2. Facts

Most food products enter the United States under the jurisdiction of the Food and Drug Administration, which relies primarily on at-the-border inspection.Footnote ⁴ But poultry (including poultry products) is under a different system, established by the Poultry Products Inspection Act (PPIA), adopted in 1957 and amended at various times. Under regulations adopted pursuant to the PPIA by the United States Department of Agriculture (USDA), the Food Safety and Inspection Service (FSIS), a component agency of the USDA, administers what the United States refers to as an ‘equivalence regime’ for poultry products. No shipment of poultry products is allowed to enter the United States unless the FSIS certifies that the country of origin has a food-safety regime for poultry products that is the equivalent of the United States' regime. Specifically, the exporting country must have a regime that can guarantee that its poultry products are ‘healthful, wholesome, fit for human food, not adulterated, and contain no dye, chemical, preservative, or ingredient which renders them unhealthful, unwholesome, adulterated, or unfit for human food’Footnote ⁵ – in a word, as the United States summarizes it, poultry products must be ‘safe’.

The procedure, in outline, is as follows: A country that wishes to export poultry to the United States applies to the FSIS for certification that its poultry-safety regime can guarantee the safety of its poultry exports. The FSIS conducts a document review and then an on-the-ground inspection of the operation of the applicant country's poultry-safety regime. If the FSIS is satisfied, it publishes in the Code of Federal Regulations (CFR) a proposed regulation certifying the applicant country. After the period for comments from interested parties, the FSIS can publish in the CFR a final regulation certifying the applicant country. The next step is for the applicant country to give the FSIS a list of producers in the country that it regards as able to guarantee the safety of poultry products. The FSIS can inspect these producers. Once the country and producers are certified, export to the United States can begin, subject always to spot-checks at the border, and subject to annual recertification reviews for both exporting country and producers.

In April 2004, China applied for certification to export poultry to the United States. The FSIS found some problems, which China partially corrected. The FSIS remained dissatisfied with China's supervision of the poultry slaughtering process, but it was willing to certify China for export of poultry products, provided the products were made from poultry slaughtered in the United States or some other country certified to export poultry carcasses to the United States.Footnote ⁶ A final rule certifying China for export of poultry products was published in the CFR in April 2006. In May 2006, the FSIS asked China for a list of producers who should be certified. No such list was forthcoming until March 2008, perhaps in part because of other developments in the interim.

In December 2007, the United States Congress adopted the Agricultural, Rural Development, Food and Drug Administration, and Related Agencies Appropriations Act (AAA) of 2008. Section 733 of this Act prohibited the FSIS from using any appropriated funds for establishing or implementing a rule to allow import of poultry products from China. Section 733 expired in September 2008, and it was not challenged in this case. But it had a successor, Section 727 of the AAA 2009, adopted in March 2009 (and expired in September 2009), which also prohibited the FSIS from using funds to establish or implement a rule to allow import of poultry products from China. Section 727 is the only measure under review in the present case.Footnote ⁷ Section 727 was adopted less than two weeks after China adopted a comprehensive new food-safety law in February 2009. China's request for consultations was made in April 2009, a month after the enactment of Section 727. In May 2009, the United States asked China for information about the new food-safety law; China did not respond, explaining that that was because it had already initiated dispute-settlement proceedings.

Those are the basic facts. It is also natural to speculate about Congress' motivation for Section 733 of the AAA 2008 and Section 727 of the AAA 2009. Congress' motivation is not formally an issue in the case, except possibly under SPS 2.3, the third prong of SPS 5.5, and the GATT Article XX chapeau, all of which refer to a ‘disguised restriction on international trade’. But what we assume, consciously or unconsciously, about Congress' motivation is likely to affect our view of all the issues. As the United States argued in US–Poultry (China), Section 727 was almost certainly not a protectionist measure. It is not clear that the poultry industry actively opposed the adoption of Section 727, because it is not clear they were involved at that point; but along with a wide coalition of US exporters, mostly food-product exporters but not limited to those, the poultry industry did actively oppose the extension of Section 727 for another year when the AAA 2010 was being considered in Congress.Footnote ⁸ Also, a similar coalition of American exporters (including the poultry industry) offered public comments to the USTR about the US–Poultry (China) dispute itself, taking China's side.Footnote ⁹ These exporters presumably feared retaliation by China or hoped, in the case of beef exporters, that if poultry products were allowed in from China, they might end their ban on US beef (motivated by worries about BSE in US cattle). Of course, even if Section 727 was not protectionist, one might suspect that it was motivated by Congressional irritation with China over a variety of trade frictions, especially Congress' view that China was engaged in currency manipulation. But there is no evidence of such motivation in the legislative history, and we should not dismiss the possibility that Section 733 and Section 727 were motivated by genuine concern for food safety, even though they involved the unusual step of Congress overruling the FSIS. In March 2007, some months before the adoption of the AAA 2008, there was a major scandal when many pets were sickened or died in the United States after eating pet food made from vegetable protein imported from China that was contaminated with melamine. China also had an internationally publicized domestic scandal involving melamine in milk and infant formula in July 2008, as well as melamine found in eggs from China in Hong Kong in October 2008 – both of these episodes occurring while versions of what became the AAA 2009 were being considered in Congress. In 2007, the Asian Development Bank issued a report which said that ‘unsafe food in the PRC remain[ed] a serious threat to public health’.Footnote ¹⁰ The United Nations Resident Coordinator in China issued a report in 2008 which said there was ‘an overall lack of resources in China for food inspection and compliance assessment services’; that enforcement of food control placed an ‘excessive reliance on end-product testing with very little use of auditing as an inspection tool’, which was ‘both inefficient and ineffective’; and that ‘in general, surveillance and monitoring systems [for food contaminants and food-borne diseases]’ were limited.Footnote ¹¹ In March 2009, days before the adoption of Section 727, Reuters reported that the Chinese Ministry of Health said: ‘At present, China's food-security situation remains grim.’ With all of this (and more)Footnote ¹² in mind, Congress might quite reasonably have worried that the FSIS had been unduly lax in reviewing China's poultry-safety regime, that it had fallen down in its duty to protect consumers against unsafe poultry products because it had caved in to political pressure.Footnote ¹³

3. Preliminaries: the Panel's jurisdiction, and the status of Section 727 as an SPS Measure

As noted in the Introduction, the United States failed to make their best argument under SPS 5.1 (which I shall discuss in Section 4). Instead, they tried to do an end-run around 5.1 and 2.2. They argued that even though Section 727 had a food-safety purpose, it was not an SPS measure, because it was a mere procedural adjustment to the United States’ ‘equivalence regime’ for poultry products; and they argued in the alternative that even if Section 727 was an SPS measure, its status as part of an equivalence regime meant it was controlled only by SPS Article 4. The first of these claims contributed to a wrangle over the Panel's jurisdiction. And both of the claims obviously needed to be adjudicated by the Panel before they could proceed to the 5.1 discussion.

4. Whether Section 727 violates SPS 5.1 and SPS 2.2

At this point, the Panel has rejected two arguments by the United States: (1) that Section 727 is not an SPS measure, because it is merely ‘procedural’, and (2) that even if Section 727 is an SPS measure, it is not subject to any SPS discipline except Article 4. The Panel was right to reject both of these arguments. Still, behind these arguments there is a genuine and serious issue. The core of the United States' position, I suggest, was the idea that Section 727 should not be subject to the risk-assessment requirement of SPS 5.1, at least as it has been developed in previous cases, because Section 727 is a different sort of measure from the measures involved in all previous cases. And Section 727 could not, in principle, have the same sort of scientific justification that we expect from those other measures.

