the effective supply of over-quota imports at the price P _T=w+T, where T is the over-quota tariff.

Figure 1. Tariff quotas and import demand

The effect of a tariff quota on trade depends on the excess demand for imports. Figure 1 shows three possible excess-demand conditions (labeled XD ₁₋₃). The intersections of XD ₁₋₃ with S fix three different import levels (points M ₁, M ₂, M ₃). At XD ₁, the quota tariff is binding, meaning that domestic excess demand is M ₁, which is less than the quota volume Q. The tariff quota functions like an ordinary tariff applied at the in-quota rate. At M ₂, the tariff quota is binding and Q limits the supply available in the domestic market at a price P _q. With a binding quota, both domestic quantity is less and prices are higher than what would prevail if the tariff quota did not exist and only a tariff were applied at the in-quota rate of t (point N). Without any tariff, quantities would be demanded at the free-trade level of F. Excess-demand curve XD ₃, finally, represents a high level of demand, sufficient to sustain imports at the over-quota tariff T. With a prevailing demand curve of XD ₃, T is binding, and M ₃ units are imported.

Some observations about tariff quotas: first, tariff quotas are inefficient whenever the tariff quota is binding. At import quantity Q, less imports are available compared to a tariff-only regime (point N) and a free-trade regime (point F). This leads to economic deadweight loss compared to the tariff-only and free-trade scenarios (illustrated by the shaded triangles ABC and BED, respectively). Secondly, tariff quotas produce exporter rents, whenever the quota Q or the out-of-quota tariff T is binding. The economic rent is illustrated by the rectangle ABP _qP _t, denoted ‘Rent’. Third, whenever there are rents to be seized, imports must be rationed.Footnote ⁹⁰ Rationing is the essence of tariff-quota administration (more on that below). Rationing is also at the core of additional inefficiencies, because (i) it allows otherwise non-competitive suppliers to export at a higher price and (ii) suppliers competing for rents may engage in wasteful activities, such as bribery.Footnote ⁹¹

As our final observation, to take up a point made above, depending on the out-of-quota tariff T, a tariff quota acts exactly like a quantitative restriction, with the only exception that at least parts of the economic rents accrue to the importing government in the form of tariff revenues (hatched rectangle AP _twE, denoted TR in Figure 1). In the original Bananas III dispute, the EC had an out-of-quota tariff of €680/mt of bananas. As was pointed out in Footnote footnote 4 above, this constituted a prohibitive tariff. Hence, the Original Banana Import Regime basically acted as a quantitative restriction, and therefore circumvented the obligations of Article XI GATT. The special status of tariff quotas is difficult to understand from a purely economic point of view, and is probably better explained by the political economy of the Uruguay Round negotiations.Footnote ⁹²

4.1.2 The distribution of trade under tariff quotas and the AB's test of Article XIII:2

The chapeau of Article XIII:2 of the GATT reads in pertinent parts:

This language implies the construction of the following counterfactual: how would the world look in the absence of such restrictions – in this case, in the absence of a tariff quota? The distribution of trade under this counterfactual shall then be used to compare the actual distribution of trade under the tariff quota to check for possible discrimination against exporting countries.Footnote ⁹³ Yet, Article XIII:2 does not state which counterfactual is pertinent. There are potentially three plausible counterfactuals implied by Article XIII:2: (i) a free-trade counterfactual, that is a situation that would exist absent any import restrictions; (ii) a tariff-equivalent counterfactual, that is a situation in which the same quantities were imported as in the presence of an in-quota tariff only; and (iii) a quota-equivalent counterfactual in which the imported quantities were limited in the same way as under the quota at issue. As will be argued below, each of these three counterfactuals could potentially be compliant with the MFN principle.

To see how these three counterfactuals differ, consider Figure 2, which is a slight variant of Figure 1, this time with an upward-sloping excess-supply curve (XS). The world price w is fixed where excess-demand and excess-supply curves intersect, setting a free-trade quantity of F. If the in-quota tariff were applied in the absence of the quantitative restriction, domestic consumers would demand a quantity of N units. Under a quota-equivalent counterfactual, the importing government would import the same quantity as under the quote (Q).

