Allocating the Burden of Proof in WTO Disputes: A Critical Analysis

INTRODUCTION

The burden of proof is a concept related to the law of evidence, which tells us which party – plaintiff or defendant – in a case must provide proof of a determinate issue at the risk of having the adjudicator rule against it with respect to that issue.¹ This concept has two main functions. First, it distributes between the parties the task of initiating the presentation of evidence that is necessary for the adjudicator to arrive at a conclusion with respect to the issues in question.² This is the organizational aspect of the burden of proof. Second, the burden of proof works as a rule of adjudication, telling the court what to do when there is not enough evidence on the record to allow the court to arrive at a conclusion or when the court is in doubt with respect to the meaning of the evidence presented. The fact that the burden of proof may determine the outcome of a case is what gives importance to this concept. Accordingly, it is crucial that the parties know in advance who bears the burden of proof with respect to the issues of the case. Such knowledge avoids later surprises and gives the party bearing the burden of proof an opportunity to make all efforts necessary to prove its case.³

At the outset, it is also important to emphasize that the burden of proof is a concept that applies to issues of fact only. In other words, ambiguities in the law must be resolved by the adjudicator through interpretation. Furthermore, the concept of the burden of proof discussed in this article relates to proof of the factual basis of a claim as a whole. In the context of the World Trade Organization (WTO), this means a claim under a provision of the WTO agreements such as Article I.1 of the General Agreement on Tariffs and Trade (GATT). The term ‘burden of proof’ is sometimes used to refer to proof of individual allegations – or individual facts – advanced by the parties in the course of the proceedings. This is the sense in which the term is used when it is said that ‘the party that asserts a fact is responsible for providing proof thereof  ’.⁴ When used in this sense, the term is merely an expression of the principle that judicial decisions must be based on verified facts rather than on unsupported allegations.

Because parties usually submit evidence in support of their position and the fraction of cases where courts remain in doubt after all the evidence is presented is small, issues regarding the burden of proof do not appear in most cases.⁵ While this statement may provide a fair account of the issue with respect to most courts and tribunals, the same cannot be said of the WTO dispute settlement system. In the WTO, questions regarding the burden of proof were raised as early as 1996⁶ – just about a year after the WTO agreements came into force – and have since then been frequently debated in the jurisprudence. In fact, it would be no exaggeration to state that questions regarding the burden of proof have been raised in almost every case since then. The universe of cases which deal directly with the issue discussed in this article is, however, much smaller but still large compared with the number of cases that have been adjudicated in the WTO.⁷ As the title suggests, the focus of this article is on the question of the allocation of the burden of proof. In very simple terms, the article analyses how adjudicators, particularly WTO panels and the Appellate Body, determine which party – complainant or defendant – must prove a certain issue or suffer the consequences of a finding against it.

Taking into account the fact that the question of the allocation of the burden of proof has not been addressed in the Understanding on Rules and Procedures Governing the Settlement of Disputes (DSU), one of the main objectives of this article is to conduct a critical analysis of the rules that panels and the Appellate Body have developed to allocate the burden of proof between the parties. The article first finds that panels and the Appellate Body have not adopted a coherent approach and that this has resulted in arbitrary allocations of the burden of proof. It then searches for solutions outside the realm of WTO jurisprudence. This article demonstrates that domestic law and general principles of law can provide only limited guidance. An alternative analytical framework is then developed on the basis of law and economics literature on the burden of proof. This framework would allow panels and the Appellate Body to choose between two alternative courses of action, implying some trade-off between predictability and flexibility.

I. CRITIQUE OF THE WTO JURISPRUDENCE

Similar to most international treaties, the WTO agreements, including the DSU, do not contain general rules regarding the allocation of the burden of proof.⁸ Panels and the Appellate Body, however, have had to deal with this question since the first disputes brought under the WTO dispute settlement system. In one of its first decisions, the Appellate Body established that ‘the burden of proof rests upon the party, whether complaining or defending, who asserts the affirmative of a particular claim or defence’.⁹ In order for this rule to operate, it was then necessary to define what constitutes a defense that must be proved by the defendant. Pre-WTO panels (GATT 1947 panels) had established that exceptions such as those contained in Articles XX and XI.2(c)(i) of the GATT are defenses that must be proved by the defendant.¹⁰ This view was embraced by panels and Appellate Body as early as 1996 in US – Gasoline.¹¹ Accordingly, it would seem that the question of the allocation of the burden of proof would have been settled: the complainant would have to prove the violations of the agreements that he alleged, and the defendant would have the burden of proving any exceptions contained in those agreements. However, this is not the case.

Panels and the Appellate Body have struggled with the definition of an exception. Provisions that are very similar – all of them containing specific rules exempting Members from compliance with more general rules – have been placed into two different categories: (1) provisions that establish an exception to a rule¹² and (2) provisions that exclude the application of other provisions.¹³ In the first case, the complainant has the burden of proving that the defendant has violated a general rule. This having been determined, the defendant has the burden of proving that it has complied with the requirements of the provision establishing an exception to that rule. In the second case, the complainant has the burden of proving that the defendant does not fall under the situation or has not complied with the requirements of a provision that excludes the application of the general rule. If the complainant successfully discharges its burden of proof with respect to the excluding provision, it may then seek to establish a violation of the general rule. In other words, in the presence of a provision which may potentially exclude a Member from the application of a general rule, the party seeking a finding of violation of the latter – the general rule – must first establish that its opponent does not fall under the situation foreseen by the excluding provision. Therefore, while the burden of proof with respect to an exception is on the defendant, in the case of a provision that excludes the application of another, the burden is on the complainant. As it will be discussed below, it is not easy to make this distinction. In fact, the confusing ways in which panels and the Appellate Body have described this distinction suggests that the distinction is artificial.¹⁴

The analysis in this section demonstrates that panels and the Appellate Body have failed to develop a consistent and effective approach to the allocation of the burden of proof. The analysis is divided into two parts. The first part identifies criteria that have been considered by panels and the Appellate Body when establishing whether a provision is a defense that must be proved by the defendant. These criteria are then deployed to analyse the findings of panels and the Appellate Body. The goal of this analysis is to demonstrate that those criteria have not been consistently applied, which, therefore, diminishes their value in predicting the outcome of decisions. In the second part, the analysis will focus on determining whether the criteria themselves are meaningful, i.e., whether the criteria are effective in distinguishing provisions that must be proved by the defendant from those that must be proved by the complainant.

A. Have panels and the Appellate Body been consistent?

1. Distinguishing between exception provisions and excluding provisions

(a) Identification of the criteria

A review of the WTO jurisprudence suggests that the following criteria have played a central role in determining whether a certain provision is an exception or a provision that excludes the application of another.

The language criterion

The ‘ordinary meaning’ of the words has been the starting point of panels and the Appellate Body in the interpretation of the WTO agreements. The focus on ‘text’ has been so prominent that on occasion it has prompted criticism against panels and the Appellate Body.¹⁵ Because of this prominence of the ordinary meaning of the text of the agreements in the WTO jurisprudence, it is assumed that this should be an important criterion.

TheUS – Wool Shirts and Blousescriteria

The ruling of the Appellate Body in US – Wool Shirts and Blouses has been cited in most subsequent cases that deal with the question of the burden of proof as the authority in the matter. First, in that case, the Appellate Body said that in order for a provision to be considered an exception, it must authorize derogation, i.e., provide an exception, from obligations, and it must not be a positive rule establishing legal obligations in itself.¹⁶ This criterion will be referred to as the positive rule criterion. Second, when developing its reasoning in support of the view that Article 6 of the Agreement on Textiles and Clothing (ATC) was not an exception, the Appellate Body said that Article 6 reflects a ‘carefully drawn balance of rights and obligations of Members’ and that ‘Article 6 of the ATC is a fundamental part of the rights and obligations of WTO Members’.¹⁷ This suggests that the importance that the drafters placed on a certain provision is a factor that must be taken into account when determining whether a certain provision is an exception; it implies the existence of a hierarchy between provisions where exceptions are placed at a lower level of the hierarchical pyramid. This criterion will be referred to as the hierarchical criterion.