In order to investigate the justification for ‘standard’ SPS measures like those in previous cases, we need to ask questions like: ‘Is Salmonella dangerous to humans?’; ‘Do chickens or other types of poultry harbor Salmonella?’; ‘Does vaccinating chickens stop or slow the spread of Salmonella?’; ‘Does heating chicken feed reduce the incidence of Salmonella?’; ‘What inspection techniques and sampling procedures do the best job of identifying Salmonella in poultry (live or slaughtered or processed)?’; ‘What quarantine procedures are necessary to stop the spread of diseases such as avian flu?’; and so on. These are all general questions about human or animal virology, statistics, epidemiology, and the like. The generality of these questions reflects the fact that ‘standard’ SPS measures are origin-neutral. But Section 727 is not origin-neutral, and the crucial question for the justification of Section 727 is: ‘How well does China do at enforcing its poultry regulations?’. This is a completely different sort of question from those listed above, and it requires a completely different sort of investigation. In asking this question, we may presuppose that China has excellent SPS regulations on paper; we may presuppose that its poultry-safety regulations on paper are equivalent to the United States' regulations. But even if China has excellent regulations on paper, the question remains how well it puts them into practice. This is a question about specific political, and social, and cultural behavior in a particular time and place. There are, of course, better and worse ways of trying to answer this question, but the methods that are appropriate to answering this question are not the methods that are appropriate to answering the questions raised by standard, origin-neutral, SPS measures. In particular, the standard understanding of a ‘risk assessment’ seems completely inapposite to investigating the question of how well China enforces its food-safety laws.Footnote ²³

Here is another way to make the same point. SPS Article 3 plainly involves the assumption that SPS measures deal with the sorts of problems we could have international standards about. Members are conditionally required to base their SPS measures on international standards, when there are such; and measures that conform to international standards are presumptively legal. But what would it mean to have an international standard on the question of whether China enforces its food regulations? We could imagine an international standard on how countries should respond to inadequate enforcement by other countries, but I am not aware of any such standards; and even such a standard would not tell us whether any particular country was falling down on enforcement. We cannot imagine an ‘international standard’ that tells us China is failing at enforcement, because such a judgment would not be a ‘standard’ in any sense. Of course, an international body might tell us that China is failing at enforcement. In fact, more than one international body had done precisely that around the time Section 727 was being adopted, as the United States pointed out. But aside from the fact that such a judgment is not a standard, it does not tell us what sort of justification the United States must have to find that China is not enforcing its rules on poultry safety. There is nothing in the SPS Agreement to suggest that a Member cannot find and act on the proposition that another Member's enforcement regime is inadequate unless there is a prior finding by a relevant international body, no more than there is anything in the Agreement that tells us a Member cannot have an SPS standard unless there is an international standard on the matter.

The point of the previous two paragraphs is not that the United States had adequate justification for Section 727 and should have prevailed in US–Poultry (China). The point is just that SPS 2.2 and 5.1 arguably should not be thought to require the same sort of justification for Section 727 that they require for the measures in previous SPS cases. ‘Scientific justification’ is simply not possible for a measure like Section 727, if by ‘science’ we mean laboratory methods, or extra-laboratory controlled experiments, or even broader epidemiological and statistical techniques as they are used to investigate general questions about nature (including the consequences of various human interventions such as vaccination, quarantine, and so on). There is, however, a broader understanding of ‘science’ available. The Appellate Body in EC–Hormones, responding to a narrow understanding of science on the part of the Panel, suggested that science should be understood as ‘a process characterized by systematic, disciplined and objective enquiry and analysis, that is, a mode of studying and sorting out facts and opinions’.Footnote ²⁴ The crucial elements here are impartial concern for facts and reasoned inference from the facts. On this understanding of science, the question about China's enforcement behavior is subject to scientific investigation, and it does make sense to ask for ‘scientific justification’ for Section 727. It could even make sense to ask for a ‘risk assessment’ properly understood – that is, provided we remember that SPS 5.1 calls for a risk assessment ‘as appropriate to the circumstances’. The crucial question about Section 727 is: What sort of risk assessment is appropriate?Footnote ²⁵

So we have two interesting and important questions the US–Poultry (China) Panel does not discuss – what sort of scientific justification is required under SPS 2.2 for a measure like Section 727, and what sort of risk assessment is required under SPS 5.1, if any? The blame for this omission rests primarily on the United States. The United States spent most of its submissions trying to establish its Article XX(b) defense to China's GATT claims. With regard to China's SPS claims, the United States did little more than reiterate its assertion that the Panel had no jurisdiction to rule on the SPS. The United States did claim that Section 727 had scientific justification, but they made the argument rather lackadaisically, essentially just pointing to the evidence they introduced in connection with GATT XX(b), without any focused argument about why it should count as ‘scientific justification’ in the SPS context, and without even suggesting that it might constitute a risk assessment ‘appropriate to the circumstances’.

In fact, as we saw in Section 2, the United States had a good deal of evidence to work with. This evidence included studies critical of the Chinese food-safety regime from the United Nations, the Asian Bank for Development, and the United States' own Department of Agriculture, as well as critical statements from the WHO. There was an extensive list of recent food-safety crises involving both exports from China and domestic Chinese markets. There was even the statement by the Chinese Health Minister that the food-safety situation in China was ‘grim’; this admission very much against interest was almost exactly contemporaneous with the adoption of Section 727. Surely, much of this evidence counts as ‘scientific’ in a sense plausibly relevant to SPS 2.2, and we know that Members can justify their SPS measures on the basis of investigations carried out by others, even after the measure was adopted. So it seems that this evidence might well be adequate scientific basis for Section 727 – and that elements of it might even be an appropriate ‘risk assessment’ in the context. The most obvious deficiency of this evidence, as noted by the Panel, is that little of it addresses the safety of poultry in particular. The evidence is almost all about other food products. On the other hand, it is not a big leap to suggest that a food-safety system that has fallen down dramatically in regard to milk, infant formula, eggs, pet food, chicken feed, spinach, pork products, and elsewhere might have problems with poultry products as well. Given the difficulty of investigating the effectiveness of enforcement on the ground, we should arguably not import the Appellate Body's stringent requirement on the specificity of risk assessments into a context like this, especially in connection with a measure whose life was limited by its own terms to six-and-a-half months.

Mentioning the time-limitation of Section 727 reminds us of another difference between origin-specific SPS measures and origin-neutral ones. The propositions that justify origin-neutral SPS measures – propositions like ‘Salmonella is dangerous to humans’ – can be assumed, once they are established, to be true for the long indefinite future. In contrast, the proposition, ‘China is doing a poor job of enforcing its food-safety rules’, even after it is established for the present, can become false in a relatively short time. Indeed, all the national and international bodies that issued reports critical of China's food-safety regime also noted that China was making serious efforts to improve things. So the question arises whether an origin-specific SPS measure should be required to be time-limited (always with the possibility of renewal, of course, after appropriate investigation). Whether and how WTO tribunals could impose and administer such a requirement of time-limitation would be one topic too many for this essay.

As I have said, the United States did not argue its claim of scientific justification under SPS 2.2 very vigorously. And it did not argue at all that its evidence counted as a risk assessment under 5.1.Footnote ²⁶ Perhaps the United States, with its arguments that Section 727 was not an SPS measure, or that it was subject only to SPS Article 4, was taking a chance on winning this case ‘on the cheap’, but it did not want to make the arguments I have suggested about SPS 2.2 and 5.1 because it thought its own future interests as a complainant in other cases might be compromised by actively pursuing the idea that SPS 5.1 could be satisfied by something less than a traditional risk assessment in any circumstances. (Similarly, the United States might have thought that in the long run, it had more to lose than to gain from encouraging appeal to SPS 5.7, which it also might have raised, but didn't.) Given the litigating posture of the United States, the Panel had little choice but to find a 5.1 violation, and hence a 2.2 violation. But from a systemic point of view, it was an opportunity missed.