Figure 2. Three interpretations of the counterfactual construed in Article XIII:2

Importantly for the Bananas III case at hand, those three counterfactual scenarios involve substantially different distributions of trade. If trade were completely unrestricted, suppliers inframarginal to world price w would be willing to export to the domestic market.Footnote ⁹⁴ The composition of this group of suppliers, that is their export shares, would then be the basis for a non-discrimination test under Article XIII:2. In the tariff-equivalent counterfactual, where the imported quantity were limited to N units, only suppliers inframarginal to a fictitious price P _te would be willing to export their goods to the importing country. The composition of suppliers in this second scenario could potentially differ, because suppliers inframarginal to w, but extramarginal to P _te, would drop out of the relevant MFN supplier group. Even fewer suppliers, namely only the most competitive ones, would manage to export, if quantity was limited according to the quota-equivalent counterfactual (quantity Q at a fictitious price P _qe). The group of suppliers and therewith counterfactual export shares will again differ from the other two scenarios.

Figure 2 lists a hypothetical group of competing importers of bananas and stacks them in their order of competitiveness, that is according to their (hypothetical) cost of production.Footnote ⁹⁵ Imagine for the sake of argument that ACP countries were the most competitive suppliers, followed by Colombia, Ecuador, Peru, and the United States. Imagine now that the zero-tariff quota would be quite small, e.g. 10,000 mt of bananas.

Depending on the counterfactual chosen, a discrimination examination under Article XIII:2 yields strikingly different results. Under the free-trade scenario, all exporters except the United States would export to the country at issue. All those exporters should thus be granted a certain in-quota share. Under the tariff-equivalent counterfactual, Ecuador and Peru would drop out of consideration, and in-quota trade should be distributed among ACP members and Colombia only. Finally, taking the quota-equivalent counterfactual as the legal benchmark, a quota allocation to ACP countries would be fully justified.

The AB in its Report justified the EC's violation of Article XIII:2 with the following words:

We submit that we believe the AB's finding to be correct in substance. However, in the light of the above, the AB's analysis could be seen as slightly too cursory, because the AB did not state which counterfactual it applied in reaching its finding, and what the competitive position of ACP suppliers really is, as compared to their competitors. The EU could have argued for the quota-equivalent counterfactual and asserted that ACP countries are the only banana suppliers inframarginal to a hypothetical price of P _qe.Footnote ⁹⁷

4.1.3 Quota-allocation methods, distribution of trade, and the inherent contradictions in Article XIII:2 GATT

As the title suggests, Article XIII regulates the non-discriminatory administration of, inter alia, tariff quotas. The AB has opined that Article XIII:2 extends the MFN principle to quota allocation, an interpretation to which the authors of this comment agree.Footnote ⁹⁸ The principle of non-discrimination among all exporting Members asserts that trade shares should be determined by the relative efficiency of suppliers, but not by alternative, discriminatory criteria. The methods with which tariff quotas are allocated among different countries have a significant impact on the distribution of trade. WTO Members have notified at least six distinct allocation methods for their tariff quotas:Footnote ⁹⁹

• • applied tariffs,

• • auctions,

• • license on demand (LoD),

• • first-come, first-served (FCFS),

• • historical allocation (HA), and

• • discretionary allocation (either through state-trading enterprises or through producer groups),

whereby applied tariffs and auctions can be grouped together as ‘market allocation’ methods, while LoD, FCFS, and HA can be categorized into the broad group of ‘quasi-market allocation’ methods.

As will be argued below, only truly market-based allocation methods, namely applied tariffs, quota auctions, and markets for quota-rights resale satisfy the principle of distributive justice among WTO Members, while the design of all other methods is inherently discriminatory against the most competitive suppliers.Footnote ¹⁰⁰ Notably, historical allocation, the so-called ‘safe haven’ mentioned in Article XIII:2(d), is inherently flawed and therewith contradicts the chapeau of Article XIII:2.

4.1.3.2 Quasi-market allocation methodsFootnote ¹⁰²

License on demand (LoD), first come, first served (FSFS), and historical allocation (HA) are a mix of market allocation and a random lottery among certain exporters. All three methods result in discriminatory allocation of quotas; in the best case, their results match that of an auction, but such an outcome would be purely coincidental.