(b) Analysis of the criteria

The language criterion

A quick look at the text of some of the provisions which were the subject of a ruling on the question of the allocation of the burden of proof reveals that most of those provisions contain some language which suggests that the drafters created a special rule deviating from other more general rules. Article XX of the GATT, and Article XIV of the General Agreement on Trade in Services (GATS) state ‘. . . nothing in this Agreement shall be construed to prevent . . .’;¹⁸ Article XI.2(c)(i) of the GATT contains the phrase ‘. . . shall not extend to the following . . .’;¹⁹ Article XXIV.5 of the GATT says that ‘. . . the provisions of this Agreement shall not prevent . . .’;²⁰ Article XV.9 of the GATT states that ‘Nothing in this Agreement shall preclude . . .’;²¹ the Enabling Clause states that ‘Notwithstanding the provisions of Article I . . .’;²² Article 27.2 of the Agreement on Subsidies and Countervailing Measures (SCM Agreement) says ‘. . . shall not apply . . .’;²³ footnote 59 of the SCM Agreement states that ‘Paragraph (e) is not intended to limit . . .’;²⁴ Article 13 of the Agreement on Agriculture says ‘. . . notwithstanding the provisions of . . .’, and Articles 5 and 6 of the SCM Agreement say ‘. . . does not apply . . .’, while Article 3 of the SCM Agreement states that ‘Except as provided in . . .’;²⁵ the words except or exception are then used in the title of Article XX of the GATT, Article XIV of the GATS, Article 2 of the ATC,²⁶ Article 30 of the Agreement on Trade Related Aspects of Intellectual Property (TRIPS Agreement),²⁷ Article 2.4 of the Agreement on Technical Barriers to Trade (TBT Agreement),²⁸ and Article 3.1 of the Agreement on the Application of Sanitary and Phytosanitary Measures (SPS Agreement).²⁹

According to the Oxford English Dictionary, the word ‘exception’ means ‘1. The action of excepting (a person or thing, a particular case) from the scope of a proposition, rule, etc. [ . . . ] 2. Something that is excepted; a particular case which comes within the terms of a rule, but to which the rule is not applicable [ . . . ]’.³⁰ It would be fair to say that all the expressions referred to above either fit this definition of the word ‘exception’ or convey the same or a very similar idea. Yet, despite the similar meaning of these expressions, not all those provisions have been found to be an exception. Of the provisions above, only Articles XI.2(c)(i), XV.9, XX, and XXIV.5³¹ of the GATT, the Enabling Clause, Article XIV of the GATS, footnote 59 of the SCM Agreement, and Article 30 of the TRIPS Agreement are exceptions that must be proved by the defendant.

Based on the wording of Article 2.4 of the ATC, the panel in US – Underwear concluded that Article 6 of the ATC is an exception to the general prohibition on the imposition of new restrictions. Therefore, claims under that Article must be proved by the defendant.³² This interpretation of Articles 2.4 and 6 of the ATC was short lived. In US – Wool Shirts and Blouses, the panel and Appellate Body did not conduct an analysis of the wording of Article 2.4 and concluded that Article 6 of the ATC is not an exception that must be proved by the defendant.

Interestingly, in Brazil – Aircraft, the panel and Appellate Body concluded on the basis of the plain meaning of ‘shall not apply’ in Article 27.2 of the SCM Agreement that developing countries falling under that provision do not fall within the scope of application of Article 3.1(a) of the SCM Agreement. Therefore, according to the panel and Appellate Body, Article 27.2 is not an exception, but rather a provision that excludes the application of another.³³ The panel and Appellate Body did not seem to consider that this fits perfectly into the definition of an exception, i.e., that an exception is ‘a particular case which comes within the terms of a rule, but to which the rule is not applicable’. Similarly, in US – Upland Cotton, the panel sought to emphasize the expression ‘does not apply’ in Articles 5 and 6 of the SCM Agreement while downplaying the word ‘except’ in Article 3 of the SCM Agreement.³⁴ When analysing the words ‘notwithstanding’ and ‘exempt’ in Article 13 of the Agreement on Agriculture, the panel seems to have realized fleetingly that the difference between an exception and something to which the rule is not applicable is artificial.³⁵ However, this did not influence the panel’s final conclusion that Article 13 is not an exception that must be proved by the defendant.

In EC – Hormones, the panel read Article 3.1 of the SPS Agreement as establishing a general obligation to base SPS measures on international standards; this obligation, however, is not absolute since Article 3.1 provides that this rule applies ‘except as otherwise provided for in [the] Agreement, and in particular Article 3.3’. The panel viewed the word ‘except’ in Article 3.1, together with the reference to Article 3.3, as an indication that Article 3.3 is an exception that must be proved by the defendant.³⁶ The Appellate Body, however, came to a different conclusion. The Appellate Body stated that the general rule that the complainant must provide a prima facie case ‘is not avoided by simply describing [a] provision as an “exception” [ . . . ]’.³⁷

Perhaps because of EC – Hormones, the panel in EC – Sardines did not rely upon the word ‘exception’ in Article 2.4 of the TBT Agreement to come to the conclusion that the defendant must bear the burden of proof with respect to the second part of Article 2.4. The Appellate Body overturned the panel’s decision on the grounds that ‘as with Articles 3.1 and 3.3 of the SPS Agreement, there is no “general rule-exception” relationship between the first and the second parts of Article 2.4’ (original emphasis).³⁸

In EC – Tariff Preferences, the panel and Appellate Body turned back to an analysis of the text. They focused on the word ‘notwithstanding’ and concluded that the ordinary meaning of this word suggests that the Enabling Clause exempts Members from complying with Article I.³⁹ Therefore, they concluded that the Enabling Clause is an exception for which the burden of proof is on the defendant.⁴⁰ The wording of Article 30 of the TRIPS Agreement also seems to have been the basis for the panel’s conclusion, in Canada – Pharmaceutical Patents, that ‘Article 30 is an exception to the obligations of the TRIPS Agreement’ and that, therefore, the burden was on Canada, the respondent, to prove that Sections 55.2(1) and 55.2(2) of Canada’s Patent Act complied with the criteria laid down in that Article.⁴¹

After this short analysis of the jurisprudence and the wording of some provisions of the agreements, it is possible to conclude that the language criterion has not been coherently applied, and, therefore, it is not a useful tool to predict the outcome of disputes regarding the allocation of the burden of proof.

TheUS – Wool Shirts and Blousescriteria. The analysis in the next few paragraphs will be divided into two parts according to the two US – Wool Shirts and Blouses criteria mentioned above. The first part focuses on the positive rule criterion, while the second part addresses the hierarchical criterion

It is not entirely clear what the Appellate Body meant when it described Articles XX and XI.2(c)(i) of the GATT as exception provisions because they provide exceptions from obligations under other provisions and they are not ‘positive rules’ establishing legal obligations in themselves.⁴² But since this statement was made against the background of Article 6 of the ATC, it may be appropriate to compare the structures of Articles XX and XI.2(c)(i) of the GATT with Article 6 of the ATC to come to a clearer definition of this criterion.⁴³ A quick comparison between those Articles suggests that provisions that provide exceptions and are not positive rules establishing obligations in themselves (1) would contain a direct reference to the obligations that they are providing an exception from and (2) would set down the conditions under which the exception could be invoked. In turn, provisions that contain positive rules establishing obligations in themselves – i.e., provisions that are not exceptions – would, similar to Article 6 of the ATC, be self-standing provisions regulating, authorizing, or prohibiting something.

In its first decision regarding the allocation of the burden of proof after US – Wool Shirts and Blouses, the Appellate Body did turn to the criterion established in that decision. In EC – Hormones, the Appellate Body overturned the panel’s characterization of Article 3.3 of the SPS Agreement as an exception. In the Appellate Body’s words, Article 3.3 of the SPS Agreement recognizes the ‘autonomous right’ of a Member to establish a higher level of protection than that implicit in an international standard provided that that Member complies with certain requirements in promulgating SPS measures.⁴⁴ The structure of Article 3.3, if read in isolation from Article 3.1, seems to support such a conclusion – there is no reference in Article 3.3 to Article 3.1. In addition, Article 3.3 contains all the elements necessary for it to stand on its own. The same, however, cannot be said of the second part of Article 2.4 of the TBT Agreement. The Appellate Body, however, did not seem to take this into account when in EC – Sardines it overturned the panel’s allocation to the defendant of the burden of proof under the second sentence of Article 2.4. In that case, the Appellate Body focused instead on the conceptual similarities between Articles 2.4 of the TBT Agreement and Article 3 of the SPS Agreement.

It was in Brazil – Aircraft that the positive rule criterion was for the first time given a different interpretation. Articles 27.2 and 27.4 of the SCM Agreement refer to Article 3.1 of the SCM Agreement and clearly depend on that provision to have any meaning. The panel, however, found that ‘although Article 27.4 would not necessarily be the legal basis for a separate claim of violation of the SCM Agreement, Article 27.4 does impose certain obligations’.⁴⁵ The panel did not seem to consider, however, that these obligations are not autonomous. Nor did the panel consider that even Article XX of the GATT imposes certain obligations in that, e.g., under certain paragraphs such as paragraph (b) measures can only be adopted if they are ‘necessary’. The Appellate Body upheld the panel’s findings on very similar grounds. According to the Appellate Body, a careful reading of paragraphs 2 and 4 of Article 27 clearly reveals that ‘the conditions set forth in paragraph 4 are positive obligations for developing country Members, not affirmative defences’.⁴⁶