5. SPS 5.5, SPS 5.6, and the ‘Appropriate Level of Protection’

We turn now to the Panel's analysis of SPS 5.5 and 5.6, where there is more to criticize. The first question under SPS 5.5 was whether the United States had a different ‘appropriate level of protection’ [ALOP] for poultry from China than it had for poultry from other countries (most particularly from Mexico, which China claimed had had food-safety crises just as China had). China claimed that the import ban of Section 727 made it clear that the United States had a zero-risk ALOP for China, while it obviously allowed for some risk for poultry from other countries that exported to the United States. The United States argued that it did not have a zero-risk ALOP for China; it had the same ALOP for China as for everyone else. This was the ALOP stated by the PPIA, which was that poultry should be wholesome, unadulterated, and fit for human consumption, in a word, ‘safe’ (but obviously not safe to a zero-risk tolerance). China might have argued that the PPIA did not specify the risk level with adequate precision, but it seems to have been accepted by both parties and the Panel that the PPIA stated a (non-zero) ALOP precisely enough to work with. Hence, the only question was whether the United States was applying that ALOP to poultry from China.

The Panel noted that the Appellate Body in Australia–Salmon Footnote ²⁷ had said that when a country did not state the ALOP on which a particular measure was based, or did not state it with sufficient precision, the Panel could infer the actual ALOP from the effects of the measure. Of course, in this case, the United States had stated an ALOP for Chinese poultry – the basic ALOP of ‘safe’ stated in the PPIA. But the Panel said that even in some cases where the regulating country had stated an ALOP, it must be possible for the Panel to look behind that statement, and to infer a different actual ALOP from the effects of the measure. As the Panel says: ‘To ignore the measure and rely solely on a Member's declared ALOP could permit a Member to evade the disciplines of Article 5.5 by simply declaring one generic ALOP for all SPS-related measures.’Footnote ²⁸ The Panel's general point is perfectly correct. Imagine, for example, that in Australia–Salmon Australia had claimed it had the same very-low-risk ALOP for diseases carried by baitfish and ornamental fish that it had for the same diseases carried by imported salmon. The Panel would have been quite right to reject this claim, because in view of the actual control measures applied to baitfish and ornamental fish, the claim would have been utterly implausible. So the Panel would have been right to reject the claimed very-low-risk ALOP for baitfish and ornamental fish and infer the actual ALOP from the effects of the actual control measures.

But the United States' claim that its ALOP for China was the same as its ALOP for Mexico is not implausible in the same way. China's recent record on food safety was worse than Mexico's, and China did not have the same track record as Mexico on working with the United States to correct deficiencies in its food-safety regime. The United States (specifically, Congress) could well have thought that the usual FSIS procedures were adequate to achieve ‘safe’ poultry from Mexico, but that they would not be adequate, at least at present, to achieve ‘safe’ poultry from China. It could well have thought that the only way to achieve ‘safe’ poultry from China involved a temporary import ban and some overshooting of the goal. But still, if the only available measure to achieve the stated ALOP happens to achieve a higher level of protection, this does not mean the Member now actually has the higher ALOP. The right to choose one's ALOP must include the right to overshoot if that is necessary.

Note that the question at this point is not whether the United States was correct in thinking that the only way to achieve their stated ALOP for China (the ALOP of ‘safe’, stated in the PPIA) was to overshoot. The question is whether they could plausibly have thought so. If they could plausibly have thought so – and they surely could – then there is no reason to go behind their claim about what their ALOP was. So the Panel should have found that the United States' ALOP, even with regard to China, was simply ‘safe’. The question whether the United States was correct in its belief that the ALOP of ‘safe’ could be achieved for Chinese poultry only by Section 727 is properly raised in a different context, a 5.6 challenge. If China can show that the stated ALOP could have been achieved by the standard FSIS procedures, without Section 727, then the standard FSIS procedures would be a less-trade-restrictive alternative than Section 727, and Section 727 would violate 5.6. But in a 5.6 challenge, the burden would clearly be on China to show that the FSIS procedures would suffice. One problem with the Panel's mode of proceeding is that it in effect relieves China of this burden that it would face in a 5.6 challenge. By saying that, for 5.5 purposes, the United States' ALOP for China is zero-risk (or at least higher than the ALOP for other countries), the Panel in effect finds for itself that Section 727 is not necessary to achieve the lower stated ALOP (even though it says correctly in its subsequent 5.6 analysis that it has no adequate ground for such a finding).

The Panel has a possible response to this criticism. The Panel says at one point that, ‘to prove that such substantially different measures were needed to achieve the same ALOP, the United States would have to demonstrate that poultry products from China presented a greater risk than poultry products from other WTO Members. The United States attempts to meet its burden’.Footnote ²⁹ The implicit suggestion is that the use by the United States of ‘substantially different measures’ shifts the burden of proving whether there are different ALOPs from China to the United States. But as a general proposition, that makes no sense. If it is not facially implausible that the substantially different measures for two cases are necessary to achieve the same ALOP in those cases – and it is not implausible here – then the burden of showing that the different measures in fact reflect different ALOPs should remain on the complainant, where SPS 5.5 puts it.

I suspect the Panel is actually relying on something more than just the fact that the United States has substantially different measures for two similar cases. It is relying on the precise nature of one of the measures. It is relying on the origin-specificity of Section 727, on the fact that there is one measure for China and a different measure for all other countries (which is quite different from having one measure for salmon and a different measure for ornamental fish). Now I agree that when the regulating Member makes a de jure distinction between the SPS measures for different countries, it is (normally) appropriate to place the burden on the regulating country to justify that distinction.Footnote ³⁰ But that suggests that 5.5 is really a red herring. China's complaint should be considered directly under 2.3, which embodies the basic MFN principle in the SPS context. SPS 5.5 is a specification of 2.3, but it is a specification designed precisely for dealing with origin-neutral measures like those in EC–Hormones or Australia–Salmon. It is for smoking out MFN violations or protectionism effected through distinctions in origin-neutral measures.

Moving then to a 2.3 analysis, note that here we must put the burden of justification for origin-specific measures on the regulating country, else the SPS would be weaker in this respect than the GATT, which it was supposed to strengthen. Under GATT Article I, an origin-specific measure distinguishing between other Members is an automatic (prima facie) violation, under the ‘hypothetical like products’ analysis. Then the burden is on the regulating country to justify the de jure distinction under Article XX. The SPS Agreement does not have the same structure of prima facie violations and defenses, and it is written so that, on its face, the burden of proving non-justification (in various ways, such as proving an arbitrary or unjustifiable distinction, or identifying a less restrictive alternative) is always on the complainant. Hence, we must create presumptions to shift the burden of justification for origin-specific measures if the SPS is not to be more favorable to regulating countries than the GATT. In sum, if the Panel was indeed responding specifically to the origin-specific nature of Section 727, then it was right to impose on the United States the burden of justifying the distinction between China and other countries. But the proper vehicle for this analysis was 2.3 itself, which makes no mention of the ALOP. The entire ALOP analysis under 5.5 was unnecessary.Footnote ³¹

Note also that even if it is right to impose the burden of justification of origin-specific measures on the regulating country under 2.3, this does not mean that the relevant justification must take the form of a risk assessment, which is not mentioned in 2.3 or in 5.5. The most distinctive feature of the SPS Agreement is the risk-assessment requirement in 5.1. But it does not follow that we should therefore read the risk-assessment requirement into every other provision. The Panel in effect reads the risk-assessment requirement into 5.5, and that is a mistake. The Panel does this in connection with the second prong of the 5.5 analysis. They find that the supposed distinction in the ALOPs for poultry from China and from other countries is ‘arbitrary or unjustifiable’ simply on the ground that there was no risk assessment for Section 727.Footnote ³² One problem with this is that it seems inconsistent with the Appellate Body's finding in EC–Hormones that the distinction between the ALOPs for artificially introduced growth hormones in beef and for naturally occurring growth hormones was not ‘arbitrary or unjustifiable’. (Or, similarly, the finding that the distinction between ALOPs for hormones introduced for growth purposes and hormones introduced for therapeutic or zootechnical purposes was not arbitrary or unjustifiable.) The Appellate Body had found that there was no risk assessment for hormones introduced for growth purposes; but it obviously did not think it followed that every case of different ALOPs, where one of the ALOPs was for hormones introduced for growth purposes, was ‘arbitrary or unjustifiable’.