Under License on demand, exporters are invited to apply for licenses at the beginning of a given quota period, specifying the quantity of imports they would like to carry out. If domestic demand is sufficient, the quota binds. The sum of all quota requests is denoted as Q*. With Q*>Q, the quota binds. To ration export licenses, every bidder's application quantities are reduced proportionally by factor λ, so that Q=λQ*, whereby λ is a rationing factor, and 0<λ<1 holds.

This tariff-quota administration method is very likely to produce a biased distribution of trade that violates the non-discrimination tenet of Article XIII:2 GATT. Consider Figure 3, which shows the supply side of imports. The domestic price is P, the world price w, and the unit quota rent is R=P−w. S is an upward-sloping supply curve of imports into the country at issue.Footnote ¹⁰³ The area under the supply curve represents payments to factors employed to the product at issue. Exporters extramarginal to w must spend resources in excess of w to produce a unit of output. Hence, extramarginal production destroys value; extramarginal suppliers would not be able to import under free-trade conditions.

Source: Based on Skully (Reference Skully2001: Figure 5).

Figure 3. Tariff-quota allocation methods

LoD allocation is a form of lottery among all exporters inframarginal to the domestic price P (that is, inframarginal and extramarginal suppliers alike). They all have an incentive to enter into the quota lottery to capture some of the riskless quota rent R. Rational applicants, cognizant of the rationing, have an incentive to overstate their license requests. Thus, quantity applied for exceeds the in-quota supply: Q*=1>Q=. To deal with quota oversubscription, each applicant is granted a pro rata share of the global quota. The effective supply curve under the tariff quota thus skews leftwards (S ⁰), to reflect the random draw among all suppliers inframarginal to P. The proportion of quota rights granted to exporters inframarginal to w is β the proportion of competitive free-trade suppliers. LoD allocation causes the displacement of λ(Λ−β) inframarginal suppliers by extramarginal suppliers, who would not have supplied the import market under free trade. Hence, the expected distribution of trade differs from the free-trade counterfactual, which consists uniquely of inframarginal suppliers. The welfare loss from LoD allocation is shown by the shaded triangle to the right of S ⁰. The loss of welfare can be interpreted as an indicator of how the expected distribution of trade under LoD differs from the tariff-equivalent counterfactual distribution of trade.

First-come, first-served allocation allows importation at the in-quota tariff until the quota is filled. To analyze the distribution of trade under FCFS, some assumptions must be made about the correspondence between an exporter's willingness to supply and his place in the FCFS queue.Footnote ¹⁰⁴ We examine four scenarios: (a) an optimistic scenario in which the lower-cost exporter is always placed ahead of the less competitive supplier; (b) a realistic scenario in which the otherwise uncompetitive higher-cost supplier queues before the competitive exporter; and two intermediate scenarios in which (c) there is a zero correlation between cost and place in the queue and (d) there is a positive, but imperfect relationship.

Consider first scenario (a): the lower-cost exporters are always placed ahead of less competitive exporters; then the effective quota supply curve is exactly the original supply curve S in the interval (0, λ). Call this segment of the S-curve S*. However, because competitive suppliers can easily sell on the world markets at w, the world price becomes the lower bound on their willingness to supply the import market. Assuming that each inframarginal supplier is equally likely to supply the import market, a random drawing from the set of inframarginal suppliers is the result, yielding the effective quota supply curve S ⁺. Under this allocation, efficient discrimination between inframarginal low-cost and extramarginal high-cost exporters is achieved and the tenet of non-discrimination is adhered to. Yet note well that this scenario presupposes a perfect correlation between competitive rank and place in the queue. Such an outcome, if it ever happens, is very unlikely.

Consider the opposite situation, scenario (b), in which the highest-cost suppliers undertake efforts – and succeed – in queuing before low-cost suppliers.Footnote ¹⁰⁵ When low cost of production is inversely correlated with the position in the queue, the expected outcome may be represented by the supply curve termed S ⁻, in the far-left corner of Figure 3 (S ⁻ is a leftward shift of the extramarginal segment of curve S). From a welfare perspective, but also from a non-discrimination point of view, such an outcome is unwelcome, since all those inframarginal suppliers, who would have supplied the import market in the absence of the tariff quota, are crowded out by high-cost exporters. A completely biased distribution of trade is the result.Footnote ¹⁰⁶

Next, consider scenario (c) a zero correlation between cost and rank in the queue. As was discussed previously under the LoD allocation method, a random selection of inframarginal (competitive) and extramarginal (non-competitive) suppliers from a population uniformly distributed over the interval (C, P) has the expected distribution represented by the supply curve S ⁰.