The last sentence of footnote 59 of the SCM Agreement contains a clear reference to item (e) of the Illustrative List of Export Subsidies, and it is not a provision that could stand on its own. Therefore, it seemed straightforward for the Appellate Body to conclude that footnote 59 contained an exception to item (e) and was not a positive rule establishing obligations in itself. Instead of mirroring its analysis on US – Wool Shirts and Blouses, however, the Appellate Body, in US – FSC (Article 21.5), rephrased the positive rule criterion. According to the rephrased version, an exception does not establish any ‘positive obligations’ relevant to determining the proper scope of the provisions they are providing an exception from.⁴⁷ The Appellate Body invoked Article XX of the GATT as an example. Article XX ‘sets out circumstances in which Members are entitled to “adopt or maintain” measures that are inconsistent with the obligations imposed under other provisions of the GATT, such as Article I and III’.⁴⁸ The Appellate Body did not seem to consider the following point. Article XX defines the scope of the obligations contained in Articles I and III of the GATT when it entitles Members to adopt or maintain measures that would otherwise be inconsistent with those Articles.⁴⁹ Furthermore, this rephrasing of the criterion has arguably given it a totally new meaning, since according to US – FSC (Article 21.5) provisions that are not self-standing will qualify as positive rules establishing obligations. The panel in US – Upland Cotton explicitly recognized that Article 13 of the Agreement on Agriculture does not impose positive obligations.⁵⁰ This finding, however, did not influence the panel’s final conclusion. The panel instead referred to the rephrased criterion from US – FSC (Article 21.5) and found that Article 13 defines the scope of positive obligations contained in Articles 3, 5, and 6 of the SCM Agreement and, therefore, it is not an exception which must be proved by the defendant.⁵¹

In EC – Tariff Preferences, the panel went back to the original US – Wool Shirts and Blouses formulation of the positive rule. The Enabling Clause clearly refers to Article I of the GATT and, therefore, is arguably not a self-standing obligation. In the panel’s view, the fact that the Enabling Clause contains requirements does not alter the ‘basic “non-obligatory” nature of these provisions’, because the conditions contained in these provisions ‘are only subsidiary obligations, dependent on the decision of the Member to take such measures’.⁵² The Appellate Body upheld the panel’s finding that the Enabling Clause is an exception which must be proved by the defendant, though it did not make an explicit determination about whether the Enabling Clause is a positive rule containing obligations in itself.

This brief analysis of the jurisprudence suggests that the US – Wool Shirts and Blouses ‘positive rule’ criterion has not been consistently applied, and for this reason it is not a useful tool to predict the outcome of disputes regarding the allocation of the burden of proof.

The analysis now turns to the second US – Wool Shirts and Blouses criterion described above, i.e., the hierarchical criterion. Similar to US – Wool Shirts and Blouses, in EC – Hormones, the Appellate Body stated that the right contained in Article 3.3 of the SPS Agreement, which gives Members the freedom to determine their own appropriate level of sanitary protection is ‘an important right’.⁵³ According to the Appellate Body, this is made clear in the sixth preambular paragraph of the SPS Agreement. In Brazil – Aircraft, the Appellate Body made a similar statement with respect to Article 27 of the SCM Agreement to the statement it had made regarding Article 6 of the ATC in US – Wool Shirts and Blouses. The Appellate Body noted that ‘paragraphs 2 and 4 of Article 27 contain a carefully negotiated balance of rights and obligations for developing country Members’.⁵⁴ The Appellate Body also referred in the following paragraph to the purpose of Article 27, i.e., to provide special and differential treatment to developing countries in recognition of the fact that subsidies may play an important role in the economic development programmes of those countries.⁵⁵

In India – Autos, the panel disregarded India’s argument that ‘[j]ust as Article 6 of the ATC [is] a fundamental part of the rights and obligations of WTO Members under the ATC, Article XVIII.B of the GATT [is] a fundamental part of the rights and obligations of developing countries under the GATT’.⁵⁶ But it was in EC – Tariff Preferences that the panel and Appellate Body categorically rejected the idea that there is any hierarchy between provisions and that this hierarchy would affect the allocation of the burden of proof.⁵⁷ The Appellate Body noted that characterizing the Enabling Clause as an exception would not undermine its importance and diminish ‘the right of Members to provide or to receive “differential and more favourable treatment” ’.⁵⁸

The analysis above, therefore, reveals that the importance the drafters placed on a certain provision has not been consistently taken into account by panels and the Appellate Body when determining whether a certain provision is an exception.

2. Decisions based on alternative grounds

There have also been cases where panels and the Appellate Body did not refer to the distinction between an exception provision and an excluding provision – though the provisions involved would clearly fall into one of these categories – or to any of the criteria mentioned above. For example, in India – Quantitative Restrictions, the panel placed the burden of proof on the party that raised each provision.⁵⁹ The panel also specifically noted that India bore the burden of proof with respect to the proviso contained in Article XVIII.11 because India’s reference to this proviso was ‘in the nature of an affirmative defence’.⁶⁰ On appeal, the Appellate Body agreed with the panel that the respondent bore the burden with respect to the proviso. Neither the panel nor the Appellate Body provided further reasons for their findings. The issue of the burden of proof with respect to Article XVIII.B was raised once again in India – Autos, a dispute which arose out of further developments related to the problem presented in India – Quantitative Restrictions. Once again, the panel allocated the burden of proof with respect to Article XVIII.B to the party that raised that provision. According to the panel, because the complainants had made no representations concerning Article XVIII.B, the burden of proof was on India, the party invoking that provision.⁶¹ It is worth noting that the criterion that the party who raises an issue must bear the burden of proving it is not followed consistently. For example, in Brazil – Aircraft, Canada had not raised any claims under Article 27 of the SCM Agreement; Article 27 of the SCM Agreement was raised by Brazil. The panel and the Appellate Body, nonetheless, imposed on Canada the burden of proving that Brazil did not meet the conditions contained in Article 27.2 and 27.4.⁶²

The panel in Dominican Republic – Cigarettes also did not discuss any of the criteria presented in the previous section. The panel noted that both parties agreed that Article XV of the GATT is an exception and that, therefore, the burden of proof with respect to that Article is on the defendant. The panel agreed with this characterization without providing further explanation.⁶³ In Brazil – Aircraft (Article 21.5), the Appellate Body also invoked the fact that a party had recognized – even though this had happened in a previous proceeding – that it bore the burden of proof with respect to a certain provision as one of the reasons to allocate the burden of proof to that party.⁶⁴ In that case, the Appellate Body again did not discuss any of the criteria presented above. It found that because Brazil was clearly asserting item (k) as a defense, it should bear the burden of proving it.⁶⁵ It is important to note that because the allocation of the burden of proof is a question of law, panels and the Appellate Body should not leave the answer to this question to be determined by the parties. However, if all the parties in a case expressly agree to a certain allocation of the burden of proof, then this agreement may be valid for the purposes of that particular dispute.⁶⁶

3. Final considerations regarding all the criteria

The discussion above demonstrates that the jurisprudence regarding the allocation of the burden of proof has not followed a clear pattern. Panels and the Appellate Body have set some criteria but have not consistently followed them. This lack of consistency makes it hard to predict the outcome of any future disputes regarding the allocation of the burden of proof.

B. Are the criteria relied upon effective?

In addition to the criticism that none of the criteria that have been discussed up until this point are consistently employed by panels and the Appellate Body, the relevance of those criteria must also be questioned. For example, the criterion that the party who raises a certain provision must bear the burden of proof with respect to the elements of that provision is easily abused and therefore unworkable. The parties to a case could easily draft their claims and submissions in ways that would avoid touching upon issues for which they would not wish to bear the burden of proof. This would force the opposing party to bring up the point and, therefore, bear the burden of proof with respect to it. For example, in a case under the now expired ATC, the complainants could choose to bring a claim only under Article 2.4 of the ATC even though they were aware of the fact that the restrictions imposed on imports originating in its territory would fall under Article 6 of the ATC.⁶⁷ In a more drastic example, the complainant, knowing that anti-dumping duties had been imposed on imports originating in its territory, could choose to challenge those duties under Articles I and II of the GATT or, alternatively, under Article VI of the GATT, therefore, leaving the defendant to raise and prove any claims under the Anti-Dumping Agreement.

In EC – Hormones and US – Wool Shirts and Blouses, the panels and the Appellate Body addressed the question of whether the prescriptive language of certain provisions has any impact on the allocation of the burden of proof.⁶⁸ In other words, the question before the panels and the Appellate Body was whether the fact that certain provisions use language such as ‘Members shall ensure that . . .’⁶⁹ suggests that the burden of proof in dispute settlement should be allocated to the party to whom the provision is addressed. The Appellate Body in EC – Hormones wisely concluded that this should not be so.⁷⁰ The prescriptive language of certain provisions is merely intended to impose an obligation on a Member to do certain things – as opposed to merely abstaining from doing certain things – when adopting certain types of measures, such as phytosanitary measures, which have a recognized impact on trade. This may impose a certain burden on Members adopting those types of measures because they have to ensure that certain procedures are followed in their domestic regulatory process. However, this burden should not be mistaken with the burden of proof in the litigation process.