The more general problem is that by relying on the absence of a risk assessment to find a 5.5 violation, the Panel seems to miss the point of 5.5, which is to identify a kind of violation that can exist even when there is a risk assessment to support the more protective measure. Imagine that in Australia–Salmon there had been a risk assessment that supported the measures imposed on imported salmon as the best or only way to achieve the stated ALOP of very high protection (very low risk) from salmon-borne diseases. Even so, given the much lower (that is, less protective, higher risk) ALOP for the same risk from baitfish and ornamental fish, we might suspect that Australia had chosen its specially high ALOP for salmon because salmon imports threatened a local industry in a way that imports of baitfish and ornamental fish did not. That is the sort of ‘discrimination or disguised restriction on trade’ that 5.5 is designed to deal with. But then it is clear that a finding of a 5.5 violation that depends essentially on the absence of a risk assessment, like the Panel's finding here, misses the point.Footnote ³³

Let us turn now to SPS 5.6. In the end, the Panel makes no finding under 5.6, for two reasons. First, it says it cannot determine the relevant ALOP for Section 727 against which to measure alternatives. In the 5.5 analysis, the Panel decided the relevant ALOP for Section 727 was the achieved ALOP, that is to say, zero risk (or nearly). But now China, which had argued for using the achieved ALOP under 5.5, was arguing that the relevant ALOP for 5.6 purposes was the United States' lower stated ALOP of ‘safe’. The Panel noted that China seemed to be adopting inconsistent positions, but, even so, it said it was unable to choose between these ALOPs for the 5.6 context, and hence it could not carry out the 5.6 analysis. As a further reason, the Panel said it did not have enough information to decide whether China's suggested less-restrictive alternative to Section 727, namely the usual FSIS process (without Congressional intervention), would suffice to achieve either of these ALOPs. Now the second of these reasons is sound, and it justifies the Panel's decision not to make a finding under 5.6. But the first reason is just a confusion. It is straightforward to argue that the relevant ALOP for 5.6 purposes is the United States' stated ALOP of ‘safe’.

The Appellate Body did a good deal to confuse this issue in Australia–Salmon, but instead of going through their analysis, let me simply explain why, if the stated ALOP for the measure under review and the achieved ALOP are different, then the relevant ALOP for 5.6 purposes is the lower of the two (that is, the less protective). Just consider two hypothetical cases. First, suppose that the achieved ALOP is lower than the stated ALOP (as might happen without any bad faith on the regulator's part if the regulator really would like the higher, stated level of protection, but just doesn't see how to achieve it with any remotely plausible measure). Now if the complainant country identifies an alternative, less-trade-restrictive measure that achieves the achieved ALOP of the measure under review, the regulator should obviously be required to use that alternative measure. The alternative does not get the regulator all that it (sincerely) wants, but, even so, it gets the regulator as much protection as the actual measure does, at less trade cost. So if the regulator uses the alternative, it is no worse off, and its trading partners are better off. Plainly, the correct ALOP to use for 5.6 purposes is the achieved ALOP. Now suppose instead that the achieved ALOP for the measure under review is higher than the stated ALOP (as might happen without bad faith if the regulator doesn't see any plausible way to achieve the lower stated ALOP without overshooting). Now if the complainant country identifies a less-trade-restrictive way to achieve exactly the stated ALOP, the regulator should be required to use it. The regulator will not get from the alternative measure all the protection it is getting from the actual measure, but it will get from the alternative measure all the protection it ever claimed to want. In effect, the complainant has shown the regulator a way to achieve the stated ALOP without overshooting. Plainly, the correct ALOP to use for 5.6 purposes in this case is the stated ALOP. These two cases together show that the relevant ALOP for 5.6 purposes is the stated ALOP or the achieved ALOP, whichever is lower (less protective). So in US–Poultry (China), the relevant ALOP for 5.6 purposes is the United States' stated ALOP of ‘safe’. (But of course, as we have noted, the Panel does not have enough information to decide the 5.6 issue even with that ALOP established.)

One final point. It follows from our arguments above that the relevant conception of the ALOP is not the same under 5.5 and under 5.6. Under 5.6, as we have just seen, the relevant ALOP is the stated ALOP or the achieved ALOP, whichever is lower. Under 5.5, as we saw earlier, the relevant ALOP is the stated ALOP, unless the regulator's claim that that is what it is seeking is utterly implausible (which does not follow just from the fact that the achieved ALOP is different).Footnote ³⁴ It should not be surprising or troubling that the ‘ALOP’ means something different under 5.5 and 5.6. It is a bastard conception anyway. The SPS is written in such a way as to suggest that a Member first selects the level of risk it wants to allow for any particular type and context of risk (the ‘ALOP’), and then looks around for a measure to achieve that level of risk. But this is not the way a rational regulator would proceed. Crudely, the rational regulator would start with an attitude to the risk in context that specifies how much it values the avoidance of that risk, and it would then consider the costs and benefits (in risk-avoidance) of various possible measures, before selecting the one with the greatest net value. To be fair to the drafters of the SPS, it would not have been easy to draft the Agreement in a way that avoided suggesting the misguided picture that the actual text suggests. But the point is not to criticize or defend the drafters; it is just to explain why we should be willing to accept that ‘ALOP’ does not mean the same thing everywhere it occurs.

6. SPS Article 8 and Annex C.1(a)

This aspect of the decision need not detain us long. SPS Article 8, in conjunction with Annex C.1(a), requires that decisions under ‘control, inspection and approval procedures’ be undertaken and completed ‘without undue delay’. The Panel finds that the normal FSIS procedures are an ‘approval procedure’, because they are used to determine whether China should be approved to export poultry to the United States. So the question is whether Section 727 resulted in ‘undue delay’ in completion of the FSIS procedures. The Panel says that the question whether delay is ‘undue’ is a question not just of the length of time, but of whether delay of that length is justified. It says that since Section 727 has already been held to be ‘arbitrary or unjustified’, because it is not based on a risk assessment, any delay caused by Section 727 is undue. Hence, since Section 727 causes delay, it violates Article 8.

This analysis misses the point of Annex C.1(a) in much the same way that the Panel's analysis of SPS 5.5 missed the point of that provision. It is true, as the Panel says, that under C.1(a), whether there is ‘undue delay’ is not determined just by how long the delay is. But, surely, the length of the delay matters. The question is whether the regulating Member is taking longer than necessary to carry out a reasonable review, given the nature of the problem. In this case, the six-and-a-half months required by Section 727 seems a perfectly reasonable amount of time for the FSIS to carry out the re-inspection of China that Congress thought was desirable in view of the two years that had passed since the previous inspection and in view of the recent adoption by China of a comprehensive new food-safety regime. But in the Panel's analysis of the case, the length of time drops out entirely. That cannot be what Annex C.1(a) is really about. The Panel is simply bootstrapping the violation of 5.1 into one more, essentially irrelevant, violation. (Innocently, no doubt. They are doing it because they were asked to, and because that is where they are led by a literalist reading of what other Panels have said about ‘undue delay’.) There is no real harm done, but it is unsatisfying nonetheless.