Finally, an FCFS process with a positive, but imperfect, correlation (scenario (d)), will yield an expected allocation of quota rights between S ⁰ and S ⁺ over the range (0, λ). Hence, some expected inframarginal displacement will occur in this case of positive but imperfect correlation too, whereby the lower the correlation between low cost and rank in the queue, the greater the expected displacement (the closer the actual supply curve will be to S ⁰).

In sum, a tariff-quota allocation method of FCFS will lead to a displacement of between none and all inframarginal suppliers, depending on the correlation between comparative advantage and place in the queue. Even if there is a zero correlation (an unrealistically optimistic suggestion), a fraction of λ(Λ−β) inframarginal suppliers will get crowded out by otherwise uncompetitive producers.

Historical allocation finally can be viewed as an extreme variant of an FCFS or LoD allocation. Whereas FCFS and LoD are annual lotteries, historical allocation of quota shares is essentially a one-time-only draw. A particular realization is sustained for many years and remains invariant to changes in underlying market conditions. Examining the US sugar import regime ever since 1934, Skully (Reference Skully1998) shows that quota rights for US sugar imports have undergone major reallocations every 15 years only.Footnote ¹⁰⁷

Bown and McCulloch (Reference Bown and McCulloch2003), in an empirical study dealing with safeguard protection, show that (tariff) quotas preserving historical market shares discriminate among foreign suppliers, as compared to safeguards implemented by simple tariffs (a regime that maintains the comparative advantage of suppliers). The authors empirically prove that safeguards implemented through quantitative restrictions (HA) discriminated against suppliers whose market shares had been growing over the ‘representative period’ leading up to the quota, in favor of suppliers whose share had been declining.Footnote ¹⁰⁸ They conclude that ‘quota and [tariff quota] safeguards used to protect domestic suppliers may also provide some relief to established trading partners whose positions in the relevant import market have been adversely affected by increased competition from other import sources’.Footnote ¹⁰⁹

To sum up our findings on quasi-market allocation mechanisms for tariff quotas: Auctions, LoD, and HA are all special cases of a general FCFS process. An auction is equivalent to a FCFS allocation with perfect correlation between competitive situation and rank in the queue (curve S ⁺ in Figure 3), LoD is equivalent to a random FCFS process with a zero correlation between cost and place (supply curve S ⁰), and HA is a ‘sticky’ random FCFS allocation somewhere between the S* and S ⁰ curves. All allocation methods, except tariff-quota auctions, inherently discriminate against those competitive (inframarginal) suppliers that would divide up the import market amongst themselves in the absence of a tariff quota.

4.2.1 The implication of ‘overlap and coincidence’ of NoI

In connection with its Article XIII appeal, the EC claimed that the compliance Panel had failed to explain how the EC's violation of Article XIII ‘caused any new or additional nullification or impairment to Ecuador … beyond the nullification under Article I’,Footnote ¹²⁷ which the EC had conceded to by not appealing it. According to the EC, the Panel had ‘double-counted’ the trade damage accruing to Ecuador.Footnote ¹²⁸ The AB responded that nullification or impairment (NoI) resulting from inconsistencies with Article I:1 and XIII may ‘coincide or overlap, and that any such overlap is relevant only to the calculation of the total level of nullification or impairment’. It concluded that it does ‘not need to pronounce on such questions in these Article 21.5 proceedings’.Footnote ¹²⁹

In the case of NoI experienced by the US, the EC held that the arbitrators in the original Article 22.6 proceedings in 1999 had found that the sources of NoI of benefits accruing to the US had uniquely been claims under the GATS and that no GATS obligations were at issue in the present 21.5 dispute.Footnote ¹³⁰ The AB replied that it had not held in the original proceeding that the United States had suffered NoI exclusively as a result of violations of the GATS and that the GATS was not the only source of trade damage.Footnote ¹³¹ Hence, it established an overlap of NoI under different agreements. The AB concluded:

We submit that we would have liked to see more guidance from the AB in the issue of sources, overlap and coincidence of NoI. True, the task of quantification of trade damages is conducted uniquely by the arbitrator and not by the dispute Panel or AB,Footnote ¹³³ but is it also the arbitrator's task to decide on the sources of NoI, and whether overlap and coincidence of NoI has occurred – and what the concept of ‘overlap and coincidence’ of NoI means in the first place?

First, it seems cursory by the AB simply to state that questions of the sources and level of NoI are ‘different’, that is distinct from questions of existence of NoI, and that it is only concerned with the latter, leaving the former questions in the hands of the arbitrator. This is not helpful, given that the reason why damage exists should bear some connection to how large it is. Hence, there is considerable overlap between the concept of existence and sources. We will deal with this relationship in subsection 4.2.3.

Next, on the issue of coincidence and overlap, the AB pronounced that overlap and coincidence of NoI exists, but shied away from discussing the implications of that finding. What is the arbitrator to do, if the same measure causes NoI under, say, Article XI GATT, and adverse effects in the form of serious prejudice under Article 5(c) of the SCM Agreement? Should he calculate one or two damage numbers? Should he then use the higher number? Or should he add both calculations and then determine and factor out areas of overlap? Generally, what is the overlap of serious prejudice and NoI? In addition, how can NoI caused by breach of two distinct GATT rules overlap? Are Members harmed more – or differently – under a violation of, say, Article III.2 than under Article XI GATT? The authors of this comment fail to see any systemic reasons why this would be the case. Yet, after the AB's pronouncement in Bananas III, arbitrators will have to start thinking about issues of overlap and coincidence, not least, since complainants may try to boost the numbers by claiming overlap instead of coincidence, whenever possible. In so doing, the arbitrator will have numerous question marks but no clarification from the AB.

4.2.2 Expired measures and the starting date for retaliation

With respect to the EC's repeal of the ACP tariff preference, the AB ruled that it ‘consider[s] it to be within the discretion of the panel to decide how it takes into account subsequent modifications or a repeal of the measure at issue’.Footnote ¹³⁴ The AB also referred to a remark made in its report in US–Upland Cotton, stating that ‘whether or not a measure is still in force is not dispositive of whether that measure is currently affecting the operation of any covered agreement’.Footnote ¹³⁵

These statements hardly give any guidance to the arbitrator, who is supposed to deal with the quantification of NoI caused by a measure that is no longer in existence. Who decides whether a repealed measure should be included in the calculation of NoI, the dispute Panel/AB, or the arbitrator? If a Panel or the AB decides to take into account the repeal of the measure, that is rules against the existence of NoI, yet the arbitrator finds that the repealed measure has persistent effects, may the arbitrator calculate trade damages of the withdrawn measure nevertheless?

If a repealed measure is to be included, what is the cut-off date for such inclusion? In the present case, the repeal happened long after the original reasonable period of time (RPT), but before the end of the Article 21.5 proceedings. May the complainant ask for retaliation rights amounting to the damage suffered until the measure was repealed (or its effects subsided)? A similar issue came up in the Article 22.6 proceedings in the US–Upland Cotton case,Footnote ¹³⁶ when the US repealed a prohibited cotton export subsidy (the so-called Step 2 payments) one full calendar year after the end of the RPT.Footnote ¹³⁷ Brazil as the complainant demanded retaliation rights for a one-time countermeasure, while the US argued that the program was terminated and that Brazil was hence seeking authorization for retroactive remedies not permitted by the DSU. The United States essentially argued that countermeasures may only be authorized as long as a Member has not come into compliance with the DSB's rulings and recommendations. According to the US, if there is no longer a measure, there is no more WTO inconsistency to be dealt with. Brazil opposed this view, arguing that retaliation rights begin with the end of the original RPT, and that Article 21.5 proceedings are a mere ‘suspension in time’. Thus, Brazil claimed, Article 21.5 proceedings are irrelevant for the sake of quantification of countermeasures (provided that the complainant wins the compliance proceedings).Footnote ¹³⁸ Hence, according to Brazil, a one-time retaliation award compensating for the year after the RPT but before the repeal of Step 2 payments does not constitute retroactive damages, but is faithful to Article 4.9 of the SCM Agreement, which obliges the DSB to ‘grant authorization to the complaining Member to take appropriate [and not disproportionate] countermeasures’.Footnote ¹³⁹