Furthermore, the distinction that panels and the Appellate Body have created in the current jurisprudence between provisions that constitute exceptions and provisions that exclude the application of others is artificial. This distinction is based on techniques of interpretation to avoid or solve conflicts between legal rules.⁷¹ Panels and the Appellate Body have resorted to conflict of norms analysis to conclude that while exception provisions and the rules they are providing an exception from are in conflict, excluding provisions and the rules that they exclude are not in conflict. Panels and the Appellate Body seem to believe that there is conflict between the provisions of the first group but that this conflict can be solved through the application of the principle of lex specialis.⁷² Pursuant to this principle, the exception provision would prevail to the extent of the conflict. As for the second group of provisions, panels and the Appellate Body maintain that these provisions do not conflict because they have different scopes of application, i.e., they do not overlap.⁷³ The main criticism to the approach followed by panels and the Appellate Body is that there is no conflict or overlap between the provisions of either of the groups. Therefore, there is no need to resort to this type of analysis. The drafters of the Agreements explicitly dealt with any possible conflict that could arise between the provisions at issue here when they included words and phrases such as ‘except’, ‘notwithstanding’, ‘nothing in this Agreement shall be construed to prevent’, and ‘shall not apply’. The language of all the provisions at issue here – exceptions as well as excluding provisions – explicitly allows for deviations from other rules. Therefore, it is clear that, e.g., due to the action of the drafters there is no conflict or overlap between Articles I and XX of the GATT just as there is no conflict or overlap between Articles 3.1 and 3.3 of the SPS Agreement.⁷⁴ Conflict of norms analysis is appropriate when due to legislative oversight courts do not know which provision applies to a certain situation. However, this is not the case when due to legislative action it is clear which provision applies.⁷⁵ The allocation of the burden of proof and determination of which party should raise a certain claim should be based on considerations specific to the question of proof as well as the efficient functioning of the legal process.⁷⁶

The criteria that have been used to differentiate an exception provision from an excluding provision lack real substance. The fact that panels and the Appellate Body have rephrased and changed the meaning of the positive rule criterion is an indication of this. The application of this criterion has created artificial differences between provisions. For example, the panel in EC – Tariff Preferences concluded that the Enabling Clause is not a positive rule because, similar to Articles XX, XXI, and XXIV of the GATT, it does not impose an obligation on Members to do something, e.g., provide preferences; the obligations contained in these provisions are only subsidiary because they depend on the decision of a Member to take such measures.⁷⁷ If this understanding is applied, many of the provisions that have been classified as provisions that exclude the application of other provisions would not meet the positive rule criterion. For instance, Article 6 of the ATC does not oblige Members to adopt safeguard measures, nor does Article 3.3 of the SPS Agreement mandate Members to adopt measures that pursue higher levels of protection than international standards. Similar to the Enabling Clause and Articles XX, XXI, XXIV of the GATT, Article 6 of the ATC and Article 3.3 of the SPS Agreement set out conditions that must be complied with if a Member decides to avail itself of the right to impose a safeguard measure or adopt an SPS measure that guarantees a higher level of protection than an existing international standard.

The distinction created between Article 3 of the SPS Agreement and Article XX of the GATT is particularly illustrative. The SPS Agreement is the product of an effort to discipline an area where the domestic regulatory process could potentially undermine the benefits of tariff bargains.⁷⁸ It is clear that WTO Members recognized that SPS regulation is often necessary, but this does not mean that Members thought that they could act unconstrainedly when enacting SPS regulation. In this context, the SPS Agreement expressed a preference that Members use international standards when possible.⁷⁹ This preference is clearly expressed in Article 3.1 of the SPS Agreement, which states that ‘Members shall base their sanitary or phytosanitary measures on international standards, guidelines or recommendations, where they exist . . .’ (emphasis added). The fact that Members expressed this preference, however, does not mean that they thought that they could not adopt measures that are not based on international standards. An authorization to adopt such measures was expressly included in Articles 3.1 and 3.3 of the SPS Agreement. This authorization, however, was not a carte blanche for Members to do whatever they saw fit. Article 3.3 contains the conditions that must be observed.

It can similarly be said that the drafters of the GATT have also recognized that some kinds of domestic regulation, such as health regulation, are necessary. They also recognized that these regulations may affect international trade by, for instance, granting imports less favourable treatment than domestic products or by restricting access to the market of the regulating country. They expressed a preference that Members do not do so in, e.g., Articles III and XI of the GATT. The fact that Members expressed this preference, however, does not mean that they thought that they could not adopt measures that, e.g., provide less favourable treatment to imports. An express authorization to do so is contained in Article XX that states that ‘nothing in this Agreement shall be construed to prevent the adoption or enforcement . . .’ (emphasis added). This authorization, however, is subject to certain conditions that are specified in the chapeau and paragraphs of Article XX. Accordingly, the distinction created between Articles 3.1 and 3.3 of the SPS Agreement, and Article XX of the GATT is not based on the substance or the purpose of those Articles.⁸⁰ Moreover, if the references to Article XX(b) contained in the SPS Agreement are taken into account, then the fragility of the distinction becomes more evident. The Preamble to the SPS Agreement notes that the SPS Agreement elaborates on the provisions of Article XX(b) of the GATT. In addition, Article 2.4 of the SPS Agreement provides that measures complying with the SPS Agreement shall be presumed to be in accordance with the obligations of the Members under the provisions of Article XX(b). Article 2.4 suggests that it might not have been the intention of Members to allocate the burden of proof differently in Article XX(b) of the GATT and in the SPS Agreement.⁸¹ As it will be discussed in the next section of this article, this is not to say that the burden of proof with respect to the SPS Agreement should be allocated to the defendant.

The application of the hierarchical criterion has created two classes of provisions that enjoy different status. The reasoning goes as follows. If Members intended that a provision be a fundamental part of the scheme of rights and obligations, then it should be concluded that they could not have intended that provision to work as an exception. That provision is then characterized as a provision that excludes the application of another, and, therefore, the burden of proof is allocated to the complainant. For obvious reasons, the application of this criterion makes the process of allocation of the burden of proof hostage to value judgments and vulnerable to criticism. The Appellate Body seemed to have realized this in EC – Tariff Preferences when it found that ‘characterizing the Enabling Clause as an exception [ . . . ] does not undermine the importance of the Enabling Clause within the overall framework of the covered agreements [ . . . ] Nor does it “discourage” developed countries from adopting measures in favour of developing countries under the Enabling Clause’.⁸² This recognition might have come too late in the process though. The prevalent perception is that certain values have been favoured over others, and that in certain situations important values have not been adequately protected. Abbott expresses this concern with respect to Articles 13 and 30 of the TRIPS Agreement, which have been classified as exceptions that must be proved by the defendant.⁸³ He notes that in EC – Hormones the Appellate Body went to some lengths to explain that the European Communities had the right to protect its public and that the burden was on the United States to demonstrate that the European Communities had failed to apply reasonable scientific standards. He then raises the question: ‘If exceptions to patent rights allow governments scope to protect the public interest, should the burden be on the Member establishing the exception to demonstrate that it is doing so?’⁸⁴ More than demonstrating that the burden of proof was allocated wrongly either in EC – Hormones or in Canada – Pharmaceutical Patents, this kind of criticism reiterates the problematic nature of the hierarchical criterion and the distinction it purports to establish between exceptions and provisions that exclude the application of other provisions. Panels and the Appellate Body are not in a position to determine the importance that Members placed on a certain provision. Nor should they infer from the importance of a provision and in the absence of an indication to the contrary that Members intended to allocate the burden of proof differently than would otherwise be the case.

Finally, relying purely on the language of a provision does not constitute a sounder and more predictable method for allocating the burden of proof. As noted before, the definition of ‘exception’ encompasses the idea expressed in the phrase ‘shall not apply’. Moreover, it is not clear why a provision that creates an exception should alter the normal allocation of the burden of proof. It is true that starting in the 1980’s GATT 1947 panels issued a stream of decisions embracing the view that the party claiming an exception had the burden of showing that it had satisfied the requirements of that exception.⁸⁵ However, WTO panels and the Appellate Body have not blindly followed GATT 1947 panels in many areas, including this issue. For example, some GATT 1947 panels considered that trade remedies constituted exceptions to the general GATT rules. Therefore, the onus was upon parties imposing anti-dumping duties, countervailing duties, and safeguards to demonstrate before a dispute settlement panel that they had not acted inconsistently with the terms of the provisions granting such exceptions.⁸⁶ WTO panels and the Appellate Body have not supported the view that trade remedies rules are exceptions that must be proved by the party making use of those rules. Accordingly, there are no reasons to blindly follow GATT 1947 panels in their interpretation that exceptions must be proved by the defendant. In fact, it is worrisome that this rule has not been questioned more vigorously, because there is no indication in the texts of the Agreements that the burden of proof with respect to provisions that provide exceptions should be allocated to the defendant. Why should we assume that this was the intention of the drafters, particularly in the light of provisions such as Article 10.3 of the Agreement on Agriculture, which clearly address the question of the burden of proof?

The Appellate Body may have been right when in EC – Hormones it found that ‘[t]he general rule in a dispute settlement proceeding requiring the complaining party to establish a prima facie case of inconsistency [ . . . ] is not avoided by simply describing that same provision as an “exception”[ . . . ]’.⁸⁷ The problem with EC – Hormones is that even though the Appellate Body seems to have stated that the fact that a provision is an exception says nothing about the allocation of the burden of proof, the Appellate Body still went to some lengths to explain that Article 3.3 of the SPS Agreement is not an exception but rather a self-standing provision that excludes the application of Article 3.1 of the SPS Agreement. In other words, the Appellate Body could instead have concluded that even though Article 3.3 is an exception – which seems undeniable from the language of Article 3.1 of the SPS Agreement – in the absence of any indication from the drafters to the contrary, there is no reason to place the burden of proof on the defendant with respect to that Article.