7. GATT Article I, SPS Article 4, and ‘equivalence regimes’

The Panel finds, without much trouble, that Section 727 violates GATT Article I. Section 727 is origin-specific, so under the ‘hypothetical like products analysis’ it treats poultry from China differently from (hypothetical) like poultry from other countries. The main focus of the United States' response was its GATT XX(b) defense. But the United States offered two arguments against finding a GATT Article I violation. First, it said the Panel should not use the hypothetical-like-products analysis when the goal of the measure is health. The United States cited EC–Asbestos Footnote ³⁵ in support of this claim, but this invocation of EC–Asbestos, which involved an origin-neutral measure, seems disingenuous at best. The United States' second argument is much more interesting. They argued that the Panel should not find a GATT Article I violation in this automatic way when dealing with an ‘equivalence regime’.Footnote ³⁶ According to the United States, it is of the essence of an equivalence regime to make distinctions between countries, and, since SPS Article 4 encourages Members to operate equivalence regimes, that Article means we should not in effect presume equivalence regimes to be illegal under GATT Article I. Specifically, SPS Article 4 places the burden on the exporting country, in this case China, to ‘objectively demonstrate’ that its SPS measures are equivalent to the United States’, and the GATT Article I analysis should not undermine SPS Article 4. This is an argument with genuine force, and the Panel offers no real response to it. Note that this argument also weighs against shifting the burden of justifying an origin-specific measure to the respondent country under SPS 2.3 (as discussed in Section 5), if the measure in question was generated by an equivalence regime.

Before going on with the discussion of GATT Article I, I should note a complication. In the next section, I shall explain why I think a reviewing Panel should not consider the GATT at all, once it has decided that the measure under review is an SPS measure. If I am right about that, then the GATT Article I issue should never have been discussed in this case. But not everyone will be persuaded by my argument about the relationship between the SPS and the GATT. And even if we forget about GATT Article I, we have just noted that the same issue about the significance of the fact that we are dealing with an equivalence regime arises in connection with SPS 2.3. So I shall go on and discuss GATT Article I, and I shall let the discussion of GATT Article I stand in also for discussion of SPS 2.3 (which I will mention only now and then to remind the reader that it is there in the background).

In the end, if the Panel was right to discuss the GATT at all, it was right to find a (prima facie) GATT Article I violation. But to see why – to give the answer to the United States' argument against finding such a violation – we need to look more closely at just what sort of ‘equivalence regime’ Article 4 should be read as encouraging. The Panel seems to concede that the US regime was the sort of regime Article 4 contemplated,Footnote ³⁷ but I think that is very doubtful.

We need to make some distinctions in the world of possible equivalence regimes that are often glossed over. To be precise, we need to make two distinctions, first between what I shall call ‘origin-neutral regimes’ and ‘origin-specific regimes’, and then, within the category of origin-specific regimes, between ‘facilitative regimes’ and ‘exclusionary regimes’.Footnote ³⁸ First, notice that one can engage with questions of equivalence in an entirely origin-neutral way. To have a crude example, imagine that Importer requires that milk be pasteurized by heating to some specified temperature for a specified time. Another Member, Exporter, who wishes to export milk to Importer, claims that heating the milk to a lower temperature for a longer time has the same effect. If Importer in fact recognizes that Exporter's method is equally effective, it is not just Exporter that will benefit. Any other Member can now use either Importer's method or Exporter's method and expect to sell its milk in Importer's market. The recognition of equivalence of the alternative method is prompted by Exporter, but there is no essential reference to Exporter in the recognition by Importer that heating milk to the lower temperature for a longer time works as well as a higher temperature for a shorter time. Anyone can benefit.Footnote ³⁹ An ‘equivalence regime’ that makes such decisions as this is one example of an ‘origin-neutral regime’.

There is nothing in Article 4 that makes it unequivocally clear that it addresses any stronger version of ‘equivalence’ than this: one Member's acknowledging the equivalence of another's substantive SPS standard. (‘Substantive standards’ include quarantine requirements, sampling and inspection methods, and so on.) Indeed, if we considered only the first sentence of 4.1, there would be a strong argument that this sort of case is all the drafters had in mind.Footnote ⁴⁰ This is the kind of case where it is by far the most likely that Exporter might be able to ‘objectively demonstrate’ that its standards are equivalent to Importer's. It is also the kind of case where it is most natural to speak of Exporter as having measures that ‘differ’ from Importer's; this locution makes it sound as if the ‘difference’ is a difference of substance, and not just the fact that each country necessarily has its own bureaucratic system for SPS control, even if they are administering the same standards. On the other hand, the second sentence of 4.1, and also of 4.2, arguably suggest a broader scope for Article 4. So let us continue.

If Importer allows importation of milk that has been processed in a specified way (or specified ways), then it needs to know how particular shipments have been processed. One way to assure itself about this is to certify individual foreign producers on the basis of inspection by Importers' inspectors, or inspection by reliable, non-governmental third parties, or self-certification by producers, or some combination of the above. The case where Importer chooses to inspect foreign producers for itself might seem to be what 4.1, second sentence, addresses (or at least, one case that sentence addresses).Footnote ⁴¹ Notice that this producer-by-producer inspection system can still be operated in an origin-neutral way, as long as Importer is willing to go through the certification process on the same terms for individual producers from any exporting country. Of course, by ‘origin-neutral’, what I really mean here, and what we usually mean, is ‘country-neutral’. A program that allows only products from certified producers obviously cares about the origin of the products – the origin in particular producers' facilities. But, as noted, the producer certification system can be open on the same terms to producers from any country. And since the relevant WTO provisions are all about discrimination between countries, such a program is ‘origin-neutral’ in the relevant sense.

Importer may prefer, however, to rely on the certification systems of the producers' home countries, at least where it regards those as reliable. So let us imagine next that Importer retains its own certification system, open to all producers from any country, but that it is also willing to accept, in place of its own certification, certification by the producer's home country, if the home country is one Importer trusts. Importer accepts certification from some other countries' governments, but not from all. Now we are dealing with an origin-specific regime. Importer distinguishes between other countries in a way that is very likely to affect trade flows. Producers from anywhere in the world have the opportunity to export to Importer if they pursue certification by Importer directly, but life is easier for producers in a country whose certifications Importer accepts. And thus, such countries are advantaged in their trade relations with Importer. This is what I refer to as a ‘facilitative’ regime: recognition by Importer of the equivalence of the exporting country's certification system for producers is not a prerequisite to exports from there, but it does facilitate them. And Article 4.2 might seem to contemplate a program such as this.Footnote ⁴²

The last possibility (for present purposes) is an ‘exclusionary’ system, where Importer declines to operate a certification system of its own for individual foreign producers and allows imports only from producers in countries whose certification system Importer recognizes as equivalent. Under such a system, the impeccably safe producer from a non-certified country is out of luck; he is summarily excluded. All milk, of whatever quality, is excluded if it comes from a non-certified country. This, of course, is the model for the United States’ ‘equivalence regime’ for poultry. It is an exclusionary regime.