Pending circulation of the arbitrator's report in US–Upland Cotton, this issue is currently unresolved. It could, however, have been resolved by the AB long before at the opportunity of these Bananas proceedings, if the AB had been more explicit about how to deal with expired measures, and, more generally, whether countermeasures are measured starting from the original RPT or only after the Article 21.5 proceedings.Footnote ¹⁴⁰

4.2.3 The nature of nullification or impairment

In its report, the AB made a series of statements about how to determine the existence of nullification or impairment. At various instances, the AB held that in order to assess whether the complainants had suffered any NoI, not only actual trade effects suffered were important. According to the AB, also ‘“competitive opportunities” and, in particular, any potential trade interest’ by the complainant matter.Footnote ¹⁴¹ The AB also referred to a statement it had made in the original proceedings, opining that if the United States did not even have a potential export interest, ‘[t]he internal market of the United States for bananas could be affected by the EC bananas regime by its effects on world supplies and world prices for bananas’.Footnote ¹⁴²

The AB's three-pronged definition of NoI (actual trade damages; competitive opportunities; internal market of the complainant) should be given meaning in the subsequent arbitration process, even though its statements were made in the context of establishing the existence of NoI, and not the level of NoI, and even though compliance proceedings are procedurally different from Article 22.6 arbitrations.Footnote ¹⁴³Existence of NoI predisposes its level.Footnote ¹⁴⁴ In other words, we believe that in Bananas III the AB has defined the nature of NoI, and proclamation on the nature should logically be the basis for any calculation or quantification exercise.

If our interpretation is correct, according to which the AB's definition of the nature of NoI cannot be logically detached from its quantification, then arbitrators will face new challenges in the quantification of NoI. The integration of potential trade interests and competitive opportunities, as well as damage suffered in the internal market of the complainant, is quite a departure from the way NoI has been interpreted (and calculated) so far, namely as actual trade damages, or ‘exports foregone’ suffered after the RPT (see Bown and Ruta, Reference Bown, Ruta, Bown and Pauwelyn2009; Schropp, Reference Schropp2009b). If we are right in believing that the definition of NoI must determine, or at least have an impact on, the methodology used for calculating the level of NoI, new ways of quantifying trade damages may become necessary.Footnote ¹⁴⁵

To be clear, the authors believe that the question of existence of NoI is different from that of the level – but not independent of it. It is the responsibility of Panels and the AB to determine whether NoI exists and the separate responsibility of Article 22.6 arbitrators to then quantify it. Nevertheless, to guide arbitrators in their already difficult job, WTO Panels or the AB should be more precise about what they believe the nature of NoI to be, and in what ways this could affect or determine the damage calculation. This, we believe, will greatly support arbitrators in their work.

To summarize, although there is no doubt that the framers of the DSU entrusted the 22.6 arbitrator with the ultimate calculation of the amount of NoI, the authors believe that it is the Panel's or the AB's obligation to make clear the nature and duration of the NoI it finds, so that arbitrators need not wonder which elements they should evaluate. In the present Bananas III (21.5) report by the AB, just as in many other reports, the AB has left too many issues for the arbitrator to deal with. In particular, the AB was slightly remiss in giving too little indication and guidance as to how the arbitrator shall deal with the issues of (i) overlap and coincidence of NoI, (ii) previously repealed measures, and (iii) the relevant definition of NoI.

Some authors have rightly argued that Article 22.6 arbitrations could benefit from Appellate Body review (e.g. Lockhart, Reference Lockhart, Bown and Pauwelyn2009), and arbitrators on occasion have lamented over the lack of guidance that the DSU holds in stock for the calculation of NoI. This should encourage the AB to spend more time and thought on the consequences that their statements (or the omission thereof) will have on this last step of WTO litigations.