One of the main problems with allocating the burden of proof to the defendant with respect to exceptions is that this allocation is based on the assumption that there is something special about an exception that would imply that the drafters made such choice of allocation. But what is so special about an exception? Does not an exception merely define the scope of a rule? As Williams explains in criminal law parlance, ‘[r]ationally regarded, an ‘exception’ merely states the limits of an offence. A person whose act falls within the exception does not commit the offence’.⁸⁸ Therefore, ‘the exception is, rationally regarded, part and parcel of the rule’.⁸⁹ In addition, it is often possible to draft provisions in a way that the exception is contained in the rule;⁹⁰ thus, the use of words indicating an exception may only have the function of making a provision easier to read and understand. For example, instead of using the word ‘except’ in Article 3.1 of the SPS Agreement to establish a relationship with Article 3.3, the drafters could have chosen the following wording:

When a Member chooses a level of sanitary and phytosanitary protection that is similar to the level implicit in an international standard it shall base its measures on that standard. If a higher level of protection is desired, a Member may adopt a measure that is not based on an international standard. In the latter case, the Member shall ensure that it observes the relevant provisions of paragraphs 1 through 8 of Article 5

The fact that an exception is stated in a separate Article should not alter this conclusion. The drafters may have preferred to create a separate provision merely for reasons of convenience, i.e., in order to avoid having to incorporate the exception into several or all Articles of an agreement.

In the light of the above, the following question seems inevitable. Is it appropriate to assume that the style of drafting chosen for a certain provision represents a considered judgement as to the allocation of the burden of proof? One might argue that the drafters were aware of the rule that the burden of proof with respect to exceptions is allocated to the defendant when they chose to express a certain rule in the form of an exception; thus, they implicitly made a choice concerning the allocation of the burden of proof. The truth of this proposition is questionable, particularly because it would be fair to state that there is no general principle of law recognized by civilized nations that an exception must be proved by the defendant.⁹¹

The jurisprudence of the International Court of Justice (ICJ) does not refer to such a principle.⁹² As for domestic law, some commentators have noted that in their quest to determine what is essential to a party’s case, courts have sometimes applied a controversial formula, i.e., they have sought to distinguish between the constituent elements of a promise or of a statutory command and matters of exception; the plaintiff would have the burden of proving the constituent elements and the defendant the burden of proving exceptions, exemptions, excuses, qualifications, and provisos.⁹³ The presence of linguistic signposts such as ‘except for’, ‘provided that’, and ‘subject to’ would help in determining whether a certain provision constitutes a matter of exception.⁹⁴ It is important to note that most examples of cases where this formula is applied refer to litigation concerning contracts⁹⁵ or cases where the application of this formula resulted from a specific statutory command.⁹⁶ Furthermore, in many cases, courts employ this formula only to allocate the burden of pleading – in which case the burden of proof remains with the plaintiff.⁹⁷ The formula has been criticized for leading to arbitrary allocations of the burden of proof, as the ‘statutory language may be due to a mere causal choice of form by the draftsman’.⁹⁸

Sullivan also notes that ‘courts sometimes suggest that statutory exceptions are to be strictly construed, or alternatively, that persons seeking to benefit of a statutory exemption or exception must establish clearly that they come within its terms’ (emphasis added).⁹⁹ She explains that the reason behind this is that the courts are ‘concerned that exceptions and exemptions be interpreted in light of their underlying rationale and not used to undermine the broad purpose of the legislation’.¹⁰⁰ In the WTO context, panels and the Appellate Body would in most cases have little legitimacy to make this kind of judgment about the broad purpose of the Agreements. Unlike domestic courts, panels and the Appellate Body cannot rely upon a set of fundamental principles such as those found in national constitutions to guide them in this process. Panels and the Appellate Body not only cannot rely upon fundamental principles but also cannot, in many cases, refer to any legislative expression of policy with respect to certain Agreements or provisions in the Agreements. For example, the Anti-Dumping Agreement and the SCM Agreement do not have a preamble that could provide some guidance as to the policy objectives of those agreements. Scarcity of legislative expressions of policy in the text of the WTO agreements likely reflect the fact that while WTO Members agreed on the text of the Agreements, they often had different views on the policy behind those Agreements. In any event, the Appellate Body has expressly rejected the view that exceptions must be strictly or narrowly construed. In EC – Hormones, the Appellate Body said that ‘merely characterizing a treaty provision as an ‘exception’ does not by itself justify a “stricter” or “narrower” interpretation of that provision than would be warranted by examination of the ordinary meaning of the actual treaty words [ . . . ]’.¹⁰¹ It may be time to do the same with the rule that exceptions must be proved by the defendant.

II. WHAT ARE THE ALTERNATIVES?

Where does this discussion leave us? Panels and the Appellate Body have not consistently followed the same rules to allocate the burden of proof; the criteria they have set out to distinguish between exceptions and provisions that exclude the application of other provisions, as well as the distinction itself, seem artificial. The criterion that the party who raises a certain provision must bear the burden of proof with respect to the elements of that provision is easily abused, and the prescriptive language of certain provisions is merely intended to impose an obligation on a Member to do certain things when adopting certain types of measures – it does not address the question of the burden of proof in the litigation process. Are there any lessons to be learned from domestic courts and other international courts and tribunals?

A. A look at domestic and international law

It is a general principle of law that the burden of proof lies on the claimant – actori incumbit probatio or actori incumbit onus probandi. This principle had its origin in the Roman judicial system and is referred nowadays as the general rule regarding the allocation of the burden of proof in domestic as well as international judicial proceedings.¹⁰² The general and abstract character of this principle has contributed to its widespread acceptance. On account of these characteristics, however, the principle does not serve as a complete guide to the allocation of the burden of proof particularly in complex situations.

It is widely understood that the reference to the claimant should not necessarily be equated with a reference to the plaintiff from the procedural standpoint¹⁰³ – though the plaintiff is indeed the claimant in most situations.¹⁰⁴ If both the plaintiff and the defendant can be the claimant, it is then necessary to define which claims must be associated with the plaintiff and which claims (or defenses stricto sensu) must be associated with the defendant. Some common law commentators have noted that when making this distinction, courts often enquire whether a certain claim is essential to the plaintiff’s or the defendant’s case.¹⁰⁵ However, as Wigmore notes, this only advances the enquiry one step in that ‘we must then ask whether there is any general principle which determines to what party’s case a fact is essential’.¹⁰⁶ Some civil law jurisdictions have pursued a similar line of enquiry; they have sought to distinguish between facts that give rise to a legal right, those that prevent the formation of that legal right, those that modify a legal right, and those that extinguish a legal right. In this system, plaintiffs who assert a legal right in judicial proceedings must prove the facts on which the right is based. Defendants may oppose by claiming the existence of certain causes preventing the legal effect claimed by the plaintiff, or by claiming that the right has been modified or extinguished; defendants must prove the facts necessary to establish the causes that prevent, modify, or extinguish the right.¹⁰⁷ Ultimately, this strategy runs into a similar problem to the one noted by Wigmore, i.e., we must ask whether there is a general principle which determines what the facts are that give rise to a legal right as opposed to those that prevent, modify, or extinguish that right.¹⁰⁸ An answer to the questions ‘What is essential to a party’s case?’ or ‘What are the facts that give rise to a legal right as opposed to those that prevent, modify or extinguish that right?’ can only be found in the specificities of the substantive law in question.¹⁰⁹ As noted before, even the distinction between a constituent element and an exception – which at first glance appears to constitute a more general formula to determine what is essential to a party’s case – is intrinsically linked to the specificities of certain statutes and some areas of substantive law such as contract law.¹¹⁰

Rules for the allocation of the burden of proof, which are based on substantive law can provide guidance or serve as models to some international tribunals adjudicating questions similar to those dealt with by domestic courts, for instance, tribunals resolving property and contractual disputes of which the Iran–United States Claims Tribunal is an example.¹¹¹ However, their relevance is limited, if not nil, to tribunals which are in charge of adjudicating unique issues. This is the case of the WTO, where, e.g., the question of who must prove force majeure in breach of contract cases is not a question that will come before panels and the Appellate Body because they have no jurisdiction to adjudicate contractual disputes. Not only are the issues not the same, but also the structure and concept of WTO law is not the same. For example, domestic civil–private litigation reflects the existence of reciprocal/mutual relations between the parties – party A demands payment, whereas party B argues that payment has already been made; party A demands damages for destruction of its property, whereas party B argues that the property was destroyed because A placed it in an inappropriate location – and these relations may lead to a variety of types of legal actions.¹¹² Panels and the Appellate Body, on the other hand, have jurisdiction to adjudicate claims under the WTO-covered agreements;¹¹³ these include violation complaints and non-violation and situation complaints.¹¹⁴ The WTO agreement is a regulatory treaty setting limits to the actions of its Members in the field of international trade. In other words, WTO Members have agreed to comply with a set of rules and when they do not do so other Members can resort to the dispute settlement mechanism to complain about the inconsistency of those actions with the WTO agreements.