I do not want to make any definitive claims here about what Article 4 has to say about ‘facilitative regimes’ versus ‘exclusionary regimes’, but I find it initially quite implausible that Article 4 contemplates, much less means to encourage, exclusionary regimes. The most important evidence here is that Article 4 requires recognition of equivalence only if Exporter ‘objectively demonstrates’ to Importer that its measure, now meaning its system for certification of producers, is equivalent. The burden is on Exporter. It seems implausible that Article 4 means to authorize the United States to announce that it will not accept any poultry at all from some other country until that country demonstrates that its national certification system is the equal of the United States' – thus allowing the United States to exclude poultry from even perfectly safe producers in that country, without having to make any showing at all about the reasons for doing this. But that would be the effect if we adopted the United States' argument in US–Poultry (China) and declined to find a GATT Article I violation.

Support for the idea that exclusionary regimes are disfavored can also be found in a non-SPS context in the Appellate Body's US–Shrimp opinions. In the first US–Shrimp opinion, the Appellate Body notes that ‘the United States did not permit imports of shrimp harvested by commercial shrimp trawl vessels using TED's comparable in effectiveness to those required in the United States if those shrimp originated in waters of countries not certified under Section 609.’ And they say that ‘the resulting situation is difficult to reconcile with the declared policy objective of protecting and conserving sea turtles’.Footnote ⁴³ In fact, the situation is not at all difficult to reconcile with the declared policy objective; it could just be that the United States is holding turtle-friendly shrimpers hostage to try to force their home country to protect turtles across its whole shrimp fleet. But the Appellate Body's observation nonetheless suggests a hostility to what I have called ‘exclusionary’ regimes. In US–Shrimp (Article 21.5–Malaysia), the Appellate Body observes that the United States now does, in practice, admit shrimp harvested by turtle-friendly shrimpers even from non-certified countries.Footnote ⁴⁴ Unfortunately, they do not make it clear just why this fact is important – specifically, whether it is important because the GATT requires that the United States offer shipment-by-shipment (or at least producer-by-producer) certification, or merely because the United States' doing so relieves the Appellate Body of the necessity of deciding at this point whether the GATT requires it.

Returning to US–Poultry (China), the problem with the United States' reliance on SPS Article 4 as a reason not to find a GATT Article I violation is not with the general form of the argument, which is potentially sound and sensible. The problem is that the United States wants to apply the argument to an exclusionary regime. The United States' argument would be much more persuasive if the United States' regime were not exclusionary but facilitative. As we have noted above, even the facilitative regime extends to some countries a trade advantage. But as long as individual producers in Exporter can get certified on their own and export to the United States, it seems reasonable to regard Article 4 as implying that the United States should not have to go further and accept Exporter's certification system as a supplement to its own until Exporter ‘objectively demonstrates’ that its certifications are reliable. Now it would indeed work at cross-purposes with Article 4 to use GATT Article I or SPS 2.3 to put the burden on the United States. In sum, the reason the Panel was right to find a GATT Article I violation, and to implicitly put the burden on the United States under SPS 2.3, is that the United States' regime was exclusionary as opposed to facilitative. But correspondingly, the Panel was wrong to suggest that the United States' equivalence regime was the sort of regime SPS Article 4 was meant to encourage. Conversely, if the United States' regime had been facilitative, or if we thought Article 4 really did mean to encourage exclusionary regimes, that would count strongly against finding a GATT Article I violation.

Actually, there is another reason the Panel was right to reject the United States' argument in this case. Even if the equivalence regime had been facilitative, there is still the problem that China was forbidden by Section 727 to secure certification for six-and-a-half months, whereas no other country seeking certification suffered that disability. China was denied the same opportunity for certification (even if it were facilitative certification) that other countries enjoyed. This sort of distinction is not essential to any kind of equivalence regime. In practice, China was probably not disadvantaged vis-à-vis any other country, since no other country was in mid-process, and the chance that any other country could have gone through the whole certification process in six-and-a-half months seems slight. Or to make the same point the other way around, Congress could almost certainly have achieved the same effect as Section 727 just by saying that in view of China's problems and the passage of time since the FSIS's first review, China had to go through the whole review process again from scratch (without any explicit six-and-a-half-month ban). Then there would have been no formal disadvantage at all vis-à-vis any other uncertified country, hence no GATT Article I violation. But, of course, the Panel reviews the actual measure, not alternative measures that might have had the same effect. And a slight chance of comparative harm to China is not the same as no chance. So the Panel would have been right to find a GATT Article I violation (which of course is still just a prima facie GATT violation) in this case, even if the United States' program had been facilitative.Footnote ⁴⁵

Let me add a final point about equivalence and SPS Article 4 that has nothing to do with US–Poultry (China). In the scholarly literature, equivalence is most often discussed in the context of ‘mutual recognition’ agreements. So it is worth noting that the SPS Agreement nowhere speaks of mutual recognition.Footnote ⁴⁶ It is right not to. ‘Mutuality’ has no essential role to play in the system; when it is present, it should be epiphenomenal. To see why, notice first that the term ‘equivalent’ is actually misleading. Plainly, Importer is required to recognize Exporter's SPS measure if it is equal to or better than (meaning more protective than) Importer's own standard. But now imagine a situation where Exporter's standard is demonstrably strictly superior to Importer's (and, conversely, Importer's standard is strictly inferior to Exporter's). Now Importer is required by SPS Article 4 to recognize Exporter's standard, but Exporter is not required to recognize Importer's. And it would be absolutely illegal for Importer to say to Exporter, ‘I'll recognize your standard if and only if you recognize mine; the recognition must be mutual.’ In practice, Importer might well offer such a deal; and even though Exporter has a right to non-mutual recognition of its standard, it might accept the deal on mutuality in preference to litigation to enforce its right, if it cares enough about access to Importer's market, and if Importer's standard is not too inferior to its own. Offering the deal would not even be abusive on Importer's part, if it sincerely, but mistakenly, believes its standard is equivalent to Exporter's. But the fact remains, Exporter is not legally required to accede to mutual recognition as the price of access to Importer's market, nor should it even be encouraged to do so by the WTO institutions. Negotiating over a possible mutual recognition agreement may be useful for a variety of reasons. But, in principle, the question whether Importer is required to accept Exporter's standard and the question whether Exporter is required to accept Importer's standard are logically independent, and we should not let a supposed concern for mutuality obfuscate that.

8. The SPS Agreement and GATT XX(b)

The United States attempted to defend against China's GATT Article I and Article XI claims (the latter of which they effectively conceded, so far as the prima facie violation was concerned) by arguing that Section 727 was justified under GATT XX(b). Since the Panel had found various violations of the SPS Agreement, it was required to consider, ‘whether it is possible to justify Section 727 under Article XX(b) of the GATT 1994’ when they had already found that Section 727 was an SPS measure and in violation of the SPS.Footnote ⁴⁷ Notice at the outset that there is a possible ambiguity in this question about whether Section 727 can be ‘justified’ under XX(b). (I do not say that the Panel is tripped up by this ambiguity, but they do not expressly disambiguate it, and they may be influenced by lack of clarity about it.) The following two questions are distinct: (1) Can GATT XX(b) be used to justify Section 727 against possible GATT violations, even though Section 727 is an SPS measure and inconsistent with the SPS?, and (2) Can GATT XX(b) be used to justify Section 727 against the SPS violations that it has been found to commit? No one is arguing that the answer to (2) is yes; even the United States is not arguing that GATT XX(b) can be used as a defense against SPS claims.Footnote ⁴⁸ So the only real question is (1) above, whether GATT XX(b) can still be used to defend against claimed GATT violations that are charged against an SPS measure that violates the SPS. The Panel answers this question in the negative.Footnote ⁴⁹ And their reason appears to be that, once we give the SPS Agreement its proper significance as ‘context’ for the interpretation of GATT XX(b), we see that, when the measure under review is an SPS measure, XX(b) should be interpreted as simply replicating all the provisions of the SPS. In effect, the meaning of XX(b) is now just read off from the meaning of the SPS – so that a measure that violates the SPS necessarily fails to qualify for a XX(b) defense. If this is indeed what the Panel is saying, it seems untenable.