Sometimes, instead of seeking to determine whether a claim is essential to the plaintiff’s or the defendant’s case, domestic courts allocate the burden of proof on the basis of other criteria. A criterion that has been subject to much criticism and has not been widely used in recent times is that according to which the burden of proof is on the party having in form the affirmative allegation.¹¹⁵ This rule is meant to reflect the alleged difficulty of providing proof of a negative. Because language can easily be manipulated so as to state most propositions either negatively or affirmatively, there is no inherent difficulty in proving negative allegations. In fact, parties often have to prove negatives. When allocating the burden of proof, courts sometimes also take into consideration the knowledge of the parties with respect to a particular fact and the availability of evidence to the parties.¹¹⁶ The fact that a party may have better knowledge and access to evidence is usually only one among other factors justifying an allocation of the burden of proof to that party, and it is not in any way an absolute criterion in that very often parties must prove matters as to which their adversaries have better knowledge and access to proof.¹¹⁷ Perhaps the two strongest reasons for not assigning more importance to this criterion are, first, that it cannot easily be turned into a rule of general application for determining the allocation of the burden of proof. This is because while there could be certain types of cases where it would be possible to determine in advance that one of the parties would in most situations be in possession of the information necessary to discharge the burden of proof, there would be many other types of cases where the information would some times be within the knowledge of one party and at other times in the possession of her opponent. Therefore, there would always be situations where a party would be required to prove a fact that was within the knowledge of her opponent. Second, whereas it seems plausible to require a party to put forward evidence which is in her domain, there is no reason in most cases to make the party, for this reason alone, bear the burden of proof.¹¹⁸ It is important to note that even after this information is placed before the court, the court may still remain in doubt about the matter and may thus have to rule against the party bearing the burden of proof in order to settle the question.¹¹⁹ One factor that has approvingly been used as a tool to allocate the burden of proof involves the drawing of an estimate of the probabilities of the situation.¹²⁰ Based on previous knowledge about the factual patterns of a case, the burden may be allocated to the party who contends that the more unusual event has occurred. Finally, many commentators note that courts often resort to policy considerations when allocating the burden of proof.¹²¹

The criteria discussed in the previous paragraph have not been coherently and systematically applied. When this is combined with rules that are based on substantive law, we are left with the impression that domestic law has failed to build a coherent analytical framework¹²² that could provide guidance to panels and the Appellate Body in their task of allocating the burden of proof. Unfortunately, international law does not provide clear guidance either. International law in general has not developed a systematic approach to the allocation of the burden of proof going beyond the general guidance provided by the principle of actori incumbit probatio.¹²³ The ICJ has only set some rules with respect to specific types of cases that have come before the court such as those where both parties lay claim to the same object – of which rival claims to the same territory are an example.¹²⁴ The Court has also often stated that ‘it is the litigant seeking to establish a fact who bears the burden of proving it’.¹²⁵ This statement, however, is more properly understood as referring to proof of allegations rather than the burden of proof as so far discussed.¹²⁶

B. Can law and economics analysis help?

Where does this leave us? In this kind of situation, it is always useful to enquire where the problem originated. It seems that panels and the Appellate Body got confused when seeking to identify an issue or ‘defense’ that must be proved by the defendant. If we know that in general the burden of proof is allocated to the plaintiff, we should then seek to determine the specific situations when the burden should be allocated to the defendant. Therefore, a more fundamental question would be: why should the burden of proof be allocated to the defendant? Law and economics can help to provide an answer to this question.

Law and economics supports the general understanding that the burden of proof should usually lie upon the plaintiff. The point of departure is the assumption that when allocating the burden of proof, courts aim at minimizing a social loss function composed of two kinds of costs: direct or process costs, and error costs.¹²⁷ Direct or process costs are those costs incurred in the process of settling a claim – cost of presenting evidence, preparing pleadings, resources devoted by the court to deliver, and enforce a judgment, among others. Error costs consist of the social costs associated with an erroneous legal judgment, e.g., the costs associated with the frustration of objectives such as deterrence¹²⁸ or other specific objectives pursued by the substantive law.

According to Hay, the costs of allocating the burden of proof to a given party are a function of five parameters: (1) the probability the party is correct, (2) the party’s estimate of that probability, (3) the party’s cost of presenting evidence, (4) the amount at stake for the party, and (5) the social costs of an erroneous ruling against the party. Hay also demonstrates that when determining the desirability of allocating the burden to a given party, an increase in parameters 1, 3, 4, and 5 ‘has the effect of making the opposite allocation more desirable’ – parameter 2 has no clear bearing on the desirability of the allocation.¹²⁹ Taking as a baseline a class of disputes where (a) given the court’s knowledge about the dispute, the probability that the plaintiff has a meritorious claim for relief is relatively small¹³⁰ and (b) the parties’ costs of presenting evidence, the amount at stake for each party, and the social costs of an error in one or the other party’s favour are all roughly the same, an analysis of the parameters suggests that the burden of proof should be allocated to the plaintiff.¹³¹ Hay argues that because most cases resemble this baseline case, as a general rule, the burden of proof on contested issues should be on the plaintiff.¹³² If under these circumstances, the burden of proof is allocated to the defendant instead, then there could be two negative consequences. First, when the parties are not certain of the merits of a claim, ‘the plaintiff would be tempted to sue on the slimmest of grounds (when her chances of success were very small), compelling the defendant to present evidence’.¹³³ In this situation, on the one hand, if the defendant presents the evidence, then process costs will be incurred. On the other hand, if the defendant does not produce evidence, then the result will be error costs. Second, the plaintiff might threaten to sue ‘even knowing she had no claim, in the hopes of inducing the defendant to pay her to drop the claim’.¹³⁴ Process and error costs in both situations would likely be greater than if the plaintiff bore the burden of proof.¹³⁵

Based on a similar analysis of cost minimization, Lee arrives at the same conclusion that the burden of proof should, as a general rule, be allocated to the plaintiff. Lee’s conclusions, however, are based on the idea that when liability is indeterminate, the plaintiff is just as likely to be correct as the defendant. In other words, in cases where the court is uncertain about which party should prevail – the kind of cases where the burden of proof determines the outcome of the case – there is a probability of 0.5 that the plaintiff deserves to prevail and an equal probability that the defendant should prevail. According to Lee, there are two justifications for a default rule that assigns the burden of proof to the plaintiff. First, a judgment for plaintiff results in post-judgment direct costs – such as remedy construction costs, enforcement costs, and transaction costs associated with payment – which would not be incurred if the judgment was for the defendant.¹³⁶ Therefore, the default rule assures that these costs will not be incurred when the plaintiff is just as likely to be correct as the defendant. In addition, when liability is indeterminate, post-judgment direct costs cannot be justified on the grounds that they would be offset by a reduction in error costs. Only when liability is not indeterminate and the plaintiff is more likely to deserve recovery will there be an economic argument for incurring the marginal increase in direct costs.¹³⁷ Second, but still related to the first point, the default rule can be justified on the grounds that ‘it economizes on direct costs by deterring claims based on indeterminate issues of liability’.¹³⁸ Lee explains that the default rule seeks to ‘encourage only those suits that serve an allocative function (i.e., that decrease error costs)’, and this will be the case ‘only if the probability that plaintiff is entitled to recovery is greater than 0.5’.¹³⁹

Departures from the general rule that the burden of proof should be allocated to the plaintiff are explained by Lee and Hay on similar cost-saving grounds¹⁴⁰:

1. The adjudicator has advance knowledge with respect to the factual pattern of a case that demonstrates that the plaintiff probably has a meritorious case, that is to say, the defendant is liable in more than 50% of the situations (theory of a priori judgments).¹⁴¹

2. The plaintiff’s costs of presenting the evidence are higher than the defendant’s; in other words, the defendant can produce core evidence at a lower cost than the complainant (relative-cost-of-proof theory).¹⁴²

3. The costs associated with an erroneous decision in favour of the defendant are higher than those associated with a decision favouring the plaintiff (error-cost theory).¹⁴³

The discussion above suggests that a cost minimization analysis of the allocation of the burden of proof supports the general rule that in most cases the burden of proof is on the plaintiff. The burden should be allocated to the defendant only in exceptional circumstances, i.e., those suggested by the theory of a priori judgments, the relative-cost-of-proof theory and the error-cost theory. Therefore, law and economics provides a systematic and coherent framework for allocating the burden of proof.

C. What are the options?

Based on the analysis described above, there are two alternative courses of action available to panels and the Appellate Body. First, panels and the Appellate Body may adopt the view that the kind of considerations required by the theory of a priori judgments, the relative-cost-of-proof theory and the error-cost theory, belong more properly to the realm of legislative activity. Thus, when WTO Members thought that any of those factors justified an allocation to the defendant, they made an express provision to that effect; this is the case, e.g., of Article 10.3 of the Agreement on Agriculture.¹⁴⁴ Accordingly, under the first option, panels and the Appellate Body would draw on law and economics analysis to support the view that the burden of proof should be allocated to the complainant unless a provision clearly provides for a special allocation.

Second, panels and the Appellate Body may embrace the view that the kind of considerations required by the theory of a priori judgments, the relative-cost-of-proof theory and the error-cost theory, are not purely in the realm of legislative activity.¹⁴⁵ Under this second option, panels and the Appellate Body would not discard the possibility that the burden of proof with respect to provisions which do not refer expressly to that question should be allocated to the defendant. Accordingly, panels and the Appellate Body could examine provisions that provide for exceptions¹⁴⁶ with a view to determining whether the a priori judgments, the relative-cost-of-proof, and the error-cost theory would justify an allocation to the defendant.¹⁴⁷ If they conclude that an analysis of those three factors indicates that the burden of proof should be allocated to the defendant, they would then have more concrete reasons to find that that was the intention of the drafters.

The advantages of this framework include not only its coherence and logic but also the fact that it takes into account concerns related to procedure and substance. Furthermore, though the economics behind it may at first suggest that the analysis required to implement the framework is totally foreign to judges, further consideration reveals that this is not the case. When courts use an estimate of the probabilities of the situation to allocate the burden of proof to the party whose claim is the most unlikely, they apply the theory of a priori judgments. The considerations embodied in the relative-cost-of-proof theory are to a certain extent the same as those taken into account by courts when they allocate the burden of proof based on a party’s knowledge and access to proof. Finally, the error-cost theory provides a framework to consider certain policy issues.

It is important to note that in the interests of predictability as well as cost saving,¹⁴⁸ these factors are not meant to be examined on a case-by-case basis, i.e., it should not matter that in a particular instance the complainant’s cost of presenting evidence is higher than the defendant’s, if in most similar cases the costs would be the same for both parties. Instead, those factors should be used to determine the allocation of the burden of proof with respect to specific provisions.

If panels and the Appellate Body decide to follow the second path, i.e., examine provisions that contain words suggesting that they are exceptions with a view to determining whether the burden of proof should be allocated to the defendant, they could develop an analysis similar to that in the next few paragraphs. The following paragraphs briefly discuss how panels and the Appellate Body could apply law and economics analysis to allocate the burden of proof with respect to two provisions, Article 3.3 of the SPS Agreement and Article XX of the GATT. These are just examples of how the analysis could be conducted; the conclusions are not intended to be generalized, i.e., they are restricted to Article 3.3 of the SPS Agreement and Article XX of the GATT in their current form.

Article 3.1 of the SPS Agreement states that except as provided in Article 3.3 of the SPS Agreement, ‘Members shall base their sanitary or phytosanitary measures on international standards, guidelines or recommendations’. In turn, Article 3.3 provides that Members may adopt measures that result in a higher level of protection than would be achieved by measures based on international standards, guidelines, or recommendations in two situations: (1) if there is a scientific justification and (2) ‘as a consequence of the level of sanitary or phytosanitary protection a Member determines to be appropriate in accordance with’ the provisions of Article 5 (‘Assessment of Risk and Determination of the Appropriate Level of Sanitary or Phytosanitary Protection’) of the SPS Agreement. Article 3.3 of the SPS Agreement, therefore, is an exception to the rule contained in Article 3.1; it establishes the conditions Members must fulfil in order to deviate from Article 3.1. If Article 3.3 is an exception, should the burden of proof with respect to that provision be allocated to the defendant? Law and economics analysis suggests that this should not be the case.

First, application of the theory of a priori judgments does not indicate that the respondent country is liable in more than half of all possible cases. In other words, there is no reason to believe that Members that apply a measure that is not based on international standards do so without a scientific justification or proper consideration of the provisions of Article 5 of the SPS Agreement in more than 50% of the cases.

Second, contrary to what seems to be the case at first glance, the respondent’s costs of presenting evidence, which concerns its own measure, are not necessarily significantly lower than the complainant’s – relative-cost-of-proof theory. The SPS Agreement contains in Article 5.8 and section 3 of Annex B obligations requiring Members to provide, when requested, information regarding their SPS measures.¹⁴⁹ It is reasonable to expect that a complainant will conduct an investigation of the situation before beginning a case against the Member applying the SPS measure. The provisions of Article 5.8 and section 3 of Annex B facilitate that investigation and, thus, are designed to ensure the even distribution between the parties of the costs of presenting evidence during the litigation process.¹⁵⁰ Moreover, when the obligations contained in Article 5.8 and section 3 of Annex B are placed in the broader context of the SPS Agreement, there is further reason to believe that the costs of presenting evidence would not justify a departure from the general rule that the burden of proof should be allocated to the complainant. The obligations in the SPS Agreement essentially require Members to take certain actions and observe certain procedures when designing and adopting SPS measures. In this context, if a Member does not provide information regarding its measure, then this may be interpreted by panels as an indication that those actions and procedures were not followed; otherwise, evidence regarding their observance would be readily available and should be made available as prescribed in Article 5.8 and section 3 of Annex B.¹⁵¹ For example, if the Member imposing an SPS measure does not provide the risk assessment supporting that measure, then this will likely be interpreted as evidence that the Member in question did not conduct such risk assessment as required in Article 5 of the SPS Agreement. It is important to remember that even after evidence that could be more easily presented by the Member imposing the measure is made available, e.g., the risk assessment, there is still a need to demonstrate that that risk assessment does or does not comply with the requirements in Article 5 of the SPS Agreement. There is no reason to believe that the costs of presenting the evidence required for such demonstration are necessarily higher for the complainant than for the defendant.¹⁵² Therefore, the relative-cost-of-proof theory would not suggest that the burden of proof with respect to Article 3.3 of the SPS Agreement should be allocated to the respondent country because the complainant’s costs of presenting evidence are generally higher than the respondent’s.

Third, application of the error-cost theory does not conclusively indicate that the costs of an erroneous decision in favour of the defendant would be higher than the costs of an erroneous decision in favour of the complainant and vice-versa. On the one hand, an erroneous decision in favour of the complainant would result in costs such as those social costs associated with the removal of a measure that achieves a Member’s correctly determined appropriate level of protection or the removal of a measure that is scientifically sound.¹⁵³ Erroneous decisions favouring the complainant might also create overdeterrence costs. In other words, it is possible that errors favouring the complainant in litigation may discourage Members from deviating from international standards in situations where it would be efficient to do so – e.g., in situations where the international standard needs to be improved. However, overdeterrence might not occur. Instead, the fact that these types of errors might occur may create an incentive for Members to try to change/improve the international standard. On the other hand, erroneous decisions in favour of the defendant – i.e., defendants win when they should lose – would result in little harmonization being achieved which in turn would result in the weakening of international standards.¹⁵⁴ The costs of unnecessary diversity may be high; trade will be disrupted because of the costs of complying with several different standards without any corresponding benefits to human, animal, or plant life or health.¹⁵⁵ This discussion suggests that there is no clear answer to the question of whether an erroneous decision in favour of the defendant would be more costly than an erroneous decision in favour of the complainant or vice-versa.¹⁵⁶ Therefore, the error-cost theory would not suggest that the burden of proof with respect to Article 3.3 of the SPS Agreement should be allocated away from the complainant.¹⁵⁷

Thus, if panels and the Appellate Body were to conduct a law and economics analysis, they would conclude that there is no reason to deviate, in the case of Article 3.3 of the SPS Agreement, from the general rule that the burden of proof should be allocated to the complainant. None of the law and economics factors – theory of a priori judgments, relative-cost-of-proof theory, and error-cost theory – provide a reason to allocate the burden of proof away from the complainant. As discussed before, this conclusion actually corresponds to the conclusion of the Appellate Body in EC – Hormones, where, contrary to the panel, the Appellate Body allocated the burden of proof with respect to Article 3.3 of the SPS Agreement to the complainant.

The discussion now turns to Article XX of the GATT. Article XX provides that Members may adopt measures which would otherwise be inconsistent with provisions of the GATT if those measures are, e.g., (1) necessary to protect public morals, or human, animal, or plant life or health and (2) provided that those measures are not applied in a manner which would constitute a means of arbitrary or unjustifiable discrimination between countries where the same conditions prevail, or a disguised restriction on international trade. In short, Article XX establishes an exception to obligations contained in other Articles of the GATT; it establishes conditions that Members must fulfil in order to deviate from other GATT rules. If Article XX is an exception, should the burden of proof with respect to that provision be allocated to the defendant? Contrary to the prevailing view in the GATT and WTO jurisprudence, law and economics analysis suggests that this should not be the case.¹⁵⁸

The starting point should be the general rule, i.e., that the complainant should bear the burden of proof with respect to Article XX. Unless any of the three law and economics factors or a combination of them indicates that the burden of proof should be allocated to the respondent, the burden should remain on the complainant.