The Panel discusses at length the connections between the SPS Agreement and GATT XX(b), as part of explaining why the SPS is important context. The Preamble to the SPS says one goal is to ‘elaborate rules for the application of the provisions of GATT 1994 which relate to the use of sanitary or phytosanitary measures, in particular the provisions of Article XX(b)’. SPS 2.4 says that SPS measures that are consistent with the SPS ‘shall be presumed to be in accordance with’ the GATT provisions relating to SPS measures, in particular XX(b). SPS 3.2 says that SPS measures that conform to international standards ‘shall be deemed to be necessary to protect human, animal, or plant life or health’, and shall be ‘presumed’ to be consistent with the rest of the SPS Agreement and the GATT. Furthermore, there are numerous passages of the SPS that repeat language verbatim, or almost verbatim, from GATT XX(b) or the chapeau of XX.Footnote ⁵⁰ This is all true, and it is all important somehow. Even so, the Panel's claim that, with regard to SPS measures, GATT XX(b) simply replicates the SPS does violence to the language of both SPS 2.4 and GATT XX(b).

As to the language of SPS 2.4, notice that 2.4 creates an implication that runs in one direction only: a measure that is consistent with the SPS is presumed consistent with the GATT, especially XX(b). But if, as the Panel claims, GATT XX(b) simply replicates the SPS, we now have implications in both directions: (1) as before, a measure that is consistent with the SPS is consistent with the GATT, in particular XX(b), and, in the other direction, (2) to be eligible for a XX(b) defense, a measure must be consistent with the SPS (from which it follows that a measure that violates the SPS is not eligible for a XX(b) defense). Note that it is the second of these implications that the Panel relies on, when it finds that because Section 727 violates the SPS, it is not eligible for a XX(b) defense. But the obvious question is, if the drafters of the SPS intended implications running in both directions, why did they state the implication explicitly in one direction only? The Panel's view simply does not fit with the way SPS 2.4 is written.Footnote ⁵¹

As to the language of GATT XX(b), the Panel purports to reach its reading of XX(b) by using the SPS as context for interpretation. But the relevance of ‘context’ under the VCLT is solely to help us establish ‘the ordinary meaning to be given to the terms of the treaty in their context and in the light of its object and purpose’.Footnote ⁵² No amount of context can make the ‘ordinary meaning’ of the word ‘necessary’ in GATT XX(b) encompass, say, the requirement to do a risk assessment. ‘Necessary’ refers to the relation that exists between the measure and its goal, whereas the SPS is about requiring specific techniques for investigating that relation. The ‘ordinary meaning’ of ‘necessary’ simply cannot be expanded to incorporate all the requirements of the SPS.Footnote ⁵³

There are two other arguments against the Panel's position that are worth mentioning. First, it is not even clear what it would mean for GATT XX(b) to simply replicate the SPS. GATT XX is a defense, which means that formally, the burden is on the respondent to prove propositions that XX(b) makes relevant.Footnote ⁵⁴ In contrast, the SPS includes nothing that is formally a defense. The formal burden of proof is always on the complainant. So if GATT XX(b) replicates the SPS, who has the burden of proof? Secondly, it is worth bearing in mind that whatever its origins as an improvement on GATT XX(b), the SPS is a free-standing agreement. It applies to some measures that would never raise issues under GATT XX at all, because they would not violate any of the primary prohibitory provisions of the GATT (for example, an origin-neutral SPS measure that reduces trade, but that has no disparate effect on market shares in any relevant product market). This ‘free-standingness’ of the SPS Agreement does not logically exclude the possibility that GATT XX(b) simply replicates the SPS, but it makes it seem much less natural.

If GATT XX(b) does not replicate the SPS, how are they related? The answer is that they are related exactly as SPS 2.4 says they are – but the practical significance of what 2.4 says is greater than we may realize at first.Footnote ⁵⁵ The consequence of 2.4 is that, even though the SPS Agreement does not rewrite the GATT, it supersedes the GATT when the measure under review is an SPS measure; it makes it unnecessary to consider the GATT at all. Suppose we are dealing with an SPS measure. If the measure is held to be consistent with all the provisions of the SPS Agreement, then by SPS 2.4, the measure is presumed to be consistent with the GATT. It is not entirely clear whether this 2.4 presumption is rebuttable or irrebutable, but let us assume for the moment that it is irrebuttable, just to get through the structure of the principal argument. Assuming the presumption is irrebuttable, then the SPS-consistent measure is also GATT-consistent, and there is no need to consider the GATT directly at this point. But what if the measure under review is held to violate the SPS? It turns out that in this case also, there is no need to consider the GATT. If the measure violates the SPS, then we already know it must be changed to make it WTO-legal. In particular, all the SPS violations must be corrected. But once all the SPS violations are corrected, the SPS 2.4 presumption will come into play and tell us that the measure is also consistent with the GATT. So, the uncorrected measure may or may not violate the GATT, but we don't need to know whether it does. Investigating whether the measure violates the GATT cannot turn up any additional changes that need to be made to make the measure WTO-legal. In sum, if we are dealing with an SPS measure, thinking about the GATT will never accomplish anything of practical significance. For such measures, the SPS supersedes the GATT.Footnote ⁵⁶

The argument does not work so neatly, of course, if the 2.4 presumption is rebuttable. The US–Poultry (China) Panel seems to assume that the presumption is irrebuttable,Footnote ⁵⁷ and this seems the most natural reading. Given that the SPS was meant to strengthen the disciplines on SPS measures, and given that the SPS generally repeats all the basic principles of the GATT, it would seem odd if the GATT were stronger in any respect. But if the GATT is not stronger in any respect, then even a formally ‘rebuttable’ 2.4 presumption would be irrebuttable in fact; so we should just say it is irrebuttable. The only real reason for doubting that the presumption is irrebuttable is that the SPS uses ‘presumed’ in 2.4, and again in 3.2, but it also uses ‘deemed’ in 3.2 to refer to a different inference. (Measures that conform to international standards are ‘deemed’ necessary for an SPS purpose.) The obvious way to make sense of these different terms is to say that a ‘deeming’ is irrebutable, while a mere ‘presumption’ is rebuttable. But another possibility is that ‘presumed’ means different things in 3.2 and 2.4 – specifically, it is rebuttable in 3.2, where it appears cheek-by-jowl with ‘deemed’, but irrebuttable in 2.4. This is actually quite plausible if we remember a difference between the 2.4 presumption and the 3.2 presumption. In 3.2, the presumption is an inference from one specific feature of the measure, namely conformity to international standards, to the conclusion that the measure is consistent with all the rest of the SPS (and the GATT); whereas in 2.4, the presumption is an inference where the premise is consistency with all the relevant provisions of the SPS. The broader premise of the 2.4 presumption makes it much more natural to assume it is irrebuttable.

To my mind, the arguments above establish that the 2.4 presumption should be read as irrebuttable. A quick look at the possibilities for rebutting the presumption, if it were rebuttable, may give further support. How could the presumption ever be rebutted? One possibility would arise, as others have pointed out, if we think that the ‘necessity’ test under GATT XX(b) involves not just a less-restrictive-alternative test, but also cost–benefit balancing, or ‘strict proportionality’. Nothing in the SPS suggests a strict proportionality test, so if GATT XX(b) does involve such a test, it would be more restrictive in this respect than the SPS.Footnote ⁵⁸ I have argued elsewhere that GATT XX(b) should not be interpreted as imposing a strict proportionality test, and more specifically that applying a strict proportionality test is inconsistent with allowing the Member to choose its own appropriate level of protection.Footnote ⁵⁹ Here, in fact, is a case where the SPS as context clearly does have a bearing on how we should interpret XX(b). Although at least one GATT Panel asserted a right of contracting parties to choose their own level of protection before the creation of the WTO,Footnote ⁶⁰ and the WTO Appellate Body has consistently done the same under the GATT (and incidentally the GATS) in non-SPS cases,Footnote ⁶¹ one might still question whether that right is firmly entrenched in the GATT. But there can be no doubt at all of its importance in the SPS, given how often it is mentioned in the text. So the SPS as context should make it abundantly clear that at least in connection with SPS measures, the GATT XX(b) ‘necessity’ test must accommodate the right to choose one's ALOP, and hence must not include a strict proportionality test. So the GATT will not be more restrictive on this ground. The other possible situation in which the GATT might be more restrictive is if we are reviewing a facilitative equivalence regime. If we simply apply the ‘hypothetical like products test’, we will find that this regime violates GATT Article I, and the respondent country will have the burden of proof of justification. But I suggested above that, under the influence of SPS Article 4, we should not shift the burden of proof in such a case under SPS 2.3, despite the origin-specificity; in a facilitative regime, the burden of showing the equivalence of the exporting country's certification system should remain on the complainant, where Article 4 puts it.Footnote ⁶² Here again, the SPS as context, and specifically SPS Article 4, ought to prevent us from finding a GATT violation; if SPS Article 4 encourages facilitative equivalence regimes, we ought not to read GATT Article I to discourage them. The reader may worry that these arguments about how to read the GATT are question-begging when they are offered in support of the claim that the 2.4 presumption is irrebuttable. It might seem that the use they make of the SPS as context simply assumes that the GATT should not be more restrictive than the SPS, which is the very same question as the question about the force of the 2.4 presumption. But that is not quite right. These arguments about the SPS as context would have force even in the absence of the explicit 2.4 presumption. So the presumption and the arguments about context work together, each reinforcing the other.Footnote ⁶³

In the end, my analysis and the Panel's may have much the same consequences in practice (aside from the fact that my approach will economize on judicial resources, often avoiding the GATT inquiry entirely). So, does it matter which we adopt? Yes, there are still reasons to prefer my analysis. First, as I have explained, my analysis is truer to the VCLT interpretive methodology; and, given the importance of that methodology in WTO adjudication, it is worth being punctilious about its application. And, second, there is some danger, small perhaps but not non-existent, that the Panel's analysis, which involves reading the SPS's complete science-based approach into the word ‘necessary’ in cases involving SPS measures, would lead later tribunals to treat ‘scientific justification’, and perhaps even some version of risk assessment, as part of ‘necessity’ in other contexts as well. That would be a major change, which should not happen unless and until the Members agree on it.

Some people may worry that both the Panel's analysis and mine make it impossible to justify an SPS measure under, say, GATT XX(a),Footnote ⁶⁴ or even conceivably GATT XXI. It is impossible, and it isn't. Strictly speaking, it is impossible to justify an SPS measure under these provisions. I have explained why the GATT provides no defense to SPS violations,Footnote ⁶⁵ and why, if we are dealing with an SPS measure, we need never look at the GATT at all. But as the Biotech Panel has taught us, a single set of words in a regulation or law that has multiple purposes may constitute more than one measure; and the words can be given legal effect if any of the measures they constitute is legal.Footnote ⁶⁶ If one of the measures constituted by such multipurpose language has a public-morals purpose or a national-security purpose, then that measure is not an SPS measure at all, and neither the Panel nor I have said anything about it.

9. ‘Judicial economy’: on not wasting judicial resources

Finally, a word about judicial economy. The US–Poultry (China) Panel exercised judicial economy by not deciding China's SPS 5.5 claim involving supposed different levels of protection between poultry and other food products from China, and by not deciding China's claim under the Agriculture Agreement. So far so good. Unfortunately, much of the remaining analysis was unproductive. The reason, in a nutshell, is this. After finding an SPS 5.1 violation, the Panel went on to find violations of SPS 2.2, 5.5, 2.3, and 8, as well as GATT I:1 and XI:1. But in the Panel's analysis, every violation other than the 5.1 violation logically depended, directly or indirectly, on the 5.1 violation. As a result, the analysis of the other violations told us nothing new about what needs to be done to make the measure WTO-legal. We know the 5.1 violation must be corrected. But once the 5.1 violation is corrected, then as far as the Panel's analysis is concerned, all the other violations are corrected as well. The edifice of violations collapses like a house of cards. So the Panel's analysis of the other violations did not identify anything else that needed to be done to bring the measure into WTO compliance beyond correcting the 5.1 violation. After the 5.1 discussion, the rest of the analysis accomplished nothing of practical significance. Notice I am not saying that after finding the 5.1 violation, there was no point in considering 2.2, or 5.5, or 2.3, or 8. To say that would be inconsistent with the Appellate Body's proceedings in EC–Hormones and Australia–Salmon. What I am saying is just that there was no point in finding violations of those other SPS provisions that depended on the 5.1 violation. The Panel should have investigated whether those other provisions were violated in any way that did not depend on the 5.1 violation (which is possible in principle for all four of them).

Let me quickly explain how it is that in the Panel's analysis, all the other violations depend on the 5.1 violation. With regard to 2.2, the Panel notes that a 5.1 violation is generally thought to entail a 2.2 violation. Having noted that, the Panel does go on to say that not only is there no risk assessment, there is no scientific justification other than a risk assessment either. Still, the statement that ‘the United States did not do a risk assessment and also did not do anything else’ plainly depends on the proposition that ‘the United States did not do a risk assessment’. And if the United States had done a risk assessment, there would have been no ground in the circumstances of this case to assert that they acted without scientific justification. So the 2.2 finding depends on the 5.1 finding. With regard to 5.5, the Panel's only ground for saying that the (supposed) different ALOPs for China and Mexico constituted an ‘arbitrary or unjustifiable distinction’ was the absence of a risk assessment, that is the 5.1 violation. I explained in Section 5 why a 5.1 violation is not the only possible ground for such a finding; but it was the only ground the Panel relied on in this case. The 2.3 violation is deduced directly from the 5.5 violation, so it also depends on the 5.1 violation. With regard to the SPS Article 8 violation, the only reason the Panel gives for thinking the delay ‘undue’, and the only reason available in this case, was the lack of a risk assessment, that is the 5.1 violation. With regard to the GATT violations, the findings of prima facie violations under GATT Articles I and XI did not depend on the SPS 5.1 violation. But we do not have all-things-considered GATT violations until we have considered and rejected the GATT XX(b) defense. The Panel's only ground for rejecting the XX(b) defense was that XX(b) replicates the SPS, and Section 727 violates the SPS. But since we have just seen that all the SPS violations depend on the SPS 5.1 violation, so does the Panel's rejection of the GATT XX(b) defense, and hence so do the all-things-considered GATT violations.Footnote ⁶⁷

So, in the Panel's analysis, every violation of the SPS or the GATT depends on the SPS 5.1 violation. As I have explained, that means that none of the Panel's analysis after the 5.1 discussion accomplishes anything of practical significance. It is not worth discussing here what other violations the Panel might have found without relying on the 5.1 violation. To my mind, the only other possible SPS violation is a 2.3 violation, and as explained in Section 8, the Panel should not have considered the GATT at all. But the real point is that, whatever the particular results, inquiries conducted without reliance on the 5.1 violation would have allowed us to find out whether anything else needed to be done to make the measure legal, other than doing a risk assessment.