It is possible that the theory of a priori judgments indicates that the respondent country is liable in more than half of all possible cases. If we examine regulations that are, e.g., inconsistent with Articles III (‘National Treatment on International Taxation and Regulation’) and Article XI (‘General Elimination of Quantitative Restrictions’) of the GATT, we may conclude that more than half of them are ‘protectionist’.¹⁵⁹ Such a conclusion is based on the assumption that governments tend to favour their domestic producers in designing their laws;¹⁶⁰ domestic regulatory and legislative processes provide greater opportunities to influence regulation and legislation for domestic producers than for foreign producers and domestic constituents which may be negatively affected. However, in the absence of statistical evidence to corroborate this assumption, and in the light of the fact that this assumption may be controversial among Members, we may opt to keep this factor neutral.

As for the costs of presenting evidence – relative-cost-of-proof theory – there seems to be no general cost imbalance between the complainant and the respondent. While it is the defendant’s measure that is at issue, it is likely that there will be information regarding that measure in the public domain. Perhaps the best evidence of this is that so far in Article XX cases both parties have submitted evidence; if the costs of producing evidence were too high, then parties would likely have refrained from presenting evidence. In addition, it is arguable that with respect to the conditions in the chapeau, complainants are in fact in a better position to provide information regarding how their exports are, e.g., being arbitrarily or unjustifiably discriminated against. Regarding the ‘necessity’ of the measure, the complainant may also in many cases be in a better position to present information regarding less-trade-restrictive measures that could substitute for the measure being challenged.¹⁶¹ Therefore, it does not seem to be the case that the defendant (or the complainant) could always present evidence at a lower cost than the complainant (or the defendant).

An analysis of the costs of an erroneous decision in favour of either the complainant or the defendant – error-cost theory – does not support the conclusion that the burden of proof should be allocated to the defendant. An erroneous decision in favour of the complainant – the defendant loses when it should win – would result in social costs such as the costs associated with the removal of a measure which fulfils an important objective such as protection of public morals, human, animal, or plant life or health.¹⁶² Erroneous decisions in favour of the complainant may also be costly for trade liberalization; such decisions may produce a negative impact on public opinion, which may lead governments to adopt a more restrictive approach to trade liberalization. However, erroneous decisions in favour of the defendant may also be costly for trade liberalization; they may lead Members to treat liberalizing offers from other Members with scepticism if they believe that concessions will later be undermined by protectionist measures of the type that would be allowed if liability under Article XX is wrongly determined. Overall, it seems that the costs of erroneous decisions in favour of the complainant would be higher than the costs of erroneous decisions in favour of the defendant. Accordingly, the error-cost theory would recommend that the burden of proof with respect to Article XX remain on the complainant.

Therefore, if panels and the Appellate Body were to conduct a law and economics analysis, then they would conclude that there is no reason to deviate, in the case of Article XX of the GATT, from the general rule that the burden of proof should be allocated to the complainant. This is so because none of the three factors suggests an allocation to the defendant; even if we considered that the theory of a priori judgments might suggest an allocation to the defendant, then this would likely have been neutralized by the fact that the error-cost theory indicates that the burden of proof should be allocated to the complainant. If all factors remain neutral, the general rule should remain unaltered, i.e., the burden of proof should rest with the complainant.¹⁶³

One could argue that it is costly to allocate the burden of proof with respect to Article XX to the complainant because the complainant does not necessarily know which of the paragraphs of Article XX the defendant relies upon. The complainant would be in a situation where it would have to prove that the defendant does not comply with the requirements of all of the paragraphs of Article XX. While it is true that it would be costly if the complainant had to prove that the defendant did not comply with the requirements of all or several of the paragraphs of Article XX, this situation can be avoided if the defendant is required to specify the paragraph(s) that it is relying upon and elaborate on the reasons why it believes it is entitled to justify its measures under that (or those) paragraph(s). In other words, the defendant would have the burden of pleading Article XX.

The requirement for parties to raise or plead certain issues has the function of facilitating efficient communication between parties in litigation or, in other words, it has the function of refining the scope of a case.¹⁶⁴ Similar to the burden of proof, as a general rule, the burden of pleading is allocated to the complainant. Lee explains that an allocation of the burden of pleading to the defendant is economically sound in two contexts. First, when certain issues are unlikely to arise – probability theory. In such cases, ‘assigning the burden of pleading to the defendant operates as a presumption that saves the cost of raising the issue in most cases’.¹⁶⁵ Second, ‘[o]n some issues assigning the burden of pleading to plaintiff would require him to make an allegation that would be much broader and more comprehensive than if the burden were assigned to defendant’.¹⁶⁶ In this situation, assigning the burden of pleading to the defendant, the party whose version of the issue in question is more discrete, economizes on direct costs – relative-cost-of-pleading theory.

The idea of allocating the burden of pleading to a different party than the party that bears the burden of proof was embraced by the Appellate Body in EC – Tariff Preferences. In EC – Tariff Preferences, the Appellate Body allocated the burden of proving the Enabling Clause to the defendant, the European Communities, and the burden of pleading that clause to the complainant, India. The rationale for allocating to different parties the burden of proving and the burden of pleading the Enabling Clause, however, is questionable. The Appellate Body invoked the importance of the Enabling Clause¹⁶⁷ and its special status in the covered agreements¹⁶⁸ to justify its decision of placing the burden of raising – i.e., burden of pleading – the Enabling Clause on the complainant. Interestingly, the Appellate Body also sought to justify its decision on the grounds that ‘[e]xposing preference schemes to open-ended challenges would be inconsistent [ . . . ] with the intention of Members [ . . . ] to “encourage” the adoption of preferential treatment for developing countries and to provide a practical means of doing so within the legal framework of the covered agreements’.¹⁶⁹ If the Appellate Body was seeking to restrain unfounded challenges to a measure taken under the framework of the Enabling Clause, placing the burden of pleading on the complaining party is hardly an effective way of achieving that objective.¹⁷⁰ The reasons provided by the Appellate Body in support of its decision to allocate the burden of pleading the Enabling Clause to the complainant would seem to justify an allocation of the burden of proof to the complainant. If Members place so much importance on encouraging preferential treatment to developing countries, then this would suggest that in cases involving the Enabling Clause, erroneous decisions in favour of complainants – defendants lose when they should win – are more costly than erroneous decisions in favour of defendants. Therefore, without a more detailed analysis of all the law and economics factors, application of the error-cost theory would suggest that the burden of proof with respect to the Enabling Clause should remain on the complainant.

It is important to note that nothing in the agreements, in particular the DSU, would prevent panels and the Appellate Body from conducting an analysis similar to that in the examples above. The DSU is completely silent with respect to the question of the allocation of the burden of proof. The absence of rules in the DSU has not prevented panels and the Appellate Body from deciding questions regarding the allocation of the burden of proof so far. In fact, because panels cannot declare non liquet, decisions regarding the burden of proof are unavoidable when the evidence is not sufficient to convince panels of the truth or fallacy of a proposition. Accordingly, irrespective of the methodology used, decisions regarding the burden of proof, when no express provision is made in the agreements, do not add to or diminish the rights and obligations of Members.

The advantage of this approach over the current one is that it provides objective reasons to support a certain allocation of the burden of proof.¹⁷¹ It can also improve predictability to the extent that the parties themselves can make an assessment of whether the a priori judgments, the relative-cost-of-proof, and the error-cost theories would suggest an allocation of the burden of proof to the defendant. Predictability, however, would obviously be stronger if panels and the Appellate Body took the first path, i.e., allocate the burden of proof away from the complainant only when the negotiators have expressly provided so.

CONCLUDING REMARKS

The preceding analysis has shown that the jurisprudence on the allocation of the burden of proof is in a confused state. Panels and the Appellate Body have failed to produce a consistent line of cases, which can be used as a predictable model to solve future cases. Furthermore, the jurisprudence has also created what appear to be artificial differences between provisions, which raises questions about the relevance of the criteria employed to distinguish provisions that must be proved by the defendant from those that must be proved by the complainant. Panels and the Appellate Body have got so entangled in the details of those distinctions that they seem to have neglected the basics. The analysis undertaken in this article suggests that it may be time to reflect upon the basic question of why the burden of proof should be allocated to a given party.

The law of proof economics literature provides a convincing explanation of why the burden should be allocated in most cases to the complainant and of the reasons that should motivate an allocation to the defendant. Following the guidance of that literature, panels and the Appellate Body could develop a clearer and more sophisticated analysis. They could adopt two alternative courses of action, assigning two different meanings to the silence of the drafters. Under the first option, silence would be interpreted as an endorsement of the rule that the burden of proof should be allocated to the complainant. Under the second option, the drafters’ silence would not be given a specific meaning, i.e., panels and the Appellate Body would be free to analyse provisions, which do not expressly refer to the question of the burden of proof, with a view to determining whether the theory of a priori judgments, the relative-cost-of-proof theory, or the error-cost theory would suggest that the burden of proof should be allocated to the defendant. However, given the importance of a concept such as the burden of proof, which may determine the outcome of a case, it would be wise for the WTO negotiators to establish clear rules for its allocation. In establishing such rules, they can of course use the analytical framework proposed by the law and economics literature. WTO negotiators would in any event be in a better position to undertake a more thorough analysis and make value judgments if necessary.

I am grateful to my SJD supervisor, Professor Michael Trebilcock, for his support and comments on earlier drafts. I thank Hiromi Yano and three anonymous referees for their helpful comments.